From d676e9a2fb533494d1558483e8266605b1d5b632 Mon Sep 17 00:00:00 2001 From: ladevieq Date: Sun, 10 Jul 2022 17:41:40 +0200 Subject: Thirdparty includes flagged as SYSTEM --- thirdparty/renderdoc.h | 688 ++ thirdparty/stb_image.h | 7897 ++++++++++++++++++ thirdparty/vk_mem_alloc.h | 19229 ++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 27814 insertions(+) create mode 100644 thirdparty/renderdoc.h create mode 100644 thirdparty/stb_image.h create mode 100644 thirdparty/vk_mem_alloc.h (limited to 'thirdparty') diff --git a/thirdparty/renderdoc.h b/thirdparty/renderdoc.h new file mode 100644 index 0000000..075306c --- /dev/null +++ b/thirdparty/renderdoc.h @@ -0,0 +1,688 @@ +/****************************************************************************** + * The MIT License (MIT) + * + * Copyright (c) 2019-2020 Baldur Karlsson + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to deal + * in the Software without restriction, including without limitation the rights + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell + * copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN + * THE SOFTWARE. + ******************************************************************************/ + +#pragma once + +////////////////////////////////////////////////////////////////////////////////////////////////// +// +// Documentation for the API is available at https://renderdoc.org/docs/in_application_api.html +// + +#if !defined(RENDERDOC_NO_STDINT) +#include +#endif + +#if defined(WIN32) || defined(__WIN32__) || defined(_WIN32) || defined(_MSC_VER) +#define RENDERDOC_CC __cdecl +#elif defined(__linux__) +#define RENDERDOC_CC +#elif defined(__APPLE__) +#define RENDERDOC_CC +#else +#error "Unknown platform" +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +////////////////////////////////////////////////////////////////////////////////////////////////// +// Constants not used directly in below API + +// This is a GUID/magic value used for when applications pass a path where shader debug +// information can be found to match up with a stripped shader. +// the define can be used like so: const GUID RENDERDOC_ShaderDebugMagicValue = +// RENDERDOC_ShaderDebugMagicValue_value +#define RENDERDOC_ShaderDebugMagicValue_struct \ + { \ + 0xeab25520, 0x6670, 0x4865, 0x84, 0x29, 0x6c, 0x8, 0x51, 0x54, 0x00, 0xff \ + } + +// as an alternative when you want a byte array (assuming x86 endianness): +#define RENDERDOC_ShaderDebugMagicValue_bytearray \ + { \ + 0x20, 0x55, 0xb2, 0xea, 0x70, 0x66, 0x65, 0x48, 0x84, 0x29, 0x6c, 0x8, 0x51, 0x54, 0x00, 0xff \ + } + +// truncated version when only a uint64_t is available (e.g. Vulkan tags): +#define RENDERDOC_ShaderDebugMagicValue_truncated 0x48656670eab25520ULL + +////////////////////////////////////////////////////////////////////////////////////////////////// +// RenderDoc capture options +// + +typedef enum RENDERDOC_CaptureOption { + // Allow the application to enable vsync + // + // Default - enabled + // + // 1 - The application can enable or disable vsync at will + // 0 - vsync is force disabled + eRENDERDOC_Option_AllowVSync = 0, + + // Allow the application to enable fullscreen + // + // Default - enabled + // + // 1 - The application can enable or disable fullscreen at will + // 0 - fullscreen is force disabled + eRENDERDOC_Option_AllowFullscreen = 1, + + // Record API debugging events and messages + // + // Default - disabled + // + // 1 - Enable built-in API debugging features and records the results into + // the capture, which is matched up with events on replay + // 0 - no API debugging is forcibly enabled + eRENDERDOC_Option_APIValidation = 2, + eRENDERDOC_Option_DebugDeviceMode = 2, // deprecated name of this enum + + // Capture CPU callstacks for API events + // + // Default - disabled + // + // 1 - Enables capturing of callstacks + // 0 - no callstacks are captured + eRENDERDOC_Option_CaptureCallstacks = 3, + + // When capturing CPU callstacks, only capture them from drawcalls. + // This option does nothing without the above option being enabled + // + // Default - disabled + // + // 1 - Only captures callstacks for drawcall type API events. + // Ignored if CaptureCallstacks is disabled + // 0 - Callstacks, if enabled, are captured for every event. + eRENDERDOC_Option_CaptureCallstacksOnlyDraws = 4, + + // Specify a delay in seconds to wait for a debugger to attach, after + // creating or injecting into a process, before continuing to allow it to run. + // + // 0 indicates no delay, and the process will run immediately after injection + // + // Default - 0 seconds + // + eRENDERDOC_Option_DelayForDebugger = 5, + + // Verify buffer access. This includes checking the memory returned by a Map() call to + // detect any out-of-bounds modification, as well as initialising buffers with undefined contents + // to a marker value to catch use of uninitialised memory. + // + // NOTE: This option is only valid for OpenGL and D3D11. Explicit APIs such as D3D12 and Vulkan do + // not do the same kind of interception & checking and undefined contents are really undefined. + // + // Default - disabled + // + // 1 - Verify buffer access + // 0 - No verification is performed, and overwriting bounds may cause crashes or corruption in + // RenderDoc. + eRENDERDOC_Option_VerifyBufferAccess = 6, + + // The old name for eRENDERDOC_Option_VerifyBufferAccess was eRENDERDOC_Option_VerifyMapWrites. + // This option now controls the filling of uninitialised buffers with 0xdddddddd which was + // previously always enabled + eRENDERDOC_Option_VerifyMapWrites = eRENDERDOC_Option_VerifyBufferAccess, + + // Hooks any system API calls that create child processes, and injects + // RenderDoc into them recursively with the same options. + // + // Default - disabled + // + // 1 - Hooks into spawned child processes + // 0 - Child processes are not hooked by RenderDoc + eRENDERDOC_Option_HookIntoChildren = 7, + + // By default RenderDoc only includes resources in the final capture necessary + // for that frame, this allows you to override that behaviour. + // + // Default - disabled + // + // 1 - all live resources at the time of capture are included in the capture + // and available for inspection + // 0 - only the resources referenced by the captured frame are included + eRENDERDOC_Option_RefAllResources = 8, + + // **NOTE**: As of RenderDoc v1.1 this option has been deprecated. Setting or + // getting it will be ignored, to allow compatibility with older versions. + // In v1.1 the option acts as if it's always enabled. + // + // By default RenderDoc skips saving initial states for resources where the + // previous contents don't appear to be used, assuming that writes before + // reads indicate previous contents aren't used. + // + // Default - disabled + // + // 1 - initial contents at the start of each captured frame are saved, even if + // they are later overwritten or cleared before being used. + // 0 - unless a read is detected, initial contents will not be saved and will + // appear as black or empty data. + eRENDERDOC_Option_SaveAllInitials = 9, + + // In APIs that allow for the recording of command lists to be replayed later, + // RenderDoc may choose to not capture command lists before a frame capture is + // triggered, to reduce overheads. This means any command lists recorded once + // and replayed many times will not be available and may cause a failure to + // capture. + // + // NOTE: This is only true for APIs where multithreading is difficult or + // discouraged. Newer APIs like Vulkan and D3D12 will ignore this option + // and always capture all command lists since the API is heavily oriented + // around it and the overheads have been reduced by API design. + // + // 1 - All command lists are captured from the start of the application + // 0 - Command lists are only captured if their recording begins during + // the period when a frame capture is in progress. + eRENDERDOC_Option_CaptureAllCmdLists = 10, + + // Mute API debugging output when the API validation mode option is enabled + // + // Default - enabled + // + // 1 - Mute any API debug messages from being displayed or passed through + // 0 - API debugging is displayed as normal + eRENDERDOC_Option_DebugOutputMute = 11, + + // Option to allow vendor extensions to be used even when they may be + // incompatible with RenderDoc and cause corrupted replays or crashes. + // + // Default - inactive + // + // No values are documented, this option should only be used when absolutely + // necessary as directed by a RenderDoc developer. + eRENDERDOC_Option_AllowUnsupportedVendorExtensions = 12, + +} RENDERDOC_CaptureOption; + +// Sets an option that controls how RenderDoc behaves on capture. +// +// Returns 1 if the option and value are valid +// Returns 0 if either is invalid and the option is unchanged +typedef int(RENDERDOC_CC *pRENDERDOC_SetCaptureOptionU32)(RENDERDOC_CaptureOption opt, uint32_t val); +typedef int(RENDERDOC_CC *pRENDERDOC_SetCaptureOptionF32)(RENDERDOC_CaptureOption opt, float val); + +// Gets the current value of an option as a uint32_t +// +// If the option is invalid, 0xffffffff is returned +typedef uint32_t(RENDERDOC_CC *pRENDERDOC_GetCaptureOptionU32)(RENDERDOC_CaptureOption opt); + +// Gets the current value of an option as a float +// +// If the option is invalid, -FLT_MAX is returned +typedef float(RENDERDOC_CC *pRENDERDOC_GetCaptureOptionF32)(RENDERDOC_CaptureOption opt); + +typedef enum RENDERDOC_InputButton { + // '0' - '9' matches ASCII values + eRENDERDOC_Key_0 = 0x30, + eRENDERDOC_Key_1 = 0x31, + eRENDERDOC_Key_2 = 0x32, + eRENDERDOC_Key_3 = 0x33, + eRENDERDOC_Key_4 = 0x34, + eRENDERDOC_Key_5 = 0x35, + eRENDERDOC_Key_6 = 0x36, + eRENDERDOC_Key_7 = 0x37, + eRENDERDOC_Key_8 = 0x38, + eRENDERDOC_Key_9 = 0x39, + + // 'A' - 'Z' matches ASCII values + eRENDERDOC_Key_A = 0x41, + eRENDERDOC_Key_B = 0x42, + eRENDERDOC_Key_C = 0x43, + eRENDERDOC_Key_D = 0x44, + eRENDERDOC_Key_E = 0x45, + eRENDERDOC_Key_F = 0x46, + eRENDERDOC_Key_G = 0x47, + eRENDERDOC_Key_H = 0x48, + eRENDERDOC_Key_I = 0x49, + eRENDERDOC_Key_J = 0x4A, + eRENDERDOC_Key_K = 0x4B, + eRENDERDOC_Key_L = 0x4C, + eRENDERDOC_Key_M = 0x4D, + eRENDERDOC_Key_N = 0x4E, + eRENDERDOC_Key_O = 0x4F, + eRENDERDOC_Key_P = 0x50, + eRENDERDOC_Key_Q = 0x51, + eRENDERDOC_Key_R = 0x52, + eRENDERDOC_Key_S = 0x53, + eRENDERDOC_Key_T = 0x54, + eRENDERDOC_Key_U = 0x55, + eRENDERDOC_Key_V = 0x56, + eRENDERDOC_Key_W = 0x57, + eRENDERDOC_Key_X = 0x58, + eRENDERDOC_Key_Y = 0x59, + eRENDERDOC_Key_Z = 0x5A, + + // leave the rest of the ASCII range free + // in case we want to use it later + eRENDERDOC_Key_NonPrintable = 0x100, + + eRENDERDOC_Key_Divide, + eRENDERDOC_Key_Multiply, + eRENDERDOC_Key_Subtract, + eRENDERDOC_Key_Plus, + + eRENDERDOC_Key_F1, + eRENDERDOC_Key_F2, + eRENDERDOC_Key_F3, + eRENDERDOC_Key_F4, + eRENDERDOC_Key_F5, + eRENDERDOC_Key_F6, + eRENDERDOC_Key_F7, + eRENDERDOC_Key_F8, + eRENDERDOC_Key_F9, + eRENDERDOC_Key_F10, + eRENDERDOC_Key_F11, + eRENDERDOC_Key_F12, + + eRENDERDOC_Key_Home, + eRENDERDOC_Key_End, + eRENDERDOC_Key_Insert, + eRENDERDOC_Key_Delete, + eRENDERDOC_Key_PageUp, + eRENDERDOC_Key_PageDn, + + eRENDERDOC_Key_Backspace, + eRENDERDOC_Key_Tab, + eRENDERDOC_Key_PrtScrn, + eRENDERDOC_Key_Pause, + + eRENDERDOC_Key_Max, +} RENDERDOC_InputButton; + +// Sets which key or keys can be used to toggle focus between multiple windows +// +// If keys is NULL or num is 0, toggle keys will be disabled +typedef void(RENDERDOC_CC *pRENDERDOC_SetFocusToggleKeys)(RENDERDOC_InputButton *keys, int num); + +// Sets which key or keys can be used to capture the next frame +// +// If keys is NULL or num is 0, captures keys will be disabled +typedef void(RENDERDOC_CC *pRENDERDOC_SetCaptureKeys)(RENDERDOC_InputButton *keys, int num); + +typedef enum RENDERDOC_OverlayBits { + // This single bit controls whether the overlay is enabled or disabled globally + eRENDERDOC_Overlay_Enabled = 0x1, + + // Show the average framerate over several seconds as well as min/max + eRENDERDOC_Overlay_FrameRate = 0x2, + + // Show the current frame number + eRENDERDOC_Overlay_FrameNumber = 0x4, + + // Show a list of recent captures, and how many captures have been made + eRENDERDOC_Overlay_CaptureList = 0x8, + + // Default values for the overlay mask + eRENDERDOC_Overlay_Default = (eRENDERDOC_Overlay_Enabled | eRENDERDOC_Overlay_FrameRate | + eRENDERDOC_Overlay_FrameNumber | eRENDERDOC_Overlay_CaptureList), + + // Enable all bits + eRENDERDOC_Overlay_All = ~0U, + + // Disable all bits + eRENDERDOC_Overlay_None = 0, +} RENDERDOC_OverlayBits; + +// returns the overlay bits that have been set +typedef uint32_t(RENDERDOC_CC *pRENDERDOC_GetOverlayBits)(); +// sets the overlay bits with an and & or mask +typedef void(RENDERDOC_CC *pRENDERDOC_MaskOverlayBits)(uint32_t And, uint32_t Or); + +// this function will attempt to remove RenderDoc's hooks in the application. +// +// Note: that this can only work correctly if done immediately after +// the module is loaded, before any API work happens. RenderDoc will remove its +// injected hooks and shut down. Behaviour is undefined if this is called +// after any API functions have been called, and there is still no guarantee of +// success. +typedef void(RENDERDOC_CC *pRENDERDOC_RemoveHooks)(); + +// DEPRECATED: compatibility for code compiled against pre-1.4.1 headers. +typedef pRENDERDOC_RemoveHooks pRENDERDOC_Shutdown; + +// This function will unload RenderDoc's crash handler. +// +// If you use your own crash handler and don't want RenderDoc's handler to +// intercede, you can call this function to unload it and any unhandled +// exceptions will pass to the next handler. +typedef void(RENDERDOC_CC *pRENDERDOC_UnloadCrashHandler)(); + +// Sets the capture file path template +// +// pathtemplate is a UTF-8 string that gives a template for how captures will be named +// and where they will be saved. +// +// Any extension is stripped off the path, and captures are saved in the directory +// specified, and named with the filename and the frame number appended. If the +// directory does not exist it will be created, including any parent directories. +// +// If pathtemplate is NULL, the template will remain unchanged +// +// Example: +// +// SetCaptureFilePathTemplate("my_captures/example"); +// +// Capture #1 -> my_captures/example_frame123.rdc +// Capture #2 -> my_captures/example_frame456.rdc +typedef void(RENDERDOC_CC *pRENDERDOC_SetCaptureFilePathTemplate)(const char *pathtemplate); + +// returns the current capture path template, see SetCaptureFileTemplate above, as a UTF-8 string +typedef const char *(RENDERDOC_CC *pRENDERDOC_GetCaptureFilePathTemplate)(); + +// DEPRECATED: compatibility for code compiled against pre-1.1.2 headers. +typedef pRENDERDOC_SetCaptureFilePathTemplate pRENDERDOC_SetLogFilePathTemplate; +typedef pRENDERDOC_GetCaptureFilePathTemplate pRENDERDOC_GetLogFilePathTemplate; + +// returns the number of captures that have been made +typedef uint32_t(RENDERDOC_CC *pRENDERDOC_GetNumCaptures)(); + +// This function returns the details of a capture, by index. New captures are added +// to the end of the list. +// +// filename will be filled with the absolute path to the capture file, as a UTF-8 string +// pathlength will be written with the length in bytes of the filename string +// timestamp will be written with the time of the capture, in seconds since the Unix epoch +// +// Any of the parameters can be NULL and they'll be skipped. +// +// The function will return 1 if the capture index is valid, or 0 if the index is invalid +// If the index is invalid, the values will be unchanged +// +// Note: when captures are deleted in the UI they will remain in this list, so the +// capture path may not exist anymore. +typedef uint32_t(RENDERDOC_CC *pRENDERDOC_GetCapture)(uint32_t idx, char *filename, + uint32_t *pathlength, uint64_t *timestamp); + +// Sets the comments associated with a capture file. These comments are displayed in the +// UI program when opening. +// +// filePath should be a path to the capture file to add comments to. If set to NULL or "" +// the most recent capture file created made will be used instead. +// comments should be a NULL-terminated UTF-8 string to add as comments. +// +// Any existing comments will be overwritten. +typedef void(RENDERDOC_CC *pRENDERDOC_SetCaptureFileComments)(const char *filePath, + const char *comments); + +// returns 1 if the RenderDoc UI is connected to this application, 0 otherwise +typedef uint32_t(RENDERDOC_CC *pRENDERDOC_IsTargetControlConnected)(); + +// DEPRECATED: compatibility for code compiled against pre-1.1.1 headers. +// This was renamed to IsTargetControlConnected in API 1.1.1, the old typedef is kept here for +// backwards compatibility with old code, it is castable either way since it's ABI compatible +// as the same function pointer type. +typedef pRENDERDOC_IsTargetControlConnected pRENDERDOC_IsRemoteAccessConnected; + +// This function will launch the Replay UI associated with the RenderDoc library injected +// into the running application. +// +// if connectTargetControl is 1, the Replay UI will be launched with a command line parameter +// to connect to this application +// cmdline is the rest of the command line, as a UTF-8 string. E.g. a captures to open +// if cmdline is NULL, the command line will be empty. +// +// returns the PID of the replay UI if successful, 0 if not successful. +typedef uint32_t(RENDERDOC_CC *pRENDERDOC_LaunchReplayUI)(uint32_t connectTargetControl, + const char *cmdline); + +// RenderDoc can return a higher version than requested if it's backwards compatible, +// this function returns the actual version returned. If a parameter is NULL, it will be +// ignored and the others will be filled out. +typedef void(RENDERDOC_CC *pRENDERDOC_GetAPIVersion)(int *major, int *minor, int *patch); + +////////////////////////////////////////////////////////////////////////// +// Capturing functions +// + +// A device pointer is a pointer to the API's root handle. +// +// This would be an ID3D11Device, HGLRC/GLXContext, ID3D12Device, etc +typedef void *RENDERDOC_DevicePointer; + +// A window handle is the OS's native window handle +// +// This would be an HWND, GLXDrawable, etc +typedef void *RENDERDOC_WindowHandle; + +// A helper macro for Vulkan, where the device handle cannot be used directly. +// +// Passing the VkInstance to this macro will return the RENDERDOC_DevicePointer to use. +// +// Specifically, the value needed is the dispatch table pointer, which sits as the first +// pointer-sized object in the memory pointed to by the VkInstance. Thus we cast to a void** and +// indirect once. +#define RENDERDOC_DEVICEPOINTER_FROM_VKINSTANCE(inst) (*((void **)(inst))) + +// This sets the RenderDoc in-app overlay in the API/window pair as 'active' and it will +// respond to keypresses. Neither parameter can be NULL +typedef void(RENDERDOC_CC *pRENDERDOC_SetActiveWindow)(RENDERDOC_DevicePointer device, + RENDERDOC_WindowHandle wndHandle); + +// capture the next frame on whichever window and API is currently considered active +typedef void(RENDERDOC_CC *pRENDERDOC_TriggerCapture)(); + +// capture the next N frames on whichever window and API is currently considered active +typedef void(RENDERDOC_CC *pRENDERDOC_TriggerMultiFrameCapture)(uint32_t numFrames); + +// When choosing either a device pointer or a window handle to capture, you can pass NULL. +// Passing NULL specifies a 'wildcard' match against anything. This allows you to specify +// any API rendering to a specific window, or a specific API instance rendering to any window, +// or in the simplest case of one window and one API, you can just pass NULL for both. +// +// In either case, if there are two or more possible matching (device,window) pairs it +// is undefined which one will be captured. +// +// Note: for headless rendering you can pass NULL for the window handle and either specify +// a device pointer or leave it NULL as above. + +// Immediately starts capturing API calls on the specified device pointer and window handle. +// +// If there is no matching thing to capture (e.g. no supported API has been initialised), +// this will do nothing. +// +// The results are undefined (including crashes) if two captures are started overlapping, +// even on separate devices and/oror windows. +typedef void(RENDERDOC_CC *pRENDERDOC_StartFrameCapture)(RENDERDOC_DevicePointer device, + RENDERDOC_WindowHandle wndHandle); + +// Returns whether or not a frame capture is currently ongoing anywhere. +// +// This will return 1 if a capture is ongoing, and 0 if there is no capture running +typedef uint32_t(RENDERDOC_CC *pRENDERDOC_IsFrameCapturing)(); + +// Ends capturing immediately. +// +// This will return 1 if the capture succeeded, and 0 if there was an error capturing. +typedef uint32_t(RENDERDOC_CC *pRENDERDOC_EndFrameCapture)(RENDERDOC_DevicePointer device, + RENDERDOC_WindowHandle wndHandle); + +// Ends capturing immediately and discard any data stored without saving to disk. +// +// This will return 1 if the capture was discarded, and 0 if there was an error or no capture +// was in progress +typedef uint32_t(RENDERDOC_CC *pRENDERDOC_DiscardFrameCapture)(RENDERDOC_DevicePointer device, + RENDERDOC_WindowHandle wndHandle); + +////////////////////////////////////////////////////////////////////////////////////////////////// +// RenderDoc API versions +// + +// RenderDoc uses semantic versioning (http://semver.org/). +// +// MAJOR version is incremented when incompatible API changes happen. +// MINOR version is incremented when functionality is added in a backwards-compatible manner. +// PATCH version is incremented when backwards-compatible bug fixes happen. +// +// Note that this means the API returned can be higher than the one you might have requested. +// e.g. if you are running against a newer RenderDoc that supports 1.0.1, it will be returned +// instead of 1.0.0. You can check this with the GetAPIVersion entry point +typedef enum RENDERDOC_Version { + eRENDERDOC_API_Version_1_0_0 = 10000, // RENDERDOC_API_1_0_0 = 1 00 00 + eRENDERDOC_API_Version_1_0_1 = 10001, // RENDERDOC_API_1_0_1 = 1 00 01 + eRENDERDOC_API_Version_1_0_2 = 10002, // RENDERDOC_API_1_0_2 = 1 00 02 + eRENDERDOC_API_Version_1_1_0 = 10100, // RENDERDOC_API_1_1_0 = 1 01 00 + eRENDERDOC_API_Version_1_1_1 = 10101, // RENDERDOC_API_1_1_1 = 1 01 01 + eRENDERDOC_API_Version_1_1_2 = 10102, // RENDERDOC_API_1_1_2 = 1 01 02 + eRENDERDOC_API_Version_1_2_0 = 10200, // RENDERDOC_API_1_2_0 = 1 02 00 + eRENDERDOC_API_Version_1_3_0 = 10300, // RENDERDOC_API_1_3_0 = 1 03 00 + eRENDERDOC_API_Version_1_4_0 = 10400, // RENDERDOC_API_1_4_0 = 1 04 00 + eRENDERDOC_API_Version_1_4_1 = 10401, // RENDERDOC_API_1_4_1 = 1 04 01 +} RENDERDOC_Version; + +// API version changelog: +// +// 1.0.0 - initial release +// 1.0.1 - Bugfix: IsFrameCapturing() was returning false for captures that were triggered +// by keypress or TriggerCapture, instead of Start/EndFrameCapture. +// 1.0.2 - Refactor: Renamed eRENDERDOC_Option_DebugDeviceMode to eRENDERDOC_Option_APIValidation +// 1.1.0 - Add feature: TriggerMultiFrameCapture(). Backwards compatible with 1.0.x since the new +// function pointer is added to the end of the struct, the original layout is identical +// 1.1.1 - Refactor: Renamed remote access to target control (to better disambiguate from remote +// replay/remote server concept in replay UI) +// 1.1.2 - Refactor: Renamed "log file" in function names to just capture, to clarify that these +// are captures and not debug logging files. This is the first API version in the v1.0 +// branch. +// 1.2.0 - Added feature: SetCaptureFileComments() to add comments to a capture file that will be +// displayed in the UI program on load. +// 1.3.0 - Added feature: New capture option eRENDERDOC_Option_AllowUnsupportedVendorExtensions +// which allows users to opt-in to allowing unsupported vendor extensions to function. +// Should be used at the user's own risk. +// Refactor: Renamed eRENDERDOC_Option_VerifyMapWrites to +// eRENDERDOC_Option_VerifyBufferAccess, which now also controls initialisation to +// 0xdddddddd of uninitialised buffer contents. +// 1.4.0 - Added feature: DiscardFrameCapture() to discard a frame capture in progress and stop +// capturing without saving anything to disk. +// 1.4.1 - Refactor: Renamed Shutdown to RemoveHooks to better clarify what is happening + +typedef struct RENDERDOC_API_1_4_1 +{ + pRENDERDOC_GetAPIVersion GetAPIVersion; + + pRENDERDOC_SetCaptureOptionU32 SetCaptureOptionU32; + pRENDERDOC_SetCaptureOptionF32 SetCaptureOptionF32; + + pRENDERDOC_GetCaptureOptionU32 GetCaptureOptionU32; + pRENDERDOC_GetCaptureOptionF32 GetCaptureOptionF32; + + pRENDERDOC_SetFocusToggleKeys SetFocusToggleKeys; + pRENDERDOC_SetCaptureKeys SetCaptureKeys; + + pRENDERDOC_GetOverlayBits GetOverlayBits; + pRENDERDOC_MaskOverlayBits MaskOverlayBits; + + // Shutdown was renamed to RemoveHooks in 1.4.1. + // These unions allow old code to continue compiling without changes + union + { + pRENDERDOC_Shutdown Shutdown; + pRENDERDOC_RemoveHooks RemoveHooks; + }; + pRENDERDOC_UnloadCrashHandler UnloadCrashHandler; + + // Get/SetLogFilePathTemplate was renamed to Get/SetCaptureFilePathTemplate in 1.1.2. + // These unions allow old code to continue compiling without changes + union + { + // deprecated name + pRENDERDOC_SetLogFilePathTemplate SetLogFilePathTemplate; + // current name + pRENDERDOC_SetCaptureFilePathTemplate SetCaptureFilePathTemplate; + }; + union + { + // deprecated name + pRENDERDOC_GetLogFilePathTemplate GetLogFilePathTemplate; + // current name + pRENDERDOC_GetCaptureFilePathTemplate GetCaptureFilePathTemplate; + }; + + pRENDERDOC_GetNumCaptures GetNumCaptures; + pRENDERDOC_GetCapture GetCapture; + + pRENDERDOC_TriggerCapture TriggerCapture; + + // IsRemoteAccessConnected was renamed to IsTargetControlConnected in 1.1.1. + // This union allows old code to continue compiling without changes + union + { + // deprecated name + pRENDERDOC_IsRemoteAccessConnected IsRemoteAccessConnected; + // current name + pRENDERDOC_IsTargetControlConnected IsTargetControlConnected; + }; + pRENDERDOC_LaunchReplayUI LaunchReplayUI; + + pRENDERDOC_SetActiveWindow SetActiveWindow; + + pRENDERDOC_StartFrameCapture StartFrameCapture; + pRENDERDOC_IsFrameCapturing IsFrameCapturing; + pRENDERDOC_EndFrameCapture EndFrameCapture; + + // new function in 1.1.0 + pRENDERDOC_TriggerMultiFrameCapture TriggerMultiFrameCapture; + + // new function in 1.2.0 + pRENDERDOC_SetCaptureFileComments SetCaptureFileComments; + + // new function in 1.4.0 + pRENDERDOC_DiscardFrameCapture DiscardFrameCapture; +} RENDERDOC_API_1_4_1; + +typedef RENDERDOC_API_1_4_1 RENDERDOC_API_1_0_0; +typedef RENDERDOC_API_1_4_1 RENDERDOC_API_1_0_1; +typedef RENDERDOC_API_1_4_1 RENDERDOC_API_1_0_2; +typedef RENDERDOC_API_1_4_1 RENDERDOC_API_1_1_0; +typedef RENDERDOC_API_1_4_1 RENDERDOC_API_1_1_1; +typedef RENDERDOC_API_1_4_1 RENDERDOC_API_1_1_2; +typedef RENDERDOC_API_1_4_1 RENDERDOC_API_1_2_0; +typedef RENDERDOC_API_1_4_1 RENDERDOC_API_1_3_0; +typedef RENDERDOC_API_1_4_1 RENDERDOC_API_1_4_0; + +////////////////////////////////////////////////////////////////////////////////////////////////// +// RenderDoc API entry point +// +// This entry point can be obtained via GetProcAddress/dlsym if RenderDoc is available. +// +// The name is the same as the typedef - "RENDERDOC_GetAPI" +// +// This function is not thread safe, and should not be called on multiple threads at once. +// Ideally, call this once as early as possible in your application's startup, before doing +// any API work, since some configuration functionality etc has to be done also before +// initialising any APIs. +// +// Parameters: +// version is a single value from the RENDERDOC_Version above. +// +// outAPIPointers will be filled out with a pointer to the corresponding struct of function +// pointers. +// +// Returns: +// 1 - if the outAPIPointers has been filled with a pointer to the API struct requested +// 0 - if the requested version is not supported or the arguments are invalid. +// +typedef int(RENDERDOC_CC *pRENDERDOC_GetAPI)(RENDERDOC_Version version, void **outAPIPointers); + +#ifdef __cplusplus +} // extern "C" +#endif diff --git a/thirdparty/stb_image.h b/thirdparty/stb_image.h new file mode 100644 index 0000000..ef76fc1 --- /dev/null +++ b/thirdparty/stb_image.h @@ -0,0 +1,7897 @@ +/* stb_image - v2.27 - public domain image loader - http://nothings.org/stb + no warranty implied; use at your own risk + + Do this: + #define STB_IMAGE_IMPLEMENTATION + before you include this file in *one* C or C++ file to create the implementation. + + // i.e. it should look like this: + #include ... + #include ... + #include ... + #define STB_IMAGE_IMPLEMENTATION + #include "stb_image.h" + + You can #define STBI_ASSERT(x) before the #include to avoid using assert.h. + And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free + + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib) + PNG 1/2/4/8/16-bit-per-channel + + TGA (not sure what subset, if a subset) + BMP non-1bpp, non-RLE + PSD (composited view only, no extra channels, 8/16 bit-per-channel) + + GIF (*comp always reports as 4-channel) + HDR (radiance rgbE format) + PIC (Softimage PIC) + PNM (PPM and PGM binary only) + + Animated GIF still needs a proper API, but here's one way to do it: + http://gist.github.com/urraka/685d9a6340b26b830d49 + + - decode from memory or through FILE (define STBI_NO_STDIO to remove code) + - decode from arbitrary I/O callbacks + - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON) + + Full documentation under "DOCUMENTATION" below. + + +LICENSE + + See end of file for license information. + +RECENT REVISION HISTORY: + + 2.27 (2021-07-11) document stbi_info better, 16-bit PNM support, bug fixes + 2.26 (2020-07-13) many minor fixes + 2.25 (2020-02-02) fix warnings + 2.24 (2020-02-02) fix warnings; thread-local failure_reason and flip_vertically + 2.23 (2019-08-11) fix clang static analysis warning + 2.22 (2019-03-04) gif fixes, fix warnings + 2.21 (2019-02-25) fix typo in comment + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) bugfix, 1-bit BMP, 16-bitness query, fix warnings + 2.16 (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes + 2.15 (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-12-04) experimental 16-bit API, only for PNG so far; fixes + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) 16-bit PNGS; enable SSE2 in non-gcc x64 + RGB-format JPEG; remove white matting in PSD; + allocate large structures on the stack; + correct channel count for PNG & BMP + 2.10 (2016-01-22) avoid warning introduced in 2.09 + 2.09 (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED + + See end of file for full revision history. + + + ============================ Contributors ========================= + + Image formats Extensions, features + Sean Barrett (jpeg, png, bmp) Jetro Lauha (stbi_info) + Nicolas Schulz (hdr, psd) Martin "SpartanJ" Golini (stbi_info) + Jonathan Dummer (tga) James "moose2000" Brown (iPhone PNG) + Jean-Marc Lienher (gif) Ben "Disch" Wenger (io callbacks) + Tom Seddon (pic) Omar Cornut (1/2/4-bit PNG) + Thatcher Ulrich (psd) Nicolas Guillemot (vertical flip) + Ken Miller (pgm, ppm) Richard Mitton (16-bit PSD) + github:urraka (animated gif) Junggon Kim (PNM comments) + Christopher Forseth (animated gif) Daniel Gibson (16-bit TGA) + socks-the-fox (16-bit PNG) + Jeremy Sawicki (handle all ImageNet JPGs) + Optimizations & bugfixes Mikhail Morozov (1-bit BMP) + Fabian "ryg" Giesen Anael Seghezzi (is-16-bit query) + Arseny Kapoulkine Simon Breuss (16-bit PNM) + John-Mark Allen + Carmelo J Fdez-Aguera + + Bug & warning fixes + Marc LeBlanc David Woo Guillaume George Martins Mozeiko + Christpher Lloyd Jerry Jansson Joseph Thomson Blazej Dariusz Roszkowski + Phil Jordan Dave Moore Roy Eltham + Hayaki Saito Nathan Reed Won Chun + Luke Graham Johan Duparc Nick Verigakis the Horde3D community + Thomas Ruf Ronny Chevalier github:rlyeh + Janez Zemva John Bartholomew Michal Cichon github:romigrou + Jonathan Blow Ken Hamada Tero Hanninen github:svdijk + Eugene Golushkov Laurent Gomila Cort Stratton github:snagar + Aruelien Pocheville Sergio Gonzalez Thibault Reuille github:Zelex + Cass Everitt Ryamond Barbiero github:grim210 + Paul Du Bois Engin Manap Aldo Culquicondor github:sammyhw + Philipp Wiesemann Dale Weiler Oriol Ferrer Mesia github:phprus + Josh Tobin Matthew Gregan github:poppolopoppo + Julian Raschke Gregory Mullen Christian Floisand github:darealshinji + Baldur Karlsson Kevin Schmidt JR Smith github:Michaelangel007 + Brad Weinberger Matvey Cherevko github:mosra + Luca Sas Alexander Veselov Zack Middleton [reserved] + Ryan C. Gordon [reserved] [reserved] + DO NOT ADD YOUR NAME HERE + + Jacko Dirks + + To add your name to the credits, pick a random blank space in the middle and fill it. + 80% of merge conflicts on stb PRs are due to people adding their name at the end + of the credits. +*/ + +#ifndef STBI_INCLUDE_STB_IMAGE_H +#define STBI_INCLUDE_STB_IMAGE_H + +// DOCUMENTATION +// +// Limitations: +// - no 12-bit-per-channel JPEG +// - no JPEGs with arithmetic coding +// - GIF always returns *comp=4 +// +// Basic usage (see HDR discussion below for HDR usage): +// int x,y,n; +// unsigned char *data = stbi_load(filename, &x, &y, &n, 0); +// // ... process data if not NULL ... +// // ... x = width, y = height, n = # 8-bit components per pixel ... +// // ... replace '0' with '1'..'4' to force that many components per pixel +// // ... but 'n' will always be the number that it would have been if you said 0 +// stbi_image_free(data) +// +// Standard parameters: +// int *x -- outputs image width in pixels +// int *y -- outputs image height in pixels +// int *channels_in_file -- outputs # of image components in image file +// int desired_channels -- if non-zero, # of image components requested in result +// +// The return value from an image loader is an 'unsigned char *' which points +// to the pixel data, or NULL on an allocation failure or if the image is +// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels, +// with each pixel consisting of N interleaved 8-bit components; the first +// pixel pointed to is top-left-most in the image. There is no padding between +// image scanlines or between pixels, regardless of format. The number of +// components N is 'desired_channels' if desired_channels is non-zero, or +// *channels_in_file otherwise. If desired_channels is non-zero, +// *channels_in_file has the number of components that _would_ have been +// output otherwise. E.g. if you set desired_channels to 4, you will always +// get RGBA output, but you can check *channels_in_file to see if it's trivially +// opaque because e.g. there were only 3 channels in the source image. +// +// An output image with N components has the following components interleaved +// in this order in each pixel: +// +// N=#comp components +// 1 grey +// 2 grey, alpha +// 3 red, green, blue +// 4 red, green, blue, alpha +// +// If image loading fails for any reason, the return value will be NULL, +// and *x, *y, *channels_in_file will be unchanged. The function +// stbi_failure_reason() can be queried for an extremely brief, end-user +// unfriendly explanation of why the load failed. Define STBI_NO_FAILURE_STRINGS +// to avoid compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly +// more user-friendly ones. +// +// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized. +// +// To query the width, height and component count of an image without having to +// decode the full file, you can use the stbi_info family of functions: +// +// int x,y,n,ok; +// ok = stbi_info(filename, &x, &y, &n); +// // returns ok=1 and sets x, y, n if image is a supported format, +// // 0 otherwise. +// +// Note that stb_image pervasively uses ints in its public API for sizes, +// including sizes of memory buffers. This is now part of the API and thus +// hard to change without causing breakage. As a result, the various image +// loaders all have certain limits on image size; these differ somewhat +// by format but generally boil down to either just under 2GB or just under +// 1GB. When the decoded image would be larger than this, stb_image decoding +// will fail. +// +// Additionally, stb_image will reject image files that have any of their +// dimensions set to a larger value than the configurable STBI_MAX_DIMENSIONS, +// which defaults to 2**24 = 16777216 pixels. Due to the above memory limit, +// the only way to have an image with such dimensions load correctly +// is for it to have a rather extreme aspect ratio. Either way, the +// assumption here is that such larger images are likely to be malformed +// or malicious. If you do need to load an image with individual dimensions +// larger than that, and it still fits in the overall size limit, you can +// #define STBI_MAX_DIMENSIONS on your own to be something larger. +// +// =========================================================================== +// +// UNICODE: +// +// If compiling for Windows and you wish to use Unicode filenames, compile +// with +// #define STBI_WINDOWS_UTF8 +// and pass utf8-encoded filenames. Call stbi_convert_wchar_to_utf8 to convert +// Windows wchar_t filenames to utf8. +// +// =========================================================================== +// +// Philosophy +// +// stb libraries are designed with the following priorities: +// +// 1. easy to use +// 2. easy to maintain +// 3. good performance +// +// Sometimes I let "good performance" creep up in priority over "easy to maintain", +// and for best performance I may provide less-easy-to-use APIs that give higher +// performance, in addition to the easy-to-use ones. Nevertheless, it's important +// to keep in mind that from the standpoint of you, a client of this library, +// all you care about is #1 and #3, and stb libraries DO NOT emphasize #3 above all. +// +// Some secondary priorities arise directly from the first two, some of which +// provide more explicit reasons why performance can't be emphasized. +// +// - Portable ("ease of use") +// - Small source code footprint ("easy to maintain") +// - No dependencies ("ease of use") +// +// =========================================================================== +// +// I/O callbacks +// +// I/O callbacks allow you to read from arbitrary sources, like packaged +// files or some other source. Data read from callbacks are processed +// through a small internal buffer (currently 128 bytes) to try to reduce +// overhead. +// +// The three functions you must define are "read" (reads some bytes of data), +// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end). +// +// =========================================================================== +// +// SIMD support +// +// The JPEG decoder will try to automatically use SIMD kernels on x86 when +// supported by the compiler. For ARM Neon support, you must explicitly +// request it. +// +// (The old do-it-yourself SIMD API is no longer supported in the current +// code.) +// +// On x86, SSE2 will automatically be used when available based on a run-time +// test; if not, the generic C versions are used as a fall-back. On ARM targets, +// the typical path is to have separate builds for NEON and non-NEON devices +// (at least this is true for iOS and Android). Therefore, the NEON support is +// toggled by a build flag: define STBI_NEON to get NEON loops. +// +// If for some reason you do not want to use any of SIMD code, or if +// you have issues compiling it, you can disable it entirely by +// defining STBI_NO_SIMD. +// +// =========================================================================== +// +// HDR image support (disable by defining STBI_NO_HDR) +// +// stb_image supports loading HDR images in general, and currently the Radiance +// .HDR file format specifically. You can still load any file through the existing +// interface; if you attempt to load an HDR file, it will be automatically remapped +// to LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1; +// both of these constants can be reconfigured through this interface: +// +// stbi_hdr_to_ldr_gamma(2.2f); +// stbi_hdr_to_ldr_scale(1.0f); +// +// (note, do not use _inverse_ constants; stbi_image will invert them +// appropriately). +// +// Additionally, there is a new, parallel interface for loading files as +// (linear) floats to preserve the full dynamic range: +// +// float *data = stbi_loadf(filename, &x, &y, &n, 0); +// +// If you load LDR images through this interface, those images will +// be promoted to floating point values, run through the inverse of +// constants corresponding to the above: +// +// stbi_ldr_to_hdr_scale(1.0f); +// stbi_ldr_to_hdr_gamma(2.2f); +// +// Finally, given a filename (or an open file or memory block--see header +// file for details) containing image data, you can query for the "most +// appropriate" interface to use (that is, whether the image is HDR or +// not), using: +// +// stbi_is_hdr(char *filename); +// +// =========================================================================== +// +// iPhone PNG support: +// +// We optionally support converting iPhone-formatted PNGs (which store +// premultiplied BGRA) back to RGB, even though they're internally encoded +// differently. To enable this conversion, call +// stbi_convert_iphone_png_to_rgb(1). +// +// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per +// pixel to remove any premultiplied alpha *only* if the image file explicitly +// says there's premultiplied data (currently only happens in iPhone images, +// and only if iPhone convert-to-rgb processing is on). +// +// =========================================================================== +// +// ADDITIONAL CONFIGURATION +// +// - You can suppress implementation of any of the decoders to reduce +// your code footprint by #defining one or more of the following +// symbols before creating the implementation. +// +// STBI_NO_JPEG +// STBI_NO_PNG +// STBI_NO_BMP +// STBI_NO_PSD +// STBI_NO_TGA +// STBI_NO_GIF +// STBI_NO_HDR +// STBI_NO_PIC +// STBI_NO_PNM (.ppm and .pgm) +// +// - You can request *only* certain decoders and suppress all other ones +// (this will be more forward-compatible, as addition of new decoders +// doesn't require you to disable them explicitly): +// +// STBI_ONLY_JPEG +// STBI_ONLY_PNG +// STBI_ONLY_BMP +// STBI_ONLY_PSD +// STBI_ONLY_TGA +// STBI_ONLY_GIF +// STBI_ONLY_HDR +// STBI_ONLY_PIC +// STBI_ONLY_PNM (.ppm and .pgm) +// +// - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still +// want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB +// +// - If you define STBI_MAX_DIMENSIONS, stb_image will reject images greater +// than that size (in either width or height) without further processing. +// This is to let programs in the wild set an upper bound to prevent +// denial-of-service attacks on untrusted data, as one could generate a +// valid image of gigantic dimensions and force stb_image to allocate a +// huge block of memory and spend disproportionate time decoding it. By +// default this is set to (1 << 24), which is 16777216, but that's still +// very big. + +#ifndef STBI_NO_STDIO +#include +#endif // STBI_NO_STDIO + +#define STBI_VERSION 1 + +enum +{ + STBI_default = 0, // only used for desired_channels + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4 +}; + +#include +typedef unsigned char stbi_uc; +typedef unsigned short stbi_us; + +#ifdef __cplusplus +extern "C" { +#endif + +#ifndef STBIDEF +#ifdef STB_IMAGE_STATIC +#define STBIDEF static +#else +#define STBIDEF extern +#endif +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// PRIMARY API - works on images of any type +// + +// +// load image by filename, open file, or memory buffer +// + +typedef struct +{ + int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read + void (*skip) (void *user,int n); // skip the next 'n' bytes, or 'unget' the last -n bytes if negative + int (*eof) (void *user); // returns nonzero if we are at end of file/data +} stbi_io_callbacks; + +//////////////////////////////////// +// +// 8-bits-per-channel interface +// + +STBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk , void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +// for stbi_load_from_file, file pointer is left pointing immediately after image +#endif + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp); +#endif + +#ifdef STBI_WINDOWS_UTF8 +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input); +#endif + +//////////////////////////////////// +// +// 16-bits-per-channel interface +// + +STBIDEF stbi_us *stbi_load_16_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_us *stbi_load_16 (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +#endif + +//////////////////////////////////// +// +// float-per-channel interface +// +#ifndef STBI_NO_LINEAR + STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + + #ifndef STBI_NO_STDIO + STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); + #endif +#endif + +#ifndef STBI_NO_HDR + STBIDEF void stbi_hdr_to_ldr_gamma(float gamma); + STBIDEF void stbi_hdr_to_ldr_scale(float scale); +#endif // STBI_NO_HDR + +#ifndef STBI_NO_LINEAR + STBIDEF void stbi_ldr_to_hdr_gamma(float gamma); + STBIDEF void stbi_ldr_to_hdr_scale(float scale); +#endif // STBI_NO_LINEAR + +// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user); +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len); +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename); +STBIDEF int stbi_is_hdr_from_file(FILE *f); +#endif // STBI_NO_STDIO + + +// get a VERY brief reason for failure +// on most compilers (and ALL modern mainstream compilers) this is threadsafe +STBIDEF const char *stbi_failure_reason (void); + +// free the loaded image -- this is just free() +STBIDEF void stbi_image_free (void *retval_from_stbi_load); + +// get image dimensions & components without fully decoding +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp); +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len); +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *clbk, void *user); + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info (char const *filename, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +STBIDEF int stbi_is_16_bit (char const *filename); +STBIDEF int stbi_is_16_bit_from_file(FILE *f); +#endif + + + +// for image formats that explicitly notate that they have premultiplied alpha, +// we just return the colors as stored in the file. set this flag to force +// unpremultiplication. results are undefined if the unpremultiply overflow. +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply); + +// indicate whether we should process iphone images back to canonical format, +// or just pass them through "as-is" +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert); + +// flip the image vertically, so the first pixel in the output array is the bottom left +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip); + +// as above, but only applies to images loaded on the thread that calls the function +// this function is only available if your compiler supports thread-local variables; +// calling it will fail to link if your compiler doesn't +STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply); +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert); +STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip); + +// ZLIB client - used by PNG, available for other purposes + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen); +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header); +STBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + + +#ifdef __cplusplus +} +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// +#endif // STBI_INCLUDE_STB_IMAGE_H + +#ifdef STB_IMAGE_IMPLEMENTATION + +#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \ + || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \ + || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \ + || defined(STBI_ONLY_ZLIB) + #ifndef STBI_ONLY_JPEG + #define STBI_NO_JPEG + #endif + #ifndef STBI_ONLY_PNG + #define STBI_NO_PNG + #endif + #ifndef STBI_ONLY_BMP + #define STBI_NO_BMP + #endif + #ifndef STBI_ONLY_PSD + #define STBI_NO_PSD + #endif + #ifndef STBI_ONLY_TGA + #define STBI_NO_TGA + #endif + #ifndef STBI_ONLY_GIF + #define STBI_NO_GIF + #endif + #ifndef STBI_ONLY_HDR + #define STBI_NO_HDR + #endif + #ifndef STBI_ONLY_PIC + #define STBI_NO_PIC + #endif + #ifndef STBI_ONLY_PNM + #define STBI_NO_PNM + #endif +#endif + +#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB) +#define STBI_NO_ZLIB +#endif + + +#include +#include // ptrdiff_t on osx +#include +#include +#include + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) +#include // ldexp, pow +#endif + +#ifndef STBI_NO_STDIO +#include +#endif + +#ifndef STBI_ASSERT +#include +#define STBI_ASSERT(x) assert(x) +#endif + +#ifdef __cplusplus +#define STBI_EXTERN extern "C" +#else +#define STBI_EXTERN extern +#endif + + +#ifndef _MSC_VER + #ifdef __cplusplus + #define stbi_inline inline + #else + #define stbi_inline + #endif +#else + #define stbi_inline __forceinline +#endif + +#ifndef STBI_NO_THREAD_LOCALS + #if defined(__cplusplus) && __cplusplus >= 201103L + #define STBI_THREAD_LOCAL thread_local + #elif defined(__GNUC__) && __GNUC__ < 5 + #define STBI_THREAD_LOCAL __thread + #elif defined(_MSC_VER) + #define STBI_THREAD_LOCAL __declspec(thread) + #elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 201112L && !defined(__STDC_NO_THREADS__) + #define STBI_THREAD_LOCAL _Thread_local + #endif + + #ifndef STBI_THREAD_LOCAL + #if defined(__GNUC__) + #define STBI_THREAD_LOCAL __thread + #endif + #endif +#endif + +#ifdef _MSC_VER +typedef unsigned short stbi__uint16; +typedef signed short stbi__int16; +typedef unsigned int stbi__uint32; +typedef signed int stbi__int32; +#else +#include +typedef uint16_t stbi__uint16; +typedef int16_t stbi__int16; +typedef uint32_t stbi__uint32; +typedef int32_t stbi__int32; +#endif + +// should produce compiler error if size is wrong +typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1]; + +#ifdef _MSC_VER +#define STBI_NOTUSED(v) (void)(v) +#else +#define STBI_NOTUSED(v) (void)sizeof(v) +#endif + +#ifdef _MSC_VER +#define STBI_HAS_LROTL +#endif + +#ifdef STBI_HAS_LROTL + #define stbi_lrot(x,y) _lrotl(x,y) +#else + #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (-(y) & 31))) +#endif + +#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED)) +// ok +#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED) +// ok +#else +#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)." +#endif + +#ifndef STBI_MALLOC +#define STBI_MALLOC(sz) malloc(sz) +#define STBI_REALLOC(p,newsz) realloc(p,newsz) +#define STBI_FREE(p) free(p) +#endif + +#ifndef STBI_REALLOC_SIZED +#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz) +#endif + +// x86/x64 detection +#if defined(__x86_64__) || defined(_M_X64) +#define STBI__X64_TARGET +#elif defined(__i386) || defined(_M_IX86) +#define STBI__X86_TARGET +#endif + +#if defined(__GNUC__) && defined(STBI__X86_TARGET) && !defined(__SSE2__) && !defined(STBI_NO_SIMD) +// gcc doesn't support sse2 intrinsics unless you compile with -msse2, +// which in turn means it gets to use SSE2 everywhere. This is unfortunate, +// but previous attempts to provide the SSE2 functions with runtime +// detection caused numerous issues. The way architecture extensions are +// exposed in GCC/Clang is, sadly, not really suited for one-file libs. +// New behavior: if compiled with -msse2, we use SSE2 without any +// detection; if not, we don't use it at all. +#define STBI_NO_SIMD +#endif + +#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD) +// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET +// +// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the +// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant. +// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not +// simultaneously enabling "-mstackrealign". +// +// See https://github.com/nothings/stb/issues/81 for more information. +// +// So default to no SSE2 on 32-bit MinGW. If you've read this far and added +// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2. +#define STBI_NO_SIMD +#endif + +#if !defined(STBI_NO_SIMD) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET)) +#define STBI_SSE2 +#include + +#ifdef _MSC_VER + +#if _MSC_VER >= 1400 // not VC6 +#include // __cpuid +static int stbi__cpuid3(void) +{ + int info[4]; + __cpuid(info,1); + return info[3]; +} +#else +static int stbi__cpuid3(void) +{ + int res; + __asm { + mov eax,1 + cpuid + mov res,edx + } + return res; +} +#endif + +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + int info3 = stbi__cpuid3(); + return ((info3 >> 26) & 1) != 0; +} +#endif + +#else // assume GCC-style if not VC++ +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + // If we're even attempting to compile this on GCC/Clang, that means + // -msse2 is on, which means the compiler is allowed to use SSE2 + // instructions at will, and so are we. + return 1; +} +#endif + +#endif +#endif + +// ARM NEON +#if defined(STBI_NO_SIMD) && defined(STBI_NEON) +#undef STBI_NEON +#endif + +#ifdef STBI_NEON +#include +#ifdef _MSC_VER +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name +#else +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) +#endif +#endif + +#ifndef STBI_SIMD_ALIGN +#define STBI_SIMD_ALIGN(type, name) type name +#endif + +#ifndef STBI_MAX_DIMENSIONS +#define STBI_MAX_DIMENSIONS (1 << 24) +#endif + +/////////////////////////////////////////////// +// +// stbi__context struct and start_xxx functions + +// stbi__context structure is our basic context used by all images, so it +// contains all the IO context, plus some basic image information +typedef struct +{ + stbi__uint32 img_x, img_y; + int img_n, img_out_n; + + stbi_io_callbacks io; + void *io_user_data; + + int read_from_callbacks; + int buflen; + stbi_uc buffer_start[128]; + int callback_already_read; + + stbi_uc *img_buffer, *img_buffer_end; + stbi_uc *img_buffer_original, *img_buffer_original_end; +} stbi__context; + + +static void stbi__refill_buffer(stbi__context *s); + +// initialize a memory-decode context +static void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len) +{ + s->io.read = NULL; + s->read_from_callbacks = 0; + s->callback_already_read = 0; + s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer; + s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len; +} + +// initialize a callback-based context +static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user) +{ + s->io = *c; + s->io_user_data = user; + s->buflen = sizeof(s->buffer_start); + s->read_from_callbacks = 1; + s->callback_already_read = 0; + s->img_buffer = s->img_buffer_original = s->buffer_start; + stbi__refill_buffer(s); + s->img_buffer_original_end = s->img_buffer_end; +} + +#ifndef STBI_NO_STDIO + +static int stbi__stdio_read(void *user, char *data, int size) +{ + return (int) fread(data,1,size,(FILE*) user); +} + +static void stbi__stdio_skip(void *user, int n) +{ + int ch; + fseek((FILE*) user, n, SEEK_CUR); + ch = fgetc((FILE*) user); /* have to read a byte to reset feof()'s flag */ + if (ch != EOF) { + ungetc(ch, (FILE *) user); /* push byte back onto stream if valid. */ + } +} + +static int stbi__stdio_eof(void *user) +{ + return feof((FILE*) user) || ferror((FILE *) user); +} + +static stbi_io_callbacks stbi__stdio_callbacks = +{ + stbi__stdio_read, + stbi__stdio_skip, + stbi__stdio_eof, +}; + +static void stbi__start_file(stbi__context *s, FILE *f) +{ + stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f); +} + +//static void stop_file(stbi__context *s) { } + +#endif // !STBI_NO_STDIO + +static void stbi__rewind(stbi__context *s) +{ + // conceptually rewind SHOULD rewind to the beginning of the stream, + // but we just rewind to the beginning of the initial buffer, because + // we only use it after doing 'test', which only ever looks at at most 92 bytes + s->img_buffer = s->img_buffer_original; + s->img_buffer_end = s->img_buffer_original_end; +} + +enum +{ + STBI_ORDER_RGB, + STBI_ORDER_BGR +}; + +typedef struct +{ + int bits_per_channel; + int num_channels; + int channel_order; +} stbi__result_info; + +#ifndef STBI_NO_JPEG +static int stbi__jpeg_test(stbi__context *s); +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNG +static int stbi__png_test(stbi__context *s); +static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__png_is16(stbi__context *s); +#endif + +#ifndef STBI_NO_BMP +static int stbi__bmp_test(stbi__context *s); +static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_TGA +static int stbi__tga_test(stbi__context *s); +static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s); +static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc); +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__psd_is16(stbi__context *s); +#endif + +#ifndef STBI_NO_HDR +static int stbi__hdr_test(stbi__context *s); +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_test(stbi__context *s); +static void *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_GIF +static int stbi__gif_test(stbi__context *s); +static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp); +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNM +static int stbi__pnm_test(stbi__context *s); +static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__pnm_is16(stbi__context *s); +#endif + +static +#ifdef STBI_THREAD_LOCAL +STBI_THREAD_LOCAL +#endif +const char *stbi__g_failure_reason; + +STBIDEF const char *stbi_failure_reason(void) +{ + return stbi__g_failure_reason; +} + +#ifndef STBI_NO_FAILURE_STRINGS +static int stbi__err(const char *str) +{ + stbi__g_failure_reason = str; + return 0; +} +#endif + +static void *stbi__malloc(size_t size) +{ + return STBI_MALLOC(size); +} + +// stb_image uses ints pervasively, including for offset calculations. +// therefore the largest decoded image size we can support with the +// current code, even on 64-bit targets, is INT_MAX. this is not a +// significant limitation for the intended use case. +// +// we do, however, need to make sure our size calculations don't +// overflow. hence a few helper functions for size calculations that +// multiply integers together, making sure that they're non-negative +// and no overflow occurs. + +// return 1 if the sum is valid, 0 on overflow. +// negative terms are considered invalid. +static int stbi__addsizes_valid(int a, int b) +{ + if (b < 0) return 0; + // now 0 <= b <= INT_MAX, hence also + // 0 <= INT_MAX - b <= INTMAX. + // And "a + b <= INT_MAX" (which might overflow) is the + // same as a <= INT_MAX - b (no overflow) + return a <= INT_MAX - b; +} + +// returns 1 if the product is valid, 0 on overflow. +// negative factors are considered invalid. +static int stbi__mul2sizes_valid(int a, int b) +{ + if (a < 0 || b < 0) return 0; + if (b == 0) return 1; // mul-by-0 is always safe + // portable way to check for no overflows in a*b + return a <= INT_MAX/b; +} + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow +static int stbi__mad2sizes_valid(int a, int b, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add); +} +#endif + +// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow +static int stbi__mad3sizes_valid(int a, int b, int c, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__addsizes_valid(a*b*c, add); +} + +// returns 1 if "a*b*c*d + add" has no negative terms/factors and doesn't overflow +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static int stbi__mad4sizes_valid(int a, int b, int c, int d, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__mul2sizes_valid(a*b*c, d) && stbi__addsizes_valid(a*b*c*d, add); +} +#endif + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// mallocs with size overflow checking +static void *stbi__malloc_mad2(int a, int b, int add) +{ + if (!stbi__mad2sizes_valid(a, b, add)) return NULL; + return stbi__malloc(a*b + add); +} +#endif + +static void *stbi__malloc_mad3(int a, int b, int c, int add) +{ + if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL; + return stbi__malloc(a*b*c + add); +} + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static void *stbi__malloc_mad4(int a, int b, int c, int d, int add) +{ + if (!stbi__mad4sizes_valid(a, b, c, d, add)) return NULL; + return stbi__malloc(a*b*c*d + add); +} +#endif + +// stbi__err - error +// stbi__errpf - error returning pointer to float +// stbi__errpuc - error returning pointer to unsigned char + +#ifdef STBI_NO_FAILURE_STRINGS + #define stbi__err(x,y) 0 +#elif defined(STBI_FAILURE_USERMSG) + #define stbi__err(x,y) stbi__err(y) +#else + #define stbi__err(x,y) stbi__err(x) +#endif + +#define stbi__errpf(x,y) ((float *)(size_t) (stbi__err(x,y)?NULL:NULL)) +#define stbi__errpuc(x,y) ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL)) + +STBIDEF void stbi_image_free(void *retval_from_stbi_load) +{ + STBI_FREE(retval_from_stbi_load); +} + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp); +#endif + +#ifndef STBI_NO_HDR +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp); +#endif + +static int stbi__vertically_flip_on_load_global = 0; + +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_global = flag_true_if_should_flip; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__vertically_flip_on_load stbi__vertically_flip_on_load_global +#else +static STBI_THREAD_LOCAL int stbi__vertically_flip_on_load_local, stbi__vertically_flip_on_load_set; + +STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_local = flag_true_if_should_flip; + stbi__vertically_flip_on_load_set = 1; +} + +#define stbi__vertically_flip_on_load (stbi__vertically_flip_on_load_set \ + ? stbi__vertically_flip_on_load_local \ + : stbi__vertically_flip_on_load_global) +#endif // STBI_THREAD_LOCAL + +static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields + ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed + ri->channel_order = STBI_ORDER_RGB; // all current input & output are this, but this is here so we can add BGR order + ri->num_channels = 0; + + // test the formats with a very explicit header first (at least a FOURCC + // or distinctive magic number first) + #ifndef STBI_NO_PNG + if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_BMP + if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_GIF + if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PSD + if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp, ri, bpc); + #else + STBI_NOTUSED(bpc); + #endif + #ifndef STBI_NO_PIC + if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp, ri); + #endif + + // then the formats that can end up attempting to load with just 1 or 2 + // bytes matching expectations; these are prone to false positives, so + // try them later + #ifndef STBI_NO_JPEG + if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PNM + if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp, ri); + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + float *hdr = stbi__hdr_load(s, x,y,comp,req_comp, ri); + return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + + #ifndef STBI_NO_TGA + // test tga last because it's a crappy test! + if (stbi__tga_test(s)) + return stbi__tga_load(s,x,y,comp,req_comp, ri); + #endif + + return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt"); +} + +static stbi_uc *stbi__convert_16_to_8(stbi__uint16 *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi_uc *reduced; + + reduced = (stbi_uc *) stbi__malloc(img_len); + if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + reduced[i] = (stbi_uc)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling + + STBI_FREE(orig); + return reduced; +} + +static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi__uint16 *enlarged; + + enlarged = (stbi__uint16 *) stbi__malloc(img_len*2); + if (enlarged == NULL) return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + enlarged[i] = (stbi__uint16)((orig[i] << 8) + orig[i]); // replicate to high and low byte, maps 0->0, 255->0xffff + + STBI_FREE(orig); + return enlarged; +} + +static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel) +{ + int row; + size_t bytes_per_row = (size_t)w * bytes_per_pixel; + stbi_uc temp[2048]; + stbi_uc *bytes = (stbi_uc *)image; + + for (row = 0; row < (h>>1); row++) { + stbi_uc *row0 = bytes + row*bytes_per_row; + stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row; + // swap row0 with row1 + size_t bytes_left = bytes_per_row; + while (bytes_left) { + size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp); + memcpy(temp, row0, bytes_copy); + memcpy(row0, row1, bytes_copy); + memcpy(row1, temp, bytes_copy); + row0 += bytes_copy; + row1 += bytes_copy; + bytes_left -= bytes_copy; + } + } +} + +#ifndef STBI_NO_GIF +static void stbi__vertical_flip_slices(void *image, int w, int h, int z, int bytes_per_pixel) +{ + int slice; + int slice_size = w * h * bytes_per_pixel; + + stbi_uc *bytes = (stbi_uc *)image; + for (slice = 0; slice < z; ++slice) { + stbi__vertical_flip(bytes, w, h, bytes_per_pixel); + bytes += slice_size; + } +} +#endif + +static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 8); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 8) { + result = stbi__convert_16_to_8((stbi__uint16 *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 8; + } + + // @TODO: move stbi__convert_format to here + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc)); + } + + return (unsigned char *) result; +} + +static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 16); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 16) { + result = stbi__convert_8_to_16((stbi_uc *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 16; + } + + // @TODO: move stbi__convert_format16 to here + // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16)); + } + + return (stbi__uint16 *) result; +} + +#if !defined(STBI_NO_HDR) && !defined(STBI_NO_LINEAR) +static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp) +{ + if (stbi__vertically_flip_on_load && result != NULL) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(float)); + } +} +#endif + +#ifndef STBI_NO_STDIO + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBI_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide); +STBI_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default); +#endif + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input) +{ + return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL); +} +#endif + +static FILE *stbi__fopen(char const *filename, char const *mode) +{ + FILE *f; +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) + wchar_t wMode[64]; + wchar_t wFilename[1024]; + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)/sizeof(*wFilename))) + return 0; + + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)/sizeof(*wMode))) + return 0; + +#if defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != _wfopen_s(&f, wFilename, wMode)) + f = 0; +#else + f = _wfopen(wFilename, wMode); +#endif + +#elif defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != fopen_s(&f, filename, mode)) + f=0; +#else + f = fopen(filename, mode); +#endif + return f; +} + + +STBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + unsigned char *result; + if (!f) return stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); + if (result) { + // need to 'unget' all the characters in the IO buffer + fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR); + } + return result; +} + +STBIDEF stbi__uint16 *stbi_load_from_file_16(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__uint16 *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_and_postprocess_16bit(&s,x,y,comp,req_comp); + if (result) { + // need to 'unget' all the characters in the IO buffer + fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR); + } + return result; +} + +STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + stbi__uint16 *result; + if (!f) return (stbi_us *) stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file_16(f,x,y,comp,req_comp); + fclose(f); + return result; +} + + +#endif //!STBI_NO_STDIO + +STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); +} + +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_mem(&s,buffer,len); + + result = (unsigned char*) stbi__load_gif_main(&s, delays, x, y, z, comp, req_comp); + if (stbi__vertically_flip_on_load) { + stbi__vertical_flip_slices( result, *x, *y, *z, *comp ); + } + + return result; +} +#endif + +#ifndef STBI_NO_LINEAR +static float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + stbi__result_info ri; + float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp, &ri); + if (hdr_data) + stbi__float_postprocess(hdr_data,x,y,comp,req_comp); + return hdr_data; + } + #endif + data = stbi__load_and_postprocess_8bit(s, x, y, comp, req_comp); + if (data) + return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return stbi__errpf("unknown image type", "Image not of any known type, or corrupt"); +} + +STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +STBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + float *result; + FILE *f = stbi__fopen(filename, "rb"); + if (!f) return stbi__errpf("can't fopen", "Unable to open file"); + result = stbi_loadf_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_file(&s,f); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} +#endif // !STBI_NO_STDIO + +#endif // !STBI_NO_LINEAR + +// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is +// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always +// reports false! + +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(buffer); + STBI_NOTUSED(len); + return 0; + #endif +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result=0; + if (f) { + result = stbi_is_hdr_from_file(f); + fclose(f); + } + return result; +} + +STBIDEF int stbi_is_hdr_from_file(FILE *f) +{ + #ifndef STBI_NO_HDR + long pos = ftell(f); + int res; + stbi__context s; + stbi__start_file(&s,f); + res = stbi__hdr_test(&s); + fseek(f, pos, SEEK_SET); + return res; + #else + STBI_NOTUSED(f); + return 0; + #endif +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(clbk); + STBI_NOTUSED(user); + return 0; + #endif +} + +#ifndef STBI_NO_LINEAR +static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f; + +STBIDEF void stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; } +STBIDEF void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; } +#endif + +static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f; + +STBIDEF void stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; } +STBIDEF void stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; } + + +////////////////////////////////////////////////////////////////////////////// +// +// Common code used by all image loaders +// + +enum +{ + STBI__SCAN_load=0, + STBI__SCAN_type, + STBI__SCAN_header +}; + +static void stbi__refill_buffer(stbi__context *s) +{ + int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen); + s->callback_already_read += (int) (s->img_buffer - s->img_buffer_original); + if (n == 0) { + // at end of file, treat same as if from memory, but need to handle case + // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file + s->read_from_callbacks = 0; + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start+1; + *s->img_buffer = 0; + } else { + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start + n; + } +} + +stbi_inline static stbi_uc stbi__get8(stbi__context *s) +{ + if (s->img_buffer < s->img_buffer_end) + return *s->img_buffer++; + if (s->read_from_callbacks) { + stbi__refill_buffer(s); + return *s->img_buffer++; + } + return 0; +} + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_HDR) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +stbi_inline static int stbi__at_eof(stbi__context *s) +{ + if (s->io.read) { + if (!(s->io.eof)(s->io_user_data)) return 0; + // if feof() is true, check if buffer = end + // special case: we've only got the special 0 character at the end + if (s->read_from_callbacks == 0) return 1; + } + + return s->img_buffer >= s->img_buffer_end; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) +// nothing +#else +static void stbi__skip(stbi__context *s, int n) +{ + if (n == 0) return; // already there! + if (n < 0) { + s->img_buffer = s->img_buffer_end; + return; + } + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + s->img_buffer = s->img_buffer_end; + (s->io.skip)(s->io_user_data, n - blen); + return; + } + } + s->img_buffer += n; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_TGA) && defined(STBI_NO_HDR) && defined(STBI_NO_PNM) +// nothing +#else +static int stbi__getn(stbi__context *s, stbi_uc *buffer, int n) +{ + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + int res, count; + + memcpy(buffer, s->img_buffer, blen); + + count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen); + res = (count == (n-blen)); + s->img_buffer = s->img_buffer_end; + return res; + } + } + + if (s->img_buffer+n <= s->img_buffer_end) { + memcpy(buffer, s->img_buffer, n); + s->img_buffer += n; + return 1; + } else + return 0; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static int stbi__get16be(stbi__context *s) +{ + int z = stbi__get8(s); + return (z << 8) + stbi__get8(s); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static stbi__uint32 stbi__get32be(stbi__context *s) +{ + stbi__uint32 z = stbi__get16be(s); + return (z << 16) + stbi__get16be(s); +} +#endif + +#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) +// nothing +#else +static int stbi__get16le(stbi__context *s) +{ + int z = stbi__get8(s); + return z + (stbi__get8(s) << 8); +} +#endif + +#ifndef STBI_NO_BMP +static stbi__uint32 stbi__get32le(stbi__context *s) +{ + stbi__uint32 z = stbi__get16le(s); + z += (stbi__uint32)stbi__get16le(s) << 16; + return z; +} +#endif + +#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +////////////////////////////////////////////////////////////////////////////// +// +// generic converter from built-in img_n to req_comp +// individual types do this automatically as much as possible (e.g. jpeg +// does all cases internally since it needs to colorspace convert anyway, +// and it never has alpha, so very few cases ). png can automatically +// interleave an alpha=255 channel, but falls back to this for other cases +// +// assume data buffer is malloced, so malloc a new one and free that one +// only failure mode is malloc failing + +static stbi_uc stbi__compute_y(int r, int g, int b) +{ + return (stbi_uc) (((r*77) + (g*150) + (29*b)) >> 8); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + unsigned char *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (unsigned char *) stbi__malloc_mad3(req_comp, x, y, 0); + if (good == NULL) { + STBI_FREE(data); + return stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + unsigned char *src = data + j * x * img_n ; + unsigned char *dest = good + j * x * req_comp; + + #define STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 stbi__compute_y_16(int r, int g, int b) +{ + return (stbi__uint16) (((r*77) + (g*150) + (29*b)) >> 8); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 *stbi__convert_format16(stbi__uint16 *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + stbi__uint16 *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (stbi__uint16 *) stbi__malloc(req_comp * x * y * 2); + if (good == NULL) { + STBI_FREE(data); + return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + stbi__uint16 *src = data + j * x * img_n ; + stbi__uint16 *dest = good + j * x * req_comp; + + #define STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return (stbi__uint16*) stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp) +{ + int i,k,n; + float *output; + if (!data) return NULL; + output = (float *) stbi__malloc_mad4(x, y, comp, sizeof(float), 0); + if (output == NULL) { STBI_FREE(data); return stbi__errpf("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale); + } + } + if (n < comp) { + for (i=0; i < x*y; ++i) { + output[i*comp + n] = data[i*comp + n]/255.0f; + } + } + STBI_FREE(data); + return output; +} +#endif + +#ifndef STBI_NO_HDR +#define stbi__float2int(x) ((int) (x)) +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp) +{ + int i,k,n; + stbi_uc *output; + if (!data) return NULL; + output = (stbi_uc *) stbi__malloc_mad3(x, y, comp, 0); + if (output == NULL) { STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + if (k < comp) { + float z = data[i*comp+k] * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + } + STBI_FREE(data); + return output; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// "baseline" JPEG/JFIF decoder +// +// simple implementation +// - doesn't support delayed output of y-dimension +// - simple interface (only one output format: 8-bit interleaved RGB) +// - doesn't try to recover corrupt jpegs +// - doesn't allow partial loading, loading multiple at once +// - still fast on x86 (copying globals into locals doesn't help x86) +// - allocates lots of intermediate memory (full size of all components) +// - non-interleaved case requires this anyway +// - allows good upsampling (see next) +// high-quality +// - upsampled channels are bilinearly interpolated, even across blocks +// - quality integer IDCT derived from IJG's 'slow' +// performance +// - fast huffman; reasonable integer IDCT +// - some SIMD kernels for common paths on targets with SSE2/NEON +// - uses a lot of intermediate memory, could cache poorly + +#ifndef STBI_NO_JPEG + +// huffman decoding acceleration +#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache + +typedef struct +{ + stbi_uc fast[1 << FAST_BITS]; + // weirdly, repacking this into AoS is a 10% speed loss, instead of a win + stbi__uint16 code[256]; + stbi_uc values[256]; + stbi_uc size[257]; + unsigned int maxcode[18]; + int delta[17]; // old 'firstsymbol' - old 'firstcode' +} stbi__huffman; + +typedef struct +{ + stbi__context *s; + stbi__huffman huff_dc[4]; + stbi__huffman huff_ac[4]; + stbi__uint16 dequant[4][64]; + stbi__int16 fast_ac[4][1 << FAST_BITS]; + +// sizes for components, interleaved MCUs + int img_h_max, img_v_max; + int img_mcu_x, img_mcu_y; + int img_mcu_w, img_mcu_h; + +// definition of jpeg image component + struct + { + int id; + int h,v; + int tq; + int hd,ha; + int dc_pred; + + int x,y,w2,h2; + stbi_uc *data; + void *raw_data, *raw_coeff; + stbi_uc *linebuf; + short *coeff; // progressive only + int coeff_w, coeff_h; // number of 8x8 coefficient blocks + } img_comp[4]; + + stbi__uint32 code_buffer; // jpeg entropy-coded buffer + int code_bits; // number of valid bits + unsigned char marker; // marker seen while filling entropy buffer + int nomore; // flag if we saw a marker so must stop + + int progressive; + int spec_start; + int spec_end; + int succ_high; + int succ_low; + int eob_run; + int jfif; + int app14_color_transform; // Adobe APP14 tag + int rgb; + + int scan_n, order[4]; + int restart_interval, todo; + +// kernels + void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]); + void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step); + stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs); +} stbi__jpeg; + +static int stbi__build_huffman(stbi__huffman *h, int *count) +{ + int i,j,k=0; + unsigned int code; + // build size list for each symbol (from JPEG spec) + for (i=0; i < 16; ++i) + for (j=0; j < count[i]; ++j) + h->size[k++] = (stbi_uc) (i+1); + h->size[k] = 0; + + // compute actual symbols (from jpeg spec) + code = 0; + k = 0; + for(j=1; j <= 16; ++j) { + // compute delta to add to code to compute symbol id + h->delta[j] = k - code; + if (h->size[k] == j) { + while (h->size[k] == j) + h->code[k++] = (stbi__uint16) (code++); + if (code-1 >= (1u << j)) return stbi__err("bad code lengths","Corrupt JPEG"); + } + // compute largest code + 1 for this size, preshifted as needed later + h->maxcode[j] = code << (16-j); + code <<= 1; + } + h->maxcode[j] = 0xffffffff; + + // build non-spec acceleration table; 255 is flag for not-accelerated + memset(h->fast, 255, 1 << FAST_BITS); + for (i=0; i < k; ++i) { + int s = h->size[i]; + if (s <= FAST_BITS) { + int c = h->code[i] << (FAST_BITS-s); + int m = 1 << (FAST_BITS-s); + for (j=0; j < m; ++j) { + h->fast[c+j] = (stbi_uc) i; + } + } + } + return 1; +} + +// build a table that decodes both magnitude and value of small ACs in +// one go. +static void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h) +{ + int i; + for (i=0; i < (1 << FAST_BITS); ++i) { + stbi_uc fast = h->fast[i]; + fast_ac[i] = 0; + if (fast < 255) { + int rs = h->values[fast]; + int run = (rs >> 4) & 15; + int magbits = rs & 15; + int len = h->size[fast]; + + if (magbits && len + magbits <= FAST_BITS) { + // magnitude code followed by receive_extend code + int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits); + int m = 1 << (magbits - 1); + if (k < m) k += (~0U << magbits) + 1; + // if the result is small enough, we can fit it in fast_ac table + if (k >= -128 && k <= 127) + fast_ac[i] = (stbi__int16) ((k * 256) + (run * 16) + (len + magbits)); + } + } + } +} + +static void stbi__grow_buffer_unsafe(stbi__jpeg *j) +{ + do { + unsigned int b = j->nomore ? 0 : stbi__get8(j->s); + if (b == 0xff) { + int c = stbi__get8(j->s); + while (c == 0xff) c = stbi__get8(j->s); // consume fill bytes + if (c != 0) { + j->marker = (unsigned char) c; + j->nomore = 1; + return; + } + } + j->code_buffer |= b << (24 - j->code_bits); + j->code_bits += 8; + } while (j->code_bits <= 24); +} + +// (1 << n) - 1 +static const stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; + +// decode a jpeg huffman value from the bitstream +stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h) +{ + unsigned int temp; + int c,k; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + // look at the top FAST_BITS and determine what symbol ID it is, + // if the code is <= FAST_BITS + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + k = h->fast[c]; + if (k < 255) { + int s = h->size[k]; + if (s > j->code_bits) + return -1; + j->code_buffer <<= s; + j->code_bits -= s; + return h->values[k]; + } + + // naive test is to shift the code_buffer down so k bits are + // valid, then test against maxcode. To speed this up, we've + // preshifted maxcode left so that it has (16-k) 0s at the + // end; in other words, regardless of the number of bits, it + // wants to be compared against something shifted to have 16; + // that way we don't need to shift inside the loop. + temp = j->code_buffer >> 16; + for (k=FAST_BITS+1 ; ; ++k) + if (temp < h->maxcode[k]) + break; + if (k == 17) { + // error! code not found + j->code_bits -= 16; + return -1; + } + + if (k > j->code_bits) + return -1; + + // convert the huffman code to the symbol id + c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k]; + STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]); + + // convert the id to a symbol + j->code_bits -= k; + j->code_buffer <<= k; + return h->values[c]; +} + +// bias[n] = (-1<code_bits < n) stbi__grow_buffer_unsafe(j); + + sgn = j->code_buffer >> 31; // sign bit always in MSB; 0 if MSB clear (positive), 1 if MSB set (negative) + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k + (stbi__jbias[n] & (sgn - 1)); +} + +// get some unsigned bits +stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n) +{ + unsigned int k; + if (j->code_bits < n) stbi__grow_buffer_unsafe(j); + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k; +} + +stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j) +{ + unsigned int k; + if (j->code_bits < 1) stbi__grow_buffer_unsafe(j); + k = j->code_buffer; + j->code_buffer <<= 1; + --j->code_bits; + return k & 0x80000000; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static const stbi_uc stbi__jpeg_dezigzag[64+15] = +{ + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + // let corrupt input sample past end + 63, 63, 63, 63, 63, 63, 63, 63, + 63, 63, 63, 63, 63, 63, 63 +}; + +// decode one 64-entry block-- +static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi__uint16 *dequant) +{ + int diff,dc,k; + int t; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + t = stbi__jpeg_huff_decode(j, hdc); + if (t < 0 || t > 15) return stbi__err("bad huffman code","Corrupt JPEG"); + + // 0 all the ac values now so we can do it 32-bits at a time + memset(data,0,64*sizeof(data[0])); + + diff = t ? stbi__extend_receive(j, t) : 0; + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + data[0] = (short) (dc * dequant[0]); + + // decode AC components, see JPEG spec + k = 1; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + j->code_buffer <<= s; + j->code_bits -= s; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) * dequant[zig]); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (rs != 0xf0) break; // end block + k += 16; + } else { + k += r; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]); + } + } + } while (k < 64); + return 1; +} + +static int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b) +{ + int diff,dc; + int t; + if (j->spec_end != 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + if (j->succ_high == 0) { + // first scan for DC coefficient, must be first + memset(data,0,64*sizeof(data[0])); // 0 all the ac values now + t = stbi__jpeg_huff_decode(j, hdc); + if (t < 0 || t > 15) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + diff = t ? stbi__extend_receive(j, t) : 0; + + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + data[0] = (short) (dc * (1 << j->succ_low)); + } else { + // refinement scan for DC coefficient + if (stbi__jpeg_get_bit(j)) + data[0] += (short) (1 << j->succ_low); + } + return 1; +} + +// @OPTIMIZE: store non-zigzagged during the decode passes, +// and only de-zigzag when dequantizing +static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac) +{ + int k; + if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->succ_high == 0) { + int shift = j->succ_low; + + if (j->eob_run) { + --j->eob_run; + return 1; + } + + k = j->spec_start; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + j->code_buffer <<= s; + j->code_bits -= s; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) * (1 << shift)); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r); + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + --j->eob_run; + break; + } + k += 16; + } else { + k += r; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) * (1 << shift)); + } + } + } while (k <= j->spec_end); + } else { + // refinement scan for these AC coefficients + + short bit = (short) (1 << j->succ_low); + + if (j->eob_run) { + --j->eob_run; + for (k = j->spec_start; k <= j->spec_end; ++k) { + short *p = &data[stbi__jpeg_dezigzag[k]]; + if (*p != 0) + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } + } else { + k = j->spec_start; + do { + int r,s; + int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r) - 1; + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + r = 64; // force end of block + } else { + // r=15 s=0 should write 16 0s, so we just do + // a run of 15 0s and then write s (which is 0), + // so we don't have to do anything special here + } + } else { + if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG"); + // sign bit + if (stbi__jpeg_get_bit(j)) + s = bit; + else + s = -bit; + } + + // advance by r + while (k <= j->spec_end) { + short *p = &data[stbi__jpeg_dezigzag[k++]]; + if (*p != 0) { + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } else { + if (r == 0) { + *p = (short) s; + break; + } + --r; + } + } + } while (k <= j->spec_end); + } + } + return 1; +} + +// take a -128..127 value and stbi__clamp it and convert to 0..255 +stbi_inline static stbi_uc stbi__clamp(int x) +{ + // trick to use a single test to catch both cases + if ((unsigned int) x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (stbi_uc) x; +} + +#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5))) +#define stbi__fsh(x) ((x) * 4096) + +// derived from jidctint -- DCT_ISLOW +#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ + int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \ + p2 = s2; \ + p3 = s6; \ + p1 = (p2+p3) * stbi__f2f(0.5411961f); \ + t2 = p1 + p3*stbi__f2f(-1.847759065f); \ + t3 = p1 + p2*stbi__f2f( 0.765366865f); \ + p2 = s0; \ + p3 = s4; \ + t0 = stbi__fsh(p2+p3); \ + t1 = stbi__fsh(p2-p3); \ + x0 = t0+t3; \ + x3 = t0-t3; \ + x1 = t1+t2; \ + x2 = t1-t2; \ + t0 = s7; \ + t1 = s5; \ + t2 = s3; \ + t3 = s1; \ + p3 = t0+t2; \ + p4 = t1+t3; \ + p1 = t0+t3; \ + p2 = t1+t2; \ + p5 = (p3+p4)*stbi__f2f( 1.175875602f); \ + t0 = t0*stbi__f2f( 0.298631336f); \ + t1 = t1*stbi__f2f( 2.053119869f); \ + t2 = t2*stbi__f2f( 3.072711026f); \ + t3 = t3*stbi__f2f( 1.501321110f); \ + p1 = p5 + p1*stbi__f2f(-0.899976223f); \ + p2 = p5 + p2*stbi__f2f(-2.562915447f); \ + p3 = p3*stbi__f2f(-1.961570560f); \ + p4 = p4*stbi__f2f(-0.390180644f); \ + t3 += p1+p4; \ + t2 += p2+p3; \ + t1 += p2+p4; \ + t0 += p1+p3; + +static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64]) +{ + int i,val[64],*v=val; + stbi_uc *o; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0]*4; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + // so we want to round that, which means adding 0.5 * 1<<17, + // aka 65536. Also, we'll end up with -128 to 127 that we want + // to encode as 0..255 by adding 128, so we'll add that before the shift + x0 += 65536 + (128<<17); + x1 += 65536 + (128<<17); + x2 += 65536 + (128<<17); + x3 += 65536 + (128<<17); + // tried computing the shifts into temps, or'ing the temps to see + // if any were out of range, but that was slower + o[0] = stbi__clamp((x0+t3) >> 17); + o[7] = stbi__clamp((x0-t3) >> 17); + o[1] = stbi__clamp((x1+t2) >> 17); + o[6] = stbi__clamp((x1-t2) >> 17); + o[2] = stbi__clamp((x2+t1) >> 17); + o[5] = stbi__clamp((x2-t1) >> 17); + o[3] = stbi__clamp((x3+t0) >> 17); + o[4] = stbi__clamp((x3-t0) >> 17); + } +} + +#ifdef STBI_SSE2 +// sse2 integer IDCT. not the fastest possible implementation but it +// produces bit-identical results to the generic C version so it's +// fully "transparent". +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + // This is constructed to match our regular (generic) integer IDCT exactly. + __m128i row0, row1, row2, row3, row4, row5, row6, row7; + __m128i tmp; + + // dot product constant: even elems=x, odd elems=y + #define dct_const(x,y) _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y)) + + // out(0) = c0[even]*x + c0[odd]*y (c0, x, y 16-bit, out 32-bit) + // out(1) = c1[even]*x + c1[odd]*y + #define dct_rot(out0,out1, x,y,c0,c1) \ + __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \ + __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \ + __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \ + __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \ + __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \ + __m128i out1##_h = _mm_madd_epi16(c0##hi, c1) + + // out = in << 12 (in 16-bit, out 32-bit) + #define dct_widen(out, in) \ + __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \ + __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4) + + // wide add + #define dct_wadd(out, a, b) \ + __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_add_epi32(a##_h, b##_h) + + // wide sub + #define dct_wsub(out, a, b) \ + __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_sub_epi32(a##_h, b##_h) + + // butterfly a/b, add bias, then shift by "s" and pack + #define dct_bfly32o(out0, out1, a,b,bias,s) \ + { \ + __m128i abiased_l = _mm_add_epi32(a##_l, bias); \ + __m128i abiased_h = _mm_add_epi32(a##_h, bias); \ + dct_wadd(sum, abiased, b); \ + dct_wsub(dif, abiased, b); \ + out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \ + out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \ + } + + // 8-bit interleave step (for transposes) + #define dct_interleave8(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi8(a, b); \ + b = _mm_unpackhi_epi8(tmp, b) + + // 16-bit interleave step (for transposes) + #define dct_interleave16(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi16(a, b); \ + b = _mm_unpackhi_epi16(tmp, b) + + #define dct_pass(bias,shift) \ + { \ + /* even part */ \ + dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \ + __m128i sum04 = _mm_add_epi16(row0, row4); \ + __m128i dif04 = _mm_sub_epi16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \ + dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \ + __m128i sum17 = _mm_add_epi16(row1, row7); \ + __m128i sum35 = _mm_add_epi16(row3, row5); \ + dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \ + dct_wadd(x4, y0o, y4o); \ + dct_wadd(x5, y1o, y5o); \ + dct_wadd(x6, y2o, y5o); \ + dct_wadd(x7, y3o, y4o); \ + dct_bfly32o(row0,row7, x0,x7,bias,shift); \ + dct_bfly32o(row1,row6, x1,x6,bias,shift); \ + dct_bfly32o(row2,row5, x2,x5,bias,shift); \ + dct_bfly32o(row3,row4, x3,x4,bias,shift); \ + } + + __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f)); + __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f)); + __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f)); + __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f)); + __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f)); + __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f)); + __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f)); + __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f)); + + // rounding biases in column/row passes, see stbi__idct_block for explanation. + __m128i bias_0 = _mm_set1_epi32(512); + __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17)); + + // load + row0 = _mm_load_si128((const __m128i *) (data + 0*8)); + row1 = _mm_load_si128((const __m128i *) (data + 1*8)); + row2 = _mm_load_si128((const __m128i *) (data + 2*8)); + row3 = _mm_load_si128((const __m128i *) (data + 3*8)); + row4 = _mm_load_si128((const __m128i *) (data + 4*8)); + row5 = _mm_load_si128((const __m128i *) (data + 5*8)); + row6 = _mm_load_si128((const __m128i *) (data + 6*8)); + row7 = _mm_load_si128((const __m128i *) (data + 7*8)); + + // column pass + dct_pass(bias_0, 10); + + { + // 16bit 8x8 transpose pass 1 + dct_interleave16(row0, row4); + dct_interleave16(row1, row5); + dct_interleave16(row2, row6); + dct_interleave16(row3, row7); + + // transpose pass 2 + dct_interleave16(row0, row2); + dct_interleave16(row1, row3); + dct_interleave16(row4, row6); + dct_interleave16(row5, row7); + + // transpose pass 3 + dct_interleave16(row0, row1); + dct_interleave16(row2, row3); + dct_interleave16(row4, row5); + dct_interleave16(row6, row7); + } + + // row pass + dct_pass(bias_1, 17); + + { + // pack + __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7 + __m128i p1 = _mm_packus_epi16(row2, row3); + __m128i p2 = _mm_packus_epi16(row4, row5); + __m128i p3 = _mm_packus_epi16(row6, row7); + + // 8bit 8x8 transpose pass 1 + dct_interleave8(p0, p2); // a0e0a1e1... + dct_interleave8(p1, p3); // c0g0c1g1... + + // transpose pass 2 + dct_interleave8(p0, p1); // a0c0e0g0... + dct_interleave8(p2, p3); // b0d0f0h0... + + // transpose pass 3 + dct_interleave8(p0, p2); // a0b0c0d0... + dct_interleave8(p1, p3); // a4b4c4d4... + + // store + _mm_storel_epi64((__m128i *) out, p0); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p2); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p1); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p3); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e)); + } + +#undef dct_const +#undef dct_rot +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_interleave8 +#undef dct_interleave16 +#undef dct_pass +} + +#endif // STBI_SSE2 + +#ifdef STBI_NEON + +// NEON integer IDCT. should produce bit-identical +// results to the generic C version. +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + int16x8_t row0, row1, row2, row3, row4, row5, row6, row7; + + int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f)); + int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f)); + int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f)); + int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f)); + int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f)); + int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f)); + int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f)); + int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f)); + int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f)); + int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f)); + int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f)); + int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f)); + +#define dct_long_mul(out, inq, coeff) \ + int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff) + +#define dct_long_mac(out, acc, inq, coeff) \ + int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff) + +#define dct_widen(out, inq) \ + int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \ + int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12) + +// wide add +#define dct_wadd(out, a, b) \ + int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vaddq_s32(a##_h, b##_h) + +// wide sub +#define dct_wsub(out, a, b) \ + int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vsubq_s32(a##_h, b##_h) + +// butterfly a/b, then shift using "shiftop" by "s" and pack +#define dct_bfly32o(out0,out1, a,b,shiftop,s) \ + { \ + dct_wadd(sum, a, b); \ + dct_wsub(dif, a, b); \ + out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \ + out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \ + } + +#define dct_pass(shiftop, shift) \ + { \ + /* even part */ \ + int16x8_t sum26 = vaddq_s16(row2, row6); \ + dct_long_mul(p1e, sum26, rot0_0); \ + dct_long_mac(t2e, p1e, row6, rot0_1); \ + dct_long_mac(t3e, p1e, row2, rot0_2); \ + int16x8_t sum04 = vaddq_s16(row0, row4); \ + int16x8_t dif04 = vsubq_s16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + int16x8_t sum15 = vaddq_s16(row1, row5); \ + int16x8_t sum17 = vaddq_s16(row1, row7); \ + int16x8_t sum35 = vaddq_s16(row3, row5); \ + int16x8_t sum37 = vaddq_s16(row3, row7); \ + int16x8_t sumodd = vaddq_s16(sum17, sum35); \ + dct_long_mul(p5o, sumodd, rot1_0); \ + dct_long_mac(p1o, p5o, sum17, rot1_1); \ + dct_long_mac(p2o, p5o, sum35, rot1_2); \ + dct_long_mul(p3o, sum37, rot2_0); \ + dct_long_mul(p4o, sum15, rot2_1); \ + dct_wadd(sump13o, p1o, p3o); \ + dct_wadd(sump24o, p2o, p4o); \ + dct_wadd(sump23o, p2o, p3o); \ + dct_wadd(sump14o, p1o, p4o); \ + dct_long_mac(x4, sump13o, row7, rot3_0); \ + dct_long_mac(x5, sump24o, row5, rot3_1); \ + dct_long_mac(x6, sump23o, row3, rot3_2); \ + dct_long_mac(x7, sump14o, row1, rot3_3); \ + dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \ + dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \ + dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \ + dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \ + } + + // load + row0 = vld1q_s16(data + 0*8); + row1 = vld1q_s16(data + 1*8); + row2 = vld1q_s16(data + 2*8); + row3 = vld1q_s16(data + 3*8); + row4 = vld1q_s16(data + 4*8); + row5 = vld1q_s16(data + 5*8); + row6 = vld1q_s16(data + 6*8); + row7 = vld1q_s16(data + 7*8); + + // add DC bias + row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0)); + + // column pass + dct_pass(vrshrn_n_s32, 10); + + // 16bit 8x8 transpose + { +// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively. +// whether compilers actually get this is another story, sadly. +#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); } +#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); } + + // pass 1 + dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6 + dct_trn16(row2, row3); + dct_trn16(row4, row5); + dct_trn16(row6, row7); + + // pass 2 + dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4 + dct_trn32(row1, row3); + dct_trn32(row4, row6); + dct_trn32(row5, row7); + + // pass 3 + dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0 + dct_trn64(row1, row5); + dct_trn64(row2, row6); + dct_trn64(row3, row7); + +#undef dct_trn16 +#undef dct_trn32 +#undef dct_trn64 + } + + // row pass + // vrshrn_n_s32 only supports shifts up to 16, we need + // 17. so do a non-rounding shift of 16 first then follow + // up with a rounding shift by 1. + dct_pass(vshrn_n_s32, 16); + + { + // pack and round + uint8x8_t p0 = vqrshrun_n_s16(row0, 1); + uint8x8_t p1 = vqrshrun_n_s16(row1, 1); + uint8x8_t p2 = vqrshrun_n_s16(row2, 1); + uint8x8_t p3 = vqrshrun_n_s16(row3, 1); + uint8x8_t p4 = vqrshrun_n_s16(row4, 1); + uint8x8_t p5 = vqrshrun_n_s16(row5, 1); + uint8x8_t p6 = vqrshrun_n_s16(row6, 1); + uint8x8_t p7 = vqrshrun_n_s16(row7, 1); + + // again, these can translate into one instruction, but often don't. +#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); } +#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); } + + // sadly can't use interleaved stores here since we only write + // 8 bytes to each scan line! + + // 8x8 8-bit transpose pass 1 + dct_trn8_8(p0, p1); + dct_trn8_8(p2, p3); + dct_trn8_8(p4, p5); + dct_trn8_8(p6, p7); + + // pass 2 + dct_trn8_16(p0, p2); + dct_trn8_16(p1, p3); + dct_trn8_16(p4, p6); + dct_trn8_16(p5, p7); + + // pass 3 + dct_trn8_32(p0, p4); + dct_trn8_32(p1, p5); + dct_trn8_32(p2, p6); + dct_trn8_32(p3, p7); + + // store + vst1_u8(out, p0); out += out_stride; + vst1_u8(out, p1); out += out_stride; + vst1_u8(out, p2); out += out_stride; + vst1_u8(out, p3); out += out_stride; + vst1_u8(out, p4); out += out_stride; + vst1_u8(out, p5); out += out_stride; + vst1_u8(out, p6); out += out_stride; + vst1_u8(out, p7); + +#undef dct_trn8_8 +#undef dct_trn8_16 +#undef dct_trn8_32 + } + +#undef dct_long_mul +#undef dct_long_mac +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_pass +} + +#endif // STBI_NEON + +#define STBI__MARKER_none 0xff +// if there's a pending marker from the entropy stream, return that +// otherwise, fetch from the stream and get a marker. if there's no +// marker, return 0xff, which is never a valid marker value +static stbi_uc stbi__get_marker(stbi__jpeg *j) +{ + stbi_uc x; + if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; } + x = stbi__get8(j->s); + if (x != 0xff) return STBI__MARKER_none; + while (x == 0xff) + x = stbi__get8(j->s); // consume repeated 0xff fill bytes + return x; +} + +// in each scan, we'll have scan_n components, and the order +// of the components is specified by order[] +#define STBI__RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7) + +// after a restart interval, stbi__jpeg_reset the entropy decoder and +// the dc prediction +static void stbi__jpeg_reset(stbi__jpeg *j) +{ + j->code_bits = 0; + j->code_buffer = 0; + j->nomore = 0; + j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = j->img_comp[3].dc_pred = 0; + j->marker = STBI__MARKER_none; + j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff; + j->eob_run = 0; + // no more than 1<<31 MCUs if no restart_interal? that's plenty safe, + // since we don't even allow 1<<30 pixels +} + +static int stbi__parse_entropy_coded_data(stbi__jpeg *z) +{ + stbi__jpeg_reset(z); + if (!z->progressive) { + if (z->scan_n == 1) { + int i,j; + STBI_SIMD_ALIGN(short, data[64]); + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + STBI_SIMD_ALIGN(short, data[64]); + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x)*8; + int y2 = (j*z->img_comp[n].v + y)*8; + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data); + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } else { + if (z->scan_n == 1) { + int i,j; + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + if (z->spec_start == 0) { + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } else { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha])) + return 0; + } + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x); + int y2 = (j*z->img_comp[n].v + y); + short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w); + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } +} + +static void stbi__jpeg_dequantize(short *data, stbi__uint16 *dequant) +{ + int i; + for (i=0; i < 64; ++i) + data[i] *= dequant[i]; +} + +static void stbi__jpeg_finish(stbi__jpeg *z) +{ + if (z->progressive) { + // dequantize and idct the data + int i,j,n; + for (n=0; n < z->s->img_n; ++n) { + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]); + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + } + } + } + } +} + +static int stbi__process_marker(stbi__jpeg *z, int m) +{ + int L; + switch (m) { + case STBI__MARKER_none: // no marker found + return stbi__err("expected marker","Corrupt JPEG"); + + case 0xDD: // DRI - specify restart interval + if (stbi__get16be(z->s) != 4) return stbi__err("bad DRI len","Corrupt JPEG"); + z->restart_interval = stbi__get16be(z->s); + return 1; + + case 0xDB: // DQT - define quantization table + L = stbi__get16be(z->s)-2; + while (L > 0) { + int q = stbi__get8(z->s); + int p = q >> 4, sixteen = (p != 0); + int t = q & 15,i; + if (p != 0 && p != 1) return stbi__err("bad DQT type","Corrupt JPEG"); + if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG"); + + for (i=0; i < 64; ++i) + z->dequant[t][stbi__jpeg_dezigzag[i]] = (stbi__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s)); + L -= (sixteen ? 129 : 65); + } + return L==0; + + case 0xC4: // DHT - define huffman table + L = stbi__get16be(z->s)-2; + while (L > 0) { + stbi_uc *v; + int sizes[16],i,n=0; + int q = stbi__get8(z->s); + int tc = q >> 4; + int th = q & 15; + if (tc > 1 || th > 3) return stbi__err("bad DHT header","Corrupt JPEG"); + for (i=0; i < 16; ++i) { + sizes[i] = stbi__get8(z->s); + n += sizes[i]; + } + L -= 17; + if (tc == 0) { + if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0; + v = z->huff_dc[th].values; + } else { + if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0; + v = z->huff_ac[th].values; + } + for (i=0; i < n; ++i) + v[i] = stbi__get8(z->s); + if (tc != 0) + stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th); + L -= n; + } + return L==0; + } + + // check for comment block or APP blocks + if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) { + L = stbi__get16be(z->s); + if (L < 2) { + if (m == 0xFE) + return stbi__err("bad COM len","Corrupt JPEG"); + else + return stbi__err("bad APP len","Corrupt JPEG"); + } + L -= 2; + + if (m == 0xE0 && L >= 5) { // JFIF APP0 segment + static const unsigned char tag[5] = {'J','F','I','F','\0'}; + int ok = 1; + int i; + for (i=0; i < 5; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 5; + if (ok) + z->jfif = 1; + } else if (m == 0xEE && L >= 12) { // Adobe APP14 segment + static const unsigned char tag[6] = {'A','d','o','b','e','\0'}; + int ok = 1; + int i; + for (i=0; i < 6; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 6; + if (ok) { + stbi__get8(z->s); // version + stbi__get16be(z->s); // flags0 + stbi__get16be(z->s); // flags1 + z->app14_color_transform = stbi__get8(z->s); // color transform + L -= 6; + } + } + + stbi__skip(z->s, L); + return 1; + } + + return stbi__err("unknown marker","Corrupt JPEG"); +} + +// after we see SOS +static int stbi__process_scan_header(stbi__jpeg *z) +{ + int i; + int Ls = stbi__get16be(z->s); + z->scan_n = stbi__get8(z->s); + if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err("bad SOS component count","Corrupt JPEG"); + if (Ls != 6+2*z->scan_n) return stbi__err("bad SOS len","Corrupt JPEG"); + for (i=0; i < z->scan_n; ++i) { + int id = stbi__get8(z->s), which; + int q = stbi__get8(z->s); + for (which = 0; which < z->s->img_n; ++which) + if (z->img_comp[which].id == id) + break; + if (which == z->s->img_n) return 0; // no match + z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return stbi__err("bad DC huff","Corrupt JPEG"); + z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return stbi__err("bad AC huff","Corrupt JPEG"); + z->order[i] = which; + } + + { + int aa; + z->spec_start = stbi__get8(z->s); + z->spec_end = stbi__get8(z->s); // should be 63, but might be 0 + aa = stbi__get8(z->s); + z->succ_high = (aa >> 4); + z->succ_low = (aa & 15); + if (z->progressive) { + if (z->spec_start > 63 || z->spec_end > 63 || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13) + return stbi__err("bad SOS", "Corrupt JPEG"); + } else { + if (z->spec_start != 0) return stbi__err("bad SOS","Corrupt JPEG"); + if (z->succ_high != 0 || z->succ_low != 0) return stbi__err("bad SOS","Corrupt JPEG"); + z->spec_end = 63; + } + } + + return 1; +} + +static int stbi__free_jpeg_components(stbi__jpeg *z, int ncomp, int why) +{ + int i; + for (i=0; i < ncomp; ++i) { + if (z->img_comp[i].raw_data) { + STBI_FREE(z->img_comp[i].raw_data); + z->img_comp[i].raw_data = NULL; + z->img_comp[i].data = NULL; + } + if (z->img_comp[i].raw_coeff) { + STBI_FREE(z->img_comp[i].raw_coeff); + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].coeff = 0; + } + if (z->img_comp[i].linebuf) { + STBI_FREE(z->img_comp[i].linebuf); + z->img_comp[i].linebuf = NULL; + } + } + return why; +} + +static int stbi__process_frame_header(stbi__jpeg *z, int scan) +{ + stbi__context *s = z->s; + int Lf,p,i,q, h_max=1,v_max=1,c; + Lf = stbi__get16be(s); if (Lf < 11) return stbi__err("bad SOF len","Corrupt JPEG"); // JPEG + p = stbi__get8(s); if (p != 8) return stbi__err("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline + s->img_y = stbi__get16be(s); if (s->img_y == 0) return stbi__err("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG + s->img_x = stbi__get16be(s); if (s->img_x == 0) return stbi__err("0 width","Corrupt JPEG"); // JPEG requires + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + c = stbi__get8(s); + if (c != 3 && c != 1 && c != 4) return stbi__err("bad component count","Corrupt JPEG"); + s->img_n = c; + for (i=0; i < c; ++i) { + z->img_comp[i].data = NULL; + z->img_comp[i].linebuf = NULL; + } + + if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG"); + + z->rgb = 0; + for (i=0; i < s->img_n; ++i) { + static const unsigned char rgb[3] = { 'R', 'G', 'B' }; + z->img_comp[i].id = stbi__get8(s); + if (s->img_n == 3 && z->img_comp[i].id == rgb[i]) + ++z->rgb; + q = stbi__get8(s); + z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err("bad H","Corrupt JPEG"); + z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err("bad V","Corrupt JPEG"); + z->img_comp[i].tq = stbi__get8(s); if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG"); + } + + if (scan != STBI__SCAN_load) return 1; + + if (!stbi__mad3sizes_valid(s->img_x, s->img_y, s->img_n, 0)) return stbi__err("too large", "Image too large to decode"); + + for (i=0; i < s->img_n; ++i) { + if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h; + if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v; + } + + // check that plane subsampling factors are integer ratios; our resamplers can't deal with fractional ratios + // and I've never seen a non-corrupted JPEG file actually use them + for (i=0; i < s->img_n; ++i) { + if (h_max % z->img_comp[i].h != 0) return stbi__err("bad H","Corrupt JPEG"); + if (v_max % z->img_comp[i].v != 0) return stbi__err("bad V","Corrupt JPEG"); + } + + // compute interleaved mcu info + z->img_h_max = h_max; + z->img_v_max = v_max; + z->img_mcu_w = h_max * 8; + z->img_mcu_h = v_max * 8; + // these sizes can't be more than 17 bits + z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w; + z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h; + + for (i=0; i < s->img_n; ++i) { + // number of effective pixels (e.g. for non-interleaved MCU) + z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max; + z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max; + // to simplify generation, we'll allocate enough memory to decode + // the bogus oversized data from using interleaved MCUs and their + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't + // discard the extra data until colorspace conversion + // + // img_mcu_x, img_mcu_y: <=17 bits; comp[i].h and .v are <=4 (checked earlier) + // so these muls can't overflow with 32-bit ints (which we require) + z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8; + z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8; + z->img_comp[i].coeff = 0; + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].linebuf = NULL; + z->img_comp[i].raw_data = stbi__malloc_mad2(z->img_comp[i].w2, z->img_comp[i].h2, 15); + if (z->img_comp[i].raw_data == NULL) + return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory")); + // align blocks for idct using mmx/sse + z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15); + if (z->progressive) { + // w2, h2 are multiples of 8 (see above) + z->img_comp[i].coeff_w = z->img_comp[i].w2 / 8; + z->img_comp[i].coeff_h = z->img_comp[i].h2 / 8; + z->img_comp[i].raw_coeff = stbi__malloc_mad3(z->img_comp[i].w2, z->img_comp[i].h2, sizeof(short), 15); + if (z->img_comp[i].raw_coeff == NULL) + return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory")); + z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15); + } + } + + return 1; +} + +// use comparisons since in some cases we handle more than one case (e.g. SOF) +#define stbi__DNL(x) ((x) == 0xdc) +#define stbi__SOI(x) ((x) == 0xd8) +#define stbi__EOI(x) ((x) == 0xd9) +#define stbi__SOF(x) ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2) +#define stbi__SOS(x) ((x) == 0xda) + +#define stbi__SOF_progressive(x) ((x) == 0xc2) + +static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan) +{ + int m; + z->jfif = 0; + z->app14_color_transform = -1; // valid values are 0,1,2 + z->marker = STBI__MARKER_none; // initialize cached marker to empty + m = stbi__get_marker(z); + if (!stbi__SOI(m)) return stbi__err("no SOI","Corrupt JPEG"); + if (scan == STBI__SCAN_type) return 1; + m = stbi__get_marker(z); + while (!stbi__SOF(m)) { + if (!stbi__process_marker(z,m)) return 0; + m = stbi__get_marker(z); + while (m == STBI__MARKER_none) { + // some files have extra padding after their blocks, so ok, we'll scan + if (stbi__at_eof(z->s)) return stbi__err("no SOF", "Corrupt JPEG"); + m = stbi__get_marker(z); + } + } + z->progressive = stbi__SOF_progressive(m); + if (!stbi__process_frame_header(z, scan)) return 0; + return 1; +} + +// decode image to YCbCr format +static int stbi__decode_jpeg_image(stbi__jpeg *j) +{ + int m; + for (m = 0; m < 4; m++) { + j->img_comp[m].raw_data = NULL; + j->img_comp[m].raw_coeff = NULL; + } + j->restart_interval = 0; + if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0; + m = stbi__get_marker(j); + while (!stbi__EOI(m)) { + if (stbi__SOS(m)) { + if (!stbi__process_scan_header(j)) return 0; + if (!stbi__parse_entropy_coded_data(j)) return 0; + if (j->marker == STBI__MARKER_none ) { + // handle 0s at the end of image data from IP Kamera 9060 + while (!stbi__at_eof(j->s)) { + int x = stbi__get8(j->s); + if (x == 255) { + j->marker = stbi__get8(j->s); + break; + } + } + // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0 + } + } else if (stbi__DNL(m)) { + int Ld = stbi__get16be(j->s); + stbi__uint32 NL = stbi__get16be(j->s); + if (Ld != 4) return stbi__err("bad DNL len", "Corrupt JPEG"); + if (NL != j->s->img_y) return stbi__err("bad DNL height", "Corrupt JPEG"); + } else { + if (!stbi__process_marker(j, m)) return 0; + } + m = stbi__get_marker(j); + } + if (j->progressive) + stbi__jpeg_finish(j); + return 1; +} + +// static jfif-centered resampling (across block boundaries) + +typedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1, + int w, int hs); + +#define stbi__div4(x) ((stbi_uc) ((x) >> 2)) + +static stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + STBI_NOTUSED(out); + STBI_NOTUSED(in_far); + STBI_NOTUSED(w); + STBI_NOTUSED(hs); + return in_near; +} + +static stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples vertically for every one in input + int i; + STBI_NOTUSED(hs); + for (i=0; i < w; ++i) + out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2); + return out; +} + +static stbi_uc* stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples horizontally for every one in input + int i; + stbi_uc *input = in_near; + + if (w == 1) { + // if only one sample, can't do any interpolation + out[0] = out[1] = input[0]; + return out; + } + + out[0] = input[0]; + out[1] = stbi__div4(input[0]*3 + input[1] + 2); + for (i=1; i < w-1; ++i) { + int n = 3*input[i]+2; + out[i*2+0] = stbi__div4(n+input[i-1]); + out[i*2+1] = stbi__div4(n+input[i+1]); + } + out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2); + out[i*2+1] = input[w-1]; + + STBI_NOTUSED(in_far); + STBI_NOTUSED(hs); + + return out; +} + +#define stbi__div16(x) ((stbi_uc) ((x) >> 4)) + +static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i,t0,t1; + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + out[0] = stbi__div4(t1+2); + for (i=1; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i=0,t0,t1; + + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + // process groups of 8 pixels for as long as we can. + // note we can't handle the last pixel in a row in this loop + // because we need to handle the filter boundary conditions. + for (; i < ((w-1) & ~7); i += 8) { +#if defined(STBI_SSE2) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + __m128i zero = _mm_setzero_si128(); + __m128i farb = _mm_loadl_epi64((__m128i *) (in_far + i)); + __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i)); + __m128i farw = _mm_unpacklo_epi8(farb, zero); + __m128i nearw = _mm_unpacklo_epi8(nearb, zero); + __m128i diff = _mm_sub_epi16(farw, nearw); + __m128i nears = _mm_slli_epi16(nearw, 2); + __m128i curr = _mm_add_epi16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + __m128i prv0 = _mm_slli_si128(curr, 2); + __m128i nxt0 = _mm_srli_si128(curr, 2); + __m128i prev = _mm_insert_epi16(prv0, t1, 0); + __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + __m128i bias = _mm_set1_epi16(8); + __m128i curs = _mm_slli_epi16(curr, 2); + __m128i prvd = _mm_sub_epi16(prev, curr); + __m128i nxtd = _mm_sub_epi16(next, curr); + __m128i curb = _mm_add_epi16(curs, bias); + __m128i even = _mm_add_epi16(prvd, curb); + __m128i odd = _mm_add_epi16(nxtd, curb); + + // interleave even and odd pixels, then undo scaling. + __m128i int0 = _mm_unpacklo_epi16(even, odd); + __m128i int1 = _mm_unpackhi_epi16(even, odd); + __m128i de0 = _mm_srli_epi16(int0, 4); + __m128i de1 = _mm_srli_epi16(int1, 4); + + // pack and write output + __m128i outv = _mm_packus_epi16(de0, de1); + _mm_storeu_si128((__m128i *) (out + i*2), outv); +#elif defined(STBI_NEON) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + uint8x8_t farb = vld1_u8(in_far + i); + uint8x8_t nearb = vld1_u8(in_near + i); + int16x8_t diff = vreinterpretq_s16_u16(vsubl_u8(farb, nearb)); + int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2)); + int16x8_t curr = vaddq_s16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + int16x8_t prv0 = vextq_s16(curr, curr, 7); + int16x8_t nxt0 = vextq_s16(curr, curr, 1); + int16x8_t prev = vsetq_lane_s16(t1, prv0, 0); + int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + int16x8_t curs = vshlq_n_s16(curr, 2); + int16x8_t prvd = vsubq_s16(prev, curr); + int16x8_t nxtd = vsubq_s16(next, curr); + int16x8_t even = vaddq_s16(curs, prvd); + int16x8_t odd = vaddq_s16(curs, nxtd); + + // undo scaling and round, then store with even/odd phases interleaved + uint8x8x2_t o; + o.val[0] = vqrshrun_n_s16(even, 4); + o.val[1] = vqrshrun_n_s16(odd, 4); + vst2_u8(out + i*2, o); +#endif + + // "previous" value for next iter + t1 = 3*in_near[i+7] + in_far[i+7]; + } + + t0 = t1; + t1 = 3*in_near[i] + in_far[i]; + out[i*2] = stbi__div16(3*t1 + t0 + 8); + + for (++i; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} +#endif + +static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // resample with nearest-neighbor + int i,j; + STBI_NOTUSED(in_far); + for (i=0; i < w; ++i) + for (j=0; j < hs; ++j) + out[i*hs+j] = in_near[i]; + return out; +} + +// this is a reduced-precision calculation of YCbCr-to-RGB introduced +// to make sure the code produces the same results in both SIMD and scalar +#define stbi__float2fixed(x) (((int) ((x) * 4096.0f + 0.5f)) << 8) +static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* stbi__float2fixed(1.40200f); + g = y_fixed + (cr*-stbi__float2fixed(0.71414f)) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step) +{ + int i = 0; + +#ifdef STBI_SSE2 + // step == 3 is pretty ugly on the final interleave, and i'm not convinced + // it's useful in practice (you wouldn't use it for textures, for example). + // so just accelerate step == 4 case. + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + __m128i signflip = _mm_set1_epi8(-0x80); + __m128i cr_const0 = _mm_set1_epi16( (short) ( 1.40200f*4096.0f+0.5f)); + __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f)); + __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f)); + __m128i cb_const1 = _mm_set1_epi16( (short) ( 1.77200f*4096.0f+0.5f)); + __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128); + __m128i xw = _mm_set1_epi16(255); // alpha channel + + for (; i+7 < count; i += 8) { + // load + __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i)); + __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i)); + __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i)); + __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128 + __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128 + + // unpack to short (and left-shift cr, cb by 8) + __m128i yw = _mm_unpacklo_epi8(y_bias, y_bytes); + __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased); + __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased); + + // color transform + __m128i yws = _mm_srli_epi16(yw, 4); + __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw); + __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw); + __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1); + __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1); + __m128i rws = _mm_add_epi16(cr0, yws); + __m128i gwt = _mm_add_epi16(cb0, yws); + __m128i bws = _mm_add_epi16(yws, cb1); + __m128i gws = _mm_add_epi16(gwt, cr1); + + // descale + __m128i rw = _mm_srai_epi16(rws, 4); + __m128i bw = _mm_srai_epi16(bws, 4); + __m128i gw = _mm_srai_epi16(gws, 4); + + // back to byte, set up for transpose + __m128i brb = _mm_packus_epi16(rw, bw); + __m128i gxb = _mm_packus_epi16(gw, xw); + + // transpose to interleave channels + __m128i t0 = _mm_unpacklo_epi8(brb, gxb); + __m128i t1 = _mm_unpackhi_epi8(brb, gxb); + __m128i o0 = _mm_unpacklo_epi16(t0, t1); + __m128i o1 = _mm_unpackhi_epi16(t0, t1); + + // store + _mm_storeu_si128((__m128i *) (out + 0), o0); + _mm_storeu_si128((__m128i *) (out + 16), o1); + out += 32; + } + } +#endif + +#ifdef STBI_NEON + // in this version, step=3 support would be easy to add. but is there demand? + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + uint8x8_t signflip = vdup_n_u8(0x80); + int16x8_t cr_const0 = vdupq_n_s16( (short) ( 1.40200f*4096.0f+0.5f)); + int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f)); + int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f)); + int16x8_t cb_const1 = vdupq_n_s16( (short) ( 1.77200f*4096.0f+0.5f)); + + for (; i+7 < count; i += 8) { + // load + uint8x8_t y_bytes = vld1_u8(y + i); + uint8x8_t cr_bytes = vld1_u8(pcr + i); + uint8x8_t cb_bytes = vld1_u8(pcb + i); + int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip)); + int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip)); + + // expand to s16 + int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4)); + int16x8_t crw = vshll_n_s8(cr_biased, 7); + int16x8_t cbw = vshll_n_s8(cb_biased, 7); + + // color transform + int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0); + int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0); + int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1); + int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1); + int16x8_t rws = vaddq_s16(yws, cr0); + int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1); + int16x8_t bws = vaddq_s16(yws, cb1); + + // undo scaling, round, convert to byte + uint8x8x4_t o; + o.val[0] = vqrshrun_n_s16(rws, 4); + o.val[1] = vqrshrun_n_s16(gws, 4); + o.val[2] = vqrshrun_n_s16(bws, 4); + o.val[3] = vdup_n_u8(255); + + // store, interleaving r/g/b/a + vst4_u8(out, o); + out += 8*4; + } + } +#endif + + for (; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* stbi__float2fixed(1.40200f); + g = y_fixed + cr*-stbi__float2fixed(0.71414f) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} +#endif + +// set up the kernels +static void stbi__setup_jpeg(stbi__jpeg *j) +{ + j->idct_block_kernel = stbi__idct_block; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2; + +#ifdef STBI_SSE2 + if (stbi__sse2_available()) { + j->idct_block_kernel = stbi__idct_simd; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; + } +#endif + +#ifdef STBI_NEON + j->idct_block_kernel = stbi__idct_simd; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; +#endif +} + +// clean up the temporary component buffers +static void stbi__cleanup_jpeg(stbi__jpeg *j) +{ + stbi__free_jpeg_components(j, j->s->img_n, 0); +} + +typedef struct +{ + resample_row_func resample; + stbi_uc *line0,*line1; + int hs,vs; // expansion factor in each axis + int w_lores; // horizontal pixels pre-expansion + int ystep; // how far through vertical expansion we are + int ypos; // which pre-expansion row we're on +} stbi__resample; + +// fast 0..255 * 0..255 => 0..255 rounded multiplication +static stbi_uc stbi__blinn_8x8(stbi_uc x, stbi_uc y) +{ + unsigned int t = x*y + 128; + return (stbi_uc) ((t + (t >>8)) >> 8); +} + +static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) +{ + int n, decode_n, is_rgb; + z->s->img_n = 0; // make stbi__cleanup_jpeg safe + + // validate req_comp + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + + // load a jpeg image from whichever source, but leave in YCbCr format + if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; } + + // determine actual number of components to generate + n = req_comp ? req_comp : z->s->img_n >= 3 ? 3 : 1; + + is_rgb = z->s->img_n == 3 && (z->rgb == 3 || (z->app14_color_transform == 0 && !z->jfif)); + + if (z->s->img_n == 3 && n < 3 && !is_rgb) + decode_n = 1; + else + decode_n = z->s->img_n; + + // nothing to do if no components requested; check this now to avoid + // accessing uninitialized coutput[0] later + if (decode_n <= 0) { stbi__cleanup_jpeg(z); return NULL; } + + // resample and color-convert + { + int k; + unsigned int i,j; + stbi_uc *output; + stbi_uc *coutput[4] = { NULL, NULL, NULL, NULL }; + + stbi__resample res_comp[4]; + + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + + // allocate line buffer big enough for upsampling off the edges + // with upsample factor of 4 + z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3); + if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + r->hs = z->img_h_max / z->img_comp[k].h; + r->vs = z->img_v_max / z->img_comp[k].v; + r->ystep = r->vs >> 1; + r->w_lores = (z->s->img_x + r->hs-1) / r->hs; + r->ypos = 0; + r->line0 = r->line1 = z->img_comp[k].data; + + if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1; + else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2; + else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2; + else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel; + else r->resample = stbi__resample_row_generic; + } + + // can't error after this so, this is safe + output = (stbi_uc *) stbi__malloc_mad3(n, z->s->img_x, z->s->img_y, 1); + if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + // now go ahead and resample + for (j=0; j < z->s->img_y; ++j) { + stbi_uc *out = output + n * z->s->img_x * j; + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + int y_bot = r->ystep >= (r->vs >> 1); + coutput[k] = r->resample(z->img_comp[k].linebuf, + y_bot ? r->line1 : r->line0, + y_bot ? r->line0 : r->line1, + r->w_lores, r->hs); + if (++r->ystep >= r->vs) { + r->ystep = 0; + r->line0 = r->line1; + if (++r->ypos < z->img_comp[k].y) + r->line1 += z->img_comp[k].w2; + } + } + if (n >= 3) { + stbi_uc *y = coutput[0]; + if (z->s->img_n == 3) { + if (is_rgb) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = y[i]; + out[1] = coutput[1][i]; + out[2] = coutput[2][i]; + out[3] = 255; + out += n; + } + } else { + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } + } else if (z->s->img_n == 4) { + if (z->app14_color_transform == 0) { // CMYK + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(coutput[0][i], m); + out[1] = stbi__blinn_8x8(coutput[1][i], m); + out[2] = stbi__blinn_8x8(coutput[2][i], m); + out[3] = 255; + out += n; + } + } else if (z->app14_color_transform == 2) { // YCCK + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(255 - out[0], m); + out[1] = stbi__blinn_8x8(255 - out[1], m); + out[2] = stbi__blinn_8x8(255 - out[2], m); + out += n; + } + } else { // YCbCr + alpha? Ignore the fourth channel for now + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } + } else + for (i=0; i < z->s->img_x; ++i) { + out[0] = out[1] = out[2] = y[i]; + out[3] = 255; // not used if n==3 + out += n; + } + } else { + if (is_rgb) { + if (n == 1) + for (i=0; i < z->s->img_x; ++i) + *out++ = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + else { + for (i=0; i < z->s->img_x; ++i, out += 2) { + out[0] = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + out[1] = 255; + } + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 0) { + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + stbi_uc r = stbi__blinn_8x8(coutput[0][i], m); + stbi_uc g = stbi__blinn_8x8(coutput[1][i], m); + stbi_uc b = stbi__blinn_8x8(coutput[2][i], m); + out[0] = stbi__compute_y(r, g, b); + out[1] = 255; + out += n; + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 2) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = stbi__blinn_8x8(255 - coutput[0][i], coutput[3][i]); + out[1] = 255; + out += n; + } + } else { + stbi_uc *y = coutput[0]; + if (n == 1) + for (i=0; i < z->s->img_x; ++i) out[i] = y[i]; + else + for (i=0; i < z->s->img_x; ++i) { *out++ = y[i]; *out++ = 255; } + } + } + } + stbi__cleanup_jpeg(z); + *out_x = z->s->img_x; + *out_y = z->s->img_y; + if (comp) *comp = z->s->img_n >= 3 ? 3 : 1; // report original components, not output + return output; + } +} + +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + unsigned char* result; + stbi__jpeg* j = (stbi__jpeg*) stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__errpuc("outofmem", "Out of memory"); + STBI_NOTUSED(ri); + j->s = s; + stbi__setup_jpeg(j); + result = load_jpeg_image(j, x,y,comp,req_comp); + STBI_FREE(j); + return result; +} + +static int stbi__jpeg_test(stbi__context *s) +{ + int r; + stbi__jpeg* j = (stbi__jpeg*)stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__err("outofmem", "Out of memory"); + j->s = s; + stbi__setup_jpeg(j); + r = stbi__decode_jpeg_header(j, STBI__SCAN_type); + stbi__rewind(s); + STBI_FREE(j); + return r; +} + +static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp) +{ + if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) { + stbi__rewind( j->s ); + return 0; + } + if (x) *x = j->s->img_x; + if (y) *y = j->s->img_y; + if (comp) *comp = j->s->img_n >= 3 ? 3 : 1; + return 1; +} + +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp) +{ + int result; + stbi__jpeg* j = (stbi__jpeg*) (stbi__malloc(sizeof(stbi__jpeg))); + if (!j) return stbi__err("outofmem", "Out of memory"); + j->s = s; + result = stbi__jpeg_info_raw(j, x, y, comp); + STBI_FREE(j); + return result; +} +#endif + +// public domain zlib decode v0.2 Sean Barrett 2006-11-18 +// simple implementation +// - all input must be provided in an upfront buffer +// - all output is written to a single output buffer (can malloc/realloc) +// performance +// - fast huffman + +#ifndef STBI_NO_ZLIB + +// fast-way is faster to check than jpeg huffman, but slow way is slower +#define STBI__ZFAST_BITS 9 // accelerate all cases in default tables +#define STBI__ZFAST_MASK ((1 << STBI__ZFAST_BITS) - 1) +#define STBI__ZNSYMS 288 // number of symbols in literal/length alphabet + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + stbi__uint16 fast[1 << STBI__ZFAST_BITS]; + stbi__uint16 firstcode[16]; + int maxcode[17]; + stbi__uint16 firstsymbol[16]; + stbi_uc size[STBI__ZNSYMS]; + stbi__uint16 value[STBI__ZNSYMS]; +} stbi__zhuffman; + +stbi_inline static int stbi__bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +stbi_inline static int stbi__bit_reverse(int v, int bits) +{ + STBI_ASSERT(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return stbi__bitreverse16(v) >> (16-bits); +} + +static int stbi__zbuild_huffman(stbi__zhuffman *z, const stbi_uc *sizelist, int num) +{ + int i,k=0; + int code, next_code[16], sizes[17]; + + // DEFLATE spec for generating codes + memset(sizes, 0, sizeof(sizes)); + memset(z->fast, 0, sizeof(z->fast)); + for (i=0; i < num; ++i) + ++sizes[sizelist[i]]; + sizes[0] = 0; + for (i=1; i < 16; ++i) + if (sizes[i] > (1 << i)) + return stbi__err("bad sizes", "Corrupt PNG"); + code = 0; + for (i=1; i < 16; ++i) { + next_code[i] = code; + z->firstcode[i] = (stbi__uint16) code; + z->firstsymbol[i] = (stbi__uint16) k; + code = (code + sizes[i]); + if (sizes[i]) + if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG"); + z->maxcode[i] = code << (16-i); // preshift for inner loop + code <<= 1; + k += sizes[i]; + } + z->maxcode[16] = 0x10000; // sentinel + for (i=0; i < num; ++i) { + int s = sizelist[i]; + if (s) { + int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s]; + stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i); + z->size [c] = (stbi_uc ) s; + z->value[c] = (stbi__uint16) i; + if (s <= STBI__ZFAST_BITS) { + int j = stbi__bit_reverse(next_code[s],s); + while (j < (1 << STBI__ZFAST_BITS)) { + z->fast[j] = fastv; + j += (1 << s); + } + } + ++next_code[s]; + } + } + return 1; +} + +// zlib-from-memory implementation for PNG reading +// because PNG allows splitting the zlib stream arbitrarily, +// and it's annoying structurally to have PNG call ZLIB call PNG, +// we require PNG read all the IDATs and combine them into a single +// memory buffer + +typedef struct +{ + stbi_uc *zbuffer, *zbuffer_end; + int num_bits; + stbi__uint32 code_buffer; + + char *zout; + char *zout_start; + char *zout_end; + int z_expandable; + + stbi__zhuffman z_length, z_distance; +} stbi__zbuf; + +stbi_inline static int stbi__zeof(stbi__zbuf *z) +{ + return (z->zbuffer >= z->zbuffer_end); +} + +stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z) +{ + return stbi__zeof(z) ? 0 : *z->zbuffer++; +} + +static void stbi__fill_bits(stbi__zbuf *z) +{ + do { + if (z->code_buffer >= (1U << z->num_bits)) { + z->zbuffer = z->zbuffer_end; /* treat this as EOF so we fail. */ + return; + } + z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits; + z->num_bits += 8; + } while (z->num_bits <= 24); +} + +stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n) +{ + unsigned int k; + if (z->num_bits < n) stbi__fill_bits(z); + k = z->code_buffer & ((1 << n) - 1); + z->code_buffer >>= n; + z->num_bits -= n; + return k; +} + +static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s,k; + // not resolved by fast table, so compute it the slow way + // use jpeg approach, which requires MSbits at top + k = stbi__bit_reverse(a->code_buffer, 16); + for (s=STBI__ZFAST_BITS+1; ; ++s) + if (k < z->maxcode[s]) + break; + if (s >= 16) return -1; // invalid code! + // code size is s, so: + b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; + if (b >= STBI__ZNSYMS) return -1; // some data was corrupt somewhere! + if (z->size[b] != s) return -1; // was originally an assert, but report failure instead. + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; +} + +stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s; + if (a->num_bits < 16) { + if (stbi__zeof(a)) { + return -1; /* report error for unexpected end of data. */ + } + stbi__fill_bits(a); + } + b = z->fast[a->code_buffer & STBI__ZFAST_MASK]; + if (b) { + s = b >> 9; + a->code_buffer >>= s; + a->num_bits -= s; + return b & 511; + } + return stbi__zhuffman_decode_slowpath(a, z); +} + +static int stbi__zexpand(stbi__zbuf *z, char *zout, int n) // need to make room for n bytes +{ + char *q; + unsigned int cur, limit, old_limit; + z->zout = zout; + if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG"); + cur = (unsigned int) (z->zout - z->zout_start); + limit = old_limit = (unsigned) (z->zout_end - z->zout_start); + if (UINT_MAX - cur < (unsigned) n) return stbi__err("outofmem", "Out of memory"); + while (cur + n > limit) { + if(limit > UINT_MAX / 2) return stbi__err("outofmem", "Out of memory"); + limit *= 2; + } + q = (char *) STBI_REALLOC_SIZED(z->zout_start, old_limit, limit); + STBI_NOTUSED(old_limit); + if (q == NULL) return stbi__err("outofmem", "Out of memory"); + z->zout_start = q; + z->zout = q + cur; + z->zout_end = q + limit; + return 1; +} + +static const int stbi__zlength_base[31] = { + 3,4,5,6,7,8,9,10,11,13, + 15,17,19,23,27,31,35,43,51,59, + 67,83,99,115,131,163,195,227,258,0,0 }; + +static const int stbi__zlength_extra[31]= +{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 }; + +static const int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, +257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0}; + +static const int stbi__zdist_extra[32] = +{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +static int stbi__parse_huffman_block(stbi__zbuf *a) +{ + char *zout = a->zout; + for(;;) { + int z = stbi__zhuffman_decode(a, &a->z_length); + if (z < 256) { + if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes + if (zout >= a->zout_end) { + if (!stbi__zexpand(a, zout, 1)) return 0; + zout = a->zout; + } + *zout++ = (char) z; + } else { + stbi_uc *p; + int len,dist; + if (z == 256) { + a->zout = zout; + return 1; + } + z -= 257; + len = stbi__zlength_base[z]; + if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]); + z = stbi__zhuffman_decode(a, &a->z_distance); + if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); + dist = stbi__zdist_base[z]; + if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]); + if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG"); + if (zout + len > a->zout_end) { + if (!stbi__zexpand(a, zout, len)) return 0; + zout = a->zout; + } + p = (stbi_uc *) (zout - dist); + if (dist == 1) { // run of one byte; common in images. + stbi_uc v = *p; + if (len) { do *zout++ = v; while (--len); } + } else { + if (len) { do *zout++ = *p++; while (--len); } + } + } + } +} + +static int stbi__compute_huffman_codes(stbi__zbuf *a) +{ + static const stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 }; + stbi__zhuffman z_codelength; + stbi_uc lencodes[286+32+137];//padding for maximum single op + stbi_uc codelength_sizes[19]; + int i,n; + + int hlit = stbi__zreceive(a,5) + 257; + int hdist = stbi__zreceive(a,5) + 1; + int hclen = stbi__zreceive(a,4) + 4; + int ntot = hlit + hdist; + + memset(codelength_sizes, 0, sizeof(codelength_sizes)); + for (i=0; i < hclen; ++i) { + int s = stbi__zreceive(a,3); + codelength_sizes[length_dezigzag[i]] = (stbi_uc) s; + } + if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0; + + n = 0; + while (n < ntot) { + int c = stbi__zhuffman_decode(a, &z_codelength); + if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG"); + if (c < 16) + lencodes[n++] = (stbi_uc) c; + else { + stbi_uc fill = 0; + if (c == 16) { + c = stbi__zreceive(a,2)+3; + if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG"); + fill = lencodes[n-1]; + } else if (c == 17) { + c = stbi__zreceive(a,3)+3; + } else if (c == 18) { + c = stbi__zreceive(a,7)+11; + } else { + return stbi__err("bad codelengths", "Corrupt PNG"); + } + if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG"); + memset(lencodes+n, fill, c); + n += c; + } + } + if (n != ntot) return stbi__err("bad codelengths","Corrupt PNG"); + if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0; + return 1; +} + +static int stbi__parse_uncompressed_block(stbi__zbuf *a) +{ + stbi_uc header[4]; + int len,nlen,k; + if (a->num_bits & 7) + stbi__zreceive(a, a->num_bits & 7); // discard + // drain the bit-packed data into header + k = 0; + while (a->num_bits > 0) { + header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check + a->code_buffer >>= 8; + a->num_bits -= 8; + } + if (a->num_bits < 0) return stbi__err("zlib corrupt","Corrupt PNG"); + // now fill header the normal way + while (k < 4) + header[k++] = stbi__zget8(a); + len = header[1] * 256 + header[0]; + nlen = header[3] * 256 + header[2]; + if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG"); + if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG"); + if (a->zout + len > a->zout_end) + if (!stbi__zexpand(a, a->zout, len)) return 0; + memcpy(a->zout, a->zbuffer, len); + a->zbuffer += len; + a->zout += len; + return 1; +} + +static int stbi__parse_zlib_header(stbi__zbuf *a) +{ + int cmf = stbi__zget8(a); + int cm = cmf & 15; + /* int cinfo = cmf >> 4; */ + int flg = stbi__zget8(a); + if (stbi__zeof(a)) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png + if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png + // window = 1 << (8 + cinfo)... but who cares, we fully buffer output + return 1; +} + +static const stbi_uc stbi__zdefault_length[STBI__ZNSYMS] = +{ + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8 +}; +static const stbi_uc stbi__zdefault_distance[32] = +{ + 5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5 +}; +/* +Init algorithm: +{ + int i; // use <= to match clearly with spec + for (i=0; i <= 143; ++i) stbi__zdefault_length[i] = 8; + for ( ; i <= 255; ++i) stbi__zdefault_length[i] = 9; + for ( ; i <= 279; ++i) stbi__zdefault_length[i] = 7; + for ( ; i <= 287; ++i) stbi__zdefault_length[i] = 8; + + for (i=0; i <= 31; ++i) stbi__zdefault_distance[i] = 5; +} +*/ + +static int stbi__parse_zlib(stbi__zbuf *a, int parse_header) +{ + int final, type; + if (parse_header) + if (!stbi__parse_zlib_header(a)) return 0; + a->num_bits = 0; + a->code_buffer = 0; + do { + final = stbi__zreceive(a,1); + type = stbi__zreceive(a,2); + if (type == 0) { + if (!stbi__parse_uncompressed_block(a)) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , STBI__ZNSYMS)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0; + } else { + if (!stbi__compute_huffman_codes(a)) return 0; + } + if (!stbi__parse_huffman_block(a)) return 0; + } + } while (!final); + return 1; +} + +static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header) +{ + a->zout_start = obuf; + a->zout = obuf; + a->zout_end = obuf + olen; + a->z_expandable = exp; + + return stbi__parse_zlib(a, parse_header); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, 1)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen) +{ + return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 1)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(16384); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer+len; + if (stbi__do_zlib(&a, p, 16384, 1, 0)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 0)) + return (int) (a.zout - a.zout_start); + else + return -1; +} +#endif + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + +#ifndef STBI_NO_PNG +typedef struct +{ + stbi__uint32 length; + stbi__uint32 type; +} stbi__pngchunk; + +static stbi__pngchunk stbi__get_chunk_header(stbi__context *s) +{ + stbi__pngchunk c; + c.length = stbi__get32be(s); + c.type = stbi__get32be(s); + return c; +} + +static int stbi__check_png_header(stbi__context *s) +{ + static const stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 }; + int i; + for (i=0; i < 8; ++i) + if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG"); + return 1; +} + +typedef struct +{ + stbi__context *s; + stbi_uc *idata, *expanded, *out; + int depth; +} stbi__png; + + +enum { + STBI__F_none=0, + STBI__F_sub=1, + STBI__F_up=2, + STBI__F_avg=3, + STBI__F_paeth=4, + // synthetic filters used for first scanline to avoid needing a dummy row of 0s + STBI__F_avg_first, + STBI__F_paeth_first +}; + +static stbi_uc first_row_filter[5] = +{ + STBI__F_none, + STBI__F_sub, + STBI__F_none, + STBI__F_avg_first, + STBI__F_paeth_first +}; + +static int stbi__paeth(int a, int b, int c) +{ + int p = a + b - c; + int pa = abs(p-a); + int pb = abs(p-b); + int pc = abs(p-c); + if (pa <= pb && pa <= pc) return a; + if (pb <= pc) return b; + return c; +} + +static const stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 }; + +// create the png data from post-deflated data +static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color) +{ + int bytes = (depth == 16? 2 : 1); + stbi__context *s = a->s; + stbi__uint32 i,j,stride = x*out_n*bytes; + stbi__uint32 img_len, img_width_bytes; + int k; + int img_n = s->img_n; // copy it into a local for later + + int output_bytes = out_n*bytes; + int filter_bytes = img_n*bytes; + int width = x; + + STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1); + a->out = (stbi_uc *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into + if (!a->out) return stbi__err("outofmem", "Out of memory"); + + if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG"); + img_width_bytes = (((img_n * x * depth) + 7) >> 3); + img_len = (img_width_bytes + 1) * y; + + // we used to check for exact match between raw_len and img_len on non-interlaced PNGs, + // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros), + // so just check for raw_len < img_len always. + if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG"); + + for (j=0; j < y; ++j) { + stbi_uc *cur = a->out + stride*j; + stbi_uc *prior; + int filter = *raw++; + + if (filter > 4) + return stbi__err("invalid filter","Corrupt PNG"); + + if (depth < 8) { + if (img_width_bytes > x) return stbi__err("invalid width","Corrupt PNG"); + cur += x*out_n - img_width_bytes; // store output to the rightmost img_len bytes, so we can decode in place + filter_bytes = 1; + width = img_width_bytes; + } + prior = cur - stride; // bugfix: need to compute this after 'cur +=' computation above + + // if first row, use special filter that doesn't sample previous row + if (j == 0) filter = first_row_filter[filter]; + + // handle first byte explicitly + for (k=0; k < filter_bytes; ++k) { + switch (filter) { + case STBI__F_none : cur[k] = raw[k]; break; + case STBI__F_sub : cur[k] = raw[k]; break; + case STBI__F_up : cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break; + case STBI__F_avg : cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); break; + case STBI__F_paeth : cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(0,prior[k],0)); break; + case STBI__F_avg_first : cur[k] = raw[k]; break; + case STBI__F_paeth_first: cur[k] = raw[k]; break; + } + } + + if (depth == 8) { + if (img_n != out_n) + cur[img_n] = 255; // first pixel + raw += img_n; + cur += out_n; + prior += out_n; + } else if (depth == 16) { + if (img_n != out_n) { + cur[filter_bytes] = 255; // first pixel top byte + cur[filter_bytes+1] = 255; // first pixel bottom byte + } + raw += filter_bytes; + cur += output_bytes; + prior += output_bytes; + } else { + raw += 1; + cur += 1; + prior += 1; + } + + // this is a little gross, so that we don't switch per-pixel or per-component + if (depth < 8 || img_n == out_n) { + int nk = (width - 1)*filter_bytes; + #define STBI__CASE(f) \ + case f: \ + for (k=0; k < nk; ++k) + switch (filter) { + // "none" filter turns into a memcpy here; make that explicit. + case STBI__F_none: memcpy(cur, raw, nk); break; + STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); } break; + STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break; + STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); } break; + STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); } break; + STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); } break; + STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); } break; + } + #undef STBI__CASE + raw += nk; + } else { + STBI_ASSERT(img_n+1 == out_n); + #define STBI__CASE(f) \ + case f: \ + for (i=x-1; i >= 1; --i, cur[filter_bytes]=255,raw+=filter_bytes,cur+=output_bytes,prior+=output_bytes) \ + for (k=0; k < filter_bytes; ++k) + switch (filter) { + STBI__CASE(STBI__F_none) { cur[k] = raw[k]; } break; + STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k- output_bytes]); } break; + STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break; + STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k- output_bytes])>>1)); } break; + STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],prior[k],prior[k- output_bytes])); } break; + STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k- output_bytes] >> 1)); } break; + STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],0,0)); } break; + } + #undef STBI__CASE + + // the loop above sets the high byte of the pixels' alpha, but for + // 16 bit png files we also need the low byte set. we'll do that here. + if (depth == 16) { + cur = a->out + stride*j; // start at the beginning of the row again + for (i=0; i < x; ++i,cur+=output_bytes) { + cur[filter_bytes+1] = 255; + } + } + } + } + + // we make a separate pass to expand bits to pixels; for performance, + // this could run two scanlines behind the above code, so it won't + // intefere with filtering but will still be in the cache. + if (depth < 8) { + for (j=0; j < y; ++j) { + stbi_uc *cur = a->out + stride*j; + stbi_uc *in = a->out + stride*j + x*out_n - img_width_bytes; + // unpack 1/2/4-bit into a 8-bit buffer. allows us to keep the common 8-bit path optimal at minimal cost for 1/2/4-bit + // png guarante byte alignment, if width is not multiple of 8/4/2 we'll decode dummy trailing data that will be skipped in the later loop + stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range + + // note that the final byte might overshoot and write more data than desired. + // we can allocate enough data that this never writes out of memory, but it + // could also overwrite the next scanline. can it overwrite non-empty data + // on the next scanline? yes, consider 1-pixel-wide scanlines with 1-bit-per-pixel. + // so we need to explicitly clamp the final ones + + if (depth == 4) { + for (k=x*img_n; k >= 2; k-=2, ++in) { + *cur++ = scale * ((*in >> 4) ); + *cur++ = scale * ((*in ) & 0x0f); + } + if (k > 0) *cur++ = scale * ((*in >> 4) ); + } else if (depth == 2) { + for (k=x*img_n; k >= 4; k-=4, ++in) { + *cur++ = scale * ((*in >> 6) ); + *cur++ = scale * ((*in >> 4) & 0x03); + *cur++ = scale * ((*in >> 2) & 0x03); + *cur++ = scale * ((*in ) & 0x03); + } + if (k > 0) *cur++ = scale * ((*in >> 6) ); + if (k > 1) *cur++ = scale * ((*in >> 4) & 0x03); + if (k > 2) *cur++ = scale * ((*in >> 2) & 0x03); + } else if (depth == 1) { + for (k=x*img_n; k >= 8; k-=8, ++in) { + *cur++ = scale * ((*in >> 7) ); + *cur++ = scale * ((*in >> 6) & 0x01); + *cur++ = scale * ((*in >> 5) & 0x01); + *cur++ = scale * ((*in >> 4) & 0x01); + *cur++ = scale * ((*in >> 3) & 0x01); + *cur++ = scale * ((*in >> 2) & 0x01); + *cur++ = scale * ((*in >> 1) & 0x01); + *cur++ = scale * ((*in ) & 0x01); + } + if (k > 0) *cur++ = scale * ((*in >> 7) ); + if (k > 1) *cur++ = scale * ((*in >> 6) & 0x01); + if (k > 2) *cur++ = scale * ((*in >> 5) & 0x01); + if (k > 3) *cur++ = scale * ((*in >> 4) & 0x01); + if (k > 4) *cur++ = scale * ((*in >> 3) & 0x01); + if (k > 5) *cur++ = scale * ((*in >> 2) & 0x01); + if (k > 6) *cur++ = scale * ((*in >> 1) & 0x01); + } + if (img_n != out_n) { + int q; + // insert alpha = 255 + cur = a->out + stride*j; + if (img_n == 1) { + for (q=x-1; q >= 0; --q) { + cur[q*2+1] = 255; + cur[q*2+0] = cur[q]; + } + } else { + STBI_ASSERT(img_n == 3); + for (q=x-1; q >= 0; --q) { + cur[q*4+3] = 255; + cur[q*4+2] = cur[q*3+2]; + cur[q*4+1] = cur[q*3+1]; + cur[q*4+0] = cur[q*3+0]; + } + } + } + } + } else if (depth == 16) { + // force the image data from big-endian to platform-native. + // this is done in a separate pass due to the decoding relying + // on the data being untouched, but could probably be done + // per-line during decode if care is taken. + stbi_uc *cur = a->out; + stbi__uint16 *cur16 = (stbi__uint16*)cur; + + for(i=0; i < x*y*out_n; ++i,cur16++,cur+=2) { + *cur16 = (cur[0] << 8) | cur[1]; + } + } + + return 1; +} + +static int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced) +{ + int bytes = (depth == 16 ? 2 : 1); + int out_bytes = out_n * bytes; + stbi_uc *final; + int p; + if (!interlaced) + return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color); + + // de-interlacing + final = (stbi_uc *) stbi__malloc_mad3(a->s->img_x, a->s->img_y, out_bytes, 0); + if (!final) return stbi__err("outofmem", "Out of memory"); + for (p=0; p < 7; ++p) { + int xorig[] = { 0,4,0,2,0,1,0 }; + int yorig[] = { 0,0,4,0,2,0,1 }; + int xspc[] = { 8,8,4,4,2,2,1 }; + int yspc[] = { 8,8,8,4,4,2,2 }; + int i,j,x,y; + // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1 + x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p]; + y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p]; + if (x && y) { + stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y; + if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) { + STBI_FREE(final); + return 0; + } + for (j=0; j < y; ++j) { + for (i=0; i < x; ++i) { + int out_y = j*yspc[p]+yorig[p]; + int out_x = i*xspc[p]+xorig[p]; + memcpy(final + out_y*a->s->img_x*out_bytes + out_x*out_bytes, + a->out + (j*x+i)*out_bytes, out_bytes); + } + } + STBI_FREE(a->out); + image_data += img_len; + image_data_len -= img_len; + } + } + a->out = final; + + return 1; +} + +static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + // compute color-based transparency, assuming we've + // already got 255 as the alpha value in the output + STBI_ASSERT(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i=0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 255); + p += 2; + } + } else { + for (i=0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int stbi__compute_transparency16(stbi__png *z, stbi__uint16 tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi__uint16 *p = (stbi__uint16*) z->out; + + // compute color-based transparency, assuming we've + // already got 65535 as the alpha value in the output + STBI_ASSERT(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i = 0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 65535); + p += 2; + } + } else { + for (i = 0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n) +{ + stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y; + stbi_uc *p, *temp_out, *orig = a->out; + + p = (stbi_uc *) stbi__malloc_mad2(pixel_count, pal_img_n, 0); + if (p == NULL) return stbi__err("outofmem", "Out of memory"); + + // between here and free(out) below, exitting would leak + temp_out = p; + + if (pal_img_n == 3) { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p += 3; + } + } else { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p[3] = palette[n+3]; + p += 4; + } + } + STBI_FREE(a->out); + a->out = temp_out; + + STBI_NOTUSED(len); + + return 1; +} + +static int stbi__unpremultiply_on_load_global = 0; +static int stbi__de_iphone_flag_global = 0; + +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_global = flag_true_if_should_unpremultiply; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_global = flag_true_if_should_convert; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__unpremultiply_on_load stbi__unpremultiply_on_load_global +#define stbi__de_iphone_flag stbi__de_iphone_flag_global +#else +static STBI_THREAD_LOCAL int stbi__unpremultiply_on_load_local, stbi__unpremultiply_on_load_set; +static STBI_THREAD_LOCAL int stbi__de_iphone_flag_local, stbi__de_iphone_flag_set; + +STBIDEF void stbi__unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_local = flag_true_if_should_unpremultiply; + stbi__unpremultiply_on_load_set = 1; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_local = flag_true_if_should_convert; + stbi__de_iphone_flag_set = 1; +} + +#define stbi__unpremultiply_on_load (stbi__unpremultiply_on_load_set \ + ? stbi__unpremultiply_on_load_local \ + : stbi__unpremultiply_on_load_global) +#define stbi__de_iphone_flag (stbi__de_iphone_flag_set \ + ? stbi__de_iphone_flag_local \ + : stbi__de_iphone_flag_global) +#endif // STBI_THREAD_LOCAL + +static void stbi__de_iphone(stbi__png *z) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + if (s->img_out_n == 3) { // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 3; + } + } else { + STBI_ASSERT(s->img_out_n == 4); + if (stbi__unpremultiply_on_load) { + // convert bgr to rgb and unpremultiply + for (i=0; i < pixel_count; ++i) { + stbi_uc a = p[3]; + stbi_uc t = p[0]; + if (a) { + stbi_uc half = a / 2; + p[0] = (p[2] * 255 + half) / a; + p[1] = (p[1] * 255 + half) / a; + p[2] = ( t * 255 + half) / a; + } else { + p[0] = p[2]; + p[2] = t; + } + p += 4; + } + } else { + // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 4; + } + } + } +} + +#define STBI__PNG_TYPE(a,b,c,d) (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (d)) + +static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) +{ + stbi_uc palette[1024], pal_img_n=0; + stbi_uc has_trans=0, tc[3]={0}; + stbi__uint16 tc16[3]; + stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k,interlace=0, color=0, is_iphone=0; + stbi__context *s = z->s; + + z->expanded = NULL; + z->idata = NULL; + z->out = NULL; + + if (!stbi__check_png_header(s)) return 0; + + if (scan == STBI__SCAN_type) return 1; + + for (;;) { + stbi__pngchunk c = stbi__get_chunk_header(s); + switch (c.type) { + case STBI__PNG_TYPE('C','g','B','I'): + is_iphone = 1; + stbi__skip(s, c.length); + break; + case STBI__PNG_TYPE('I','H','D','R'): { + int comp,filter; + if (!first) return stbi__err("multiple IHDR","Corrupt PNG"); + first = 0; + if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG"); + s->img_x = stbi__get32be(s); + s->img_y = stbi__get32be(s); + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + z->depth = stbi__get8(s); if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16) return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only"); + color = stbi__get8(s); if (color > 6) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3 && z->depth == 16) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG"); + comp = stbi__get8(s); if (comp) return stbi__err("bad comp method","Corrupt PNG"); + filter= stbi__get8(s); if (filter) return stbi__err("bad filter method","Corrupt PNG"); + interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG"); + if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG"); + if (!pal_img_n) { + s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0); + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode"); + if (scan == STBI__SCAN_header) return 1; + } else { + // if paletted, then pal_n is our final components, and + // img_n is # components to decompress/filter. + s->img_n = 1; + if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG"); + // if SCAN_header, have to scan to see if we have a tRNS + } + break; + } + + case STBI__PNG_TYPE('P','L','T','E'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG"); + pal_len = c.length / 3; + if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG"); + for (i=0; i < pal_len; ++i) { + palette[i*4+0] = stbi__get8(s); + palette[i*4+1] = stbi__get8(s); + palette[i*4+2] = stbi__get8(s); + palette[i*4+3] = 255; + } + break; + } + + case STBI__PNG_TYPE('t','R','N','S'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG"); + if (pal_img_n) { + if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; } + if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG"); + if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG"); + pal_img_n = 4; + for (i=0; i < c.length; ++i) + palette[i*4+3] = stbi__get8(s); + } else { + if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG"); + if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG"); + has_trans = 1; + if (z->depth == 16) { + for (k = 0; k < s->img_n; ++k) tc16[k] = (stbi__uint16)stbi__get16be(s); // copy the values as-is + } else { + for (k = 0; k < s->img_n; ++k) tc[k] = (stbi_uc)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->depth]; // non 8-bit images will be larger + } + } + break; + } + + case STBI__PNG_TYPE('I','D','A','T'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG"); + if (scan == STBI__SCAN_header) { s->img_n = pal_img_n; return 1; } + if ((int)(ioff + c.length) < (int)ioff) return 0; + if (ioff + c.length > idata_limit) { + stbi__uint32 idata_limit_old = idata_limit; + stbi_uc *p; + if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096; + while (ioff + c.length > idata_limit) + idata_limit *= 2; + STBI_NOTUSED(idata_limit_old); + p = (stbi_uc *) STBI_REALLOC_SIZED(z->idata, idata_limit_old, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory"); + z->idata = p; + } + if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG"); + ioff += c.length; + break; + } + + case STBI__PNG_TYPE('I','E','N','D'): { + stbi__uint32 raw_len, bpl; + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (scan != STBI__SCAN_load) return 1; + if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG"); + // initial guess for decoded data size to avoid unnecessary reallocs + bpl = (s->img_x * z->depth + 7) / 8; // bytes per line, per component + raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */; + z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone); + if (z->expanded == NULL) return 0; // zlib should set error + STBI_FREE(z->idata); z->idata = NULL; + if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans) + s->img_out_n = s->img_n+1; + else + s->img_out_n = s->img_n; + if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, z->depth, color, interlace)) return 0; + if (has_trans) { + if (z->depth == 16) { + if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0; + } else { + if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0; + } + } + if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2) + stbi__de_iphone(z); + if (pal_img_n) { + // pal_img_n == 3 or 4 + s->img_n = pal_img_n; // record the actual colors we had + s->img_out_n = pal_img_n; + if (req_comp >= 3) s->img_out_n = req_comp; + if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n)) + return 0; + } else if (has_trans) { + // non-paletted image with tRNS -> source image has (constant) alpha + ++s->img_n; + } + STBI_FREE(z->expanded); z->expanded = NULL; + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + return 1; + } + + default: + // if critical, fail + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if ((c.type & (1 << 29)) == 0) { + #ifndef STBI_NO_FAILURE_STRINGS + // not threadsafe + static char invalid_chunk[] = "XXXX PNG chunk not known"; + invalid_chunk[0] = STBI__BYTECAST(c.type >> 24); + invalid_chunk[1] = STBI__BYTECAST(c.type >> 16); + invalid_chunk[2] = STBI__BYTECAST(c.type >> 8); + invalid_chunk[3] = STBI__BYTECAST(c.type >> 0); + #endif + return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type"); + } + stbi__skip(s, c.length); + break; + } + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + } +} + +static void *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp, stbi__result_info *ri) +{ + void *result=NULL; + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) { + if (p->depth <= 8) + ri->bits_per_channel = 8; + else if (p->depth == 16) + ri->bits_per_channel = 16; + else + return stbi__errpuc("bad bits_per_channel", "PNG not supported: unsupported color depth"); + result = p->out; + p->out = NULL; + if (req_comp && req_comp != p->s->img_out_n) { + if (ri->bits_per_channel == 8) + result = stbi__convert_format((unsigned char *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + else + result = stbi__convert_format16((stbi__uint16 *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + p->s->img_out_n = req_comp; + if (result == NULL) return result; + } + *x = p->s->img_x; + *y = p->s->img_y; + if (n) *n = p->s->img_n; + } + STBI_FREE(p->out); p->out = NULL; + STBI_FREE(p->expanded); p->expanded = NULL; + STBI_FREE(p->idata); p->idata = NULL; + + return result; +} + +static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi__png p; + p.s = s; + return stbi__do_png(&p, x,y,comp,req_comp, ri); +} + +static int stbi__png_test(stbi__context *s) +{ + int r; + r = stbi__check_png_header(s); + stbi__rewind(s); + return r; +} + +static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp) +{ + if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) { + stbi__rewind( p->s ); + return 0; + } + if (x) *x = p->s->img_x; + if (y) *y = p->s->img_y; + if (comp) *comp = p->s->img_n; + return 1; +} + +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__png p; + p.s = s; + return stbi__png_info_raw(&p, x, y, comp); +} + +static int stbi__png_is16(stbi__context *s) +{ + stbi__png p; + p.s = s; + if (!stbi__png_info_raw(&p, NULL, NULL, NULL)) + return 0; + if (p.depth != 16) { + stbi__rewind(p.s); + return 0; + } + return 1; +} +#endif + +// Microsoft/Windows BMP image + +#ifndef STBI_NO_BMP +static int stbi__bmp_test_raw(stbi__context *s) +{ + int r; + int sz; + if (stbi__get8(s) != 'B') return 0; + if (stbi__get8(s) != 'M') return 0; + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + stbi__get32le(s); // discard data offset + sz = stbi__get32le(s); + r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124); + return r; +} + +static int stbi__bmp_test(stbi__context *s) +{ + int r = stbi__bmp_test_raw(s); + stbi__rewind(s); + return r; +} + + +// returns 0..31 for the highest set bit +static int stbi__high_bit(unsigned int z) +{ + int n=0; + if (z == 0) return -1; + if (z >= 0x10000) { n += 16; z >>= 16; } + if (z >= 0x00100) { n += 8; z >>= 8; } + if (z >= 0x00010) { n += 4; z >>= 4; } + if (z >= 0x00004) { n += 2; z >>= 2; } + if (z >= 0x00002) { n += 1;/* >>= 1;*/ } + return n; +} + +static int stbi__bitcount(unsigned int a) +{ + a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2 + a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4 + a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits + a = (a + (a >> 8)); // max 16 per 8 bits + a = (a + (a >> 16)); // max 32 per 8 bits + return a & 0xff; +} + +// extract an arbitrarily-aligned N-bit value (N=bits) +// from v, and then make it 8-bits long and fractionally +// extend it to full full range. +static int stbi__shiftsigned(unsigned int v, int shift, int bits) +{ + static unsigned int mul_table[9] = { + 0, + 0xff/*0b11111111*/, 0x55/*0b01010101*/, 0x49/*0b01001001*/, 0x11/*0b00010001*/, + 0x21/*0b00100001*/, 0x41/*0b01000001*/, 0x81/*0b10000001*/, 0x01/*0b00000001*/, + }; + static unsigned int shift_table[9] = { + 0, 0,0,1,0,2,4,6,0, + }; + if (shift < 0) + v <<= -shift; + else + v >>= shift; + STBI_ASSERT(v < 256); + v >>= (8-bits); + STBI_ASSERT(bits >= 0 && bits <= 8); + return (int) ((unsigned) v * mul_table[bits]) >> shift_table[bits]; +} + +typedef struct +{ + int bpp, offset, hsz; + unsigned int mr,mg,mb,ma, all_a; + int extra_read; +} stbi__bmp_data; + +static int stbi__bmp_set_mask_defaults(stbi__bmp_data *info, int compress) +{ + // BI_BITFIELDS specifies masks explicitly, don't override + if (compress == 3) + return 1; + + if (compress == 0) { + if (info->bpp == 16) { + info->mr = 31u << 10; + info->mg = 31u << 5; + info->mb = 31u << 0; + } else if (info->bpp == 32) { + info->mr = 0xffu << 16; + info->mg = 0xffu << 8; + info->mb = 0xffu << 0; + info->ma = 0xffu << 24; + info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0 + } else { + // otherwise, use defaults, which is all-0 + info->mr = info->mg = info->mb = info->ma = 0; + } + return 1; + } + return 0; // error +} + +static void *stbi__bmp_parse_header(stbi__context *s, stbi__bmp_data *info) +{ + int hsz; + if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc("not BMP", "Corrupt BMP"); + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + info->offset = stbi__get32le(s); + info->hsz = hsz = stbi__get32le(s); + info->mr = info->mg = info->mb = info->ma = 0; + info->extra_read = 14; + + if (info->offset < 0) return stbi__errpuc("bad BMP", "bad BMP"); + + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc("unknown BMP", "BMP type not supported: unknown"); + if (hsz == 12) { + s->img_x = stbi__get16le(s); + s->img_y = stbi__get16le(s); + } else { + s->img_x = stbi__get32le(s); + s->img_y = stbi__get32le(s); + } + if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP"); + info->bpp = stbi__get16le(s); + if (hsz != 12) { + int compress = stbi__get32le(s); + if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE"); + if (compress >= 4) return stbi__errpuc("BMP JPEG/PNG", "BMP type not supported: unsupported compression"); // this includes PNG/JPEG modes + if (compress == 3 && info->bpp != 16 && info->bpp != 32) return stbi__errpuc("bad BMP", "bad BMP"); // bitfields requires 16 or 32 bits/pixel + stbi__get32le(s); // discard sizeof + stbi__get32le(s); // discard hres + stbi__get32le(s); // discard vres + stbi__get32le(s); // discard colorsused + stbi__get32le(s); // discard max important + if (hsz == 40 || hsz == 56) { + if (hsz == 56) { + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + } + if (info->bpp == 16 || info->bpp == 32) { + if (compress == 0) { + stbi__bmp_set_mask_defaults(info, compress); + } else if (compress == 3) { + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->extra_read += 12; + // not documented, but generated by photoshop and handled by mspaint + if (info->mr == info->mg && info->mg == info->mb) { + // ?!?!? + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else { + // V4/V5 header + int i; + if (hsz != 108 && hsz != 124) + return stbi__errpuc("bad BMP", "bad BMP"); + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->ma = stbi__get32le(s); + if (compress != 3) // override mr/mg/mb unless in BI_BITFIELDS mode, as per docs + stbi__bmp_set_mask_defaults(info, compress); + stbi__get32le(s); // discard color space + for (i=0; i < 12; ++i) + stbi__get32le(s); // discard color space parameters + if (hsz == 124) { + stbi__get32le(s); // discard rendering intent + stbi__get32le(s); // discard offset of profile data + stbi__get32le(s); // discard size of profile data + stbi__get32le(s); // discard reserved + } + } + } + return (void *) 1; +} + + +static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *out; + unsigned int mr=0,mg=0,mb=0,ma=0, all_a; + stbi_uc pal[256][4]; + int psize=0,i,j,width; + int flip_vertically, pad, target; + stbi__bmp_data info; + STBI_NOTUSED(ri); + + info.all_a = 255; + if (stbi__bmp_parse_header(s, &info) == NULL) + return NULL; // error code already set + + flip_vertically = ((int) s->img_y) > 0; + s->img_y = abs((int) s->img_y); + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + mr = info.mr; + mg = info.mg; + mb = info.mb; + ma = info.ma; + all_a = info.all_a; + + if (info.hsz == 12) { + if (info.bpp < 24) + psize = (info.offset - info.extra_read - 24) / 3; + } else { + if (info.bpp < 16) + psize = (info.offset - info.extra_read - info.hsz) >> 2; + } + if (psize == 0) { + if (info.offset != s->callback_already_read + (s->img_buffer - s->img_buffer_original)) { + return stbi__errpuc("bad offset", "Corrupt BMP"); + } + } + + if (info.bpp == 24 && ma == 0xff000000) + s->img_n = 3; + else + s->img_n = ma ? 4 : 3; + if (req_comp && req_comp >= 3) // we can directly decode 3 or 4 + target = req_comp; + else + target = s->img_n; // if they want monochrome, we'll post-convert + + // sanity-check size + if (!stbi__mad3sizes_valid(target, s->img_x, s->img_y, 0)) + return stbi__errpuc("too large", "Corrupt BMP"); + + out = (stbi_uc *) stbi__malloc_mad3(target, s->img_x, s->img_y, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + if (info.bpp < 16) { + int z=0; + if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc("invalid", "Corrupt BMP"); } + for (i=0; i < psize; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + if (info.hsz != 12) stbi__get8(s); + pal[i][3] = 255; + } + stbi__skip(s, info.offset - info.extra_read - info.hsz - psize * (info.hsz == 12 ? 3 : 4)); + if (info.bpp == 1) width = (s->img_x + 7) >> 3; + else if (info.bpp == 4) width = (s->img_x + 1) >> 1; + else if (info.bpp == 8) width = s->img_x; + else { STBI_FREE(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); } + pad = (-width)&3; + if (info.bpp == 1) { + for (j=0; j < (int) s->img_y; ++j) { + int bit_offset = 7, v = stbi__get8(s); + for (i=0; i < (int) s->img_x; ++i) { + int color = (v>>bit_offset)&0x1; + out[z++] = pal[color][0]; + out[z++] = pal[color][1]; + out[z++] = pal[color][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + if((--bit_offset) < 0) { + bit_offset = 7; + v = stbi__get8(s); + } + } + stbi__skip(s, pad); + } + } else { + for (j=0; j < (int) s->img_y; ++j) { + for (i=0; i < (int) s->img_x; i += 2) { + int v=stbi__get8(s),v2=0; + if (info.bpp == 4) { + v2 = v & 15; + v >>= 4; + } + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + v = (info.bpp == 8) ? stbi__get8(s) : v2; + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + } + stbi__skip(s, pad); + } + } + } else { + int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0; + int z = 0; + int easy=0; + stbi__skip(s, info.offset - info.extra_read - info.hsz); + if (info.bpp == 24) width = 3 * s->img_x; + else if (info.bpp == 16) width = 2*s->img_x; + else /* bpp = 32 and pad = 0 */ width=0; + pad = (-width) & 3; + if (info.bpp == 24) { + easy = 1; + } else if (info.bpp == 32) { + if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000) + easy = 2; + } + if (!easy) { + if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + // right shift amt to put high bit in position #7 + rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr); + gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg); + bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb); + ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma); + if (rcount > 8 || gcount > 8 || bcount > 8 || acount > 8) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + } + for (j=0; j < (int) s->img_y; ++j) { + if (easy) { + for (i=0; i < (int) s->img_x; ++i) { + unsigned char a; + out[z+2] = stbi__get8(s); + out[z+1] = stbi__get8(s); + out[z+0] = stbi__get8(s); + z += 3; + a = (easy == 2 ? stbi__get8(s) : 255); + all_a |= a; + if (target == 4) out[z++] = a; + } + } else { + int bpp = info.bpp; + for (i=0; i < (int) s->img_x; ++i) { + stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s)); + unsigned int a; + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount)); + a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255); + all_a |= a; + if (target == 4) out[z++] = STBI__BYTECAST(a); + } + } + stbi__skip(s, pad); + } + } + + // if alpha channel is all 0s, replace with all 255s + if (target == 4 && all_a == 0) + for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4) + out[i] = 255; + + if (flip_vertically) { + stbi_uc t; + for (j=0; j < (int) s->img_y>>1; ++j) { + stbi_uc *p1 = out + j *s->img_x*target; + stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target; + for (i=0; i < (int) s->img_x*target; ++i) { + t = p1[i]; p1[i] = p2[i]; p2[i] = t; + } + } + } + + if (req_comp && req_comp != target) { + out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + return out; +} +#endif + +// Targa Truevision - TGA +// by Jonathan Dummer +#ifndef STBI_NO_TGA +// returns STBI_rgb or whatever, 0 on error +static int stbi__tga_get_comp(int bits_per_pixel, int is_grey, int* is_rgb16) +{ + // only RGB or RGBA (incl. 16bit) or grey allowed + if (is_rgb16) *is_rgb16 = 0; + switch(bits_per_pixel) { + case 8: return STBI_grey; + case 16: if(is_grey) return STBI_grey_alpha; + // fallthrough + case 15: if(is_rgb16) *is_rgb16 = 1; + return STBI_rgb; + case 24: // fallthrough + case 32: return bits_per_pixel/8; + default: return 0; + } +} + +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp) +{ + int tga_w, tga_h, tga_comp, tga_image_type, tga_bits_per_pixel, tga_colormap_bpp; + int sz, tga_colormap_type; + stbi__get8(s); // discard Offset + tga_colormap_type = stbi__get8(s); // colormap type + if( tga_colormap_type > 1 ) { + stbi__rewind(s); + return 0; // only RGB or indexed allowed + } + tga_image_type = stbi__get8(s); // image type + if ( tga_colormap_type == 1 ) { // colormapped (paletted) image + if (tga_image_type != 1 && tga_image_type != 9) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip image x and y origin + tga_colormap_bpp = sz; + } else { // "normal" image w/o colormap - only RGB or grey allowed, +/- RLE + if ( (tga_image_type != 2) && (tga_image_type != 3) && (tga_image_type != 10) && (tga_image_type != 11) ) { + stbi__rewind(s); + return 0; // only RGB or grey allowed, +/- RLE + } + stbi__skip(s,9); // skip colormap specification and image x/y origin + tga_colormap_bpp = 0; + } + tga_w = stbi__get16le(s); + if( tga_w < 1 ) { + stbi__rewind(s); + return 0; // test width + } + tga_h = stbi__get16le(s); + if( tga_h < 1 ) { + stbi__rewind(s); + return 0; // test height + } + tga_bits_per_pixel = stbi__get8(s); // bits per pixel + stbi__get8(s); // ignore alpha bits + if (tga_colormap_bpp != 0) { + if((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16)) { + // when using a colormap, tga_bits_per_pixel is the size of the indexes + // I don't think anything but 8 or 16bit indexes makes sense + stbi__rewind(s); + return 0; + } + tga_comp = stbi__tga_get_comp(tga_colormap_bpp, 0, NULL); + } else { + tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3) || (tga_image_type == 11), NULL); + } + if(!tga_comp) { + stbi__rewind(s); + return 0; + } + if (x) *x = tga_w; + if (y) *y = tga_h; + if (comp) *comp = tga_comp; + return 1; // seems to have passed everything +} + +static int stbi__tga_test(stbi__context *s) +{ + int res = 0; + int sz, tga_color_type; + stbi__get8(s); // discard Offset + tga_color_type = stbi__get8(s); // color type + if ( tga_color_type > 1 ) goto errorEnd; // only RGB or indexed allowed + sz = stbi__get8(s); // image type + if ( tga_color_type == 1 ) { // colormapped (paletted) image + if (sz != 1 && sz != 9) goto errorEnd; // colortype 1 demands image type 1 or 9 + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + stbi__skip(s,4); // skip image x and y origin + } else { // "normal" image w/o colormap + if ( (sz != 2) && (sz != 3) && (sz != 10) && (sz != 11) ) goto errorEnd; // only RGB or grey allowed, +/- RLE + stbi__skip(s,9); // skip colormap specification and image x/y origin + } + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test width + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test height + sz = stbi__get8(s); // bits per pixel + if ( (tga_color_type == 1) && (sz != 8) && (sz != 16) ) goto errorEnd; // for colormapped images, bpp is size of an index + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + + res = 1; // if we got this far, everything's good and we can return 1 instead of 0 + +errorEnd: + stbi__rewind(s); + return res; +} + +// read 16bit value and convert to 24bit RGB +static void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out) +{ + stbi__uint16 px = (stbi__uint16)stbi__get16le(s); + stbi__uint16 fiveBitMask = 31; + // we have 3 channels with 5bits each + int r = (px >> 10) & fiveBitMask; + int g = (px >> 5) & fiveBitMask; + int b = px & fiveBitMask; + // Note that this saves the data in RGB(A) order, so it doesn't need to be swapped later + out[0] = (stbi_uc)((r * 255)/31); + out[1] = (stbi_uc)((g * 255)/31); + out[2] = (stbi_uc)((b * 255)/31); + + // some people claim that the most significant bit might be used for alpha + // (possibly if an alpha-bit is set in the "image descriptor byte") + // but that only made 16bit test images completely translucent.. + // so let's treat all 15 and 16bit TGAs as RGB with no alpha. +} + +static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + // read in the TGA header stuff + int tga_offset = stbi__get8(s); + int tga_indexed = stbi__get8(s); + int tga_image_type = stbi__get8(s); + int tga_is_RLE = 0; + int tga_palette_start = stbi__get16le(s); + int tga_palette_len = stbi__get16le(s); + int tga_palette_bits = stbi__get8(s); + int tga_x_origin = stbi__get16le(s); + int tga_y_origin = stbi__get16le(s); + int tga_width = stbi__get16le(s); + int tga_height = stbi__get16le(s); + int tga_bits_per_pixel = stbi__get8(s); + int tga_comp, tga_rgb16=0; + int tga_inverted = stbi__get8(s); + // int tga_alpha_bits = tga_inverted & 15; // the 4 lowest bits - unused (useless?) + // image data + unsigned char *tga_data; + unsigned char *tga_palette = NULL; + int i, j; + unsigned char raw_data[4] = {0}; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + STBI_NOTUSED(ri); + STBI_NOTUSED(tga_x_origin); // @TODO + STBI_NOTUSED(tga_y_origin); // @TODO + + if (tga_height > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (tga_width > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // do a tiny bit of precessing + if ( tga_image_type >= 8 ) + { + tga_image_type -= 8; + tga_is_RLE = 1; + } + tga_inverted = 1 - ((tga_inverted >> 5) & 1); + + // If I'm paletted, then I'll use the number of bits from the palette + if ( tga_indexed ) tga_comp = stbi__tga_get_comp(tga_palette_bits, 0, &tga_rgb16); + else tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3), &tga_rgb16); + + if(!tga_comp) // shouldn't really happen, stbi__tga_test() should have ensured basic consistency + return stbi__errpuc("bad format", "Can't find out TGA pixelformat"); + + // tga info + *x = tga_width; + *y = tga_height; + if (comp) *comp = tga_comp; + + if (!stbi__mad3sizes_valid(tga_width, tga_height, tga_comp, 0)) + return stbi__errpuc("too large", "Corrupt TGA"); + + tga_data = (unsigned char*)stbi__malloc_mad3(tga_width, tga_height, tga_comp, 0); + if (!tga_data) return stbi__errpuc("outofmem", "Out of memory"); + + // skip to the data's starting position (offset usually = 0) + stbi__skip(s, tga_offset ); + + if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) { + for (i=0; i < tga_height; ++i) { + int row = tga_inverted ? tga_height -i - 1 : i; + stbi_uc *tga_row = tga_data + row*tga_width*tga_comp; + stbi__getn(s, tga_row, tga_width * tga_comp); + } + } else { + // do I need to load a palette? + if ( tga_indexed) + { + if (tga_palette_len == 0) { /* you have to have at least one entry! */ + STBI_FREE(tga_data); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + + // any data to skip? (offset usually = 0) + stbi__skip(s, tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)stbi__malloc_mad2(tga_palette_len, tga_comp, 0); + if (!tga_palette) { + STBI_FREE(tga_data); + return stbi__errpuc("outofmem", "Out of memory"); + } + if (tga_rgb16) { + stbi_uc *pal_entry = tga_palette; + STBI_ASSERT(tga_comp == STBI_rgb); + for (i=0; i < tga_palette_len; ++i) { + stbi__tga_read_rgb16(s, pal_entry); + pal_entry += tga_comp; + } + } else if (!stbi__getn(s, tga_palette, tga_palette_len * tga_comp)) { + STBI_FREE(tga_data); + STBI_FREE(tga_palette); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + } + // load the data + for (i=0; i < tga_width * tga_height; ++i) + { + // if I'm in RLE mode, do I need to get a RLE stbi__pngchunk? + if ( tga_is_RLE ) + { + if ( RLE_count == 0 ) + { + // yep, get the next byte as a RLE command + int RLE_cmd = stbi__get8(s); + RLE_count = 1 + (RLE_cmd & 127); + RLE_repeating = RLE_cmd >> 7; + read_next_pixel = 1; + } else if ( !RLE_repeating ) + { + read_next_pixel = 1; + } + } else + { + read_next_pixel = 1; + } + // OK, if I need to read a pixel, do it now + if ( read_next_pixel ) + { + // load however much data we did have + if ( tga_indexed ) + { + // read in index, then perform the lookup + int pal_idx = (tga_bits_per_pixel == 8) ? stbi__get8(s) : stbi__get16le(s); + if ( pal_idx >= tga_palette_len ) { + // invalid index + pal_idx = 0; + } + pal_idx *= tga_comp; + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = tga_palette[pal_idx+j]; + } + } else if(tga_rgb16) { + STBI_ASSERT(tga_comp == STBI_rgb); + stbi__tga_read_rgb16(s, raw_data); + } else { + // read in the data raw + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = stbi__get8(s); + } + } + // clear the reading flag for the next pixel + read_next_pixel = 0; + } // end of reading a pixel + + // copy data + for (j = 0; j < tga_comp; ++j) + tga_data[i*tga_comp+j] = raw_data[j]; + + // in case we're in RLE mode, keep counting down + --RLE_count; + } + // do I need to invert the image? + if ( tga_inverted ) + { + for (j = 0; j*2 < tga_height; ++j) + { + int index1 = j * tga_width * tga_comp; + int index2 = (tga_height - 1 - j) * tga_width * tga_comp; + for (i = tga_width * tga_comp; i > 0; --i) + { + unsigned char temp = tga_data[index1]; + tga_data[index1] = tga_data[index2]; + tga_data[index2] = temp; + ++index1; + ++index2; + } + } + } + // clear my palette, if I had one + if ( tga_palette != NULL ) + { + STBI_FREE( tga_palette ); + } + } + + // swap RGB - if the source data was RGB16, it already is in the right order + if (tga_comp >= 3 && !tga_rgb16) + { + unsigned char* tga_pixel = tga_data; + for (i=0; i < tga_width * tga_height; ++i) + { + unsigned char temp = tga_pixel[0]; + tga_pixel[0] = tga_pixel[2]; + tga_pixel[2] = temp; + tga_pixel += tga_comp; + } + } + + // convert to target component count + if (req_comp && req_comp != tga_comp) + tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height); + + // the things I do to get rid of an error message, and yet keep + // Microsoft's C compilers happy... [8^( + tga_palette_start = tga_palette_len = tga_palette_bits = + tga_x_origin = tga_y_origin = 0; + STBI_NOTUSED(tga_palette_start); + // OK, done + return tga_data; +} +#endif + +// ************************************************************************************************* +// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s) +{ + int r = (stbi__get32be(s) == 0x38425053); + stbi__rewind(s); + return r; +} + +static int stbi__psd_decode_rle(stbi__context *s, stbi_uc *p, int pixelCount) +{ + int count, nleft, len; + + count = 0; + while ((nleft = pixelCount - count) > 0) { + len = stbi__get8(s); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + if (len > nleft) return 0; // corrupt data + count += len; + while (len) { + *p = stbi__get8(s); + p += 4; + len--; + } + } else if (len > 128) { + stbi_uc val; + // Next -len+1 bytes in the dest are replicated from next source byte. + // (Interpret len as a negative 8-bit int.) + len = 257 - len; + if (len > nleft) return 0; // corrupt data + val = stbi__get8(s); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + + return 1; +} + +static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + int pixelCount; + int channelCount, compression; + int channel, i; + int bitdepth; + int w,h; + stbi_uc *out; + STBI_NOTUSED(ri); + + // Check identifier + if (stbi__get32be(s) != 0x38425053) // "8BPS" + return stbi__errpuc("not PSD", "Corrupt PSD image"); + + // Check file type version. + if (stbi__get16be(s) != 1) + return stbi__errpuc("wrong version", "Unsupported version of PSD image"); + + // Skip 6 reserved bytes. + stbi__skip(s, 6 ); + + // Read the number of channels (R, G, B, A, etc). + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) + return stbi__errpuc("wrong channel count", "Unsupported number of channels in PSD image"); + + // Read the rows and columns of the image. + h = stbi__get32be(s); + w = stbi__get32be(s); + + if (h > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (w > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // Make sure the depth is 8 bits. + bitdepth = stbi__get16be(s); + if (bitdepth != 8 && bitdepth != 16) + return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 or 16 bit"); + + // Make sure the color mode is RGB. + // Valid options are: + // 0: Bitmap + // 1: Grayscale + // 2: Indexed color + // 3: RGB color + // 4: CMYK color + // 7: Multichannel + // 8: Duotone + // 9: Lab color + if (stbi__get16be(s) != 3) + return stbi__errpuc("wrong color format", "PSD is not in RGB color format"); + + // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.) + stbi__skip(s,stbi__get32be(s) ); + + // Skip the image resources. (resolution, pen tool paths, etc) + stbi__skip(s, stbi__get32be(s) ); + + // Skip the reserved data. + stbi__skip(s, stbi__get32be(s) ); + + // Find out if the data is compressed. + // Known values: + // 0: no compression + // 1: RLE compressed + compression = stbi__get16be(s); + if (compression > 1) + return stbi__errpuc("bad compression", "PSD has an unknown compression format"); + + // Check size + if (!stbi__mad3sizes_valid(4, w, h, 0)) + return stbi__errpuc("too large", "Corrupt PSD"); + + // Create the destination image. + + if (!compression && bitdepth == 16 && bpc == 16) { + out = (stbi_uc *) stbi__malloc_mad3(8, w, h, 0); + ri->bits_per_channel = 16; + } else + out = (stbi_uc *) stbi__malloc(4 * w*h); + + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + pixelCount = w*h; + + // Initialize the data to zero. + //memset( out, 0, pixelCount * 4 ); + + // Finally, the image data. + if (compression) { + // RLE as used by .PSD and .TIFF + // Loop until you get the number of unpacked bytes you are expecting: + // Read the next source byte into n. + // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally. + // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times. + // Else if n is 128, noop. + // Endloop + + // The RLE-compressed data is preceded by a 2-byte data count for each row in the data, + // which we're going to just skip. + stbi__skip(s, h * channelCount * 2 ); + + // Read the RLE data by channel. + for (channel = 0; channel < 4; channel++) { + stbi_uc *p; + + p = out+channel; + if (channel >= channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++, p += 4) + *p = (channel == 3 ? 255 : 0); + } else { + // Read the RLE data. + if (!stbi__psd_decode_rle(s, p, pixelCount)) { + STBI_FREE(out); + return stbi__errpuc("corrupt", "bad RLE data"); + } + } + } + + } else { + // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...) + // where each channel consists of an 8-bit (or 16-bit) value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + if (channel >= channelCount) { + // Fill this channel with default data. + if (bitdepth == 16 && bpc == 16) { + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + stbi__uint16 val = channel == 3 ? 65535 : 0; + for (i = 0; i < pixelCount; i++, q += 4) + *q = val; + } else { + stbi_uc *p = out+channel; + stbi_uc val = channel == 3 ? 255 : 0; + for (i = 0; i < pixelCount; i++, p += 4) + *p = val; + } + } else { + if (ri->bits_per_channel == 16) { // output bpc + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + for (i = 0; i < pixelCount; i++, q += 4) + *q = (stbi__uint16) stbi__get16be(s); + } else { + stbi_uc *p = out+channel; + if (bitdepth == 16) { // input bpc + for (i = 0; i < pixelCount; i++, p += 4) + *p = (stbi_uc) (stbi__get16be(s) >> 8); + } else { + for (i = 0; i < pixelCount; i++, p += 4) + *p = stbi__get8(s); + } + } + } + } + } + + // remove weird white matte from PSD + if (channelCount >= 4) { + if (ri->bits_per_channel == 16) { + for (i=0; i < w*h; ++i) { + stbi__uint16 *pixel = (stbi__uint16 *) out + 4*i; + if (pixel[3] != 0 && pixel[3] != 65535) { + float a = pixel[3] / 65535.0f; + float ra = 1.0f / a; + float inv_a = 65535.0f * (1 - ra); + pixel[0] = (stbi__uint16) (pixel[0]*ra + inv_a); + pixel[1] = (stbi__uint16) (pixel[1]*ra + inv_a); + pixel[2] = (stbi__uint16) (pixel[2]*ra + inv_a); + } + } + } else { + for (i=0; i < w*h; ++i) { + unsigned char *pixel = out + 4*i; + if (pixel[3] != 0 && pixel[3] != 255) { + float a = pixel[3] / 255.0f; + float ra = 1.0f / a; + float inv_a = 255.0f * (1 - ra); + pixel[0] = (unsigned char) (pixel[0]*ra + inv_a); + pixel[1] = (unsigned char) (pixel[1]*ra + inv_a); + pixel[2] = (unsigned char) (pixel[2]*ra + inv_a); + } + } + } + } + + // convert to desired output format + if (req_comp && req_comp != 4) { + if (ri->bits_per_channel == 16) + out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, 4, req_comp, w, h); + else + out = stbi__convert_format(out, 4, req_comp, w, h); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + if (comp) *comp = 4; + *y = h; + *x = w; + + return out; +} +#endif + +// ************************************************************************************************* +// Softimage PIC loader +// by Tom Seddon +// +// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format +// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/ + +#ifndef STBI_NO_PIC +static int stbi__pic_is4(stbi__context *s,const char *str) +{ + int i; + for (i=0; i<4; ++i) + if (stbi__get8(s) != (stbi_uc)str[i]) + return 0; + + return 1; +} + +static int stbi__pic_test_core(stbi__context *s) +{ + int i; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) + return 0; + + for(i=0;i<84;++i) + stbi__get8(s); + + if (!stbi__pic_is4(s,"PICT")) + return 0; + + return 1; +} + +typedef struct +{ + stbi_uc size,type,channel; +} stbi__pic_packet; + +static stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest) +{ + int mask=0x80, i; + + for (i=0; i<4; ++i, mask>>=1) { + if (channel & mask) { + if (stbi__at_eof(s)) return stbi__errpuc("bad file","PIC file too short"); + dest[i]=stbi__get8(s); + } + } + + return dest; +} + +static void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src) +{ + int mask=0x80,i; + + for (i=0;i<4; ++i, mask>>=1) + if (channel&mask) + dest[i]=src[i]; +} + +static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result) +{ + int act_comp=0,num_packets=0,y,chained; + stbi__pic_packet packets[10]; + + // this will (should...) cater for even some bizarre stuff like having data + // for the same channel in multiple packets. + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return stbi__errpuc("bad format","too many packets"); + + packet = &packets[num_packets++]; + + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + + act_comp |= packet->channel; + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (reading packets)"); + if (packet->size != 8) return stbi__errpuc("bad format","packet isn't 8bpp"); + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel? + + for(y=0; ytype) { + default: + return stbi__errpuc("bad format","packet has bad compression type"); + + case 0: {//uncompressed + int x; + + for(x=0;xchannel,dest)) + return 0; + break; + } + + case 1://Pure RLE + { + int left=width, i; + + while (left>0) { + stbi_uc count,value[4]; + + count=stbi__get8(s); + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pure read count)"); + + if (count > left) + count = (stbi_uc) left; + + if (!stbi__readval(s,packet->channel,value)) return 0; + + for(i=0; ichannel,dest,value); + left -= count; + } + } + break; + + case 2: {//Mixed RLE + int left=width; + while (left>0) { + int count = stbi__get8(s), i; + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (mixed read count)"); + + if (count >= 128) { // Repeated + stbi_uc value[4]; + + if (count==128) + count = stbi__get16be(s); + else + count -= 127; + if (count > left) + return stbi__errpuc("bad file","scanline overrun"); + + if (!stbi__readval(s,packet->channel,value)) + return 0; + + for(i=0;ichannel,dest,value); + } else { // Raw + ++count; + if (count>left) return stbi__errpuc("bad file","scanline overrun"); + + for(i=0;ichannel,dest)) + return 0; + } + left-=count; + } + break; + } + } + } + } + + return result; +} + +static void *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp, stbi__result_info *ri) +{ + stbi_uc *result; + int i, x,y, internal_comp; + STBI_NOTUSED(ri); + + if (!comp) comp = &internal_comp; + + for (i=0; i<92; ++i) + stbi__get8(s); + + x = stbi__get16be(s); + y = stbi__get16be(s); + + if (y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pic header)"); + if (!stbi__mad3sizes_valid(x, y, 4, 0)) return stbi__errpuc("too large", "PIC image too large to decode"); + + stbi__get32be(s); //skip `ratio' + stbi__get16be(s); //skip `fields' + stbi__get16be(s); //skip `pad' + + // intermediate buffer is RGBA + result = (stbi_uc *) stbi__malloc_mad3(x, y, 4, 0); + if (!result) return stbi__errpuc("outofmem", "Out of memory"); + memset(result, 0xff, x*y*4); + + if (!stbi__pic_load_core(s,x,y,comp, result)) { + STBI_FREE(result); + result=0; + } + *px = x; + *py = y; + if (req_comp == 0) req_comp = *comp; + result=stbi__convert_format(result,4,req_comp,x,y); + + return result; +} + +static int stbi__pic_test(stbi__context *s) +{ + int r = stbi__pic_test_core(s); + stbi__rewind(s); + return r; +} +#endif + +// ************************************************************************************************* +// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb + +#ifndef STBI_NO_GIF +typedef struct +{ + stbi__int16 prefix; + stbi_uc first; + stbi_uc suffix; +} stbi__gif_lzw; + +typedef struct +{ + int w,h; + stbi_uc *out; // output buffer (always 4 components) + stbi_uc *background; // The current "background" as far as a gif is concerned + stbi_uc *history; + int flags, bgindex, ratio, transparent, eflags; + stbi_uc pal[256][4]; + stbi_uc lpal[256][4]; + stbi__gif_lzw codes[8192]; + stbi_uc *color_table; + int parse, step; + int lflags; + int start_x, start_y; + int max_x, max_y; + int cur_x, cur_y; + int line_size; + int delay; +} stbi__gif; + +static int stbi__gif_test_raw(stbi__context *s) +{ + int sz; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0; + sz = stbi__get8(s); + if (sz != '9' && sz != '7') return 0; + if (stbi__get8(s) != 'a') return 0; + return 1; +} + +static int stbi__gif_test(stbi__context *s) +{ + int r = stbi__gif_test_raw(s); + stbi__rewind(s); + return r; +} + +static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp) +{ + int i; + for (i=0; i < num_entries; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + pal[i][3] = transp == i ? 0 : 255; + } +} + +static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info) +{ + stbi_uc version; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') + return stbi__err("not GIF", "Corrupt GIF"); + + version = stbi__get8(s); + if (version != '7' && version != '9') return stbi__err("not GIF", "Corrupt GIF"); + if (stbi__get8(s) != 'a') return stbi__err("not GIF", "Corrupt GIF"); + + stbi__g_failure_reason = ""; + g->w = stbi__get16le(s); + g->h = stbi__get16le(s); + g->flags = stbi__get8(s); + g->bgindex = stbi__get8(s); + g->ratio = stbi__get8(s); + g->transparent = -1; + + if (g->w > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (g->h > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + + if (comp != 0) *comp = 4; // can't actually tell whether it's 3 or 4 until we parse the comments + + if (is_info) return 1; + + if (g->flags & 0x80) + stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1); + + return 1; +} + +static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__gif* g = (stbi__gif*) stbi__malloc(sizeof(stbi__gif)); + if (!g) return stbi__err("outofmem", "Out of memory"); + if (!stbi__gif_header(s, g, comp, 1)) { + STBI_FREE(g); + stbi__rewind( s ); + return 0; + } + if (x) *x = g->w; + if (y) *y = g->h; + STBI_FREE(g); + return 1; +} + +static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code) +{ + stbi_uc *p, *c; + int idx; + + // recurse to decode the prefixes, since the linked-list is backwards, + // and working backwards through an interleaved image would be nasty + if (g->codes[code].prefix >= 0) + stbi__out_gif_code(g, g->codes[code].prefix); + + if (g->cur_y >= g->max_y) return; + + idx = g->cur_x + g->cur_y; + p = &g->out[idx]; + g->history[idx / 4] = 1; + + c = &g->color_table[g->codes[code].suffix * 4]; + if (c[3] > 128) { // don't render transparent pixels; + p[0] = c[2]; + p[1] = c[1]; + p[2] = c[0]; + p[3] = c[3]; + } + g->cur_x += 4; + + if (g->cur_x >= g->max_x) { + g->cur_x = g->start_x; + g->cur_y += g->step; + + while (g->cur_y >= g->max_y && g->parse > 0) { + g->step = (1 << g->parse) * g->line_size; + g->cur_y = g->start_y + (g->step >> 1); + --g->parse; + } + } +} + +static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g) +{ + stbi_uc lzw_cs; + stbi__int32 len, init_code; + stbi__uint32 first; + stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear; + stbi__gif_lzw *p; + + lzw_cs = stbi__get8(s); + if (lzw_cs > 12) return NULL; + clear = 1 << lzw_cs; + first = 1; + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + bits = 0; + valid_bits = 0; + for (init_code = 0; init_code < clear; init_code++) { + g->codes[init_code].prefix = -1; + g->codes[init_code].first = (stbi_uc) init_code; + g->codes[init_code].suffix = (stbi_uc) init_code; + } + + // support no starting clear code + avail = clear+2; + oldcode = -1; + + len = 0; + for(;;) { + if (valid_bits < codesize) { + if (len == 0) { + len = stbi__get8(s); // start new block + if (len == 0) + return g->out; + } + --len; + bits |= (stbi__int32) stbi__get8(s) << valid_bits; + valid_bits += 8; + } else { + stbi__int32 code = bits & codemask; + bits >>= codesize; + valid_bits -= codesize; + // @OPTIMIZE: is there some way we can accelerate the non-clear path? + if (code == clear) { // clear code + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + avail = clear + 2; + oldcode = -1; + first = 0; + } else if (code == clear + 1) { // end of stream code + stbi__skip(s, len); + while ((len = stbi__get8(s)) > 0) + stbi__skip(s,len); + return g->out; + } else if (code <= avail) { + if (first) { + return stbi__errpuc("no clear code", "Corrupt GIF"); + } + + if (oldcode >= 0) { + p = &g->codes[avail++]; + if (avail > 8192) { + return stbi__errpuc("too many codes", "Corrupt GIF"); + } + + p->prefix = (stbi__int16) oldcode; + p->first = g->codes[oldcode].first; + p->suffix = (code == avail) ? p->first : g->codes[code].first; + } else if (code == avail) + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + + stbi__out_gif_code(g, (stbi__uint16) code); + + if ((avail & codemask) == 0 && avail <= 0x0FFF) { + codesize++; + codemask = (1 << codesize) - 1; + } + + oldcode = code; + } else { + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + } + } + } +} + +// this function is designed to support animated gifs, although stb_image doesn't support it +// two back is the image from two frames ago, used for a very specific disposal format +static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp, stbi_uc *two_back) +{ + int dispose; + int first_frame; + int pi; + int pcount; + STBI_NOTUSED(req_comp); + + // on first frame, any non-written pixels get the background colour (non-transparent) + first_frame = 0; + if (g->out == 0) { + if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header + if (!stbi__mad3sizes_valid(4, g->w, g->h, 0)) + return stbi__errpuc("too large", "GIF image is too large"); + pcount = g->w * g->h; + g->out = (stbi_uc *) stbi__malloc(4 * pcount); + g->background = (stbi_uc *) stbi__malloc(4 * pcount); + g->history = (stbi_uc *) stbi__malloc(pcount); + if (!g->out || !g->background || !g->history) + return stbi__errpuc("outofmem", "Out of memory"); + + // image is treated as "transparent" at the start - ie, nothing overwrites the current background; + // background colour is only used for pixels that are not rendered first frame, after that "background" + // color refers to the color that was there the previous frame. + memset(g->out, 0x00, 4 * pcount); + memset(g->background, 0x00, 4 * pcount); // state of the background (starts transparent) + memset(g->history, 0x00, pcount); // pixels that were affected previous frame + first_frame = 1; + } else { + // second frame - how do we dispose of the previous one? + dispose = (g->eflags & 0x1C) >> 2; + pcount = g->w * g->h; + + if ((dispose == 3) && (two_back == 0)) { + dispose = 2; // if I don't have an image to revert back to, default to the old background + } + + if (dispose == 3) { // use previous graphic + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &two_back[pi * 4], 4 ); + } + } + } else if (dispose == 2) { + // restore what was changed last frame to background before that frame; + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &g->background[pi * 4], 4 ); + } + } + } else { + // This is a non-disposal case eithe way, so just + // leave the pixels as is, and they will become the new background + // 1: do not dispose + // 0: not specified. + } + + // background is what out is after the undoing of the previou frame; + memcpy( g->background, g->out, 4 * g->w * g->h ); + } + + // clear my history; + memset( g->history, 0x00, g->w * g->h ); // pixels that were affected previous frame + + for (;;) { + int tag = stbi__get8(s); + switch (tag) { + case 0x2C: /* Image Descriptor */ + { + stbi__int32 x, y, w, h; + stbi_uc *o; + + x = stbi__get16le(s); + y = stbi__get16le(s); + w = stbi__get16le(s); + h = stbi__get16le(s); + if (((x + w) > (g->w)) || ((y + h) > (g->h))) + return stbi__errpuc("bad Image Descriptor", "Corrupt GIF"); + + g->line_size = g->w * 4; + g->start_x = x * 4; + g->start_y = y * g->line_size; + g->max_x = g->start_x + w * 4; + g->max_y = g->start_y + h * g->line_size; + g->cur_x = g->start_x; + g->cur_y = g->start_y; + + // if the width of the specified rectangle is 0, that means + // we may not see *any* pixels or the image is malformed; + // to make sure this is caught, move the current y down to + // max_y (which is what out_gif_code checks). + if (w == 0) + g->cur_y = g->max_y; + + g->lflags = stbi__get8(s); + + if (g->lflags & 0x40) { + g->step = 8 * g->line_size; // first interlaced spacing + g->parse = 3; + } else { + g->step = g->line_size; + g->parse = 0; + } + + if (g->lflags & 0x80) { + stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1); + g->color_table = (stbi_uc *) g->lpal; + } else if (g->flags & 0x80) { + g->color_table = (stbi_uc *) g->pal; + } else + return stbi__errpuc("missing color table", "Corrupt GIF"); + + o = stbi__process_gif_raster(s, g); + if (!o) return NULL; + + // if this was the first frame, + pcount = g->w * g->h; + if (first_frame && (g->bgindex > 0)) { + // if first frame, any pixel not drawn to gets the background color + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi] == 0) { + g->pal[g->bgindex][3] = 255; // just in case it was made transparent, undo that; It will be reset next frame if need be; + memcpy( &g->out[pi * 4], &g->pal[g->bgindex], 4 ); + } + } + } + + return o; + } + + case 0x21: // Comment Extension. + { + int len; + int ext = stbi__get8(s); + if (ext == 0xF9) { // Graphic Control Extension. + len = stbi__get8(s); + if (len == 4) { + g->eflags = stbi__get8(s); + g->delay = 10 * stbi__get16le(s); // delay - 1/100th of a second, saving as 1/1000ths. + + // unset old transparent + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 255; + } + if (g->eflags & 0x01) { + g->transparent = stbi__get8(s); + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 0; + } + } else { + // don't need transparent + stbi__skip(s, 1); + g->transparent = -1; + } + } else { + stbi__skip(s, len); + break; + } + } + while ((len = stbi__get8(s)) != 0) { + stbi__skip(s, len); + } + break; + } + + case 0x3B: // gif stream termination code + return (stbi_uc *) s; // using '1' causes warning on some compilers + + default: + return stbi__errpuc("unknown code", "Corrupt GIF"); + } + } +} + +static void *stbi__load_gif_main_outofmem(stbi__gif *g, stbi_uc *out, int **delays) +{ + STBI_FREE(g->out); + STBI_FREE(g->history); + STBI_FREE(g->background); + + if (out) STBI_FREE(out); + if (delays && *delays) STBI_FREE(*delays); + return stbi__errpuc("outofmem", "Out of memory"); +} + +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + if (stbi__gif_test(s)) { + int layers = 0; + stbi_uc *u = 0; + stbi_uc *out = 0; + stbi_uc *two_back = 0; + stbi__gif g; + int stride; + int out_size = 0; + int delays_size = 0; + + STBI_NOTUSED(out_size); + STBI_NOTUSED(delays_size); + + memset(&g, 0, sizeof(g)); + if (delays) { + *delays = 0; + } + + do { + u = stbi__gif_load_next(s, &g, comp, req_comp, two_back); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + + if (u) { + *x = g.w; + *y = g.h; + ++layers; + stride = g.w * g.h * 4; + + if (out) { + void *tmp = (stbi_uc*) STBI_REALLOC_SIZED( out, out_size, layers * stride ); + if (!tmp) + return stbi__load_gif_main_outofmem(&g, out, delays); + else { + out = (stbi_uc*) tmp; + out_size = layers * stride; + } + + if (delays) { + int *new_delays = (int*) STBI_REALLOC_SIZED( *delays, delays_size, sizeof(int) * layers ); + if (!new_delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + *delays = new_delays; + delays_size = layers * sizeof(int); + } + } else { + out = (stbi_uc*)stbi__malloc( layers * stride ); + if (!out) + return stbi__load_gif_main_outofmem(&g, out, delays); + out_size = layers * stride; + if (delays) { + *delays = (int*) stbi__malloc( layers * sizeof(int) ); + if (!*delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + delays_size = layers * sizeof(int); + } + } + memcpy( out + ((layers - 1) * stride), u, stride ); + if (layers >= 2) { + two_back = out - 2 * stride; + } + + if (delays) { + (*delays)[layers - 1U] = g.delay; + } + } + } while (u != 0); + + // free temp buffer; + STBI_FREE(g.out); + STBI_FREE(g.history); + STBI_FREE(g.background); + + // do the final conversion after loading everything; + if (req_comp && req_comp != 4) + out = stbi__convert_format(out, 4, req_comp, layers * g.w, g.h); + + *z = layers; + return out; + } else { + return stbi__errpuc("not GIF", "Image was not as a gif type."); + } +} + +static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *u = 0; + stbi__gif g; + memset(&g, 0, sizeof(g)); + STBI_NOTUSED(ri); + + u = stbi__gif_load_next(s, &g, comp, req_comp, 0); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + if (u) { + *x = g.w; + *y = g.h; + + // moved conversion to after successful load so that the same + // can be done for multiple frames. + if (req_comp && req_comp != 4) + u = stbi__convert_format(u, 4, req_comp, g.w, g.h); + } else if (g.out) { + // if there was an error and we allocated an image buffer, free it! + STBI_FREE(g.out); + } + + // free buffers needed for multiple frame loading; + STBI_FREE(g.history); + STBI_FREE(g.background); + + return u; +} + +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp) +{ + return stbi__gif_info_raw(s,x,y,comp); +} +#endif + +// ************************************************************************************************* +// Radiance RGBE HDR loader +// originally by Nicolas Schulz +#ifndef STBI_NO_HDR +static int stbi__hdr_test_core(stbi__context *s, const char *signature) +{ + int i; + for (i=0; signature[i]; ++i) + if (stbi__get8(s) != signature[i]) + return 0; + stbi__rewind(s); + return 1; +} + +static int stbi__hdr_test(stbi__context* s) +{ + int r = stbi__hdr_test_core(s, "#?RADIANCE\n"); + stbi__rewind(s); + if(!r) { + r = stbi__hdr_test_core(s, "#?RGBE\n"); + stbi__rewind(s); + } + return r; +} + +#define STBI__HDR_BUFLEN 1024 +static char *stbi__hdr_gettoken(stbi__context *z, char *buffer) +{ + int len=0; + char c = '\0'; + + c = (char) stbi__get8(z); + + while (!stbi__at_eof(z) && c != '\n') { + buffer[len++] = c; + if (len == STBI__HDR_BUFLEN-1) { + // flush to end of line + while (!stbi__at_eof(z) && stbi__get8(z) != '\n') + ; + break; + } + c = (char) stbi__get8(z); + } + + buffer[len] = 0; + return buffer; +} + +static void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp) +{ + if ( input[3] != 0 ) { + float f1; + // Exponent + f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8)); + if (req_comp <= 2) + output[0] = (input[0] + input[1] + input[2]) * f1 / 3; + else { + output[0] = input[0] * f1; + output[1] = input[1] * f1; + output[2] = input[2] * f1; + } + if (req_comp == 2) output[1] = 1; + if (req_comp == 4) output[3] = 1; + } else { + switch (req_comp) { + case 4: output[3] = 1; /* fallthrough */ + case 3: output[0] = output[1] = output[2] = 0; + break; + case 2: output[1] = 1; /* fallthrough */ + case 1: output[0] = 0; + break; + } + } +} + +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + float *hdr_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + const char *headerToken; + STBI_NOTUSED(ri); + + // Check identifier + headerToken = stbi__hdr_gettoken(s,buffer); + if (strcmp(headerToken, "#?RADIANCE") != 0 && strcmp(headerToken, "#?RGBE") != 0) + return stbi__errpf("not HDR", "Corrupt HDR image"); + + // Parse header + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return stbi__errpf("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = (int) strtol(token, NULL, 10); + + if (height > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + if (width > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + + *x = width; + *y = height; + + if (comp) *comp = 3; + if (req_comp == 0) req_comp = 3; + + if (!stbi__mad4sizes_valid(width, height, req_comp, sizeof(float), 0)) + return stbi__errpf("too large", "HDR image is too large"); + + // Read data + hdr_data = (float *) stbi__malloc_mad4(width, height, req_comp, sizeof(float), 0); + if (!hdr_data) + return stbi__errpf("outofmem", "Out of memory"); + + // Load image data + // image data is stored as some number of sca + if ( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + stbi_uc rgbe[4]; + main_decode_loop: + stbi__getn(s, rgbe, 4); + stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp); + } + } + } else { + // Read RLE-encoded data + scanline = NULL; + + for (j = 0; j < height; ++j) { + c1 = stbi__get8(s); + c2 = stbi__get8(s); + len = stbi__get8(s); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + stbi_uc rgbe[4]; + rgbe[0] = (stbi_uc) c1; + rgbe[1] = (stbi_uc) c2; + rgbe[2] = (stbi_uc) len; + rgbe[3] = (stbi_uc) stbi__get8(s); + stbi__hdr_convert(hdr_data, rgbe, req_comp); + i = 1; + j = 0; + STBI_FREE(scanline); + goto main_decode_loop; // yes, this makes no sense + } + len <<= 8; + len |= stbi__get8(s); + if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); } + if (scanline == NULL) { + scanline = (stbi_uc *) stbi__malloc_mad2(width, 4, 0); + if (!scanline) { + STBI_FREE(hdr_data); + return stbi__errpf("outofmem", "Out of memory"); + } + } + + for (k = 0; k < 4; ++k) { + int nleft; + i = 0; + while ((nleft = width - i) > 0) { + count = stbi__get8(s); + if (count > 128) { + // Run + value = stbi__get8(s); + count -= 128; + if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + if (count > nleft) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = stbi__get8(s); + } + } + } + for (i=0; i < width; ++i) + stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); + } + if (scanline) + STBI_FREE(scanline); + } + + return hdr_data; +} + +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + int dummy; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + if (stbi__hdr_test(s) == 0) { + stbi__rewind( s ); + return 0; + } + + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) { + stbi__rewind( s ); + return 0; + } + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *y = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *x = (int) strtol(token, NULL, 10); + *comp = 3; + return 1; +} +#endif // STBI_NO_HDR + +#ifndef STBI_NO_BMP +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp) +{ + void *p; + stbi__bmp_data info; + + info.all_a = 255; + p = stbi__bmp_parse_header(s, &info); + if (p == NULL) { + stbi__rewind( s ); + return 0; + } + if (x) *x = s->img_x; + if (y) *y = s->img_y; + if (comp) { + if (info.bpp == 24 && info.ma == 0xff000000) + *comp = 3; + else + *comp = info.ma ? 4 : 3; + } + return 1; +} +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp) +{ + int channelCount, dummy, depth; + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + if (stbi__get32be(s) != 0x38425053) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 1) { + stbi__rewind( s ); + return 0; + } + stbi__skip(s, 6); + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) { + stbi__rewind( s ); + return 0; + } + *y = stbi__get32be(s); + *x = stbi__get32be(s); + depth = stbi__get16be(s); + if (depth != 8 && depth != 16) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 3) { + stbi__rewind( s ); + return 0; + } + *comp = 4; + return 1; +} + +static int stbi__psd_is16(stbi__context *s) +{ + int channelCount, depth; + if (stbi__get32be(s) != 0x38425053) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 1) { + stbi__rewind( s ); + return 0; + } + stbi__skip(s, 6); + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) { + stbi__rewind( s ); + return 0; + } + STBI_NOTUSED(stbi__get32be(s)); + STBI_NOTUSED(stbi__get32be(s)); + depth = stbi__get16be(s); + if (depth != 16) { + stbi__rewind( s ); + return 0; + } + return 1; +} +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp) +{ + int act_comp=0,num_packets=0,chained,dummy; + stbi__pic_packet packets[10]; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) { + stbi__rewind(s); + return 0; + } + + stbi__skip(s, 88); + + *x = stbi__get16be(s); + *y = stbi__get16be(s); + if (stbi__at_eof(s)) { + stbi__rewind( s); + return 0; + } + if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) { + stbi__rewind( s ); + return 0; + } + + stbi__skip(s, 8); + + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return 0; + + packet = &packets[num_packets++]; + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + act_comp |= packet->channel; + + if (stbi__at_eof(s)) { + stbi__rewind( s ); + return 0; + } + if (packet->size != 8) { + stbi__rewind( s ); + return 0; + } + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); + + return 1; +} +#endif + +// ************************************************************************************************* +// Portable Gray Map and Portable Pixel Map loader +// by Ken Miller +// +// PGM: http://netpbm.sourceforge.net/doc/pgm.html +// PPM: http://netpbm.sourceforge.net/doc/ppm.html +// +// Known limitations: +// Does not support comments in the header section +// Does not support ASCII image data (formats P2 and P3) + +#ifndef STBI_NO_PNM + +static int stbi__pnm_test(stbi__context *s) +{ + char p, t; + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind( s ); + return 0; + } + return 1; +} + +static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *out; + STBI_NOTUSED(ri); + + ri->bits_per_channel = stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n); + if (ri->bits_per_channel == 0) + return 0; + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + + if (!stbi__mad4sizes_valid(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0)) + return stbi__errpuc("too large", "PNM too large"); + + out = (stbi_uc *) stbi__malloc_mad4(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + stbi__getn(s, out, s->img_n * s->img_x * s->img_y * (ri->bits_per_channel / 8)); + + if (req_comp && req_comp != s->img_n) { + out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + return out; +} + +static int stbi__pnm_isspace(char c) +{ + return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r'; +} + +static void stbi__pnm_skip_whitespace(stbi__context *s, char *c) +{ + for (;;) { + while (!stbi__at_eof(s) && stbi__pnm_isspace(*c)) + *c = (char) stbi__get8(s); + + if (stbi__at_eof(s) || *c != '#') + break; + + while (!stbi__at_eof(s) && *c != '\n' && *c != '\r' ) + *c = (char) stbi__get8(s); + } +} + +static int stbi__pnm_isdigit(char c) +{ + return c >= '0' && c <= '9'; +} + +static int stbi__pnm_getinteger(stbi__context *s, char *c) +{ + int value = 0; + + while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) { + value = value*10 + (*c - '0'); + *c = (char) stbi__get8(s); + } + + return value; +} + +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp) +{ + int maxv, dummy; + char c, p, t; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + stbi__rewind(s); + + // Get identifier + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind(s); + return 0; + } + + *comp = (t == '6') ? 3 : 1; // '5' is 1-component .pgm; '6' is 3-component .ppm + + c = (char) stbi__get8(s); + stbi__pnm_skip_whitespace(s, &c); + + *x = stbi__pnm_getinteger(s, &c); // read width + stbi__pnm_skip_whitespace(s, &c); + + *y = stbi__pnm_getinteger(s, &c); // read height + stbi__pnm_skip_whitespace(s, &c); + + maxv = stbi__pnm_getinteger(s, &c); // read max value + if (maxv > 65535) + return stbi__err("max value > 65535", "PPM image supports only 8-bit and 16-bit images"); + else if (maxv > 255) + return 16; + else + return 8; +} + +static int stbi__pnm_is16(stbi__context *s) +{ + if (stbi__pnm_info(s, NULL, NULL, NULL) == 16) + return 1; + return 0; +} +#endif + +static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp) +{ + #ifndef STBI_NO_JPEG + if (stbi__jpeg_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNG + if (stbi__png_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_GIF + if (stbi__gif_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_BMP + if (stbi__bmp_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PIC + if (stbi__pic_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_info(s, x, y, comp)) return 1; + #endif + + // test tga last because it's a crappy test! + #ifndef STBI_NO_TGA + if (stbi__tga_info(s, x, y, comp)) + return 1; + #endif + return stbi__err("unknown image type", "Image not of any known type, or corrupt"); +} + +static int stbi__is_16_main(stbi__context *s) +{ + #ifndef STBI_NO_PNG + if (stbi__png_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_is16(s)) return 1; + #endif + return 0; +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_info_from_file(f, x, y, comp); + fclose(f); + return result; +} + +STBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__info_main(&s,x,y,comp); + fseek(f,pos,SEEK_SET); + return r; +} + +STBIDEF int stbi_is_16_bit(char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_is_16_bit_from_file(f); + fclose(f); + return result; +} + +STBIDEF int stbi_is_16_bit_from_file(FILE *f) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__is_16_main(&s); + fseek(f,pos,SEEK_SET); + return r; +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__info_main(&s,x,y,comp); +} + +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__info_main(&s,x,y,comp); +} + +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__is_16_main(&s); +} + +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *c, void *user) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__is_16_main(&s); +} + +#endif // STB_IMAGE_IMPLEMENTATION + +/* + revision history: + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) change sbti__shiftsigned to avoid clang -O2 bug + 1-bit BMP + *_is_16_bit api + avoid warnings + 2.16 (2017-07-23) all functions have 16-bit variants; + STBI_NO_STDIO works again; + compilation fixes; + fix rounding in unpremultiply; + optimize vertical flip; + disable raw_len validation; + documentation fixes + 2.15 (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode; + warning fixes; disable run-time SSE detection on gcc; + uniform handling of optional "return" values; + thread-safe initialization of zlib tables + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-11-29) add 16-bit API, only supported for PNG right now + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) allocate large structures on the stack + remove white matting for transparent PSD + fix reported channel count for PNG & BMP + re-enable SSE2 in non-gcc 64-bit + support RGB-formatted JPEG + read 16-bit PNGs (only as 8-bit) + 2.10 (2016-01-22) avoid warning introduced in 2.09 by STBI_REALLOC_SIZED + 2.09 (2016-01-16) allow comments in PNM files + 16-bit-per-pixel TGA (not bit-per-component) + info() for TGA could break due to .hdr handling + info() for BMP to shares code instead of sloppy parse + can use STBI_REALLOC_SIZED if allocator doesn't support realloc + code cleanup + 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA + 2.07 (2015-09-13) fix compiler warnings + partial animated GIF support + limited 16-bpc PSD support + #ifdef unused functions + bug with < 92 byte PIC,PNM,HDR,TGA + 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value + 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning + 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit + 2.03 (2015-04-12) extra corruption checking (mmozeiko) + stbi_set_flip_vertically_on_load (nguillemot) + fix NEON support; fix mingw support + 2.02 (2015-01-19) fix incorrect assert, fix warning + 2.01 (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2 + 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG + 2.00 (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg) + progressive JPEG (stb) + PGM/PPM support (Ken Miller) + STBI_MALLOC,STBI_REALLOC,STBI_FREE + GIF bugfix -- seemingly never worked + STBI_NO_*, STBI_ONLY_* + 1.48 (2014-12-14) fix incorrectly-named assert() + 1.47 (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb) + optimize PNG (ryg) + fix bug in interlaced PNG with user-specified channel count (stb) + 1.46 (2014-08-26) + fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG + 1.45 (2014-08-16) + fix MSVC-ARM internal compiler error by wrapping malloc + 1.44 (2014-08-07) + various warning fixes from Ronny Chevalier + 1.43 (2014-07-15) + fix MSVC-only compiler problem in code changed in 1.42 + 1.42 (2014-07-09) + don't define _CRT_SECURE_NO_WARNINGS (affects user code) + fixes to stbi__cleanup_jpeg path + added STBI_ASSERT to avoid requiring assert.h + 1.41 (2014-06-25) + fix search&replace from 1.36 that messed up comments/error messages + 1.40 (2014-06-22) + fix gcc struct-initialization warning + 1.39 (2014-06-15) + fix to TGA optimization when req_comp != number of components in TGA; + fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite) + add support for BMP version 5 (more ignored fields) + 1.38 (2014-06-06) + suppress MSVC warnings on integer casts truncating values + fix accidental rename of 'skip' field of I/O + 1.37 (2014-06-04) + remove duplicate typedef + 1.36 (2014-06-03) + convert to header file single-file library + if de-iphone isn't set, load iphone images color-swapped instead of returning NULL + 1.35 (2014-05-27) + various warnings + fix broken STBI_SIMD path + fix bug where stbi_load_from_file no longer left file pointer in correct place + fix broken non-easy path for 32-bit BMP (possibly never used) + TGA optimization by Arseny Kapoulkine + 1.34 (unknown) + use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case + 1.33 (2011-07-14) + make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements + 1.32 (2011-07-13) + support for "info" function for all supported filetypes (SpartanJ) + 1.31 (2011-06-20) + a few more leak fixes, bug in PNG handling (SpartanJ) + 1.30 (2011-06-11) + added ability to load files via callbacks to accomidate custom input streams (Ben Wenger) + removed deprecated format-specific test/load functions + removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway + error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha) + fix inefficiency in decoding 32-bit BMP (David Woo) + 1.29 (2010-08-16) + various warning fixes from Aurelien Pocheville + 1.28 (2010-08-01) + fix bug in GIF palette transparency (SpartanJ) + 1.27 (2010-08-01) + cast-to-stbi_uc to fix warnings + 1.26 (2010-07-24) + fix bug in file buffering for PNG reported by SpartanJ + 1.25 (2010-07-17) + refix trans_data warning (Won Chun) + 1.24 (2010-07-12) + perf improvements reading from files on platforms with lock-heavy fgetc() + minor perf improvements for jpeg + deprecated type-specific functions so we'll get feedback if they're needed + attempt to fix trans_data warning (Won Chun) + 1.23 fixed bug in iPhone support + 1.22 (2010-07-10) + removed image *writing* support + stbi_info support from Jetro Lauha + GIF support from Jean-Marc Lienher + iPhone PNG-extensions from James Brown + warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva) + 1.21 fix use of 'stbi_uc' in header (reported by jon blow) + 1.20 added support for Softimage PIC, by Tom Seddon + 1.19 bug in interlaced PNG corruption check (found by ryg) + 1.18 (2008-08-02) + fix a threading bug (local mutable static) + 1.17 support interlaced PNG + 1.16 major bugfix - stbi__convert_format converted one too many pixels + 1.15 initialize some fields for thread safety + 1.14 fix threadsafe conversion bug + header-file-only version (#define STBI_HEADER_FILE_ONLY before including) + 1.13 threadsafe + 1.12 const qualifiers in the API + 1.11 Support installable IDCT, colorspace conversion routines + 1.10 Fixes for 64-bit (don't use "unsigned long") + optimized upsampling by Fabian "ryg" Giesen + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant + 0.50 (2006-11-19) + first released version +*/ + + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/thirdparty/vk_mem_alloc.h b/thirdparty/vk_mem_alloc.h new file mode 100644 index 0000000..6687871 --- /dev/null +++ b/thirdparty/vk_mem_alloc.h @@ -0,0 +1,19229 @@ +// +// Copyright (c) 2017-2020 Advanced Micro Devices, Inc. All rights reserved. +// +// Permission is hereby granted, free of charge, to any person obtaining a copy +// of this software and associated documentation files (the "Software"), to deal +// in the Software without restriction, including without limitation the rights +// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +// copies of the Software, and to permit persons to whom the Software is +// furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in +// all copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN +// THE SOFTWARE. +// + +#ifndef AMD_VULKAN_MEMORY_ALLOCATOR_H +#define AMD_VULKAN_MEMORY_ALLOCATOR_H + +/** \mainpage Vulkan Memory Allocator + +Version 3.0.0-development (2020-11-03) + +Copyright (c) 2017-2020 Advanced Micro Devices, Inc. All rights reserved. \n +License: MIT + +Documentation of all members: vk_mem_alloc.h + +\section main_table_of_contents Table of contents + +- User guide + - \subpage quick_start + - [Project setup](@ref quick_start_project_setup) + - [Initialization](@ref quick_start_initialization) + - [Resource allocation](@ref quick_start_resource_allocation) + - \subpage choosing_memory_type + - [Usage](@ref choosing_memory_type_usage) + - [Required and preferred flags](@ref choosing_memory_type_required_preferred_flags) + - [Explicit memory types](@ref choosing_memory_type_explicit_memory_types) + - [Custom memory pools](@ref choosing_memory_type_custom_memory_pools) + - [Dedicated allocations](@ref choosing_memory_type_dedicated_allocations) + - \subpage memory_mapping + - [Mapping functions](@ref memory_mapping_mapping_functions) + - [Persistently mapped memory](@ref memory_mapping_persistently_mapped_memory) + - [Cache flush and invalidate](@ref memory_mapping_cache_control) + - [Finding out if memory is mappable](@ref memory_mapping_finding_if_memory_mappable) + - \subpage staying_within_budget + - [Querying for budget](@ref staying_within_budget_querying_for_budget) + - [Controlling memory usage](@ref staying_within_budget_controlling_memory_usage) + - \subpage resource_aliasing + - \subpage custom_memory_pools + - [Choosing memory type index](@ref custom_memory_pools_MemTypeIndex) + - [Linear allocation algorithm](@ref linear_algorithm) + - [Free-at-once](@ref linear_algorithm_free_at_once) + - [Stack](@ref linear_algorithm_stack) + - [Double stack](@ref linear_algorithm_double_stack) + - [Ring buffer](@ref linear_algorithm_ring_buffer) + - [Buddy allocation algorithm](@ref buddy_algorithm) + - \subpage defragmentation + - [Defragmenting CPU memory](@ref defragmentation_cpu) + - [Defragmenting GPU memory](@ref defragmentation_gpu) + - [Additional notes](@ref defragmentation_additional_notes) + - [Writing custom allocation algorithm](@ref defragmentation_custom_algorithm) + - \subpage lost_allocations + - \subpage statistics + - [Numeric statistics](@ref statistics_numeric_statistics) + - [JSON dump](@ref statistics_json_dump) + - \subpage allocation_annotation + - [Allocation user data](@ref allocation_user_data) + - [Allocation names](@ref allocation_names) + - \subpage debugging_memory_usage + - [Memory initialization](@ref debugging_memory_usage_initialization) + - [Margins](@ref debugging_memory_usage_margins) + - [Corruption detection](@ref debugging_memory_usage_corruption_detection) + - \subpage record_and_replay +- \subpage usage_patterns + - [Common mistakes](@ref usage_patterns_common_mistakes) + - [Simple patterns](@ref usage_patterns_simple) + - [Advanced patterns](@ref usage_patterns_advanced) +- \subpage configuration + - [Pointers to Vulkan functions](@ref config_Vulkan_functions) + - [Custom host memory allocator](@ref custom_memory_allocator) + - [Device memory allocation callbacks](@ref allocation_callbacks) + - [Device heap memory limit](@ref heap_memory_limit) + - \subpage vk_khr_dedicated_allocation + - \subpage enabling_buffer_device_address + - \subpage vk_amd_device_coherent_memory +- \subpage general_considerations + - [Thread safety](@ref general_considerations_thread_safety) + - [Validation layer warnings](@ref general_considerations_validation_layer_warnings) + - [Allocation algorithm](@ref general_considerations_allocation_algorithm) + - [Features not supported](@ref general_considerations_features_not_supported) + +\section main_see_also See also + +- [Product page on GPUOpen](https://gpuopen.com/gaming-product/vulkan-memory-allocator/) +- [Source repository on GitHub](https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator) + + + + +\page quick_start Quick start + +\section quick_start_project_setup Project setup + +Vulkan Memory Allocator comes in form of a "stb-style" single header file. +You don't need to build it as a separate library project. +You can add this file directly to your project and submit it to code repository next to your other source files. + +"Single header" doesn't mean that everything is contained in C/C++ declarations, +like it tends to be in case of inline functions or C++ templates. +It means that implementation is bundled with interface in a single file and needs to be extracted using preprocessor macro. +If you don't do it properly, you will get linker errors. + +To do it properly: + +-# Include "vk_mem_alloc.h" file in each CPP file where you want to use the library. + This includes declarations of all members of the library. +-# In exacly one CPP file define following macro before this include. + It enables also internal definitions. + +\code +#define VMA_IMPLEMENTATION +#include "vk_mem_alloc.h" +\endcode + +It may be a good idea to create dedicated CPP file just for this purpose. + +Note on language: This library is written in C++, but has C-compatible interface. +Thus you can include and use vk_mem_alloc.h in C or C++ code, but full +implementation with `VMA_IMPLEMENTATION` macro must be compiled as C++, NOT as C. + +Please note that this library includes header ``, which in turn +includes `` on Windows. If you need some specific macros defined +before including these headers (like `WIN32_LEAN_AND_MEAN` or +`WINVER` for Windows, `VK_USE_PLATFORM_WIN32_KHR` for Vulkan), you must define +them before every `#include` of this library. + +You may need to configure the way you import Vulkan functions. + +- By default, VMA assumes you you link statically with Vulkan API. If this is not the case, + `#define VMA_STATIC_VULKAN_FUNCTIONS 0` before `#include` of the VMA implementation and use another way. +- You can `#define VMA_DYNAMIC_VULKAN_FUNCTIONS 1` and make sure `vkGetInstanceProcAddr` and `vkGetDeviceProcAddr` globals are defined. + All the remaining Vulkan functions will be fetched automatically. +- Finally, you can provide your own pointers to all Vulkan functions needed by VMA using structure member + VmaAllocatorCreateInfo::pVulkanFunctions, if you fetched them in some custom way e.g. using some loader like [Volk](https://github.com/zeux/volk). + + +\section quick_start_initialization Initialization + +At program startup: + +-# Initialize Vulkan to have `VkPhysicalDevice`, `VkDevice` and `VkInstance` object. +-# Fill VmaAllocatorCreateInfo structure and create #VmaAllocator object by + calling vmaCreateAllocator(). + +\code +VmaAllocatorCreateInfo allocatorInfo = {}; +allocatorInfo.vulkanApiVersion = VK_API_VERSION_1_2; +allocatorInfo.physicalDevice = physicalDevice; +allocatorInfo.device = device; +allocatorInfo.instance = instance; + +VmaAllocator allocator; +vmaCreateAllocator(&allocatorInfo, &allocator); +\endcode + +Only members `physicalDevice`, `device`, `instance` are required. +However, you should inform the library which Vulkan version do you use by setting +VmaAllocatorCreateInfo::vulkanApiVersion and which extensions did you enable +by setting VmaAllocatorCreateInfo::flags (like #VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT for VK_KHR_buffer_device_address). +Otherwise, VMA would use only features of Vulkan 1.0 core with no extensions. + + +\section quick_start_resource_allocation Resource allocation + +When you want to create a buffer or image: + +-# Fill `VkBufferCreateInfo` / `VkImageCreateInfo` structure. +-# Fill VmaAllocationCreateInfo structure. +-# Call vmaCreateBuffer() / vmaCreateImage() to get `VkBuffer`/`VkImage` with memory + already allocated and bound to it. + +\code +VkBufferCreateInfo bufferInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; +bufferInfo.size = 65536; +bufferInfo.usage = VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; + +VmaAllocationCreateInfo allocInfo = {}; +allocInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; + +VkBuffer buffer; +VmaAllocation allocation; +vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &buffer, &allocation, nullptr); +\endcode + +Don't forget to destroy your objects when no longer needed: + +\code +vmaDestroyBuffer(allocator, buffer, allocation); +vmaDestroyAllocator(allocator); +\endcode + + +\page choosing_memory_type Choosing memory type + +Physical devices in Vulkan support various combinations of memory heaps and +types. Help with choosing correct and optimal memory type for your specific +resource is one of the key features of this library. You can use it by filling +appropriate members of VmaAllocationCreateInfo structure, as described below. +You can also combine multiple methods. + +-# If you just want to find memory type index that meets your requirements, you + can use function: vmaFindMemoryTypeIndex(), vmaFindMemoryTypeIndexForBufferInfo(), + vmaFindMemoryTypeIndexForImageInfo(). +-# If you want to allocate a region of device memory without association with any + specific image or buffer, you can use function vmaAllocateMemory(). Usage of + this function is not recommended and usually not needed. + vmaAllocateMemoryPages() function is also provided for creating multiple allocations at once, + which may be useful for sparse binding. +-# If you already have a buffer or an image created, you want to allocate memory + for it and then you will bind it yourself, you can use function + vmaAllocateMemoryForBuffer(), vmaAllocateMemoryForImage(). + For binding you should use functions: vmaBindBufferMemory(), vmaBindImageMemory() + or their extended versions: vmaBindBufferMemory2(), vmaBindImageMemory2(). +-# If you want to create a buffer or an image, allocate memory for it and bind + them together, all in one call, you can use function vmaCreateBuffer(), + vmaCreateImage(). This is the easiest and recommended way to use this library. + +When using 3. or 4., the library internally queries Vulkan for memory types +supported for that buffer or image (function `vkGetBufferMemoryRequirements()`) +and uses only one of these types. + +If no memory type can be found that meets all the requirements, these functions +return `VK_ERROR_FEATURE_NOT_PRESENT`. + +You can leave VmaAllocationCreateInfo structure completely filled with zeros. +It means no requirements are specified for memory type. +It is valid, although not very useful. + +\section choosing_memory_type_usage Usage + +The easiest way to specify memory requirements is to fill member +VmaAllocationCreateInfo::usage using one of the values of enum #VmaMemoryUsage. +It defines high level, common usage types. +For more details, see description of this enum. + +For example, if you want to create a uniform buffer that will be filled using +transfer only once or infrequently and used for rendering every frame, you can +do it using following code: + +\code +VkBufferCreateInfo bufferInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; +bufferInfo.size = 65536; +bufferInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; + +VmaAllocationCreateInfo allocInfo = {}; +allocInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; + +VkBuffer buffer; +VmaAllocation allocation; +vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &buffer, &allocation, nullptr); +\endcode + +\section choosing_memory_type_required_preferred_flags Required and preferred flags + +You can specify more detailed requirements by filling members +VmaAllocationCreateInfo::requiredFlags and VmaAllocationCreateInfo::preferredFlags +with a combination of bits from enum `VkMemoryPropertyFlags`. For example, +if you want to create a buffer that will be persistently mapped on host (so it +must be `HOST_VISIBLE`) and preferably will also be `HOST_COHERENT` and `HOST_CACHED`, +use following code: + +\code +VmaAllocationCreateInfo allocInfo = {}; +allocInfo.requiredFlags = VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; +allocInfo.preferredFlags = VK_MEMORY_PROPERTY_HOST_COHERENT_BIT | VK_MEMORY_PROPERTY_HOST_CACHED_BIT; +allocInfo.flags = VMA_ALLOCATION_CREATE_MAPPED_BIT; + +VkBuffer buffer; +VmaAllocation allocation; +vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &buffer, &allocation, nullptr); +\endcode + +A memory type is chosen that has all the required flags and as many preferred +flags set as possible. + +If you use VmaAllocationCreateInfo::usage, it is just internally converted to +a set of required and preferred flags. + +\section choosing_memory_type_explicit_memory_types Explicit memory types + +If you inspected memory types available on the physical device and you have +a preference for memory types that you want to use, you can fill member +VmaAllocationCreateInfo::memoryTypeBits. It is a bit mask, where each bit set +means that a memory type with that index is allowed to be used for the +allocation. Special value 0, just like `UINT32_MAX`, means there are no +restrictions to memory type index. + +Please note that this member is NOT just a memory type index. +Still you can use it to choose just one, specific memory type. +For example, if you already determined that your buffer should be created in +memory type 2, use following code: + +\code +uint32_t memoryTypeIndex = 2; + +VmaAllocationCreateInfo allocInfo = {}; +allocInfo.memoryTypeBits = 1u << memoryTypeIndex; + +VkBuffer buffer; +VmaAllocation allocation; +vmaCreateBuffer(allocator, &bufferInfo, &allocInfo, &buffer, &allocation, nullptr); +\endcode + + +\section choosing_memory_type_custom_memory_pools Custom memory pools + +If you allocate from custom memory pool, all the ways of specifying memory +requirements described above are not applicable and the aforementioned members +of VmaAllocationCreateInfo structure are ignored. Memory type is selected +explicitly when creating the pool and then used to make all the allocations from +that pool. For further details, see \ref custom_memory_pools. + +\section choosing_memory_type_dedicated_allocations Dedicated allocations + +Memory for allocations is reserved out of larger block of `VkDeviceMemory` +allocated from Vulkan internally. That's the main feature of this whole library. +You can still request a separate memory block to be created for an allocation, +just like you would do in a trivial solution without using any allocator. +In that case, a buffer or image is always bound to that memory at offset 0. +This is called a "dedicated allocation". +You can explicitly request it by using flag #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. +The library can also internally decide to use dedicated allocation in some cases, e.g.: + +- When the size of the allocation is large. +- When [VK_KHR_dedicated_allocation](@ref vk_khr_dedicated_allocation) extension is enabled + and it reports that dedicated allocation is required or recommended for the resource. +- When allocation of next big memory block fails due to not enough device memory, + but allocation with the exact requested size succeeds. + + +\page memory_mapping Memory mapping + +To "map memory" in Vulkan means to obtain a CPU pointer to `VkDeviceMemory`, +to be able to read from it or write to it in CPU code. +Mapping is possible only of memory allocated from a memory type that has +`VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT` flag. +Functions `vkMapMemory()`, `vkUnmapMemory()` are designed for this purpose. +You can use them directly with memory allocated by this library, +but it is not recommended because of following issue: +Mapping the same `VkDeviceMemory` block multiple times is illegal - only one mapping at a time is allowed. +This includes mapping disjoint regions. Mapping is not reference-counted internally by Vulkan. +Because of this, Vulkan Memory Allocator provides following facilities: + +\section memory_mapping_mapping_functions Mapping functions + +The library provides following functions for mapping of a specific #VmaAllocation: vmaMapMemory(), vmaUnmapMemory(). +They are safer and more convenient to use than standard Vulkan functions. +You can map an allocation multiple times simultaneously - mapping is reference-counted internally. +You can also map different allocations simultaneously regardless of whether they use the same `VkDeviceMemory` block. +The way it's implemented is that the library always maps entire memory block, not just region of the allocation. +For further details, see description of vmaMapMemory() function. +Example: + +\code +// Having these objects initialized: + +struct ConstantBuffer +{ + ... +}; +ConstantBuffer constantBufferData; + +VmaAllocator allocator; +VkBuffer constantBuffer; +VmaAllocation constantBufferAllocation; + +// You can map and fill your buffer using following code: + +void* mappedData; +vmaMapMemory(allocator, constantBufferAllocation, &mappedData); +memcpy(mappedData, &constantBufferData, sizeof(constantBufferData)); +vmaUnmapMemory(allocator, constantBufferAllocation); +\endcode + +When mapping, you may see a warning from Vulkan validation layer similar to this one: + +Mapping an image with layout VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL can result in undefined behavior if this memory is used by the device. Only GENERAL or PREINITIALIZED should be used. + +It happens because the library maps entire `VkDeviceMemory` block, where different +types of images and buffers may end up together, especially on GPUs with unified memory like Intel. +You can safely ignore it if you are sure you access only memory of the intended +object that you wanted to map. + + +\section memory_mapping_persistently_mapped_memory Persistently mapped memory + +Kepping your memory persistently mapped is generally OK in Vulkan. +You don't need to unmap it before using its data on the GPU. +The library provides a special feature designed for that: +Allocations made with #VMA_ALLOCATION_CREATE_MAPPED_BIT flag set in +VmaAllocationCreateInfo::flags stay mapped all the time, +so you can just access CPU pointer to it any time +without a need to call any "map" or "unmap" function. +Example: + +\code +VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; +bufCreateInfo.size = sizeof(ConstantBuffer); +bufCreateInfo.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT; + +VmaAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.usage = VMA_MEMORY_USAGE_CPU_ONLY; +allocCreateInfo.flags = VMA_ALLOCATION_CREATE_MAPPED_BIT; + +VkBuffer buf; +VmaAllocation alloc; +VmaAllocationInfo allocInfo; +vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &buf, &alloc, &allocInfo); + +// Buffer is already mapped. You can access its memory. +memcpy(allocInfo.pMappedData, &constantBufferData, sizeof(constantBufferData)); +\endcode + +There are some exceptions though, when you should consider mapping memory only for a short period of time: + +- When operating system is Windows 7 or 8.x (Windows 10 is not affected because it uses WDDM2), + device is discrete AMD GPU, + and memory type is the special 256 MiB pool of `DEVICE_LOCAL + HOST_VISIBLE` memory + (selected when you use #VMA_MEMORY_USAGE_CPU_TO_GPU), + then whenever a memory block allocated from this memory type stays mapped + for the time of any call to `vkQueueSubmit()` or `vkQueuePresentKHR()`, this + block is migrated by WDDM to system RAM, which degrades performance. It doesn't + matter if that particular memory block is actually used by the command buffer + being submitted. +- On Mac/MoltenVK there is a known bug - [Issue #175](https://github.com/KhronosGroup/MoltenVK/issues/175) + which requires unmapping before GPU can see updated texture. +- Keeping many large memory blocks mapped may impact performance or stability of some debugging tools. + +\section memory_mapping_cache_control Cache flush and invalidate + +Memory in Vulkan doesn't need to be unmapped before using it on GPU, +but unless a memory types has `VK_MEMORY_PROPERTY_HOST_COHERENT_BIT` flag set, +you need to manually **invalidate** cache before reading of mapped pointer +and **flush** cache after writing to mapped pointer. +Map/unmap operations don't do that automatically. +Vulkan provides following functions for this purpose `vkFlushMappedMemoryRanges()`, +`vkInvalidateMappedMemoryRanges()`, but this library provides more convenient +functions that refer to given allocation object: vmaFlushAllocation(), +vmaInvalidateAllocation(), +or multiple objects at once: vmaFlushAllocations(), vmaInvalidateAllocations(). + +Regions of memory specified for flush/invalidate must be aligned to +`VkPhysicalDeviceLimits::nonCoherentAtomSize`. This is automatically ensured by the library. +In any memory type that is `HOST_VISIBLE` but not `HOST_COHERENT`, all allocations +within blocks are aligned to this value, so their offsets are always multiply of +`nonCoherentAtomSize` and two different allocations never share same "line" of this size. + +Please note that memory allocated with #VMA_MEMORY_USAGE_CPU_ONLY is guaranteed to be `HOST_COHERENT`. + +Also, Windows drivers from all 3 **PC** GPU vendors (AMD, Intel, NVIDIA) +currently provide `HOST_COHERENT` flag on all memory types that are +`HOST_VISIBLE`, so on this platform you may not need to bother. + +\section memory_mapping_finding_if_memory_mappable Finding out if memory is mappable + +It may happen that your allocation ends up in memory that is `HOST_VISIBLE` (available for mapping) +despite it wasn't explicitly requested. +For example, application may work on integrated graphics with unified memory (like Intel) or +allocation from video memory might have failed, so the library chose system memory as fallback. + +You can detect this case and map such allocation to access its memory on CPU directly, +instead of launching a transfer operation. +In order to do that: inspect `allocInfo.memoryType`, call vmaGetMemoryTypeProperties(), +and look for `VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT` flag in properties of that memory type. + +\code +VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; +bufCreateInfo.size = sizeof(ConstantBuffer); +bufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; + +VmaAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; +allocCreateInfo.preferredFlags = VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; + +VkBuffer buf; +VmaAllocation alloc; +VmaAllocationInfo allocInfo; +vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &buf, &alloc, &allocInfo); + +VkMemoryPropertyFlags memFlags; +vmaGetMemoryTypeProperties(allocator, allocInfo.memoryType, &memFlags); +if((memFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) != 0) +{ + // Allocation ended up in mappable memory. You can map it and access it directly. + void* mappedData; + vmaMapMemory(allocator, alloc, &mappedData); + memcpy(mappedData, &constantBufferData, sizeof(constantBufferData)); + vmaUnmapMemory(allocator, alloc); +} +else +{ + // Allocation ended up in non-mappable memory. + // You need to create CPU-side buffer in VMA_MEMORY_USAGE_CPU_ONLY and make a transfer. +} +\endcode + +You can even use #VMA_ALLOCATION_CREATE_MAPPED_BIT flag while creating allocations +that are not necessarily `HOST_VISIBLE` (e.g. using #VMA_MEMORY_USAGE_GPU_ONLY). +If the allocation ends up in memory type that is `HOST_VISIBLE`, it will be persistently mapped and you can use it directly. +If not, the flag is just ignored. +Example: + +\code +VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; +bufCreateInfo.size = sizeof(ConstantBuffer); +bufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; + +VmaAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; +allocCreateInfo.flags = VMA_ALLOCATION_CREATE_MAPPED_BIT; + +VkBuffer buf; +VmaAllocation alloc; +VmaAllocationInfo allocInfo; +vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &buf, &alloc, &allocInfo); + +if(allocInfo.pMappedData != nullptr) +{ + // Allocation ended up in mappable memory. + // It's persistently mapped. You can access it directly. + memcpy(allocInfo.pMappedData, &constantBufferData, sizeof(constantBufferData)); +} +else +{ + // Allocation ended up in non-mappable memory. + // You need to create CPU-side buffer in VMA_MEMORY_USAGE_CPU_ONLY and make a transfer. +} +\endcode + + +\page staying_within_budget Staying within budget + +When developing a graphics-intensive game or program, it is important to avoid allocating +more GPU memory than it's physically available. When the memory is over-committed, +various bad things can happen, depending on the specific GPU, graphics driver, and +operating system: + +- It may just work without any problems. +- The application may slow down because some memory blocks are moved to system RAM + and the GPU has to access them through PCI Express bus. +- A new allocation may take very long time to complete, even few seconds, and possibly + freeze entire system. +- The new allocation may fail with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. +- It may even result in GPU crash (TDR), observed as `VK_ERROR_DEVICE_LOST` + returned somewhere later. + +\section staying_within_budget_querying_for_budget Querying for budget + +To query for current memory usage and available budget, use function vmaGetBudget(). +Returned structure #VmaBudget contains quantities expressed in bytes, per Vulkan memory heap. + +Please note that this function returns different information and works faster than +vmaCalculateStats(). vmaGetBudget() can be called every frame or even before every +allocation, while vmaCalculateStats() is intended to be used rarely, +only to obtain statistical information, e.g. for debugging purposes. + +It is recommended to use VK_EXT_memory_budget device extension to obtain information +about the budget from Vulkan device. VMA is able to use this extension automatically. +When not enabled, the allocator behaves same way, but then it estimates current usage +and available budget based on its internal information and Vulkan memory heap sizes, +which may be less precise. In order to use this extension: + +1. Make sure extensions VK_EXT_memory_budget and VK_KHR_get_physical_device_properties2 + required by it are available and enable them. Please note that the first is a device + extension and the second is instance extension! +2. Use flag #VMA_ALLOCATOR_CREATE_EXT_MEMORY_BUDGET_BIT when creating #VmaAllocator object. +3. Make sure to call vmaSetCurrentFrameIndex() every frame. Budget is queried from + Vulkan inside of it to avoid overhead of querying it with every allocation. + +\section staying_within_budget_controlling_memory_usage Controlling memory usage + +There are many ways in which you can try to stay within the budget. + +First, when making new allocation requires allocating a new memory block, the library +tries not to exceed the budget automatically. If a block with default recommended size +(e.g. 256 MB) would go over budget, a smaller block is allocated, possibly even +dedicated memory for just this resource. + +If the size of the requested resource plus current memory usage is more than the +budget, by default the library still tries to create it, leaving it to the Vulkan +implementation whether the allocation succeeds or fails. You can change this behavior +by using #VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT flag. With it, the allocation is +not made if it would exceed the budget or if the budget is already exceeded. +Some other allocations become lost instead to make room for it, if the mechanism of +[lost allocations](@ref lost_allocations) is used. +If that is not possible, the allocation fails with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. +Example usage pattern may be to pass the #VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT flag +when creating resources that are not essential for the application (e.g. the texture +of a specific object) and not to pass it when creating critically important resources +(e.g. render targets). + +Finally, you can also use #VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT flag to make sure +a new allocation is created only when it fits inside one of the existing memory blocks. +If it would require to allocate a new block, if fails instead with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. +This also ensures that the function call is very fast because it never goes to Vulkan +to obtain a new block. + +Please note that creating \ref custom_memory_pools with VmaPoolCreateInfo::minBlockCount +set to more than 0 will try to allocate memory blocks without checking whether they +fit within budget. + + +\page resource_aliasing Resource aliasing (overlap) + +New explicit graphics APIs (Vulkan and Direct3D 12), thanks to manual memory +management, give an opportunity to alias (overlap) multiple resources in the +same region of memory - a feature not available in the old APIs (Direct3D 11, OpenGL). +It can be useful to save video memory, but it must be used with caution. + +For example, if you know the flow of your whole render frame in advance, you +are going to use some intermediate textures or buffers only during a small range of render passes, +and you know these ranges don't overlap in time, you can bind these resources to +the same place in memory, even if they have completely different parameters (width, height, format etc.). + +![Resource aliasing (overlap)](../gfx/Aliasing.png) + +Such scenario is possible using VMA, but you need to create your images manually. +Then you need to calculate parameters of an allocation to be made using formula: + +- allocation size = max(size of each image) +- allocation alignment = max(alignment of each image) +- allocation memoryTypeBits = bitwise AND(memoryTypeBits of each image) + +Following example shows two different images bound to the same place in memory, +allocated to fit largest of them. + +\code +// A 512x512 texture to be sampled. +VkImageCreateInfo img1CreateInfo = { VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO }; +img1CreateInfo.imageType = VK_IMAGE_TYPE_2D; +img1CreateInfo.extent.width = 512; +img1CreateInfo.extent.height = 512; +img1CreateInfo.extent.depth = 1; +img1CreateInfo.mipLevels = 10; +img1CreateInfo.arrayLayers = 1; +img1CreateInfo.format = VK_FORMAT_R8G8B8A8_SRGB; +img1CreateInfo.tiling = VK_IMAGE_TILING_OPTIMAL; +img1CreateInfo.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; +img1CreateInfo.usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT; +img1CreateInfo.samples = VK_SAMPLE_COUNT_1_BIT; + +// A full screen texture to be used as color attachment. +VkImageCreateInfo img2CreateInfo = { VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO }; +img2CreateInfo.imageType = VK_IMAGE_TYPE_2D; +img2CreateInfo.extent.width = 1920; +img2CreateInfo.extent.height = 1080; +img2CreateInfo.extent.depth = 1; +img2CreateInfo.mipLevels = 1; +img2CreateInfo.arrayLayers = 1; +img2CreateInfo.format = VK_FORMAT_R8G8B8A8_UNORM; +img2CreateInfo.tiling = VK_IMAGE_TILING_OPTIMAL; +img2CreateInfo.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; +img2CreateInfo.usage = VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; +img2CreateInfo.samples = VK_SAMPLE_COUNT_1_BIT; + +VkImage img1; +res = vkCreateImage(device, &img1CreateInfo, nullptr, &img1); +VkImage img2; +res = vkCreateImage(device, &img2CreateInfo, nullptr, &img2); + +VkMemoryRequirements img1MemReq; +vkGetImageMemoryRequirements(device, img1, &img1MemReq); +VkMemoryRequirements img2MemReq; +vkGetImageMemoryRequirements(device, img2, &img2MemReq); + +VkMemoryRequirements finalMemReq = {}; +finalMemReq.size = std::max(img1MemReq.size, img2MemReq.size); +finalMemReq.alignment = std::max(img1MemReq.alignment, img2MemReq.alignment); +finalMemReq.memoryTypeBits = img1MemReq.memoryTypeBits & img2MemReq.memoryTypeBits; +// Validate if(finalMemReq.memoryTypeBits != 0) + +VmaAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; + +VmaAllocation alloc; +res = vmaAllocateMemory(allocator, &finalMemReq, &allocCreateInfo, &alloc, nullptr); + +res = vmaBindImageMemory(allocator, alloc, img1); +res = vmaBindImageMemory(allocator, alloc, img2); + +// You can use img1, img2 here, but not at the same time! + +vmaFreeMemory(allocator, alloc); +vkDestroyImage(allocator, img2, nullptr); +vkDestroyImage(allocator, img1, nullptr); +\endcode + +Remember that using resouces that alias in memory requires proper synchronization. +You need to issue a memory barrier to make sure commands that use `img1` and `img2` +don't overlap on GPU timeline. +You also need to treat a resource after aliasing as uninitialized - containing garbage data. +For example, if you use `img1` and then want to use `img2`, you need to issue +an image memory barrier for `img2` with `oldLayout` = `VK_IMAGE_LAYOUT_UNDEFINED`. + +Additional considerations: + +- Vulkan also allows to interpret contents of memory between aliasing resources consistently in some cases. +See chapter 11.8. "Memory Aliasing" of Vulkan specification or `VK_IMAGE_CREATE_ALIAS_BIT` flag. +- You can create more complex layout where different images and buffers are bound +at different offsets inside one large allocation. For example, one can imagine +a big texture used in some render passes, aliasing with a set of many small buffers +used between in some further passes. To bind a resource at non-zero offset of an allocation, +use vmaBindBufferMemory2() / vmaBindImageMemory2(). +- Before allocating memory for the resources you want to alias, check `memoryTypeBits` +returned in memory requirements of each resource to make sure the bits overlap. +Some GPUs may expose multiple memory types suitable e.g. only for buffers or +images with `COLOR_ATTACHMENT` usage, so the sets of memory types supported by your +resources may be disjoint. Aliasing them is not possible in that case. + + +\page custom_memory_pools Custom memory pools + +A memory pool contains a number of `VkDeviceMemory` blocks. +The library automatically creates and manages default pool for each memory type available on the device. +Default memory pool automatically grows in size. +Size of allocated blocks is also variable and managed automatically. + +You can create custom pool and allocate memory out of it. +It can be useful if you want to: + +- Keep certain kind of allocations separate from others. +- Enforce particular, fixed size of Vulkan memory blocks. +- Limit maximum amount of Vulkan memory allocated for that pool. +- Reserve minimum or fixed amount of Vulkan memory always preallocated for that pool. + +To use custom memory pools: + +-# Fill VmaPoolCreateInfo structure. +-# Call vmaCreatePool() to obtain #VmaPool handle. +-# When making an allocation, set VmaAllocationCreateInfo::pool to this handle. + You don't need to specify any other parameters of this structure, like `usage`. + +Example: + +\code +// Create a pool that can have at most 2 blocks, 128 MiB each. +VmaPoolCreateInfo poolCreateInfo = {}; +poolCreateInfo.memoryTypeIndex = ... +poolCreateInfo.blockSize = 128ull * 1024 * 1024; +poolCreateInfo.maxBlockCount = 2; + +VmaPool pool; +vmaCreatePool(allocator, &poolCreateInfo, &pool); + +// Allocate a buffer out of it. +VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; +bufCreateInfo.size = 1024; +bufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; + +VmaAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.pool = pool; + +VkBuffer buf; +VmaAllocation alloc; +VmaAllocationInfo allocInfo; +vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &buf, &alloc, &allocInfo); +\endcode + +You have to free all allocations made from this pool before destroying it. + +\code +vmaDestroyBuffer(allocator, buf, alloc); +vmaDestroyPool(allocator, pool); +\endcode + +\section custom_memory_pools_MemTypeIndex Choosing memory type index + +When creating a pool, you must explicitly specify memory type index. +To find the one suitable for your buffers or images, you can use helper functions +vmaFindMemoryTypeIndexForBufferInfo(), vmaFindMemoryTypeIndexForImageInfo(). +You need to provide structures with example parameters of buffers or images +that you are going to create in that pool. + +\code +VkBufferCreateInfo exampleBufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; +exampleBufCreateInfo.size = 1024; // Whatever. +exampleBufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; // Change if needed. + +VmaAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; // Change if needed. + +uint32_t memTypeIndex; +vmaFindMemoryTypeIndexForBufferInfo(allocator, &exampleBufCreateInfo, &allocCreateInfo, &memTypeIndex); + +VmaPoolCreateInfo poolCreateInfo = {}; +poolCreateInfo.memoryTypeIndex = memTypeIndex; +// ... +\endcode + +When creating buffers/images allocated in that pool, provide following parameters: + +- `VkBufferCreateInfo`: Prefer to pass same parameters as above. + Otherwise you risk creating resources in a memory type that is not suitable for them, which may result in undefined behavior. + Using different `VK_BUFFER_USAGE_` flags may work, but you shouldn't create images in a pool intended for buffers + or the other way around. +- VmaAllocationCreateInfo: You don't need to pass same parameters. Fill only `pool` member. + Other members are ignored anyway. + +\section linear_algorithm Linear allocation algorithm + +Each Vulkan memory block managed by this library has accompanying metadata that +keeps track of used and unused regions. By default, the metadata structure and +algorithm tries to find best place for new allocations among free regions to +optimize memory usage. This way you can allocate and free objects in any order. + +![Default allocation algorithm](../gfx/Linear_allocator_1_algo_default.png) + +Sometimes there is a need to use simpler, linear allocation algorithm. You can +create custom pool that uses such algorithm by adding flag +#VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT to VmaPoolCreateInfo::flags while creating +#VmaPool object. Then an alternative metadata management is used. It always +creates new allocations after last one and doesn't reuse free regions after +allocations freed in the middle. It results in better allocation performance and +less memory consumed by metadata. + +![Linear allocation algorithm](../gfx/Linear_allocator_2_algo_linear.png) + +With this one flag, you can create a custom pool that can be used in many ways: +free-at-once, stack, double stack, and ring buffer. See below for details. + +\subsection linear_algorithm_free_at_once Free-at-once + +In a pool that uses linear algorithm, you still need to free all the allocations +individually, e.g. by using vmaFreeMemory() or vmaDestroyBuffer(). You can free +them in any order. New allocations are always made after last one - free space +in the middle is not reused. However, when you release all the allocation and +the pool becomes empty, allocation starts from the beginning again. This way you +can use linear algorithm to speed up creation of allocations that you are going +to release all at once. + +![Free-at-once](../gfx/Linear_allocator_3_free_at_once.png) + +This mode is also available for pools created with VmaPoolCreateInfo::maxBlockCount +value that allows multiple memory blocks. + +\subsection linear_algorithm_stack Stack + +When you free an allocation that was created last, its space can be reused. +Thanks to this, if you always release allocations in the order opposite to their +creation (LIFO - Last In First Out), you can achieve behavior of a stack. + +![Stack](../gfx/Linear_allocator_4_stack.png) + +This mode is also available for pools created with VmaPoolCreateInfo::maxBlockCount +value that allows multiple memory blocks. + +\subsection linear_algorithm_double_stack Double stack + +The space reserved by a custom pool with linear algorithm may be used by two +stacks: + +- First, default one, growing up from offset 0. +- Second, "upper" one, growing down from the end towards lower offsets. + +To make allocation from upper stack, add flag #VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT +to VmaAllocationCreateInfo::flags. + +![Double stack](../gfx/Linear_allocator_7_double_stack.png) + +Double stack is available only in pools with one memory block - +VmaPoolCreateInfo::maxBlockCount must be 1. Otherwise behavior is undefined. + +When the two stacks' ends meet so there is not enough space between them for a +new allocation, such allocation fails with usual +`VK_ERROR_OUT_OF_DEVICE_MEMORY` error. + +\subsection linear_algorithm_ring_buffer Ring buffer + +When you free some allocations from the beginning and there is not enough free space +for a new one at the end of a pool, allocator's "cursor" wraps around to the +beginning and starts allocation there. Thanks to this, if you always release +allocations in the same order as you created them (FIFO - First In First Out), +you can achieve behavior of a ring buffer / queue. + +![Ring buffer](../gfx/Linear_allocator_5_ring_buffer.png) + +Pools with linear algorithm support [lost allocations](@ref lost_allocations) when used as ring buffer. +If there is not enough free space for a new allocation, but existing allocations +from the front of the queue can become lost, they become lost and the allocation +succeeds. + +![Ring buffer with lost allocations](../gfx/Linear_allocator_6_ring_buffer_lost.png) + +Ring buffer is available only in pools with one memory block - +VmaPoolCreateInfo::maxBlockCount must be 1. Otherwise behavior is undefined. + +\section buddy_algorithm Buddy allocation algorithm + +There is another allocation algorithm that can be used with custom pools, called +"buddy". Its internal data structure is based on a tree of blocks, each having +size that is a power of two and a half of its parent's size. When you want to +allocate memory of certain size, a free node in the tree is located. If it's too +large, it is recursively split into two halves (called "buddies"). However, if +requested allocation size is not a power of two, the size of a tree node is +aligned up to the nearest power of two and the remaining space is wasted. When +two buddy nodes become free, they are merged back into one larger node. + +![Buddy allocator](../gfx/Buddy_allocator.png) + +The advantage of buddy allocation algorithm over default algorithm is faster +allocation and deallocation, as well as smaller external fragmentation. The +disadvantage is more wasted space (internal fragmentation). + +For more information, please read ["Buddy memory allocation" on Wikipedia](https://en.wikipedia.org/wiki/Buddy_memory_allocation) +or other sources that describe this concept in general. + +To use buddy allocation algorithm with a custom pool, add flag +#VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT to VmaPoolCreateInfo::flags while creating +#VmaPool object. + +Several limitations apply to pools that use buddy algorithm: + +- It is recommended to use VmaPoolCreateInfo::blockSize that is a power of two. + Otherwise, only largest power of two smaller than the size is used for + allocations. The remaining space always stays unused. +- [Margins](@ref debugging_memory_usage_margins) and + [corruption detection](@ref debugging_memory_usage_corruption_detection) + don't work in such pools. +- [Lost allocations](@ref lost_allocations) don't work in such pools. You can + use them, but they never become lost. Support may be added in the future. +- [Defragmentation](@ref defragmentation) doesn't work with allocations made from + such pool. + +\page defragmentation Defragmentation + +Interleaved allocations and deallocations of many objects of varying size can +cause fragmentation over time, which can lead to a situation where the library is unable +to find a continuous range of free memory for a new allocation despite there is +enough free space, just scattered across many small free ranges between existing +allocations. + +To mitigate this problem, you can use defragmentation feature: +structure #VmaDefragmentationInfo2, function vmaDefragmentationBegin(), vmaDefragmentationEnd(). +Given set of allocations, +this function can move them to compact used memory, ensure more continuous free +space and possibly also free some `VkDeviceMemory` blocks. + +What the defragmentation does is: + +- Updates #VmaAllocation objects to point to new `VkDeviceMemory` and offset. + After allocation has been moved, its VmaAllocationInfo::deviceMemory and/or + VmaAllocationInfo::offset changes. You must query them again using + vmaGetAllocationInfo() if you need them. +- Moves actual data in memory. + +What it doesn't do, so you need to do it yourself: + +- Recreate buffers and images that were bound to allocations that were defragmented and + bind them with their new places in memory. + You must use `vkDestroyBuffer()`, `vkDestroyImage()`, + `vkCreateBuffer()`, `vkCreateImage()`, vmaBindBufferMemory(), vmaBindImageMemory() + for that purpose and NOT vmaDestroyBuffer(), + vmaDestroyImage(), vmaCreateBuffer(), vmaCreateImage(), because you don't need to + destroy or create allocation objects! +- Recreate views and update descriptors that point to these buffers and images. + +\section defragmentation_cpu Defragmenting CPU memory + +Following example demonstrates how you can run defragmentation on CPU. +Only allocations created in memory types that are `HOST_VISIBLE` can be defragmented. +Others are ignored. + +The way it works is: + +- It temporarily maps entire memory blocks when necessary. +- It moves data using `memmove()` function. + +\code +// Given following variables already initialized: +VkDevice device; +VmaAllocator allocator; +std::vector buffers; +std::vector allocations; + + +const uint32_t allocCount = (uint32_t)allocations.size(); +std::vector allocationsChanged(allocCount); + +VmaDefragmentationInfo2 defragInfo = {}; +defragInfo.allocationCount = allocCount; +defragInfo.pAllocations = allocations.data(); +defragInfo.pAllocationsChanged = allocationsChanged.data(); +defragInfo.maxCpuBytesToMove = VK_WHOLE_SIZE; // No limit. +defragInfo.maxCpuAllocationsToMove = UINT32_MAX; // No limit. + +VmaDefragmentationContext defragCtx; +vmaDefragmentationBegin(allocator, &defragInfo, nullptr, &defragCtx); +vmaDefragmentationEnd(allocator, defragCtx); + +for(uint32_t i = 0; i < allocCount; ++i) +{ + if(allocationsChanged[i]) + { + // Destroy buffer that is immutably bound to memory region which is no longer valid. + vkDestroyBuffer(device, buffers[i], nullptr); + + // Create new buffer with same parameters. + VkBufferCreateInfo bufferInfo = ...; + vkCreateBuffer(device, &bufferInfo, nullptr, &buffers[i]); + + // You can make dummy call to vkGetBufferMemoryRequirements here to silence validation layer warning. + + // Bind new buffer to new memory region. Data contained in it is already moved. + VmaAllocationInfo allocInfo; + vmaGetAllocationInfo(allocator, allocations[i], &allocInfo); + vmaBindBufferMemory(allocator, allocations[i], buffers[i]); + } +} +\endcode + +Setting VmaDefragmentationInfo2::pAllocationsChanged is optional. +This output array tells whether particular allocation in VmaDefragmentationInfo2::pAllocations at the same index +has been modified during defragmentation. +You can pass null, but you then need to query every allocation passed to defragmentation +for new parameters using vmaGetAllocationInfo() if you might need to recreate and rebind a buffer or image associated with it. + +If you use [Custom memory pools](@ref choosing_memory_type_custom_memory_pools), +you can fill VmaDefragmentationInfo2::poolCount and VmaDefragmentationInfo2::pPools +instead of VmaDefragmentationInfo2::allocationCount and VmaDefragmentationInfo2::pAllocations +to defragment all allocations in given pools. +You cannot use VmaDefragmentationInfo2::pAllocationsChanged in that case. +You can also combine both methods. + +\section defragmentation_gpu Defragmenting GPU memory + +It is also possible to defragment allocations created in memory types that are not `HOST_VISIBLE`. +To do that, you need to pass a command buffer that meets requirements as described in +VmaDefragmentationInfo2::commandBuffer. The way it works is: + +- It creates temporary buffers and binds them to entire memory blocks when necessary. +- It issues `vkCmdCopyBuffer()` to passed command buffer. + +Example: + +\code +// Given following variables already initialized: +VkDevice device; +VmaAllocator allocator; +VkCommandBuffer commandBuffer; +std::vector buffers; +std::vector allocations; + + +const uint32_t allocCount = (uint32_t)allocations.size(); +std::vector allocationsChanged(allocCount); + +VkCommandBufferBeginInfo cmdBufBeginInfo = ...; +vkBeginCommandBuffer(commandBuffer, &cmdBufBeginInfo); + +VmaDefragmentationInfo2 defragInfo = {}; +defragInfo.allocationCount = allocCount; +defragInfo.pAllocations = allocations.data(); +defragInfo.pAllocationsChanged = allocationsChanged.data(); +defragInfo.maxGpuBytesToMove = VK_WHOLE_SIZE; // Notice it's "GPU" this time. +defragInfo.maxGpuAllocationsToMove = UINT32_MAX; // Notice it's "GPU" this time. +defragInfo.commandBuffer = commandBuffer; + +VmaDefragmentationContext defragCtx; +vmaDefragmentationBegin(allocator, &defragInfo, nullptr, &defragCtx); + +vkEndCommandBuffer(commandBuffer); + +// Submit commandBuffer. +// Wait for a fence that ensures commandBuffer execution finished. + +vmaDefragmentationEnd(allocator, defragCtx); + +for(uint32_t i = 0; i < allocCount; ++i) +{ + if(allocationsChanged[i]) + { + // Destroy buffer that is immutably bound to memory region which is no longer valid. + vkDestroyBuffer(device, buffers[i], nullptr); + + // Create new buffer with same parameters. + VkBufferCreateInfo bufferInfo = ...; + vkCreateBuffer(device, &bufferInfo, nullptr, &buffers[i]); + + // You can make dummy call to vkGetBufferMemoryRequirements here to silence validation layer warning. + + // Bind new buffer to new memory region. Data contained in it is already moved. + VmaAllocationInfo allocInfo; + vmaGetAllocationInfo(allocator, allocations[i], &allocInfo); + vmaBindBufferMemory(allocator, allocations[i], buffers[i]); + } +} +\endcode + +You can combine these two methods by specifying non-zero `maxGpu*` as well as `maxCpu*` parameters. +The library automatically chooses best method to defragment each memory pool. + +You may try not to block your entire program to wait until defragmentation finishes, +but do it in the background, as long as you carefully fullfill requirements described +in function vmaDefragmentationBegin(). + +\section defragmentation_additional_notes Additional notes + +It is only legal to defragment allocations bound to: + +- buffers +- images created with `VK_IMAGE_CREATE_ALIAS_BIT`, `VK_IMAGE_TILING_LINEAR`, and + being currently in `VK_IMAGE_LAYOUT_GENERAL` or `VK_IMAGE_LAYOUT_PREINITIALIZED`. + +Defragmentation of images created with `VK_IMAGE_TILING_OPTIMAL` or in any other +layout may give undefined results. + +If you defragment allocations bound to images, new images to be bound to new +memory region after defragmentation should be created with `VK_IMAGE_LAYOUT_PREINITIALIZED` +and then transitioned to their original layout from before defragmentation if +needed using an image memory barrier. + +While using defragmentation, you may experience validation layer warnings, which you just need to ignore. +See [Validation layer warnings](@ref general_considerations_validation_layer_warnings). + +Please don't expect memory to be fully compacted after defragmentation. +Algorithms inside are based on some heuristics that try to maximize number of Vulkan +memory blocks to make totally empty to release them, as well as to maximimze continuous +empty space inside remaining blocks, while minimizing the number and size of allocations that +need to be moved. Some fragmentation may still remain - this is normal. + +\section defragmentation_custom_algorithm Writing custom defragmentation algorithm + +If you want to implement your own, custom defragmentation algorithm, +there is infrastructure prepared for that, +but it is not exposed through the library API - you need to hack its source code. +Here are steps needed to do this: + +-# Main thing you need to do is to define your own class derived from base abstract + class `VmaDefragmentationAlgorithm` and implement your version of its pure virtual methods. + See definition and comments of this class for details. +-# Your code needs to interact with device memory block metadata. + If you need more access to its data than it's provided by its public interface, + declare your new class as a friend class e.g. in class `VmaBlockMetadata_Generic`. +-# If you want to create a flag that would enable your algorithm or pass some additional + flags to configure it, add them to `VmaDefragmentationFlagBits` and use them in + VmaDefragmentationInfo2::flags. +-# Modify function `VmaBlockVectorDefragmentationContext::Begin` to create object + of your new class whenever needed. + + +\page lost_allocations Lost allocations + +If your game oversubscribes video memory, if may work OK in previous-generation +graphics APIs (DirectX 9, 10, 11, OpenGL) because resources are automatically +paged to system RAM. In Vulkan you can't do it because when you run out of +memory, an allocation just fails. If you have more data (e.g. textures) that can +fit into VRAM and you don't need it all at once, you may want to upload them to +GPU on demand and "push out" ones that are not used for a long time to make room +for the new ones, effectively using VRAM (or a cartain memory pool) as a form of +cache. Vulkan Memory Allocator can help you with that by supporting a concept of +"lost allocations". + +To create an allocation that can become lost, include #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT +flag in VmaAllocationCreateInfo::flags. Before using a buffer or image bound to +such allocation in every new frame, you need to query it if it's not lost. +To check it, call vmaTouchAllocation(). +If the allocation is lost, you should not use it or buffer/image bound to it. +You mustn't forget to destroy this allocation and this buffer/image. +vmaGetAllocationInfo() can also be used for checking status of the allocation. +Allocation is lost when returned VmaAllocationInfo::deviceMemory == `VK_NULL_HANDLE`. + +To create an allocation that can make some other allocations lost to make room +for it, use #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT flag. You will +usually use both flags #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT and +#VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT at the same time. + +Warning! Current implementation uses quite naive, brute force algorithm, +which can make allocation calls that use #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT +flag quite slow. A new, more optimal algorithm and data structure to speed this +up is planned for the future. + +Q: When interleaving creation of new allocations with usage of existing ones, +how do you make sure that an allocation won't become lost while it's used in the +current frame? + +It is ensured because vmaTouchAllocation() / vmaGetAllocationInfo() not only returns allocation +status/parameters and checks whether it's not lost, but when it's not, it also +atomically marks it as used in the current frame, which makes it impossible to +become lost in that frame. It uses lockless algorithm, so it works fast and +doesn't involve locking any internal mutex. + +Q: What if my allocation may still be in use by the GPU when it's rendering a +previous frame while I already submit new frame on the CPU? + +You can make sure that allocations "touched" by vmaTouchAllocation() / vmaGetAllocationInfo() will not +become lost for a number of additional frames back from the current one by +specifying this number as VmaAllocatorCreateInfo::frameInUseCount (for default +memory pool) and VmaPoolCreateInfo::frameInUseCount (for custom pool). + +Q: How do you inform the library when new frame starts? + +You need to call function vmaSetCurrentFrameIndex(). + +Example code: + +\code +struct MyBuffer +{ + VkBuffer m_Buf = nullptr; + VmaAllocation m_Alloc = nullptr; + + // Called when the buffer is really needed in the current frame. + void EnsureBuffer(); +}; + +void MyBuffer::EnsureBuffer() +{ + // Buffer has been created. + if(m_Buf != VK_NULL_HANDLE) + { + // Check if its allocation is not lost + mark it as used in current frame. + if(vmaTouchAllocation(allocator, m_Alloc)) + { + // It's all OK - safe to use m_Buf. + return; + } + } + + // Buffer not yet exists or lost - destroy and recreate it. + + vmaDestroyBuffer(allocator, m_Buf, m_Alloc); + + VkBufferCreateInfo bufCreateInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; + bufCreateInfo.size = 1024; + bufCreateInfo.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; + + VmaAllocationCreateInfo allocCreateInfo = {}; + allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; + allocCreateInfo.flags = VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT | + VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT; + + vmaCreateBuffer(allocator, &bufCreateInfo, &allocCreateInfo, &m_Buf, &m_Alloc, nullptr); +} +\endcode + +When using lost allocations, you may see some Vulkan validation layer warnings +about overlapping regions of memory bound to different kinds of buffers and +images. This is still valid as long as you implement proper handling of lost +allocations (like in the example above) and don't use them. + +You can create an allocation that is already in lost state from the beginning using function +vmaCreateLostAllocation(). It may be useful if you need a "dummy" allocation that is not null. + +You can call function vmaMakePoolAllocationsLost() to set all eligible allocations +in a specified custom pool to lost state. +Allocations that have been "touched" in current frame or VmaPoolCreateInfo::frameInUseCount frames back +cannot become lost. + +Q: Can I touch allocation that cannot become lost? + +Yes, although it has no visible effect. +Calls to vmaGetAllocationInfo() and vmaTouchAllocation() update last use frame index +also for allocations that cannot become lost, but the only way to observe it is to dump +internal allocator state using vmaBuildStatsString(). +You can use this feature for debugging purposes to explicitly mark allocations that you use +in current frame and then analyze JSON dump to see for how long each allocation stays unused. + + +\page statistics Statistics + +This library contains functions that return information about its internal state, +especially the amount of memory allocated from Vulkan. +Please keep in mind that these functions need to traverse all internal data structures +to gather these information, so they may be quite time-consuming. +Don't call them too often. + +\section statistics_numeric_statistics Numeric statistics + +You can query for overall statistics of the allocator using function vmaCalculateStats(). +Information are returned using structure #VmaStats. +It contains #VmaStatInfo - number of allocated blocks, number of allocations +(occupied ranges in these blocks), number of unused (free) ranges in these blocks, +number of bytes used and unused (but still allocated from Vulkan) and other information. +They are summed across memory heaps, memory types and total for whole allocator. + +You can query for statistics of a custom pool using function vmaGetPoolStats(). +Information are returned using structure #VmaPoolStats. + +You can query for information about specific allocation using function vmaGetAllocationInfo(). +It fill structure #VmaAllocationInfo. + +\section statistics_json_dump JSON dump + +You can dump internal state of the allocator to a string in JSON format using function vmaBuildStatsString(). +The result is guaranteed to be correct JSON. +It uses ANSI encoding. +Any strings provided by user (see [Allocation names](@ref allocation_names)) +are copied as-is and properly escaped for JSON, so if they use UTF-8, ISO-8859-2 or any other encoding, +this JSON string can be treated as using this encoding. +It must be freed using function vmaFreeStatsString(). + +The format of this JSON string is not part of official documentation of the library, +but it will not change in backward-incompatible way without increasing library major version number +and appropriate mention in changelog. + +The JSON string contains all the data that can be obtained using vmaCalculateStats(). +It can also contain detailed map of allocated memory blocks and their regions - +free and occupied by allocations. +This allows e.g. to visualize the memory or assess fragmentation. + + +\page allocation_annotation Allocation names and user data + +\section allocation_user_data Allocation user data + +You can annotate allocations with your own information, e.g. for debugging purposes. +To do that, fill VmaAllocationCreateInfo::pUserData field when creating +an allocation. It's an opaque `void*` pointer. You can use it e.g. as a pointer, +some handle, index, key, ordinal number or any other value that would associate +the allocation with your custom metadata. + +\code +VkBufferCreateInfo bufferInfo = { VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO }; +// Fill bufferInfo... + +MyBufferMetadata* pMetadata = CreateBufferMetadata(); + +VmaAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; +allocCreateInfo.pUserData = pMetadata; + +VkBuffer buffer; +VmaAllocation allocation; +vmaCreateBuffer(allocator, &bufferInfo, &allocCreateInfo, &buffer, &allocation, nullptr); +\endcode + +The pointer may be later retrieved as VmaAllocationInfo::pUserData: + +\code +VmaAllocationInfo allocInfo; +vmaGetAllocationInfo(allocator, allocation, &allocInfo); +MyBufferMetadata* pMetadata = (MyBufferMetadata*)allocInfo.pUserData; +\endcode + +It can also be changed using function vmaSetAllocationUserData(). + +Values of (non-zero) allocations' `pUserData` are printed in JSON report created by +vmaBuildStatsString(), in hexadecimal form. + +\section allocation_names Allocation names + +There is alternative mode available where `pUserData` pointer is used to point to +a null-terminated string, giving a name to the allocation. To use this mode, +set #VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT flag in VmaAllocationCreateInfo::flags. +Then `pUserData` passed as VmaAllocationCreateInfo::pUserData or argument to +vmaSetAllocationUserData() must be either null or pointer to a null-terminated string. +The library creates internal copy of the string, so the pointer you pass doesn't need +to be valid for whole lifetime of the allocation. You can free it after the call. + +\code +VkImageCreateInfo imageInfo = { VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO }; +// Fill imageInfo... + +std::string imageName = "Texture: "; +imageName += fileName; + +VmaAllocationCreateInfo allocCreateInfo = {}; +allocCreateInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; +allocCreateInfo.flags = VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT; +allocCreateInfo.pUserData = imageName.c_str(); + +VkImage image; +VmaAllocation allocation; +vmaCreateImage(allocator, &imageInfo, &allocCreateInfo, &image, &allocation, nullptr); +\endcode + +The value of `pUserData` pointer of the allocation will be different than the one +you passed when setting allocation's name - pointing to a buffer managed +internally that holds copy of the string. + +\code +VmaAllocationInfo allocInfo; +vmaGetAllocationInfo(allocator, allocation, &allocInfo); +const char* imageName = (const char*)allocInfo.pUserData; +printf("Image name: %s\n", imageName); +\endcode + +That string is also printed in JSON report created by vmaBuildStatsString(). + +\note Passing string name to VMA allocation doesn't automatically set it to the Vulkan buffer or image created with it. +You must do it manually using an extension like VK_EXT_debug_utils, which is independent of this library. + + +\page debugging_memory_usage Debugging incorrect memory usage + +If you suspect a bug with memory usage, like usage of uninitialized memory or +memory being overwritten out of bounds of an allocation, +you can use debug features of this library to verify this. + +\section debugging_memory_usage_initialization Memory initialization + +If you experience a bug with incorrect and nondeterministic data in your program and you suspect uninitialized memory to be used, +you can enable automatic memory initialization to verify this. +To do it, define macro `VMA_DEBUG_INITIALIZE_ALLOCATIONS` to 1. + +\code +#define VMA_DEBUG_INITIALIZE_ALLOCATIONS 1 +#include "vk_mem_alloc.h" +\endcode + +It makes memory of all new allocations initialized to bit pattern `0xDCDCDCDC`. +Before an allocation is destroyed, its memory is filled with bit pattern `0xEFEFEFEF`. +Memory is automatically mapped and unmapped if necessary. + +If you find these values while debugging your program, good chances are that you incorrectly +read Vulkan memory that is allocated but not initialized, or already freed, respectively. + +Memory initialization works only with memory types that are `HOST_VISIBLE`. +It works also with dedicated allocations. +It doesn't work with allocations created with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag, +as they cannot be mapped. + +\section debugging_memory_usage_margins Margins + +By default, allocations are laid out in memory blocks next to each other if possible +(considering required alignment, `bufferImageGranularity`, and `nonCoherentAtomSize`). + +![Allocations without margin](../gfx/Margins_1.png) + +Define macro `VMA_DEBUG_MARGIN` to some non-zero value (e.g. 16) to enforce specified +number of bytes as a margin before and after every allocation. + +\code +#define VMA_DEBUG_MARGIN 16 +#include "vk_mem_alloc.h" +\endcode + +![Allocations with margin](../gfx/Margins_2.png) + +If your bug goes away after enabling margins, it means it may be caused by memory +being overwritten outside of allocation boundaries. It is not 100% certain though. +Change in application behavior may also be caused by different order and distribution +of allocations across memory blocks after margins are applied. + +The margin is applied also before first and after last allocation in a block. +It may occur only once between two adjacent allocations. + +Margins work with all types of memory. + +Margin is applied only to allocations made out of memory blocks and not to dedicated +allocations, which have their own memory block of specific size. +It is thus not applied to allocations made using #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT flag +or those automatically decided to put into dedicated allocations, e.g. due to its +large size or recommended by VK_KHR_dedicated_allocation extension. +Margins are also not active in custom pools created with #VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT flag. + +Margins appear in [JSON dump](@ref statistics_json_dump) as part of free space. + +Note that enabling margins increases memory usage and fragmentation. + +\section debugging_memory_usage_corruption_detection Corruption detection + +You can additionally define macro `VMA_DEBUG_DETECT_CORRUPTION` to 1 to enable validation +of contents of the margins. + +\code +#define VMA_DEBUG_MARGIN 16 +#define VMA_DEBUG_DETECT_CORRUPTION 1 +#include "vk_mem_alloc.h" +\endcode + +When this feature is enabled, number of bytes specified as `VMA_DEBUG_MARGIN` +(it must be multiply of 4) before and after every allocation is filled with a magic number. +This idea is also know as "canary". +Memory is automatically mapped and unmapped if necessary. + +This number is validated automatically when the allocation is destroyed. +If it's not equal to the expected value, `VMA_ASSERT()` is executed. +It clearly means that either CPU or GPU overwritten the memory outside of boundaries of the allocation, +which indicates a serious bug. + +You can also explicitly request checking margins of all allocations in all memory blocks +that belong to specified memory types by using function vmaCheckCorruption(), +or in memory blocks that belong to specified custom pool, by using function +vmaCheckPoolCorruption(). + +Margin validation (corruption detection) works only for memory types that are +`HOST_VISIBLE` and `HOST_COHERENT`. + + +\page record_and_replay Record and replay + +\section record_and_replay_introduction Introduction + +While using the library, sequence of calls to its functions together with their +parameters can be recorded to a file and later replayed using standalone player +application. It can be useful to: + +- Test correctness - check if same sequence of calls will not cause crash or + failures on a target platform. +- Gather statistics - see number of allocations, peak memory usage, number of + calls etc. +- Benchmark performance - see how much time it takes to replay the whole + sequence. + +\section record_and_replay_usage Usage + +Recording functionality is disabled by default. +To enable it, define following macro before every include of this library: + +\code +#define VMA_RECORDING_ENABLED 1 +\endcode + +To record sequence of calls to a file: Fill in +VmaAllocatorCreateInfo::pRecordSettings member while creating #VmaAllocator +object. File is opened and written during whole lifetime of the allocator. + +To replay file: Use VmaReplay - standalone command-line program. +Precompiled binary can be found in "bin" directory. +Its source can be found in "src/VmaReplay" directory. +Its project is generated by Premake. +Command line syntax is printed when the program is launched without parameters. +Basic usage: + + VmaReplay.exe MyRecording.csv + +Documentation of file format can be found in file: "docs/Recording file format.md". +It's a human-readable, text file in CSV format (Comma Separated Values). + +\section record_and_replay_additional_considerations Additional considerations + +- Replaying file that was recorded on a different GPU (with different parameters + like `bufferImageGranularity`, `nonCoherentAtomSize`, and especially different + set of memory heaps and types) may give different performance and memory usage + results, as well as issue some warnings and errors. +- Current implementation of recording in VMA, as well as VmaReplay application, is + coded and tested only on Windows. Inclusion of recording code is driven by + `VMA_RECORDING_ENABLED` macro. Support for other platforms should be easy to + add. Contributions are welcomed. + + +\page usage_patterns Recommended usage patterns + +See also slides from talk: +[Sawicki, Adam. Advanced Graphics Techniques Tutorial: Memory management in Vulkan and DX12. Game Developers Conference, 2018](https://www.gdcvault.com/play/1025458/Advanced-Graphics-Techniques-Tutorial-New) + + +\section usage_patterns_common_mistakes Common mistakes + +Use of CPU_TO_GPU instead of CPU_ONLY memory + +#VMA_MEMORY_USAGE_CPU_TO_GPU is recommended only for resources that will be +mapped and written by the CPU, as well as read directly by the GPU - like some +buffers or textures updated every frame (dynamic). If you create a staging copy +of a resource to be written by CPU and then used as a source of transfer to +another resource placed in the GPU memory, that staging resource should be +created with #VMA_MEMORY_USAGE_CPU_ONLY. Please read the descriptions of these +enums carefully for details. + +Unnecessary use of custom pools + +\ref custom_memory_pools may be useful for special purposes - when you want to +keep certain type of resources separate e.g. to reserve minimum amount of memory +for them, limit maximum amount of memory they can occupy, or make some of them +push out the other through the mechanism of \ref lost_allocations. For most +resources this is not needed and so it is not recommended to create #VmaPool +objects and allocations out of them. Allocating from the default pool is sufficient. + +\section usage_patterns_simple Simple patterns + +\subsection usage_patterns_simple_render_targets Render targets + +When: +Any resources that you frequently write and read on GPU, +e.g. images used as color attachments (aka "render targets"), depth-stencil attachments, +images/buffers used as storage image/buffer (aka "Unordered Access View (UAV)"). + +What to do: +Create them in video memory that is fastest to access from GPU using +#VMA_MEMORY_USAGE_GPU_ONLY. + +Consider using [VK_KHR_dedicated_allocation](@ref vk_khr_dedicated_allocation) extension +and/or manually creating them as dedicated allocations using #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT, +especially if they are large or if you plan to destroy and recreate them e.g. when +display resolution changes. +Prefer to create such resources first and all other GPU resources (like textures and vertex buffers) later. + +\subsection usage_patterns_simple_immutable_resources Immutable resources + +When: +Any resources that you fill on CPU only once (aka "immutable") or infrequently +and then read frequently on GPU, +e.g. textures, vertex and index buffers, constant buffers that don't change often. + +What to do: +Create them in video memory that is fastest to access from GPU using +#VMA_MEMORY_USAGE_GPU_ONLY. + +To initialize content of such resource, create a CPU-side (aka "staging") copy of it +in system memory - #VMA_MEMORY_USAGE_CPU_ONLY, map it, fill it, +and submit a transfer from it to the GPU resource. +You can keep the staging copy if you need it for another upload transfer in the future. +If you don't, you can destroy it or reuse this buffer for uploading different resource +after the transfer finishes. + +Prefer to create just buffers in system memory rather than images, even for uploading textures. +Use `vkCmdCopyBufferToImage()`. +Dont use images with `VK_IMAGE_TILING_LINEAR`. + +\subsection usage_patterns_dynamic_resources Dynamic resources + +When: +Any resources that change frequently (aka "dynamic"), e.g. every frame or every draw call, +written on CPU, read on GPU. + +What to do: +Create them using #VMA_MEMORY_USAGE_CPU_TO_GPU. +You can map it and write to it directly on CPU, as well as read from it on GPU. + +This is a more complex situation. Different solutions are possible, +and the best one depends on specific GPU type, but you can use this simple approach for the start. +Prefer to write to such resource sequentially (e.g. using `memcpy`). +Don't perform random access or any reads from it on CPU, as it may be very slow. +Also note that textures written directly from the host through a mapped pointer need to be in LINEAR not OPTIMAL layout. + +\subsection usage_patterns_readback Readback + +When: +Resources that contain data written by GPU that you want to read back on CPU, +e.g. results of some computations. + +What to do: +Create them using #VMA_MEMORY_USAGE_GPU_TO_CPU. +You can write to them directly on GPU, as well as map and read them on CPU. + +\section usage_patterns_advanced Advanced patterns + +\subsection usage_patterns_integrated_graphics Detecting integrated graphics + +You can support integrated graphics (like Intel HD Graphics, AMD APU) better +by detecting it in Vulkan. +To do it, call `vkGetPhysicalDeviceProperties()`, inspect +`VkPhysicalDeviceProperties::deviceType` and look for `VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU`. +When you find it, you can assume that memory is unified and all memory types are comparably fast +to access from GPU, regardless of `VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT`. + +You can then sum up sizes of all available memory heaps and treat them as useful for +your GPU resources, instead of only `DEVICE_LOCAL` ones. +You can also prefer to create your resources in memory types that are `HOST_VISIBLE` to map them +directly instead of submitting explicit transfer (see below). + +\subsection usage_patterns_direct_vs_transfer Direct access versus transfer + +For resources that you frequently write on CPU and read on GPU, many solutions are possible: + +-# Create one copy in video memory using #VMA_MEMORY_USAGE_GPU_ONLY, + second copy in system memory using #VMA_MEMORY_USAGE_CPU_ONLY and submit explicit transfer each time. +-# Create just a single copy using #VMA_MEMORY_USAGE_CPU_TO_GPU, map it and fill it on CPU, + read it directly on GPU. +-# Create just a single copy using #VMA_MEMORY_USAGE_CPU_ONLY, map it and fill it on CPU, + read it directly on GPU. + +Which solution is the most efficient depends on your resource and especially on the GPU. +It is best to measure it and then make the decision. +Some general recommendations: + +- On integrated graphics use (2) or (3) to avoid unnecesary time and memory overhead + related to using a second copy and making transfer. +- For small resources (e.g. constant buffers) use (2). + Discrete AMD cards have special 256 MiB pool of video memory that is directly mappable. + Even if the resource ends up in system memory, its data may be cached on GPU after first + fetch over PCIe bus. +- For larger resources (e.g. textures), decide between (1) and (2). + You may want to differentiate NVIDIA and AMD, e.g. by looking for memory type that is + both `DEVICE_LOCAL` and `HOST_VISIBLE`. When you find it, use (2), otherwise use (1). + +Similarly, for resources that you frequently write on GPU and read on CPU, multiple +solutions are possible: + +-# Create one copy in video memory using #VMA_MEMORY_USAGE_GPU_ONLY, + second copy in system memory using #VMA_MEMORY_USAGE_GPU_TO_CPU and submit explicit tranfer each time. +-# Create just single copy using #VMA_MEMORY_USAGE_GPU_TO_CPU, write to it directly on GPU, + map it and read it on CPU. + +You should take some measurements to decide which option is faster in case of your specific +resource. + +Note that textures accessed directly from the host through a mapped pointer need to be in LINEAR layout, +which may slow down their usage on the device. +Textures accessed only by the device and transfer operations can use OPTIMAL layout. + +If you don't want to specialize your code for specific types of GPUs, you can still make +an simple optimization for cases when your resource ends up in mappable memory to use it +directly in this case instead of creating CPU-side staging copy. +For details see [Finding out if memory is mappable](@ref memory_mapping_finding_if_memory_mappable). + + +\page configuration Configuration + +Please check "CONFIGURATION SECTION" in the code to find macros that you can define +before each include of this file or change directly in this file to provide +your own implementation of basic facilities like assert, `min()` and `max()` functions, +mutex, atomic etc. +The library uses its own implementation of containers by default, but you can switch to using +STL containers instead. + +For example, define `VMA_ASSERT(expr)` before including the library to provide +custom implementation of the assertion, compatible with your project. +By default it is defined to standard C `assert(expr)` in `_DEBUG` configuration +and empty otherwise. + +\section config_Vulkan_functions Pointers to Vulkan functions + +There are multiple ways to import pointers to Vulkan functions in the library. +In the simplest case you don't need to do anything. +If the compilation or linking of your program or the initialization of the #VmaAllocator +doesn't work for you, you can try to reconfigure it. + +First, the allocator tries to fetch pointers to Vulkan functions linked statically, +like this: + +\code +m_VulkanFunctions.vkAllocateMemory = (PFN_vkAllocateMemory)vkAllocateMemory; +\endcode + +If you want to disable this feature, set configuration macro: `#define VMA_STATIC_VULKAN_FUNCTIONS 0`. + +Second, you can provide the pointers yourself by setting member VmaAllocatorCreateInfo::pVulkanFunctions. +You can fetch them e.g. using functions `vkGetInstanceProcAddr` and `vkGetDeviceProcAddr` or +by using a helper library like [volk](https://github.com/zeux/volk). + +Third, VMA tries to fetch remaining pointers that are still null by calling +`vkGetInstanceProcAddr` and `vkGetDeviceProcAddr` on its own. +If you want to disable this feature, set configuration macro: `#define VMA_DYNAMIC_VULKAN_FUNCTIONS 0`. + +Finally, all the function pointers required by the library (considering selected +Vulkan version and enabled extensions) are checked with `VMA_ASSERT` if they are not null. + + +\section custom_memory_allocator Custom host memory allocator + +If you use custom allocator for CPU memory rather than default operator `new` +and `delete` from C++, you can make this library using your allocator as well +by filling optional member VmaAllocatorCreateInfo::pAllocationCallbacks. These +functions will be passed to Vulkan, as well as used by the library itself to +make any CPU-side allocations. + +\section allocation_callbacks Device memory allocation callbacks + +The library makes calls to `vkAllocateMemory()` and `vkFreeMemory()` internally. +You can setup callbacks to be informed about these calls, e.g. for the purpose +of gathering some statistics. To do it, fill optional member +VmaAllocatorCreateInfo::pDeviceMemoryCallbacks. + +\section heap_memory_limit Device heap memory limit + +When device memory of certain heap runs out of free space, new allocations may +fail (returning error code) or they may succeed, silently pushing some existing +memory blocks from GPU VRAM to system RAM (which degrades performance). This +behavior is implementation-dependent - it depends on GPU vendor and graphics +driver. + +On AMD cards it can be controlled while creating Vulkan device object by using +VK_AMD_memory_overallocation_behavior extension, if available. + +Alternatively, if you want to test how your program behaves with limited amount of Vulkan device +memory available without switching your graphics card to one that really has +smaller VRAM, you can use a feature of this library intended for this purpose. +To do it, fill optional member VmaAllocatorCreateInfo::pHeapSizeLimit. + + + +\page vk_khr_dedicated_allocation VK_KHR_dedicated_allocation + +VK_KHR_dedicated_allocation is a Vulkan extension which can be used to improve +performance on some GPUs. It augments Vulkan API with possibility to query +driver whether it prefers particular buffer or image to have its own, dedicated +allocation (separate `VkDeviceMemory` block) for better efficiency - to be able +to do some internal optimizations. + +The extension is supported by this library. It will be used automatically when +enabled. To enable it: + +1 . When creating Vulkan device, check if following 2 device extensions are +supported (call `vkEnumerateDeviceExtensionProperties()`). +If yes, enable them (fill `VkDeviceCreateInfo::ppEnabledExtensionNames`). + +- VK_KHR_get_memory_requirements2 +- VK_KHR_dedicated_allocation + +If you enabled these extensions: + +2 . Use #VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT flag when creating +your #VmaAllocator`to inform the library that you enabled required extensions +and you want the library to use them. + +\code +allocatorInfo.flags |= VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT; + +vmaCreateAllocator(&allocatorInfo, &allocator); +\endcode + +That's all. The extension will be automatically used whenever you create a +buffer using vmaCreateBuffer() or image using vmaCreateImage(). + +When using the extension together with Vulkan Validation Layer, you will receive +warnings like this: + + vkBindBufferMemory(): Binding memory to buffer 0x33 but vkGetBufferMemoryRequirements() has not been called on that buffer. + +It is OK, you should just ignore it. It happens because you use function +`vkGetBufferMemoryRequirements2KHR()` instead of standard +`vkGetBufferMemoryRequirements()`, while the validation layer seems to be +unaware of it. + +To learn more about this extension, see: + +- [VK_KHR_dedicated_allocation in Vulkan specification](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/html/chap44.html#VK_KHR_dedicated_allocation) +- [VK_KHR_dedicated_allocation unofficial manual](http://asawicki.info/articles/VK_KHR_dedicated_allocation.php5) + + + +\page vk_amd_device_coherent_memory VK_AMD_device_coherent_memory + +VK_AMD_device_coherent_memory is a device extension that enables access to +additional memory types with `VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD` and +`VK_MEMORY_PROPERTY_DEVICE_UNCACHED_BIT_AMD` flag. It is useful mostly for +allocation of buffers intended for writing "breadcrumb markers" in between passes +or draw calls, which in turn are useful for debugging GPU crash/hang/TDR cases. + +When the extension is available but has not been enabled, Vulkan physical device +still exposes those memory types, but their usage is forbidden. VMA automatically +takes care of that - it returns `VK_ERROR_FEATURE_NOT_PRESENT` when an attempt +to allocate memory of such type is made. + +If you want to use this extension in connection with VMA, follow these steps: + +\section vk_amd_device_coherent_memory_initialization Initialization + +1) Call `vkEnumerateDeviceExtensionProperties` for the physical device. +Check if the extension is supported - if returned array of `VkExtensionProperties` contains "VK_AMD_device_coherent_memory". + +2) Call `vkGetPhysicalDeviceFeatures2` for the physical device instead of old `vkGetPhysicalDeviceFeatures`. +Attach additional structure `VkPhysicalDeviceCoherentMemoryFeaturesAMD` to `VkPhysicalDeviceFeatures2::pNext` to be returned. +Check if the device feature is really supported - check if `VkPhysicalDeviceCoherentMemoryFeaturesAMD::deviceCoherentMemory` is true. + +3) While creating device with `vkCreateDevice`, enable this extension - add "VK_AMD_device_coherent_memory" +to the list passed as `VkDeviceCreateInfo::ppEnabledExtensionNames`. + +4) While creating the device, also don't set `VkDeviceCreateInfo::pEnabledFeatures`. +Fill in `VkPhysicalDeviceFeatures2` structure instead and pass it as `VkDeviceCreateInfo::pNext`. +Enable this device feature - attach additional structure `VkPhysicalDeviceCoherentMemoryFeaturesAMD` to +`VkPhysicalDeviceFeatures2::pNext` and set its member `deviceCoherentMemory` to `VK_TRUE`. + +5) While creating #VmaAllocator with vmaCreateAllocator() inform VMA that you +have enabled this extension and feature - add #VMA_ALLOCATOR_CREATE_AMD_DEVICE_COHERENT_MEMORY_BIT +to VmaAllocatorCreateInfo::flags. + +\section vk_amd_device_coherent_memory_usage Usage + +After following steps described above, you can create VMA allocations and custom pools +out of the special `DEVICE_COHERENT` and `DEVICE_UNCACHED` memory types on eligible +devices. There are multiple ways to do it, for example: + +- You can request or prefer to allocate out of such memory types by adding + `VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD` to VmaAllocationCreateInfo::requiredFlags + or VmaAllocationCreateInfo::preferredFlags. Those flags can be freely mixed with + other ways of \ref choosing_memory_type, like setting VmaAllocationCreateInfo::usage. +- If you manually found memory type index to use for this purpose, force allocation + from this specific index by setting VmaAllocationCreateInfo::memoryTypeBits `= 1u << index`. + +\section vk_amd_device_coherent_memory_more_information More information + +To learn more about this extension, see [VK_AMD_device_coherent_memory in Vulkan specification](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/html/chap44.html#VK_AMD_device_coherent_memory) + +Example use of this extension can be found in the code of the sample and test suite +accompanying this library. + + +\page enabling_buffer_device_address Enabling buffer device address + +Device extension VK_KHR_buffer_device_address +allow to fetch raw GPU pointer to a buffer and pass it for usage in a shader code. +It is promoted to core Vulkan 1.2. + +If you want to use this feature in connection with VMA, follow these steps: + +\section enabling_buffer_device_address_initialization Initialization + +1) (For Vulkan version < 1.2) Call `vkEnumerateDeviceExtensionProperties` for the physical device. +Check if the extension is supported - if returned array of `VkExtensionProperties` contains +"VK_KHR_buffer_device_address". + +2) Call `vkGetPhysicalDeviceFeatures2` for the physical device instead of old `vkGetPhysicalDeviceFeatures`. +Attach additional structure `VkPhysicalDeviceBufferDeviceAddressFeatures*` to `VkPhysicalDeviceFeatures2::pNext` to be returned. +Check if the device feature is really supported - check if `VkPhysicalDeviceBufferDeviceAddressFeatures*::bufferDeviceAddress` is true. + +3) (For Vulkan version < 1.2) While creating device with `vkCreateDevice`, enable this extension - add +"VK_KHR_buffer_device_address" to the list passed as `VkDeviceCreateInfo::ppEnabledExtensionNames`. + +4) While creating the device, also don't set `VkDeviceCreateInfo::pEnabledFeatures`. +Fill in `VkPhysicalDeviceFeatures2` structure instead and pass it as `VkDeviceCreateInfo::pNext`. +Enable this device feature - attach additional structure `VkPhysicalDeviceBufferDeviceAddressFeatures*` to +`VkPhysicalDeviceFeatures2::pNext` and set its member `bufferDeviceAddress` to `VK_TRUE`. + +5) While creating #VmaAllocator with vmaCreateAllocator() inform VMA that you +have enabled this feature - add #VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT +to VmaAllocatorCreateInfo::flags. + +\section enabling_buffer_device_address_usage Usage + +After following steps described above, you can create buffers with `VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT*` using VMA. +The library automatically adds `VK_MEMORY_ALLOCATE_DEVICE_ADDRESS_BIT*` to +allocated memory blocks wherever it might be needed. + +Please note that the library supports only `VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT*`. +The second part of this functionality related to "capture and replay" is not supported, +as it is intended for usage in debugging tools like RenderDoc, not in everyday Vulkan usage. + +\section enabling_buffer_device_address_more_information More information + +To learn more about this extension, see [VK_KHR_buffer_device_address in Vulkan specification](https://www.khronos.org/registry/vulkan/specs/1.2-extensions/html/chap46.html#VK_KHR_buffer_device_address) + +Example use of this extension can be found in the code of the sample and test suite +accompanying this library. + +\page general_considerations General considerations + +\section general_considerations_thread_safety Thread safety + +- The library has no global state, so separate #VmaAllocator objects can be used + independently. + There should be no need to create multiple such objects though - one per `VkDevice` is enough. +- By default, all calls to functions that take #VmaAllocator as first parameter + are safe to call from multiple threads simultaneously because they are + synchronized internally when needed. +- When the allocator is created with #VMA_ALLOCATOR_CREATE_EXTERNALLY_SYNCHRONIZED_BIT + flag, calls to functions that take such #VmaAllocator object must be + synchronized externally. +- Access to a #VmaAllocation object must be externally synchronized. For example, + you must not call vmaGetAllocationInfo() and vmaMapMemory() from different + threads at the same time if you pass the same #VmaAllocation object to these + functions. + +\section general_considerations_validation_layer_warnings Validation layer warnings + +When using this library, you can meet following types of warnings issued by +Vulkan validation layer. They don't necessarily indicate a bug, so you may need +to just ignore them. + +- *vkBindBufferMemory(): Binding memory to buffer 0xeb8e4 but vkGetBufferMemoryRequirements() has not been called on that buffer.* + - It happens when VK_KHR_dedicated_allocation extension is enabled. + `vkGetBufferMemoryRequirements2KHR` function is used instead, while validation layer seems to be unaware of it. +- *Mapping an image with layout VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL can result in undefined behavior if this memory is used by the device. Only GENERAL or PREINITIALIZED should be used.* + - It happens when you map a buffer or image, because the library maps entire + `VkDeviceMemory` block, where different types of images and buffers may end + up together, especially on GPUs with unified memory like Intel. +- *Non-linear image 0xebc91 is aliased with linear buffer 0xeb8e4 which may indicate a bug.* + - It happens when you use lost allocations, and a new image or buffer is + created in place of an existing object that bacame lost. + - It may happen also when you use [defragmentation](@ref defragmentation). + +\section general_considerations_allocation_algorithm Allocation algorithm + +The library uses following algorithm for allocation, in order: + +-# Try to find free range of memory in existing blocks. +-# If failed, try to create a new block of `VkDeviceMemory`, with preferred block size. +-# If failed, try to create such block with size/2, size/4, size/8. +-# If failed and #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT flag was + specified, try to find space in existing blocks, possilby making some other + allocations lost. +-# If failed, try to allocate separate `VkDeviceMemory` for this allocation, + just like when you use #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. +-# If failed, choose other memory type that meets the requirements specified in + VmaAllocationCreateInfo and go to point 1. +-# If failed, return `VK_ERROR_OUT_OF_DEVICE_MEMORY`. + +\section general_considerations_features_not_supported Features not supported + +Features deliberately excluded from the scope of this library: + +- Data transfer. Uploading (straming) and downloading data of buffers and images + between CPU and GPU memory and related synchronization is responsibility of the user. + Defining some "texture" object that would automatically stream its data from a + staging copy in CPU memory to GPU memory would rather be a feature of another, + higher-level library implemented on top of VMA. +- Allocations for imported/exported external memory. They tend to require + explicit memory type index and dedicated allocation anyway, so they don't + interact with main features of this library. Such special purpose allocations + should be made manually, using `vkCreateBuffer()` and `vkAllocateMemory()`. +- Sub-allocation of parts of one large buffer. Although recommended as a good practice, + it is the user's responsibility to implement such logic on top of VMA. +- Recreation of buffers and images. Although the library has functions for + buffer and image creation (vmaCreateBuffer(), vmaCreateImage()), you need to + recreate these objects yourself after defragmentation. That's because the big + structures `VkBufferCreateInfo`, `VkImageCreateInfo` are not stored in + #VmaAllocation object. +- Handling CPU memory allocation failures. When dynamically creating small C++ + objects in CPU memory (not Vulkan memory), allocation failures are not checked + and handled gracefully, because that would complicate code significantly and + is usually not needed in desktop PC applications anyway. + Success of an allocation is just checked with an assert. +- Code free of any compiler warnings. Maintaining the library to compile and + work correctly on so many different platforms is hard enough. Being free of + any warnings, on any version of any compiler, is simply not feasible. +- This is a C++ library with C interface. + Bindings or ports to any other programming languages are welcomed as external projects and + are not going to be included into this repository. + +*/ + +#ifdef __cplusplus +extern "C" { +#endif + +/* +Define this macro to 0/1 to disable/enable support for recording functionality, +available through VmaAllocatorCreateInfo::pRecordSettings. +*/ +#ifndef VMA_RECORDING_ENABLED + #define VMA_RECORDING_ENABLED 0 +#endif + +#if !defined(NOMINMAX) && defined(VMA_IMPLEMENTATION) + #define NOMINMAX // For windows.h +#endif + +#if defined(__ANDROID__) && defined(VK_NO_PROTOTYPES) && VMA_STATIC_VULKAN_FUNCTIONS + extern PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr; + extern PFN_vkGetDeviceProcAddr vkGetDeviceProcAddr; + extern PFN_vkGetPhysicalDeviceProperties vkGetPhysicalDeviceProperties; + extern PFN_vkGetPhysicalDeviceMemoryProperties vkGetPhysicalDeviceMemoryProperties; + extern PFN_vkAllocateMemory vkAllocateMemory; + extern PFN_vkFreeMemory vkFreeMemory; + extern PFN_vkMapMemory vkMapMemory; + extern PFN_vkUnmapMemory vkUnmapMemory; + extern PFN_vkFlushMappedMemoryRanges vkFlushMappedMemoryRanges; + extern PFN_vkInvalidateMappedMemoryRanges vkInvalidateMappedMemoryRanges; + extern PFN_vkBindBufferMemory vkBindBufferMemory; + extern PFN_vkBindImageMemory vkBindImageMemory; + extern PFN_vkGetBufferMemoryRequirements vkGetBufferMemoryRequirements; + extern PFN_vkGetImageMemoryRequirements vkGetImageMemoryRequirements; + extern PFN_vkCreateBuffer vkCreateBuffer; + extern PFN_vkDestroyBuffer vkDestroyBuffer; + extern PFN_vkCreateImage vkCreateImage; + extern PFN_vkDestroyImage vkDestroyImage; + extern PFN_vkCmdCopyBuffer vkCmdCopyBuffer; + #if VMA_VULKAN_VERSION >= 1001000 + extern PFN_vkGetBufferMemoryRequirements2 vkGetBufferMemoryRequirements2; + extern PFN_vkGetImageMemoryRequirements2 vkGetImageMemoryRequirements2; + extern PFN_vkBindBufferMemory2 vkBindBufferMemory2; + extern PFN_vkBindImageMemory2 vkBindImageMemory2; + extern PFN_vkGetPhysicalDeviceMemoryProperties2 vkGetPhysicalDeviceMemoryProperties2; + #endif // #if VMA_VULKAN_VERSION >= 1001000 +#endif // #if defined(__ANDROID__) && VMA_STATIC_VULKAN_FUNCTIONS && VK_NO_PROTOTYPES + +#ifndef VULKAN_H_ + #include +#endif + +// Define this macro to declare maximum supported Vulkan version in format AAABBBCCC, +// where AAA = major, BBB = minor, CCC = patch. +// If you want to use version > 1.0, it still needs to be enabled via VmaAllocatorCreateInfo::vulkanApiVersion. +#if !defined(VMA_VULKAN_VERSION) + #if defined(VK_VERSION_1_2) + #define VMA_VULKAN_VERSION 1002000 + #elif defined(VK_VERSION_1_1) + #define VMA_VULKAN_VERSION 1001000 + #else + #define VMA_VULKAN_VERSION 1000000 + #endif +#endif + +#if !defined(VMA_DEDICATED_ALLOCATION) + #if VK_KHR_get_memory_requirements2 && VK_KHR_dedicated_allocation + #define VMA_DEDICATED_ALLOCATION 1 + #else + #define VMA_DEDICATED_ALLOCATION 0 + #endif +#endif + +#if !defined(VMA_BIND_MEMORY2) + #if VK_KHR_bind_memory2 + #define VMA_BIND_MEMORY2 1 + #else + #define VMA_BIND_MEMORY2 0 + #endif +#endif + +#if !defined(VMA_MEMORY_BUDGET) + #if VK_EXT_memory_budget && (VK_KHR_get_physical_device_properties2 || VMA_VULKAN_VERSION >= 1001000) + #define VMA_MEMORY_BUDGET 1 + #else + #define VMA_MEMORY_BUDGET 0 + #endif +#endif + +// Defined to 1 when VK_KHR_buffer_device_address device extension or equivalent core Vulkan 1.2 feature is defined in its headers. +#if !defined(VMA_BUFFER_DEVICE_ADDRESS) + #if VK_KHR_buffer_device_address || VMA_VULKAN_VERSION >= 1002000 + #define VMA_BUFFER_DEVICE_ADDRESS 1 + #else + #define VMA_BUFFER_DEVICE_ADDRESS 0 + #endif +#endif + +// Define these macros to decorate all public functions with additional code, +// before and after returned type, appropriately. This may be useful for +// exporting the functions when compiling VMA as a separate library. Example: +// #define VMA_CALL_PRE __declspec(dllexport) +// #define VMA_CALL_POST __cdecl +#ifndef VMA_CALL_PRE + #define VMA_CALL_PRE +#endif +#ifndef VMA_CALL_POST + #define VMA_CALL_POST +#endif + +// Define this macro to decorate pointers with an attribute specifying the +// length of the array they point to if they are not null. +// +// The length may be one of +// - The name of another parameter in the argument list where the pointer is declared +// - The name of another member in the struct where the pointer is declared +// - The name of a member of a struct type, meaning the value of that member in +// the context of the call. For example +// VMA_LEN_IF_NOT_NULL("VkPhysicalDeviceMemoryProperties::memoryHeapCount"), +// this means the number of memory heaps available in the device associated +// with the VmaAllocator being dealt with. +#ifndef VMA_LEN_IF_NOT_NULL + #define VMA_LEN_IF_NOT_NULL(len) +#endif + +// The VMA_NULLABLE macro is defined to be _Nullable when compiling with Clang. +// see: https://clang.llvm.org/docs/AttributeReference.html#nullable +#ifndef VMA_NULLABLE + #ifdef __clang__ + #define VMA_NULLABLE _Nullable + #else + #define VMA_NULLABLE + #endif +#endif + +// The VMA_NOT_NULL macro is defined to be _Nonnull when compiling with Clang. +// see: https://clang.llvm.org/docs/AttributeReference.html#nonnull +#ifndef VMA_NOT_NULL + #ifdef __clang__ + #define VMA_NOT_NULL _Nonnull + #else + #define VMA_NOT_NULL + #endif +#endif + +// If non-dispatchable handles are represented as pointers then we can give +// then nullability annotations +#ifndef VMA_NOT_NULL_NON_DISPATCHABLE + #if defined(__LP64__) || defined(_WIN64) || (defined(__x86_64__) && !defined(__ILP32__) ) || defined(_M_X64) || defined(__ia64) || defined (_M_IA64) || defined(__aarch64__) || defined(__powerpc64__) + #define VMA_NOT_NULL_NON_DISPATCHABLE VMA_NOT_NULL + #else + #define VMA_NOT_NULL_NON_DISPATCHABLE + #endif +#endif + +#ifndef VMA_NULLABLE_NON_DISPATCHABLE + #if defined(__LP64__) || defined(_WIN64) || (defined(__x86_64__) && !defined(__ILP32__) ) || defined(_M_X64) || defined(__ia64) || defined (_M_IA64) || defined(__aarch64__) || defined(__powerpc64__) + #define VMA_NULLABLE_NON_DISPATCHABLE VMA_NULLABLE + #else + #define VMA_NULLABLE_NON_DISPATCHABLE + #endif +#endif + +/** \struct VmaAllocator +\brief Represents main object of this library initialized. + +Fill structure #VmaAllocatorCreateInfo and call function vmaCreateAllocator() to create it. +Call function vmaDestroyAllocator() to destroy it. + +It is recommended to create just one object of this type per `VkDevice` object, +right after Vulkan is initialized and keep it alive until before Vulkan device is destroyed. +*/ +VK_DEFINE_HANDLE(VmaAllocator) + +/// Callback function called after successful vkAllocateMemory. +typedef void (VKAPI_PTR *PFN_vmaAllocateDeviceMemoryFunction)( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t memoryType, + VkDeviceMemory VMA_NOT_NULL_NON_DISPATCHABLE memory, + VkDeviceSize size, + void* VMA_NULLABLE pUserData); +/// Callback function called before vkFreeMemory. +typedef void (VKAPI_PTR *PFN_vmaFreeDeviceMemoryFunction)( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t memoryType, + VkDeviceMemory VMA_NOT_NULL_NON_DISPATCHABLE memory, + VkDeviceSize size, + void* VMA_NULLABLE pUserData); + +/** \brief Set of callbacks that the library will call for `vkAllocateMemory` and `vkFreeMemory`. + +Provided for informative purpose, e.g. to gather statistics about number of +allocations or total amount of memory allocated in Vulkan. + +Used in VmaAllocatorCreateInfo::pDeviceMemoryCallbacks. +*/ +typedef struct VmaDeviceMemoryCallbacks { + /// Optional, can be null. + PFN_vmaAllocateDeviceMemoryFunction VMA_NULLABLE pfnAllocate; + /// Optional, can be null. + PFN_vmaFreeDeviceMemoryFunction VMA_NULLABLE pfnFree; + /// Optional, can be null. + void* VMA_NULLABLE pUserData; +} VmaDeviceMemoryCallbacks; + +/// Flags for created #VmaAllocator. +typedef enum VmaAllocatorCreateFlagBits { + /** \brief Allocator and all objects created from it will not be synchronized internally, so you must guarantee they are used from only one thread at a time or synchronized externally by you. + + Using this flag may increase performance because internal mutexes are not used. + */ + VMA_ALLOCATOR_CREATE_EXTERNALLY_SYNCHRONIZED_BIT = 0x00000001, + /** \brief Enables usage of VK_KHR_dedicated_allocation extension. + + The flag works only if VmaAllocatorCreateInfo::vulkanApiVersion `== VK_API_VERSION_1_0`. + When it's `VK_API_VERSION_1_1`, the flag is ignored because the extension has been promoted to Vulkan 1.1. + + Using this extenion will automatically allocate dedicated blocks of memory for + some buffers and images instead of suballocating place for them out of bigger + memory blocks (as if you explicitly used #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT + flag) when it is recommended by the driver. It may improve performance on some + GPUs. + + You may set this flag only if you found out that following device extensions are + supported, you enabled them while creating Vulkan device passed as + VmaAllocatorCreateInfo::device, and you want them to be used internally by this + library: + + - VK_KHR_get_memory_requirements2 (device extension) + - VK_KHR_dedicated_allocation (device extension) + + When this flag is set, you can experience following warnings reported by Vulkan + validation layer. You can ignore them. + + > vkBindBufferMemory(): Binding memory to buffer 0x2d but vkGetBufferMemoryRequirements() has not been called on that buffer. + */ + VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT = 0x00000002, + /** + Enables usage of VK_KHR_bind_memory2 extension. + + The flag works only if VmaAllocatorCreateInfo::vulkanApiVersion `== VK_API_VERSION_1_0`. + When it's `VK_API_VERSION_1_1`, the flag is ignored because the extension has been promoted to Vulkan 1.1. + + You may set this flag only if you found out that this device extension is supported, + you enabled it while creating Vulkan device passed as VmaAllocatorCreateInfo::device, + and you want it to be used internally by this library. + + The extension provides functions `vkBindBufferMemory2KHR` and `vkBindImageMemory2KHR`, + which allow to pass a chain of `pNext` structures while binding. + This flag is required if you use `pNext` parameter in vmaBindBufferMemory2() or vmaBindImageMemory2(). + */ + VMA_ALLOCATOR_CREATE_KHR_BIND_MEMORY2_BIT = 0x00000004, + /** + Enables usage of VK_EXT_memory_budget extension. + + You may set this flag only if you found out that this device extension is supported, + you enabled it while creating Vulkan device passed as VmaAllocatorCreateInfo::device, + and you want it to be used internally by this library, along with another instance extension + VK_KHR_get_physical_device_properties2, which is required by it (or Vulkan 1.1, where this extension is promoted). + + The extension provides query for current memory usage and budget, which will probably + be more accurate than an estimation used by the library otherwise. + */ + VMA_ALLOCATOR_CREATE_EXT_MEMORY_BUDGET_BIT = 0x00000008, + /** + Enables usage of VK_AMD_device_coherent_memory extension. + + You may set this flag only if you: + + - found out that this device extension is supported and enabled it while creating Vulkan device passed as VmaAllocatorCreateInfo::device, + - checked that `VkPhysicalDeviceCoherentMemoryFeaturesAMD::deviceCoherentMemory` is true and set it while creating the Vulkan device, + - want it to be used internally by this library. + + The extension and accompanying device feature provide access to memory types with + `VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD` and `VK_MEMORY_PROPERTY_DEVICE_UNCACHED_BIT_AMD` flags. + They are useful mostly for writing breadcrumb markers - a common method for debugging GPU crash/hang/TDR. + + When the extension is not enabled, such memory types are still enumerated, but their usage is illegal. + To protect from this error, if you don't create the allocator with this flag, it will refuse to allocate any memory or create a custom pool in such memory type, + returning `VK_ERROR_FEATURE_NOT_PRESENT`. + */ + VMA_ALLOCATOR_CREATE_AMD_DEVICE_COHERENT_MEMORY_BIT = 0x00000010, + /** + Enables usage of "buffer device address" feature, which allows you to use function + `vkGetBufferDeviceAddress*` to get raw GPU pointer to a buffer and pass it for usage inside a shader. + + You may set this flag only if you: + + 1. (For Vulkan version < 1.2) Found as available and enabled device extension + VK_KHR_buffer_device_address. + This extension is promoted to core Vulkan 1.2. + 2. Found as available and enabled device feature `VkPhysicalDeviceBufferDeviceAddressFeatures::bufferDeviceAddress`. + + When this flag is set, you can create buffers with `VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT` using VMA. + The library automatically adds `VK_MEMORY_ALLOCATE_DEVICE_ADDRESS_BIT` to + allocated memory blocks wherever it might be needed. + + For more information, see documentation chapter \ref enabling_buffer_device_address. + */ + VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT = 0x00000020, + + VMA_ALLOCATOR_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaAllocatorCreateFlagBits; +typedef VkFlags VmaAllocatorCreateFlags; + +/** \brief Pointers to some Vulkan functions - a subset used by the library. + +Used in VmaAllocatorCreateInfo::pVulkanFunctions. +*/ +typedef struct VmaVulkanFunctions { + PFN_vkGetPhysicalDeviceProperties VMA_NULLABLE vkGetPhysicalDeviceProperties; + PFN_vkGetPhysicalDeviceMemoryProperties VMA_NULLABLE vkGetPhysicalDeviceMemoryProperties; + PFN_vkAllocateMemory VMA_NULLABLE vkAllocateMemory; + PFN_vkFreeMemory VMA_NULLABLE vkFreeMemory; + PFN_vkMapMemory VMA_NULLABLE vkMapMemory; + PFN_vkUnmapMemory VMA_NULLABLE vkUnmapMemory; + PFN_vkFlushMappedMemoryRanges VMA_NULLABLE vkFlushMappedMemoryRanges; + PFN_vkInvalidateMappedMemoryRanges VMA_NULLABLE vkInvalidateMappedMemoryRanges; + PFN_vkBindBufferMemory VMA_NULLABLE vkBindBufferMemory; + PFN_vkBindImageMemory VMA_NULLABLE vkBindImageMemory; + PFN_vkGetBufferMemoryRequirements VMA_NULLABLE vkGetBufferMemoryRequirements; + PFN_vkGetImageMemoryRequirements VMA_NULLABLE vkGetImageMemoryRequirements; + PFN_vkCreateBuffer VMA_NULLABLE vkCreateBuffer; + PFN_vkDestroyBuffer VMA_NULLABLE vkDestroyBuffer; + PFN_vkCreateImage VMA_NULLABLE vkCreateImage; + PFN_vkDestroyImage VMA_NULLABLE vkDestroyImage; + PFN_vkCmdCopyBuffer VMA_NULLABLE vkCmdCopyBuffer; +#if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + PFN_vkGetBufferMemoryRequirements2KHR VMA_NULLABLE vkGetBufferMemoryRequirements2KHR; + PFN_vkGetImageMemoryRequirements2KHR VMA_NULLABLE vkGetImageMemoryRequirements2KHR; +#endif +#if VMA_BIND_MEMORY2 || VMA_VULKAN_VERSION >= 1001000 + PFN_vkBindBufferMemory2KHR VMA_NULLABLE vkBindBufferMemory2KHR; + PFN_vkBindImageMemory2KHR VMA_NULLABLE vkBindImageMemory2KHR; +#endif +#if VMA_MEMORY_BUDGET || VMA_VULKAN_VERSION >= 1001000 + PFN_vkGetPhysicalDeviceMemoryProperties2KHR VMA_NULLABLE vkGetPhysicalDeviceMemoryProperties2KHR; +#endif +} VmaVulkanFunctions; + +/// Flags to be used in VmaRecordSettings::flags. +typedef enum VmaRecordFlagBits { + /** \brief Enables flush after recording every function call. + + Enable it if you expect your application to crash, which may leave recording file truncated. + It may degrade performance though. + */ + VMA_RECORD_FLUSH_AFTER_CALL_BIT = 0x00000001, + + VMA_RECORD_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaRecordFlagBits; +typedef VkFlags VmaRecordFlags; + +/// Parameters for recording calls to VMA functions. To be used in VmaAllocatorCreateInfo::pRecordSettings. +typedef struct VmaRecordSettings +{ + /// Flags for recording. Use #VmaRecordFlagBits enum. + VmaRecordFlags flags; + /** \brief Path to the file that should be written by the recording. + + Suggested extension: "csv". + If the file already exists, it will be overwritten. + It will be opened for the whole time #VmaAllocator object is alive. + If opening this file fails, creation of the whole allocator object fails. + */ + const char* VMA_NOT_NULL pFilePath; +} VmaRecordSettings; + +/// Description of a Allocator to be created. +typedef struct VmaAllocatorCreateInfo +{ + /// Flags for created allocator. Use #VmaAllocatorCreateFlagBits enum. + VmaAllocatorCreateFlags flags; + /// Vulkan physical device. + /** It must be valid throughout whole lifetime of created allocator. */ + VkPhysicalDevice VMA_NOT_NULL physicalDevice; + /// Vulkan device. + /** It must be valid throughout whole lifetime of created allocator. */ + VkDevice VMA_NOT_NULL device; + /// Preferred size of a single `VkDeviceMemory` block to be allocated from large heaps > 1 GiB. Optional. + /** Set to 0 to use default, which is currently 256 MiB. */ + VkDeviceSize preferredLargeHeapBlockSize; + /// Custom CPU memory allocation callbacks. Optional. + /** Optional, can be null. When specified, will also be used for all CPU-side memory allocations. */ + const VkAllocationCallbacks* VMA_NULLABLE pAllocationCallbacks; + /// Informative callbacks for `vkAllocateMemory`, `vkFreeMemory`. Optional. + /** Optional, can be null. */ + const VmaDeviceMemoryCallbacks* VMA_NULLABLE pDeviceMemoryCallbacks; + /** \brief Maximum number of additional frames that are in use at the same time as current frame. + + This value is used only when you make allocations with + VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag. Such allocation cannot become + lost if allocation.lastUseFrameIndex >= allocator.currentFrameIndex - frameInUseCount. + + For example, if you double-buffer your command buffers, so resources used for + rendering in previous frame may still be in use by the GPU at the moment you + allocate resources needed for the current frame, set this value to 1. + + If you want to allow any allocations other than used in the current frame to + become lost, set this value to 0. + */ + uint32_t frameInUseCount; + /** \brief Either null or a pointer to an array of limits on maximum number of bytes that can be allocated out of particular Vulkan memory heap. + + If not NULL, it must be a pointer to an array of + `VkPhysicalDeviceMemoryProperties::memoryHeapCount` elements, defining limit on + maximum number of bytes that can be allocated out of particular Vulkan memory + heap. + + Any of the elements may be equal to `VK_WHOLE_SIZE`, which means no limit on that + heap. This is also the default in case of `pHeapSizeLimit` = NULL. + + If there is a limit defined for a heap: + + - If user tries to allocate more memory from that heap using this allocator, + the allocation fails with `VK_ERROR_OUT_OF_DEVICE_MEMORY`. + - If the limit is smaller than heap size reported in `VkMemoryHeap::size`, the + value of this limit will be reported instead when using vmaGetMemoryProperties(). + + Warning! Using this feature may not be equivalent to installing a GPU with + smaller amount of memory, because graphics driver doesn't necessary fail new + allocations with `VK_ERROR_OUT_OF_DEVICE_MEMORY` result when memory capacity is + exceeded. It may return success and just silently migrate some device memory + blocks to system RAM. This driver behavior can also be controlled using + VK_AMD_memory_overallocation_behavior extension. + */ + const VkDeviceSize* VMA_NULLABLE VMA_LEN_IF_NOT_NULL("VkPhysicalDeviceMemoryProperties::memoryHeapCount") pHeapSizeLimit; + + /** \brief Pointers to Vulkan functions. Can be null. + + For details see [Pointers to Vulkan functions](@ref config_Vulkan_functions). + */ + const VmaVulkanFunctions* VMA_NULLABLE pVulkanFunctions; + /** \brief Parameters for recording of VMA calls. Can be null. + + If not null, it enables recording of calls to VMA functions to a file. + If support for recording is not enabled using `VMA_RECORDING_ENABLED` macro, + creation of the allocator object fails with `VK_ERROR_FEATURE_NOT_PRESENT`. + */ + const VmaRecordSettings* VMA_NULLABLE pRecordSettings; + /** \brief Handle to Vulkan instance object. + + Starting from version 3.0.0 this member is no longer optional, it must be set! + */ + VkInstance VMA_NOT_NULL instance; + /** \brief Optional. The highest version of Vulkan that the application is designed to use. + + It must be a value in the format as created by macro `VK_MAKE_VERSION` or a constant like: `VK_API_VERSION_1_1`, `VK_API_VERSION_1_0`. + The patch version number specified is ignored. Only the major and minor versions are considered. + It must be less or equal (preferably equal) to value as passed to `vkCreateInstance` as `VkApplicationInfo::apiVersion`. + Only versions 1.0, 1.1, 1.2 are supported by the current implementation. + Leaving it initialized to zero is equivalent to `VK_API_VERSION_1_0`. + */ + uint32_t vulkanApiVersion; +} VmaAllocatorCreateInfo; + +/// Creates Allocator object. +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateAllocator( + const VmaAllocatorCreateInfo* VMA_NOT_NULL pCreateInfo, + VmaAllocator VMA_NULLABLE * VMA_NOT_NULL pAllocator); + +/// Destroys allocator object. +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyAllocator( + VmaAllocator VMA_NULLABLE allocator); + +/** \brief Information about existing #VmaAllocator object. +*/ +typedef struct VmaAllocatorInfo +{ + /** \brief Handle to Vulkan instance object. + + This is the same value as has been passed through VmaAllocatorCreateInfo::instance. + */ + VkInstance VMA_NOT_NULL instance; + /** \brief Handle to Vulkan physical device object. + + This is the same value as has been passed through VmaAllocatorCreateInfo::physicalDevice. + */ + VkPhysicalDevice VMA_NOT_NULL physicalDevice; + /** \brief Handle to Vulkan device object. + + This is the same value as has been passed through VmaAllocatorCreateInfo::device. + */ + VkDevice VMA_NOT_NULL device; +} VmaAllocatorInfo; + +/** \brief Returns information about existing #VmaAllocator object - handle to Vulkan device etc. + +It might be useful if you want to keep just the #VmaAllocator handle and fetch other required handles to +`VkPhysicalDevice`, `VkDevice` etc. every time using this function. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocatorInfo(VmaAllocator VMA_NOT_NULL allocator, VmaAllocatorInfo* VMA_NOT_NULL pAllocatorInfo); + +/** +PhysicalDeviceProperties are fetched from physicalDevice by the allocator. +You can access it here, without fetching it again on your own. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetPhysicalDeviceProperties( + VmaAllocator VMA_NOT_NULL allocator, + const VkPhysicalDeviceProperties* VMA_NULLABLE * VMA_NOT_NULL ppPhysicalDeviceProperties); + +/** +PhysicalDeviceMemoryProperties are fetched from physicalDevice by the allocator. +You can access it here, without fetching it again on your own. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetMemoryProperties( + VmaAllocator VMA_NOT_NULL allocator, + const VkPhysicalDeviceMemoryProperties* VMA_NULLABLE * VMA_NOT_NULL ppPhysicalDeviceMemoryProperties); + +/** +\brief Given Memory Type Index, returns Property Flags of this memory type. + +This is just a convenience function. Same information can be obtained using +vmaGetMemoryProperties(). +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetMemoryTypeProperties( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t memoryTypeIndex, + VkMemoryPropertyFlags* VMA_NOT_NULL pFlags); + +/** \brief Sets index of the current frame. + +This function must be used if you make allocations with +#VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT and +#VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT flags to inform the allocator +when a new frame begins. Allocations queried using vmaGetAllocationInfo() cannot +become lost in the current frame. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaSetCurrentFrameIndex( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t frameIndex); + +/** \brief Calculated statistics of memory usage in entire allocator. +*/ +typedef struct VmaStatInfo +{ + /// Number of `VkDeviceMemory` Vulkan memory blocks allocated. + uint32_t blockCount; + /// Number of #VmaAllocation allocation objects allocated. + uint32_t allocationCount; + /// Number of free ranges of memory between allocations. + uint32_t unusedRangeCount; + /// Total number of bytes occupied by all allocations. + VkDeviceSize usedBytes; + /// Total number of bytes occupied by unused ranges. + VkDeviceSize unusedBytes; + VkDeviceSize allocationSizeMin, allocationSizeAvg, allocationSizeMax; + VkDeviceSize unusedRangeSizeMin, unusedRangeSizeAvg, unusedRangeSizeMax; +} VmaStatInfo; + +/// General statistics from current state of Allocator. +typedef struct VmaStats +{ + VmaStatInfo memoryType[VK_MAX_MEMORY_TYPES]; + VmaStatInfo memoryHeap[VK_MAX_MEMORY_HEAPS]; + VmaStatInfo total; +} VmaStats; + +/** \brief Retrieves statistics from current state of the Allocator. + +This function is called "calculate" not "get" because it has to traverse all +internal data structures, so it may be quite slow. For faster but more brief statistics +suitable to be called every frame or every allocation, use vmaGetBudget(). + +Note that when using allocator from multiple threads, returned information may immediately +become outdated. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaCalculateStats( + VmaAllocator VMA_NOT_NULL allocator, + VmaStats* VMA_NOT_NULL pStats); + +/** \brief Statistics of current memory usage and available budget, in bytes, for specific memory heap. +*/ +typedef struct VmaBudget +{ + /** \brief Sum size of all `VkDeviceMemory` blocks allocated from particular heap, in bytes. + */ + VkDeviceSize blockBytes; + + /** \brief Sum size of all allocations created in particular heap, in bytes. + + Usually less or equal than `blockBytes`. + Difference `blockBytes - allocationBytes` is the amount of memory allocated but unused - + available for new allocations or wasted due to fragmentation. + + It might be greater than `blockBytes` if there are some allocations in lost state, as they account + to this value as well. + */ + VkDeviceSize allocationBytes; + + /** \brief Estimated current memory usage of the program, in bytes. + + Fetched from system using `VK_EXT_memory_budget` extension if enabled. + + It might be different than `blockBytes` (usually higher) due to additional implicit objects + also occupying the memory, like swapchain, pipelines, descriptor heaps, command buffers, or + `VkDeviceMemory` blocks allocated outside of this library, if any. + */ + VkDeviceSize usage; + + /** \brief Estimated amount of memory available to the program, in bytes. + + Fetched from system using `VK_EXT_memory_budget` extension if enabled. + + It might be different (most probably smaller) than `VkMemoryHeap::size[heapIndex]` due to factors + external to the program, like other programs also consuming system resources. + Difference `budget - usage` is the amount of additional memory that can probably + be allocated without problems. Exceeding the budget may result in various problems. + */ + VkDeviceSize budget; +} VmaBudget; + +/** \brief Retrieves information about current memory budget for all memory heaps. + +\param[out] pBudget Must point to array with number of elements at least equal to number of memory heaps in physical device used. + +This function is called "get" not "calculate" because it is very fast, suitable to be called +every frame or every allocation. For more detailed statistics use vmaCalculateStats(). + +Note that when using allocator from multiple threads, returned information may immediately +become outdated. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetBudget( + VmaAllocator VMA_NOT_NULL allocator, + VmaBudget* VMA_NOT_NULL pBudget); + +#ifndef VMA_STATS_STRING_ENABLED +#define VMA_STATS_STRING_ENABLED 1 +#endif + +#if VMA_STATS_STRING_ENABLED + +/// Builds and returns statistics as string in JSON format. +/** @param[out] ppStatsString Must be freed using vmaFreeStatsString() function. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaBuildStatsString( + VmaAllocator VMA_NOT_NULL allocator, + char* VMA_NULLABLE * VMA_NOT_NULL ppStatsString, + VkBool32 detailedMap); + +VMA_CALL_PRE void VMA_CALL_POST vmaFreeStatsString( + VmaAllocator VMA_NOT_NULL allocator, + char* VMA_NULLABLE pStatsString); + +#endif // #if VMA_STATS_STRING_ENABLED + +/** \struct VmaPool +\brief Represents custom memory pool + +Fill structure VmaPoolCreateInfo and call function vmaCreatePool() to create it. +Call function vmaDestroyPool() to destroy it. + +For more information see [Custom memory pools](@ref choosing_memory_type_custom_memory_pools). +*/ +VK_DEFINE_HANDLE(VmaPool) + +typedef enum VmaMemoryUsage +{ + /** No intended memory usage specified. + Use other members of VmaAllocationCreateInfo to specify your requirements. + */ + VMA_MEMORY_USAGE_UNKNOWN = 0, + /** Memory will be used on device only, so fast access from the device is preferred. + It usually means device-local GPU (video) memory. + No need to be mappable on host. + It is roughly equivalent of `D3D12_HEAP_TYPE_DEFAULT`. + + Usage: + + - Resources written and read by device, e.g. images used as attachments. + - Resources transferred from host once (immutable) or infrequently and read by + device multiple times, e.g. textures to be sampled, vertex buffers, uniform + (constant) buffers, and majority of other types of resources used on GPU. + + Allocation may still end up in `HOST_VISIBLE` memory on some implementations. + In such case, you are free to map it. + You can use #VMA_ALLOCATION_CREATE_MAPPED_BIT with this usage type. + */ + VMA_MEMORY_USAGE_GPU_ONLY = 1, + /** Memory will be mappable on host. + It usually means CPU (system) memory. + Guarantees to be `HOST_VISIBLE` and `HOST_COHERENT`. + CPU access is typically uncached. Writes may be write-combined. + Resources created in this pool may still be accessible to the device, but access to them can be slow. + It is roughly equivalent of `D3D12_HEAP_TYPE_UPLOAD`. + + Usage: Staging copy of resources used as transfer source. + */ + VMA_MEMORY_USAGE_CPU_ONLY = 2, + /** + Memory that is both mappable on host (guarantees to be `HOST_VISIBLE`) and preferably fast to access by GPU. + CPU access is typically uncached. Writes may be write-combined. + + Usage: Resources written frequently by host (dynamic), read by device. E.g. textures (with LINEAR layout), vertex buffers, uniform buffers updated every frame or every draw call. + */ + VMA_MEMORY_USAGE_CPU_TO_GPU = 3, + /** Memory mappable on host (guarantees to be `HOST_VISIBLE`) and cached. + It is roughly equivalent of `D3D12_HEAP_TYPE_READBACK`. + + Usage: + + - Resources written by device, read by host - results of some computations, e.g. screen capture, average scene luminance for HDR tone mapping. + - Any resources read or accessed randomly on host, e.g. CPU-side copy of vertex buffer used as source of transfer, but also used for collision detection. + */ + VMA_MEMORY_USAGE_GPU_TO_CPU = 4, + /** CPU memory - memory that is preferably not `DEVICE_LOCAL`, but also not guaranteed to be `HOST_VISIBLE`. + + Usage: Staging copy of resources moved from GPU memory to CPU memory as part + of custom paging/residency mechanism, to be moved back to GPU memory when needed. + */ + VMA_MEMORY_USAGE_CPU_COPY = 5, + /** Lazily allocated GPU memory having `VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT`. + Exists mostly on mobile platforms. Using it on desktop PC or other GPUs with no such memory type present will fail the allocation. + + Usage: Memory for transient attachment images (color attachments, depth attachments etc.), created with `VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT`. + + Allocations with this usage are always created as dedicated - it implies #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. + */ + VMA_MEMORY_USAGE_GPU_LAZILY_ALLOCATED = 6, + + VMA_MEMORY_USAGE_MAX_ENUM = 0x7FFFFFFF +} VmaMemoryUsage; + +/// Flags to be passed as VmaAllocationCreateInfo::flags. +typedef enum VmaAllocationCreateFlagBits { + /** \brief Set this flag if the allocation should have its own memory block. + + Use it for special, big resources, like fullscreen images used as attachments. + + You should not use this flag if VmaAllocationCreateInfo::pool is not null. + */ + VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT = 0x00000001, + + /** \brief Set this flag to only try to allocate from existing `VkDeviceMemory` blocks and never create new such block. + + If new allocation cannot be placed in any of the existing blocks, allocation + fails with `VK_ERROR_OUT_OF_DEVICE_MEMORY` error. + + You should not use #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT and + #VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT at the same time. It makes no sense. + + If VmaAllocationCreateInfo::pool is not null, this flag is implied and ignored. */ + VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT = 0x00000002, + /** \brief Set this flag to use a memory that will be persistently mapped and retrieve pointer to it. + + Pointer to mapped memory will be returned through VmaAllocationInfo::pMappedData. + + It is valid to use this flag for allocation made from memory type that is not + `HOST_VISIBLE`. This flag is then ignored and memory is not mapped. This is + useful if you need an allocation that is efficient to use on GPU + (`DEVICE_LOCAL`) and still want to map it directly if possible on platforms that + support it (e.g. Intel GPU). + + You should not use this flag together with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT. + */ + VMA_ALLOCATION_CREATE_MAPPED_BIT = 0x00000004, + /** Allocation created with this flag can become lost as a result of another + allocation with #VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT flag, so you + must check it before use. + + To check if allocation is not lost, call vmaGetAllocationInfo() and check if + VmaAllocationInfo::deviceMemory is not `VK_NULL_HANDLE`. + + For details about supporting lost allocations, see Lost Allocations + chapter of User Guide on Main Page. + + You should not use this flag together with #VMA_ALLOCATION_CREATE_MAPPED_BIT. + */ + VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT = 0x00000008, + /** While creating allocation using this flag, other allocations that were + created with flag #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT can become lost. + + For details about supporting lost allocations, see Lost Allocations + chapter of User Guide on Main Page. + */ + VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT = 0x00000010, + /** Set this flag to treat VmaAllocationCreateInfo::pUserData as pointer to a + null-terminated string. Instead of copying pointer value, a local copy of the + string is made and stored in allocation's `pUserData`. The string is automatically + freed together with the allocation. It is also used in vmaBuildStatsString(). + */ + VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT = 0x00000020, + /** Allocation will be created from upper stack in a double stack pool. + + This flag is only allowed for custom pools created with #VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT flag. + */ + VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT = 0x00000040, + /** Create both buffer/image and allocation, but don't bind them together. + It is useful when you want to bind yourself to do some more advanced binding, e.g. using some extensions. + The flag is meaningful only with functions that bind by default: vmaCreateBuffer(), vmaCreateImage(). + Otherwise it is ignored. + */ + VMA_ALLOCATION_CREATE_DONT_BIND_BIT = 0x00000080, + /** Create allocation only if additional device memory required for it, if any, won't exceed + memory budget. Otherwise return `VK_ERROR_OUT_OF_DEVICE_MEMORY`. + */ + VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT = 0x00000100, + + /** Allocation strategy that chooses smallest possible free range for the + allocation. + */ + VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT = 0x00010000, + /** Allocation strategy that chooses biggest possible free range for the + allocation. + */ + VMA_ALLOCATION_CREATE_STRATEGY_WORST_FIT_BIT = 0x00020000, + /** Allocation strategy that chooses first suitable free range for the + allocation. + + "First" doesn't necessarily means the one with smallest offset in memory, + but rather the one that is easiest and fastest to find. + */ + VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT = 0x00040000, + + /** Allocation strategy that tries to minimize memory usage. + */ + VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT = VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT, + /** Allocation strategy that tries to minimize allocation time. + */ + VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT = VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT, + /** Allocation strategy that tries to minimize memory fragmentation. + */ + VMA_ALLOCATION_CREATE_STRATEGY_MIN_FRAGMENTATION_BIT = VMA_ALLOCATION_CREATE_STRATEGY_WORST_FIT_BIT, + + /** A bit mask to extract only `STRATEGY` bits from entire set of flags. + */ + VMA_ALLOCATION_CREATE_STRATEGY_MASK = + VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT | + VMA_ALLOCATION_CREATE_STRATEGY_WORST_FIT_BIT | + VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT, + + VMA_ALLOCATION_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaAllocationCreateFlagBits; +typedef VkFlags VmaAllocationCreateFlags; + +typedef struct VmaAllocationCreateInfo +{ + /// Use #VmaAllocationCreateFlagBits enum. + VmaAllocationCreateFlags flags; + /** \brief Intended usage of memory. + + You can leave #VMA_MEMORY_USAGE_UNKNOWN if you specify memory requirements in other way. \n + If `pool` is not null, this member is ignored. + */ + VmaMemoryUsage usage; + /** \brief Flags that must be set in a Memory Type chosen for an allocation. + + Leave 0 if you specify memory requirements in other way. \n + If `pool` is not null, this member is ignored.*/ + VkMemoryPropertyFlags requiredFlags; + /** \brief Flags that preferably should be set in a memory type chosen for an allocation. + + Set to 0 if no additional flags are preferred. \n + If `pool` is not null, this member is ignored. */ + VkMemoryPropertyFlags preferredFlags; + /** \brief Bitmask containing one bit set for every memory type acceptable for this allocation. + + Value 0 is equivalent to `UINT32_MAX` - it means any memory type is accepted if + it meets other requirements specified by this structure, with no further + restrictions on memory type index. \n + If `pool` is not null, this member is ignored. + */ + uint32_t memoryTypeBits; + /** \brief Pool that this allocation should be created in. + + Leave `VK_NULL_HANDLE` to allocate from default pool. If not null, members: + `usage`, `requiredFlags`, `preferredFlags`, `memoryTypeBits` are ignored. + */ + VmaPool VMA_NULLABLE pool; + /** \brief Custom general-purpose pointer that will be stored in #VmaAllocation, can be read as VmaAllocationInfo::pUserData and changed using vmaSetAllocationUserData(). + + If #VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT is used, it must be either + null or pointer to a null-terminated string. The string will be then copied to + internal buffer, so it doesn't need to be valid after allocation call. + */ + void* VMA_NULLABLE pUserData; +} VmaAllocationCreateInfo; + +/** +\brief Helps to find memoryTypeIndex, given memoryTypeBits and VmaAllocationCreateInfo. + +This algorithm tries to find a memory type that: + +- Is allowed by memoryTypeBits. +- Contains all the flags from pAllocationCreateInfo->requiredFlags. +- Matches intended usage. +- Has as many flags from pAllocationCreateInfo->preferredFlags as possible. + +\return Returns VK_ERROR_FEATURE_NOT_PRESENT if not found. Receiving such result +from this function or any other allocating function probably means that your +device doesn't support any memory type with requested features for the specific +type of resource you want to use it for. Please check parameters of your +resource, like image layout (OPTIMAL versus LINEAR) or mip level count. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndex( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t memoryTypeBits, + const VmaAllocationCreateInfo* VMA_NOT_NULL pAllocationCreateInfo, + uint32_t* VMA_NOT_NULL pMemoryTypeIndex); + +/** +\brief Helps to find memoryTypeIndex, given VkBufferCreateInfo and VmaAllocationCreateInfo. + +It can be useful e.g. to determine value to be used as VmaPoolCreateInfo::memoryTypeIndex. +It internally creates a temporary, dummy buffer that never has memory bound. +It is just a convenience function, equivalent to calling: + +- `vkCreateBuffer` +- `vkGetBufferMemoryRequirements` +- `vmaFindMemoryTypeIndex` +- `vkDestroyBuffer` +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndexForBufferInfo( + VmaAllocator VMA_NOT_NULL allocator, + const VkBufferCreateInfo* VMA_NOT_NULL pBufferCreateInfo, + const VmaAllocationCreateInfo* VMA_NOT_NULL pAllocationCreateInfo, + uint32_t* VMA_NOT_NULL pMemoryTypeIndex); + +/** +\brief Helps to find memoryTypeIndex, given VkImageCreateInfo and VmaAllocationCreateInfo. + +It can be useful e.g. to determine value to be used as VmaPoolCreateInfo::memoryTypeIndex. +It internally creates a temporary, dummy image that never has memory bound. +It is just a convenience function, equivalent to calling: + +- `vkCreateImage` +- `vkGetImageMemoryRequirements` +- `vmaFindMemoryTypeIndex` +- `vkDestroyImage` +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndexForImageInfo( + VmaAllocator VMA_NOT_NULL allocator, + const VkImageCreateInfo* VMA_NOT_NULL pImageCreateInfo, + const VmaAllocationCreateInfo* VMA_NOT_NULL pAllocationCreateInfo, + uint32_t* VMA_NOT_NULL pMemoryTypeIndex); + +/// Flags to be passed as VmaPoolCreateInfo::flags. +typedef enum VmaPoolCreateFlagBits { + /** \brief Use this flag if you always allocate only buffers and linear images or only optimal images out of this pool and so Buffer-Image Granularity can be ignored. + + This is an optional optimization flag. + + If you always allocate using vmaCreateBuffer(), vmaCreateImage(), + vmaAllocateMemoryForBuffer(), then you don't need to use it because allocator + knows exact type of your allocations so it can handle Buffer-Image Granularity + in the optimal way. + + If you also allocate using vmaAllocateMemoryForImage() or vmaAllocateMemory(), + exact type of such allocations is not known, so allocator must be conservative + in handling Buffer-Image Granularity, which can lead to suboptimal allocation + (wasted memory). In that case, if you can make sure you always allocate only + buffers and linear images or only optimal images out of this pool, use this flag + to make allocator disregard Buffer-Image Granularity and so make allocations + faster and more optimal. + */ + VMA_POOL_CREATE_IGNORE_BUFFER_IMAGE_GRANULARITY_BIT = 0x00000002, + + /** \brief Enables alternative, linear allocation algorithm in this pool. + + Specify this flag to enable linear allocation algorithm, which always creates + new allocations after last one and doesn't reuse space from allocations freed in + between. It trades memory consumption for simplified algorithm and data + structure, which has better performance and uses less memory for metadata. + + By using this flag, you can achieve behavior of free-at-once, stack, + ring buffer, and double stack. For details, see documentation chapter + \ref linear_algorithm. + + When using this flag, you must specify VmaPoolCreateInfo::maxBlockCount == 1 (or 0 for default). + + For more details, see [Linear allocation algorithm](@ref linear_algorithm). + */ + VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT = 0x00000004, + + /** \brief Enables alternative, buddy allocation algorithm in this pool. + + It operates on a tree of blocks, each having size that is a power of two and + a half of its parent's size. Comparing to default algorithm, this one provides + faster allocation and deallocation and decreased external fragmentation, + at the expense of more memory wasted (internal fragmentation). + + For more details, see [Buddy allocation algorithm](@ref buddy_algorithm). + */ + VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT = 0x00000008, + + /** Bit mask to extract only `ALGORITHM` bits from entire set of flags. + */ + VMA_POOL_CREATE_ALGORITHM_MASK = + VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT | + VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT, + + VMA_POOL_CREATE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaPoolCreateFlagBits; +typedef VkFlags VmaPoolCreateFlags; + +/** \brief Describes parameter of created #VmaPool. +*/ +typedef struct VmaPoolCreateInfo { + /** \brief Vulkan memory type index to allocate this pool from. + */ + uint32_t memoryTypeIndex; + /** \brief Use combination of #VmaPoolCreateFlagBits. + */ + VmaPoolCreateFlags flags; + /** \brief Size of a single `VkDeviceMemory` block to be allocated as part of this pool, in bytes. Optional. + + Specify nonzero to set explicit, constant size of memory blocks used by this + pool. + + Leave 0 to use default and let the library manage block sizes automatically. + Sizes of particular blocks may vary. + */ + VkDeviceSize blockSize; + /** \brief Minimum number of blocks to be always allocated in this pool, even if they stay empty. + + Set to 0 to have no preallocated blocks and allow the pool be completely empty. + */ + size_t minBlockCount; + /** \brief Maximum number of blocks that can be allocated in this pool. Optional. + + Set to 0 to use default, which is `SIZE_MAX`, which means no limit. + + Set to same value as VmaPoolCreateInfo::minBlockCount to have fixed amount of memory allocated + throughout whole lifetime of this pool. + */ + size_t maxBlockCount; + /** \brief Maximum number of additional frames that are in use at the same time as current frame. + + This value is used only when you make allocations with + #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag. Such allocation cannot become + lost if allocation.lastUseFrameIndex >= allocator.currentFrameIndex - frameInUseCount. + + For example, if you double-buffer your command buffers, so resources used for + rendering in previous frame may still be in use by the GPU at the moment you + allocate resources needed for the current frame, set this value to 1. + + If you want to allow any allocations other than used in the current frame to + become lost, set this value to 0. + */ + uint32_t frameInUseCount; +} VmaPoolCreateInfo; + +/** \brief Describes parameter of existing #VmaPool. +*/ +typedef struct VmaPoolStats { + /** \brief Total amount of `VkDeviceMemory` allocated from Vulkan for this pool, in bytes. + */ + VkDeviceSize size; + /** \brief Total number of bytes in the pool not used by any #VmaAllocation. + */ + VkDeviceSize unusedSize; + /** \brief Number of #VmaAllocation objects created from this pool that were not destroyed or lost. + */ + size_t allocationCount; + /** \brief Number of continuous memory ranges in the pool not used by any #VmaAllocation. + */ + size_t unusedRangeCount; + /** \brief Size of the largest continuous free memory region available for new allocation. + + Making a new allocation of that size is not guaranteed to succeed because of + possible additional margin required to respect alignment and buffer/image + granularity. + */ + VkDeviceSize unusedRangeSizeMax; + /** \brief Number of `VkDeviceMemory` blocks allocated for this pool. + */ + size_t blockCount; +} VmaPoolStats; + +/** \brief Allocates Vulkan device memory and creates #VmaPool object. + +@param allocator Allocator object. +@param pCreateInfo Parameters of pool to create. +@param[out] pPool Handle to created pool. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreatePool( + VmaAllocator VMA_NOT_NULL allocator, + const VmaPoolCreateInfo* VMA_NOT_NULL pCreateInfo, + VmaPool VMA_NULLABLE * VMA_NOT_NULL pPool); + +/** \brief Destroys #VmaPool object and frees Vulkan device memory. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyPool( + VmaAllocator VMA_NOT_NULL allocator, + VmaPool VMA_NULLABLE pool); + +/** \brief Retrieves statistics of existing #VmaPool object. + +@param allocator Allocator object. +@param pool Pool object. +@param[out] pPoolStats Statistics of specified pool. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetPoolStats( + VmaAllocator VMA_NOT_NULL allocator, + VmaPool VMA_NOT_NULL pool, + VmaPoolStats* VMA_NOT_NULL pPoolStats); + +/** \brief Marks all allocations in given pool as lost if they are not used in current frame or VmaPoolCreateInfo::frameInUseCount back from now. + +@param allocator Allocator object. +@param pool Pool. +@param[out] pLostAllocationCount Number of allocations marked as lost. Optional - pass null if you don't need this information. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaMakePoolAllocationsLost( + VmaAllocator VMA_NOT_NULL allocator, + VmaPool VMA_NOT_NULL pool, + size_t* VMA_NULLABLE pLostAllocationCount); + +/** \brief Checks magic number in margins around all allocations in given memory pool in search for corruptions. + +Corruption detection is enabled only when `VMA_DEBUG_DETECT_CORRUPTION` macro is defined to nonzero, +`VMA_DEBUG_MARGIN` is defined to nonzero and the pool is created in memory type that is +`HOST_VISIBLE` and `HOST_COHERENT`. For more information, see [Corruption detection](@ref debugging_memory_usage_corruption_detection). + +Possible return values: + +- `VK_ERROR_FEATURE_NOT_PRESENT` - corruption detection is not enabled for specified pool. +- `VK_SUCCESS` - corruption detection has been performed and succeeded. +- `VK_ERROR_VALIDATION_FAILED_EXT` - corruption detection has been performed and found memory corruptions around one of the allocations. + `VMA_ASSERT` is also fired in that case. +- Other value: Error returned by Vulkan, e.g. memory mapping failure. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckPoolCorruption(VmaAllocator VMA_NOT_NULL allocator, VmaPool VMA_NOT_NULL pool); + +/** \brief Retrieves name of a custom pool. + +After the call `ppName` is either null or points to an internally-owned null-terminated string +containing name of the pool that was previously set. The pointer becomes invalid when the pool is +destroyed or its name is changed using vmaSetPoolName(). +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetPoolName( + VmaAllocator VMA_NOT_NULL allocator, + VmaPool VMA_NOT_NULL pool, + const char* VMA_NULLABLE * VMA_NOT_NULL ppName); + +/** \brief Sets name of a custom pool. + +`pName` can be either null or pointer to a null-terminated string with new name for the pool. +Function makes internal copy of the string, so it can be changed or freed immediately after this call. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaSetPoolName( + VmaAllocator VMA_NOT_NULL allocator, + VmaPool VMA_NOT_NULL pool, + const char* VMA_NULLABLE pName); + +/** \struct VmaAllocation +\brief Represents single memory allocation. + +It may be either dedicated block of `VkDeviceMemory` or a specific region of a bigger block of this type +plus unique offset. + +There are multiple ways to create such object. +You need to fill structure VmaAllocationCreateInfo. +For more information see [Choosing memory type](@ref choosing_memory_type). + +Although the library provides convenience functions that create Vulkan buffer or image, +allocate memory for it and bind them together, +binding of the allocation to a buffer or an image is out of scope of the allocation itself. +Allocation object can exist without buffer/image bound, +binding can be done manually by the user, and destruction of it can be done +independently of destruction of the allocation. + +The object also remembers its size and some other information. +To retrieve this information, use function vmaGetAllocationInfo() and inspect +returned structure VmaAllocationInfo. + +Some kinds allocations can be in lost state. +For more information, see [Lost allocations](@ref lost_allocations). +*/ +VK_DEFINE_HANDLE(VmaAllocation) + +/** \brief Parameters of #VmaAllocation objects, that can be retrieved using function vmaGetAllocationInfo(). +*/ +typedef struct VmaAllocationInfo { + /** \brief Memory type index that this allocation was allocated from. + + It never changes. + */ + uint32_t memoryType; + /** \brief Handle to Vulkan memory object. + + Same memory object can be shared by multiple allocations. + + It can change after call to vmaDefragment() if this allocation is passed to the function, or if allocation is lost. + + If the allocation is lost, it is equal to `VK_NULL_HANDLE`. + */ + VkDeviceMemory VMA_NULLABLE_NON_DISPATCHABLE deviceMemory; + /** \brief Offset in `VkDeviceMemory` object to the beginning of this allocation, in bytes. `(deviceMemory, offset)` pair is unique to this allocation. + + You usually don't need to use this offset. If you create a buffer or an image together with the allocation using e.g. function + vmaCreateBuffer(), vmaCreateImage(), functions that operate on these resources refer to the beginning of the buffer or image, + not entire device memory block. Functions like vmaMapMemory(), vmaBindBufferMemory() also refer to the beginning of the allocation + and apply this offset automatically. + + It can change after call to vmaDefragment() if this allocation is passed to the function, or if allocation is lost. + */ + VkDeviceSize offset; + /** \brief Size of this allocation, in bytes. + + It never changes, unless allocation is lost. + + \note Allocation size returned in this variable may be greater than the size + requested for the resource e.g. as `VkBufferCreateInfo::size`. Whole size of the + allocation is accessible for operations on memory e.g. using a pointer after + mapping with vmaMapMemory(), but operations on the resource e.g. using + `vkCmdCopyBuffer` must be limited to the size of the resource. + */ + VkDeviceSize size; + /** \brief Pointer to the beginning of this allocation as mapped data. + + If the allocation hasn't been mapped using vmaMapMemory() and hasn't been + created with #VMA_ALLOCATION_CREATE_MAPPED_BIT flag, this value is null. + + It can change after call to vmaMapMemory(), vmaUnmapMemory(). + It can also change after call to vmaDefragment() if this allocation is passed to the function. + */ + void* VMA_NULLABLE pMappedData; + /** \brief Custom general-purpose pointer that was passed as VmaAllocationCreateInfo::pUserData or set using vmaSetAllocationUserData(). + + It can change after call to vmaSetAllocationUserData() for this allocation. + */ + void* VMA_NULLABLE pUserData; +} VmaAllocationInfo; + +/** \brief General purpose memory allocation. + +@param[out] pAllocation Handle to allocated memory. +@param[out] pAllocationInfo Optional. Information about allocated memory. It can be later fetched using function vmaGetAllocationInfo(). + +You should free the memory using vmaFreeMemory() or vmaFreeMemoryPages(). + +It is recommended to use vmaAllocateMemoryForBuffer(), vmaAllocateMemoryForImage(), +vmaCreateBuffer(), vmaCreateImage() instead whenever possible. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemory( + VmaAllocator VMA_NOT_NULL allocator, + const VkMemoryRequirements* VMA_NOT_NULL pVkMemoryRequirements, + const VmaAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, + VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); + +/** \brief General purpose memory allocation for multiple allocation objects at once. + +@param allocator Allocator object. +@param pVkMemoryRequirements Memory requirements for each allocation. +@param pCreateInfo Creation parameters for each alloction. +@param allocationCount Number of allocations to make. +@param[out] pAllocations Pointer to array that will be filled with handles to created allocations. +@param[out] pAllocationInfo Optional. Pointer to array that will be filled with parameters of created allocations. + +You should free the memory using vmaFreeMemory() or vmaFreeMemoryPages(). + +Word "pages" is just a suggestion to use this function to allocate pieces of memory needed for sparse binding. +It is just a general purpose allocation function able to make multiple allocations at once. +It may be internally optimized to be more efficient than calling vmaAllocateMemory() `allocationCount` times. + +All allocations are made using same parameters. All of them are created out of the same memory pool and type. +If any allocation fails, all allocations already made within this function call are also freed, so that when +returned result is not `VK_SUCCESS`, `pAllocation` array is always entirely filled with `VK_NULL_HANDLE`. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryPages( + VmaAllocator VMA_NOT_NULL allocator, + const VkMemoryRequirements* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pVkMemoryRequirements, + const VmaAllocationCreateInfo* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pCreateInfo, + size_t allocationCount, + VmaAllocation VMA_NULLABLE * VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations, + VmaAllocationInfo* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocationInfo); + +/** +@param[out] pAllocation Handle to allocated memory. +@param[out] pAllocationInfo Optional. Information about allocated memory. It can be later fetched using function vmaGetAllocationInfo(). + +You should free the memory using vmaFreeMemory(). +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryForBuffer( + VmaAllocator VMA_NOT_NULL allocator, + VkBuffer VMA_NOT_NULL_NON_DISPATCHABLE buffer, + const VmaAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, + VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); + +/// Function similar to vmaAllocateMemoryForBuffer(). +VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryForImage( + VmaAllocator VMA_NOT_NULL allocator, + VkImage VMA_NOT_NULL_NON_DISPATCHABLE image, + const VmaAllocationCreateInfo* VMA_NOT_NULL pCreateInfo, + VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); + +/** \brief Frees memory previously allocated using vmaAllocateMemory(), vmaAllocateMemoryForBuffer(), or vmaAllocateMemoryForImage(). + +Passing `VK_NULL_HANDLE` as `allocation` is valid. Such function call is just skipped. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaFreeMemory( + VmaAllocator VMA_NOT_NULL allocator, + const VmaAllocation VMA_NULLABLE allocation); + +/** \brief Frees memory and destroys multiple allocations. + +Word "pages" is just a suggestion to use this function to free pieces of memory used for sparse binding. +It is just a general purpose function to free memory and destroy allocations made using e.g. vmaAllocateMemory(), +vmaAllocateMemoryPages() and other functions. +It may be internally optimized to be more efficient than calling vmaFreeMemory() `allocationCount` times. + +Allocations in `pAllocations` array can come from any memory pools and types. +Passing `VK_NULL_HANDLE` as elements of `pAllocations` array is valid. Such entries are just skipped. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaFreeMemoryPages( + VmaAllocator VMA_NOT_NULL allocator, + size_t allocationCount, + const VmaAllocation VMA_NULLABLE * VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations); + +/** \brief Deprecated. + +\deprecated +In version 2.2.0 it used to try to change allocation's size without moving or reallocating it. +In current version it returns `VK_SUCCESS` only if `newSize` equals current allocation's size. +Otherwise returns `VK_ERROR_OUT_OF_POOL_MEMORY`, indicating that allocation's size could not be changed. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaResizeAllocation( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + VkDeviceSize newSize); + +/** \brief Returns current information about specified allocation and atomically marks it as used in current frame. + +Current paramteres of given allocation are returned in `pAllocationInfo`. + +This function also atomically "touches" allocation - marks it as used in current frame, +just like vmaTouchAllocation(). +If the allocation is in lost state, `pAllocationInfo->deviceMemory == VK_NULL_HANDLE`. + +Although this function uses atomics and doesn't lock any mutex, so it should be quite efficient, +you can avoid calling it too often. + +- You can retrieve same VmaAllocationInfo structure while creating your resource, from function + vmaCreateBuffer(), vmaCreateImage(). You can remember it if you are sure parameters don't change + (e.g. due to defragmentation or allocation becoming lost). +- If you just want to check if allocation is not lost, vmaTouchAllocation() will work faster. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocationInfo( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + VmaAllocationInfo* VMA_NOT_NULL pAllocationInfo); + +/** \brief Returns `VK_TRUE` if allocation is not lost and atomically marks it as used in current frame. + +If the allocation has been created with #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag, +this function returns `VK_TRUE` if it's not in lost state, so it can still be used. +It then also atomically "touches" the allocation - marks it as used in current frame, +so that you can be sure it won't become lost in current frame or next `frameInUseCount` frames. + +If the allocation is in lost state, the function returns `VK_FALSE`. +Memory of such allocation, as well as buffer or image bound to it, should not be used. +Lost allocation and the buffer/image still need to be destroyed. + +If the allocation has been created without #VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag, +this function always returns `VK_TRUE`. +*/ +VMA_CALL_PRE VkBool32 VMA_CALL_POST vmaTouchAllocation( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation); + +/** \brief Sets pUserData in given allocation to new value. + +If the allocation was created with VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT, +pUserData must be either null, or pointer to a null-terminated string. The function +makes local copy of the string and sets it as allocation's `pUserData`. String +passed as pUserData doesn't need to be valid for whole lifetime of the allocation - +you can free it after this call. String previously pointed by allocation's +pUserData is freed from memory. + +If the flag was not used, the value of pointer `pUserData` is just copied to +allocation's `pUserData`. It is opaque, so you can use it however you want - e.g. +as a pointer, ordinal number or some handle to you own data. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaSetAllocationUserData( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + void* VMA_NULLABLE pUserData); + +/** \brief Creates new allocation that is in lost state from the beginning. + +It can be useful if you need a dummy, non-null allocation. + +You still need to destroy created object using vmaFreeMemory(). + +Returned allocation is not tied to any specific memory pool or memory type and +not bound to any image or buffer. It has size = 0. It cannot be turned into +a real, non-empty allocation. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaCreateLostAllocation( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation); + +/** \brief Maps memory represented by given allocation and returns pointer to it. + +Maps memory represented by given allocation to make it accessible to CPU code. +When succeeded, `*ppData` contains pointer to first byte of this memory. +If the allocation is part of bigger `VkDeviceMemory` block, the pointer is +correctly offseted to the beginning of region assigned to this particular +allocation. + +Mapping is internally reference-counted and synchronized, so despite raw Vulkan +function `vkMapMemory()` cannot be used to map same block of `VkDeviceMemory` +multiple times simultaneously, it is safe to call this function on allocations +assigned to the same memory block. Actual Vulkan memory will be mapped on first +mapping and unmapped on last unmapping. + +If the function succeeded, you must call vmaUnmapMemory() to unmap the +allocation when mapping is no longer needed or before freeing the allocation, at +the latest. + +It also safe to call this function multiple times on the same allocation. You +must call vmaUnmapMemory() same number of times as you called vmaMapMemory(). + +It is also safe to call this function on allocation created with +#VMA_ALLOCATION_CREATE_MAPPED_BIT flag. Its memory stays mapped all the time. +You must still call vmaUnmapMemory() same number of times as you called +vmaMapMemory(). You must not call vmaUnmapMemory() additional time to free the +"0-th" mapping made automatically due to #VMA_ALLOCATION_CREATE_MAPPED_BIT flag. + +This function fails when used on allocation made in memory type that is not +`HOST_VISIBLE`. + +This function always fails when called for allocation that was created with +#VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT flag. Such allocations cannot be +mapped. + +This function doesn't automatically flush or invalidate caches. +If the allocation is made from a memory types that is not `HOST_COHERENT`, +you also need to use vmaInvalidateAllocation() / vmaFlushAllocation(), as required by Vulkan specification. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaMapMemory( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + void* VMA_NULLABLE * VMA_NOT_NULL ppData); + +/** \brief Unmaps memory represented by given allocation, mapped previously using vmaMapMemory(). + +For details, see description of vmaMapMemory(). + +This function doesn't automatically flush or invalidate caches. +If the allocation is made from a memory types that is not `HOST_COHERENT`, +you also need to use vmaInvalidateAllocation() / vmaFlushAllocation(), as required by Vulkan specification. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaUnmapMemory( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation); + +/** \brief Flushes memory of given allocation. + +Calls `vkFlushMappedMemoryRanges()` for memory associated with given range of given allocation. +It needs to be called after writing to a mapped memory for memory types that are not `HOST_COHERENT`. +Unmap operation doesn't do that automatically. + +- `offset` must be relative to the beginning of allocation. +- `size` can be `VK_WHOLE_SIZE`. It means all memory from `offset` the the end of given allocation. +- `offset` and `size` don't have to be aligned. + They are internally rounded down/up to multiply of `nonCoherentAtomSize`. +- If `size` is 0, this call is ignored. +- If memory type that the `allocation` belongs to is not `HOST_VISIBLE` or it is `HOST_COHERENT`, + this call is ignored. + +Warning! `offset` and `size` are relative to the contents of given `allocation`. +If you mean whole allocation, you can pass 0 and `VK_WHOLE_SIZE`, respectively. +Do not pass allocation's offset as `offset`!!! + +This function returns the `VkResult` from `vkFlushMappedMemoryRanges` if it is +called, otherwise `VK_SUCCESS`. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFlushAllocation( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + VkDeviceSize offset, + VkDeviceSize size); + +/** \brief Invalidates memory of given allocation. + +Calls `vkInvalidateMappedMemoryRanges()` for memory associated with given range of given allocation. +It needs to be called before reading from a mapped memory for memory types that are not `HOST_COHERENT`. +Map operation doesn't do that automatically. + +- `offset` must be relative to the beginning of allocation. +- `size` can be `VK_WHOLE_SIZE`. It means all memory from `offset` the the end of given allocation. +- `offset` and `size` don't have to be aligned. + They are internally rounded down/up to multiply of `nonCoherentAtomSize`. +- If `size` is 0, this call is ignored. +- If memory type that the `allocation` belongs to is not `HOST_VISIBLE` or it is `HOST_COHERENT`, + this call is ignored. + +Warning! `offset` and `size` are relative to the contents of given `allocation`. +If you mean whole allocation, you can pass 0 and `VK_WHOLE_SIZE`, respectively. +Do not pass allocation's offset as `offset`!!! + +This function returns the `VkResult` from `vkInvalidateMappedMemoryRanges` if +it is called, otherwise `VK_SUCCESS`. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaInvalidateAllocation( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + VkDeviceSize offset, + VkDeviceSize size); + +/** \brief Flushes memory of given set of allocations. + +Calls `vkFlushMappedMemoryRanges()` for memory associated with given ranges of given allocations. +For more information, see documentation of vmaFlushAllocation(). + +\param allocator +\param allocationCount +\param allocations +\param offsets If not null, it must point to an array of offsets of regions to flush, relative to the beginning of respective allocations. Null means all ofsets are zero. +\param sizes If not null, it must point to an array of sizes of regions to flush in respective allocations. Null means `VK_WHOLE_SIZE` for all allocations. + +This function returns the `VkResult` from `vkFlushMappedMemoryRanges` if it is +called, otherwise `VK_SUCCESS`. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFlushAllocations( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t allocationCount, + const VmaAllocation VMA_NOT_NULL * VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) allocations, + const VkDeviceSize* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) offsets, + const VkDeviceSize* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) sizes); + +/** \brief Invalidates memory of given set of allocations. + +Calls `vkInvalidateMappedMemoryRanges()` for memory associated with given ranges of given allocations. +For more information, see documentation of vmaInvalidateAllocation(). + +\param allocator +\param allocationCount +\param allocations +\param offsets If not null, it must point to an array of offsets of regions to flush, relative to the beginning of respective allocations. Null means all ofsets are zero. +\param sizes If not null, it must point to an array of sizes of regions to flush in respective allocations. Null means `VK_WHOLE_SIZE` for all allocations. + +This function returns the `VkResult` from `vkInvalidateMappedMemoryRanges` if it is +called, otherwise `VK_SUCCESS`. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaInvalidateAllocations( + VmaAllocator VMA_NOT_NULL allocator, + uint32_t allocationCount, + const VmaAllocation VMA_NOT_NULL * VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) allocations, + const VkDeviceSize* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) offsets, + const VkDeviceSize* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) sizes); + +/** \brief Checks magic number in margins around all allocations in given memory types (in both default and custom pools) in search for corruptions. + +@param memoryTypeBits Bit mask, where each bit set means that a memory type with that index should be checked. + +Corruption detection is enabled only when `VMA_DEBUG_DETECT_CORRUPTION` macro is defined to nonzero, +`VMA_DEBUG_MARGIN` is defined to nonzero and only for memory types that are +`HOST_VISIBLE` and `HOST_COHERENT`. For more information, see [Corruption detection](@ref debugging_memory_usage_corruption_detection). + +Possible return values: + +- `VK_ERROR_FEATURE_NOT_PRESENT` - corruption detection is not enabled for any of specified memory types. +- `VK_SUCCESS` - corruption detection has been performed and succeeded. +- `VK_ERROR_VALIDATION_FAILED_EXT` - corruption detection has been performed and found memory corruptions around one of the allocations. + `VMA_ASSERT` is also fired in that case. +- Other value: Error returned by Vulkan, e.g. memory mapping failure. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckCorruption(VmaAllocator VMA_NOT_NULL allocator, uint32_t memoryTypeBits); + +/** \struct VmaDefragmentationContext +\brief Represents Opaque object that represents started defragmentation process. + +Fill structure #VmaDefragmentationInfo2 and call function vmaDefragmentationBegin() to create it. +Call function vmaDefragmentationEnd() to destroy it. +*/ +VK_DEFINE_HANDLE(VmaDefragmentationContext) + +/// Flags to be used in vmaDefragmentationBegin(). None at the moment. Reserved for future use. +typedef enum VmaDefragmentationFlagBits { + VMA_DEFRAGMENTATION_FLAG_INCREMENTAL = 0x1, + VMA_DEFRAGMENTATION_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF +} VmaDefragmentationFlagBits; +typedef VkFlags VmaDefragmentationFlags; + +/** \brief Parameters for defragmentation. + +To be used with function vmaDefragmentationBegin(). +*/ +typedef struct VmaDefragmentationInfo2 { + /** \brief Reserved for future use. Should be 0. + */ + VmaDefragmentationFlags flags; + /** \brief Number of allocations in `pAllocations` array. + */ + uint32_t allocationCount; + /** \brief Pointer to array of allocations that can be defragmented. + + The array should have `allocationCount` elements. + The array should not contain nulls. + Elements in the array should be unique - same allocation cannot occur twice. + It is safe to pass allocations that are in the lost state - they are ignored. + All allocations not present in this array are considered non-moveable during this defragmentation. + */ + const VmaAllocation VMA_NOT_NULL * VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations; + /** \brief Optional, output. Pointer to array that will be filled with information whether the allocation at certain index has been changed during defragmentation. + + The array should have `allocationCount` elements. + You can pass null if you are not interested in this information. + */ + VkBool32* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocationsChanged; + /** \brief Numer of pools in `pPools` array. + */ + uint32_t poolCount; + /** \brief Either null or pointer to array of pools to be defragmented. + + All the allocations in the specified pools can be moved during defragmentation + and there is no way to check if they were really moved as in `pAllocationsChanged`, + so you must query all the allocations in all these pools for new `VkDeviceMemory` + and offset using vmaGetAllocationInfo() if you might need to recreate buffers + and images bound to them. + + The array should have `poolCount` elements. + The array should not contain nulls. + Elements in the array should be unique - same pool cannot occur twice. + + Using this array is equivalent to specifying all allocations from the pools in `pAllocations`. + It might be more efficient. + */ + const VmaPool VMA_NOT_NULL * VMA_NULLABLE VMA_LEN_IF_NOT_NULL(poolCount) pPools; + /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places using transfers on CPU side, like `memcpy()`, `memmove()`. + + `VK_WHOLE_SIZE` means no limit. + */ + VkDeviceSize maxCpuBytesToMove; + /** \brief Maximum number of allocations that can be moved to a different place using transfers on CPU side, like `memcpy()`, `memmove()`. + + `UINT32_MAX` means no limit. + */ + uint32_t maxCpuAllocationsToMove; + /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places using transfers on GPU side, posted to `commandBuffer`. + + `VK_WHOLE_SIZE` means no limit. + */ + VkDeviceSize maxGpuBytesToMove; + /** \brief Maximum number of allocations that can be moved to a different place using transfers on GPU side, posted to `commandBuffer`. + + `UINT32_MAX` means no limit. + */ + uint32_t maxGpuAllocationsToMove; + /** \brief Optional. Command buffer where GPU copy commands will be posted. + + If not null, it must be a valid command buffer handle that supports Transfer queue type. + It must be in the recording state and outside of a render pass instance. + You need to submit it and make sure it finished execution before calling vmaDefragmentationEnd(). + + Passing null means that only CPU defragmentation will be performed. + */ + VkCommandBuffer VMA_NULLABLE commandBuffer; +} VmaDefragmentationInfo2; + +typedef struct VmaDefragmentationPassMoveInfo { + VmaAllocation VMA_NOT_NULL allocation; + VkDeviceMemory VMA_NOT_NULL_NON_DISPATCHABLE memory; + VkDeviceSize offset; +} VmaDefragmentationPassMoveInfo; + +/** \brief Parameters for incremental defragmentation steps. + +To be used with function vmaBeginDefragmentationPass(). +*/ +typedef struct VmaDefragmentationPassInfo { + uint32_t moveCount; + VmaDefragmentationPassMoveInfo* VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(moveCount) pMoves; +} VmaDefragmentationPassInfo; + +/** \brief Deprecated. Optional configuration parameters to be passed to function vmaDefragment(). + +\deprecated This is a part of the old interface. It is recommended to use structure #VmaDefragmentationInfo2 and function vmaDefragmentationBegin() instead. +*/ +typedef struct VmaDefragmentationInfo { + /** \brief Maximum total numbers of bytes that can be copied while moving allocations to different places. + + Default is `VK_WHOLE_SIZE`, which means no limit. + */ + VkDeviceSize maxBytesToMove; + /** \brief Maximum number of allocations that can be moved to different place. + + Default is `UINT32_MAX`, which means no limit. + */ + uint32_t maxAllocationsToMove; +} VmaDefragmentationInfo; + +/** \brief Statistics returned by function vmaDefragment(). */ +typedef struct VmaDefragmentationStats { + /// Total number of bytes that have been copied while moving allocations to different places. + VkDeviceSize bytesMoved; + /// Total number of bytes that have been released to the system by freeing empty `VkDeviceMemory` objects. + VkDeviceSize bytesFreed; + /// Number of allocations that have been moved to different places. + uint32_t allocationsMoved; + /// Number of empty `VkDeviceMemory` objects that have been released to the system. + uint32_t deviceMemoryBlocksFreed; +} VmaDefragmentationStats; + +/** \brief Begins defragmentation process. + +@param allocator Allocator object. +@param pInfo Structure filled with parameters of defragmentation. +@param[out] pStats Optional. Statistics of defragmentation. You can pass null if you are not interested in this information. +@param[out] pContext Context object that must be passed to vmaDefragmentationEnd() to finish defragmentation. +@return `VK_SUCCESS` and `*pContext == null` if defragmentation finished within this function call. `VK_NOT_READY` and `*pContext != null` if defragmentation has been started and you need to call vmaDefragmentationEnd() to finish it. Negative value in case of error. + +Use this function instead of old, deprecated vmaDefragment(). + +Warning! Between the call to vmaDefragmentationBegin() and vmaDefragmentationEnd(): + +- You should not use any of allocations passed as `pInfo->pAllocations` or + any allocations that belong to pools passed as `pInfo->pPools`, + including calling vmaGetAllocationInfo(), vmaTouchAllocation(), or access + their data. +- Some mutexes protecting internal data structures may be locked, so trying to + make or free any allocations, bind buffers or images, map memory, or launch + another simultaneous defragmentation in between may cause stall (when done on + another thread) or deadlock (when done on the same thread), unless you are + 100% sure that defragmented allocations are in different pools. +- Information returned via `pStats` and `pInfo->pAllocationsChanged` are undefined. + They become valid after call to vmaDefragmentationEnd(). +- If `pInfo->commandBuffer` is not null, you must submit that command buffer + and make sure it finished execution before calling vmaDefragmentationEnd(). + +For more information and important limitations regarding defragmentation, see documentation chapter: +[Defragmentation](@ref defragmentation). +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragmentationBegin( + VmaAllocator VMA_NOT_NULL allocator, + const VmaDefragmentationInfo2* VMA_NOT_NULL pInfo, + VmaDefragmentationStats* VMA_NULLABLE pStats, + VmaDefragmentationContext VMA_NULLABLE * VMA_NOT_NULL pContext); + +/** \brief Ends defragmentation process. + +Use this function to finish defragmentation started by vmaDefragmentationBegin(). +It is safe to pass `context == null`. The function then does nothing. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragmentationEnd( + VmaAllocator VMA_NOT_NULL allocator, + VmaDefragmentationContext VMA_NULLABLE context); + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBeginDefragmentationPass( + VmaAllocator VMA_NOT_NULL allocator, + VmaDefragmentationContext VMA_NULLABLE context, + VmaDefragmentationPassInfo* VMA_NOT_NULL pInfo +); +VMA_CALL_PRE VkResult VMA_CALL_POST vmaEndDefragmentationPass( + VmaAllocator VMA_NOT_NULL allocator, + VmaDefragmentationContext VMA_NULLABLE context +); + +/** \brief Deprecated. Compacts memory by moving allocations. + +@param pAllocations Array of allocations that can be moved during this compation. +@param allocationCount Number of elements in pAllocations and pAllocationsChanged arrays. +@param[out] pAllocationsChanged Array of boolean values that will indicate whether matching allocation in pAllocations array has been moved. This parameter is optional. Pass null if you don't need this information. +@param pDefragmentationInfo Configuration parameters. Optional - pass null to use default values. +@param[out] pDefragmentationStats Statistics returned by the function. Optional - pass null if you don't need this information. +@return `VK_SUCCESS` if completed, negative error code in case of error. + +\deprecated This is a part of the old interface. It is recommended to use structure #VmaDefragmentationInfo2 and function vmaDefragmentationBegin() instead. + +This function works by moving allocations to different places (different +`VkDeviceMemory` objects and/or different offsets) in order to optimize memory +usage. Only allocations that are in `pAllocations` array can be moved. All other +allocations are considered nonmovable in this call. Basic rules: + +- Only allocations made in memory types that have + `VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT` and `VK_MEMORY_PROPERTY_HOST_COHERENT_BIT` + flags can be compacted. You may pass other allocations but it makes no sense - + these will never be moved. +- Custom pools created with #VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT or + #VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT flag are not defragmented. Allocations + passed to this function that come from such pools are ignored. +- Allocations created with #VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT or + created as dedicated allocations for any other reason are also ignored. +- Both allocations made with or without #VMA_ALLOCATION_CREATE_MAPPED_BIT + flag can be compacted. If not persistently mapped, memory will be mapped + temporarily inside this function if needed. +- You must not pass same #VmaAllocation object multiple times in `pAllocations` array. + +The function also frees empty `VkDeviceMemory` blocks. + +Warning: This function may be time-consuming, so you shouldn't call it too often +(like after every resource creation/destruction). +You can call it on special occasions (like when reloading a game level or +when you just destroyed a lot of objects). Calling it every frame may be OK, but +you should measure that on your platform. + +For more information, see [Defragmentation](@ref defragmentation) chapter. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragment( + VmaAllocator VMA_NOT_NULL allocator, + const VmaAllocation VMA_NOT_NULL * VMA_NOT_NULL VMA_LEN_IF_NOT_NULL(allocationCount) pAllocations, + size_t allocationCount, + VkBool32* VMA_NULLABLE VMA_LEN_IF_NOT_NULL(allocationCount) pAllocationsChanged, + const VmaDefragmentationInfo* VMA_NULLABLE pDefragmentationInfo, + VmaDefragmentationStats* VMA_NULLABLE pDefragmentationStats); + +/** \brief Binds buffer to allocation. + +Binds specified buffer to region of memory represented by specified allocation. +Gets `VkDeviceMemory` handle and offset from the allocation. +If you want to create a buffer, allocate memory for it and bind them together separately, +you should use this function for binding instead of standard `vkBindBufferMemory()`, +because it ensures proper synchronization so that when a `VkDeviceMemory` object is used by multiple +allocations, calls to `vkBind*Memory()` or `vkMapMemory()` won't happen from multiple threads simultaneously +(which is illegal in Vulkan). + +It is recommended to use function vmaCreateBuffer() instead of this one. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindBufferMemory( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + VkBuffer VMA_NOT_NULL_NON_DISPATCHABLE buffer); + +/** \brief Binds buffer to allocation with additional parameters. + +@param allocationLocalOffset Additional offset to be added while binding, relative to the beginnig of the `allocation`. Normally it should be 0. +@param pNext A chain of structures to be attached to `VkBindBufferMemoryInfoKHR` structure used internally. Normally it should be null. + +This function is similar to vmaBindBufferMemory(), but it provides additional parameters. + +If `pNext` is not null, #VmaAllocator object must have been created with #VMA_ALLOCATOR_CREATE_KHR_BIND_MEMORY2_BIT flag +or with VmaAllocatorCreateInfo::vulkanApiVersion `>= VK_API_VERSION_1_1`. Otherwise the call fails. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindBufferMemory2( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + VkDeviceSize allocationLocalOffset, + VkBuffer VMA_NOT_NULL_NON_DISPATCHABLE buffer, + const void* VMA_NULLABLE pNext); + +/** \brief Binds image to allocation. + +Binds specified image to region of memory represented by specified allocation. +Gets `VkDeviceMemory` handle and offset from the allocation. +If you want to create an image, allocate memory for it and bind them together separately, +you should use this function for binding instead of standard `vkBindImageMemory()`, +because it ensures proper synchronization so that when a `VkDeviceMemory` object is used by multiple +allocations, calls to `vkBind*Memory()` or `vkMapMemory()` won't happen from multiple threads simultaneously +(which is illegal in Vulkan). + +It is recommended to use function vmaCreateImage() instead of this one. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindImageMemory( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + VkImage VMA_NOT_NULL_NON_DISPATCHABLE image); + +/** \brief Binds image to allocation with additional parameters. + +@param allocationLocalOffset Additional offset to be added while binding, relative to the beginnig of the `allocation`. Normally it should be 0. +@param pNext A chain of structures to be attached to `VkBindImageMemoryInfoKHR` structure used internally. Normally it should be null. + +This function is similar to vmaBindImageMemory(), but it provides additional parameters. + +If `pNext` is not null, #VmaAllocator object must have been created with #VMA_ALLOCATOR_CREATE_KHR_BIND_MEMORY2_BIT flag +or with VmaAllocatorCreateInfo::vulkanApiVersion `>= VK_API_VERSION_1_1`. Otherwise the call fails. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindImageMemory2( + VmaAllocator VMA_NOT_NULL allocator, + VmaAllocation VMA_NOT_NULL allocation, + VkDeviceSize allocationLocalOffset, + VkImage VMA_NOT_NULL_NON_DISPATCHABLE image, + const void* VMA_NULLABLE pNext); + +/** +@param[out] pBuffer Buffer that was created. +@param[out] pAllocation Allocation that was created. +@param[out] pAllocationInfo Optional. Information about allocated memory. It can be later fetched using function vmaGetAllocationInfo(). + +This function automatically: + +-# Creates buffer. +-# Allocates appropriate memory for it. +-# Binds the buffer with the memory. + +If any of these operations fail, buffer and allocation are not created, +returned value is negative error code, *pBuffer and *pAllocation are null. + +If the function succeeded, you must destroy both buffer and allocation when you +no longer need them using either convenience function vmaDestroyBuffer() or +separately, using `vkDestroyBuffer()` and vmaFreeMemory(). + +If #VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT flag was used, +VK_KHR_dedicated_allocation extension is used internally to query driver whether +it requires or prefers the new buffer to have dedicated allocation. If yes, +and if dedicated allocation is possible (VmaAllocationCreateInfo::pool is null +and #VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT is not used), it creates dedicated +allocation for this buffer, just like when using +#VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT. + +\note This function creates a new `VkBuffer`. Sub-allocation of parts of one large buffer, +although recommended as a good practice, is out of scope of this library and could be implemented +by the user as a higher-level logic on top of VMA. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateBuffer( + VmaAllocator VMA_NOT_NULL allocator, + const VkBufferCreateInfo* VMA_NOT_NULL pBufferCreateInfo, + const VmaAllocationCreateInfo* VMA_NOT_NULL pAllocationCreateInfo, + VkBuffer VMA_NULLABLE_NON_DISPATCHABLE * VMA_NOT_NULL pBuffer, + VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); + +/** \brief Destroys Vulkan buffer and frees allocated memory. + +This is just a convenience function equivalent to: + +\code +vkDestroyBuffer(device, buffer, allocationCallbacks); +vmaFreeMemory(allocator, allocation); +\endcode + +It it safe to pass null as buffer and/or allocation. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyBuffer( + VmaAllocator VMA_NOT_NULL allocator, + VkBuffer VMA_NULLABLE_NON_DISPATCHABLE buffer, + VmaAllocation VMA_NULLABLE allocation); + +/// Function similar to vmaCreateBuffer(). +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateImage( + VmaAllocator VMA_NOT_NULL allocator, + const VkImageCreateInfo* VMA_NOT_NULL pImageCreateInfo, + const VmaAllocationCreateInfo* VMA_NOT_NULL pAllocationCreateInfo, + VkImage VMA_NULLABLE_NON_DISPATCHABLE * VMA_NOT_NULL pImage, + VmaAllocation VMA_NULLABLE * VMA_NOT_NULL pAllocation, + VmaAllocationInfo* VMA_NULLABLE pAllocationInfo); + +/** \brief Destroys Vulkan image and frees allocated memory. + +This is just a convenience function equivalent to: + +\code +vkDestroyImage(device, image, allocationCallbacks); +vmaFreeMemory(allocator, allocation); +\endcode + +It it safe to pass null as image and/or allocation. +*/ +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyImage( + VmaAllocator VMA_NOT_NULL allocator, + VkImage VMA_NULLABLE_NON_DISPATCHABLE image, + VmaAllocation VMA_NULLABLE allocation); + +#ifdef __cplusplus +} +#endif + +#endif // AMD_VULKAN_MEMORY_ALLOCATOR_H + +// For Visual Studio IntelliSense. +#if defined(__cplusplus) && defined(__INTELLISENSE__) +#define VMA_IMPLEMENTATION +#endif + +#ifdef VMA_IMPLEMENTATION +#undef VMA_IMPLEMENTATION + +#include +#include +#include +#include + +#if VMA_RECORDING_ENABLED + #include + #if defined(_WIN32) + #include + #else + #include + #include + #endif +#endif + +/******************************************************************************* +CONFIGURATION SECTION + +Define some of these macros before each #include of this header or change them +here if you need other then default behavior depending on your environment. +*/ + +/* +Define this macro to 1 to make the library fetch pointers to Vulkan functions +internally, like: + + vulkanFunctions.vkAllocateMemory = &vkAllocateMemory; +*/ +#if !defined(VMA_STATIC_VULKAN_FUNCTIONS) && !defined(VK_NO_PROTOTYPES) + #define VMA_STATIC_VULKAN_FUNCTIONS 1 +#endif + +/* +Define this macro to 1 to make the library fetch pointers to Vulkan functions +internally, like: + + vulkanFunctions.vkAllocateMemory = (PFN_vkAllocateMemory)vkGetDeviceProcAddr(m_hDevice, vkAllocateMemory); +*/ +#if !defined(VMA_DYNAMIC_VULKAN_FUNCTIONS) + #define VMA_DYNAMIC_VULKAN_FUNCTIONS 1 + #if defined(VK_NO_PROTOTYPES) + extern PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr; + extern PFN_vkGetDeviceProcAddr vkGetDeviceProcAddr; + #endif +#endif + +// Define this macro to 1 to make the library use STL containers instead of its own implementation. +//#define VMA_USE_STL_CONTAINERS 1 + +/* Set this macro to 1 to make the library including and using STL containers: +std::pair, std::vector, std::list, std::unordered_map. + +Set it to 0 or undefined to make the library using its own implementation of +the containers. +*/ +#if VMA_USE_STL_CONTAINERS + #define VMA_USE_STL_VECTOR 1 + #define VMA_USE_STL_UNORDERED_MAP 1 + #define VMA_USE_STL_LIST 1 +#endif + +#ifndef VMA_USE_STL_SHARED_MUTEX + // Compiler conforms to C++17. + #if __cplusplus >= 201703L + #define VMA_USE_STL_SHARED_MUTEX 1 + // Visual studio defines __cplusplus properly only when passed additional parameter: /Zc:__cplusplus + // Otherwise it's always 199711L, despite shared_mutex works since Visual Studio 2015 Update 2. + // See: https://blogs.msdn.microsoft.com/vcblog/2018/04/09/msvc-now-correctly-reports-__cplusplus/ + #elif defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 190023918 && __cplusplus == 199711L && _MSVC_LANG >= 201703L + #define VMA_USE_STL_SHARED_MUTEX 1 + #else + #define VMA_USE_STL_SHARED_MUTEX 0 + #endif +#endif + +/* +THESE INCLUDES ARE NOT ENABLED BY DEFAULT. +Library has its own container implementation. +*/ +#if VMA_USE_STL_VECTOR + #include +#endif + +#if VMA_USE_STL_UNORDERED_MAP + #include +#endif + +#if VMA_USE_STL_LIST + #include +#endif + +/* +Following headers are used in this CONFIGURATION section only, so feel free to +remove them if not needed. +*/ +#include // for assert +#include // for min, max +#include + +#ifndef VMA_NULL + // Value used as null pointer. Define it to e.g.: nullptr, NULL, 0, (void*)0. + #define VMA_NULL nullptr +#endif + +#if defined(__ANDROID_API__) && (__ANDROID_API__ < 16) +#include +static void* vma_aligned_alloc(size_t alignment, size_t size) +{ + // alignment must be >= sizeof(void*) + if(alignment < sizeof(void*)) + { + alignment = sizeof(void*); + } + + return memalign(alignment, size); +} +#elif defined(__APPLE__) || defined(__ANDROID__) || (defined(__linux__) && defined(__GLIBCXX__) && !defined(_GLIBCXX_HAVE_ALIGNED_ALLOC)) +#include + +#if defined(__APPLE__) +#include +#endif + +static void* vma_aligned_alloc(size_t alignment, size_t size) +{ +#if defined(__APPLE__) && (defined(MAC_OS_X_VERSION_10_16) || defined(__IPHONE_14_0)) +#if MAC_OS_X_VERSION_MAX_ALLOWED >= MAC_OS_X_VERSION_10_16 || __IPHONE_OS_VERSION_MAX_ALLOWED >= __IPHONE_14_0 + // For C++14, usr/include/malloc/_malloc.h declares aligned_alloc()) only + // with the MacOSX11.0 SDK in Xcode 12 (which is what adds + // MAC_OS_X_VERSION_10_16), even though the function is marked + // availabe for 10.15. That's why the preprocessor checks for 10.16 but + // the __builtin_available checks for 10.15. + // People who use C++17 could call aligned_alloc with the 10.15 SDK already. + if (__builtin_available(macOS 10.15, iOS 13, *)) + return aligned_alloc(alignment, size); +#endif +#endif + // alignment must be >= sizeof(void*) + if(alignment < sizeof(void*)) + { + alignment = sizeof(void*); + } + + void *pointer; + if(posix_memalign(&pointer, alignment, size) == 0) + return pointer; + return VMA_NULL; +} +#elif defined(_WIN32) +static void* vma_aligned_alloc(size_t alignment, size_t size) +{ + return _aligned_malloc(size, alignment); +} +#else +static void* vma_aligned_alloc(size_t alignment, size_t size) +{ + return aligned_alloc(alignment, size); +} +#endif + +#if defined(_WIN32) +static void vma_aligned_free(void* ptr) +{ + _aligned_free(ptr); +} +#else +static void vma_aligned_free(void* ptr) +{ + free(ptr); +} +#endif + +// If your compiler is not compatible with C++11 and definition of +// aligned_alloc() function is missing, uncommeting following line may help: + +//#include + +// Normal assert to check for programmer's errors, especially in Debug configuration. +#ifndef VMA_ASSERT + #ifdef NDEBUG + #define VMA_ASSERT(expr) + #else + #define VMA_ASSERT(expr) assert(expr) + #endif +#endif + +// Assert that will be called very often, like inside data structures e.g. operator[]. +// Making it non-empty can make program slow. +#ifndef VMA_HEAVY_ASSERT + #ifdef NDEBUG + #define VMA_HEAVY_ASSERT(expr) + #else + #define VMA_HEAVY_ASSERT(expr) //VMA_ASSERT(expr) + #endif +#endif + +#ifndef VMA_ALIGN_OF + #define VMA_ALIGN_OF(type) (__alignof(type)) +#endif + +#ifndef VMA_SYSTEM_ALIGNED_MALLOC + #define VMA_SYSTEM_ALIGNED_MALLOC(size, alignment) vma_aligned_alloc((alignment), (size)) +#endif + +#ifndef VMA_SYSTEM_ALIGNED_FREE + // VMA_SYSTEM_FREE is the old name, but might have been defined by the user + #if defined(VMA_SYSTEM_FREE) + #define VMA_SYSTEM_ALIGNED_FREE(ptr) VMA_SYSTEM_FREE(ptr) + #else + #define VMA_SYSTEM_ALIGNED_FREE(ptr) vma_aligned_free(ptr) + #endif +#endif + +#ifndef VMA_MIN + #define VMA_MIN(v1, v2) (std::min((v1), (v2))) +#endif + +#ifndef VMA_MAX + #define VMA_MAX(v1, v2) (std::max((v1), (v2))) +#endif + +#ifndef VMA_SWAP + #define VMA_SWAP(v1, v2) std::swap((v1), (v2)) +#endif + +#ifndef VMA_SORT + #define VMA_SORT(beg, end, cmp) std::sort(beg, end, cmp) +#endif + +#ifndef VMA_DEBUG_LOG + #define VMA_DEBUG_LOG(format, ...) + /* + #define VMA_DEBUG_LOG(format, ...) do { \ + printf(format, __VA_ARGS__); \ + printf("\n"); \ + } while(false) + */ +#endif + +// Define this macro to 1 to enable functions: vmaBuildStatsString, vmaFreeStatsString. +#if VMA_STATS_STRING_ENABLED + static inline void VmaUint32ToStr(char* outStr, size_t strLen, uint32_t num) + { + snprintf(outStr, strLen, "%u", static_cast(num)); + } + static inline void VmaUint64ToStr(char* outStr, size_t strLen, uint64_t num) + { + snprintf(outStr, strLen, "%llu", static_cast(num)); + } + static inline void VmaPtrToStr(char* outStr, size_t strLen, const void* ptr) + { + snprintf(outStr, strLen, "%p", ptr); + } +#endif + +#ifndef VMA_MUTEX + class VmaMutex + { + public: + void Lock() { m_Mutex.lock(); } + void Unlock() { m_Mutex.unlock(); } + bool TryLock() { return m_Mutex.try_lock(); } + private: + std::mutex m_Mutex; + }; + #define VMA_MUTEX VmaMutex +#endif + +// Read-write mutex, where "read" is shared access, "write" is exclusive access. +#ifndef VMA_RW_MUTEX + #if VMA_USE_STL_SHARED_MUTEX + // Use std::shared_mutex from C++17. + #include + class VmaRWMutex + { + public: + void LockRead() { m_Mutex.lock_shared(); } + void UnlockRead() { m_Mutex.unlock_shared(); } + bool TryLockRead() { return m_Mutex.try_lock_shared(); } + void LockWrite() { m_Mutex.lock(); } + void UnlockWrite() { m_Mutex.unlock(); } + bool TryLockWrite() { return m_Mutex.try_lock(); } + private: + std::shared_mutex m_Mutex; + }; + #define VMA_RW_MUTEX VmaRWMutex + #elif defined(_WIN32) && defined(WINVER) && WINVER >= 0x0600 + // Use SRWLOCK from WinAPI. + // Minimum supported client = Windows Vista, server = Windows Server 2008. + class VmaRWMutex + { + public: + VmaRWMutex() { InitializeSRWLock(&m_Lock); } + void LockRead() { AcquireSRWLockShared(&m_Lock); } + void UnlockRead() { ReleaseSRWLockShared(&m_Lock); } + bool TryLockRead() { return TryAcquireSRWLockShared(&m_Lock) != FALSE; } + void LockWrite() { AcquireSRWLockExclusive(&m_Lock); } + void UnlockWrite() { ReleaseSRWLockExclusive(&m_Lock); } + bool TryLockWrite() { return TryAcquireSRWLockExclusive(&m_Lock) != FALSE; } + private: + SRWLOCK m_Lock; + }; + #define VMA_RW_MUTEX VmaRWMutex + #else + // Less efficient fallback: Use normal mutex. + class VmaRWMutex + { + public: + void LockRead() { m_Mutex.Lock(); } + void UnlockRead() { m_Mutex.Unlock(); } + bool TryLockRead() { return m_Mutex.TryLock(); } + void LockWrite() { m_Mutex.Lock(); } + void UnlockWrite() { m_Mutex.Unlock(); } + bool TryLockWrite() { return m_Mutex.TryLock(); } + private: + VMA_MUTEX m_Mutex; + }; + #define VMA_RW_MUTEX VmaRWMutex + #endif // #if VMA_USE_STL_SHARED_MUTEX +#endif // #ifndef VMA_RW_MUTEX + +/* +If providing your own implementation, you need to implement a subset of std::atomic. +*/ +#ifndef VMA_ATOMIC_UINT32 + #include + #define VMA_ATOMIC_UINT32 std::atomic +#endif + +#ifndef VMA_ATOMIC_UINT64 + #include + #define VMA_ATOMIC_UINT64 std::atomic +#endif + +#ifndef VMA_DEBUG_ALWAYS_DEDICATED_MEMORY + /** + Every allocation will have its own memory block. + Define to 1 for debugging purposes only. + */ + #define VMA_DEBUG_ALWAYS_DEDICATED_MEMORY (0) +#endif + +#ifndef VMA_DEBUG_ALIGNMENT + /** + Minimum alignment of all allocations, in bytes. + Set to more than 1 for debugging purposes only. Must be power of two. + */ + #define VMA_DEBUG_ALIGNMENT (1) +#endif + +#ifndef VMA_DEBUG_MARGIN + /** + Minimum margin before and after every allocation, in bytes. + Set nonzero for debugging purposes only. + */ + #define VMA_DEBUG_MARGIN (0) +#endif + +#ifndef VMA_DEBUG_INITIALIZE_ALLOCATIONS + /** + Define this macro to 1 to automatically fill new allocations and destroyed + allocations with some bit pattern. + */ + #define VMA_DEBUG_INITIALIZE_ALLOCATIONS (0) +#endif + +#ifndef VMA_DEBUG_DETECT_CORRUPTION + /** + Define this macro to 1 together with non-zero value of VMA_DEBUG_MARGIN to + enable writing magic value to the margin before and after every allocation and + validating it, so that memory corruptions (out-of-bounds writes) are detected. + */ + #define VMA_DEBUG_DETECT_CORRUPTION (0) +#endif + +#ifndef VMA_DEBUG_GLOBAL_MUTEX + /** + Set this to 1 for debugging purposes only, to enable single mutex protecting all + entry calls to the library. Can be useful for debugging multithreading issues. + */ + #define VMA_DEBUG_GLOBAL_MUTEX (0) +#endif + +#ifndef VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY + /** + Minimum value for VkPhysicalDeviceLimits::bufferImageGranularity. + Set to more than 1 for debugging purposes only. Must be power of two. + */ + #define VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY (1) +#endif + +#ifndef VMA_SMALL_HEAP_MAX_SIZE + /// Maximum size of a memory heap in Vulkan to consider it "small". + #define VMA_SMALL_HEAP_MAX_SIZE (1024ull * 1024 * 1024) +#endif + +#ifndef VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE + /// Default size of a block allocated as single VkDeviceMemory from a "large" heap. + #define VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE (256ull * 1024 * 1024) +#endif + +#ifndef VMA_CLASS_NO_COPY + #define VMA_CLASS_NO_COPY(className) \ + private: \ + className(const className&) = delete; \ + className& operator=(const className&) = delete; +#endif + +static const uint32_t VMA_FRAME_INDEX_LOST = UINT32_MAX; + +// Decimal 2139416166, float NaN, little-endian binary 66 E6 84 7F. +static const uint32_t VMA_CORRUPTION_DETECTION_MAGIC_VALUE = 0x7F84E666; + +static const uint8_t VMA_ALLOCATION_FILL_PATTERN_CREATED = 0xDC; +static const uint8_t VMA_ALLOCATION_FILL_PATTERN_DESTROYED = 0xEF; + +/******************************************************************************* +END OF CONFIGURATION +*/ + +// # Copy of some Vulkan definitions so we don't need to check their existence just to handle few constants. + +static const uint32_t VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD_COPY = 0x00000040; +static const uint32_t VK_MEMORY_PROPERTY_DEVICE_UNCACHED_BIT_AMD_COPY = 0x00000080; +static const uint32_t VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT_COPY = 0x00020000; + +static const uint32_t VMA_ALLOCATION_INTERNAL_STRATEGY_MIN_OFFSET = 0x10000000u; + +static VkAllocationCallbacks VmaEmptyAllocationCallbacks = { + VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL, VMA_NULL }; + +// Returns number of bits set to 1 in (v). +static inline uint32_t VmaCountBitsSet(uint32_t v) +{ + uint32_t c = v - ((v >> 1) & 0x55555555); + c = ((c >> 2) & 0x33333333) + (c & 0x33333333); + c = ((c >> 4) + c) & 0x0F0F0F0F; + c = ((c >> 8) + c) & 0x00FF00FF; + c = ((c >> 16) + c) & 0x0000FFFF; + return c; +} + +/* +Returns true if given number is a power of two. +T must be unsigned integer number or signed integer but always nonnegative. +For 0 returns true. +*/ +template +inline bool VmaIsPow2(T x) +{ + return (x & (x-1)) == 0; +} + +// Aligns given value up to nearest multiply of align value. For example: VmaAlignUp(11, 8) = 16. +// Use types like uint32_t, uint64_t as T. +template +static inline T VmaAlignUp(T val, T alignment) +{ + VMA_HEAVY_ASSERT(VmaIsPow2(alignment)); + return (val + alignment - 1) & ~(alignment - 1); +} +// Aligns given value down to nearest multiply of align value. For example: VmaAlignUp(11, 8) = 8. +// Use types like uint32_t, uint64_t as T. +template +static inline T VmaAlignDown(T val, T alignment) +{ + VMA_HEAVY_ASSERT(VmaIsPow2(alignment)); + return val & ~(alignment - 1); +} + +// Division with mathematical rounding to nearest number. +template +static inline T VmaRoundDiv(T x, T y) +{ + return (x + (y / (T)2)) / y; +} + +// Returns smallest power of 2 greater or equal to v. +static inline uint32_t VmaNextPow2(uint32_t v) +{ + v--; + v |= v >> 1; + v |= v >> 2; + v |= v >> 4; + v |= v >> 8; + v |= v >> 16; + v++; + return v; +} +static inline uint64_t VmaNextPow2(uint64_t v) +{ + v--; + v |= v >> 1; + v |= v >> 2; + v |= v >> 4; + v |= v >> 8; + v |= v >> 16; + v |= v >> 32; + v++; + return v; +} + +// Returns largest power of 2 less or equal to v. +static inline uint32_t VmaPrevPow2(uint32_t v) +{ + v |= v >> 1; + v |= v >> 2; + v |= v >> 4; + v |= v >> 8; + v |= v >> 16; + v = v ^ (v >> 1); + return v; +} +static inline uint64_t VmaPrevPow2(uint64_t v) +{ + v |= v >> 1; + v |= v >> 2; + v |= v >> 4; + v |= v >> 8; + v |= v >> 16; + v |= v >> 32; + v = v ^ (v >> 1); + return v; +} + +static inline bool VmaStrIsEmpty(const char* pStr) +{ + return pStr == VMA_NULL || *pStr == '\0'; +} + +#if VMA_STATS_STRING_ENABLED + +static const char* VmaAlgorithmToStr(uint32_t algorithm) +{ + switch(algorithm) + { + case VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT: + return "Linear"; + case VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT: + return "Buddy"; + case 0: + return "Default"; + default: + VMA_ASSERT(0); + return ""; + } +} + +#endif // #if VMA_STATS_STRING_ENABLED + +#ifndef VMA_SORT + +template +Iterator VmaQuickSortPartition(Iterator beg, Iterator end, Compare cmp) +{ + Iterator centerValue = end; --centerValue; + Iterator insertIndex = beg; + for(Iterator memTypeIndex = beg; memTypeIndex < centerValue; ++memTypeIndex) + { + if(cmp(*memTypeIndex, *centerValue)) + { + if(insertIndex != memTypeIndex) + { + VMA_SWAP(*memTypeIndex, *insertIndex); + } + ++insertIndex; + } + } + if(insertIndex != centerValue) + { + VMA_SWAP(*insertIndex, *centerValue); + } + return insertIndex; +} + +template +void VmaQuickSort(Iterator beg, Iterator end, Compare cmp) +{ + if(beg < end) + { + Iterator it = VmaQuickSortPartition(beg, end, cmp); + VmaQuickSort(beg, it, cmp); + VmaQuickSort(it + 1, end, cmp); + } +} + +#define VMA_SORT(beg, end, cmp) VmaQuickSort(beg, end, cmp) + +#endif // #ifndef VMA_SORT + +/* +Returns true if two memory blocks occupy overlapping pages. +ResourceA must be in less memory offset than ResourceB. + +Algorithm is based on "Vulkan 1.0.39 - A Specification (with all registered Vulkan extensions)" +chapter 11.6 "Resource Memory Association", paragraph "Buffer-Image Granularity". +*/ +static inline bool VmaBlocksOnSamePage( + VkDeviceSize resourceAOffset, + VkDeviceSize resourceASize, + VkDeviceSize resourceBOffset, + VkDeviceSize pageSize) +{ + VMA_ASSERT(resourceAOffset + resourceASize <= resourceBOffset && resourceASize > 0 && pageSize > 0); + VkDeviceSize resourceAEnd = resourceAOffset + resourceASize - 1; + VkDeviceSize resourceAEndPage = resourceAEnd & ~(pageSize - 1); + VkDeviceSize resourceBStart = resourceBOffset; + VkDeviceSize resourceBStartPage = resourceBStart & ~(pageSize - 1); + return resourceAEndPage == resourceBStartPage; +} + +enum VmaSuballocationType +{ + VMA_SUBALLOCATION_TYPE_FREE = 0, + VMA_SUBALLOCATION_TYPE_UNKNOWN = 1, + VMA_SUBALLOCATION_TYPE_BUFFER = 2, + VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN = 3, + VMA_SUBALLOCATION_TYPE_IMAGE_LINEAR = 4, + VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL = 5, + VMA_SUBALLOCATION_TYPE_MAX_ENUM = 0x7FFFFFFF +}; + +/* +Returns true if given suballocation types could conflict and must respect +VkPhysicalDeviceLimits::bufferImageGranularity. They conflict if one is buffer +or linear image and another one is optimal image. If type is unknown, behave +conservatively. +*/ +static inline bool VmaIsBufferImageGranularityConflict( + VmaSuballocationType suballocType1, + VmaSuballocationType suballocType2) +{ + if(suballocType1 > suballocType2) + { + VMA_SWAP(suballocType1, suballocType2); + } + + switch(suballocType1) + { + case VMA_SUBALLOCATION_TYPE_FREE: + return false; + case VMA_SUBALLOCATION_TYPE_UNKNOWN: + return true; + case VMA_SUBALLOCATION_TYPE_BUFFER: + return + suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN || + suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL; + case VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN: + return + suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN || + suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_LINEAR || + suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL; + case VMA_SUBALLOCATION_TYPE_IMAGE_LINEAR: + return + suballocType2 == VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL; + case VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL: + return false; + default: + VMA_ASSERT(0); + return true; + } +} + +static void VmaWriteMagicValue(void* pData, VkDeviceSize offset) +{ +#if VMA_DEBUG_MARGIN > 0 && VMA_DEBUG_DETECT_CORRUPTION + uint32_t* pDst = (uint32_t*)((char*)pData + offset); + const size_t numberCount = VMA_DEBUG_MARGIN / sizeof(uint32_t); + for(size_t i = 0; i < numberCount; ++i, ++pDst) + { + *pDst = VMA_CORRUPTION_DETECTION_MAGIC_VALUE; + } +#else + // no-op +#endif +} + +static bool VmaValidateMagicValue(const void* pData, VkDeviceSize offset) +{ +#if VMA_DEBUG_MARGIN > 0 && VMA_DEBUG_DETECT_CORRUPTION + const uint32_t* pSrc = (const uint32_t*)((const char*)pData + offset); + const size_t numberCount = VMA_DEBUG_MARGIN / sizeof(uint32_t); + for(size_t i = 0; i < numberCount; ++i, ++pSrc) + { + if(*pSrc != VMA_CORRUPTION_DETECTION_MAGIC_VALUE) + { + return false; + } + } +#endif + return true; +} + +/* +Fills structure with parameters of an example buffer to be used for transfers +during GPU memory defragmentation. +*/ +static void VmaFillGpuDefragmentationBufferCreateInfo(VkBufferCreateInfo& outBufCreateInfo) +{ + memset(&outBufCreateInfo, 0, sizeof(outBufCreateInfo)); + outBufCreateInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; + outBufCreateInfo.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT; + outBufCreateInfo.size = (VkDeviceSize)VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE; // Example size. +} + +// Helper RAII class to lock a mutex in constructor and unlock it in destructor (at the end of scope). +struct VmaMutexLock +{ + VMA_CLASS_NO_COPY(VmaMutexLock) +public: + VmaMutexLock(VMA_MUTEX& mutex, bool useMutex = true) : + m_pMutex(useMutex ? &mutex : VMA_NULL) + { if(m_pMutex) { m_pMutex->Lock(); } } + ~VmaMutexLock() + { if(m_pMutex) { m_pMutex->Unlock(); } } +private: + VMA_MUTEX* m_pMutex; +}; + +// Helper RAII class to lock a RW mutex in constructor and unlock it in destructor (at the end of scope), for reading. +struct VmaMutexLockRead +{ + VMA_CLASS_NO_COPY(VmaMutexLockRead) +public: + VmaMutexLockRead(VMA_RW_MUTEX& mutex, bool useMutex) : + m_pMutex(useMutex ? &mutex : VMA_NULL) + { if(m_pMutex) { m_pMutex->LockRead(); } } + ~VmaMutexLockRead() { if(m_pMutex) { m_pMutex->UnlockRead(); } } +private: + VMA_RW_MUTEX* m_pMutex; +}; + +// Helper RAII class to lock a RW mutex in constructor and unlock it in destructor (at the end of scope), for writing. +struct VmaMutexLockWrite +{ + VMA_CLASS_NO_COPY(VmaMutexLockWrite) +public: + VmaMutexLockWrite(VMA_RW_MUTEX& mutex, bool useMutex) : + m_pMutex(useMutex ? &mutex : VMA_NULL) + { if(m_pMutex) { m_pMutex->LockWrite(); } } + ~VmaMutexLockWrite() { if(m_pMutex) { m_pMutex->UnlockWrite(); } } +private: + VMA_RW_MUTEX* m_pMutex; +}; + +#if VMA_DEBUG_GLOBAL_MUTEX + static VMA_MUTEX gDebugGlobalMutex; + #define VMA_DEBUG_GLOBAL_MUTEX_LOCK VmaMutexLock debugGlobalMutexLock(gDebugGlobalMutex, true); +#else + #define VMA_DEBUG_GLOBAL_MUTEX_LOCK +#endif + +// Minimum size of a free suballocation to register it in the free suballocation collection. +static const VkDeviceSize VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER = 16; + +/* +Performs binary search and returns iterator to first element that is greater or +equal to (key), according to comparison (cmp). + +Cmp should return true if first argument is less than second argument. + +Returned value is the found element, if present in the collection or place where +new element with value (key) should be inserted. +*/ +template +static IterT VmaBinaryFindFirstNotLess(IterT beg, IterT end, const KeyT &key, const CmpLess& cmp) +{ + size_t down = 0, up = (end - beg); + while(down < up) + { + const size_t mid = down + (up - down) / 2; // Overflow-safe midpoint calculation + if(cmp(*(beg+mid), key)) + { + down = mid + 1; + } + else + { + up = mid; + } + } + return beg + down; +} + +template +IterT VmaBinaryFindSorted(const IterT& beg, const IterT& end, const KeyT& value, const CmpLess& cmp) +{ + IterT it = VmaBinaryFindFirstNotLess( + beg, end, value, cmp); + if(it == end || + (!cmp(*it, value) && !cmp(value, *it))) + { + return it; + } + return end; +} + +/* +Returns true if all pointers in the array are not-null and unique. +Warning! O(n^2) complexity. Use only inside VMA_HEAVY_ASSERT. +T must be pointer type, e.g. VmaAllocation, VmaPool. +*/ +template +static bool VmaValidatePointerArray(uint32_t count, const T* arr) +{ + for(uint32_t i = 0; i < count; ++i) + { + const T iPtr = arr[i]; + if(iPtr == VMA_NULL) + { + return false; + } + for(uint32_t j = i + 1; j < count; ++j) + { + if(iPtr == arr[j]) + { + return false; + } + } + } + return true; +} + +template +static inline void VmaPnextChainPushFront(MainT* mainStruct, NewT* newStruct) +{ + newStruct->pNext = mainStruct->pNext; + mainStruct->pNext = newStruct; +} + +//////////////////////////////////////////////////////////////////////////////// +// Memory allocation + +static void* VmaMalloc(const VkAllocationCallbacks* pAllocationCallbacks, size_t size, size_t alignment) +{ + void* result = VMA_NULL; + if((pAllocationCallbacks != VMA_NULL) && + (pAllocationCallbacks->pfnAllocation != VMA_NULL)) + { + result = (*pAllocationCallbacks->pfnAllocation)( + pAllocationCallbacks->pUserData, + size, + alignment, + VK_SYSTEM_ALLOCATION_SCOPE_OBJECT); + } + else + { + result = VMA_SYSTEM_ALIGNED_MALLOC(size, alignment); + } + VMA_ASSERT(result != VMA_NULL && "CPU memory allocation failed."); + return result; +} + +static void VmaFree(const VkAllocationCallbacks* pAllocationCallbacks, void* ptr) +{ + if((pAllocationCallbacks != VMA_NULL) && + (pAllocationCallbacks->pfnFree != VMA_NULL)) + { + (*pAllocationCallbacks->pfnFree)(pAllocationCallbacks->pUserData, ptr); + } + else + { + VMA_SYSTEM_ALIGNED_FREE(ptr); + } +} + +template +static T* VmaAllocate(const VkAllocationCallbacks* pAllocationCallbacks) +{ + return (T*)VmaMalloc(pAllocationCallbacks, sizeof(T), VMA_ALIGN_OF(T)); +} + +template +static T* VmaAllocateArray(const VkAllocationCallbacks* pAllocationCallbacks, size_t count) +{ + return (T*)VmaMalloc(pAllocationCallbacks, sizeof(T) * count, VMA_ALIGN_OF(T)); +} + +#define vma_new(allocator, type) new(VmaAllocate(allocator))(type) + +#define vma_new_array(allocator, type, count) new(VmaAllocateArray((allocator), (count)))(type) + +template +static void vma_delete(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr) +{ + ptr->~T(); + VmaFree(pAllocationCallbacks, ptr); +} + +template +static void vma_delete_array(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr, size_t count) +{ + if(ptr != VMA_NULL) + { + for(size_t i = count; i--; ) + { + ptr[i].~T(); + } + VmaFree(pAllocationCallbacks, ptr); + } +} + +static char* VmaCreateStringCopy(const VkAllocationCallbacks* allocs, const char* srcStr) +{ + if(srcStr != VMA_NULL) + { + const size_t len = strlen(srcStr); + char* const result = vma_new_array(allocs, char, len + 1); + memcpy(result, srcStr, len + 1); + return result; + } + else + { + return VMA_NULL; + } +} + +static void VmaFreeString(const VkAllocationCallbacks* allocs, char* str) +{ + if(str != VMA_NULL) + { + const size_t len = strlen(str); + vma_delete_array(allocs, str, len + 1); + } +} + +// STL-compatible allocator. +template +class VmaStlAllocator +{ +public: + const VkAllocationCallbacks* const m_pCallbacks; + typedef T value_type; + + VmaStlAllocator(const VkAllocationCallbacks* pCallbacks) : m_pCallbacks(pCallbacks) { } + template VmaStlAllocator(const VmaStlAllocator& src) : m_pCallbacks(src.m_pCallbacks) { } + + T* allocate(size_t n) { return VmaAllocateArray(m_pCallbacks, n); } + void deallocate(T* p, size_t n) { VmaFree(m_pCallbacks, p); } + + template + bool operator==(const VmaStlAllocator& rhs) const + { + return m_pCallbacks == rhs.m_pCallbacks; + } + template + bool operator!=(const VmaStlAllocator& rhs) const + { + return m_pCallbacks != rhs.m_pCallbacks; + } + + VmaStlAllocator& operator=(const VmaStlAllocator& x) = delete; +}; + +#if VMA_USE_STL_VECTOR + +#define VmaVector std::vector + +template +static void VmaVectorInsert(std::vector& vec, size_t index, const T& item) +{ + vec.insert(vec.begin() + index, item); +} + +template +static void VmaVectorRemove(std::vector& vec, size_t index) +{ + vec.erase(vec.begin() + index); +} + +#else // #if VMA_USE_STL_VECTOR + +/* Class with interface compatible with subset of std::vector. +T must be POD because constructors and destructors are not called and memcpy is +used for these objects. */ +template +class VmaVector +{ +public: + typedef T value_type; + + VmaVector(const AllocatorT& allocator) : + m_Allocator(allocator), + m_pArray(VMA_NULL), + m_Count(0), + m_Capacity(0) + { + } + + VmaVector(size_t count, const AllocatorT& allocator) : + m_Allocator(allocator), + m_pArray(count ? (T*)VmaAllocateArray(allocator.m_pCallbacks, count) : VMA_NULL), + m_Count(count), + m_Capacity(count) + { + } + + // This version of the constructor is here for compatibility with pre-C++14 std::vector. + // value is unused. + VmaVector(size_t count, const T& value, const AllocatorT& allocator) + : VmaVector(count, allocator) {} + + VmaVector(const VmaVector& src) : + m_Allocator(src.m_Allocator), + m_pArray(src.m_Count ? (T*)VmaAllocateArray(src.m_Allocator.m_pCallbacks, src.m_Count) : VMA_NULL), + m_Count(src.m_Count), + m_Capacity(src.m_Count) + { + if(m_Count != 0) + { + memcpy(m_pArray, src.m_pArray, m_Count * sizeof(T)); + } + } + + ~VmaVector() + { + VmaFree(m_Allocator.m_pCallbacks, m_pArray); + } + + VmaVector& operator=(const VmaVector& rhs) + { + if(&rhs != this) + { + resize(rhs.m_Count); + if(m_Count != 0) + { + memcpy(m_pArray, rhs.m_pArray, m_Count * sizeof(T)); + } + } + return *this; + } + + bool empty() const { return m_Count == 0; } + size_t size() const { return m_Count; } + T* data() { return m_pArray; } + const T* data() const { return m_pArray; } + + T& operator[](size_t index) + { + VMA_HEAVY_ASSERT(index < m_Count); + return m_pArray[index]; + } + const T& operator[](size_t index) const + { + VMA_HEAVY_ASSERT(index < m_Count); + return m_pArray[index]; + } + + T& front() + { + VMA_HEAVY_ASSERT(m_Count > 0); + return m_pArray[0]; + } + const T& front() const + { + VMA_HEAVY_ASSERT(m_Count > 0); + return m_pArray[0]; + } + T& back() + { + VMA_HEAVY_ASSERT(m_Count > 0); + return m_pArray[m_Count - 1]; + } + const T& back() const + { + VMA_HEAVY_ASSERT(m_Count > 0); + return m_pArray[m_Count - 1]; + } + + void reserve(size_t newCapacity, bool freeMemory = false) + { + newCapacity = VMA_MAX(newCapacity, m_Count); + + if((newCapacity < m_Capacity) && !freeMemory) + { + newCapacity = m_Capacity; + } + + if(newCapacity != m_Capacity) + { + T* const newArray = newCapacity ? VmaAllocateArray(m_Allocator, newCapacity) : VMA_NULL; + if(m_Count != 0) + { + memcpy(newArray, m_pArray, m_Count * sizeof(T)); + } + VmaFree(m_Allocator.m_pCallbacks, m_pArray); + m_Capacity = newCapacity; + m_pArray = newArray; + } + } + + void resize(size_t newCount, bool freeMemory = false) + { + size_t newCapacity = m_Capacity; + if(newCount > m_Capacity) + { + newCapacity = VMA_MAX(newCount, VMA_MAX(m_Capacity * 3 / 2, (size_t)8)); + } + else if(freeMemory) + { + newCapacity = newCount; + } + + if(newCapacity != m_Capacity) + { + T* const newArray = newCapacity ? VmaAllocateArray(m_Allocator.m_pCallbacks, newCapacity) : VMA_NULL; + const size_t elementsToCopy = VMA_MIN(m_Count, newCount); + if(elementsToCopy != 0) + { + memcpy(newArray, m_pArray, elementsToCopy * sizeof(T)); + } + VmaFree(m_Allocator.m_pCallbacks, m_pArray); + m_Capacity = newCapacity; + m_pArray = newArray; + } + + m_Count = newCount; + } + + void clear(bool freeMemory = false) + { + resize(0, freeMemory); + } + + void insert(size_t index, const T& src) + { + VMA_HEAVY_ASSERT(index <= m_Count); + const size_t oldCount = size(); + resize(oldCount + 1); + if(index < oldCount) + { + memmove(m_pArray + (index + 1), m_pArray + index, (oldCount - index) * sizeof(T)); + } + m_pArray[index] = src; + } + + void remove(size_t index) + { + VMA_HEAVY_ASSERT(index < m_Count); + const size_t oldCount = size(); + if(index < oldCount - 1) + { + memmove(m_pArray + index, m_pArray + (index + 1), (oldCount - index - 1) * sizeof(T)); + } + resize(oldCount - 1); + } + + void push_back(const T& src) + { + const size_t newIndex = size(); + resize(newIndex + 1); + m_pArray[newIndex] = src; + } + + void pop_back() + { + VMA_HEAVY_ASSERT(m_Count > 0); + resize(size() - 1); + } + + void push_front(const T& src) + { + insert(0, src); + } + + void pop_front() + { + VMA_HEAVY_ASSERT(m_Count > 0); + remove(0); + } + + typedef T* iterator; + + iterator begin() { return m_pArray; } + iterator end() { return m_pArray + m_Count; } + +private: + AllocatorT m_Allocator; + T* m_pArray; + size_t m_Count; + size_t m_Capacity; +}; + +template +static void VmaVectorInsert(VmaVector& vec, size_t index, const T& item) +{ + vec.insert(index, item); +} + +template +static void VmaVectorRemove(VmaVector& vec, size_t index) +{ + vec.remove(index); +} + +#endif // #if VMA_USE_STL_VECTOR + +template +size_t VmaVectorInsertSorted(VectorT& vector, const typename VectorT::value_type& value) +{ + const size_t indexToInsert = VmaBinaryFindFirstNotLess( + vector.data(), + vector.data() + vector.size(), + value, + CmpLess()) - vector.data(); + VmaVectorInsert(vector, indexToInsert, value); + return indexToInsert; +} + +template +bool VmaVectorRemoveSorted(VectorT& vector, const typename VectorT::value_type& value) +{ + CmpLess comparator; + typename VectorT::iterator it = VmaBinaryFindFirstNotLess( + vector.begin(), + vector.end(), + value, + comparator); + if((it != vector.end()) && !comparator(*it, value) && !comparator(value, *it)) + { + size_t indexToRemove = it - vector.begin(); + VmaVectorRemove(vector, indexToRemove); + return true; + } + return false; +} + +//////////////////////////////////////////////////////////////////////////////// +// class VmaSmallVector + +/* +This is a vector (a variable-sized array), optimized for the case when the array is small. + +It contains some number of elements in-place, which allows it to avoid heap allocation +when the actual number of elements is below that threshold. This allows normal "small" +cases to be fast without losing generality for large inputs. +*/ + +template +class VmaSmallVector +{ +public: + typedef T value_type; + + VmaSmallVector(const AllocatorT& allocator) : + m_Count(0), + m_DynamicArray(allocator) + { + } + VmaSmallVector(size_t count, const AllocatorT& allocator) : + m_Count(count), + m_DynamicArray(count > N ? count : 0, allocator) + { + } + template + VmaSmallVector(const VmaSmallVector& src) = delete; + template + VmaSmallVector& operator=(const VmaSmallVector& rhs) = delete; + + bool empty() const { return m_Count == 0; } + size_t size() const { return m_Count; } + T* data() { return m_Count > N ? m_DynamicArray.data() : m_StaticArray; } + const T* data() const { return m_Count > N ? m_DynamicArray.data() : m_StaticArray; } + + T& operator[](size_t index) + { + VMA_HEAVY_ASSERT(index < m_Count); + return data()[index]; + } + const T& operator[](size_t index) const + { + VMA_HEAVY_ASSERT(index < m_Count); + return data()[index]; + } + + T& front() + { + VMA_HEAVY_ASSERT(m_Count > 0); + return data()[0]; + } + const T& front() const + { + VMA_HEAVY_ASSERT(m_Count > 0); + return data()[0]; + } + T& back() + { + VMA_HEAVY_ASSERT(m_Count > 0); + return data()[m_Count - 1]; + } + const T& back() const + { + VMA_HEAVY_ASSERT(m_Count > 0); + return data()[m_Count - 1]; + } + + void resize(size_t newCount, bool freeMemory = false) + { + if(newCount > N && m_Count > N) + { + // Any direction, staying in m_DynamicArray + m_DynamicArray.resize(newCount, freeMemory); + } + else if(newCount > N && m_Count <= N) + { + // Growing, moving from m_StaticArray to m_DynamicArray + m_DynamicArray.resize(newCount, freeMemory); + if(m_Count > 0) + { + memcpy(m_DynamicArray.data(), m_StaticArray, m_Count * sizeof(T)); + } + } + else if(newCount <= N && m_Count > N) + { + // Shrinking, moving from m_DynamicArray to m_StaticArray + if(newCount > 0) + { + memcpy(m_StaticArray, m_DynamicArray.data(), newCount * sizeof(T)); + } + m_DynamicArray.resize(0, freeMemory); + } + else + { + // Any direction, staying in m_StaticArray - nothing to do here + } + m_Count = newCount; + } + + void clear(bool freeMemory = false) + { + m_DynamicArray.clear(freeMemory); + m_Count = 0; + } + + void insert(size_t index, const T& src) + { + VMA_HEAVY_ASSERT(index <= m_Count); + const size_t oldCount = size(); + resize(oldCount + 1); + T* const dataPtr = data(); + if(index < oldCount) + { + // I know, this could be more optimal for case where memmove can be memcpy directly from m_StaticArray to m_DynamicArray. + memmove(dataPtr + (index + 1), dataPtr + index, (oldCount - index) * sizeof(T)); + } + dataPtr[index] = src; + } + + void remove(size_t index) + { + VMA_HEAVY_ASSERT(index < m_Count); + const size_t oldCount = size(); + if(index < oldCount - 1) + { + // I know, this could be more optimal for case where memmove can be memcpy directly from m_DynamicArray to m_StaticArray. + T* const dataPtr = data(); + memmove(dataPtr + index, dataPtr + (index + 1), (oldCount - index - 1) * sizeof(T)); + } + resize(oldCount - 1); + } + + void push_back(const T& src) + { + const size_t newIndex = size(); + resize(newIndex + 1); + data()[newIndex] = src; + } + + void pop_back() + { + VMA_HEAVY_ASSERT(m_Count > 0); + resize(size() - 1); + } + + void push_front(const T& src) + { + insert(0, src); + } + + void pop_front() + { + VMA_HEAVY_ASSERT(m_Count > 0); + remove(0); + } + + typedef T* iterator; + + iterator begin() { return data(); } + iterator end() { return data() + m_Count; } + +private: + size_t m_Count; + T m_StaticArray[N]; // Used when m_Size <= N + VmaVector m_DynamicArray; // Used when m_Size > N +}; + +//////////////////////////////////////////////////////////////////////////////// +// class VmaPoolAllocator + +/* +Allocator for objects of type T using a list of arrays (pools) to speed up +allocation. Number of elements that can be allocated is not bounded because +allocator can create multiple blocks. +*/ +template +class VmaPoolAllocator +{ + VMA_CLASS_NO_COPY(VmaPoolAllocator) +public: + VmaPoolAllocator(const VkAllocationCallbacks* pAllocationCallbacks, uint32_t firstBlockCapacity); + ~VmaPoolAllocator(); + template T* Alloc(Types... args); + void Free(T* ptr); + +private: + union Item + { + uint32_t NextFreeIndex; + alignas(T) char Value[sizeof(T)]; + }; + + struct ItemBlock + { + Item* pItems; + uint32_t Capacity; + uint32_t FirstFreeIndex; + }; + + const VkAllocationCallbacks* m_pAllocationCallbacks; + const uint32_t m_FirstBlockCapacity; + VmaVector< ItemBlock, VmaStlAllocator > m_ItemBlocks; + + ItemBlock& CreateNewBlock(); +}; + +template +VmaPoolAllocator::VmaPoolAllocator(const VkAllocationCallbacks* pAllocationCallbacks, uint32_t firstBlockCapacity) : + m_pAllocationCallbacks(pAllocationCallbacks), + m_FirstBlockCapacity(firstBlockCapacity), + m_ItemBlocks(VmaStlAllocator(pAllocationCallbacks)) +{ + VMA_ASSERT(m_FirstBlockCapacity > 1); +} + +template +VmaPoolAllocator::~VmaPoolAllocator() +{ + for(size_t i = m_ItemBlocks.size(); i--; ) + vma_delete_array(m_pAllocationCallbacks, m_ItemBlocks[i].pItems, m_ItemBlocks[i].Capacity); + m_ItemBlocks.clear(); +} + +template +template T* VmaPoolAllocator::Alloc(Types... args) +{ + for(size_t i = m_ItemBlocks.size(); i--; ) + { + ItemBlock& block = m_ItemBlocks[i]; + // This block has some free items: Use first one. + if(block.FirstFreeIndex != UINT32_MAX) + { + Item* const pItem = &block.pItems[block.FirstFreeIndex]; + block.FirstFreeIndex = pItem->NextFreeIndex; + T* result = (T*)&pItem->Value; + new(result)T(std::forward(args)...); // Explicit constructor call. + return result; + } + } + + // No block has free item: Create new one and use it. + ItemBlock& newBlock = CreateNewBlock(); + Item* const pItem = &newBlock.pItems[0]; + newBlock.FirstFreeIndex = pItem->NextFreeIndex; + T* result = (T*)&pItem->Value; + new(result)T(std::forward(args)...); // Explicit constructor call. + return result; +} + +template +void VmaPoolAllocator::Free(T* ptr) +{ + // Search all memory blocks to find ptr. + for(size_t i = m_ItemBlocks.size(); i--; ) + { + ItemBlock& block = m_ItemBlocks[i]; + + // Casting to union. + Item* pItemPtr; + memcpy(&pItemPtr, &ptr, sizeof(pItemPtr)); + + // Check if pItemPtr is in address range of this block. + if((pItemPtr >= block.pItems) && (pItemPtr < block.pItems + block.Capacity)) + { + ptr->~T(); // Explicit destructor call. + const uint32_t index = static_cast(pItemPtr - block.pItems); + pItemPtr->NextFreeIndex = block.FirstFreeIndex; + block.FirstFreeIndex = index; + return; + } + } + VMA_ASSERT(0 && "Pointer doesn't belong to this memory pool."); +} + +template +typename VmaPoolAllocator::ItemBlock& VmaPoolAllocator::CreateNewBlock() +{ + const uint32_t newBlockCapacity = m_ItemBlocks.empty() ? + m_FirstBlockCapacity : m_ItemBlocks.back().Capacity * 3 / 2; + + const ItemBlock newBlock = { + vma_new_array(m_pAllocationCallbacks, Item, newBlockCapacity), + newBlockCapacity, + 0 }; + + m_ItemBlocks.push_back(newBlock); + + // Setup singly-linked list of all free items in this block. + for(uint32_t i = 0; i < newBlockCapacity - 1; ++i) + newBlock.pItems[i].NextFreeIndex = i + 1; + newBlock.pItems[newBlockCapacity - 1].NextFreeIndex = UINT32_MAX; + return m_ItemBlocks.back(); +} + +//////////////////////////////////////////////////////////////////////////////// +// class VmaRawList, VmaList + +#if VMA_USE_STL_LIST + +#define VmaList std::list + +#else // #if VMA_USE_STL_LIST + +template +struct VmaListItem +{ + VmaListItem* pPrev; + VmaListItem* pNext; + T Value; +}; + +// Doubly linked list. +template +class VmaRawList +{ + VMA_CLASS_NO_COPY(VmaRawList) +public: + typedef VmaListItem ItemType; + + VmaRawList(const VkAllocationCallbacks* pAllocationCallbacks); + ~VmaRawList(); + void Clear(); + + size_t GetCount() const { return m_Count; } + bool IsEmpty() const { return m_Count == 0; } + + ItemType* Front() { return m_pFront; } + const ItemType* Front() const { return m_pFront; } + ItemType* Back() { return m_pBack; } + const ItemType* Back() const { return m_pBack; } + + ItemType* PushBack(); + ItemType* PushFront(); + ItemType* PushBack(const T& value); + ItemType* PushFront(const T& value); + void PopBack(); + void PopFront(); + + // Item can be null - it means PushBack. + ItemType* InsertBefore(ItemType* pItem); + // Item can be null - it means PushFront. + ItemType* InsertAfter(ItemType* pItem); + + ItemType* InsertBefore(ItemType* pItem, const T& value); + ItemType* InsertAfter(ItemType* pItem, const T& value); + + void Remove(ItemType* pItem); + +private: + const VkAllocationCallbacks* const m_pAllocationCallbacks; + VmaPoolAllocator m_ItemAllocator; + ItemType* m_pFront; + ItemType* m_pBack; + size_t m_Count; +}; + +template +VmaRawList::VmaRawList(const VkAllocationCallbacks* pAllocationCallbacks) : + m_pAllocationCallbacks(pAllocationCallbacks), + m_ItemAllocator(pAllocationCallbacks, 128), + m_pFront(VMA_NULL), + m_pBack(VMA_NULL), + m_Count(0) +{ +} + +template +VmaRawList::~VmaRawList() +{ + // Intentionally not calling Clear, because that would be unnecessary + // computations to return all items to m_ItemAllocator as free. +} + +template +void VmaRawList::Clear() +{ + if(IsEmpty() == false) + { + ItemType* pItem = m_pBack; + while(pItem != VMA_NULL) + { + ItemType* const pPrevItem = pItem->pPrev; + m_ItemAllocator.Free(pItem); + pItem = pPrevItem; + } + m_pFront = VMA_NULL; + m_pBack = VMA_NULL; + m_Count = 0; + } +} + +template +VmaListItem* VmaRawList::PushBack() +{ + ItemType* const pNewItem = m_ItemAllocator.Alloc(); + pNewItem->pNext = VMA_NULL; + if(IsEmpty()) + { + pNewItem->pPrev = VMA_NULL; + m_pFront = pNewItem; + m_pBack = pNewItem; + m_Count = 1; + } + else + { + pNewItem->pPrev = m_pBack; + m_pBack->pNext = pNewItem; + m_pBack = pNewItem; + ++m_Count; + } + return pNewItem; +} + +template +VmaListItem* VmaRawList::PushFront() +{ + ItemType* const pNewItem = m_ItemAllocator.Alloc(); + pNewItem->pPrev = VMA_NULL; + if(IsEmpty()) + { + pNewItem->pNext = VMA_NULL; + m_pFront = pNewItem; + m_pBack = pNewItem; + m_Count = 1; + } + else + { + pNewItem->pNext = m_pFront; + m_pFront->pPrev = pNewItem; + m_pFront = pNewItem; + ++m_Count; + } + return pNewItem; +} + +template +VmaListItem* VmaRawList::PushBack(const T& value) +{ + ItemType* const pNewItem = PushBack(); + pNewItem->Value = value; + return pNewItem; +} + +template +VmaListItem* VmaRawList::PushFront(const T& value) +{ + ItemType* const pNewItem = PushFront(); + pNewItem->Value = value; + return pNewItem; +} + +template +void VmaRawList::PopBack() +{ + VMA_HEAVY_ASSERT(m_Count > 0); + ItemType* const pBackItem = m_pBack; + ItemType* const pPrevItem = pBackItem->pPrev; + if(pPrevItem != VMA_NULL) + { + pPrevItem->pNext = VMA_NULL; + } + m_pBack = pPrevItem; + m_ItemAllocator.Free(pBackItem); + --m_Count; +} + +template +void VmaRawList::PopFront() +{ + VMA_HEAVY_ASSERT(m_Count > 0); + ItemType* const pFrontItem = m_pFront; + ItemType* const pNextItem = pFrontItem->pNext; + if(pNextItem != VMA_NULL) + { + pNextItem->pPrev = VMA_NULL; + } + m_pFront = pNextItem; + m_ItemAllocator.Free(pFrontItem); + --m_Count; +} + +template +void VmaRawList::Remove(ItemType* pItem) +{ + VMA_HEAVY_ASSERT(pItem != VMA_NULL); + VMA_HEAVY_ASSERT(m_Count > 0); + + if(pItem->pPrev != VMA_NULL) + { + pItem->pPrev->pNext = pItem->pNext; + } + else + { + VMA_HEAVY_ASSERT(m_pFront == pItem); + m_pFront = pItem->pNext; + } + + if(pItem->pNext != VMA_NULL) + { + pItem->pNext->pPrev = pItem->pPrev; + } + else + { + VMA_HEAVY_ASSERT(m_pBack == pItem); + m_pBack = pItem->pPrev; + } + + m_ItemAllocator.Free(pItem); + --m_Count; +} + +template +VmaListItem* VmaRawList::InsertBefore(ItemType* pItem) +{ + if(pItem != VMA_NULL) + { + ItemType* const prevItem = pItem->pPrev; + ItemType* const newItem = m_ItemAllocator.Alloc(); + newItem->pPrev = prevItem; + newItem->pNext = pItem; + pItem->pPrev = newItem; + if(prevItem != VMA_NULL) + { + prevItem->pNext = newItem; + } + else + { + VMA_HEAVY_ASSERT(m_pFront == pItem); + m_pFront = newItem; + } + ++m_Count; + return newItem; + } + else + return PushBack(); +} + +template +VmaListItem* VmaRawList::InsertAfter(ItemType* pItem) +{ + if(pItem != VMA_NULL) + { + ItemType* const nextItem = pItem->pNext; + ItemType* const newItem = m_ItemAllocator.Alloc(); + newItem->pNext = nextItem; + newItem->pPrev = pItem; + pItem->pNext = newItem; + if(nextItem != VMA_NULL) + { + nextItem->pPrev = newItem; + } + else + { + VMA_HEAVY_ASSERT(m_pBack == pItem); + m_pBack = newItem; + } + ++m_Count; + return newItem; + } + else + return PushFront(); +} + +template +VmaListItem* VmaRawList::InsertBefore(ItemType* pItem, const T& value) +{ + ItemType* const newItem = InsertBefore(pItem); + newItem->Value = value; + return newItem; +} + +template +VmaListItem* VmaRawList::InsertAfter(ItemType* pItem, const T& value) +{ + ItemType* const newItem = InsertAfter(pItem); + newItem->Value = value; + return newItem; +} + +template +class VmaList +{ + VMA_CLASS_NO_COPY(VmaList) +public: + class iterator + { + public: + iterator() : + m_pList(VMA_NULL), + m_pItem(VMA_NULL) + { + } + + T& operator*() const + { + VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); + return m_pItem->Value; + } + T* operator->() const + { + VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); + return &m_pItem->Value; + } + + iterator& operator++() + { + VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); + m_pItem = m_pItem->pNext; + return *this; + } + iterator& operator--() + { + if(m_pItem != VMA_NULL) + { + m_pItem = m_pItem->pPrev; + } + else + { + VMA_HEAVY_ASSERT(!m_pList->IsEmpty()); + m_pItem = m_pList->Back(); + } + return *this; + } + + iterator operator++(int) + { + iterator result = *this; + ++*this; + return result; + } + iterator operator--(int) + { + iterator result = *this; + --*this; + return result; + } + + bool operator==(const iterator& rhs) const + { + VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); + return m_pItem == rhs.m_pItem; + } + bool operator!=(const iterator& rhs) const + { + VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); + return m_pItem != rhs.m_pItem; + } + + private: + VmaRawList* m_pList; + VmaListItem* m_pItem; + + iterator(VmaRawList* pList, VmaListItem* pItem) : + m_pList(pList), + m_pItem(pItem) + { + } + + friend class VmaList; + }; + + class const_iterator + { + public: + const_iterator() : + m_pList(VMA_NULL), + m_pItem(VMA_NULL) + { + } + + const_iterator(const iterator& src) : + m_pList(src.m_pList), + m_pItem(src.m_pItem) + { + } + + const T& operator*() const + { + VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); + return m_pItem->Value; + } + const T* operator->() const + { + VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); + return &m_pItem->Value; + } + + const_iterator& operator++() + { + VMA_HEAVY_ASSERT(m_pItem != VMA_NULL); + m_pItem = m_pItem->pNext; + return *this; + } + const_iterator& operator--() + { + if(m_pItem != VMA_NULL) + { + m_pItem = m_pItem->pPrev; + } + else + { + VMA_HEAVY_ASSERT(!m_pList->IsEmpty()); + m_pItem = m_pList->Back(); + } + return *this; + } + + const_iterator operator++(int) + { + const_iterator result = *this; + ++*this; + return result; + } + const_iterator operator--(int) + { + const_iterator result = *this; + --*this; + return result; + } + + bool operator==(const const_iterator& rhs) const + { + VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); + return m_pItem == rhs.m_pItem; + } + bool operator!=(const const_iterator& rhs) const + { + VMA_HEAVY_ASSERT(m_pList == rhs.m_pList); + return m_pItem != rhs.m_pItem; + } + + private: + const_iterator(const VmaRawList* pList, const VmaListItem* pItem) : + m_pList(pList), + m_pItem(pItem) + { + } + + const VmaRawList* m_pList; + const VmaListItem* m_pItem; + + friend class VmaList; + }; + + VmaList(const AllocatorT& allocator) : m_RawList(allocator.m_pCallbacks) { } + + bool empty() const { return m_RawList.IsEmpty(); } + size_t size() const { return m_RawList.GetCount(); } + + iterator begin() { return iterator(&m_RawList, m_RawList.Front()); } + iterator end() { return iterator(&m_RawList, VMA_NULL); } + + const_iterator cbegin() const { return const_iterator(&m_RawList, m_RawList.Front()); } + const_iterator cend() const { return const_iterator(&m_RawList, VMA_NULL); } + + void clear() { m_RawList.Clear(); } + void push_back(const T& value) { m_RawList.PushBack(value); } + void erase(iterator it) { m_RawList.Remove(it.m_pItem); } + iterator insert(iterator it, const T& value) { return iterator(&m_RawList, m_RawList.InsertBefore(it.m_pItem, value)); } + +private: + VmaRawList m_RawList; +}; + +#endif // #if VMA_USE_STL_LIST + +//////////////////////////////////////////////////////////////////////////////// +// class VmaMap + +// Unused in this version. +#if 0 + +#if VMA_USE_STL_UNORDERED_MAP + +#define VmaPair std::pair + +#define VMA_MAP_TYPE(KeyT, ValueT) \ + std::unordered_map< KeyT, ValueT, std::hash, std::equal_to, VmaStlAllocator< std::pair > > + +#else // #if VMA_USE_STL_UNORDERED_MAP + +template +struct VmaPair +{ + T1 first; + T2 second; + + VmaPair() : first(), second() { } + VmaPair(const T1& firstSrc, const T2& secondSrc) : first(firstSrc), second(secondSrc) { } +}; + +/* Class compatible with subset of interface of std::unordered_map. +KeyT, ValueT must be POD because they will be stored in VmaVector. +*/ +template +class VmaMap +{ +public: + typedef VmaPair PairType; + typedef PairType* iterator; + + VmaMap(const VmaStlAllocator& allocator) : m_Vector(allocator) { } + + iterator begin() { return m_Vector.begin(); } + iterator end() { return m_Vector.end(); } + + void insert(const PairType& pair); + iterator find(const KeyT& key); + void erase(iterator it); + +private: + VmaVector< PairType, VmaStlAllocator > m_Vector; +}; + +#define VMA_MAP_TYPE(KeyT, ValueT) VmaMap + +template +struct VmaPairFirstLess +{ + bool operator()(const VmaPair& lhs, const VmaPair& rhs) const + { + return lhs.first < rhs.first; + } + bool operator()(const VmaPair& lhs, const FirstT& rhsFirst) const + { + return lhs.first < rhsFirst; + } +}; + +template +void VmaMap::insert(const PairType& pair) +{ + const size_t indexToInsert = VmaBinaryFindFirstNotLess( + m_Vector.data(), + m_Vector.data() + m_Vector.size(), + pair, + VmaPairFirstLess()) - m_Vector.data(); + VmaVectorInsert(m_Vector, indexToInsert, pair); +} + +template +VmaPair* VmaMap::find(const KeyT& key) +{ + PairType* it = VmaBinaryFindFirstNotLess( + m_Vector.data(), + m_Vector.data() + m_Vector.size(), + key, + VmaPairFirstLess()); + if((it != m_Vector.end()) && (it->first == key)) + { + return it; + } + else + { + return m_Vector.end(); + } +} + +template +void VmaMap::erase(iterator it) +{ + VmaVectorRemove(m_Vector, it - m_Vector.begin()); +} + +#endif // #if VMA_USE_STL_UNORDERED_MAP + +#endif // #if 0 + +//////////////////////////////////////////////////////////////////////////////// + +class VmaDeviceMemoryBlock; + +enum VMA_CACHE_OPERATION { VMA_CACHE_FLUSH, VMA_CACHE_INVALIDATE }; + +struct VmaAllocation_T +{ +private: + static const uint8_t MAP_COUNT_FLAG_PERSISTENT_MAP = 0x80; + + enum FLAGS + { + FLAG_USER_DATA_STRING = 0x01, + }; + +public: + enum ALLOCATION_TYPE + { + ALLOCATION_TYPE_NONE, + ALLOCATION_TYPE_BLOCK, + ALLOCATION_TYPE_DEDICATED, + }; + + /* + This struct is allocated using VmaPoolAllocator. + */ + + VmaAllocation_T(uint32_t currentFrameIndex, bool userDataString) : + m_Alignment{1}, + m_Size{0}, + m_pUserData{VMA_NULL}, + m_LastUseFrameIndex{currentFrameIndex}, + m_MemoryTypeIndex{0}, + m_Type{(uint8_t)ALLOCATION_TYPE_NONE}, + m_SuballocationType{(uint8_t)VMA_SUBALLOCATION_TYPE_UNKNOWN}, + m_MapCount{0}, + m_Flags{userDataString ? (uint8_t)FLAG_USER_DATA_STRING : (uint8_t)0} + { +#if VMA_STATS_STRING_ENABLED + m_CreationFrameIndex = currentFrameIndex; + m_BufferImageUsage = 0; +#endif + } + + ~VmaAllocation_T() + { + VMA_ASSERT((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) == 0 && "Allocation was not unmapped before destruction."); + + // Check if owned string was freed. + VMA_ASSERT(m_pUserData == VMA_NULL); + } + + void InitBlockAllocation( + VmaDeviceMemoryBlock* block, + VkDeviceSize offset, + VkDeviceSize alignment, + VkDeviceSize size, + uint32_t memoryTypeIndex, + VmaSuballocationType suballocationType, + bool mapped, + bool canBecomeLost) + { + VMA_ASSERT(m_Type == ALLOCATION_TYPE_NONE); + VMA_ASSERT(block != VMA_NULL); + m_Type = (uint8_t)ALLOCATION_TYPE_BLOCK; + m_Alignment = alignment; + m_Size = size; + m_MemoryTypeIndex = memoryTypeIndex; + m_MapCount = mapped ? MAP_COUNT_FLAG_PERSISTENT_MAP : 0; + m_SuballocationType = (uint8_t)suballocationType; + m_BlockAllocation.m_Block = block; + m_BlockAllocation.m_Offset = offset; + m_BlockAllocation.m_CanBecomeLost = canBecomeLost; + } + + void InitLost() + { + VMA_ASSERT(m_Type == ALLOCATION_TYPE_NONE); + VMA_ASSERT(m_LastUseFrameIndex.load() == VMA_FRAME_INDEX_LOST); + m_Type = (uint8_t)ALLOCATION_TYPE_BLOCK; + m_MemoryTypeIndex = 0; + m_BlockAllocation.m_Block = VMA_NULL; + m_BlockAllocation.m_Offset = 0; + m_BlockAllocation.m_CanBecomeLost = true; + } + + void ChangeBlockAllocation( + VmaAllocator hAllocator, + VmaDeviceMemoryBlock* block, + VkDeviceSize offset); + + void ChangeOffset(VkDeviceSize newOffset); + + // pMappedData not null means allocation is created with MAPPED flag. + void InitDedicatedAllocation( + uint32_t memoryTypeIndex, + VkDeviceMemory hMemory, + VmaSuballocationType suballocationType, + void* pMappedData, + VkDeviceSize size) + { + VMA_ASSERT(m_Type == ALLOCATION_TYPE_NONE); + VMA_ASSERT(hMemory != VK_NULL_HANDLE); + m_Type = (uint8_t)ALLOCATION_TYPE_DEDICATED; + m_Alignment = 0; + m_Size = size; + m_MemoryTypeIndex = memoryTypeIndex; + m_SuballocationType = (uint8_t)suballocationType; + m_MapCount = (pMappedData != VMA_NULL) ? MAP_COUNT_FLAG_PERSISTENT_MAP : 0; + m_DedicatedAllocation.m_hMemory = hMemory; + m_DedicatedAllocation.m_pMappedData = pMappedData; + } + + ALLOCATION_TYPE GetType() const { return (ALLOCATION_TYPE)m_Type; } + VkDeviceSize GetAlignment() const { return m_Alignment; } + VkDeviceSize GetSize() const { return m_Size; } + bool IsUserDataString() const { return (m_Flags & FLAG_USER_DATA_STRING) != 0; } + void* GetUserData() const { return m_pUserData; } + void SetUserData(VmaAllocator hAllocator, void* pUserData); + VmaSuballocationType GetSuballocationType() const { return (VmaSuballocationType)m_SuballocationType; } + + VmaDeviceMemoryBlock* GetBlock() const + { + VMA_ASSERT(m_Type == ALLOCATION_TYPE_BLOCK); + return m_BlockAllocation.m_Block; + } + VkDeviceSize GetOffset() const; + VkDeviceMemory GetMemory() const; + uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } + bool IsPersistentMap() const { return (m_MapCount & MAP_COUNT_FLAG_PERSISTENT_MAP) != 0; } + void* GetMappedData() const; + bool CanBecomeLost() const; + + uint32_t GetLastUseFrameIndex() const + { + return m_LastUseFrameIndex.load(); + } + bool CompareExchangeLastUseFrameIndex(uint32_t& expected, uint32_t desired) + { + return m_LastUseFrameIndex.compare_exchange_weak(expected, desired); + } + /* + - If hAllocation.LastUseFrameIndex + frameInUseCount < allocator.CurrentFrameIndex, + makes it lost by setting LastUseFrameIndex = VMA_FRAME_INDEX_LOST and returns true. + - Else, returns false. + + If hAllocation is already lost, assert - you should not call it then. + If hAllocation was not created with CAN_BECOME_LOST_BIT, assert. + */ + bool MakeLost(uint32_t currentFrameIndex, uint32_t frameInUseCount); + + void DedicatedAllocCalcStatsInfo(VmaStatInfo& outInfo) + { + VMA_ASSERT(m_Type == ALLOCATION_TYPE_DEDICATED); + outInfo.blockCount = 1; + outInfo.allocationCount = 1; + outInfo.unusedRangeCount = 0; + outInfo.usedBytes = m_Size; + outInfo.unusedBytes = 0; + outInfo.allocationSizeMin = outInfo.allocationSizeMax = m_Size; + outInfo.unusedRangeSizeMin = UINT64_MAX; + outInfo.unusedRangeSizeMax = 0; + } + + void BlockAllocMap(); + void BlockAllocUnmap(); + VkResult DedicatedAllocMap(VmaAllocator hAllocator, void** ppData); + void DedicatedAllocUnmap(VmaAllocator hAllocator); + +#if VMA_STATS_STRING_ENABLED + uint32_t GetCreationFrameIndex() const { return m_CreationFrameIndex; } + uint32_t GetBufferImageUsage() const { return m_BufferImageUsage; } + + void InitBufferImageUsage(uint32_t bufferImageUsage) + { + VMA_ASSERT(m_BufferImageUsage == 0); + m_BufferImageUsage = bufferImageUsage; + } + + void PrintParameters(class VmaJsonWriter& json) const; +#endif + +private: + VkDeviceSize m_Alignment; + VkDeviceSize m_Size; + void* m_pUserData; + VMA_ATOMIC_UINT32 m_LastUseFrameIndex; + uint32_t m_MemoryTypeIndex; + uint8_t m_Type; // ALLOCATION_TYPE + uint8_t m_SuballocationType; // VmaSuballocationType + // Bit 0x80 is set when allocation was created with VMA_ALLOCATION_CREATE_MAPPED_BIT. + // Bits with mask 0x7F are reference counter for vmaMapMemory()/vmaUnmapMemory(). + uint8_t m_MapCount; + uint8_t m_Flags; // enum FLAGS + + // Allocation out of VmaDeviceMemoryBlock. + struct BlockAllocation + { + VmaDeviceMemoryBlock* m_Block; + VkDeviceSize m_Offset; + bool m_CanBecomeLost; + }; + + // Allocation for an object that has its own private VkDeviceMemory. + struct DedicatedAllocation + { + VkDeviceMemory m_hMemory; + void* m_pMappedData; // Not null means memory is mapped. + }; + + union + { + // Allocation out of VmaDeviceMemoryBlock. + BlockAllocation m_BlockAllocation; + // Allocation for an object that has its own private VkDeviceMemory. + DedicatedAllocation m_DedicatedAllocation; + }; + +#if VMA_STATS_STRING_ENABLED + uint32_t m_CreationFrameIndex; + uint32_t m_BufferImageUsage; // 0 if unknown. +#endif + + void FreeUserDataString(VmaAllocator hAllocator); +}; + +/* +Represents a region of VmaDeviceMemoryBlock that is either assigned and returned as +allocated memory block or free. +*/ +struct VmaSuballocation +{ + VkDeviceSize offset; + VkDeviceSize size; + VmaAllocation hAllocation; + VmaSuballocationType type; +}; + +// Comparator for offsets. +struct VmaSuballocationOffsetLess +{ + bool operator()(const VmaSuballocation& lhs, const VmaSuballocation& rhs) const + { + return lhs.offset < rhs.offset; + } +}; +struct VmaSuballocationOffsetGreater +{ + bool operator()(const VmaSuballocation& lhs, const VmaSuballocation& rhs) const + { + return lhs.offset > rhs.offset; + } +}; + +typedef VmaList< VmaSuballocation, VmaStlAllocator > VmaSuballocationList; + +// Cost of one additional allocation lost, as equivalent in bytes. +static const VkDeviceSize VMA_LOST_ALLOCATION_COST = 1048576; + +enum class VmaAllocationRequestType +{ + Normal, + // Used by "Linear" algorithm. + UpperAddress, + EndOf1st, + EndOf2nd, +}; + +/* +Parameters of planned allocation inside a VmaDeviceMemoryBlock. + +If canMakeOtherLost was false: +- item points to a FREE suballocation. +- itemsToMakeLostCount is 0. + +If canMakeOtherLost was true: +- item points to first of sequence of suballocations, which are either FREE, + or point to VmaAllocations that can become lost. +- itemsToMakeLostCount is the number of VmaAllocations that need to be made lost for + the requested allocation to succeed. +*/ +struct VmaAllocationRequest +{ + VkDeviceSize offset; + VkDeviceSize sumFreeSize; // Sum size of free items that overlap with proposed allocation. + VkDeviceSize sumItemSize; // Sum size of items to make lost that overlap with proposed allocation. + VmaSuballocationList::iterator item; + size_t itemsToMakeLostCount; + void* customData; + VmaAllocationRequestType type; + + VkDeviceSize CalcCost() const + { + return sumItemSize + itemsToMakeLostCount * VMA_LOST_ALLOCATION_COST; + } +}; + +/* +Data structure used for bookkeeping of allocations and unused ranges of memory +in a single VkDeviceMemory block. +*/ +class VmaBlockMetadata +{ +public: + VmaBlockMetadata(VmaAllocator hAllocator); + virtual ~VmaBlockMetadata() { } + virtual void Init(VkDeviceSize size) { m_Size = size; } + + // Validates all data structures inside this object. If not valid, returns false. + virtual bool Validate() const = 0; + VkDeviceSize GetSize() const { return m_Size; } + virtual size_t GetAllocationCount() const = 0; + virtual VkDeviceSize GetSumFreeSize() const = 0; + virtual VkDeviceSize GetUnusedRangeSizeMax() const = 0; + // Returns true if this block is empty - contains only single free suballocation. + virtual bool IsEmpty() const = 0; + + virtual void CalcAllocationStatInfo(VmaStatInfo& outInfo) const = 0; + // Shouldn't modify blockCount. + virtual void AddPoolStats(VmaPoolStats& inoutStats) const = 0; + +#if VMA_STATS_STRING_ENABLED + virtual void PrintDetailedMap(class VmaJsonWriter& json) const = 0; +#endif + + // Tries to find a place for suballocation with given parameters inside this block. + // If succeeded, fills pAllocationRequest and returns true. + // If failed, returns false. + virtual bool CreateAllocationRequest( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + bool canMakeOtherLost, + // Always one of VMA_ALLOCATION_CREATE_STRATEGY_* or VMA_ALLOCATION_INTERNAL_STRATEGY_* flags. + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) = 0; + + virtual bool MakeRequestedAllocationsLost( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VmaAllocationRequest* pAllocationRequest) = 0; + + virtual uint32_t MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) = 0; + + virtual VkResult CheckCorruption(const void* pBlockData) = 0; + + // Makes actual allocation based on request. Request must already be checked and valid. + virtual void Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + VkDeviceSize allocSize, + VmaAllocation hAllocation) = 0; + + // Frees suballocation assigned to given memory region. + virtual void Free(const VmaAllocation allocation) = 0; + virtual void FreeAtOffset(VkDeviceSize offset) = 0; + +protected: + const VkAllocationCallbacks* GetAllocationCallbacks() const { return m_pAllocationCallbacks; } + +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap_Begin(class VmaJsonWriter& json, + VkDeviceSize unusedBytes, + size_t allocationCount, + size_t unusedRangeCount) const; + void PrintDetailedMap_Allocation(class VmaJsonWriter& json, + VkDeviceSize offset, + VmaAllocation hAllocation) const; + void PrintDetailedMap_UnusedRange(class VmaJsonWriter& json, + VkDeviceSize offset, + VkDeviceSize size) const; + void PrintDetailedMap_End(class VmaJsonWriter& json) const; +#endif + +private: + VkDeviceSize m_Size; + const VkAllocationCallbacks* m_pAllocationCallbacks; +}; + +#define VMA_VALIDATE(cond) do { if(!(cond)) { \ + VMA_ASSERT(0 && "Validation failed: " #cond); \ + return false; \ + } } while(false) + +class VmaBlockMetadata_Generic : public VmaBlockMetadata +{ + VMA_CLASS_NO_COPY(VmaBlockMetadata_Generic) +public: + VmaBlockMetadata_Generic(VmaAllocator hAllocator); + virtual ~VmaBlockMetadata_Generic(); + virtual void Init(VkDeviceSize size); + + virtual bool Validate() const; + virtual size_t GetAllocationCount() const { return m_Suballocations.size() - m_FreeCount; } + virtual VkDeviceSize GetSumFreeSize() const { return m_SumFreeSize; } + virtual VkDeviceSize GetUnusedRangeSizeMax() const; + virtual bool IsEmpty() const; + + virtual void CalcAllocationStatInfo(VmaStatInfo& outInfo) const; + virtual void AddPoolStats(VmaPoolStats& inoutStats) const; + +#if VMA_STATS_STRING_ENABLED + virtual void PrintDetailedMap(class VmaJsonWriter& json) const; +#endif + + virtual bool CreateAllocationRequest( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest); + + virtual bool MakeRequestedAllocationsLost( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VmaAllocationRequest* pAllocationRequest); + + virtual uint32_t MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount); + + virtual VkResult CheckCorruption(const void* pBlockData); + + virtual void Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + VkDeviceSize allocSize, + VmaAllocation hAllocation); + + virtual void Free(const VmaAllocation allocation); + virtual void FreeAtOffset(VkDeviceSize offset); + + //////////////////////////////////////////////////////////////////////////////// + // For defragmentation + + bool IsBufferImageGranularityConflictPossible( + VkDeviceSize bufferImageGranularity, + VmaSuballocationType& inOutPrevSuballocType) const; + +private: + friend class VmaDefragmentationAlgorithm_Generic; + friend class VmaDefragmentationAlgorithm_Fast; + + uint32_t m_FreeCount; + VkDeviceSize m_SumFreeSize; + VmaSuballocationList m_Suballocations; + // Suballocations that are free and have size greater than certain threshold. + // Sorted by size, ascending. + VmaVector< VmaSuballocationList::iterator, VmaStlAllocator< VmaSuballocationList::iterator > > m_FreeSuballocationsBySize; + + bool ValidateFreeSuballocationList() const; + + // Checks if requested suballocation with given parameters can be placed in given pFreeSuballocItem. + // If yes, fills pOffset and returns true. If no, returns false. + bool CheckAllocation( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + VmaSuballocationList::const_iterator suballocItem, + bool canMakeOtherLost, + VkDeviceSize* pOffset, + size_t* itemsToMakeLostCount, + VkDeviceSize* pSumFreeSize, + VkDeviceSize* pSumItemSize) const; + // Given free suballocation, it merges it with following one, which must also be free. + void MergeFreeWithNext(VmaSuballocationList::iterator item); + // Releases given suballocation, making it free. + // Merges it with adjacent free suballocations if applicable. + // Returns iterator to new free suballocation at this place. + VmaSuballocationList::iterator FreeSuballocation(VmaSuballocationList::iterator suballocItem); + // Given free suballocation, it inserts it into sorted list of + // m_FreeSuballocationsBySize if it's suitable. + void RegisterFreeSuballocation(VmaSuballocationList::iterator item); + // Given free suballocation, it removes it from sorted list of + // m_FreeSuballocationsBySize if it's suitable. + void UnregisterFreeSuballocation(VmaSuballocationList::iterator item); +}; + +/* +Allocations and their references in internal data structure look like this: + +if(m_2ndVectorMode == SECOND_VECTOR_EMPTY): + + 0 +-------+ + | | + | | + | | + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount] + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount + 1] + +-------+ + | ... | + +-------+ + | Alloc | 1st[1st.size() - 1] + +-------+ + | | + | | + | | +GetSize() +-------+ + +if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER): + + 0 +-------+ + | Alloc | 2nd[0] + +-------+ + | Alloc | 2nd[1] + +-------+ + | ... | + +-------+ + | Alloc | 2nd[2nd.size() - 1] + +-------+ + | | + | | + | | + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount] + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount + 1] + +-------+ + | ... | + +-------+ + | Alloc | 1st[1st.size() - 1] + +-------+ + | | +GetSize() +-------+ + +if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK): + + 0 +-------+ + | | + | | + | | + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount] + +-------+ + | Alloc | 1st[m_1stNullItemsBeginCount + 1] + +-------+ + | ... | + +-------+ + | Alloc | 1st[1st.size() - 1] + +-------+ + | | + | | + | | + +-------+ + | Alloc | 2nd[2nd.size() - 1] + +-------+ + | ... | + +-------+ + | Alloc | 2nd[1] + +-------+ + | Alloc | 2nd[0] +GetSize() +-------+ + +*/ +class VmaBlockMetadata_Linear : public VmaBlockMetadata +{ + VMA_CLASS_NO_COPY(VmaBlockMetadata_Linear) +public: + VmaBlockMetadata_Linear(VmaAllocator hAllocator); + virtual ~VmaBlockMetadata_Linear(); + virtual void Init(VkDeviceSize size); + + virtual bool Validate() const; + virtual size_t GetAllocationCount() const; + virtual VkDeviceSize GetSumFreeSize() const { return m_SumFreeSize; } + virtual VkDeviceSize GetUnusedRangeSizeMax() const; + virtual bool IsEmpty() const { return GetAllocationCount() == 0; } + + virtual void CalcAllocationStatInfo(VmaStatInfo& outInfo) const; + virtual void AddPoolStats(VmaPoolStats& inoutStats) const; + +#if VMA_STATS_STRING_ENABLED + virtual void PrintDetailedMap(class VmaJsonWriter& json) const; +#endif + + virtual bool CreateAllocationRequest( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest); + + virtual bool MakeRequestedAllocationsLost( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VmaAllocationRequest* pAllocationRequest); + + virtual uint32_t MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount); + + virtual VkResult CheckCorruption(const void* pBlockData); + + virtual void Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + VkDeviceSize allocSize, + VmaAllocation hAllocation); + + virtual void Free(const VmaAllocation allocation); + virtual void FreeAtOffset(VkDeviceSize offset); + +private: + /* + There are two suballocation vectors, used in ping-pong way. + The one with index m_1stVectorIndex is called 1st. + The one with index (m_1stVectorIndex ^ 1) is called 2nd. + 2nd can be non-empty only when 1st is not empty. + When 2nd is not empty, m_2ndVectorMode indicates its mode of operation. + */ + typedef VmaVector< VmaSuballocation, VmaStlAllocator > SuballocationVectorType; + + enum SECOND_VECTOR_MODE + { + SECOND_VECTOR_EMPTY, + /* + Suballocations in 2nd vector are created later than the ones in 1st, but they + all have smaller offset. + */ + SECOND_VECTOR_RING_BUFFER, + /* + Suballocations in 2nd vector are upper side of double stack. + They all have offsets higher than those in 1st vector. + Top of this stack means smaller offsets, but higher indices in this vector. + */ + SECOND_VECTOR_DOUBLE_STACK, + }; + + VkDeviceSize m_SumFreeSize; + SuballocationVectorType m_Suballocations0, m_Suballocations1; + uint32_t m_1stVectorIndex; + SECOND_VECTOR_MODE m_2ndVectorMode; + + SuballocationVectorType& AccessSuballocations1st() { return m_1stVectorIndex ? m_Suballocations1 : m_Suballocations0; } + SuballocationVectorType& AccessSuballocations2nd() { return m_1stVectorIndex ? m_Suballocations0 : m_Suballocations1; } + const SuballocationVectorType& AccessSuballocations1st() const { return m_1stVectorIndex ? m_Suballocations1 : m_Suballocations0; } + const SuballocationVectorType& AccessSuballocations2nd() const { return m_1stVectorIndex ? m_Suballocations0 : m_Suballocations1; } + + // Number of items in 1st vector with hAllocation = null at the beginning. + size_t m_1stNullItemsBeginCount; + // Number of other items in 1st vector with hAllocation = null somewhere in the middle. + size_t m_1stNullItemsMiddleCount; + // Number of items in 2nd vector with hAllocation = null. + size_t m_2ndNullItemsCount; + + bool ShouldCompact1st() const; + void CleanupAfterFree(); + + bool CreateAllocationRequest_LowerAddress( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest); + bool CreateAllocationRequest_UpperAddress( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest); +}; + +/* +- GetSize() is the original size of allocated memory block. +- m_UsableSize is this size aligned down to a power of two. + All allocations and calculations happen relative to m_UsableSize. +- GetUnusableSize() is the difference between them. + It is repoted as separate, unused range, not available for allocations. + +Node at level 0 has size = m_UsableSize. +Each next level contains nodes with size 2 times smaller than current level. +m_LevelCount is the maximum number of levels to use in the current object. +*/ +class VmaBlockMetadata_Buddy : public VmaBlockMetadata +{ + VMA_CLASS_NO_COPY(VmaBlockMetadata_Buddy) +public: + VmaBlockMetadata_Buddy(VmaAllocator hAllocator); + virtual ~VmaBlockMetadata_Buddy(); + virtual void Init(VkDeviceSize size); + + virtual bool Validate() const; + virtual size_t GetAllocationCount() const { return m_AllocationCount; } + virtual VkDeviceSize GetSumFreeSize() const { return m_SumFreeSize + GetUnusableSize(); } + virtual VkDeviceSize GetUnusedRangeSizeMax() const; + virtual bool IsEmpty() const { return m_Root->type == Node::TYPE_FREE; } + + virtual void CalcAllocationStatInfo(VmaStatInfo& outInfo) const; + virtual void AddPoolStats(VmaPoolStats& inoutStats) const; + +#if VMA_STATS_STRING_ENABLED + virtual void PrintDetailedMap(class VmaJsonWriter& json) const; +#endif + + virtual bool CreateAllocationRequest( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest); + + virtual bool MakeRequestedAllocationsLost( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VmaAllocationRequest* pAllocationRequest); + + virtual uint32_t MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount); + + virtual VkResult CheckCorruption(const void* pBlockData) { return VK_ERROR_FEATURE_NOT_PRESENT; } + + virtual void Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + VkDeviceSize allocSize, + VmaAllocation hAllocation); + + virtual void Free(const VmaAllocation allocation) { FreeAtOffset(allocation, allocation->GetOffset()); } + virtual void FreeAtOffset(VkDeviceSize offset) { FreeAtOffset(VMA_NULL, offset); } + +private: + static const VkDeviceSize MIN_NODE_SIZE = 32; + static const size_t MAX_LEVELS = 30; + + struct ValidationContext + { + size_t calculatedAllocationCount; + size_t calculatedFreeCount; + VkDeviceSize calculatedSumFreeSize; + + ValidationContext() : + calculatedAllocationCount(0), + calculatedFreeCount(0), + calculatedSumFreeSize(0) { } + }; + + struct Node + { + VkDeviceSize offset; + enum TYPE + { + TYPE_FREE, + TYPE_ALLOCATION, + TYPE_SPLIT, + TYPE_COUNT + } type; + Node* parent; + Node* buddy; + + union + { + struct + { + Node* prev; + Node* next; + } free; + struct + { + VmaAllocation alloc; + } allocation; + struct + { + Node* leftChild; + } split; + }; + }; + + // Size of the memory block aligned down to a power of two. + VkDeviceSize m_UsableSize; + uint32_t m_LevelCount; + + Node* m_Root; + struct { + Node* front; + Node* back; + } m_FreeList[MAX_LEVELS]; + // Number of nodes in the tree with type == TYPE_ALLOCATION. + size_t m_AllocationCount; + // Number of nodes in the tree with type == TYPE_FREE. + size_t m_FreeCount; + // This includes space wasted due to internal fragmentation. Doesn't include unusable size. + VkDeviceSize m_SumFreeSize; + + VkDeviceSize GetUnusableSize() const { return GetSize() - m_UsableSize; } + void DeleteNode(Node* node); + bool ValidateNode(ValidationContext& ctx, const Node* parent, const Node* curr, uint32_t level, VkDeviceSize levelNodeSize) const; + uint32_t AllocSizeToLevel(VkDeviceSize allocSize) const; + inline VkDeviceSize LevelToNodeSize(uint32_t level) const { return m_UsableSize >> level; } + // Alloc passed just for validation. Can be null. + void FreeAtOffset(VmaAllocation alloc, VkDeviceSize offset); + void CalcAllocationStatInfoNode(VmaStatInfo& outInfo, const Node* node, VkDeviceSize levelNodeSize) const; + // Adds node to the front of FreeList at given level. + // node->type must be FREE. + // node->free.prev, next can be undefined. + void AddToFreeListFront(uint32_t level, Node* node); + // Removes node from FreeList at given level. + // node->type must be FREE. + // node->free.prev, next stay untouched. + void RemoveFromFreeList(uint32_t level, Node* node); + +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMapNode(class VmaJsonWriter& json, const Node* node, VkDeviceSize levelNodeSize) const; +#endif +}; + +/* +Represents a single block of device memory (`VkDeviceMemory`) with all the +data about its regions (aka suballocations, #VmaAllocation), assigned and free. + +Thread-safety: This class must be externally synchronized. +*/ +class VmaDeviceMemoryBlock +{ + VMA_CLASS_NO_COPY(VmaDeviceMemoryBlock) +public: + VmaBlockMetadata* m_pMetadata; + + VmaDeviceMemoryBlock(VmaAllocator hAllocator); + + ~VmaDeviceMemoryBlock() + { + VMA_ASSERT(m_MapCount == 0 && "VkDeviceMemory block is being destroyed while it is still mapped."); + VMA_ASSERT(m_hMemory == VK_NULL_HANDLE); + } + + // Always call after construction. + void Init( + VmaAllocator hAllocator, + VmaPool hParentPool, + uint32_t newMemoryTypeIndex, + VkDeviceMemory newMemory, + VkDeviceSize newSize, + uint32_t id, + uint32_t algorithm); + // Always call before destruction. + void Destroy(VmaAllocator allocator); + + VmaPool GetParentPool() const { return m_hParentPool; } + VkDeviceMemory GetDeviceMemory() const { return m_hMemory; } + uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } + uint32_t GetId() const { return m_Id; } + void* GetMappedData() const { return m_pMappedData; } + + // Validates all data structures inside this object. If not valid, returns false. + bool Validate() const; + + VkResult CheckCorruption(VmaAllocator hAllocator); + + // ppData can be null. + VkResult Map(VmaAllocator hAllocator, uint32_t count, void** ppData); + void Unmap(VmaAllocator hAllocator, uint32_t count); + + VkResult WriteMagicValueAroundAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize); + VkResult ValidateMagicValueAroundAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize); + + VkResult BindBufferMemory( + const VmaAllocator hAllocator, + const VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkBuffer hBuffer, + const void* pNext); + VkResult BindImageMemory( + const VmaAllocator hAllocator, + const VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkImage hImage, + const void* pNext); + +private: + VmaPool m_hParentPool; // VK_NULL_HANDLE if not belongs to custom pool. + uint32_t m_MemoryTypeIndex; + uint32_t m_Id; + VkDeviceMemory m_hMemory; + + /* + Protects access to m_hMemory so it's not used by multiple threads simultaneously, e.g. vkMapMemory, vkBindBufferMemory. + Also protects m_MapCount, m_pMappedData. + Allocations, deallocations, any change in m_pMetadata is protected by parent's VmaBlockVector::m_Mutex. + */ + VMA_MUTEX m_Mutex; + uint32_t m_MapCount; + void* m_pMappedData; +}; + +struct VmaPointerLess +{ + bool operator()(const void* lhs, const void* rhs) const + { + return lhs < rhs; + } +}; + +struct VmaDefragmentationMove +{ + size_t srcBlockIndex; + size_t dstBlockIndex; + VkDeviceSize srcOffset; + VkDeviceSize dstOffset; + VkDeviceSize size; + VmaAllocation hAllocation; + VmaDeviceMemoryBlock* pSrcBlock; + VmaDeviceMemoryBlock* pDstBlock; +}; + +class VmaDefragmentationAlgorithm; + +/* +Sequence of VmaDeviceMemoryBlock. Represents memory blocks allocated for a specific +Vulkan memory type. + +Synchronized internally with a mutex. +*/ +struct VmaBlockVector +{ + VMA_CLASS_NO_COPY(VmaBlockVector) +public: + VmaBlockVector( + VmaAllocator hAllocator, + VmaPool hParentPool, + uint32_t memoryTypeIndex, + VkDeviceSize preferredBlockSize, + size_t minBlockCount, + size_t maxBlockCount, + VkDeviceSize bufferImageGranularity, + uint32_t frameInUseCount, + bool explicitBlockSize, + uint32_t algorithm); + ~VmaBlockVector(); + + VkResult CreateMinBlocks(); + + VmaAllocator GetAllocator() const { return m_hAllocator; } + VmaPool GetParentPool() const { return m_hParentPool; } + bool IsCustomPool() const { return m_hParentPool != VMA_NULL; } + uint32_t GetMemoryTypeIndex() const { return m_MemoryTypeIndex; } + VkDeviceSize GetPreferredBlockSize() const { return m_PreferredBlockSize; } + VkDeviceSize GetBufferImageGranularity() const { return m_BufferImageGranularity; } + uint32_t GetFrameInUseCount() const { return m_FrameInUseCount; } + uint32_t GetAlgorithm() const { return m_Algorithm; } + + void GetPoolStats(VmaPoolStats* pStats); + + bool IsEmpty(); + bool IsCorruptionDetectionEnabled() const; + + VkResult Allocate( + uint32_t currentFrameIndex, + VkDeviceSize size, + VkDeviceSize alignment, + const VmaAllocationCreateInfo& createInfo, + VmaSuballocationType suballocType, + size_t allocationCount, + VmaAllocation* pAllocations); + + void Free(const VmaAllocation hAllocation); + + // Adds statistics of this BlockVector to pStats. + void AddStats(VmaStats* pStats); + +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap(class VmaJsonWriter& json); +#endif + + void MakePoolAllocationsLost( + uint32_t currentFrameIndex, + size_t* pLostAllocationCount); + VkResult CheckCorruption(); + + // Saves results in pCtx->res. + void Defragment( + class VmaBlockVectorDefragmentationContext* pCtx, + VmaDefragmentationStats* pStats, VmaDefragmentationFlags flags, + VkDeviceSize& maxCpuBytesToMove, uint32_t& maxCpuAllocationsToMove, + VkDeviceSize& maxGpuBytesToMove, uint32_t& maxGpuAllocationsToMove, + VkCommandBuffer commandBuffer); + void DefragmentationEnd( + class VmaBlockVectorDefragmentationContext* pCtx, + uint32_t flags, + VmaDefragmentationStats* pStats); + + uint32_t ProcessDefragmentations( + class VmaBlockVectorDefragmentationContext *pCtx, + VmaDefragmentationPassMoveInfo* pMove, uint32_t maxMoves); + + void CommitDefragmentations( + class VmaBlockVectorDefragmentationContext *pCtx, + VmaDefragmentationStats* pStats); + + //////////////////////////////////////////////////////////////////////////////// + // To be used only while the m_Mutex is locked. Used during defragmentation. + + size_t GetBlockCount() const { return m_Blocks.size(); } + VmaDeviceMemoryBlock* GetBlock(size_t index) const { return m_Blocks[index]; } + size_t CalcAllocationCount() const; + bool IsBufferImageGranularityConflictPossible() const; + +private: + friend class VmaDefragmentationAlgorithm_Generic; + + const VmaAllocator m_hAllocator; + const VmaPool m_hParentPool; + const uint32_t m_MemoryTypeIndex; + const VkDeviceSize m_PreferredBlockSize; + const size_t m_MinBlockCount; + const size_t m_MaxBlockCount; + const VkDeviceSize m_BufferImageGranularity; + const uint32_t m_FrameInUseCount; + const bool m_ExplicitBlockSize; + const uint32_t m_Algorithm; + VMA_RW_MUTEX m_Mutex; + + /* There can be at most one allocation that is completely empty (except when minBlockCount > 0) - + a hysteresis to avoid pessimistic case of alternating creation and destruction of a VkDeviceMemory. */ + bool m_HasEmptyBlock; + // Incrementally sorted by sumFreeSize, ascending. + VmaVector< VmaDeviceMemoryBlock*, VmaStlAllocator > m_Blocks; + uint32_t m_NextBlockId; + + VkDeviceSize CalcMaxBlockSize() const; + + // Finds and removes given block from vector. + void Remove(VmaDeviceMemoryBlock* pBlock); + + // Performs single step in sorting m_Blocks. They may not be fully sorted + // after this call. + void IncrementallySortBlocks(); + + VkResult AllocatePage( + uint32_t currentFrameIndex, + VkDeviceSize size, + VkDeviceSize alignment, + const VmaAllocationCreateInfo& createInfo, + VmaSuballocationType suballocType, + VmaAllocation* pAllocation); + + // To be used only without CAN_MAKE_OTHER_LOST flag. + VkResult AllocateFromBlock( + VmaDeviceMemoryBlock* pBlock, + uint32_t currentFrameIndex, + VkDeviceSize size, + VkDeviceSize alignment, + VmaAllocationCreateFlags allocFlags, + void* pUserData, + VmaSuballocationType suballocType, + uint32_t strategy, + VmaAllocation* pAllocation); + + VkResult CreateBlock(VkDeviceSize blockSize, size_t* pNewBlockIndex); + + // Saves result to pCtx->res. + void ApplyDefragmentationMovesCpu( + class VmaBlockVectorDefragmentationContext* pDefragCtx, + const VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves); + // Saves result to pCtx->res. + void ApplyDefragmentationMovesGpu( + class VmaBlockVectorDefragmentationContext* pDefragCtx, + VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves, + VkCommandBuffer commandBuffer); + + /* + Used during defragmentation. pDefragmentationStats is optional. It's in/out + - updated with new data. + */ + void FreeEmptyBlocks(VmaDefragmentationStats* pDefragmentationStats); + + void UpdateHasEmptyBlock(); +}; + +struct VmaPool_T +{ + VMA_CLASS_NO_COPY(VmaPool_T) +public: + VmaBlockVector m_BlockVector; + + VmaPool_T( + VmaAllocator hAllocator, + const VmaPoolCreateInfo& createInfo, + VkDeviceSize preferredBlockSize); + ~VmaPool_T(); + + uint32_t GetId() const { return m_Id; } + void SetId(uint32_t id) { VMA_ASSERT(m_Id == 0); m_Id = id; } + + const char* GetName() const { return m_Name; } + void SetName(const char* pName); + +#if VMA_STATS_STRING_ENABLED + //void PrintDetailedMap(class VmaStringBuilder& sb); +#endif + +private: + uint32_t m_Id; + char* m_Name; +}; + +/* +Performs defragmentation: + +- Updates `pBlockVector->m_pMetadata`. +- Updates allocations by calling ChangeBlockAllocation() or ChangeOffset(). +- Does not move actual data, only returns requested moves as `moves`. +*/ +class VmaDefragmentationAlgorithm +{ + VMA_CLASS_NO_COPY(VmaDefragmentationAlgorithm) +public: + VmaDefragmentationAlgorithm( + VmaAllocator hAllocator, + VmaBlockVector* pBlockVector, + uint32_t currentFrameIndex) : + m_hAllocator(hAllocator), + m_pBlockVector(pBlockVector), + m_CurrentFrameIndex(currentFrameIndex) + { + } + virtual ~VmaDefragmentationAlgorithm() + { + } + + virtual void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged) = 0; + virtual void AddAll() = 0; + + virtual VkResult Defragment( + VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + VmaDefragmentationFlags flags) = 0; + + virtual VkDeviceSize GetBytesMoved() const = 0; + virtual uint32_t GetAllocationsMoved() const = 0; + +protected: + VmaAllocator const m_hAllocator; + VmaBlockVector* const m_pBlockVector; + const uint32_t m_CurrentFrameIndex; + + struct AllocationInfo + { + VmaAllocation m_hAllocation; + VkBool32* m_pChanged; + + AllocationInfo() : + m_hAllocation(VK_NULL_HANDLE), + m_pChanged(VMA_NULL) + { + } + AllocationInfo(VmaAllocation hAlloc, VkBool32* pChanged) : + m_hAllocation(hAlloc), + m_pChanged(pChanged) + { + } + }; +}; + +class VmaDefragmentationAlgorithm_Generic : public VmaDefragmentationAlgorithm +{ + VMA_CLASS_NO_COPY(VmaDefragmentationAlgorithm_Generic) +public: + VmaDefragmentationAlgorithm_Generic( + VmaAllocator hAllocator, + VmaBlockVector* pBlockVector, + uint32_t currentFrameIndex, + bool overlappingMoveSupported); + virtual ~VmaDefragmentationAlgorithm_Generic(); + + virtual void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged); + virtual void AddAll() { m_AllAllocations = true; } + + virtual VkResult Defragment( + VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + VmaDefragmentationFlags flags); + + virtual VkDeviceSize GetBytesMoved() const { return m_BytesMoved; } + virtual uint32_t GetAllocationsMoved() const { return m_AllocationsMoved; } + +private: + uint32_t m_AllocationCount; + bool m_AllAllocations; + + VkDeviceSize m_BytesMoved; + uint32_t m_AllocationsMoved; + + struct AllocationInfoSizeGreater + { + bool operator()(const AllocationInfo& lhs, const AllocationInfo& rhs) const + { + return lhs.m_hAllocation->GetSize() > rhs.m_hAllocation->GetSize(); + } + }; + + struct AllocationInfoOffsetGreater + { + bool operator()(const AllocationInfo& lhs, const AllocationInfo& rhs) const + { + return lhs.m_hAllocation->GetOffset() > rhs.m_hAllocation->GetOffset(); + } + }; + + struct BlockInfo + { + size_t m_OriginalBlockIndex; + VmaDeviceMemoryBlock* m_pBlock; + bool m_HasNonMovableAllocations; + VmaVector< AllocationInfo, VmaStlAllocator > m_Allocations; + + BlockInfo(const VkAllocationCallbacks* pAllocationCallbacks) : + m_OriginalBlockIndex(SIZE_MAX), + m_pBlock(VMA_NULL), + m_HasNonMovableAllocations(true), + m_Allocations(pAllocationCallbacks) + { + } + + void CalcHasNonMovableAllocations() + { + const size_t blockAllocCount = m_pBlock->m_pMetadata->GetAllocationCount(); + const size_t defragmentAllocCount = m_Allocations.size(); + m_HasNonMovableAllocations = blockAllocCount != defragmentAllocCount; + } + + void SortAllocationsBySizeDescending() + { + VMA_SORT(m_Allocations.begin(), m_Allocations.end(), AllocationInfoSizeGreater()); + } + + void SortAllocationsByOffsetDescending() + { + VMA_SORT(m_Allocations.begin(), m_Allocations.end(), AllocationInfoOffsetGreater()); + } + }; + + struct BlockPointerLess + { + bool operator()(const BlockInfo* pLhsBlockInfo, const VmaDeviceMemoryBlock* pRhsBlock) const + { + return pLhsBlockInfo->m_pBlock < pRhsBlock; + } + bool operator()(const BlockInfo* pLhsBlockInfo, const BlockInfo* pRhsBlockInfo) const + { + return pLhsBlockInfo->m_pBlock < pRhsBlockInfo->m_pBlock; + } + }; + + // 1. Blocks with some non-movable allocations go first. + // 2. Blocks with smaller sumFreeSize go first. + struct BlockInfoCompareMoveDestination + { + bool operator()(const BlockInfo* pLhsBlockInfo, const BlockInfo* pRhsBlockInfo) const + { + if(pLhsBlockInfo->m_HasNonMovableAllocations && !pRhsBlockInfo->m_HasNonMovableAllocations) + { + return true; + } + if(!pLhsBlockInfo->m_HasNonMovableAllocations && pRhsBlockInfo->m_HasNonMovableAllocations) + { + return false; + } + if(pLhsBlockInfo->m_pBlock->m_pMetadata->GetSumFreeSize() < pRhsBlockInfo->m_pBlock->m_pMetadata->GetSumFreeSize()) + { + return true; + } + return false; + } + }; + + typedef VmaVector< BlockInfo*, VmaStlAllocator > BlockInfoVector; + BlockInfoVector m_Blocks; + + VkResult DefragmentRound( + VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + bool freeOldAllocations); + + size_t CalcBlocksWithNonMovableCount() const; + + static bool MoveMakesSense( + size_t dstBlockIndex, VkDeviceSize dstOffset, + size_t srcBlockIndex, VkDeviceSize srcOffset); +}; + +class VmaDefragmentationAlgorithm_Fast : public VmaDefragmentationAlgorithm +{ + VMA_CLASS_NO_COPY(VmaDefragmentationAlgorithm_Fast) +public: + VmaDefragmentationAlgorithm_Fast( + VmaAllocator hAllocator, + VmaBlockVector* pBlockVector, + uint32_t currentFrameIndex, + bool overlappingMoveSupported); + virtual ~VmaDefragmentationAlgorithm_Fast(); + + virtual void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged) { ++m_AllocationCount; } + virtual void AddAll() { m_AllAllocations = true; } + + virtual VkResult Defragment( + VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + VmaDefragmentationFlags flags); + + virtual VkDeviceSize GetBytesMoved() const { return m_BytesMoved; } + virtual uint32_t GetAllocationsMoved() const { return m_AllocationsMoved; } + +private: + struct BlockInfo + { + size_t origBlockIndex; + }; + + class FreeSpaceDatabase + { + public: + FreeSpaceDatabase() + { + FreeSpace s = {}; + s.blockInfoIndex = SIZE_MAX; + for(size_t i = 0; i < MAX_COUNT; ++i) + { + m_FreeSpaces[i] = s; + } + } + + void Register(size_t blockInfoIndex, VkDeviceSize offset, VkDeviceSize size) + { + if(size < VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) + { + return; + } + + // Find first invalid or the smallest structure. + size_t bestIndex = SIZE_MAX; + for(size_t i = 0; i < MAX_COUNT; ++i) + { + // Empty structure. + if(m_FreeSpaces[i].blockInfoIndex == SIZE_MAX) + { + bestIndex = i; + break; + } + if(m_FreeSpaces[i].size < size && + (bestIndex == SIZE_MAX || m_FreeSpaces[bestIndex].size > m_FreeSpaces[i].size)) + { + bestIndex = i; + } + } + + if(bestIndex != SIZE_MAX) + { + m_FreeSpaces[bestIndex].blockInfoIndex = blockInfoIndex; + m_FreeSpaces[bestIndex].offset = offset; + m_FreeSpaces[bestIndex].size = size; + } + } + + bool Fetch(VkDeviceSize alignment, VkDeviceSize size, + size_t& outBlockInfoIndex, VkDeviceSize& outDstOffset) + { + size_t bestIndex = SIZE_MAX; + VkDeviceSize bestFreeSpaceAfter = 0; + for(size_t i = 0; i < MAX_COUNT; ++i) + { + // Structure is valid. + if(m_FreeSpaces[i].blockInfoIndex != SIZE_MAX) + { + const VkDeviceSize dstOffset = VmaAlignUp(m_FreeSpaces[i].offset, alignment); + // Allocation fits into this structure. + if(dstOffset + size <= m_FreeSpaces[i].offset + m_FreeSpaces[i].size) + { + const VkDeviceSize freeSpaceAfter = (m_FreeSpaces[i].offset + m_FreeSpaces[i].size) - + (dstOffset + size); + if(bestIndex == SIZE_MAX || freeSpaceAfter > bestFreeSpaceAfter) + { + bestIndex = i; + bestFreeSpaceAfter = freeSpaceAfter; + } + } + } + } + + if(bestIndex != SIZE_MAX) + { + outBlockInfoIndex = m_FreeSpaces[bestIndex].blockInfoIndex; + outDstOffset = VmaAlignUp(m_FreeSpaces[bestIndex].offset, alignment); + + if(bestFreeSpaceAfter >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) + { + // Leave this structure for remaining empty space. + const VkDeviceSize alignmentPlusSize = (outDstOffset - m_FreeSpaces[bestIndex].offset) + size; + m_FreeSpaces[bestIndex].offset += alignmentPlusSize; + m_FreeSpaces[bestIndex].size -= alignmentPlusSize; + } + else + { + // This structure becomes invalid. + m_FreeSpaces[bestIndex].blockInfoIndex = SIZE_MAX; + } + + return true; + } + + return false; + } + + private: + static const size_t MAX_COUNT = 4; + + struct FreeSpace + { + size_t blockInfoIndex; // SIZE_MAX means this structure is invalid. + VkDeviceSize offset; + VkDeviceSize size; + } m_FreeSpaces[MAX_COUNT]; + }; + + const bool m_OverlappingMoveSupported; + + uint32_t m_AllocationCount; + bool m_AllAllocations; + + VkDeviceSize m_BytesMoved; + uint32_t m_AllocationsMoved; + + VmaVector< BlockInfo, VmaStlAllocator > m_BlockInfos; + + void PreprocessMetadata(); + void PostprocessMetadata(); + void InsertSuballoc(VmaBlockMetadata_Generic* pMetadata, const VmaSuballocation& suballoc); +}; + +struct VmaBlockDefragmentationContext +{ + enum BLOCK_FLAG + { + BLOCK_FLAG_USED = 0x00000001, + }; + uint32_t flags; + VkBuffer hBuffer; +}; + +class VmaBlockVectorDefragmentationContext +{ + VMA_CLASS_NO_COPY(VmaBlockVectorDefragmentationContext) +public: + VkResult res; + bool mutexLocked; + VmaVector< VmaBlockDefragmentationContext, VmaStlAllocator > blockContexts; + VmaVector< VmaDefragmentationMove, VmaStlAllocator > defragmentationMoves; + uint32_t defragmentationMovesProcessed; + uint32_t defragmentationMovesCommitted; + bool hasDefragmentationPlan; + + VmaBlockVectorDefragmentationContext( + VmaAllocator hAllocator, + VmaPool hCustomPool, // Optional. + VmaBlockVector* pBlockVector, + uint32_t currFrameIndex); + ~VmaBlockVectorDefragmentationContext(); + + VmaPool GetCustomPool() const { return m_hCustomPool; } + VmaBlockVector* GetBlockVector() const { return m_pBlockVector; } + VmaDefragmentationAlgorithm* GetAlgorithm() const { return m_pAlgorithm; } + + void AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged); + void AddAll() { m_AllAllocations = true; } + + void Begin(bool overlappingMoveSupported, VmaDefragmentationFlags flags); + +private: + const VmaAllocator m_hAllocator; + // Null if not from custom pool. + const VmaPool m_hCustomPool; + // Redundant, for convenience not to fetch from m_hCustomPool->m_BlockVector or m_hAllocator->m_pBlockVectors. + VmaBlockVector* const m_pBlockVector; + const uint32_t m_CurrFrameIndex; + // Owner of this object. + VmaDefragmentationAlgorithm* m_pAlgorithm; + + struct AllocInfo + { + VmaAllocation hAlloc; + VkBool32* pChanged; + }; + // Used between constructor and Begin. + VmaVector< AllocInfo, VmaStlAllocator > m_Allocations; + bool m_AllAllocations; +}; + +struct VmaDefragmentationContext_T +{ +private: + VMA_CLASS_NO_COPY(VmaDefragmentationContext_T) +public: + VmaDefragmentationContext_T( + VmaAllocator hAllocator, + uint32_t currFrameIndex, + uint32_t flags, + VmaDefragmentationStats* pStats); + ~VmaDefragmentationContext_T(); + + void AddPools(uint32_t poolCount, const VmaPool* pPools); + void AddAllocations( + uint32_t allocationCount, + const VmaAllocation* pAllocations, + VkBool32* pAllocationsChanged); + + /* + Returns: + - `VK_SUCCESS` if succeeded and object can be destroyed immediately. + - `VK_NOT_READY` if succeeded but the object must remain alive until vmaDefragmentationEnd(). + - Negative value if error occured and object can be destroyed immediately. + */ + VkResult Defragment( + VkDeviceSize maxCpuBytesToMove, uint32_t maxCpuAllocationsToMove, + VkDeviceSize maxGpuBytesToMove, uint32_t maxGpuAllocationsToMove, + VkCommandBuffer commandBuffer, VmaDefragmentationStats* pStats, VmaDefragmentationFlags flags); + + VkResult DefragmentPassBegin(VmaDefragmentationPassInfo* pInfo); + VkResult DefragmentPassEnd(); + +private: + const VmaAllocator m_hAllocator; + const uint32_t m_CurrFrameIndex; + const uint32_t m_Flags; + VmaDefragmentationStats* const m_pStats; + + VkDeviceSize m_MaxCpuBytesToMove; + uint32_t m_MaxCpuAllocationsToMove; + VkDeviceSize m_MaxGpuBytesToMove; + uint32_t m_MaxGpuAllocationsToMove; + + // Owner of these objects. + VmaBlockVectorDefragmentationContext* m_DefaultPoolContexts[VK_MAX_MEMORY_TYPES]; + // Owner of these objects. + VmaVector< VmaBlockVectorDefragmentationContext*, VmaStlAllocator > m_CustomPoolContexts; +}; + +#if VMA_RECORDING_ENABLED + +class VmaRecorder +{ +public: + VmaRecorder(); + VkResult Init(const VmaRecordSettings& settings, bool useMutex); + void WriteConfiguration( + const VkPhysicalDeviceProperties& devProps, + const VkPhysicalDeviceMemoryProperties& memProps, + uint32_t vulkanApiVersion, + bool dedicatedAllocationExtensionEnabled, + bool bindMemory2ExtensionEnabled, + bool memoryBudgetExtensionEnabled, + bool deviceCoherentMemoryExtensionEnabled); + ~VmaRecorder(); + + void RecordCreateAllocator(uint32_t frameIndex); + void RecordDestroyAllocator(uint32_t frameIndex); + void RecordCreatePool(uint32_t frameIndex, + const VmaPoolCreateInfo& createInfo, + VmaPool pool); + void RecordDestroyPool(uint32_t frameIndex, VmaPool pool); + void RecordAllocateMemory(uint32_t frameIndex, + const VkMemoryRequirements& vkMemReq, + const VmaAllocationCreateInfo& createInfo, + VmaAllocation allocation); + void RecordAllocateMemoryPages(uint32_t frameIndex, + const VkMemoryRequirements& vkMemReq, + const VmaAllocationCreateInfo& createInfo, + uint64_t allocationCount, + const VmaAllocation* pAllocations); + void RecordAllocateMemoryForBuffer(uint32_t frameIndex, + const VkMemoryRequirements& vkMemReq, + bool requiresDedicatedAllocation, + bool prefersDedicatedAllocation, + const VmaAllocationCreateInfo& createInfo, + VmaAllocation allocation); + void RecordAllocateMemoryForImage(uint32_t frameIndex, + const VkMemoryRequirements& vkMemReq, + bool requiresDedicatedAllocation, + bool prefersDedicatedAllocation, + const VmaAllocationCreateInfo& createInfo, + VmaAllocation allocation); + void RecordFreeMemory(uint32_t frameIndex, + VmaAllocation allocation); + void RecordFreeMemoryPages(uint32_t frameIndex, + uint64_t allocationCount, + const VmaAllocation* pAllocations); + void RecordSetAllocationUserData(uint32_t frameIndex, + VmaAllocation allocation, + const void* pUserData); + void RecordCreateLostAllocation(uint32_t frameIndex, + VmaAllocation allocation); + void RecordMapMemory(uint32_t frameIndex, + VmaAllocation allocation); + void RecordUnmapMemory(uint32_t frameIndex, + VmaAllocation allocation); + void RecordFlushAllocation(uint32_t frameIndex, + VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size); + void RecordInvalidateAllocation(uint32_t frameIndex, + VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size); + void RecordCreateBuffer(uint32_t frameIndex, + const VkBufferCreateInfo& bufCreateInfo, + const VmaAllocationCreateInfo& allocCreateInfo, + VmaAllocation allocation); + void RecordCreateImage(uint32_t frameIndex, + const VkImageCreateInfo& imageCreateInfo, + const VmaAllocationCreateInfo& allocCreateInfo, + VmaAllocation allocation); + void RecordDestroyBuffer(uint32_t frameIndex, + VmaAllocation allocation); + void RecordDestroyImage(uint32_t frameIndex, + VmaAllocation allocation); + void RecordTouchAllocation(uint32_t frameIndex, + VmaAllocation allocation); + void RecordGetAllocationInfo(uint32_t frameIndex, + VmaAllocation allocation); + void RecordMakePoolAllocationsLost(uint32_t frameIndex, + VmaPool pool); + void RecordDefragmentationBegin(uint32_t frameIndex, + const VmaDefragmentationInfo2& info, + VmaDefragmentationContext ctx); + void RecordDefragmentationEnd(uint32_t frameIndex, + VmaDefragmentationContext ctx); + void RecordSetPoolName(uint32_t frameIndex, + VmaPool pool, + const char* name); + +private: + struct CallParams + { + uint32_t threadId; + double time; + }; + + class UserDataString + { + public: + UserDataString(VmaAllocationCreateFlags allocFlags, const void* pUserData); + const char* GetString() const { return m_Str; } + + private: + char m_PtrStr[17]; + const char* m_Str; + }; + + bool m_UseMutex; + VmaRecordFlags m_Flags; + FILE* m_File; + VMA_MUTEX m_FileMutex; + std::chrono::time_point m_RecordingStartTime; + + void GetBasicParams(CallParams& outParams); + + // T must be a pointer type, e.g. VmaAllocation, VmaPool. + template + void PrintPointerList(uint64_t count, const T* pItems) + { + if(count) + { + fprintf(m_File, "%p", pItems[0]); + for(uint64_t i = 1; i < count; ++i) + { + fprintf(m_File, " %p", pItems[i]); + } + } + } + + void PrintPointerList(uint64_t count, const VmaAllocation* pItems); + void Flush(); +}; + +#endif // #if VMA_RECORDING_ENABLED + +/* +Thread-safe wrapper over VmaPoolAllocator free list, for allocation of VmaAllocation_T objects. +*/ +class VmaAllocationObjectAllocator +{ + VMA_CLASS_NO_COPY(VmaAllocationObjectAllocator) +public: + VmaAllocationObjectAllocator(const VkAllocationCallbacks* pAllocationCallbacks); + + template VmaAllocation Allocate(Types... args); + void Free(VmaAllocation hAlloc); + +private: + VMA_MUTEX m_Mutex; + VmaPoolAllocator m_Allocator; +}; + +struct VmaCurrentBudgetData +{ + VMA_ATOMIC_UINT64 m_BlockBytes[VK_MAX_MEMORY_HEAPS]; + VMA_ATOMIC_UINT64 m_AllocationBytes[VK_MAX_MEMORY_HEAPS]; + +#if VMA_MEMORY_BUDGET + VMA_ATOMIC_UINT32 m_OperationsSinceBudgetFetch; + VMA_RW_MUTEX m_BudgetMutex; + uint64_t m_VulkanUsage[VK_MAX_MEMORY_HEAPS]; + uint64_t m_VulkanBudget[VK_MAX_MEMORY_HEAPS]; + uint64_t m_BlockBytesAtBudgetFetch[VK_MAX_MEMORY_HEAPS]; +#endif // #if VMA_MEMORY_BUDGET + + VmaCurrentBudgetData() + { + for(uint32_t heapIndex = 0; heapIndex < VK_MAX_MEMORY_HEAPS; ++heapIndex) + { + m_BlockBytes[heapIndex] = 0; + m_AllocationBytes[heapIndex] = 0; +#if VMA_MEMORY_BUDGET + m_VulkanUsage[heapIndex] = 0; + m_VulkanBudget[heapIndex] = 0; + m_BlockBytesAtBudgetFetch[heapIndex] = 0; +#endif + } + +#if VMA_MEMORY_BUDGET + m_OperationsSinceBudgetFetch = 0; +#endif + } + + void AddAllocation(uint32_t heapIndex, VkDeviceSize allocationSize) + { + m_AllocationBytes[heapIndex] += allocationSize; +#if VMA_MEMORY_BUDGET + ++m_OperationsSinceBudgetFetch; +#endif + } + + void RemoveAllocation(uint32_t heapIndex, VkDeviceSize allocationSize) + { + VMA_ASSERT(m_AllocationBytes[heapIndex] >= allocationSize); // DELME + m_AllocationBytes[heapIndex] -= allocationSize; +#if VMA_MEMORY_BUDGET + ++m_OperationsSinceBudgetFetch; +#endif + } +}; + +// Main allocator object. +struct VmaAllocator_T +{ + VMA_CLASS_NO_COPY(VmaAllocator_T) +public: + bool m_UseMutex; + uint32_t m_VulkanApiVersion; + bool m_UseKhrDedicatedAllocation; // Can be set only if m_VulkanApiVersion < VK_MAKE_VERSION(1, 1, 0). + bool m_UseKhrBindMemory2; // Can be set only if m_VulkanApiVersion < VK_MAKE_VERSION(1, 1, 0). + bool m_UseExtMemoryBudget; + bool m_UseAmdDeviceCoherentMemory; + bool m_UseKhrBufferDeviceAddress; + VkDevice m_hDevice; + VkInstance m_hInstance; + bool m_AllocationCallbacksSpecified; + VkAllocationCallbacks m_AllocationCallbacks; + VmaDeviceMemoryCallbacks m_DeviceMemoryCallbacks; + VmaAllocationObjectAllocator m_AllocationObjectAllocator; + + // Each bit (1 << i) is set if HeapSizeLimit is enabled for that heap, so cannot allocate more than the heap size. + uint32_t m_HeapSizeLimitMask; + + VkPhysicalDeviceProperties m_PhysicalDeviceProperties; + VkPhysicalDeviceMemoryProperties m_MemProps; + + // Default pools. + VmaBlockVector* m_pBlockVectors[VK_MAX_MEMORY_TYPES]; + + // Each vector is sorted by memory (handle value). + typedef VmaVector< VmaAllocation, VmaStlAllocator > AllocationVectorType; + AllocationVectorType* m_pDedicatedAllocations[VK_MAX_MEMORY_TYPES]; + VMA_RW_MUTEX m_DedicatedAllocationsMutex[VK_MAX_MEMORY_TYPES]; + + VmaCurrentBudgetData m_Budget; + + VmaAllocator_T(const VmaAllocatorCreateInfo* pCreateInfo); + VkResult Init(const VmaAllocatorCreateInfo* pCreateInfo); + ~VmaAllocator_T(); + + const VkAllocationCallbacks* GetAllocationCallbacks() const + { + return m_AllocationCallbacksSpecified ? &m_AllocationCallbacks : 0; + } + const VmaVulkanFunctions& GetVulkanFunctions() const + { + return m_VulkanFunctions; + } + + VkPhysicalDevice GetPhysicalDevice() const { return m_PhysicalDevice; } + + VkDeviceSize GetBufferImageGranularity() const + { + return VMA_MAX( + static_cast(VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY), + m_PhysicalDeviceProperties.limits.bufferImageGranularity); + } + + uint32_t GetMemoryHeapCount() const { return m_MemProps.memoryHeapCount; } + uint32_t GetMemoryTypeCount() const { return m_MemProps.memoryTypeCount; } + + uint32_t MemoryTypeIndexToHeapIndex(uint32_t memTypeIndex) const + { + VMA_ASSERT(memTypeIndex < m_MemProps.memoryTypeCount); + return m_MemProps.memoryTypes[memTypeIndex].heapIndex; + } + // True when specific memory type is HOST_VISIBLE but not HOST_COHERENT. + bool IsMemoryTypeNonCoherent(uint32_t memTypeIndex) const + { + return (m_MemProps.memoryTypes[memTypeIndex].propertyFlags & (VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT)) == + VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; + } + // Minimum alignment for all allocations in specific memory type. + VkDeviceSize GetMemoryTypeMinAlignment(uint32_t memTypeIndex) const + { + return IsMemoryTypeNonCoherent(memTypeIndex) ? + VMA_MAX((VkDeviceSize)VMA_DEBUG_ALIGNMENT, m_PhysicalDeviceProperties.limits.nonCoherentAtomSize) : + (VkDeviceSize)VMA_DEBUG_ALIGNMENT; + } + + bool IsIntegratedGpu() const + { + return m_PhysicalDeviceProperties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU; + } + + uint32_t GetGlobalMemoryTypeBits() const { return m_GlobalMemoryTypeBits; } + +#if VMA_RECORDING_ENABLED + VmaRecorder* GetRecorder() const { return m_pRecorder; } +#endif + + void GetBufferMemoryRequirements( + VkBuffer hBuffer, + VkMemoryRequirements& memReq, + bool& requiresDedicatedAllocation, + bool& prefersDedicatedAllocation) const; + void GetImageMemoryRequirements( + VkImage hImage, + VkMemoryRequirements& memReq, + bool& requiresDedicatedAllocation, + bool& prefersDedicatedAllocation) const; + + // Main allocation function. + VkResult AllocateMemory( + const VkMemoryRequirements& vkMemReq, + bool requiresDedicatedAllocation, + bool prefersDedicatedAllocation, + VkBuffer dedicatedBuffer, + VkBufferUsageFlags dedicatedBufferUsage, // UINT32_MAX when unknown. + VkImage dedicatedImage, + const VmaAllocationCreateInfo& createInfo, + VmaSuballocationType suballocType, + size_t allocationCount, + VmaAllocation* pAllocations); + + // Main deallocation function. + void FreeMemory( + size_t allocationCount, + const VmaAllocation* pAllocations); + + VkResult ResizeAllocation( + const VmaAllocation alloc, + VkDeviceSize newSize); + + void CalculateStats(VmaStats* pStats); + + void GetBudget( + VmaBudget* outBudget, uint32_t firstHeap, uint32_t heapCount); + +#if VMA_STATS_STRING_ENABLED + void PrintDetailedMap(class VmaJsonWriter& json); +#endif + + VkResult DefragmentationBegin( + const VmaDefragmentationInfo2& info, + VmaDefragmentationStats* pStats, + VmaDefragmentationContext* pContext); + VkResult DefragmentationEnd( + VmaDefragmentationContext context); + + VkResult DefragmentationPassBegin( + VmaDefragmentationPassInfo* pInfo, + VmaDefragmentationContext context); + VkResult DefragmentationPassEnd( + VmaDefragmentationContext context); + + void GetAllocationInfo(VmaAllocation hAllocation, VmaAllocationInfo* pAllocationInfo); + bool TouchAllocation(VmaAllocation hAllocation); + + VkResult CreatePool(const VmaPoolCreateInfo* pCreateInfo, VmaPool* pPool); + void DestroyPool(VmaPool pool); + void GetPoolStats(VmaPool pool, VmaPoolStats* pPoolStats); + + void SetCurrentFrameIndex(uint32_t frameIndex); + uint32_t GetCurrentFrameIndex() const { return m_CurrentFrameIndex.load(); } + + void MakePoolAllocationsLost( + VmaPool hPool, + size_t* pLostAllocationCount); + VkResult CheckPoolCorruption(VmaPool hPool); + VkResult CheckCorruption(uint32_t memoryTypeBits); + + void CreateLostAllocation(VmaAllocation* pAllocation); + + // Call to Vulkan function vkAllocateMemory with accompanying bookkeeping. + VkResult AllocateVulkanMemory(const VkMemoryAllocateInfo* pAllocateInfo, VkDeviceMemory* pMemory); + // Call to Vulkan function vkFreeMemory with accompanying bookkeeping. + void FreeVulkanMemory(uint32_t memoryType, VkDeviceSize size, VkDeviceMemory hMemory); + // Call to Vulkan function vkBindBufferMemory or vkBindBufferMemory2KHR. + VkResult BindVulkanBuffer( + VkDeviceMemory memory, + VkDeviceSize memoryOffset, + VkBuffer buffer, + const void* pNext); + // Call to Vulkan function vkBindImageMemory or vkBindImageMemory2KHR. + VkResult BindVulkanImage( + VkDeviceMemory memory, + VkDeviceSize memoryOffset, + VkImage image, + const void* pNext); + + VkResult Map(VmaAllocation hAllocation, void** ppData); + void Unmap(VmaAllocation hAllocation); + + VkResult BindBufferMemory( + VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkBuffer hBuffer, + const void* pNext); + VkResult BindImageMemory( + VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkImage hImage, + const void* pNext); + + VkResult FlushOrInvalidateAllocation( + VmaAllocation hAllocation, + VkDeviceSize offset, VkDeviceSize size, + VMA_CACHE_OPERATION op); + VkResult FlushOrInvalidateAllocations( + uint32_t allocationCount, + const VmaAllocation* allocations, + const VkDeviceSize* offsets, const VkDeviceSize* sizes, + VMA_CACHE_OPERATION op); + + void FillAllocation(const VmaAllocation hAllocation, uint8_t pattern); + + /* + Returns bit mask of memory types that can support defragmentation on GPU as + they support creation of required buffer for copy operations. + */ + uint32_t GetGpuDefragmentationMemoryTypeBits(); + +private: + VkDeviceSize m_PreferredLargeHeapBlockSize; + + VkPhysicalDevice m_PhysicalDevice; + VMA_ATOMIC_UINT32 m_CurrentFrameIndex; + VMA_ATOMIC_UINT32 m_GpuDefragmentationMemoryTypeBits; // UINT32_MAX means uninitialized. + + VMA_RW_MUTEX m_PoolsMutex; + // Protected by m_PoolsMutex. Sorted by pointer value. + VmaVector > m_Pools; + uint32_t m_NextPoolId; + + VmaVulkanFunctions m_VulkanFunctions; + + // Global bit mask AND-ed with any memoryTypeBits to disallow certain memory types. + uint32_t m_GlobalMemoryTypeBits; + +#if VMA_RECORDING_ENABLED + VmaRecorder* m_pRecorder; +#endif + + void ImportVulkanFunctions(const VmaVulkanFunctions* pVulkanFunctions); + +#if VMA_STATIC_VULKAN_FUNCTIONS == 1 + void ImportVulkanFunctions_Static(); +#endif + + void ImportVulkanFunctions_Custom(const VmaVulkanFunctions* pVulkanFunctions); + +#if VMA_DYNAMIC_VULKAN_FUNCTIONS == 1 + void ImportVulkanFunctions_Dynamic(); +#endif + + void ValidateVulkanFunctions(); + + VkDeviceSize CalcPreferredBlockSize(uint32_t memTypeIndex); + + VkResult AllocateMemoryOfType( + VkDeviceSize size, + VkDeviceSize alignment, + bool dedicatedAllocation, + VkBuffer dedicatedBuffer, + VkBufferUsageFlags dedicatedBufferUsage, + VkImage dedicatedImage, + const VmaAllocationCreateInfo& createInfo, + uint32_t memTypeIndex, + VmaSuballocationType suballocType, + size_t allocationCount, + VmaAllocation* pAllocations); + + // Helper function only to be used inside AllocateDedicatedMemory. + VkResult AllocateDedicatedMemoryPage( + VkDeviceSize size, + VmaSuballocationType suballocType, + uint32_t memTypeIndex, + const VkMemoryAllocateInfo& allocInfo, + bool map, + bool isUserDataString, + void* pUserData, + VmaAllocation* pAllocation); + + // Allocates and registers new VkDeviceMemory specifically for dedicated allocations. + VkResult AllocateDedicatedMemory( + VkDeviceSize size, + VmaSuballocationType suballocType, + uint32_t memTypeIndex, + bool withinBudget, + bool map, + bool isUserDataString, + void* pUserData, + VkBuffer dedicatedBuffer, + VkBufferUsageFlags dedicatedBufferUsage, + VkImage dedicatedImage, + size_t allocationCount, + VmaAllocation* pAllocations); + + void FreeDedicatedMemory(const VmaAllocation allocation); + + /* + Calculates and returns bit mask of memory types that can support defragmentation + on GPU as they support creation of required buffer for copy operations. + */ + uint32_t CalculateGpuDefragmentationMemoryTypeBits() const; + + uint32_t CalculateGlobalMemoryTypeBits() const; + + bool GetFlushOrInvalidateRange( + VmaAllocation allocation, + VkDeviceSize offset, VkDeviceSize size, + VkMappedMemoryRange& outRange) const; + +#if VMA_MEMORY_BUDGET + void UpdateVulkanBudget(); +#endif // #if VMA_MEMORY_BUDGET +}; + +//////////////////////////////////////////////////////////////////////////////// +// Memory allocation #2 after VmaAllocator_T definition + +static void* VmaMalloc(VmaAllocator hAllocator, size_t size, size_t alignment) +{ + return VmaMalloc(&hAllocator->m_AllocationCallbacks, size, alignment); +} + +static void VmaFree(VmaAllocator hAllocator, void* ptr) +{ + VmaFree(&hAllocator->m_AllocationCallbacks, ptr); +} + +template +static T* VmaAllocate(VmaAllocator hAllocator) +{ + return (T*)VmaMalloc(hAllocator, sizeof(T), VMA_ALIGN_OF(T)); +} + +template +static T* VmaAllocateArray(VmaAllocator hAllocator, size_t count) +{ + return (T*)VmaMalloc(hAllocator, sizeof(T) * count, VMA_ALIGN_OF(T)); +} + +template +static void vma_delete(VmaAllocator hAllocator, T* ptr) +{ + if(ptr != VMA_NULL) + { + ptr->~T(); + VmaFree(hAllocator, ptr); + } +} + +template +static void vma_delete_array(VmaAllocator hAllocator, T* ptr, size_t count) +{ + if(ptr != VMA_NULL) + { + for(size_t i = count; i--; ) + ptr[i].~T(); + VmaFree(hAllocator, ptr); + } +} + +//////////////////////////////////////////////////////////////////////////////// +// VmaStringBuilder + +#if VMA_STATS_STRING_ENABLED + +class VmaStringBuilder +{ +public: + VmaStringBuilder(VmaAllocator alloc) : m_Data(VmaStlAllocator(alloc->GetAllocationCallbacks())) { } + size_t GetLength() const { return m_Data.size(); } + const char* GetData() const { return m_Data.data(); } + + void Add(char ch) { m_Data.push_back(ch); } + void Add(const char* pStr); + void AddNewLine() { Add('\n'); } + void AddNumber(uint32_t num); + void AddNumber(uint64_t num); + void AddPointer(const void* ptr); + +private: + VmaVector< char, VmaStlAllocator > m_Data; +}; + +void VmaStringBuilder::Add(const char* pStr) +{ + const size_t strLen = strlen(pStr); + if(strLen > 0) + { + const size_t oldCount = m_Data.size(); + m_Data.resize(oldCount + strLen); + memcpy(m_Data.data() + oldCount, pStr, strLen); + } +} + +void VmaStringBuilder::AddNumber(uint32_t num) +{ + char buf[11]; + buf[10] = '\0'; + char *p = &buf[10]; + do + { + *--p = '0' + (num % 10); + num /= 10; + } + while(num); + Add(p); +} + +void VmaStringBuilder::AddNumber(uint64_t num) +{ + char buf[21]; + buf[20] = '\0'; + char *p = &buf[20]; + do + { + *--p = '0' + (num % 10); + num /= 10; + } + while(num); + Add(p); +} + +void VmaStringBuilder::AddPointer(const void* ptr) +{ + char buf[21]; + VmaPtrToStr(buf, sizeof(buf), ptr); + Add(buf); +} + +#endif // #if VMA_STATS_STRING_ENABLED + +//////////////////////////////////////////////////////////////////////////////// +// VmaJsonWriter + +#if VMA_STATS_STRING_ENABLED + +class VmaJsonWriter +{ + VMA_CLASS_NO_COPY(VmaJsonWriter) +public: + VmaJsonWriter(const VkAllocationCallbacks* pAllocationCallbacks, VmaStringBuilder& sb); + ~VmaJsonWriter(); + + void BeginObject(bool singleLine = false); + void EndObject(); + + void BeginArray(bool singleLine = false); + void EndArray(); + + void WriteString(const char* pStr); + void BeginString(const char* pStr = VMA_NULL); + void ContinueString(const char* pStr); + void ContinueString(uint32_t n); + void ContinueString(uint64_t n); + void ContinueString_Pointer(const void* ptr); + void EndString(const char* pStr = VMA_NULL); + + void WriteNumber(uint32_t n); + void WriteNumber(uint64_t n); + void WriteBool(bool b); + void WriteNull(); + +private: + static const char* const INDENT; + + enum COLLECTION_TYPE + { + COLLECTION_TYPE_OBJECT, + COLLECTION_TYPE_ARRAY, + }; + struct StackItem + { + COLLECTION_TYPE type; + uint32_t valueCount; + bool singleLineMode; + }; + + VmaStringBuilder& m_SB; + VmaVector< StackItem, VmaStlAllocator > m_Stack; + bool m_InsideString; + + void BeginValue(bool isString); + void WriteIndent(bool oneLess = false); +}; + +const char* const VmaJsonWriter::INDENT = " "; + +VmaJsonWriter::VmaJsonWriter(const VkAllocationCallbacks* pAllocationCallbacks, VmaStringBuilder& sb) : + m_SB(sb), + m_Stack(VmaStlAllocator(pAllocationCallbacks)), + m_InsideString(false) +{ +} + +VmaJsonWriter::~VmaJsonWriter() +{ + VMA_ASSERT(!m_InsideString); + VMA_ASSERT(m_Stack.empty()); +} + +void VmaJsonWriter::BeginObject(bool singleLine) +{ + VMA_ASSERT(!m_InsideString); + + BeginValue(false); + m_SB.Add('{'); + + StackItem item; + item.type = COLLECTION_TYPE_OBJECT; + item.valueCount = 0; + item.singleLineMode = singleLine; + m_Stack.push_back(item); +} + +void VmaJsonWriter::EndObject() +{ + VMA_ASSERT(!m_InsideString); + + WriteIndent(true); + m_SB.Add('}'); + + VMA_ASSERT(!m_Stack.empty() && m_Stack.back().type == COLLECTION_TYPE_OBJECT); + m_Stack.pop_back(); +} + +void VmaJsonWriter::BeginArray(bool singleLine) +{ + VMA_ASSERT(!m_InsideString); + + BeginValue(false); + m_SB.Add('['); + + StackItem item; + item.type = COLLECTION_TYPE_ARRAY; + item.valueCount = 0; + item.singleLineMode = singleLine; + m_Stack.push_back(item); +} + +void VmaJsonWriter::EndArray() +{ + VMA_ASSERT(!m_InsideString); + + WriteIndent(true); + m_SB.Add(']'); + + VMA_ASSERT(!m_Stack.empty() && m_Stack.back().type == COLLECTION_TYPE_ARRAY); + m_Stack.pop_back(); +} + +void VmaJsonWriter::WriteString(const char* pStr) +{ + BeginString(pStr); + EndString(); +} + +void VmaJsonWriter::BeginString(const char* pStr) +{ + VMA_ASSERT(!m_InsideString); + + BeginValue(true); + m_SB.Add('"'); + m_InsideString = true; + if(pStr != VMA_NULL && pStr[0] != '\0') + { + ContinueString(pStr); + } +} + +void VmaJsonWriter::ContinueString(const char* pStr) +{ + VMA_ASSERT(m_InsideString); + + const size_t strLen = strlen(pStr); + for(size_t i = 0; i < strLen; ++i) + { + char ch = pStr[i]; + if(ch == '\\') + { + m_SB.Add("\\\\"); + } + else if(ch == '"') + { + m_SB.Add("\\\""); + } + else if(ch >= 32) + { + m_SB.Add(ch); + } + else switch(ch) + { + case '\b': + m_SB.Add("\\b"); + break; + case '\f': + m_SB.Add("\\f"); + break; + case '\n': + m_SB.Add("\\n"); + break; + case '\r': + m_SB.Add("\\r"); + break; + case '\t': + m_SB.Add("\\t"); + break; + default: + VMA_ASSERT(0 && "Character not currently supported."); + break; + } + } +} + +void VmaJsonWriter::ContinueString(uint32_t n) +{ + VMA_ASSERT(m_InsideString); + m_SB.AddNumber(n); +} + +void VmaJsonWriter::ContinueString(uint64_t n) +{ + VMA_ASSERT(m_InsideString); + m_SB.AddNumber(n); +} + +void VmaJsonWriter::ContinueString_Pointer(const void* ptr) +{ + VMA_ASSERT(m_InsideString); + m_SB.AddPointer(ptr); +} + +void VmaJsonWriter::EndString(const char* pStr) +{ + VMA_ASSERT(m_InsideString); + if(pStr != VMA_NULL && pStr[0] != '\0') + { + ContinueString(pStr); + } + m_SB.Add('"'); + m_InsideString = false; +} + +void VmaJsonWriter::WriteNumber(uint32_t n) +{ + VMA_ASSERT(!m_InsideString); + BeginValue(false); + m_SB.AddNumber(n); +} + +void VmaJsonWriter::WriteNumber(uint64_t n) +{ + VMA_ASSERT(!m_InsideString); + BeginValue(false); + m_SB.AddNumber(n); +} + +void VmaJsonWriter::WriteBool(bool b) +{ + VMA_ASSERT(!m_InsideString); + BeginValue(false); + m_SB.Add(b ? "true" : "false"); +} + +void VmaJsonWriter::WriteNull() +{ + VMA_ASSERT(!m_InsideString); + BeginValue(false); + m_SB.Add("null"); +} + +void VmaJsonWriter::BeginValue(bool isString) +{ + if(!m_Stack.empty()) + { + StackItem& currItem = m_Stack.back(); + if(currItem.type == COLLECTION_TYPE_OBJECT && + currItem.valueCount % 2 == 0) + { + VMA_ASSERT(isString); + } + + if(currItem.type == COLLECTION_TYPE_OBJECT && + currItem.valueCount % 2 != 0) + { + m_SB.Add(": "); + } + else if(currItem.valueCount > 0) + { + m_SB.Add(", "); + WriteIndent(); + } + else + { + WriteIndent(); + } + ++currItem.valueCount; + } +} + +void VmaJsonWriter::WriteIndent(bool oneLess) +{ + if(!m_Stack.empty() && !m_Stack.back().singleLineMode) + { + m_SB.AddNewLine(); + + size_t count = m_Stack.size(); + if(count > 0 && oneLess) + { + --count; + } + for(size_t i = 0; i < count; ++i) + { + m_SB.Add(INDENT); + } + } +} + +#endif // #if VMA_STATS_STRING_ENABLED + +//////////////////////////////////////////////////////////////////////////////// + +void VmaAllocation_T::SetUserData(VmaAllocator hAllocator, void* pUserData) +{ + if(IsUserDataString()) + { + VMA_ASSERT(pUserData == VMA_NULL || pUserData != m_pUserData); + + FreeUserDataString(hAllocator); + + if(pUserData != VMA_NULL) + { + m_pUserData = VmaCreateStringCopy(hAllocator->GetAllocationCallbacks(), (const char*)pUserData); + } + } + else + { + m_pUserData = pUserData; + } +} + +void VmaAllocation_T::ChangeBlockAllocation( + VmaAllocator hAllocator, + VmaDeviceMemoryBlock* block, + VkDeviceSize offset) +{ + VMA_ASSERT(block != VMA_NULL); + VMA_ASSERT(m_Type == ALLOCATION_TYPE_BLOCK); + + // Move mapping reference counter from old block to new block. + if(block != m_BlockAllocation.m_Block) + { + uint32_t mapRefCount = m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP; + if(IsPersistentMap()) + ++mapRefCount; + m_BlockAllocation.m_Block->Unmap(hAllocator, mapRefCount); + block->Map(hAllocator, mapRefCount, VMA_NULL); + } + + m_BlockAllocation.m_Block = block; + m_BlockAllocation.m_Offset = offset; +} + +void VmaAllocation_T::ChangeOffset(VkDeviceSize newOffset) +{ + VMA_ASSERT(m_Type == ALLOCATION_TYPE_BLOCK); + m_BlockAllocation.m_Offset = newOffset; +} + +VkDeviceSize VmaAllocation_T::GetOffset() const +{ + switch(m_Type) + { + case ALLOCATION_TYPE_BLOCK: + return m_BlockAllocation.m_Offset; + case ALLOCATION_TYPE_DEDICATED: + return 0; + default: + VMA_ASSERT(0); + return 0; + } +} + +VkDeviceMemory VmaAllocation_T::GetMemory() const +{ + switch(m_Type) + { + case ALLOCATION_TYPE_BLOCK: + return m_BlockAllocation.m_Block->GetDeviceMemory(); + case ALLOCATION_TYPE_DEDICATED: + return m_DedicatedAllocation.m_hMemory; + default: + VMA_ASSERT(0); + return VK_NULL_HANDLE; + } +} + +void* VmaAllocation_T::GetMappedData() const +{ + switch(m_Type) + { + case ALLOCATION_TYPE_BLOCK: + if(m_MapCount != 0) + { + void* pBlockData = m_BlockAllocation.m_Block->GetMappedData(); + VMA_ASSERT(pBlockData != VMA_NULL); + return (char*)pBlockData + m_BlockAllocation.m_Offset; + } + else + { + return VMA_NULL; + } + break; + case ALLOCATION_TYPE_DEDICATED: + VMA_ASSERT((m_DedicatedAllocation.m_pMappedData != VMA_NULL) == (m_MapCount != 0)); + return m_DedicatedAllocation.m_pMappedData; + default: + VMA_ASSERT(0); + return VMA_NULL; + } +} + +bool VmaAllocation_T::CanBecomeLost() const +{ + switch(m_Type) + { + case ALLOCATION_TYPE_BLOCK: + return m_BlockAllocation.m_CanBecomeLost; + case ALLOCATION_TYPE_DEDICATED: + return false; + default: + VMA_ASSERT(0); + return false; + } +} + +bool VmaAllocation_T::MakeLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) +{ + VMA_ASSERT(CanBecomeLost()); + + /* + Warning: This is a carefully designed algorithm. + Do not modify unless you really know what you're doing :) + */ + uint32_t localLastUseFrameIndex = GetLastUseFrameIndex(); + for(;;) + { + if(localLastUseFrameIndex == VMA_FRAME_INDEX_LOST) + { + VMA_ASSERT(0); + return false; + } + else if(localLastUseFrameIndex + frameInUseCount >= currentFrameIndex) + { + return false; + } + else // Last use time earlier than current time. + { + if(CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, VMA_FRAME_INDEX_LOST)) + { + // Setting hAllocation.LastUseFrameIndex atomic to VMA_FRAME_INDEX_LOST is enough to mark it as LOST. + // Calling code just needs to unregister this allocation in owning VmaDeviceMemoryBlock. + return true; + } + } + } +} + +#if VMA_STATS_STRING_ENABLED + +// Correspond to values of enum VmaSuballocationType. +static const char* VMA_SUBALLOCATION_TYPE_NAMES[] = { + "FREE", + "UNKNOWN", + "BUFFER", + "IMAGE_UNKNOWN", + "IMAGE_LINEAR", + "IMAGE_OPTIMAL", +}; + +void VmaAllocation_T::PrintParameters(class VmaJsonWriter& json) const +{ + json.WriteString("Type"); + json.WriteString(VMA_SUBALLOCATION_TYPE_NAMES[m_SuballocationType]); + + json.WriteString("Size"); + json.WriteNumber(m_Size); + + if(m_pUserData != VMA_NULL) + { + json.WriteString("UserData"); + if(IsUserDataString()) + { + json.WriteString((const char*)m_pUserData); + } + else + { + json.BeginString(); + json.ContinueString_Pointer(m_pUserData); + json.EndString(); + } + } + + json.WriteString("CreationFrameIndex"); + json.WriteNumber(m_CreationFrameIndex); + + json.WriteString("LastUseFrameIndex"); + json.WriteNumber(GetLastUseFrameIndex()); + + if(m_BufferImageUsage != 0) + { + json.WriteString("Usage"); + json.WriteNumber(m_BufferImageUsage); + } +} + +#endif + +void VmaAllocation_T::FreeUserDataString(VmaAllocator hAllocator) +{ + VMA_ASSERT(IsUserDataString()); + VmaFreeString(hAllocator->GetAllocationCallbacks(), (char*)m_pUserData); + m_pUserData = VMA_NULL; +} + +void VmaAllocation_T::BlockAllocMap() +{ + VMA_ASSERT(GetType() == ALLOCATION_TYPE_BLOCK); + + if((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) < 0x7F) + { + ++m_MapCount; + } + else + { + VMA_ASSERT(0 && "Allocation mapped too many times simultaneously."); + } +} + +void VmaAllocation_T::BlockAllocUnmap() +{ + VMA_ASSERT(GetType() == ALLOCATION_TYPE_BLOCK); + + if((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) != 0) + { + --m_MapCount; + } + else + { + VMA_ASSERT(0 && "Unmapping allocation not previously mapped."); + } +} + +VkResult VmaAllocation_T::DedicatedAllocMap(VmaAllocator hAllocator, void** ppData) +{ + VMA_ASSERT(GetType() == ALLOCATION_TYPE_DEDICATED); + + if(m_MapCount != 0) + { + if((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) < 0x7F) + { + VMA_ASSERT(m_DedicatedAllocation.m_pMappedData != VMA_NULL); + *ppData = m_DedicatedAllocation.m_pMappedData; + ++m_MapCount; + return VK_SUCCESS; + } + else + { + VMA_ASSERT(0 && "Dedicated allocation mapped too many times simultaneously."); + return VK_ERROR_MEMORY_MAP_FAILED; + } + } + else + { + VkResult result = (*hAllocator->GetVulkanFunctions().vkMapMemory)( + hAllocator->m_hDevice, + m_DedicatedAllocation.m_hMemory, + 0, // offset + VK_WHOLE_SIZE, + 0, // flags + ppData); + if(result == VK_SUCCESS) + { + m_DedicatedAllocation.m_pMappedData = *ppData; + m_MapCount = 1; + } + return result; + } +} + +void VmaAllocation_T::DedicatedAllocUnmap(VmaAllocator hAllocator) +{ + VMA_ASSERT(GetType() == ALLOCATION_TYPE_DEDICATED); + + if((m_MapCount & ~MAP_COUNT_FLAG_PERSISTENT_MAP) != 0) + { + --m_MapCount; + if(m_MapCount == 0) + { + m_DedicatedAllocation.m_pMappedData = VMA_NULL; + (*hAllocator->GetVulkanFunctions().vkUnmapMemory)( + hAllocator->m_hDevice, + m_DedicatedAllocation.m_hMemory); + } + } + else + { + VMA_ASSERT(0 && "Unmapping dedicated allocation not previously mapped."); + } +} + +#if VMA_STATS_STRING_ENABLED + +static void VmaPrintStatInfo(VmaJsonWriter& json, const VmaStatInfo& stat) +{ + json.BeginObject(); + + json.WriteString("Blocks"); + json.WriteNumber(stat.blockCount); + + json.WriteString("Allocations"); + json.WriteNumber(stat.allocationCount); + + json.WriteString("UnusedRanges"); + json.WriteNumber(stat.unusedRangeCount); + + json.WriteString("UsedBytes"); + json.WriteNumber(stat.usedBytes); + + json.WriteString("UnusedBytes"); + json.WriteNumber(stat.unusedBytes); + + if(stat.allocationCount > 1) + { + json.WriteString("AllocationSize"); + json.BeginObject(true); + json.WriteString("Min"); + json.WriteNumber(stat.allocationSizeMin); + json.WriteString("Avg"); + json.WriteNumber(stat.allocationSizeAvg); + json.WriteString("Max"); + json.WriteNumber(stat.allocationSizeMax); + json.EndObject(); + } + + if(stat.unusedRangeCount > 1) + { + json.WriteString("UnusedRangeSize"); + json.BeginObject(true); + json.WriteString("Min"); + json.WriteNumber(stat.unusedRangeSizeMin); + json.WriteString("Avg"); + json.WriteNumber(stat.unusedRangeSizeAvg); + json.WriteString("Max"); + json.WriteNumber(stat.unusedRangeSizeMax); + json.EndObject(); + } + + json.EndObject(); +} + +#endif // #if VMA_STATS_STRING_ENABLED + +struct VmaSuballocationItemSizeLess +{ + bool operator()( + const VmaSuballocationList::iterator lhs, + const VmaSuballocationList::iterator rhs) const + { + return lhs->size < rhs->size; + } + bool operator()( + const VmaSuballocationList::iterator lhs, + VkDeviceSize rhsSize) const + { + return lhs->size < rhsSize; + } +}; + + +//////////////////////////////////////////////////////////////////////////////// +// class VmaBlockMetadata + +VmaBlockMetadata::VmaBlockMetadata(VmaAllocator hAllocator) : + m_Size(0), + m_pAllocationCallbacks(hAllocator->GetAllocationCallbacks()) +{ +} + +#if VMA_STATS_STRING_ENABLED + +void VmaBlockMetadata::PrintDetailedMap_Begin(class VmaJsonWriter& json, + VkDeviceSize unusedBytes, + size_t allocationCount, + size_t unusedRangeCount) const +{ + json.BeginObject(); + + json.WriteString("TotalBytes"); + json.WriteNumber(GetSize()); + + json.WriteString("UnusedBytes"); + json.WriteNumber(unusedBytes); + + json.WriteString("Allocations"); + json.WriteNumber((uint64_t)allocationCount); + + json.WriteString("UnusedRanges"); + json.WriteNumber((uint64_t)unusedRangeCount); + + json.WriteString("Suballocations"); + json.BeginArray(); +} + +void VmaBlockMetadata::PrintDetailedMap_Allocation(class VmaJsonWriter& json, + VkDeviceSize offset, + VmaAllocation hAllocation) const +{ + json.BeginObject(true); + + json.WriteString("Offset"); + json.WriteNumber(offset); + + hAllocation->PrintParameters(json); + + json.EndObject(); +} + +void VmaBlockMetadata::PrintDetailedMap_UnusedRange(class VmaJsonWriter& json, + VkDeviceSize offset, + VkDeviceSize size) const +{ + json.BeginObject(true); + + json.WriteString("Offset"); + json.WriteNumber(offset); + + json.WriteString("Type"); + json.WriteString(VMA_SUBALLOCATION_TYPE_NAMES[VMA_SUBALLOCATION_TYPE_FREE]); + + json.WriteString("Size"); + json.WriteNumber(size); + + json.EndObject(); +} + +void VmaBlockMetadata::PrintDetailedMap_End(class VmaJsonWriter& json) const +{ + json.EndArray(); + json.EndObject(); +} + +#endif // #if VMA_STATS_STRING_ENABLED + +//////////////////////////////////////////////////////////////////////////////// +// class VmaBlockMetadata_Generic + +VmaBlockMetadata_Generic::VmaBlockMetadata_Generic(VmaAllocator hAllocator) : + VmaBlockMetadata(hAllocator), + m_FreeCount(0), + m_SumFreeSize(0), + m_Suballocations(VmaStlAllocator(hAllocator->GetAllocationCallbacks())), + m_FreeSuballocationsBySize(VmaStlAllocator(hAllocator->GetAllocationCallbacks())) +{ +} + +VmaBlockMetadata_Generic::~VmaBlockMetadata_Generic() +{ +} + +void VmaBlockMetadata_Generic::Init(VkDeviceSize size) +{ + VmaBlockMetadata::Init(size); + + m_FreeCount = 1; + m_SumFreeSize = size; + + VmaSuballocation suballoc = {}; + suballoc.offset = 0; + suballoc.size = size; + suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + suballoc.hAllocation = VK_NULL_HANDLE; + + VMA_ASSERT(size > VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER); + m_Suballocations.push_back(suballoc); + VmaSuballocationList::iterator suballocItem = m_Suballocations.end(); + --suballocItem; + m_FreeSuballocationsBySize.push_back(suballocItem); +} + +bool VmaBlockMetadata_Generic::Validate() const +{ + VMA_VALIDATE(!m_Suballocations.empty()); + + // Expected offset of new suballocation as calculated from previous ones. + VkDeviceSize calculatedOffset = 0; + // Expected number of free suballocations as calculated from traversing their list. + uint32_t calculatedFreeCount = 0; + // Expected sum size of free suballocations as calculated from traversing their list. + VkDeviceSize calculatedSumFreeSize = 0; + // Expected number of free suballocations that should be registered in + // m_FreeSuballocationsBySize calculated from traversing their list. + size_t freeSuballocationsToRegister = 0; + // True if previous visited suballocation was free. + bool prevFree = false; + + for(VmaSuballocationList::const_iterator suballocItem = m_Suballocations.cbegin(); + suballocItem != m_Suballocations.cend(); + ++suballocItem) + { + const VmaSuballocation& subAlloc = *suballocItem; + + // Actual offset of this suballocation doesn't match expected one. + VMA_VALIDATE(subAlloc.offset == calculatedOffset); + + const bool currFree = (subAlloc.type == VMA_SUBALLOCATION_TYPE_FREE); + // Two adjacent free suballocations are invalid. They should be merged. + VMA_VALIDATE(!prevFree || !currFree); + + VMA_VALIDATE(currFree == (subAlloc.hAllocation == VK_NULL_HANDLE)); + + if(currFree) + { + calculatedSumFreeSize += subAlloc.size; + ++calculatedFreeCount; + if(subAlloc.size >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) + { + ++freeSuballocationsToRegister; + } + + // Margin required between allocations - every free space must be at least that large. + VMA_VALIDATE(subAlloc.size >= VMA_DEBUG_MARGIN); + } + else + { + VMA_VALIDATE(subAlloc.hAllocation->GetOffset() == subAlloc.offset); + VMA_VALIDATE(subAlloc.hAllocation->GetSize() == subAlloc.size); + + // Margin required between allocations - previous allocation must be free. + VMA_VALIDATE(VMA_DEBUG_MARGIN == 0 || prevFree); + } + + calculatedOffset += subAlloc.size; + prevFree = currFree; + } + + // Number of free suballocations registered in m_FreeSuballocationsBySize doesn't + // match expected one. + VMA_VALIDATE(m_FreeSuballocationsBySize.size() == freeSuballocationsToRegister); + + VkDeviceSize lastSize = 0; + for(size_t i = 0; i < m_FreeSuballocationsBySize.size(); ++i) + { + VmaSuballocationList::iterator suballocItem = m_FreeSuballocationsBySize[i]; + + // Only free suballocations can be registered in m_FreeSuballocationsBySize. + VMA_VALIDATE(suballocItem->type == VMA_SUBALLOCATION_TYPE_FREE); + // They must be sorted by size ascending. + VMA_VALIDATE(suballocItem->size >= lastSize); + + lastSize = suballocItem->size; + } + + // Check if totals match calculacted values. + VMA_VALIDATE(ValidateFreeSuballocationList()); + VMA_VALIDATE(calculatedOffset == GetSize()); + VMA_VALIDATE(calculatedSumFreeSize == m_SumFreeSize); + VMA_VALIDATE(calculatedFreeCount == m_FreeCount); + + return true; +} + +VkDeviceSize VmaBlockMetadata_Generic::GetUnusedRangeSizeMax() const +{ + if(!m_FreeSuballocationsBySize.empty()) + { + return m_FreeSuballocationsBySize.back()->size; + } + else + { + return 0; + } +} + +bool VmaBlockMetadata_Generic::IsEmpty() const +{ + return (m_Suballocations.size() == 1) && (m_FreeCount == 1); +} + +void VmaBlockMetadata_Generic::CalcAllocationStatInfo(VmaStatInfo& outInfo) const +{ + outInfo.blockCount = 1; + + const uint32_t rangeCount = (uint32_t)m_Suballocations.size(); + outInfo.allocationCount = rangeCount - m_FreeCount; + outInfo.unusedRangeCount = m_FreeCount; + + outInfo.unusedBytes = m_SumFreeSize; + outInfo.usedBytes = GetSize() - outInfo.unusedBytes; + + outInfo.allocationSizeMin = UINT64_MAX; + outInfo.allocationSizeMax = 0; + outInfo.unusedRangeSizeMin = UINT64_MAX; + outInfo.unusedRangeSizeMax = 0; + + for(VmaSuballocationList::const_iterator suballocItem = m_Suballocations.cbegin(); + suballocItem != m_Suballocations.cend(); + ++suballocItem) + { + const VmaSuballocation& suballoc = *suballocItem; + if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + { + outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); + outInfo.allocationSizeMax = VMA_MAX(outInfo.allocationSizeMax, suballoc.size); + } + else + { + outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, suballoc.size); + outInfo.unusedRangeSizeMax = VMA_MAX(outInfo.unusedRangeSizeMax, suballoc.size); + } + } +} + +void VmaBlockMetadata_Generic::AddPoolStats(VmaPoolStats& inoutStats) const +{ + const uint32_t rangeCount = (uint32_t)m_Suballocations.size(); + + inoutStats.size += GetSize(); + inoutStats.unusedSize += m_SumFreeSize; + inoutStats.allocationCount += rangeCount - m_FreeCount; + inoutStats.unusedRangeCount += m_FreeCount; + inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, GetUnusedRangeSizeMax()); +} + +#if VMA_STATS_STRING_ENABLED + +void VmaBlockMetadata_Generic::PrintDetailedMap(class VmaJsonWriter& json) const +{ + PrintDetailedMap_Begin(json, + m_SumFreeSize, // unusedBytes + m_Suballocations.size() - (size_t)m_FreeCount, // allocationCount + m_FreeCount); // unusedRangeCount + + size_t i = 0; + for(VmaSuballocationList::const_iterator suballocItem = m_Suballocations.cbegin(); + suballocItem != m_Suballocations.cend(); + ++suballocItem, ++i) + { + if(suballocItem->type == VMA_SUBALLOCATION_TYPE_FREE) + { + PrintDetailedMap_UnusedRange(json, suballocItem->offset, suballocItem->size); + } + else + { + PrintDetailedMap_Allocation(json, suballocItem->offset, suballocItem->hAllocation); + } + } + + PrintDetailedMap_End(json); +} + +#endif // #if VMA_STATS_STRING_ENABLED + +bool VmaBlockMetadata_Generic::CreateAllocationRequest( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) +{ + VMA_ASSERT(allocSize > 0); + VMA_ASSERT(!upperAddress); + VMA_ASSERT(allocType != VMA_SUBALLOCATION_TYPE_FREE); + VMA_ASSERT(pAllocationRequest != VMA_NULL); + VMA_HEAVY_ASSERT(Validate()); + + pAllocationRequest->type = VmaAllocationRequestType::Normal; + + // There is not enough total free space in this block to fullfill the request: Early return. + if(canMakeOtherLost == false && + m_SumFreeSize < allocSize + 2 * VMA_DEBUG_MARGIN) + { + return false; + } + + // New algorithm, efficiently searching freeSuballocationsBySize. + const size_t freeSuballocCount = m_FreeSuballocationsBySize.size(); + if(freeSuballocCount > 0) + { + if(strategy == VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT) + { + // Find first free suballocation with size not less than allocSize + 2 * VMA_DEBUG_MARGIN. + VmaSuballocationList::iterator* const it = VmaBinaryFindFirstNotLess( + m_FreeSuballocationsBySize.data(), + m_FreeSuballocationsBySize.data() + freeSuballocCount, + allocSize + 2 * VMA_DEBUG_MARGIN, + VmaSuballocationItemSizeLess()); + size_t index = it - m_FreeSuballocationsBySize.data(); + for(; index < freeSuballocCount; ++index) + { + if(CheckAllocation( + currentFrameIndex, + frameInUseCount, + bufferImageGranularity, + allocSize, + allocAlignment, + allocType, + m_FreeSuballocationsBySize[index], + false, // canMakeOtherLost + &pAllocationRequest->offset, + &pAllocationRequest->itemsToMakeLostCount, + &pAllocationRequest->sumFreeSize, + &pAllocationRequest->sumItemSize)) + { + pAllocationRequest->item = m_FreeSuballocationsBySize[index]; + return true; + } + } + } + else if(strategy == VMA_ALLOCATION_INTERNAL_STRATEGY_MIN_OFFSET) + { + for(VmaSuballocationList::iterator it = m_Suballocations.begin(); + it != m_Suballocations.end(); + ++it) + { + if(it->type == VMA_SUBALLOCATION_TYPE_FREE && CheckAllocation( + currentFrameIndex, + frameInUseCount, + bufferImageGranularity, + allocSize, + allocAlignment, + allocType, + it, + false, // canMakeOtherLost + &pAllocationRequest->offset, + &pAllocationRequest->itemsToMakeLostCount, + &pAllocationRequest->sumFreeSize, + &pAllocationRequest->sumItemSize)) + { + pAllocationRequest->item = it; + return true; + } + } + } + else // WORST_FIT, FIRST_FIT + { + // Search staring from biggest suballocations. + for(size_t index = freeSuballocCount; index--; ) + { + if(CheckAllocation( + currentFrameIndex, + frameInUseCount, + bufferImageGranularity, + allocSize, + allocAlignment, + allocType, + m_FreeSuballocationsBySize[index], + false, // canMakeOtherLost + &pAllocationRequest->offset, + &pAllocationRequest->itemsToMakeLostCount, + &pAllocationRequest->sumFreeSize, + &pAllocationRequest->sumItemSize)) + { + pAllocationRequest->item = m_FreeSuballocationsBySize[index]; + return true; + } + } + } + } + + if(canMakeOtherLost) + { + // Brute-force algorithm. TODO: Come up with something better. + + bool found = false; + VmaAllocationRequest tmpAllocRequest = {}; + tmpAllocRequest.type = VmaAllocationRequestType::Normal; + for(VmaSuballocationList::iterator suballocIt = m_Suballocations.begin(); + suballocIt != m_Suballocations.end(); + ++suballocIt) + { + if(suballocIt->type == VMA_SUBALLOCATION_TYPE_FREE || + suballocIt->hAllocation->CanBecomeLost()) + { + if(CheckAllocation( + currentFrameIndex, + frameInUseCount, + bufferImageGranularity, + allocSize, + allocAlignment, + allocType, + suballocIt, + canMakeOtherLost, + &tmpAllocRequest.offset, + &tmpAllocRequest.itemsToMakeLostCount, + &tmpAllocRequest.sumFreeSize, + &tmpAllocRequest.sumItemSize)) + { + if(strategy == VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT) + { + *pAllocationRequest = tmpAllocRequest; + pAllocationRequest->item = suballocIt; + break; + } + if(!found || tmpAllocRequest.CalcCost() < pAllocationRequest->CalcCost()) + { + *pAllocationRequest = tmpAllocRequest; + pAllocationRequest->item = suballocIt; + found = true; + } + } + } + } + + return found; + } + + return false; +} + +bool VmaBlockMetadata_Generic::MakeRequestedAllocationsLost( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VmaAllocationRequest* pAllocationRequest) +{ + VMA_ASSERT(pAllocationRequest && pAllocationRequest->type == VmaAllocationRequestType::Normal); + + while(pAllocationRequest->itemsToMakeLostCount > 0) + { + if(pAllocationRequest->item->type == VMA_SUBALLOCATION_TYPE_FREE) + { + ++pAllocationRequest->item; + } + VMA_ASSERT(pAllocationRequest->item != m_Suballocations.end()); + VMA_ASSERT(pAllocationRequest->item->hAllocation != VK_NULL_HANDLE); + VMA_ASSERT(pAllocationRequest->item->hAllocation->CanBecomeLost()); + if(pAllocationRequest->item->hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) + { + pAllocationRequest->item = FreeSuballocation(pAllocationRequest->item); + --pAllocationRequest->itemsToMakeLostCount; + } + else + { + return false; + } + } + + VMA_HEAVY_ASSERT(Validate()); + VMA_ASSERT(pAllocationRequest->item != m_Suballocations.end()); + VMA_ASSERT(pAllocationRequest->item->type == VMA_SUBALLOCATION_TYPE_FREE); + + return true; +} + +uint32_t VmaBlockMetadata_Generic::MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) +{ + uint32_t lostAllocationCount = 0; + for(VmaSuballocationList::iterator it = m_Suballocations.begin(); + it != m_Suballocations.end(); + ++it) + { + if(it->type != VMA_SUBALLOCATION_TYPE_FREE && + it->hAllocation->CanBecomeLost() && + it->hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) + { + it = FreeSuballocation(it); + ++lostAllocationCount; + } + } + return lostAllocationCount; +} + +VkResult VmaBlockMetadata_Generic::CheckCorruption(const void* pBlockData) +{ + for(VmaSuballocationList::iterator it = m_Suballocations.begin(); + it != m_Suballocations.end(); + ++it) + { + if(it->type != VMA_SUBALLOCATION_TYPE_FREE) + { + if(!VmaValidateMagicValue(pBlockData, it->offset - VMA_DEBUG_MARGIN)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED BEFORE VALIDATED ALLOCATION!"); + return VK_ERROR_VALIDATION_FAILED_EXT; + } + if(!VmaValidateMagicValue(pBlockData, it->offset + it->size)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER VALIDATED ALLOCATION!"); + return VK_ERROR_VALIDATION_FAILED_EXT; + } + } + } + + return VK_SUCCESS; +} + +void VmaBlockMetadata_Generic::Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + VkDeviceSize allocSize, + VmaAllocation hAllocation) +{ + VMA_ASSERT(request.type == VmaAllocationRequestType::Normal); + VMA_ASSERT(request.item != m_Suballocations.end()); + VmaSuballocation& suballoc = *request.item; + // Given suballocation is a free block. + VMA_ASSERT(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); + // Given offset is inside this suballocation. + VMA_ASSERT(request.offset >= suballoc.offset); + const VkDeviceSize paddingBegin = request.offset - suballoc.offset; + VMA_ASSERT(suballoc.size >= paddingBegin + allocSize); + const VkDeviceSize paddingEnd = suballoc.size - paddingBegin - allocSize; + + // Unregister this free suballocation from m_FreeSuballocationsBySize and update + // it to become used. + UnregisterFreeSuballocation(request.item); + + suballoc.offset = request.offset; + suballoc.size = allocSize; + suballoc.type = type; + suballoc.hAllocation = hAllocation; + + // If there are any free bytes remaining at the end, insert new free suballocation after current one. + if(paddingEnd) + { + VmaSuballocation paddingSuballoc = {}; + paddingSuballoc.offset = request.offset + allocSize; + paddingSuballoc.size = paddingEnd; + paddingSuballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + VmaSuballocationList::iterator next = request.item; + ++next; + const VmaSuballocationList::iterator paddingEndItem = + m_Suballocations.insert(next, paddingSuballoc); + RegisterFreeSuballocation(paddingEndItem); + } + + // If there are any free bytes remaining at the beginning, insert new free suballocation before current one. + if(paddingBegin) + { + VmaSuballocation paddingSuballoc = {}; + paddingSuballoc.offset = request.offset - paddingBegin; + paddingSuballoc.size = paddingBegin; + paddingSuballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + const VmaSuballocationList::iterator paddingBeginItem = + m_Suballocations.insert(request.item, paddingSuballoc); + RegisterFreeSuballocation(paddingBeginItem); + } + + // Update totals. + m_FreeCount = m_FreeCount - 1; + if(paddingBegin > 0) + { + ++m_FreeCount; + } + if(paddingEnd > 0) + { + ++m_FreeCount; + } + m_SumFreeSize -= allocSize; +} + +void VmaBlockMetadata_Generic::Free(const VmaAllocation allocation) +{ + for(VmaSuballocationList::iterator suballocItem = m_Suballocations.begin(); + suballocItem != m_Suballocations.end(); + ++suballocItem) + { + VmaSuballocation& suballoc = *suballocItem; + if(suballoc.hAllocation == allocation) + { + FreeSuballocation(suballocItem); + VMA_HEAVY_ASSERT(Validate()); + return; + } + } + VMA_ASSERT(0 && "Not found!"); +} + +void VmaBlockMetadata_Generic::FreeAtOffset(VkDeviceSize offset) +{ + for(VmaSuballocationList::iterator suballocItem = m_Suballocations.begin(); + suballocItem != m_Suballocations.end(); + ++suballocItem) + { + VmaSuballocation& suballoc = *suballocItem; + if(suballoc.offset == offset) + { + FreeSuballocation(suballocItem); + return; + } + } + VMA_ASSERT(0 && "Not found!"); +} + +bool VmaBlockMetadata_Generic::ValidateFreeSuballocationList() const +{ + VkDeviceSize lastSize = 0; + for(size_t i = 0, count = m_FreeSuballocationsBySize.size(); i < count; ++i) + { + const VmaSuballocationList::iterator it = m_FreeSuballocationsBySize[i]; + + VMA_VALIDATE(it->type == VMA_SUBALLOCATION_TYPE_FREE); + VMA_VALIDATE(it->size >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER); + VMA_VALIDATE(it->size >= lastSize); + lastSize = it->size; + } + return true; +} + +bool VmaBlockMetadata_Generic::CheckAllocation( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + VmaSuballocationList::const_iterator suballocItem, + bool canMakeOtherLost, + VkDeviceSize* pOffset, + size_t* itemsToMakeLostCount, + VkDeviceSize* pSumFreeSize, + VkDeviceSize* pSumItemSize) const +{ + VMA_ASSERT(allocSize > 0); + VMA_ASSERT(allocType != VMA_SUBALLOCATION_TYPE_FREE); + VMA_ASSERT(suballocItem != m_Suballocations.cend()); + VMA_ASSERT(pOffset != VMA_NULL); + + *itemsToMakeLostCount = 0; + *pSumFreeSize = 0; + *pSumItemSize = 0; + + if(canMakeOtherLost) + { + if(suballocItem->type == VMA_SUBALLOCATION_TYPE_FREE) + { + *pSumFreeSize = suballocItem->size; + } + else + { + if(suballocItem->hAllocation->CanBecomeLost() && + suballocItem->hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) + { + ++*itemsToMakeLostCount; + *pSumItemSize = suballocItem->size; + } + else + { + return false; + } + } + + // Remaining size is too small for this request: Early return. + if(GetSize() - suballocItem->offset < allocSize) + { + return false; + } + + // Start from offset equal to beginning of this suballocation. + *pOffset = suballocItem->offset; + + // Apply VMA_DEBUG_MARGIN at the beginning. + if(VMA_DEBUG_MARGIN > 0) + { + *pOffset += VMA_DEBUG_MARGIN; + } + + // Apply alignment. + *pOffset = VmaAlignUp(*pOffset, allocAlignment); + + // Check previous suballocations for BufferImageGranularity conflicts. + // Make bigger alignment if necessary. + if(bufferImageGranularity > 1) + { + bool bufferImageGranularityConflict = false; + VmaSuballocationList::const_iterator prevSuballocItem = suballocItem; + while(prevSuballocItem != m_Suballocations.cbegin()) + { + --prevSuballocItem; + const VmaSuballocation& prevSuballoc = *prevSuballocItem; + if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, *pOffset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) + { + bufferImageGranularityConflict = true; + break; + } + } + else + // Already on previous page. + break; + } + if(bufferImageGranularityConflict) + { + *pOffset = VmaAlignUp(*pOffset, bufferImageGranularity); + } + } + + // Now that we have final *pOffset, check if we are past suballocItem. + // If yes, return false - this function should be called for another suballocItem as starting point. + if(*pOffset >= suballocItem->offset + suballocItem->size) + { + return false; + } + + // Calculate padding at the beginning based on current offset. + const VkDeviceSize paddingBegin = *pOffset - suballocItem->offset; + + // Calculate required margin at the end. + const VkDeviceSize requiredEndMargin = VMA_DEBUG_MARGIN; + + const VkDeviceSize totalSize = paddingBegin + allocSize + requiredEndMargin; + // Another early return check. + if(suballocItem->offset + totalSize > GetSize()) + { + return false; + } + + // Advance lastSuballocItem until desired size is reached. + // Update itemsToMakeLostCount. + VmaSuballocationList::const_iterator lastSuballocItem = suballocItem; + if(totalSize > suballocItem->size) + { + VkDeviceSize remainingSize = totalSize - suballocItem->size; + while(remainingSize > 0) + { + ++lastSuballocItem; + if(lastSuballocItem == m_Suballocations.cend()) + { + return false; + } + if(lastSuballocItem->type == VMA_SUBALLOCATION_TYPE_FREE) + { + *pSumFreeSize += lastSuballocItem->size; + } + else + { + VMA_ASSERT(lastSuballocItem->hAllocation != VK_NULL_HANDLE); + if(lastSuballocItem->hAllocation->CanBecomeLost() && + lastSuballocItem->hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) + { + ++*itemsToMakeLostCount; + *pSumItemSize += lastSuballocItem->size; + } + else + { + return false; + } + } + remainingSize = (lastSuballocItem->size < remainingSize) ? + remainingSize - lastSuballocItem->size : 0; + } + } + + // Check next suballocations for BufferImageGranularity conflicts. + // If conflict exists, we must mark more allocations lost or fail. + if(bufferImageGranularity > 1) + { + VmaSuballocationList::const_iterator nextSuballocItem = lastSuballocItem; + ++nextSuballocItem; + while(nextSuballocItem != m_Suballocations.cend()) + { + const VmaSuballocation& nextSuballoc = *nextSuballocItem; + if(VmaBlocksOnSamePage(*pOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) + { + VMA_ASSERT(nextSuballoc.hAllocation != VK_NULL_HANDLE); + if(nextSuballoc.hAllocation->CanBecomeLost() && + nextSuballoc.hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) + { + ++*itemsToMakeLostCount; + } + else + { + return false; + } + } + } + else + { + // Already on next page. + break; + } + ++nextSuballocItem; + } + } + } + else + { + const VmaSuballocation& suballoc = *suballocItem; + VMA_ASSERT(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); + + *pSumFreeSize = suballoc.size; + + // Size of this suballocation is too small for this request: Early return. + if(suballoc.size < allocSize) + { + return false; + } + + // Start from offset equal to beginning of this suballocation. + *pOffset = suballoc.offset; + + // Apply VMA_DEBUG_MARGIN at the beginning. + if(VMA_DEBUG_MARGIN > 0) + { + *pOffset += VMA_DEBUG_MARGIN; + } + + // Apply alignment. + *pOffset = VmaAlignUp(*pOffset, allocAlignment); + + // Check previous suballocations for BufferImageGranularity conflicts. + // Make bigger alignment if necessary. + if(bufferImageGranularity > 1) + { + bool bufferImageGranularityConflict = false; + VmaSuballocationList::const_iterator prevSuballocItem = suballocItem; + while(prevSuballocItem != m_Suballocations.cbegin()) + { + --prevSuballocItem; + const VmaSuballocation& prevSuballoc = *prevSuballocItem; + if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, *pOffset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) + { + bufferImageGranularityConflict = true; + break; + } + } + else + // Already on previous page. + break; + } + if(bufferImageGranularityConflict) + { + *pOffset = VmaAlignUp(*pOffset, bufferImageGranularity); + } + } + + // Calculate padding at the beginning based on current offset. + const VkDeviceSize paddingBegin = *pOffset - suballoc.offset; + + // Calculate required margin at the end. + const VkDeviceSize requiredEndMargin = VMA_DEBUG_MARGIN; + + // Fail if requested size plus margin before and after is bigger than size of this suballocation. + if(paddingBegin + allocSize + requiredEndMargin > suballoc.size) + { + return false; + } + + // Check next suballocations for BufferImageGranularity conflicts. + // If conflict exists, allocation cannot be made here. + if(bufferImageGranularity > 1) + { + VmaSuballocationList::const_iterator nextSuballocItem = suballocItem; + ++nextSuballocItem; + while(nextSuballocItem != m_Suballocations.cend()) + { + const VmaSuballocation& nextSuballoc = *nextSuballocItem; + if(VmaBlocksOnSamePage(*pOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) + { + return false; + } + } + else + { + // Already on next page. + break; + } + ++nextSuballocItem; + } + } + } + + // All tests passed: Success. pOffset is already filled. + return true; +} + +void VmaBlockMetadata_Generic::MergeFreeWithNext(VmaSuballocationList::iterator item) +{ + VMA_ASSERT(item != m_Suballocations.end()); + VMA_ASSERT(item->type == VMA_SUBALLOCATION_TYPE_FREE); + + VmaSuballocationList::iterator nextItem = item; + ++nextItem; + VMA_ASSERT(nextItem != m_Suballocations.end()); + VMA_ASSERT(nextItem->type == VMA_SUBALLOCATION_TYPE_FREE); + + item->size += nextItem->size; + --m_FreeCount; + m_Suballocations.erase(nextItem); +} + +VmaSuballocationList::iterator VmaBlockMetadata_Generic::FreeSuballocation(VmaSuballocationList::iterator suballocItem) +{ + // Change this suballocation to be marked as free. + VmaSuballocation& suballoc = *suballocItem; + suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + suballoc.hAllocation = VK_NULL_HANDLE; + + // Update totals. + ++m_FreeCount; + m_SumFreeSize += suballoc.size; + + // Merge with previous and/or next suballocation if it's also free. + bool mergeWithNext = false; + bool mergeWithPrev = false; + + VmaSuballocationList::iterator nextItem = suballocItem; + ++nextItem; + if((nextItem != m_Suballocations.end()) && (nextItem->type == VMA_SUBALLOCATION_TYPE_FREE)) + { + mergeWithNext = true; + } + + VmaSuballocationList::iterator prevItem = suballocItem; + if(suballocItem != m_Suballocations.begin()) + { + --prevItem; + if(prevItem->type == VMA_SUBALLOCATION_TYPE_FREE) + { + mergeWithPrev = true; + } + } + + if(mergeWithNext) + { + UnregisterFreeSuballocation(nextItem); + MergeFreeWithNext(suballocItem); + } + + if(mergeWithPrev) + { + UnregisterFreeSuballocation(prevItem); + MergeFreeWithNext(prevItem); + RegisterFreeSuballocation(prevItem); + return prevItem; + } + else + { + RegisterFreeSuballocation(suballocItem); + return suballocItem; + } +} + +void VmaBlockMetadata_Generic::RegisterFreeSuballocation(VmaSuballocationList::iterator item) +{ + VMA_ASSERT(item->type == VMA_SUBALLOCATION_TYPE_FREE); + VMA_ASSERT(item->size > 0); + + // You may want to enable this validation at the beginning or at the end of + // this function, depending on what do you want to check. + VMA_HEAVY_ASSERT(ValidateFreeSuballocationList()); + + if(item->size >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) + { + if(m_FreeSuballocationsBySize.empty()) + { + m_FreeSuballocationsBySize.push_back(item); + } + else + { + VmaVectorInsertSorted(m_FreeSuballocationsBySize, item); + } + } + + //VMA_HEAVY_ASSERT(ValidateFreeSuballocationList()); +} + + +void VmaBlockMetadata_Generic::UnregisterFreeSuballocation(VmaSuballocationList::iterator item) +{ + VMA_ASSERT(item->type == VMA_SUBALLOCATION_TYPE_FREE); + VMA_ASSERT(item->size > 0); + + // You may want to enable this validation at the beginning or at the end of + // this function, depending on what do you want to check. + VMA_HEAVY_ASSERT(ValidateFreeSuballocationList()); + + if(item->size >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) + { + VmaSuballocationList::iterator* const it = VmaBinaryFindFirstNotLess( + m_FreeSuballocationsBySize.data(), + m_FreeSuballocationsBySize.data() + m_FreeSuballocationsBySize.size(), + item, + VmaSuballocationItemSizeLess()); + for(size_t index = it - m_FreeSuballocationsBySize.data(); + index < m_FreeSuballocationsBySize.size(); + ++index) + { + if(m_FreeSuballocationsBySize[index] == item) + { + VmaVectorRemove(m_FreeSuballocationsBySize, index); + return; + } + VMA_ASSERT((m_FreeSuballocationsBySize[index]->size == item->size) && "Not found."); + } + VMA_ASSERT(0 && "Not found."); + } + + //VMA_HEAVY_ASSERT(ValidateFreeSuballocationList()); +} + +bool VmaBlockMetadata_Generic::IsBufferImageGranularityConflictPossible( + VkDeviceSize bufferImageGranularity, + VmaSuballocationType& inOutPrevSuballocType) const +{ + if(bufferImageGranularity == 1 || IsEmpty()) + { + return false; + } + + VkDeviceSize minAlignment = VK_WHOLE_SIZE; + bool typeConflictFound = false; + for(VmaSuballocationList::const_iterator it = m_Suballocations.cbegin(); + it != m_Suballocations.cend(); + ++it) + { + const VmaSuballocationType suballocType = it->type; + if(suballocType != VMA_SUBALLOCATION_TYPE_FREE) + { + minAlignment = VMA_MIN(minAlignment, it->hAllocation->GetAlignment()); + if(VmaIsBufferImageGranularityConflict(inOutPrevSuballocType, suballocType)) + { + typeConflictFound = true; + } + inOutPrevSuballocType = suballocType; + } + } + + return typeConflictFound || minAlignment >= bufferImageGranularity; +} + +//////////////////////////////////////////////////////////////////////////////// +// class VmaBlockMetadata_Linear + +VmaBlockMetadata_Linear::VmaBlockMetadata_Linear(VmaAllocator hAllocator) : + VmaBlockMetadata(hAllocator), + m_SumFreeSize(0), + m_Suballocations0(VmaStlAllocator(hAllocator->GetAllocationCallbacks())), + m_Suballocations1(VmaStlAllocator(hAllocator->GetAllocationCallbacks())), + m_1stVectorIndex(0), + m_2ndVectorMode(SECOND_VECTOR_EMPTY), + m_1stNullItemsBeginCount(0), + m_1stNullItemsMiddleCount(0), + m_2ndNullItemsCount(0) +{ +} + +VmaBlockMetadata_Linear::~VmaBlockMetadata_Linear() +{ +} + +void VmaBlockMetadata_Linear::Init(VkDeviceSize size) +{ + VmaBlockMetadata::Init(size); + m_SumFreeSize = size; +} + +bool VmaBlockMetadata_Linear::Validate() const +{ + const SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + const SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + + VMA_VALIDATE(suballocations2nd.empty() == (m_2ndVectorMode == SECOND_VECTOR_EMPTY)); + VMA_VALIDATE(!suballocations1st.empty() || + suballocations2nd.empty() || + m_2ndVectorMode != SECOND_VECTOR_RING_BUFFER); + + if(!suballocations1st.empty()) + { + // Null item at the beginning should be accounted into m_1stNullItemsBeginCount. + VMA_VALIDATE(suballocations1st[m_1stNullItemsBeginCount].hAllocation != VK_NULL_HANDLE); + // Null item at the end should be just pop_back(). + VMA_VALIDATE(suballocations1st.back().hAllocation != VK_NULL_HANDLE); + } + if(!suballocations2nd.empty()) + { + // Null item at the end should be just pop_back(). + VMA_VALIDATE(suballocations2nd.back().hAllocation != VK_NULL_HANDLE); + } + + VMA_VALIDATE(m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount <= suballocations1st.size()); + VMA_VALIDATE(m_2ndNullItemsCount <= suballocations2nd.size()); + + VkDeviceSize sumUsedSize = 0; + const size_t suballoc1stCount = suballocations1st.size(); + VkDeviceSize offset = VMA_DEBUG_MARGIN; + + if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + const size_t suballoc2ndCount = suballocations2nd.size(); + size_t nullItem2ndCount = 0; + for(size_t i = 0; i < suballoc2ndCount; ++i) + { + const VmaSuballocation& suballoc = suballocations2nd[i]; + const bool currFree = (suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); + + VMA_VALIDATE(currFree == (suballoc.hAllocation == VK_NULL_HANDLE)); + VMA_VALIDATE(suballoc.offset >= offset); + + if(!currFree) + { + VMA_VALIDATE(suballoc.hAllocation->GetOffset() == suballoc.offset); + VMA_VALIDATE(suballoc.hAllocation->GetSize() == suballoc.size); + sumUsedSize += suballoc.size; + } + else + { + ++nullItem2ndCount; + } + + offset = suballoc.offset + suballoc.size + VMA_DEBUG_MARGIN; + } + + VMA_VALIDATE(nullItem2ndCount == m_2ndNullItemsCount); + } + + for(size_t i = 0; i < m_1stNullItemsBeginCount; ++i) + { + const VmaSuballocation& suballoc = suballocations1st[i]; + VMA_VALIDATE(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE && + suballoc.hAllocation == VK_NULL_HANDLE); + } + + size_t nullItem1stCount = m_1stNullItemsBeginCount; + + for(size_t i = m_1stNullItemsBeginCount; i < suballoc1stCount; ++i) + { + const VmaSuballocation& suballoc = suballocations1st[i]; + const bool currFree = (suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); + + VMA_VALIDATE(currFree == (suballoc.hAllocation == VK_NULL_HANDLE)); + VMA_VALIDATE(suballoc.offset >= offset); + VMA_VALIDATE(i >= m_1stNullItemsBeginCount || currFree); + + if(!currFree) + { + VMA_VALIDATE(suballoc.hAllocation->GetOffset() == suballoc.offset); + VMA_VALIDATE(suballoc.hAllocation->GetSize() == suballoc.size); + sumUsedSize += suballoc.size; + } + else + { + ++nullItem1stCount; + } + + offset = suballoc.offset + suballoc.size + VMA_DEBUG_MARGIN; + } + VMA_VALIDATE(nullItem1stCount == m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount); + + if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + { + const size_t suballoc2ndCount = suballocations2nd.size(); + size_t nullItem2ndCount = 0; + for(size_t i = suballoc2ndCount; i--; ) + { + const VmaSuballocation& suballoc = suballocations2nd[i]; + const bool currFree = (suballoc.type == VMA_SUBALLOCATION_TYPE_FREE); + + VMA_VALIDATE(currFree == (suballoc.hAllocation == VK_NULL_HANDLE)); + VMA_VALIDATE(suballoc.offset >= offset); + + if(!currFree) + { + VMA_VALIDATE(suballoc.hAllocation->GetOffset() == suballoc.offset); + VMA_VALIDATE(suballoc.hAllocation->GetSize() == suballoc.size); + sumUsedSize += suballoc.size; + } + else + { + ++nullItem2ndCount; + } + + offset = suballoc.offset + suballoc.size + VMA_DEBUG_MARGIN; + } + + VMA_VALIDATE(nullItem2ndCount == m_2ndNullItemsCount); + } + + VMA_VALIDATE(offset <= GetSize()); + VMA_VALIDATE(m_SumFreeSize == GetSize() - sumUsedSize); + + return true; +} + +size_t VmaBlockMetadata_Linear::GetAllocationCount() const +{ + return AccessSuballocations1st().size() - (m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount) + + AccessSuballocations2nd().size() - m_2ndNullItemsCount; +} + +VkDeviceSize VmaBlockMetadata_Linear::GetUnusedRangeSizeMax() const +{ + const VkDeviceSize size = GetSize(); + + /* + We don't consider gaps inside allocation vectors with freed allocations because + they are not suitable for reuse in linear allocator. We consider only space that + is available for new allocations. + */ + if(IsEmpty()) + { + return size; + } + + const SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + + switch(m_2ndVectorMode) + { + case SECOND_VECTOR_EMPTY: + /* + Available space is after end of 1st, as well as before beginning of 1st (which + whould make it a ring buffer). + */ + { + const size_t suballocations1stCount = suballocations1st.size(); + VMA_ASSERT(suballocations1stCount > m_1stNullItemsBeginCount); + const VmaSuballocation& firstSuballoc = suballocations1st[m_1stNullItemsBeginCount]; + const VmaSuballocation& lastSuballoc = suballocations1st[suballocations1stCount - 1]; + return VMA_MAX( + firstSuballoc.offset, + size - (lastSuballoc.offset + lastSuballoc.size)); + } + break; + + case SECOND_VECTOR_RING_BUFFER: + /* + Available space is only between end of 2nd and beginning of 1st. + */ + { + const SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + const VmaSuballocation& lastSuballoc2nd = suballocations2nd.back(); + const VmaSuballocation& firstSuballoc1st = suballocations1st[m_1stNullItemsBeginCount]; + return firstSuballoc1st.offset - (lastSuballoc2nd.offset + lastSuballoc2nd.size); + } + break; + + case SECOND_VECTOR_DOUBLE_STACK: + /* + Available space is only between end of 1st and top of 2nd. + */ + { + const SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + const VmaSuballocation& topSuballoc2nd = suballocations2nd.back(); + const VmaSuballocation& lastSuballoc1st = suballocations1st.back(); + return topSuballoc2nd.offset - (lastSuballoc1st.offset + lastSuballoc1st.size); + } + break; + + default: + VMA_ASSERT(0); + return 0; + } +} + +void VmaBlockMetadata_Linear::CalcAllocationStatInfo(VmaStatInfo& outInfo) const +{ + const VkDeviceSize size = GetSize(); + const SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + const SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + const size_t suballoc1stCount = suballocations1st.size(); + const size_t suballoc2ndCount = suballocations2nd.size(); + + outInfo.blockCount = 1; + outInfo.allocationCount = (uint32_t)GetAllocationCount(); + outInfo.unusedRangeCount = 0; + outInfo.usedBytes = 0; + outInfo.allocationSizeMin = UINT64_MAX; + outInfo.allocationSizeMax = 0; + outInfo.unusedRangeSizeMin = UINT64_MAX; + outInfo.unusedRangeSizeMax = 0; + + VkDeviceSize lastOffset = 0; + + if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + const VkDeviceSize freeSpace2ndTo1stEnd = suballocations1st[m_1stNullItemsBeginCount].offset; + size_t nextAlloc2ndIndex = 0; + while(lastOffset < freeSpace2ndTo1stEnd) + { + // Find next non-null allocation or move nextAllocIndex to the end. + while(nextAlloc2ndIndex < suballoc2ndCount && + suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + { + ++nextAlloc2ndIndex; + } + + // Found non-null allocation. + if(nextAlloc2ndIndex < suballoc2ndCount) + { + const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; + ++outInfo.unusedRangeCount; + outInfo.unusedBytes += unusedRangeSize; + outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); + outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + outInfo.usedBytes += suballoc.size; + outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); + outInfo.allocationSizeMax = VMA_MIN(outInfo.allocationSizeMax, suballoc.size); + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + ++nextAlloc2ndIndex; + } + // We are at the end. + else + { + // There is free space from lastOffset to freeSpace2ndTo1stEnd. + if(lastOffset < freeSpace2ndTo1stEnd) + { + const VkDeviceSize unusedRangeSize = freeSpace2ndTo1stEnd - lastOffset; + ++outInfo.unusedRangeCount; + outInfo.unusedBytes += unusedRangeSize; + outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); + outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); + } + + // End of loop. + lastOffset = freeSpace2ndTo1stEnd; + } + } + } + + size_t nextAlloc1stIndex = m_1stNullItemsBeginCount; + const VkDeviceSize freeSpace1stTo2ndEnd = + m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK ? suballocations2nd.back().offset : size; + while(lastOffset < freeSpace1stTo2ndEnd) + { + // Find next non-null allocation or move nextAllocIndex to the end. + while(nextAlloc1stIndex < suballoc1stCount && + suballocations1st[nextAlloc1stIndex].hAllocation == VK_NULL_HANDLE) + { + ++nextAlloc1stIndex; + } + + // Found non-null allocation. + if(nextAlloc1stIndex < suballoc1stCount) + { + const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; + ++outInfo.unusedRangeCount; + outInfo.unusedBytes += unusedRangeSize; + outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); + outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + outInfo.usedBytes += suballoc.size; + outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); + outInfo.allocationSizeMax = VMA_MIN(outInfo.allocationSizeMax, suballoc.size); + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + ++nextAlloc1stIndex; + } + // We are at the end. + else + { + // There is free space from lastOffset to freeSpace1stTo2ndEnd. + if(lastOffset < freeSpace1stTo2ndEnd) + { + const VkDeviceSize unusedRangeSize = freeSpace1stTo2ndEnd - lastOffset; + ++outInfo.unusedRangeCount; + outInfo.unusedBytes += unusedRangeSize; + outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); + outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); + } + + // End of loop. + lastOffset = freeSpace1stTo2ndEnd; + } + } + + if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + { + size_t nextAlloc2ndIndex = suballocations2nd.size() - 1; + while(lastOffset < size) + { + // Find next non-null allocation or move nextAllocIndex to the end. + while(nextAlloc2ndIndex != SIZE_MAX && + suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + { + --nextAlloc2ndIndex; + } + + // Found non-null allocation. + if(nextAlloc2ndIndex != SIZE_MAX) + { + const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; + ++outInfo.unusedRangeCount; + outInfo.unusedBytes += unusedRangeSize; + outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); + outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + outInfo.usedBytes += suballoc.size; + outInfo.allocationSizeMin = VMA_MIN(outInfo.allocationSizeMin, suballoc.size); + outInfo.allocationSizeMax = VMA_MIN(outInfo.allocationSizeMax, suballoc.size); + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + --nextAlloc2ndIndex; + } + // We are at the end. + else + { + // There is free space from lastOffset to size. + if(lastOffset < size) + { + const VkDeviceSize unusedRangeSize = size - lastOffset; + ++outInfo.unusedRangeCount; + outInfo.unusedBytes += unusedRangeSize; + outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusedRangeSize); + outInfo.unusedRangeSizeMax = VMA_MIN(outInfo.unusedRangeSizeMax, unusedRangeSize); + } + + // End of loop. + lastOffset = size; + } + } + } + + outInfo.unusedBytes = size - outInfo.usedBytes; +} + +void VmaBlockMetadata_Linear::AddPoolStats(VmaPoolStats& inoutStats) const +{ + const SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + const SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + const VkDeviceSize size = GetSize(); + const size_t suballoc1stCount = suballocations1st.size(); + const size_t suballoc2ndCount = suballocations2nd.size(); + + inoutStats.size += size; + + VkDeviceSize lastOffset = 0; + + if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + const VkDeviceSize freeSpace2ndTo1stEnd = suballocations1st[m_1stNullItemsBeginCount].offset; + size_t nextAlloc2ndIndex = m_1stNullItemsBeginCount; + while(lastOffset < freeSpace2ndTo1stEnd) + { + // Find next non-null allocation or move nextAlloc2ndIndex to the end. + while(nextAlloc2ndIndex < suballoc2ndCount && + suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + { + ++nextAlloc2ndIndex; + } + + // Found non-null allocation. + if(nextAlloc2ndIndex < suballoc2ndCount) + { + const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; + inoutStats.unusedSize += unusedRangeSize; + ++inoutStats.unusedRangeCount; + inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + ++inoutStats.allocationCount; + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + ++nextAlloc2ndIndex; + } + // We are at the end. + else + { + if(lastOffset < freeSpace2ndTo1stEnd) + { + // There is free space from lastOffset to freeSpace2ndTo1stEnd. + const VkDeviceSize unusedRangeSize = freeSpace2ndTo1stEnd - lastOffset; + inoutStats.unusedSize += unusedRangeSize; + ++inoutStats.unusedRangeCount; + inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); + } + + // End of loop. + lastOffset = freeSpace2ndTo1stEnd; + } + } + } + + size_t nextAlloc1stIndex = m_1stNullItemsBeginCount; + const VkDeviceSize freeSpace1stTo2ndEnd = + m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK ? suballocations2nd.back().offset : size; + while(lastOffset < freeSpace1stTo2ndEnd) + { + // Find next non-null allocation or move nextAllocIndex to the end. + while(nextAlloc1stIndex < suballoc1stCount && + suballocations1st[nextAlloc1stIndex].hAllocation == VK_NULL_HANDLE) + { + ++nextAlloc1stIndex; + } + + // Found non-null allocation. + if(nextAlloc1stIndex < suballoc1stCount) + { + const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; + inoutStats.unusedSize += unusedRangeSize; + ++inoutStats.unusedRangeCount; + inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + ++inoutStats.allocationCount; + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + ++nextAlloc1stIndex; + } + // We are at the end. + else + { + if(lastOffset < freeSpace1stTo2ndEnd) + { + // There is free space from lastOffset to freeSpace1stTo2ndEnd. + const VkDeviceSize unusedRangeSize = freeSpace1stTo2ndEnd - lastOffset; + inoutStats.unusedSize += unusedRangeSize; + ++inoutStats.unusedRangeCount; + inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); + } + + // End of loop. + lastOffset = freeSpace1stTo2ndEnd; + } + } + + if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + { + size_t nextAlloc2ndIndex = suballocations2nd.size() - 1; + while(lastOffset < size) + { + // Find next non-null allocation or move nextAlloc2ndIndex to the end. + while(nextAlloc2ndIndex != SIZE_MAX && + suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + { + --nextAlloc2ndIndex; + } + + // Found non-null allocation. + if(nextAlloc2ndIndex != SIZE_MAX) + { + const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; + inoutStats.unusedSize += unusedRangeSize; + ++inoutStats.unusedRangeCount; + inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + ++inoutStats.allocationCount; + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + --nextAlloc2ndIndex; + } + // We are at the end. + else + { + if(lastOffset < size) + { + // There is free space from lastOffset to size. + const VkDeviceSize unusedRangeSize = size - lastOffset; + inoutStats.unusedSize += unusedRangeSize; + ++inoutStats.unusedRangeCount; + inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, unusedRangeSize); + } + + // End of loop. + lastOffset = size; + } + } + } +} + +#if VMA_STATS_STRING_ENABLED +void VmaBlockMetadata_Linear::PrintDetailedMap(class VmaJsonWriter& json) const +{ + const VkDeviceSize size = GetSize(); + const SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + const SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + const size_t suballoc1stCount = suballocations1st.size(); + const size_t suballoc2ndCount = suballocations2nd.size(); + + // FIRST PASS + + size_t unusedRangeCount = 0; + VkDeviceSize usedBytes = 0; + + VkDeviceSize lastOffset = 0; + + size_t alloc2ndCount = 0; + if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + const VkDeviceSize freeSpace2ndTo1stEnd = suballocations1st[m_1stNullItemsBeginCount].offset; + size_t nextAlloc2ndIndex = 0; + while(lastOffset < freeSpace2ndTo1stEnd) + { + // Find next non-null allocation or move nextAlloc2ndIndex to the end. + while(nextAlloc2ndIndex < suballoc2ndCount && + suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + { + ++nextAlloc2ndIndex; + } + + // Found non-null allocation. + if(nextAlloc2ndIndex < suballoc2ndCount) + { + const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + ++unusedRangeCount; + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + ++alloc2ndCount; + usedBytes += suballoc.size; + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + ++nextAlloc2ndIndex; + } + // We are at the end. + else + { + if(lastOffset < freeSpace2ndTo1stEnd) + { + // There is free space from lastOffset to freeSpace2ndTo1stEnd. + ++unusedRangeCount; + } + + // End of loop. + lastOffset = freeSpace2ndTo1stEnd; + } + } + } + + size_t nextAlloc1stIndex = m_1stNullItemsBeginCount; + size_t alloc1stCount = 0; + const VkDeviceSize freeSpace1stTo2ndEnd = + m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK ? suballocations2nd.back().offset : size; + while(lastOffset < freeSpace1stTo2ndEnd) + { + // Find next non-null allocation or move nextAllocIndex to the end. + while(nextAlloc1stIndex < suballoc1stCount && + suballocations1st[nextAlloc1stIndex].hAllocation == VK_NULL_HANDLE) + { + ++nextAlloc1stIndex; + } + + // Found non-null allocation. + if(nextAlloc1stIndex < suballoc1stCount) + { + const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + ++unusedRangeCount; + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + ++alloc1stCount; + usedBytes += suballoc.size; + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + ++nextAlloc1stIndex; + } + // We are at the end. + else + { + if(lastOffset < size) + { + // There is free space from lastOffset to freeSpace1stTo2ndEnd. + ++unusedRangeCount; + } + + // End of loop. + lastOffset = freeSpace1stTo2ndEnd; + } + } + + if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + { + size_t nextAlloc2ndIndex = suballocations2nd.size() - 1; + while(lastOffset < size) + { + // Find next non-null allocation or move nextAlloc2ndIndex to the end. + while(nextAlloc2ndIndex != SIZE_MAX && + suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + { + --nextAlloc2ndIndex; + } + + // Found non-null allocation. + if(nextAlloc2ndIndex != SIZE_MAX) + { + const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + ++unusedRangeCount; + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + ++alloc2ndCount; + usedBytes += suballoc.size; + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + --nextAlloc2ndIndex; + } + // We are at the end. + else + { + if(lastOffset < size) + { + // There is free space from lastOffset to size. + ++unusedRangeCount; + } + + // End of loop. + lastOffset = size; + } + } + } + + const VkDeviceSize unusedBytes = size - usedBytes; + PrintDetailedMap_Begin(json, unusedBytes, alloc1stCount + alloc2ndCount, unusedRangeCount); + + // SECOND PASS + lastOffset = 0; + + if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + const VkDeviceSize freeSpace2ndTo1stEnd = suballocations1st[m_1stNullItemsBeginCount].offset; + size_t nextAlloc2ndIndex = 0; + while(lastOffset < freeSpace2ndTo1stEnd) + { + // Find next non-null allocation or move nextAlloc2ndIndex to the end. + while(nextAlloc2ndIndex < suballoc2ndCount && + suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + { + ++nextAlloc2ndIndex; + } + + // Found non-null allocation. + if(nextAlloc2ndIndex < suballoc2ndCount) + { + const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; + PrintDetailedMap_UnusedRange(json, lastOffset, unusedRangeSize); + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + ++nextAlloc2ndIndex; + } + // We are at the end. + else + { + if(lastOffset < freeSpace2ndTo1stEnd) + { + // There is free space from lastOffset to freeSpace2ndTo1stEnd. + const VkDeviceSize unusedRangeSize = freeSpace2ndTo1stEnd - lastOffset; + PrintDetailedMap_UnusedRange(json, lastOffset, unusedRangeSize); + } + + // End of loop. + lastOffset = freeSpace2ndTo1stEnd; + } + } + } + + nextAlloc1stIndex = m_1stNullItemsBeginCount; + while(lastOffset < freeSpace1stTo2ndEnd) + { + // Find next non-null allocation or move nextAllocIndex to the end. + while(nextAlloc1stIndex < suballoc1stCount && + suballocations1st[nextAlloc1stIndex].hAllocation == VK_NULL_HANDLE) + { + ++nextAlloc1stIndex; + } + + // Found non-null allocation. + if(nextAlloc1stIndex < suballoc1stCount) + { + const VmaSuballocation& suballoc = suballocations1st[nextAlloc1stIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; + PrintDetailedMap_UnusedRange(json, lastOffset, unusedRangeSize); + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + ++nextAlloc1stIndex; + } + // We are at the end. + else + { + if(lastOffset < freeSpace1stTo2ndEnd) + { + // There is free space from lastOffset to freeSpace1stTo2ndEnd. + const VkDeviceSize unusedRangeSize = freeSpace1stTo2ndEnd - lastOffset; + PrintDetailedMap_UnusedRange(json, lastOffset, unusedRangeSize); + } + + // End of loop. + lastOffset = freeSpace1stTo2ndEnd; + } + } + + if(m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + { + size_t nextAlloc2ndIndex = suballocations2nd.size() - 1; + while(lastOffset < size) + { + // Find next non-null allocation or move nextAlloc2ndIndex to the end. + while(nextAlloc2ndIndex != SIZE_MAX && + suballocations2nd[nextAlloc2ndIndex].hAllocation == VK_NULL_HANDLE) + { + --nextAlloc2ndIndex; + } + + // Found non-null allocation. + if(nextAlloc2ndIndex != SIZE_MAX) + { + const VmaSuballocation& suballoc = suballocations2nd[nextAlloc2ndIndex]; + + // 1. Process free space before this allocation. + if(lastOffset < suballoc.offset) + { + // There is free space from lastOffset to suballoc.offset. + const VkDeviceSize unusedRangeSize = suballoc.offset - lastOffset; + PrintDetailedMap_UnusedRange(json, lastOffset, unusedRangeSize); + } + + // 2. Process this allocation. + // There is allocation with suballoc.offset, suballoc.size. + PrintDetailedMap_Allocation(json, suballoc.offset, suballoc.hAllocation); + + // 3. Prepare for next iteration. + lastOffset = suballoc.offset + suballoc.size; + --nextAlloc2ndIndex; + } + // We are at the end. + else + { + if(lastOffset < size) + { + // There is free space from lastOffset to size. + const VkDeviceSize unusedRangeSize = size - lastOffset; + PrintDetailedMap_UnusedRange(json, lastOffset, unusedRangeSize); + } + + // End of loop. + lastOffset = size; + } + } + } + + PrintDetailedMap_End(json); +} +#endif // #if VMA_STATS_STRING_ENABLED + +bool VmaBlockMetadata_Linear::CreateAllocationRequest( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) +{ + VMA_ASSERT(allocSize > 0); + VMA_ASSERT(allocType != VMA_SUBALLOCATION_TYPE_FREE); + VMA_ASSERT(pAllocationRequest != VMA_NULL); + VMA_HEAVY_ASSERT(Validate()); + return upperAddress ? + CreateAllocationRequest_UpperAddress( + currentFrameIndex, frameInUseCount, bufferImageGranularity, + allocSize, allocAlignment, allocType, canMakeOtherLost, strategy, pAllocationRequest) : + CreateAllocationRequest_LowerAddress( + currentFrameIndex, frameInUseCount, bufferImageGranularity, + allocSize, allocAlignment, allocType, canMakeOtherLost, strategy, pAllocationRequest); +} + +bool VmaBlockMetadata_Linear::CreateAllocationRequest_UpperAddress( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) +{ + const VkDeviceSize size = GetSize(); + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + + if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + VMA_ASSERT(0 && "Trying to use pool with linear algorithm as double stack, while it is already being used as ring buffer."); + return false; + } + + // Try to allocate before 2nd.back(), or end of block if 2nd.empty(). + if(allocSize > size) + { + return false; + } + VkDeviceSize resultBaseOffset = size - allocSize; + if(!suballocations2nd.empty()) + { + const VmaSuballocation& lastSuballoc = suballocations2nd.back(); + resultBaseOffset = lastSuballoc.offset - allocSize; + if(allocSize > lastSuballoc.offset) + { + return false; + } + } + + // Start from offset equal to end of free space. + VkDeviceSize resultOffset = resultBaseOffset; + + // Apply VMA_DEBUG_MARGIN at the end. + if(VMA_DEBUG_MARGIN > 0) + { + if(resultOffset < VMA_DEBUG_MARGIN) + { + return false; + } + resultOffset -= VMA_DEBUG_MARGIN; + } + + // Apply alignment. + resultOffset = VmaAlignDown(resultOffset, allocAlignment); + + // Check next suballocations from 2nd for BufferImageGranularity conflicts. + // Make bigger alignment if necessary. + if(bufferImageGranularity > 1 && !suballocations2nd.empty()) + { + bool bufferImageGranularityConflict = false; + for(size_t nextSuballocIndex = suballocations2nd.size(); nextSuballocIndex--; ) + { + const VmaSuballocation& nextSuballoc = suballocations2nd[nextSuballocIndex]; + if(VmaBlocksOnSamePage(resultOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(nextSuballoc.type, allocType)) + { + bufferImageGranularityConflict = true; + break; + } + } + else + // Already on previous page. + break; + } + if(bufferImageGranularityConflict) + { + resultOffset = VmaAlignDown(resultOffset, bufferImageGranularity); + } + } + + // There is enough free space. + const VkDeviceSize endOf1st = !suballocations1st.empty() ? + suballocations1st.back().offset + suballocations1st.back().size : + 0; + if(endOf1st + VMA_DEBUG_MARGIN <= resultOffset) + { + // Check previous suballocations for BufferImageGranularity conflicts. + // If conflict exists, allocation cannot be made here. + if(bufferImageGranularity > 1) + { + for(size_t prevSuballocIndex = suballocations1st.size(); prevSuballocIndex--; ) + { + const VmaSuballocation& prevSuballoc = suballocations1st[prevSuballocIndex]; + if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, resultOffset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(allocType, prevSuballoc.type)) + { + return false; + } + } + else + { + // Already on next page. + break; + } + } + } + + // All tests passed: Success. + pAllocationRequest->offset = resultOffset; + pAllocationRequest->sumFreeSize = resultBaseOffset + allocSize - endOf1st; + pAllocationRequest->sumItemSize = 0; + // pAllocationRequest->item unused. + pAllocationRequest->itemsToMakeLostCount = 0; + pAllocationRequest->type = VmaAllocationRequestType::UpperAddress; + return true; + } + + return false; +} + +bool VmaBlockMetadata_Linear::CreateAllocationRequest_LowerAddress( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) +{ + const VkDeviceSize size = GetSize(); + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + + if(m_2ndVectorMode == SECOND_VECTOR_EMPTY || m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + { + // Try to allocate at the end of 1st vector. + + VkDeviceSize resultBaseOffset = 0; + if(!suballocations1st.empty()) + { + const VmaSuballocation& lastSuballoc = suballocations1st.back(); + resultBaseOffset = lastSuballoc.offset + lastSuballoc.size; + } + + // Start from offset equal to beginning of free space. + VkDeviceSize resultOffset = resultBaseOffset; + + // Apply VMA_DEBUG_MARGIN at the beginning. + if(VMA_DEBUG_MARGIN > 0) + { + resultOffset += VMA_DEBUG_MARGIN; + } + + // Apply alignment. + resultOffset = VmaAlignUp(resultOffset, allocAlignment); + + // Check previous suballocations for BufferImageGranularity conflicts. + // Make bigger alignment if necessary. + if(bufferImageGranularity > 1 && !suballocations1st.empty()) + { + bool bufferImageGranularityConflict = false; + for(size_t prevSuballocIndex = suballocations1st.size(); prevSuballocIndex--; ) + { + const VmaSuballocation& prevSuballoc = suballocations1st[prevSuballocIndex]; + if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, resultOffset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) + { + bufferImageGranularityConflict = true; + break; + } + } + else + // Already on previous page. + break; + } + if(bufferImageGranularityConflict) + { + resultOffset = VmaAlignUp(resultOffset, bufferImageGranularity); + } + } + + const VkDeviceSize freeSpaceEnd = m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK ? + suballocations2nd.back().offset : size; + + // There is enough free space at the end after alignment. + if(resultOffset + allocSize + VMA_DEBUG_MARGIN <= freeSpaceEnd) + { + // Check next suballocations for BufferImageGranularity conflicts. + // If conflict exists, allocation cannot be made here. + if(bufferImageGranularity > 1 && m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + { + for(size_t nextSuballocIndex = suballocations2nd.size(); nextSuballocIndex--; ) + { + const VmaSuballocation& nextSuballoc = suballocations2nd[nextSuballocIndex]; + if(VmaBlocksOnSamePage(resultOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) + { + return false; + } + } + else + { + // Already on previous page. + break; + } + } + } + + // All tests passed: Success. + pAllocationRequest->offset = resultOffset; + pAllocationRequest->sumFreeSize = freeSpaceEnd - resultBaseOffset; + pAllocationRequest->sumItemSize = 0; + // pAllocationRequest->item, customData unused. + pAllocationRequest->type = VmaAllocationRequestType::EndOf1st; + pAllocationRequest->itemsToMakeLostCount = 0; + return true; + } + } + + // Wrap-around to end of 2nd vector. Try to allocate there, watching for the + // beginning of 1st vector as the end of free space. + if(m_2ndVectorMode == SECOND_VECTOR_EMPTY || m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + VMA_ASSERT(!suballocations1st.empty()); + + VkDeviceSize resultBaseOffset = 0; + if(!suballocations2nd.empty()) + { + const VmaSuballocation& lastSuballoc = suballocations2nd.back(); + resultBaseOffset = lastSuballoc.offset + lastSuballoc.size; + } + + // Start from offset equal to beginning of free space. + VkDeviceSize resultOffset = resultBaseOffset; + + // Apply VMA_DEBUG_MARGIN at the beginning. + if(VMA_DEBUG_MARGIN > 0) + { + resultOffset += VMA_DEBUG_MARGIN; + } + + // Apply alignment. + resultOffset = VmaAlignUp(resultOffset, allocAlignment); + + // Check previous suballocations for BufferImageGranularity conflicts. + // Make bigger alignment if necessary. + if(bufferImageGranularity > 1 && !suballocations2nd.empty()) + { + bool bufferImageGranularityConflict = false; + for(size_t prevSuballocIndex = suballocations2nd.size(); prevSuballocIndex--; ) + { + const VmaSuballocation& prevSuballoc = suballocations2nd[prevSuballocIndex]; + if(VmaBlocksOnSamePage(prevSuballoc.offset, prevSuballoc.size, resultOffset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(prevSuballoc.type, allocType)) + { + bufferImageGranularityConflict = true; + break; + } + } + else + // Already on previous page. + break; + } + if(bufferImageGranularityConflict) + { + resultOffset = VmaAlignUp(resultOffset, bufferImageGranularity); + } + } + + pAllocationRequest->itemsToMakeLostCount = 0; + pAllocationRequest->sumItemSize = 0; + size_t index1st = m_1stNullItemsBeginCount; + + if(canMakeOtherLost) + { + while(index1st < suballocations1st.size() && + resultOffset + allocSize + VMA_DEBUG_MARGIN > suballocations1st[index1st].offset) + { + // Next colliding allocation at the beginning of 1st vector found. Try to make it lost. + const VmaSuballocation& suballoc = suballocations1st[index1st]; + if(suballoc.type == VMA_SUBALLOCATION_TYPE_FREE) + { + // No problem. + } + else + { + VMA_ASSERT(suballoc.hAllocation != VK_NULL_HANDLE); + if(suballoc.hAllocation->CanBecomeLost() && + suballoc.hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) + { + ++pAllocationRequest->itemsToMakeLostCount; + pAllocationRequest->sumItemSize += suballoc.size; + } + else + { + return false; + } + } + ++index1st; + } + + // Check next suballocations for BufferImageGranularity conflicts. + // If conflict exists, we must mark more allocations lost or fail. + if(bufferImageGranularity > 1) + { + while(index1st < suballocations1st.size()) + { + const VmaSuballocation& suballoc = suballocations1st[index1st]; + if(VmaBlocksOnSamePage(resultOffset, allocSize, suballoc.offset, bufferImageGranularity)) + { + if(suballoc.hAllocation != VK_NULL_HANDLE) + { + // Not checking actual VmaIsBufferImageGranularityConflict(allocType, suballoc.type). + if(suballoc.hAllocation->CanBecomeLost() && + suballoc.hAllocation->GetLastUseFrameIndex() + frameInUseCount < currentFrameIndex) + { + ++pAllocationRequest->itemsToMakeLostCount; + pAllocationRequest->sumItemSize += suballoc.size; + } + else + { + return false; + } + } + } + else + { + // Already on next page. + break; + } + ++index1st; + } + } + + // Special case: There is not enough room at the end for this allocation, even after making all from the 1st lost. + if(index1st == suballocations1st.size() && + resultOffset + allocSize + VMA_DEBUG_MARGIN > size) + { + // TODO: This is a known bug that it's not yet implemented and the allocation is failing. + VMA_DEBUG_LOG("Unsupported special case in custom pool with linear allocation algorithm used as ring buffer with allocations that can be lost."); + } + } + + // There is enough free space at the end after alignment. + if((index1st == suballocations1st.size() && resultOffset + allocSize + VMA_DEBUG_MARGIN <= size) || + (index1st < suballocations1st.size() && resultOffset + allocSize + VMA_DEBUG_MARGIN <= suballocations1st[index1st].offset)) + { + // Check next suballocations for BufferImageGranularity conflicts. + // If conflict exists, allocation cannot be made here. + if(bufferImageGranularity > 1) + { + for(size_t nextSuballocIndex = index1st; + nextSuballocIndex < suballocations1st.size(); + nextSuballocIndex++) + { + const VmaSuballocation& nextSuballoc = suballocations1st[nextSuballocIndex]; + if(VmaBlocksOnSamePage(resultOffset, allocSize, nextSuballoc.offset, bufferImageGranularity)) + { + if(VmaIsBufferImageGranularityConflict(allocType, nextSuballoc.type)) + { + return false; + } + } + else + { + // Already on next page. + break; + } + } + } + + // All tests passed: Success. + pAllocationRequest->offset = resultOffset; + pAllocationRequest->sumFreeSize = + (index1st < suballocations1st.size() ? suballocations1st[index1st].offset : size) + - resultBaseOffset + - pAllocationRequest->sumItemSize; + pAllocationRequest->type = VmaAllocationRequestType::EndOf2nd; + // pAllocationRequest->item, customData unused. + return true; + } + } + + return false; +} + +bool VmaBlockMetadata_Linear::MakeRequestedAllocationsLost( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VmaAllocationRequest* pAllocationRequest) +{ + if(pAllocationRequest->itemsToMakeLostCount == 0) + { + return true; + } + + VMA_ASSERT(m_2ndVectorMode == SECOND_VECTOR_EMPTY || m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER); + + // We always start from 1st. + SuballocationVectorType* suballocations = &AccessSuballocations1st(); + size_t index = m_1stNullItemsBeginCount; + size_t madeLostCount = 0; + while(madeLostCount < pAllocationRequest->itemsToMakeLostCount) + { + if(index == suballocations->size()) + { + index = 0; + // If we get to the end of 1st, we wrap around to beginning of 2nd of 1st. + if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + suballocations = &AccessSuballocations2nd(); + } + // else: m_2ndVectorMode == SECOND_VECTOR_EMPTY: + // suballocations continues pointing at AccessSuballocations1st(). + VMA_ASSERT(!suballocations->empty()); + } + VmaSuballocation& suballoc = (*suballocations)[index]; + if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + { + VMA_ASSERT(suballoc.hAllocation != VK_NULL_HANDLE); + VMA_ASSERT(suballoc.hAllocation->CanBecomeLost()); + if(suballoc.hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) + { + suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + suballoc.hAllocation = VK_NULL_HANDLE; + m_SumFreeSize += suballoc.size; + if(suballocations == &AccessSuballocations1st()) + { + ++m_1stNullItemsMiddleCount; + } + else + { + ++m_2ndNullItemsCount; + } + ++madeLostCount; + } + else + { + return false; + } + } + ++index; + } + + CleanupAfterFree(); + //VMA_HEAVY_ASSERT(Validate()); // Already called by ClanupAfterFree(). + + return true; +} + +uint32_t VmaBlockMetadata_Linear::MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) +{ + uint32_t lostAllocationCount = 0; + + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + for(size_t i = m_1stNullItemsBeginCount, count = suballocations1st.size(); i < count; ++i) + { + VmaSuballocation& suballoc = suballocations1st[i]; + if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE && + suballoc.hAllocation->CanBecomeLost() && + suballoc.hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) + { + suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + suballoc.hAllocation = VK_NULL_HANDLE; + ++m_1stNullItemsMiddleCount; + m_SumFreeSize += suballoc.size; + ++lostAllocationCount; + } + } + + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + for(size_t i = 0, count = suballocations2nd.size(); i < count; ++i) + { + VmaSuballocation& suballoc = suballocations2nd[i]; + if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE && + suballoc.hAllocation->CanBecomeLost() && + suballoc.hAllocation->MakeLost(currentFrameIndex, frameInUseCount)) + { + suballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + suballoc.hAllocation = VK_NULL_HANDLE; + ++m_2ndNullItemsCount; + m_SumFreeSize += suballoc.size; + ++lostAllocationCount; + } + } + + if(lostAllocationCount) + { + CleanupAfterFree(); + } + + return lostAllocationCount; +} + +VkResult VmaBlockMetadata_Linear::CheckCorruption(const void* pBlockData) +{ + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + for(size_t i = m_1stNullItemsBeginCount, count = suballocations1st.size(); i < count; ++i) + { + const VmaSuballocation& suballoc = suballocations1st[i]; + if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + { + if(!VmaValidateMagicValue(pBlockData, suballoc.offset - VMA_DEBUG_MARGIN)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED BEFORE VALIDATED ALLOCATION!"); + return VK_ERROR_VALIDATION_FAILED_EXT; + } + if(!VmaValidateMagicValue(pBlockData, suballoc.offset + suballoc.size)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER VALIDATED ALLOCATION!"); + return VK_ERROR_VALIDATION_FAILED_EXT; + } + } + } + + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + for(size_t i = 0, count = suballocations2nd.size(); i < count; ++i) + { + const VmaSuballocation& suballoc = suballocations2nd[i]; + if(suballoc.type != VMA_SUBALLOCATION_TYPE_FREE) + { + if(!VmaValidateMagicValue(pBlockData, suballoc.offset - VMA_DEBUG_MARGIN)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED BEFORE VALIDATED ALLOCATION!"); + return VK_ERROR_VALIDATION_FAILED_EXT; + } + if(!VmaValidateMagicValue(pBlockData, suballoc.offset + suballoc.size)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER VALIDATED ALLOCATION!"); + return VK_ERROR_VALIDATION_FAILED_EXT; + } + } + } + + return VK_SUCCESS; +} + +void VmaBlockMetadata_Linear::Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + VkDeviceSize allocSize, + VmaAllocation hAllocation) +{ + const VmaSuballocation newSuballoc = { request.offset, allocSize, hAllocation, type }; + + switch(request.type) + { + case VmaAllocationRequestType::UpperAddress: + { + VMA_ASSERT(m_2ndVectorMode != SECOND_VECTOR_RING_BUFFER && + "CRITICAL ERROR: Trying to use linear allocator as double stack while it was already used as ring buffer."); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + suballocations2nd.push_back(newSuballoc); + m_2ndVectorMode = SECOND_VECTOR_DOUBLE_STACK; + } + break; + case VmaAllocationRequestType::EndOf1st: + { + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + + VMA_ASSERT(suballocations1st.empty() || + request.offset >= suballocations1st.back().offset + suballocations1st.back().size); + // Check if it fits before the end of the block. + VMA_ASSERT(request.offset + allocSize <= GetSize()); + + suballocations1st.push_back(newSuballoc); + } + break; + case VmaAllocationRequestType::EndOf2nd: + { + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + // New allocation at the end of 2-part ring buffer, so before first allocation from 1st vector. + VMA_ASSERT(!suballocations1st.empty() && + request.offset + allocSize <= suballocations1st[m_1stNullItemsBeginCount].offset); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + + switch(m_2ndVectorMode) + { + case SECOND_VECTOR_EMPTY: + // First allocation from second part ring buffer. + VMA_ASSERT(suballocations2nd.empty()); + m_2ndVectorMode = SECOND_VECTOR_RING_BUFFER; + break; + case SECOND_VECTOR_RING_BUFFER: + // 2-part ring buffer is already started. + VMA_ASSERT(!suballocations2nd.empty()); + break; + case SECOND_VECTOR_DOUBLE_STACK: + VMA_ASSERT(0 && "CRITICAL ERROR: Trying to use linear allocator as ring buffer while it was already used as double stack."); + break; + default: + VMA_ASSERT(0); + } + + suballocations2nd.push_back(newSuballoc); + } + break; + default: + VMA_ASSERT(0 && "CRITICAL INTERNAL ERROR."); + } + + m_SumFreeSize -= newSuballoc.size; +} + +void VmaBlockMetadata_Linear::Free(const VmaAllocation allocation) +{ + FreeAtOffset(allocation->GetOffset()); +} + +void VmaBlockMetadata_Linear::FreeAtOffset(VkDeviceSize offset) +{ + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + + if(!suballocations1st.empty()) + { + // First allocation: Mark it as next empty at the beginning. + VmaSuballocation& firstSuballoc = suballocations1st[m_1stNullItemsBeginCount]; + if(firstSuballoc.offset == offset) + { + firstSuballoc.type = VMA_SUBALLOCATION_TYPE_FREE; + firstSuballoc.hAllocation = VK_NULL_HANDLE; + m_SumFreeSize += firstSuballoc.size; + ++m_1stNullItemsBeginCount; + CleanupAfterFree(); + return; + } + } + + // Last allocation in 2-part ring buffer or top of upper stack (same logic). + if(m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER || + m_2ndVectorMode == SECOND_VECTOR_DOUBLE_STACK) + { + VmaSuballocation& lastSuballoc = suballocations2nd.back(); + if(lastSuballoc.offset == offset) + { + m_SumFreeSize += lastSuballoc.size; + suballocations2nd.pop_back(); + CleanupAfterFree(); + return; + } + } + // Last allocation in 1st vector. + else if(m_2ndVectorMode == SECOND_VECTOR_EMPTY) + { + VmaSuballocation& lastSuballoc = suballocations1st.back(); + if(lastSuballoc.offset == offset) + { + m_SumFreeSize += lastSuballoc.size; + suballocations1st.pop_back(); + CleanupAfterFree(); + return; + } + } + + // Item from the middle of 1st vector. + { + VmaSuballocation refSuballoc; + refSuballoc.offset = offset; + // Rest of members stays uninitialized intentionally for better performance. + SuballocationVectorType::iterator it = VmaBinaryFindSorted( + suballocations1st.begin() + m_1stNullItemsBeginCount, + suballocations1st.end(), + refSuballoc, + VmaSuballocationOffsetLess()); + if(it != suballocations1st.end()) + { + it->type = VMA_SUBALLOCATION_TYPE_FREE; + it->hAllocation = VK_NULL_HANDLE; + ++m_1stNullItemsMiddleCount; + m_SumFreeSize += it->size; + CleanupAfterFree(); + return; + } + } + + if(m_2ndVectorMode != SECOND_VECTOR_EMPTY) + { + // Item from the middle of 2nd vector. + VmaSuballocation refSuballoc; + refSuballoc.offset = offset; + // Rest of members stays uninitialized intentionally for better performance. + SuballocationVectorType::iterator it = m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER ? + VmaBinaryFindSorted(suballocations2nd.begin(), suballocations2nd.end(), refSuballoc, VmaSuballocationOffsetLess()) : + VmaBinaryFindSorted(suballocations2nd.begin(), suballocations2nd.end(), refSuballoc, VmaSuballocationOffsetGreater()); + if(it != suballocations2nd.end()) + { + it->type = VMA_SUBALLOCATION_TYPE_FREE; + it->hAllocation = VK_NULL_HANDLE; + ++m_2ndNullItemsCount; + m_SumFreeSize += it->size; + CleanupAfterFree(); + return; + } + } + + VMA_ASSERT(0 && "Allocation to free not found in linear allocator!"); +} + +bool VmaBlockMetadata_Linear::ShouldCompact1st() const +{ + const size_t nullItemCount = m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount; + const size_t suballocCount = AccessSuballocations1st().size(); + return suballocCount > 32 && nullItemCount * 2 >= (suballocCount - nullItemCount) * 3; +} + +void VmaBlockMetadata_Linear::CleanupAfterFree() +{ + SuballocationVectorType& suballocations1st = AccessSuballocations1st(); + SuballocationVectorType& suballocations2nd = AccessSuballocations2nd(); + + if(IsEmpty()) + { + suballocations1st.clear(); + suballocations2nd.clear(); + m_1stNullItemsBeginCount = 0; + m_1stNullItemsMiddleCount = 0; + m_2ndNullItemsCount = 0; + m_2ndVectorMode = SECOND_VECTOR_EMPTY; + } + else + { + const size_t suballoc1stCount = suballocations1st.size(); + const size_t nullItem1stCount = m_1stNullItemsBeginCount + m_1stNullItemsMiddleCount; + VMA_ASSERT(nullItem1stCount <= suballoc1stCount); + + // Find more null items at the beginning of 1st vector. + while(m_1stNullItemsBeginCount < suballoc1stCount && + suballocations1st[m_1stNullItemsBeginCount].hAllocation == VK_NULL_HANDLE) + { + ++m_1stNullItemsBeginCount; + --m_1stNullItemsMiddleCount; + } + + // Find more null items at the end of 1st vector. + while(m_1stNullItemsMiddleCount > 0 && + suballocations1st.back().hAllocation == VK_NULL_HANDLE) + { + --m_1stNullItemsMiddleCount; + suballocations1st.pop_back(); + } + + // Find more null items at the end of 2nd vector. + while(m_2ndNullItemsCount > 0 && + suballocations2nd.back().hAllocation == VK_NULL_HANDLE) + { + --m_2ndNullItemsCount; + suballocations2nd.pop_back(); + } + + // Find more null items at the beginning of 2nd vector. + while(m_2ndNullItemsCount > 0 && + suballocations2nd[0].hAllocation == VK_NULL_HANDLE) + { + --m_2ndNullItemsCount; + VmaVectorRemove(suballocations2nd, 0); + } + + if(ShouldCompact1st()) + { + const size_t nonNullItemCount = suballoc1stCount - nullItem1stCount; + size_t srcIndex = m_1stNullItemsBeginCount; + for(size_t dstIndex = 0; dstIndex < nonNullItemCount; ++dstIndex) + { + while(suballocations1st[srcIndex].hAllocation == VK_NULL_HANDLE) + { + ++srcIndex; + } + if(dstIndex != srcIndex) + { + suballocations1st[dstIndex] = suballocations1st[srcIndex]; + } + ++srcIndex; + } + suballocations1st.resize(nonNullItemCount); + m_1stNullItemsBeginCount = 0; + m_1stNullItemsMiddleCount = 0; + } + + // 2nd vector became empty. + if(suballocations2nd.empty()) + { + m_2ndVectorMode = SECOND_VECTOR_EMPTY; + } + + // 1st vector became empty. + if(suballocations1st.size() - m_1stNullItemsBeginCount == 0) + { + suballocations1st.clear(); + m_1stNullItemsBeginCount = 0; + + if(!suballocations2nd.empty() && m_2ndVectorMode == SECOND_VECTOR_RING_BUFFER) + { + // Swap 1st with 2nd. Now 2nd is empty. + m_2ndVectorMode = SECOND_VECTOR_EMPTY; + m_1stNullItemsMiddleCount = m_2ndNullItemsCount; + while(m_1stNullItemsBeginCount < suballocations2nd.size() && + suballocations2nd[m_1stNullItemsBeginCount].hAllocation == VK_NULL_HANDLE) + { + ++m_1stNullItemsBeginCount; + --m_1stNullItemsMiddleCount; + } + m_2ndNullItemsCount = 0; + m_1stVectorIndex ^= 1; + } + } + } + + VMA_HEAVY_ASSERT(Validate()); +} + + +//////////////////////////////////////////////////////////////////////////////// +// class VmaBlockMetadata_Buddy + +VmaBlockMetadata_Buddy::VmaBlockMetadata_Buddy(VmaAllocator hAllocator) : + VmaBlockMetadata(hAllocator), + m_Root(VMA_NULL), + m_AllocationCount(0), + m_FreeCount(1), + m_SumFreeSize(0) +{ + memset(m_FreeList, 0, sizeof(m_FreeList)); +} + +VmaBlockMetadata_Buddy::~VmaBlockMetadata_Buddy() +{ + DeleteNode(m_Root); +} + +void VmaBlockMetadata_Buddy::Init(VkDeviceSize size) +{ + VmaBlockMetadata::Init(size); + + m_UsableSize = VmaPrevPow2(size); + m_SumFreeSize = m_UsableSize; + + // Calculate m_LevelCount. + m_LevelCount = 1; + while(m_LevelCount < MAX_LEVELS && + LevelToNodeSize(m_LevelCount) >= MIN_NODE_SIZE) + { + ++m_LevelCount; + } + + Node* rootNode = vma_new(GetAllocationCallbacks(), Node)(); + rootNode->offset = 0; + rootNode->type = Node::TYPE_FREE; + rootNode->parent = VMA_NULL; + rootNode->buddy = VMA_NULL; + + m_Root = rootNode; + AddToFreeListFront(0, rootNode); +} + +bool VmaBlockMetadata_Buddy::Validate() const +{ + // Validate tree. + ValidationContext ctx; + if(!ValidateNode(ctx, VMA_NULL, m_Root, 0, LevelToNodeSize(0))) + { + VMA_VALIDATE(false && "ValidateNode failed."); + } + VMA_VALIDATE(m_AllocationCount == ctx.calculatedAllocationCount); + VMA_VALIDATE(m_SumFreeSize == ctx.calculatedSumFreeSize); + + // Validate free node lists. + for(uint32_t level = 0; level < m_LevelCount; ++level) + { + VMA_VALIDATE(m_FreeList[level].front == VMA_NULL || + m_FreeList[level].front->free.prev == VMA_NULL); + + for(Node* node = m_FreeList[level].front; + node != VMA_NULL; + node = node->free.next) + { + VMA_VALIDATE(node->type == Node::TYPE_FREE); + + if(node->free.next == VMA_NULL) + { + VMA_VALIDATE(m_FreeList[level].back == node); + } + else + { + VMA_VALIDATE(node->free.next->free.prev == node); + } + } + } + + // Validate that free lists ar higher levels are empty. + for(uint32_t level = m_LevelCount; level < MAX_LEVELS; ++level) + { + VMA_VALIDATE(m_FreeList[level].front == VMA_NULL && m_FreeList[level].back == VMA_NULL); + } + + return true; +} + +VkDeviceSize VmaBlockMetadata_Buddy::GetUnusedRangeSizeMax() const +{ + for(uint32_t level = 0; level < m_LevelCount; ++level) + { + if(m_FreeList[level].front != VMA_NULL) + { + return LevelToNodeSize(level); + } + } + return 0; +} + +void VmaBlockMetadata_Buddy::CalcAllocationStatInfo(VmaStatInfo& outInfo) const +{ + const VkDeviceSize unusableSize = GetUnusableSize(); + + outInfo.blockCount = 1; + + outInfo.allocationCount = outInfo.unusedRangeCount = 0; + outInfo.usedBytes = outInfo.unusedBytes = 0; + + outInfo.allocationSizeMax = outInfo.unusedRangeSizeMax = 0; + outInfo.allocationSizeMin = outInfo.unusedRangeSizeMin = UINT64_MAX; + outInfo.allocationSizeAvg = outInfo.unusedRangeSizeAvg = 0; // Unused. + + CalcAllocationStatInfoNode(outInfo, m_Root, LevelToNodeSize(0)); + + if(unusableSize > 0) + { + ++outInfo.unusedRangeCount; + outInfo.unusedBytes += unusableSize; + outInfo.unusedRangeSizeMax = VMA_MAX(outInfo.unusedRangeSizeMax, unusableSize); + outInfo.unusedRangeSizeMin = VMA_MIN(outInfo.unusedRangeSizeMin, unusableSize); + } +} + +void VmaBlockMetadata_Buddy::AddPoolStats(VmaPoolStats& inoutStats) const +{ + const VkDeviceSize unusableSize = GetUnusableSize(); + + inoutStats.size += GetSize(); + inoutStats.unusedSize += m_SumFreeSize + unusableSize; + inoutStats.allocationCount += m_AllocationCount; + inoutStats.unusedRangeCount += m_FreeCount; + inoutStats.unusedRangeSizeMax = VMA_MAX(inoutStats.unusedRangeSizeMax, GetUnusedRangeSizeMax()); + + if(unusableSize > 0) + { + ++inoutStats.unusedRangeCount; + // Not updating inoutStats.unusedRangeSizeMax with unusableSize because this space is not available for allocations. + } +} + +#if VMA_STATS_STRING_ENABLED + +void VmaBlockMetadata_Buddy::PrintDetailedMap(class VmaJsonWriter& json) const +{ + // TODO optimize + VmaStatInfo stat; + CalcAllocationStatInfo(stat); + + PrintDetailedMap_Begin( + json, + stat.unusedBytes, + stat.allocationCount, + stat.unusedRangeCount); + + PrintDetailedMapNode(json, m_Root, LevelToNodeSize(0)); + + const VkDeviceSize unusableSize = GetUnusableSize(); + if(unusableSize > 0) + { + PrintDetailedMap_UnusedRange(json, + m_UsableSize, // offset + unusableSize); // size + } + + PrintDetailedMap_End(json); +} + +#endif // #if VMA_STATS_STRING_ENABLED + +bool VmaBlockMetadata_Buddy::CreateAllocationRequest( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VkDeviceSize bufferImageGranularity, + VkDeviceSize allocSize, + VkDeviceSize allocAlignment, + bool upperAddress, + VmaSuballocationType allocType, + bool canMakeOtherLost, + uint32_t strategy, + VmaAllocationRequest* pAllocationRequest) +{ + VMA_ASSERT(!upperAddress && "VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT can be used only with linear algorithm."); + + // Simple way to respect bufferImageGranularity. May be optimized some day. + // Whenever it might be an OPTIMAL image... + if(allocType == VMA_SUBALLOCATION_TYPE_UNKNOWN || + allocType == VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN || + allocType == VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL) + { + allocAlignment = VMA_MAX(allocAlignment, bufferImageGranularity); + allocSize = VMA_MAX(allocSize, bufferImageGranularity); + } + + if(allocSize > m_UsableSize) + { + return false; + } + + const uint32_t targetLevel = AllocSizeToLevel(allocSize); + for(uint32_t level = targetLevel + 1; level--; ) + { + for(Node* freeNode = m_FreeList[level].front; + freeNode != VMA_NULL; + freeNode = freeNode->free.next) + { + if(freeNode->offset % allocAlignment == 0) + { + pAllocationRequest->type = VmaAllocationRequestType::Normal; + pAllocationRequest->offset = freeNode->offset; + pAllocationRequest->sumFreeSize = LevelToNodeSize(level); + pAllocationRequest->sumItemSize = 0; + pAllocationRequest->itemsToMakeLostCount = 0; + pAllocationRequest->customData = (void*)(uintptr_t)level; + return true; + } + } + } + + return false; +} + +bool VmaBlockMetadata_Buddy::MakeRequestedAllocationsLost( + uint32_t currentFrameIndex, + uint32_t frameInUseCount, + VmaAllocationRequest* pAllocationRequest) +{ + /* + Lost allocations are not supported in buddy allocator at the moment. + Support might be added in the future. + */ + return pAllocationRequest->itemsToMakeLostCount == 0; +} + +uint32_t VmaBlockMetadata_Buddy::MakeAllocationsLost(uint32_t currentFrameIndex, uint32_t frameInUseCount) +{ + /* + Lost allocations are not supported in buddy allocator at the moment. + Support might be added in the future. + */ + return 0; +} + +void VmaBlockMetadata_Buddy::Alloc( + const VmaAllocationRequest& request, + VmaSuballocationType type, + VkDeviceSize allocSize, + VmaAllocation hAllocation) +{ + VMA_ASSERT(request.type == VmaAllocationRequestType::Normal); + + const uint32_t targetLevel = AllocSizeToLevel(allocSize); + uint32_t currLevel = (uint32_t)(uintptr_t)request.customData; + + Node* currNode = m_FreeList[currLevel].front; + VMA_ASSERT(currNode != VMA_NULL && currNode->type == Node::TYPE_FREE); + while(currNode->offset != request.offset) + { + currNode = currNode->free.next; + VMA_ASSERT(currNode != VMA_NULL && currNode->type == Node::TYPE_FREE); + } + + // Go down, splitting free nodes. + while(currLevel < targetLevel) + { + // currNode is already first free node at currLevel. + // Remove it from list of free nodes at this currLevel. + RemoveFromFreeList(currLevel, currNode); + + const uint32_t childrenLevel = currLevel + 1; + + // Create two free sub-nodes. + Node* leftChild = vma_new(GetAllocationCallbacks(), Node)(); + Node* rightChild = vma_new(GetAllocationCallbacks(), Node)(); + + leftChild->offset = currNode->offset; + leftChild->type = Node::TYPE_FREE; + leftChild->parent = currNode; + leftChild->buddy = rightChild; + + rightChild->offset = currNode->offset + LevelToNodeSize(childrenLevel); + rightChild->type = Node::TYPE_FREE; + rightChild->parent = currNode; + rightChild->buddy = leftChild; + + // Convert current currNode to split type. + currNode->type = Node::TYPE_SPLIT; + currNode->split.leftChild = leftChild; + + // Add child nodes to free list. Order is important! + AddToFreeListFront(childrenLevel, rightChild); + AddToFreeListFront(childrenLevel, leftChild); + + ++m_FreeCount; + //m_SumFreeSize -= LevelToNodeSize(currLevel) % 2; // Useful only when level node sizes can be non power of 2. + ++currLevel; + currNode = m_FreeList[currLevel].front; + + /* + We can be sure that currNode, as left child of node previously split, + also fullfills the alignment requirement. + */ + } + + // Remove from free list. + VMA_ASSERT(currLevel == targetLevel && + currNode != VMA_NULL && + currNode->type == Node::TYPE_FREE); + RemoveFromFreeList(currLevel, currNode); + + // Convert to allocation node. + currNode->type = Node::TYPE_ALLOCATION; + currNode->allocation.alloc = hAllocation; + + ++m_AllocationCount; + --m_FreeCount; + m_SumFreeSize -= allocSize; +} + +void VmaBlockMetadata_Buddy::DeleteNode(Node* node) +{ + if(node->type == Node::TYPE_SPLIT) + { + DeleteNode(node->split.leftChild->buddy); + DeleteNode(node->split.leftChild); + } + + vma_delete(GetAllocationCallbacks(), node); +} + +bool VmaBlockMetadata_Buddy::ValidateNode(ValidationContext& ctx, const Node* parent, const Node* curr, uint32_t level, VkDeviceSize levelNodeSize) const +{ + VMA_VALIDATE(level < m_LevelCount); + VMA_VALIDATE(curr->parent == parent); + VMA_VALIDATE((curr->buddy == VMA_NULL) == (parent == VMA_NULL)); + VMA_VALIDATE(curr->buddy == VMA_NULL || curr->buddy->buddy == curr); + switch(curr->type) + { + case Node::TYPE_FREE: + // curr->free.prev, next are validated separately. + ctx.calculatedSumFreeSize += levelNodeSize; + ++ctx.calculatedFreeCount; + break; + case Node::TYPE_ALLOCATION: + ++ctx.calculatedAllocationCount; + ctx.calculatedSumFreeSize += levelNodeSize - curr->allocation.alloc->GetSize(); + VMA_VALIDATE(curr->allocation.alloc != VK_NULL_HANDLE); + break; + case Node::TYPE_SPLIT: + { + const uint32_t childrenLevel = level + 1; + const VkDeviceSize childrenLevelNodeSize = levelNodeSize / 2; + const Node* const leftChild = curr->split.leftChild; + VMA_VALIDATE(leftChild != VMA_NULL); + VMA_VALIDATE(leftChild->offset == curr->offset); + if(!ValidateNode(ctx, curr, leftChild, childrenLevel, childrenLevelNodeSize)) + { + VMA_VALIDATE(false && "ValidateNode for left child failed."); + } + const Node* const rightChild = leftChild->buddy; + VMA_VALIDATE(rightChild->offset == curr->offset + childrenLevelNodeSize); + if(!ValidateNode(ctx, curr, rightChild, childrenLevel, childrenLevelNodeSize)) + { + VMA_VALIDATE(false && "ValidateNode for right child failed."); + } + } + break; + default: + return false; + } + + return true; +} + +uint32_t VmaBlockMetadata_Buddy::AllocSizeToLevel(VkDeviceSize allocSize) const +{ + // I know this could be optimized somehow e.g. by using std::log2p1 from C++20. + uint32_t level = 0; + VkDeviceSize currLevelNodeSize = m_UsableSize; + VkDeviceSize nextLevelNodeSize = currLevelNodeSize >> 1; + while(allocSize <= nextLevelNodeSize && level + 1 < m_LevelCount) + { + ++level; + currLevelNodeSize = nextLevelNodeSize; + nextLevelNodeSize = currLevelNodeSize >> 1; + } + return level; +} + +void VmaBlockMetadata_Buddy::FreeAtOffset(VmaAllocation alloc, VkDeviceSize offset) +{ + // Find node and level. + Node* node = m_Root; + VkDeviceSize nodeOffset = 0; + uint32_t level = 0; + VkDeviceSize levelNodeSize = LevelToNodeSize(0); + while(node->type == Node::TYPE_SPLIT) + { + const VkDeviceSize nextLevelSize = levelNodeSize >> 1; + if(offset < nodeOffset + nextLevelSize) + { + node = node->split.leftChild; + } + else + { + node = node->split.leftChild->buddy; + nodeOffset += nextLevelSize; + } + ++level; + levelNodeSize = nextLevelSize; + } + + VMA_ASSERT(node != VMA_NULL && node->type == Node::TYPE_ALLOCATION); + VMA_ASSERT(alloc == VK_NULL_HANDLE || node->allocation.alloc == alloc); + + ++m_FreeCount; + --m_AllocationCount; + m_SumFreeSize += alloc->GetSize(); + + node->type = Node::TYPE_FREE; + + // Join free nodes if possible. + while(level > 0 && node->buddy->type == Node::TYPE_FREE) + { + RemoveFromFreeList(level, node->buddy); + Node* const parent = node->parent; + + vma_delete(GetAllocationCallbacks(), node->buddy); + vma_delete(GetAllocationCallbacks(), node); + parent->type = Node::TYPE_FREE; + + node = parent; + --level; + //m_SumFreeSize += LevelToNodeSize(level) % 2; // Useful only when level node sizes can be non power of 2. + --m_FreeCount; + } + + AddToFreeListFront(level, node); +} + +void VmaBlockMetadata_Buddy::CalcAllocationStatInfoNode(VmaStatInfo& outInfo, const Node* node, VkDeviceSize levelNodeSize) const +{ + switch(node->type) + { + case Node::TYPE_FREE: + ++outInfo.unusedRangeCount; + outInfo.unusedBytes += levelNodeSize; + outInfo.unusedRangeSizeMax = VMA_MAX(outInfo.unusedRangeSizeMax, levelNodeSize); + outInfo.unusedRangeSizeMin = VMA_MAX(outInfo.unusedRangeSizeMin, levelNodeSize); + break; + case Node::TYPE_ALLOCATION: + { + const VkDeviceSize allocSize = node->allocation.alloc->GetSize(); + ++outInfo.allocationCount; + outInfo.usedBytes += allocSize; + outInfo.allocationSizeMax = VMA_MAX(outInfo.allocationSizeMax, allocSize); + outInfo.allocationSizeMin = VMA_MAX(outInfo.allocationSizeMin, allocSize); + + const VkDeviceSize unusedRangeSize = levelNodeSize - allocSize; + if(unusedRangeSize > 0) + { + ++outInfo.unusedRangeCount; + outInfo.unusedBytes += unusedRangeSize; + outInfo.unusedRangeSizeMax = VMA_MAX(outInfo.unusedRangeSizeMax, unusedRangeSize); + outInfo.unusedRangeSizeMin = VMA_MAX(outInfo.unusedRangeSizeMin, unusedRangeSize); + } + } + break; + case Node::TYPE_SPLIT: + { + const VkDeviceSize childrenNodeSize = levelNodeSize / 2; + const Node* const leftChild = node->split.leftChild; + CalcAllocationStatInfoNode(outInfo, leftChild, childrenNodeSize); + const Node* const rightChild = leftChild->buddy; + CalcAllocationStatInfoNode(outInfo, rightChild, childrenNodeSize); + } + break; + default: + VMA_ASSERT(0); + } +} + +void VmaBlockMetadata_Buddy::AddToFreeListFront(uint32_t level, Node* node) +{ + VMA_ASSERT(node->type == Node::TYPE_FREE); + + // List is empty. + Node* const frontNode = m_FreeList[level].front; + if(frontNode == VMA_NULL) + { + VMA_ASSERT(m_FreeList[level].back == VMA_NULL); + node->free.prev = node->free.next = VMA_NULL; + m_FreeList[level].front = m_FreeList[level].back = node; + } + else + { + VMA_ASSERT(frontNode->free.prev == VMA_NULL); + node->free.prev = VMA_NULL; + node->free.next = frontNode; + frontNode->free.prev = node; + m_FreeList[level].front = node; + } +} + +void VmaBlockMetadata_Buddy::RemoveFromFreeList(uint32_t level, Node* node) +{ + VMA_ASSERT(m_FreeList[level].front != VMA_NULL); + + // It is at the front. + if(node->free.prev == VMA_NULL) + { + VMA_ASSERT(m_FreeList[level].front == node); + m_FreeList[level].front = node->free.next; + } + else + { + Node* const prevFreeNode = node->free.prev; + VMA_ASSERT(prevFreeNode->free.next == node); + prevFreeNode->free.next = node->free.next; + } + + // It is at the back. + if(node->free.next == VMA_NULL) + { + VMA_ASSERT(m_FreeList[level].back == node); + m_FreeList[level].back = node->free.prev; + } + else + { + Node* const nextFreeNode = node->free.next; + VMA_ASSERT(nextFreeNode->free.prev == node); + nextFreeNode->free.prev = node->free.prev; + } +} + +#if VMA_STATS_STRING_ENABLED +void VmaBlockMetadata_Buddy::PrintDetailedMapNode(class VmaJsonWriter& json, const Node* node, VkDeviceSize levelNodeSize) const +{ + switch(node->type) + { + case Node::TYPE_FREE: + PrintDetailedMap_UnusedRange(json, node->offset, levelNodeSize); + break; + case Node::TYPE_ALLOCATION: + { + PrintDetailedMap_Allocation(json, node->offset, node->allocation.alloc); + const VkDeviceSize allocSize = node->allocation.alloc->GetSize(); + if(allocSize < levelNodeSize) + { + PrintDetailedMap_UnusedRange(json, node->offset + allocSize, levelNodeSize - allocSize); + } + } + break; + case Node::TYPE_SPLIT: + { + const VkDeviceSize childrenNodeSize = levelNodeSize / 2; + const Node* const leftChild = node->split.leftChild; + PrintDetailedMapNode(json, leftChild, childrenNodeSize); + const Node* const rightChild = leftChild->buddy; + PrintDetailedMapNode(json, rightChild, childrenNodeSize); + } + break; + default: + VMA_ASSERT(0); + } +} +#endif // #if VMA_STATS_STRING_ENABLED + + +//////////////////////////////////////////////////////////////////////////////// +// class VmaDeviceMemoryBlock + +VmaDeviceMemoryBlock::VmaDeviceMemoryBlock(VmaAllocator hAllocator) : + m_pMetadata(VMA_NULL), + m_MemoryTypeIndex(UINT32_MAX), + m_Id(0), + m_hMemory(VK_NULL_HANDLE), + m_MapCount(0), + m_pMappedData(VMA_NULL) +{ +} + +void VmaDeviceMemoryBlock::Init( + VmaAllocator hAllocator, + VmaPool hParentPool, + uint32_t newMemoryTypeIndex, + VkDeviceMemory newMemory, + VkDeviceSize newSize, + uint32_t id, + uint32_t algorithm) +{ + VMA_ASSERT(m_hMemory == VK_NULL_HANDLE); + + m_hParentPool = hParentPool; + m_MemoryTypeIndex = newMemoryTypeIndex; + m_Id = id; + m_hMemory = newMemory; + + switch(algorithm) + { + case VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT: + m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_Linear)(hAllocator); + break; + case VMA_POOL_CREATE_BUDDY_ALGORITHM_BIT: + m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_Buddy)(hAllocator); + break; + default: + VMA_ASSERT(0); + // Fall-through. + case 0: + m_pMetadata = vma_new(hAllocator, VmaBlockMetadata_Generic)(hAllocator); + } + m_pMetadata->Init(newSize); +} + +void VmaDeviceMemoryBlock::Destroy(VmaAllocator allocator) +{ + // This is the most important assert in the entire library. + // Hitting it means you have some memory leak - unreleased VmaAllocation objects. + VMA_ASSERT(m_pMetadata->IsEmpty() && "Some allocations were not freed before destruction of this memory block!"); + + VMA_ASSERT(m_hMemory != VK_NULL_HANDLE); + allocator->FreeVulkanMemory(m_MemoryTypeIndex, m_pMetadata->GetSize(), m_hMemory); + m_hMemory = VK_NULL_HANDLE; + + vma_delete(allocator, m_pMetadata); + m_pMetadata = VMA_NULL; +} + +bool VmaDeviceMemoryBlock::Validate() const +{ + VMA_VALIDATE((m_hMemory != VK_NULL_HANDLE) && + (m_pMetadata->GetSize() != 0)); + + return m_pMetadata->Validate(); +} + +VkResult VmaDeviceMemoryBlock::CheckCorruption(VmaAllocator hAllocator) +{ + void* pData = nullptr; + VkResult res = Map(hAllocator, 1, &pData); + if(res != VK_SUCCESS) + { + return res; + } + + res = m_pMetadata->CheckCorruption(pData); + + Unmap(hAllocator, 1); + + return res; +} + +VkResult VmaDeviceMemoryBlock::Map(VmaAllocator hAllocator, uint32_t count, void** ppData) +{ + if(count == 0) + { + return VK_SUCCESS; + } + + VmaMutexLock lock(m_Mutex, hAllocator->m_UseMutex); + if(m_MapCount != 0) + { + m_MapCount += count; + VMA_ASSERT(m_pMappedData != VMA_NULL); + if(ppData != VMA_NULL) + { + *ppData = m_pMappedData; + } + return VK_SUCCESS; + } + else + { + VkResult result = (*hAllocator->GetVulkanFunctions().vkMapMemory)( + hAllocator->m_hDevice, + m_hMemory, + 0, // offset + VK_WHOLE_SIZE, + 0, // flags + &m_pMappedData); + if(result == VK_SUCCESS) + { + if(ppData != VMA_NULL) + { + *ppData = m_pMappedData; + } + m_MapCount = count; + } + return result; + } +} + +void VmaDeviceMemoryBlock::Unmap(VmaAllocator hAllocator, uint32_t count) +{ + if(count == 0) + { + return; + } + + VmaMutexLock lock(m_Mutex, hAllocator->m_UseMutex); + if(m_MapCount >= count) + { + m_MapCount -= count; + if(m_MapCount == 0) + { + m_pMappedData = VMA_NULL; + (*hAllocator->GetVulkanFunctions().vkUnmapMemory)(hAllocator->m_hDevice, m_hMemory); + } + } + else + { + VMA_ASSERT(0 && "VkDeviceMemory block is being unmapped while it was not previously mapped."); + } +} + +VkResult VmaDeviceMemoryBlock::WriteMagicValueAroundAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize) +{ + VMA_ASSERT(VMA_DEBUG_MARGIN > 0 && VMA_DEBUG_MARGIN % 4 == 0 && VMA_DEBUG_DETECT_CORRUPTION); + VMA_ASSERT(allocOffset >= VMA_DEBUG_MARGIN); + + void* pData; + VkResult res = Map(hAllocator, 1, &pData); + if(res != VK_SUCCESS) + { + return res; + } + + VmaWriteMagicValue(pData, allocOffset - VMA_DEBUG_MARGIN); + VmaWriteMagicValue(pData, allocOffset + allocSize); + + Unmap(hAllocator, 1); + + return VK_SUCCESS; +} + +VkResult VmaDeviceMemoryBlock::ValidateMagicValueAroundAllocation(VmaAllocator hAllocator, VkDeviceSize allocOffset, VkDeviceSize allocSize) +{ + VMA_ASSERT(VMA_DEBUG_MARGIN > 0 && VMA_DEBUG_MARGIN % 4 == 0 && VMA_DEBUG_DETECT_CORRUPTION); + VMA_ASSERT(allocOffset >= VMA_DEBUG_MARGIN); + + void* pData; + VkResult res = Map(hAllocator, 1, &pData); + if(res != VK_SUCCESS) + { + return res; + } + + if(!VmaValidateMagicValue(pData, allocOffset - VMA_DEBUG_MARGIN)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED BEFORE FREED ALLOCATION!"); + } + else if(!VmaValidateMagicValue(pData, allocOffset + allocSize)) + { + VMA_ASSERT(0 && "MEMORY CORRUPTION DETECTED AFTER FREED ALLOCATION!"); + } + + Unmap(hAllocator, 1); + + return VK_SUCCESS; +} + +VkResult VmaDeviceMemoryBlock::BindBufferMemory( + const VmaAllocator hAllocator, + const VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkBuffer hBuffer, + const void* pNext) +{ + VMA_ASSERT(hAllocation->GetType() == VmaAllocation_T::ALLOCATION_TYPE_BLOCK && + hAllocation->GetBlock() == this); + VMA_ASSERT(allocationLocalOffset < hAllocation->GetSize() && + "Invalid allocationLocalOffset. Did you forget that this offset is relative to the beginning of the allocation, not the whole memory block?"); + const VkDeviceSize memoryOffset = hAllocation->GetOffset() + allocationLocalOffset; + // This lock is important so that we don't call vkBind... and/or vkMap... simultaneously on the same VkDeviceMemory from multiple threads. + VmaMutexLock lock(m_Mutex, hAllocator->m_UseMutex); + return hAllocator->BindVulkanBuffer(m_hMemory, memoryOffset, hBuffer, pNext); +} + +VkResult VmaDeviceMemoryBlock::BindImageMemory( + const VmaAllocator hAllocator, + const VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkImage hImage, + const void* pNext) +{ + VMA_ASSERT(hAllocation->GetType() == VmaAllocation_T::ALLOCATION_TYPE_BLOCK && + hAllocation->GetBlock() == this); + VMA_ASSERT(allocationLocalOffset < hAllocation->GetSize() && + "Invalid allocationLocalOffset. Did you forget that this offset is relative to the beginning of the allocation, not the whole memory block?"); + const VkDeviceSize memoryOffset = hAllocation->GetOffset() + allocationLocalOffset; + // This lock is important so that we don't call vkBind... and/or vkMap... simultaneously on the same VkDeviceMemory from multiple threads. + VmaMutexLock lock(m_Mutex, hAllocator->m_UseMutex); + return hAllocator->BindVulkanImage(m_hMemory, memoryOffset, hImage, pNext); +} + +static void InitStatInfo(VmaStatInfo& outInfo) +{ + memset(&outInfo, 0, sizeof(outInfo)); + outInfo.allocationSizeMin = UINT64_MAX; + outInfo.unusedRangeSizeMin = UINT64_MAX; +} + +// Adds statistics srcInfo into inoutInfo, like: inoutInfo += srcInfo. +static void VmaAddStatInfo(VmaStatInfo& inoutInfo, const VmaStatInfo& srcInfo) +{ + inoutInfo.blockCount += srcInfo.blockCount; + inoutInfo.allocationCount += srcInfo.allocationCount; + inoutInfo.unusedRangeCount += srcInfo.unusedRangeCount; + inoutInfo.usedBytes += srcInfo.usedBytes; + inoutInfo.unusedBytes += srcInfo.unusedBytes; + inoutInfo.allocationSizeMin = VMA_MIN(inoutInfo.allocationSizeMin, srcInfo.allocationSizeMin); + inoutInfo.allocationSizeMax = VMA_MAX(inoutInfo.allocationSizeMax, srcInfo.allocationSizeMax); + inoutInfo.unusedRangeSizeMin = VMA_MIN(inoutInfo.unusedRangeSizeMin, srcInfo.unusedRangeSizeMin); + inoutInfo.unusedRangeSizeMax = VMA_MAX(inoutInfo.unusedRangeSizeMax, srcInfo.unusedRangeSizeMax); +} + +static void VmaPostprocessCalcStatInfo(VmaStatInfo& inoutInfo) +{ + inoutInfo.allocationSizeAvg = (inoutInfo.allocationCount > 0) ? + VmaRoundDiv(inoutInfo.usedBytes, inoutInfo.allocationCount) : 0; + inoutInfo.unusedRangeSizeAvg = (inoutInfo.unusedRangeCount > 0) ? + VmaRoundDiv(inoutInfo.unusedBytes, inoutInfo.unusedRangeCount) : 0; +} + +VmaPool_T::VmaPool_T( + VmaAllocator hAllocator, + const VmaPoolCreateInfo& createInfo, + VkDeviceSize preferredBlockSize) : + m_BlockVector( + hAllocator, + this, // hParentPool + createInfo.memoryTypeIndex, + createInfo.blockSize != 0 ? createInfo.blockSize : preferredBlockSize, + createInfo.minBlockCount, + createInfo.maxBlockCount, + (createInfo.flags & VMA_POOL_CREATE_IGNORE_BUFFER_IMAGE_GRANULARITY_BIT) != 0 ? 1 : hAllocator->GetBufferImageGranularity(), + createInfo.frameInUseCount, + createInfo.blockSize != 0, // explicitBlockSize + createInfo.flags & VMA_POOL_CREATE_ALGORITHM_MASK), // algorithm + m_Id(0), + m_Name(VMA_NULL) +{ +} + +VmaPool_T::~VmaPool_T() +{ +} + +void VmaPool_T::SetName(const char* pName) +{ + const VkAllocationCallbacks* allocs = m_BlockVector.GetAllocator()->GetAllocationCallbacks(); + VmaFreeString(allocs, m_Name); + + if(pName != VMA_NULL) + { + m_Name = VmaCreateStringCopy(allocs, pName); + } + else + { + m_Name = VMA_NULL; + } +} + +#if VMA_STATS_STRING_ENABLED + +#endif // #if VMA_STATS_STRING_ENABLED + +VmaBlockVector::VmaBlockVector( + VmaAllocator hAllocator, + VmaPool hParentPool, + uint32_t memoryTypeIndex, + VkDeviceSize preferredBlockSize, + size_t minBlockCount, + size_t maxBlockCount, + VkDeviceSize bufferImageGranularity, + uint32_t frameInUseCount, + bool explicitBlockSize, + uint32_t algorithm) : + m_hAllocator(hAllocator), + m_hParentPool(hParentPool), + m_MemoryTypeIndex(memoryTypeIndex), + m_PreferredBlockSize(preferredBlockSize), + m_MinBlockCount(minBlockCount), + m_MaxBlockCount(maxBlockCount), + m_BufferImageGranularity(bufferImageGranularity), + m_FrameInUseCount(frameInUseCount), + m_ExplicitBlockSize(explicitBlockSize), + m_Algorithm(algorithm), + m_HasEmptyBlock(false), + m_Blocks(VmaStlAllocator(hAllocator->GetAllocationCallbacks())), + m_NextBlockId(0) +{ +} + +VmaBlockVector::~VmaBlockVector() +{ + for(size_t i = m_Blocks.size(); i--; ) + { + m_Blocks[i]->Destroy(m_hAllocator); + vma_delete(m_hAllocator, m_Blocks[i]); + } +} + +VkResult VmaBlockVector::CreateMinBlocks() +{ + for(size_t i = 0; i < m_MinBlockCount; ++i) + { + VkResult res = CreateBlock(m_PreferredBlockSize, VMA_NULL); + if(res != VK_SUCCESS) + { + return res; + } + } + return VK_SUCCESS; +} + +void VmaBlockVector::GetPoolStats(VmaPoolStats* pStats) +{ + VmaMutexLockRead lock(m_Mutex, m_hAllocator->m_UseMutex); + + const size_t blockCount = m_Blocks.size(); + + pStats->size = 0; + pStats->unusedSize = 0; + pStats->allocationCount = 0; + pStats->unusedRangeCount = 0; + pStats->unusedRangeSizeMax = 0; + pStats->blockCount = blockCount; + + for(uint32_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + { + const VmaDeviceMemoryBlock* const pBlock = m_Blocks[blockIndex]; + VMA_ASSERT(pBlock); + VMA_HEAVY_ASSERT(pBlock->Validate()); + pBlock->m_pMetadata->AddPoolStats(*pStats); + } +} + +bool VmaBlockVector::IsEmpty() +{ + VmaMutexLockRead lock(m_Mutex, m_hAllocator->m_UseMutex); + return m_Blocks.empty(); +} + +bool VmaBlockVector::IsCorruptionDetectionEnabled() const +{ + const uint32_t requiredMemFlags = VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT; + return (VMA_DEBUG_DETECT_CORRUPTION != 0) && + (VMA_DEBUG_MARGIN > 0) && + (m_Algorithm == 0 || m_Algorithm == VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT) && + (m_hAllocator->m_MemProps.memoryTypes[m_MemoryTypeIndex].propertyFlags & requiredMemFlags) == requiredMemFlags; +} + +static const uint32_t VMA_ALLOCATION_TRY_COUNT = 32; + +VkResult VmaBlockVector::Allocate( + uint32_t currentFrameIndex, + VkDeviceSize size, + VkDeviceSize alignment, + const VmaAllocationCreateInfo& createInfo, + VmaSuballocationType suballocType, + size_t allocationCount, + VmaAllocation* pAllocations) +{ + size_t allocIndex; + VkResult res = VK_SUCCESS; + + if(IsCorruptionDetectionEnabled()) + { + size = VmaAlignUp(size, sizeof(VMA_CORRUPTION_DETECTION_MAGIC_VALUE)); + alignment = VmaAlignUp(alignment, sizeof(VMA_CORRUPTION_DETECTION_MAGIC_VALUE)); + } + + { + VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); + for(allocIndex = 0; allocIndex < allocationCount; ++allocIndex) + { + res = AllocatePage( + currentFrameIndex, + size, + alignment, + createInfo, + suballocType, + pAllocations + allocIndex); + if(res != VK_SUCCESS) + { + break; + } + } + } + + if(res != VK_SUCCESS) + { + // Free all already created allocations. + while(allocIndex--) + { + Free(pAllocations[allocIndex]); + } + memset(pAllocations, 0, sizeof(VmaAllocation) * allocationCount); + } + + return res; +} + +VkResult VmaBlockVector::AllocatePage( + uint32_t currentFrameIndex, + VkDeviceSize size, + VkDeviceSize alignment, + const VmaAllocationCreateInfo& createInfo, + VmaSuballocationType suballocType, + VmaAllocation* pAllocation) +{ + const bool isUpperAddress = (createInfo.flags & VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0; + bool canMakeOtherLost = (createInfo.flags & VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT) != 0; + const bool mapped = (createInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0; + const bool isUserDataString = (createInfo.flags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0; + + VkDeviceSize freeMemory; + { + const uint32_t heapIndex = m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex); + VmaBudget heapBudget = {}; + m_hAllocator->GetBudget(&heapBudget, heapIndex, 1); + freeMemory = (heapBudget.usage < heapBudget.budget) ? (heapBudget.budget - heapBudget.usage) : 0; + } + + const bool canFallbackToDedicated = !IsCustomPool(); + const bool canCreateNewBlock = + ((createInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) == 0) && + (m_Blocks.size() < m_MaxBlockCount) && + (freeMemory >= size || !canFallbackToDedicated); + uint32_t strategy = createInfo.flags & VMA_ALLOCATION_CREATE_STRATEGY_MASK; + + // If linearAlgorithm is used, canMakeOtherLost is available only when used as ring buffer. + // Which in turn is available only when maxBlockCount = 1. + if(m_Algorithm == VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT && m_MaxBlockCount > 1) + { + canMakeOtherLost = false; + } + + // Upper address can only be used with linear allocator and within single memory block. + if(isUpperAddress && + (m_Algorithm != VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT || m_MaxBlockCount > 1)) + { + return VK_ERROR_FEATURE_NOT_PRESENT; + } + + // Validate strategy. + switch(strategy) + { + case 0: + strategy = VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT; + break; + case VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT: + case VMA_ALLOCATION_CREATE_STRATEGY_WORST_FIT_BIT: + case VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT: + break; + default: + return VK_ERROR_FEATURE_NOT_PRESENT; + } + + // Early reject: requested allocation size is larger that maximum block size for this block vector. + if(size + 2 * VMA_DEBUG_MARGIN > m_PreferredBlockSize) + { + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + + /* + Under certain condition, this whole section can be skipped for optimization, so + we move on directly to trying to allocate with canMakeOtherLost. That's the case + e.g. for custom pools with linear algorithm. + */ + if(!canMakeOtherLost || canCreateNewBlock) + { + // 1. Search existing allocations. Try to allocate without making other allocations lost. + VmaAllocationCreateFlags allocFlagsCopy = createInfo.flags; + allocFlagsCopy &= ~VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT; + + if(m_Algorithm == VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT) + { + // Use only last block. + if(!m_Blocks.empty()) + { + VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks.back(); + VMA_ASSERT(pCurrBlock); + VkResult res = AllocateFromBlock( + pCurrBlock, + currentFrameIndex, + size, + alignment, + allocFlagsCopy, + createInfo.pUserData, + suballocType, + strategy, + pAllocation); + if(res == VK_SUCCESS) + { + VMA_DEBUG_LOG(" Returned from last block #%u", pCurrBlock->GetId()); + return VK_SUCCESS; + } + } + } + else + { + if(strategy == VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT) + { + // Forward order in m_Blocks - prefer blocks with smallest amount of free space. + for(size_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex ) + { + VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; + VMA_ASSERT(pCurrBlock); + VkResult res = AllocateFromBlock( + pCurrBlock, + currentFrameIndex, + size, + alignment, + allocFlagsCopy, + createInfo.pUserData, + suballocType, + strategy, + pAllocation); + if(res == VK_SUCCESS) + { + VMA_DEBUG_LOG(" Returned from existing block #%u", pCurrBlock->GetId()); + return VK_SUCCESS; + } + } + } + else // WORST_FIT, FIRST_FIT + { + // Backward order in m_Blocks - prefer blocks with largest amount of free space. + for(size_t blockIndex = m_Blocks.size(); blockIndex--; ) + { + VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; + VMA_ASSERT(pCurrBlock); + VkResult res = AllocateFromBlock( + pCurrBlock, + currentFrameIndex, + size, + alignment, + allocFlagsCopy, + createInfo.pUserData, + suballocType, + strategy, + pAllocation); + if(res == VK_SUCCESS) + { + VMA_DEBUG_LOG(" Returned from existing block #%u", pCurrBlock->GetId()); + return VK_SUCCESS; + } + } + } + } + + // 2. Try to create new block. + if(canCreateNewBlock) + { + // Calculate optimal size for new block. + VkDeviceSize newBlockSize = m_PreferredBlockSize; + uint32_t newBlockSizeShift = 0; + const uint32_t NEW_BLOCK_SIZE_SHIFT_MAX = 3; + + if(!m_ExplicitBlockSize) + { + // Allocate 1/8, 1/4, 1/2 as first blocks. + const VkDeviceSize maxExistingBlockSize = CalcMaxBlockSize(); + for(uint32_t i = 0; i < NEW_BLOCK_SIZE_SHIFT_MAX; ++i) + { + const VkDeviceSize smallerNewBlockSize = newBlockSize / 2; + if(smallerNewBlockSize > maxExistingBlockSize && smallerNewBlockSize >= size * 2) + { + newBlockSize = smallerNewBlockSize; + ++newBlockSizeShift; + } + else + { + break; + } + } + } + + size_t newBlockIndex = 0; + VkResult res = (newBlockSize <= freeMemory || !canFallbackToDedicated) ? + CreateBlock(newBlockSize, &newBlockIndex) : VK_ERROR_OUT_OF_DEVICE_MEMORY; + // Allocation of this size failed? Try 1/2, 1/4, 1/8 of m_PreferredBlockSize. + if(!m_ExplicitBlockSize) + { + while(res < 0 && newBlockSizeShift < NEW_BLOCK_SIZE_SHIFT_MAX) + { + const VkDeviceSize smallerNewBlockSize = newBlockSize / 2; + if(smallerNewBlockSize >= size) + { + newBlockSize = smallerNewBlockSize; + ++newBlockSizeShift; + res = (newBlockSize <= freeMemory || !canFallbackToDedicated) ? + CreateBlock(newBlockSize, &newBlockIndex) : VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + else + { + break; + } + } + } + + if(res == VK_SUCCESS) + { + VmaDeviceMemoryBlock* const pBlock = m_Blocks[newBlockIndex]; + VMA_ASSERT(pBlock->m_pMetadata->GetSize() >= size); + + res = AllocateFromBlock( + pBlock, + currentFrameIndex, + size, + alignment, + allocFlagsCopy, + createInfo.pUserData, + suballocType, + strategy, + pAllocation); + if(res == VK_SUCCESS) + { + VMA_DEBUG_LOG(" Created new block #%u Size=%llu", pBlock->GetId(), newBlockSize); + return VK_SUCCESS; + } + else + { + // Allocation from new block failed, possibly due to VMA_DEBUG_MARGIN or alignment. + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + } + } + } + + // 3. Try to allocate from existing blocks with making other allocations lost. + if(canMakeOtherLost) + { + uint32_t tryIndex = 0; + for(; tryIndex < VMA_ALLOCATION_TRY_COUNT; ++tryIndex) + { + VmaDeviceMemoryBlock* pBestRequestBlock = VMA_NULL; + VmaAllocationRequest bestRequest = {}; + VkDeviceSize bestRequestCost = VK_WHOLE_SIZE; + + // 1. Search existing allocations. + if(strategy == VMA_ALLOCATION_CREATE_STRATEGY_BEST_FIT_BIT) + { + // Forward order in m_Blocks - prefer blocks with smallest amount of free space. + for(size_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex ) + { + VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; + VMA_ASSERT(pCurrBlock); + VmaAllocationRequest currRequest = {}; + if(pCurrBlock->m_pMetadata->CreateAllocationRequest( + currentFrameIndex, + m_FrameInUseCount, + m_BufferImageGranularity, + size, + alignment, + (createInfo.flags & VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0, + suballocType, + canMakeOtherLost, + strategy, + &currRequest)) + { + const VkDeviceSize currRequestCost = currRequest.CalcCost(); + if(pBestRequestBlock == VMA_NULL || + currRequestCost < bestRequestCost) + { + pBestRequestBlock = pCurrBlock; + bestRequest = currRequest; + bestRequestCost = currRequestCost; + + if(bestRequestCost == 0) + { + break; + } + } + } + } + } + else // WORST_FIT, FIRST_FIT + { + // Backward order in m_Blocks - prefer blocks with largest amount of free space. + for(size_t blockIndex = m_Blocks.size(); blockIndex--; ) + { + VmaDeviceMemoryBlock* const pCurrBlock = m_Blocks[blockIndex]; + VMA_ASSERT(pCurrBlock); + VmaAllocationRequest currRequest = {}; + if(pCurrBlock->m_pMetadata->CreateAllocationRequest( + currentFrameIndex, + m_FrameInUseCount, + m_BufferImageGranularity, + size, + alignment, + (createInfo.flags & VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0, + suballocType, + canMakeOtherLost, + strategy, + &currRequest)) + { + const VkDeviceSize currRequestCost = currRequest.CalcCost(); + if(pBestRequestBlock == VMA_NULL || + currRequestCost < bestRequestCost || + strategy == VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT) + { + pBestRequestBlock = pCurrBlock; + bestRequest = currRequest; + bestRequestCost = currRequestCost; + + if(bestRequestCost == 0 || + strategy == VMA_ALLOCATION_CREATE_STRATEGY_FIRST_FIT_BIT) + { + break; + } + } + } + } + } + + if(pBestRequestBlock != VMA_NULL) + { + if(mapped) + { + VkResult res = pBestRequestBlock->Map(m_hAllocator, 1, VMA_NULL); + if(res != VK_SUCCESS) + { + return res; + } + } + + if(pBestRequestBlock->m_pMetadata->MakeRequestedAllocationsLost( + currentFrameIndex, + m_FrameInUseCount, + &bestRequest)) + { + // Allocate from this pBlock. + *pAllocation = m_hAllocator->m_AllocationObjectAllocator.Allocate(currentFrameIndex, isUserDataString); + pBestRequestBlock->m_pMetadata->Alloc(bestRequest, suballocType, size, *pAllocation); + UpdateHasEmptyBlock(); + (*pAllocation)->InitBlockAllocation( + pBestRequestBlock, + bestRequest.offset, + alignment, + size, + m_MemoryTypeIndex, + suballocType, + mapped, + (createInfo.flags & VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT) != 0); + VMA_HEAVY_ASSERT(pBestRequestBlock->Validate()); + VMA_DEBUG_LOG(" Returned from existing block"); + (*pAllocation)->SetUserData(m_hAllocator, createInfo.pUserData); + m_hAllocator->m_Budget.AddAllocation(m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex), size); + if(VMA_DEBUG_INITIALIZE_ALLOCATIONS) + { + m_hAllocator->FillAllocation(*pAllocation, VMA_ALLOCATION_FILL_PATTERN_CREATED); + } + if(IsCorruptionDetectionEnabled()) + { + VkResult res = pBestRequestBlock->WriteMagicValueAroundAllocation(m_hAllocator, bestRequest.offset, size); + VMA_ASSERT(res == VK_SUCCESS && "Couldn't map block memory to write magic value."); + } + return VK_SUCCESS; + } + // else: Some allocations must have been touched while we are here. Next try. + } + else + { + // Could not find place in any of the blocks - break outer loop. + break; + } + } + /* Maximum number of tries exceeded - a very unlike event when many other + threads are simultaneously touching allocations making it impossible to make + lost at the same time as we try to allocate. */ + if(tryIndex == VMA_ALLOCATION_TRY_COUNT) + { + return VK_ERROR_TOO_MANY_OBJECTS; + } + } + + return VK_ERROR_OUT_OF_DEVICE_MEMORY; +} + +void VmaBlockVector::Free( + const VmaAllocation hAllocation) +{ + VmaDeviceMemoryBlock* pBlockToDelete = VMA_NULL; + + bool budgetExceeded = false; + { + const uint32_t heapIndex = m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex); + VmaBudget heapBudget = {}; + m_hAllocator->GetBudget(&heapBudget, heapIndex, 1); + budgetExceeded = heapBudget.usage >= heapBudget.budget; + } + + // Scope for lock. + { + VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); + + VmaDeviceMemoryBlock* pBlock = hAllocation->GetBlock(); + + if(IsCorruptionDetectionEnabled()) + { + VkResult res = pBlock->ValidateMagicValueAroundAllocation(m_hAllocator, hAllocation->GetOffset(), hAllocation->GetSize()); + VMA_ASSERT(res == VK_SUCCESS && "Couldn't map block memory to validate magic value."); + } + + if(hAllocation->IsPersistentMap()) + { + pBlock->Unmap(m_hAllocator, 1); + } + + pBlock->m_pMetadata->Free(hAllocation); + VMA_HEAVY_ASSERT(pBlock->Validate()); + + VMA_DEBUG_LOG(" Freed from MemoryTypeIndex=%u", m_MemoryTypeIndex); + + const bool canDeleteBlock = m_Blocks.size() > m_MinBlockCount; + // pBlock became empty after this deallocation. + if(pBlock->m_pMetadata->IsEmpty()) + { + // Already has empty block. We don't want to have two, so delete this one. + if((m_HasEmptyBlock || budgetExceeded) && canDeleteBlock) + { + pBlockToDelete = pBlock; + Remove(pBlock); + } + // else: We now have an empty block - leave it. + } + // pBlock didn't become empty, but we have another empty block - find and free that one. + // (This is optional, heuristics.) + else if(m_HasEmptyBlock && canDeleteBlock) + { + VmaDeviceMemoryBlock* pLastBlock = m_Blocks.back(); + if(pLastBlock->m_pMetadata->IsEmpty()) + { + pBlockToDelete = pLastBlock; + m_Blocks.pop_back(); + } + } + + UpdateHasEmptyBlock(); + IncrementallySortBlocks(); + } + + // Destruction of a free block. Deferred until this point, outside of mutex + // lock, for performance reason. + if(pBlockToDelete != VMA_NULL) + { + VMA_DEBUG_LOG(" Deleted empty block"); + pBlockToDelete->Destroy(m_hAllocator); + vma_delete(m_hAllocator, pBlockToDelete); + } +} + +VkDeviceSize VmaBlockVector::CalcMaxBlockSize() const +{ + VkDeviceSize result = 0; + for(size_t i = m_Blocks.size(); i--; ) + { + result = VMA_MAX(result, m_Blocks[i]->m_pMetadata->GetSize()); + if(result >= m_PreferredBlockSize) + { + break; + } + } + return result; +} + +void VmaBlockVector::Remove(VmaDeviceMemoryBlock* pBlock) +{ + for(uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) + { + if(m_Blocks[blockIndex] == pBlock) + { + VmaVectorRemove(m_Blocks, blockIndex); + return; + } + } + VMA_ASSERT(0); +} + +void VmaBlockVector::IncrementallySortBlocks() +{ + if(m_Algorithm != VMA_POOL_CREATE_LINEAR_ALGORITHM_BIT) + { + // Bubble sort only until first swap. + for(size_t i = 1; i < m_Blocks.size(); ++i) + { + if(m_Blocks[i - 1]->m_pMetadata->GetSumFreeSize() > m_Blocks[i]->m_pMetadata->GetSumFreeSize()) + { + VMA_SWAP(m_Blocks[i - 1], m_Blocks[i]); + return; + } + } + } +} + +VkResult VmaBlockVector::AllocateFromBlock( + VmaDeviceMemoryBlock* pBlock, + uint32_t currentFrameIndex, + VkDeviceSize size, + VkDeviceSize alignment, + VmaAllocationCreateFlags allocFlags, + void* pUserData, + VmaSuballocationType suballocType, + uint32_t strategy, + VmaAllocation* pAllocation) +{ + VMA_ASSERT((allocFlags & VMA_ALLOCATION_CREATE_CAN_MAKE_OTHER_LOST_BIT) == 0); + const bool isUpperAddress = (allocFlags & VMA_ALLOCATION_CREATE_UPPER_ADDRESS_BIT) != 0; + const bool mapped = (allocFlags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0; + const bool isUserDataString = (allocFlags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0; + + VmaAllocationRequest currRequest = {}; + if(pBlock->m_pMetadata->CreateAllocationRequest( + currentFrameIndex, + m_FrameInUseCount, + m_BufferImageGranularity, + size, + alignment, + isUpperAddress, + suballocType, + false, // canMakeOtherLost + strategy, + &currRequest)) + { + // Allocate from pCurrBlock. + VMA_ASSERT(currRequest.itemsToMakeLostCount == 0); + + if(mapped) + { + VkResult res = pBlock->Map(m_hAllocator, 1, VMA_NULL); + if(res != VK_SUCCESS) + { + return res; + } + } + + *pAllocation = m_hAllocator->m_AllocationObjectAllocator.Allocate(currentFrameIndex, isUserDataString); + pBlock->m_pMetadata->Alloc(currRequest, suballocType, size, *pAllocation); + UpdateHasEmptyBlock(); + (*pAllocation)->InitBlockAllocation( + pBlock, + currRequest.offset, + alignment, + size, + m_MemoryTypeIndex, + suballocType, + mapped, + (allocFlags & VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT) != 0); + VMA_HEAVY_ASSERT(pBlock->Validate()); + (*pAllocation)->SetUserData(m_hAllocator, pUserData); + m_hAllocator->m_Budget.AddAllocation(m_hAllocator->MemoryTypeIndexToHeapIndex(m_MemoryTypeIndex), size); + if(VMA_DEBUG_INITIALIZE_ALLOCATIONS) + { + m_hAllocator->FillAllocation(*pAllocation, VMA_ALLOCATION_FILL_PATTERN_CREATED); + } + if(IsCorruptionDetectionEnabled()) + { + VkResult res = pBlock->WriteMagicValueAroundAllocation(m_hAllocator, currRequest.offset, size); + VMA_ASSERT(res == VK_SUCCESS && "Couldn't map block memory to write magic value."); + } + return VK_SUCCESS; + } + return VK_ERROR_OUT_OF_DEVICE_MEMORY; +} + +VkResult VmaBlockVector::CreateBlock(VkDeviceSize blockSize, size_t* pNewBlockIndex) +{ + VkMemoryAllocateInfo allocInfo = { VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO }; + allocInfo.memoryTypeIndex = m_MemoryTypeIndex; + allocInfo.allocationSize = blockSize; + +#if VMA_BUFFER_DEVICE_ADDRESS + // Every standalone block can potentially contain a buffer with VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT - always enable the feature. + VkMemoryAllocateFlagsInfoKHR allocFlagsInfo = { VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_FLAGS_INFO_KHR }; + if(m_hAllocator->m_UseKhrBufferDeviceAddress) + { + allocFlagsInfo.flags = VK_MEMORY_ALLOCATE_DEVICE_ADDRESS_BIT_KHR; + VmaPnextChainPushFront(&allocInfo, &allocFlagsInfo); + } +#endif // #if VMA_BUFFER_DEVICE_ADDRESS + + VkDeviceMemory mem = VK_NULL_HANDLE; + VkResult res = m_hAllocator->AllocateVulkanMemory(&allocInfo, &mem); + if(res < 0) + { + return res; + } + + // New VkDeviceMemory successfully created. + + // Create new Allocation for it. + VmaDeviceMemoryBlock* const pBlock = vma_new(m_hAllocator, VmaDeviceMemoryBlock)(m_hAllocator); + pBlock->Init( + m_hAllocator, + m_hParentPool, + m_MemoryTypeIndex, + mem, + allocInfo.allocationSize, + m_NextBlockId++, + m_Algorithm); + + m_Blocks.push_back(pBlock); + if(pNewBlockIndex != VMA_NULL) + { + *pNewBlockIndex = m_Blocks.size() - 1; + } + + return VK_SUCCESS; +} + +void VmaBlockVector::ApplyDefragmentationMovesCpu( + class VmaBlockVectorDefragmentationContext* pDefragCtx, + const VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves) +{ + const size_t blockCount = m_Blocks.size(); + const bool isNonCoherent = m_hAllocator->IsMemoryTypeNonCoherent(m_MemoryTypeIndex); + + enum BLOCK_FLAG + { + BLOCK_FLAG_USED = 0x00000001, + BLOCK_FLAG_MAPPED_FOR_DEFRAGMENTATION = 0x00000002, + }; + + struct BlockInfo + { + uint32_t flags; + void* pMappedData; + }; + VmaVector< BlockInfo, VmaStlAllocator > + blockInfo(blockCount, BlockInfo(), VmaStlAllocator(m_hAllocator->GetAllocationCallbacks())); + memset(blockInfo.data(), 0, blockCount * sizeof(BlockInfo)); + + // Go over all moves. Mark blocks that are used with BLOCK_FLAG_USED. + const size_t moveCount = moves.size(); + for(size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) + { + const VmaDefragmentationMove& move = moves[moveIndex]; + blockInfo[move.srcBlockIndex].flags |= BLOCK_FLAG_USED; + blockInfo[move.dstBlockIndex].flags |= BLOCK_FLAG_USED; + } + + VMA_ASSERT(pDefragCtx->res == VK_SUCCESS); + + // Go over all blocks. Get mapped pointer or map if necessary. + for(size_t blockIndex = 0; pDefragCtx->res == VK_SUCCESS && blockIndex < blockCount; ++blockIndex) + { + BlockInfo& currBlockInfo = blockInfo[blockIndex]; + VmaDeviceMemoryBlock* pBlock = m_Blocks[blockIndex]; + if((currBlockInfo.flags & BLOCK_FLAG_USED) != 0) + { + currBlockInfo.pMappedData = pBlock->GetMappedData(); + // It is not originally mapped - map it. + if(currBlockInfo.pMappedData == VMA_NULL) + { + pDefragCtx->res = pBlock->Map(m_hAllocator, 1, &currBlockInfo.pMappedData); + if(pDefragCtx->res == VK_SUCCESS) + { + currBlockInfo.flags |= BLOCK_FLAG_MAPPED_FOR_DEFRAGMENTATION; + } + } + } + } + + // Go over all moves. Do actual data transfer. + if(pDefragCtx->res == VK_SUCCESS) + { + const VkDeviceSize nonCoherentAtomSize = m_hAllocator->m_PhysicalDeviceProperties.limits.nonCoherentAtomSize; + VkMappedMemoryRange memRange = { VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE }; + + for(size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) + { + const VmaDefragmentationMove& move = moves[moveIndex]; + + const BlockInfo& srcBlockInfo = blockInfo[move.srcBlockIndex]; + const BlockInfo& dstBlockInfo = blockInfo[move.dstBlockIndex]; + + VMA_ASSERT(srcBlockInfo.pMappedData && dstBlockInfo.pMappedData); + + // Invalidate source. + if(isNonCoherent) + { + VmaDeviceMemoryBlock* const pSrcBlock = m_Blocks[move.srcBlockIndex]; + memRange.memory = pSrcBlock->GetDeviceMemory(); + memRange.offset = VmaAlignDown(move.srcOffset, nonCoherentAtomSize); + memRange.size = VMA_MIN( + VmaAlignUp(move.size + (move.srcOffset - memRange.offset), nonCoherentAtomSize), + pSrcBlock->m_pMetadata->GetSize() - memRange.offset); + (*m_hAllocator->GetVulkanFunctions().vkInvalidateMappedMemoryRanges)(m_hAllocator->m_hDevice, 1, &memRange); + } + + // THE PLACE WHERE ACTUAL DATA COPY HAPPENS. + memmove( + reinterpret_cast(dstBlockInfo.pMappedData) + move.dstOffset, + reinterpret_cast(srcBlockInfo.pMappedData) + move.srcOffset, + static_cast(move.size)); + + if(IsCorruptionDetectionEnabled()) + { + VmaWriteMagicValue(dstBlockInfo.pMappedData, move.dstOffset - VMA_DEBUG_MARGIN); + VmaWriteMagicValue(dstBlockInfo.pMappedData, move.dstOffset + move.size); + } + + // Flush destination. + if(isNonCoherent) + { + VmaDeviceMemoryBlock* const pDstBlock = m_Blocks[move.dstBlockIndex]; + memRange.memory = pDstBlock->GetDeviceMemory(); + memRange.offset = VmaAlignDown(move.dstOffset, nonCoherentAtomSize); + memRange.size = VMA_MIN( + VmaAlignUp(move.size + (move.dstOffset - memRange.offset), nonCoherentAtomSize), + pDstBlock->m_pMetadata->GetSize() - memRange.offset); + (*m_hAllocator->GetVulkanFunctions().vkFlushMappedMemoryRanges)(m_hAllocator->m_hDevice, 1, &memRange); + } + } + } + + // Go over all blocks in reverse order. Unmap those that were mapped just for defragmentation. + // Regardless of pCtx->res == VK_SUCCESS. + for(size_t blockIndex = blockCount; blockIndex--; ) + { + const BlockInfo& currBlockInfo = blockInfo[blockIndex]; + if((currBlockInfo.flags & BLOCK_FLAG_MAPPED_FOR_DEFRAGMENTATION) != 0) + { + VmaDeviceMemoryBlock* pBlock = m_Blocks[blockIndex]; + pBlock->Unmap(m_hAllocator, 1); + } + } +} + +void VmaBlockVector::ApplyDefragmentationMovesGpu( + class VmaBlockVectorDefragmentationContext* pDefragCtx, + VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves, + VkCommandBuffer commandBuffer) +{ + const size_t blockCount = m_Blocks.size(); + + pDefragCtx->blockContexts.resize(blockCount); + memset(pDefragCtx->blockContexts.data(), 0, blockCount * sizeof(VmaBlockDefragmentationContext)); + + // Go over all moves. Mark blocks that are used with BLOCK_FLAG_USED. + const size_t moveCount = moves.size(); + for(size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) + { + const VmaDefragmentationMove& move = moves[moveIndex]; + + //if(move.type == VMA_ALLOCATION_TYPE_UNKNOWN) + { + // Old school move still require us to map the whole block + pDefragCtx->blockContexts[move.srcBlockIndex].flags |= VmaBlockDefragmentationContext::BLOCK_FLAG_USED; + pDefragCtx->blockContexts[move.dstBlockIndex].flags |= VmaBlockDefragmentationContext::BLOCK_FLAG_USED; + } + } + + VMA_ASSERT(pDefragCtx->res == VK_SUCCESS); + + // Go over all blocks. Create and bind buffer for whole block if necessary. + { + VkBufferCreateInfo bufCreateInfo; + VmaFillGpuDefragmentationBufferCreateInfo(bufCreateInfo); + + for(size_t blockIndex = 0; pDefragCtx->res == VK_SUCCESS && blockIndex < blockCount; ++blockIndex) + { + VmaBlockDefragmentationContext& currBlockCtx = pDefragCtx->blockContexts[blockIndex]; + VmaDeviceMemoryBlock* pBlock = m_Blocks[blockIndex]; + if((currBlockCtx.flags & VmaBlockDefragmentationContext::BLOCK_FLAG_USED) != 0) + { + bufCreateInfo.size = pBlock->m_pMetadata->GetSize(); + pDefragCtx->res = (*m_hAllocator->GetVulkanFunctions().vkCreateBuffer)( + m_hAllocator->m_hDevice, &bufCreateInfo, m_hAllocator->GetAllocationCallbacks(), &currBlockCtx.hBuffer); + if(pDefragCtx->res == VK_SUCCESS) + { + pDefragCtx->res = (*m_hAllocator->GetVulkanFunctions().vkBindBufferMemory)( + m_hAllocator->m_hDevice, currBlockCtx.hBuffer, pBlock->GetDeviceMemory(), 0); + } + } + } + } + + // Go over all moves. Post data transfer commands to command buffer. + if(pDefragCtx->res == VK_SUCCESS) + { + for(size_t moveIndex = 0; moveIndex < moveCount; ++moveIndex) + { + const VmaDefragmentationMove& move = moves[moveIndex]; + + const VmaBlockDefragmentationContext& srcBlockCtx = pDefragCtx->blockContexts[move.srcBlockIndex]; + const VmaBlockDefragmentationContext& dstBlockCtx = pDefragCtx->blockContexts[move.dstBlockIndex]; + + VMA_ASSERT(srcBlockCtx.hBuffer && dstBlockCtx.hBuffer); + + VkBufferCopy region = { + move.srcOffset, + move.dstOffset, + move.size }; + (*m_hAllocator->GetVulkanFunctions().vkCmdCopyBuffer)( + commandBuffer, srcBlockCtx.hBuffer, dstBlockCtx.hBuffer, 1, ®ion); + } + } + + // Save buffers to defrag context for later destruction. + if(pDefragCtx->res == VK_SUCCESS && moveCount > 0) + { + pDefragCtx->res = VK_NOT_READY; + } +} + +void VmaBlockVector::FreeEmptyBlocks(VmaDefragmentationStats* pDefragmentationStats) +{ + for(size_t blockIndex = m_Blocks.size(); blockIndex--; ) + { + VmaDeviceMemoryBlock* pBlock = m_Blocks[blockIndex]; + if(pBlock->m_pMetadata->IsEmpty()) + { + if(m_Blocks.size() > m_MinBlockCount) + { + if(pDefragmentationStats != VMA_NULL) + { + ++pDefragmentationStats->deviceMemoryBlocksFreed; + pDefragmentationStats->bytesFreed += pBlock->m_pMetadata->GetSize(); + } + + VmaVectorRemove(m_Blocks, blockIndex); + pBlock->Destroy(m_hAllocator); + vma_delete(m_hAllocator, pBlock); + } + else + { + break; + } + } + } + UpdateHasEmptyBlock(); +} + +void VmaBlockVector::UpdateHasEmptyBlock() +{ + m_HasEmptyBlock = false; + for(size_t index = 0, count = m_Blocks.size(); index < count; ++index) + { + VmaDeviceMemoryBlock* const pBlock = m_Blocks[index]; + if(pBlock->m_pMetadata->IsEmpty()) + { + m_HasEmptyBlock = true; + break; + } + } +} + +#if VMA_STATS_STRING_ENABLED + +void VmaBlockVector::PrintDetailedMap(class VmaJsonWriter& json) +{ + VmaMutexLockRead lock(m_Mutex, m_hAllocator->m_UseMutex); + + json.BeginObject(); + + if(IsCustomPool()) + { + const char* poolName = m_hParentPool->GetName(); + if(poolName != VMA_NULL && poolName[0] != '\0') + { + json.WriteString("Name"); + json.WriteString(poolName); + } + + json.WriteString("MemoryTypeIndex"); + json.WriteNumber(m_MemoryTypeIndex); + + json.WriteString("BlockSize"); + json.WriteNumber(m_PreferredBlockSize); + + json.WriteString("BlockCount"); + json.BeginObject(true); + if(m_MinBlockCount > 0) + { + json.WriteString("Min"); + json.WriteNumber((uint64_t)m_MinBlockCount); + } + if(m_MaxBlockCount < SIZE_MAX) + { + json.WriteString("Max"); + json.WriteNumber((uint64_t)m_MaxBlockCount); + } + json.WriteString("Cur"); + json.WriteNumber((uint64_t)m_Blocks.size()); + json.EndObject(); + + if(m_FrameInUseCount > 0) + { + json.WriteString("FrameInUseCount"); + json.WriteNumber(m_FrameInUseCount); + } + + if(m_Algorithm != 0) + { + json.WriteString("Algorithm"); + json.WriteString(VmaAlgorithmToStr(m_Algorithm)); + } + } + else + { + json.WriteString("PreferredBlockSize"); + json.WriteNumber(m_PreferredBlockSize); + } + + json.WriteString("Blocks"); + json.BeginObject(); + for(size_t i = 0; i < m_Blocks.size(); ++i) + { + json.BeginString(); + json.ContinueString(m_Blocks[i]->GetId()); + json.EndString(); + + m_Blocks[i]->m_pMetadata->PrintDetailedMap(json); + } + json.EndObject(); + + json.EndObject(); +} + +#endif // #if VMA_STATS_STRING_ENABLED + +void VmaBlockVector::Defragment( + class VmaBlockVectorDefragmentationContext* pCtx, + VmaDefragmentationStats* pStats, VmaDefragmentationFlags flags, + VkDeviceSize& maxCpuBytesToMove, uint32_t& maxCpuAllocationsToMove, + VkDeviceSize& maxGpuBytesToMove, uint32_t& maxGpuAllocationsToMove, + VkCommandBuffer commandBuffer) +{ + pCtx->res = VK_SUCCESS; + + const VkMemoryPropertyFlags memPropFlags = + m_hAllocator->m_MemProps.memoryTypes[m_MemoryTypeIndex].propertyFlags; + const bool isHostVisible = (memPropFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) != 0; + + const bool canDefragmentOnCpu = maxCpuBytesToMove > 0 && maxCpuAllocationsToMove > 0 && + isHostVisible; + const bool canDefragmentOnGpu = maxGpuBytesToMove > 0 && maxGpuAllocationsToMove > 0 && + !IsCorruptionDetectionEnabled() && + ((1u << m_MemoryTypeIndex) & m_hAllocator->GetGpuDefragmentationMemoryTypeBits()) != 0; + + // There are options to defragment this memory type. + if(canDefragmentOnCpu || canDefragmentOnGpu) + { + bool defragmentOnGpu; + // There is only one option to defragment this memory type. + if(canDefragmentOnGpu != canDefragmentOnCpu) + { + defragmentOnGpu = canDefragmentOnGpu; + } + // Both options are available: Heuristics to choose the best one. + else + { + defragmentOnGpu = (memPropFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT) != 0 || + m_hAllocator->IsIntegratedGpu(); + } + + bool overlappingMoveSupported = !defragmentOnGpu; + + if(m_hAllocator->m_UseMutex) + { + if(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL) + { + if(!m_Mutex.TryLockWrite()) + { + pCtx->res = VK_ERROR_INITIALIZATION_FAILED; + return; + } + } + else + { + m_Mutex.LockWrite(); + pCtx->mutexLocked = true; + } + } + + pCtx->Begin(overlappingMoveSupported, flags); + + // Defragment. + + const VkDeviceSize maxBytesToMove = defragmentOnGpu ? maxGpuBytesToMove : maxCpuBytesToMove; + const uint32_t maxAllocationsToMove = defragmentOnGpu ? maxGpuAllocationsToMove : maxCpuAllocationsToMove; + pCtx->res = pCtx->GetAlgorithm()->Defragment(pCtx->defragmentationMoves, maxBytesToMove, maxAllocationsToMove, flags); + + // Accumulate statistics. + if(pStats != VMA_NULL) + { + const VkDeviceSize bytesMoved = pCtx->GetAlgorithm()->GetBytesMoved(); + const uint32_t allocationsMoved = pCtx->GetAlgorithm()->GetAllocationsMoved(); + pStats->bytesMoved += bytesMoved; + pStats->allocationsMoved += allocationsMoved; + VMA_ASSERT(bytesMoved <= maxBytesToMove); + VMA_ASSERT(allocationsMoved <= maxAllocationsToMove); + if(defragmentOnGpu) + { + maxGpuBytesToMove -= bytesMoved; + maxGpuAllocationsToMove -= allocationsMoved; + } + else + { + maxCpuBytesToMove -= bytesMoved; + maxCpuAllocationsToMove -= allocationsMoved; + } + } + + if(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL) + { + if(m_hAllocator->m_UseMutex) + m_Mutex.UnlockWrite(); + + if(pCtx->res >= VK_SUCCESS && !pCtx->defragmentationMoves.empty()) + pCtx->res = VK_NOT_READY; + + return; + } + + if(pCtx->res >= VK_SUCCESS) + { + if(defragmentOnGpu) + { + ApplyDefragmentationMovesGpu(pCtx, pCtx->defragmentationMoves, commandBuffer); + } + else + { + ApplyDefragmentationMovesCpu(pCtx, pCtx->defragmentationMoves); + } + } + } +} + +void VmaBlockVector::DefragmentationEnd( + class VmaBlockVectorDefragmentationContext* pCtx, + uint32_t flags, + VmaDefragmentationStats* pStats) +{ + if(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL && m_hAllocator->m_UseMutex) + { + VMA_ASSERT(pCtx->mutexLocked == false); + + // Incremental defragmentation doesn't hold the lock, so when we enter here we don't actually have any + // lock protecting us. Since we mutate state here, we have to take the lock out now + m_Mutex.LockWrite(); + pCtx->mutexLocked = true; + } + + // If the mutex isn't locked we didn't do any work and there is nothing to delete. + if(pCtx->mutexLocked || !m_hAllocator->m_UseMutex) + { + // Destroy buffers. + for(size_t blockIndex = pCtx->blockContexts.size(); blockIndex--;) + { + VmaBlockDefragmentationContext &blockCtx = pCtx->blockContexts[blockIndex]; + if(blockCtx.hBuffer) + { + (*m_hAllocator->GetVulkanFunctions().vkDestroyBuffer)(m_hAllocator->m_hDevice, blockCtx.hBuffer, m_hAllocator->GetAllocationCallbacks()); + } + } + + if(pCtx->res >= VK_SUCCESS) + { + FreeEmptyBlocks(pStats); + } + } + + if(pCtx->mutexLocked) + { + VMA_ASSERT(m_hAllocator->m_UseMutex); + m_Mutex.UnlockWrite(); + } +} + +uint32_t VmaBlockVector::ProcessDefragmentations( + class VmaBlockVectorDefragmentationContext *pCtx, + VmaDefragmentationPassMoveInfo* pMove, uint32_t maxMoves) +{ + VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); + + const uint32_t moveCount = VMA_MIN(uint32_t(pCtx->defragmentationMoves.size()) - pCtx->defragmentationMovesProcessed, maxMoves); + + for(uint32_t i = 0; i < moveCount; ++ i) + { + VmaDefragmentationMove& move = pCtx->defragmentationMoves[pCtx->defragmentationMovesProcessed + i]; + + pMove->allocation = move.hAllocation; + pMove->memory = move.pDstBlock->GetDeviceMemory(); + pMove->offset = move.dstOffset; + + ++ pMove; + } + + pCtx->defragmentationMovesProcessed += moveCount; + + return moveCount; +} + +void VmaBlockVector::CommitDefragmentations( + class VmaBlockVectorDefragmentationContext *pCtx, + VmaDefragmentationStats* pStats) +{ + VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); + + for(uint32_t i = pCtx->defragmentationMovesCommitted; i < pCtx->defragmentationMovesProcessed; ++ i) + { + const VmaDefragmentationMove &move = pCtx->defragmentationMoves[i]; + + move.pSrcBlock->m_pMetadata->FreeAtOffset(move.srcOffset); + move.hAllocation->ChangeBlockAllocation(m_hAllocator, move.pDstBlock, move.dstOffset); + } + + pCtx->defragmentationMovesCommitted = pCtx->defragmentationMovesProcessed; + FreeEmptyBlocks(pStats); +} + +size_t VmaBlockVector::CalcAllocationCount() const +{ + size_t result = 0; + for(size_t i = 0; i < m_Blocks.size(); ++i) + { + result += m_Blocks[i]->m_pMetadata->GetAllocationCount(); + } + return result; +} + +bool VmaBlockVector::IsBufferImageGranularityConflictPossible() const +{ + if(m_BufferImageGranularity == 1) + { + return false; + } + VmaSuballocationType lastSuballocType = VMA_SUBALLOCATION_TYPE_FREE; + for(size_t i = 0, count = m_Blocks.size(); i < count; ++i) + { + VmaDeviceMemoryBlock* const pBlock = m_Blocks[i]; + VMA_ASSERT(m_Algorithm == 0); + VmaBlockMetadata_Generic* const pMetadata = (VmaBlockMetadata_Generic*)pBlock->m_pMetadata; + if(pMetadata->IsBufferImageGranularityConflictPossible(m_BufferImageGranularity, lastSuballocType)) + { + return true; + } + } + return false; +} + +void VmaBlockVector::MakePoolAllocationsLost( + uint32_t currentFrameIndex, + size_t* pLostAllocationCount) +{ + VmaMutexLockWrite lock(m_Mutex, m_hAllocator->m_UseMutex); + size_t lostAllocationCount = 0; + for(uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) + { + VmaDeviceMemoryBlock* const pBlock = m_Blocks[blockIndex]; + VMA_ASSERT(pBlock); + lostAllocationCount += pBlock->m_pMetadata->MakeAllocationsLost(currentFrameIndex, m_FrameInUseCount); + } + if(pLostAllocationCount != VMA_NULL) + { + *pLostAllocationCount = lostAllocationCount; + } +} + +VkResult VmaBlockVector::CheckCorruption() +{ + if(!IsCorruptionDetectionEnabled()) + { + return VK_ERROR_FEATURE_NOT_PRESENT; + } + + VmaMutexLockRead lock(m_Mutex, m_hAllocator->m_UseMutex); + for(uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) + { + VmaDeviceMemoryBlock* const pBlock = m_Blocks[blockIndex]; + VMA_ASSERT(pBlock); + VkResult res = pBlock->CheckCorruption(m_hAllocator); + if(res != VK_SUCCESS) + { + return res; + } + } + return VK_SUCCESS; +} + +void VmaBlockVector::AddStats(VmaStats* pStats) +{ + const uint32_t memTypeIndex = m_MemoryTypeIndex; + const uint32_t memHeapIndex = m_hAllocator->MemoryTypeIndexToHeapIndex(memTypeIndex); + + VmaMutexLockRead lock(m_Mutex, m_hAllocator->m_UseMutex); + + for(uint32_t blockIndex = 0; blockIndex < m_Blocks.size(); ++blockIndex) + { + const VmaDeviceMemoryBlock* const pBlock = m_Blocks[blockIndex]; + VMA_ASSERT(pBlock); + VMA_HEAVY_ASSERT(pBlock->Validate()); + VmaStatInfo allocationStatInfo; + pBlock->m_pMetadata->CalcAllocationStatInfo(allocationStatInfo); + VmaAddStatInfo(pStats->total, allocationStatInfo); + VmaAddStatInfo(pStats->memoryType[memTypeIndex], allocationStatInfo); + VmaAddStatInfo(pStats->memoryHeap[memHeapIndex], allocationStatInfo); + } +} + +//////////////////////////////////////////////////////////////////////////////// +// VmaDefragmentationAlgorithm_Generic members definition + +VmaDefragmentationAlgorithm_Generic::VmaDefragmentationAlgorithm_Generic( + VmaAllocator hAllocator, + VmaBlockVector* pBlockVector, + uint32_t currentFrameIndex, + bool overlappingMoveSupported) : + VmaDefragmentationAlgorithm(hAllocator, pBlockVector, currentFrameIndex), + m_AllocationCount(0), + m_AllAllocations(false), + m_BytesMoved(0), + m_AllocationsMoved(0), + m_Blocks(VmaStlAllocator(hAllocator->GetAllocationCallbacks())) +{ + // Create block info for each block. + const size_t blockCount = m_pBlockVector->m_Blocks.size(); + for(size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + { + BlockInfo* pBlockInfo = vma_new(m_hAllocator, BlockInfo)(m_hAllocator->GetAllocationCallbacks()); + pBlockInfo->m_OriginalBlockIndex = blockIndex; + pBlockInfo->m_pBlock = m_pBlockVector->m_Blocks[blockIndex]; + m_Blocks.push_back(pBlockInfo); + } + + // Sort them by m_pBlock pointer value. + VMA_SORT(m_Blocks.begin(), m_Blocks.end(), BlockPointerLess()); +} + +VmaDefragmentationAlgorithm_Generic::~VmaDefragmentationAlgorithm_Generic() +{ + for(size_t i = m_Blocks.size(); i--; ) + { + vma_delete(m_hAllocator, m_Blocks[i]); + } +} + +void VmaDefragmentationAlgorithm_Generic::AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged) +{ + // Now as we are inside VmaBlockVector::m_Mutex, we can make final check if this allocation was not lost. + if(hAlloc->GetLastUseFrameIndex() != VMA_FRAME_INDEX_LOST) + { + VmaDeviceMemoryBlock* pBlock = hAlloc->GetBlock(); + BlockInfoVector::iterator it = VmaBinaryFindFirstNotLess(m_Blocks.begin(), m_Blocks.end(), pBlock, BlockPointerLess()); + if(it != m_Blocks.end() && (*it)->m_pBlock == pBlock) + { + AllocationInfo allocInfo = AllocationInfo(hAlloc, pChanged); + (*it)->m_Allocations.push_back(allocInfo); + } + else + { + VMA_ASSERT(0); + } + + ++m_AllocationCount; + } +} + +VkResult VmaDefragmentationAlgorithm_Generic::DefragmentRound( + VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + bool freeOldAllocations) +{ + if(m_Blocks.empty()) + { + return VK_SUCCESS; + } + + // This is a choice based on research. + // Option 1: + uint32_t strategy = VMA_ALLOCATION_CREATE_STRATEGY_MIN_TIME_BIT; + // Option 2: + //uint32_t strategy = VMA_ALLOCATION_CREATE_STRATEGY_MIN_MEMORY_BIT; + // Option 3: + //uint32_t strategy = VMA_ALLOCATION_CREATE_STRATEGY_MIN_FRAGMENTATION_BIT; + + size_t srcBlockMinIndex = 0; + // When FAST_ALGORITHM, move allocations from only last out of blocks that contain non-movable allocations. + /* + if(m_AlgorithmFlags & VMA_DEFRAGMENTATION_FAST_ALGORITHM_BIT) + { + const size_t blocksWithNonMovableCount = CalcBlocksWithNonMovableCount(); + if(blocksWithNonMovableCount > 0) + { + srcBlockMinIndex = blocksWithNonMovableCount - 1; + } + } + */ + + size_t srcBlockIndex = m_Blocks.size() - 1; + size_t srcAllocIndex = SIZE_MAX; + for(;;) + { + // 1. Find next allocation to move. + // 1.1. Start from last to first m_Blocks - they are sorted from most "destination" to most "source". + // 1.2. Then start from last to first m_Allocations. + while(srcAllocIndex >= m_Blocks[srcBlockIndex]->m_Allocations.size()) + { + if(m_Blocks[srcBlockIndex]->m_Allocations.empty()) + { + // Finished: no more allocations to process. + if(srcBlockIndex == srcBlockMinIndex) + { + return VK_SUCCESS; + } + else + { + --srcBlockIndex; + srcAllocIndex = SIZE_MAX; + } + } + else + { + srcAllocIndex = m_Blocks[srcBlockIndex]->m_Allocations.size() - 1; + } + } + + BlockInfo* pSrcBlockInfo = m_Blocks[srcBlockIndex]; + AllocationInfo& allocInfo = pSrcBlockInfo->m_Allocations[srcAllocIndex]; + + const VkDeviceSize size = allocInfo.m_hAllocation->GetSize(); + const VkDeviceSize srcOffset = allocInfo.m_hAllocation->GetOffset(); + const VkDeviceSize alignment = allocInfo.m_hAllocation->GetAlignment(); + const VmaSuballocationType suballocType = allocInfo.m_hAllocation->GetSuballocationType(); + + // 2. Try to find new place for this allocation in preceding or current block. + for(size_t dstBlockIndex = 0; dstBlockIndex <= srcBlockIndex; ++dstBlockIndex) + { + BlockInfo* pDstBlockInfo = m_Blocks[dstBlockIndex]; + VmaAllocationRequest dstAllocRequest; + if(pDstBlockInfo->m_pBlock->m_pMetadata->CreateAllocationRequest( + m_CurrentFrameIndex, + m_pBlockVector->GetFrameInUseCount(), + m_pBlockVector->GetBufferImageGranularity(), + size, + alignment, + false, // upperAddress + suballocType, + false, // canMakeOtherLost + strategy, + &dstAllocRequest) && + MoveMakesSense( + dstBlockIndex, dstAllocRequest.offset, srcBlockIndex, srcOffset)) + { + VMA_ASSERT(dstAllocRequest.itemsToMakeLostCount == 0); + + // Reached limit on number of allocations or bytes to move. + if((m_AllocationsMoved + 1 > maxAllocationsToMove) || + (m_BytesMoved + size > maxBytesToMove)) + { + return VK_SUCCESS; + } + + VmaDefragmentationMove move = {}; + move.srcBlockIndex = pSrcBlockInfo->m_OriginalBlockIndex; + move.dstBlockIndex = pDstBlockInfo->m_OriginalBlockIndex; + move.srcOffset = srcOffset; + move.dstOffset = dstAllocRequest.offset; + move.size = size; + move.hAllocation = allocInfo.m_hAllocation; + move.pSrcBlock = pSrcBlockInfo->m_pBlock; + move.pDstBlock = pDstBlockInfo->m_pBlock; + + moves.push_back(move); + + pDstBlockInfo->m_pBlock->m_pMetadata->Alloc( + dstAllocRequest, + suballocType, + size, + allocInfo.m_hAllocation); + + if(freeOldAllocations) + { + pSrcBlockInfo->m_pBlock->m_pMetadata->FreeAtOffset(srcOffset); + allocInfo.m_hAllocation->ChangeBlockAllocation(m_hAllocator, pDstBlockInfo->m_pBlock, dstAllocRequest.offset); + } + + if(allocInfo.m_pChanged != VMA_NULL) + { + *allocInfo.m_pChanged = VK_TRUE; + } + + ++m_AllocationsMoved; + m_BytesMoved += size; + + VmaVectorRemove(pSrcBlockInfo->m_Allocations, srcAllocIndex); + + break; + } + } + + // If not processed, this allocInfo remains in pBlockInfo->m_Allocations for next round. + + if(srcAllocIndex > 0) + { + --srcAllocIndex; + } + else + { + if(srcBlockIndex > 0) + { + --srcBlockIndex; + srcAllocIndex = SIZE_MAX; + } + else + { + return VK_SUCCESS; + } + } + } +} + +size_t VmaDefragmentationAlgorithm_Generic::CalcBlocksWithNonMovableCount() const +{ + size_t result = 0; + for(size_t i = 0; i < m_Blocks.size(); ++i) + { + if(m_Blocks[i]->m_HasNonMovableAllocations) + { + ++result; + } + } + return result; +} + +VkResult VmaDefragmentationAlgorithm_Generic::Defragment( + VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + VmaDefragmentationFlags flags) +{ + if(!m_AllAllocations && m_AllocationCount == 0) + { + return VK_SUCCESS; + } + + const size_t blockCount = m_Blocks.size(); + for(size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + { + BlockInfo* pBlockInfo = m_Blocks[blockIndex]; + + if(m_AllAllocations) + { + VmaBlockMetadata_Generic* pMetadata = (VmaBlockMetadata_Generic*)pBlockInfo->m_pBlock->m_pMetadata; + for(VmaSuballocationList::const_iterator it = pMetadata->m_Suballocations.begin(); + it != pMetadata->m_Suballocations.end(); + ++it) + { + if(it->type != VMA_SUBALLOCATION_TYPE_FREE) + { + AllocationInfo allocInfo = AllocationInfo(it->hAllocation, VMA_NULL); + pBlockInfo->m_Allocations.push_back(allocInfo); + } + } + } + + pBlockInfo->CalcHasNonMovableAllocations(); + + // This is a choice based on research. + // Option 1: + pBlockInfo->SortAllocationsByOffsetDescending(); + // Option 2: + //pBlockInfo->SortAllocationsBySizeDescending(); + } + + // Sort m_Blocks this time by the main criterium, from most "destination" to most "source" blocks. + VMA_SORT(m_Blocks.begin(), m_Blocks.end(), BlockInfoCompareMoveDestination()); + + // This is a choice based on research. + const uint32_t roundCount = 2; + + // Execute defragmentation rounds (the main part). + VkResult result = VK_SUCCESS; + for(uint32_t round = 0; (round < roundCount) && (result == VK_SUCCESS); ++round) + { + result = DefragmentRound(moves, maxBytesToMove, maxAllocationsToMove, !(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL)); + } + + return result; +} + +bool VmaDefragmentationAlgorithm_Generic::MoveMakesSense( + size_t dstBlockIndex, VkDeviceSize dstOffset, + size_t srcBlockIndex, VkDeviceSize srcOffset) +{ + if(dstBlockIndex < srcBlockIndex) + { + return true; + } + if(dstBlockIndex > srcBlockIndex) + { + return false; + } + if(dstOffset < srcOffset) + { + return true; + } + return false; +} + +//////////////////////////////////////////////////////////////////////////////// +// VmaDefragmentationAlgorithm_Fast + +VmaDefragmentationAlgorithm_Fast::VmaDefragmentationAlgorithm_Fast( + VmaAllocator hAllocator, + VmaBlockVector* pBlockVector, + uint32_t currentFrameIndex, + bool overlappingMoveSupported) : + VmaDefragmentationAlgorithm(hAllocator, pBlockVector, currentFrameIndex), + m_OverlappingMoveSupported(overlappingMoveSupported), + m_AllocationCount(0), + m_AllAllocations(false), + m_BytesMoved(0), + m_AllocationsMoved(0), + m_BlockInfos(VmaStlAllocator(hAllocator->GetAllocationCallbacks())) +{ + VMA_ASSERT(VMA_DEBUG_MARGIN == 0); + +} + +VmaDefragmentationAlgorithm_Fast::~VmaDefragmentationAlgorithm_Fast() +{ +} + +VkResult VmaDefragmentationAlgorithm_Fast::Defragment( + VmaVector< VmaDefragmentationMove, VmaStlAllocator >& moves, + VkDeviceSize maxBytesToMove, + uint32_t maxAllocationsToMove, + VmaDefragmentationFlags flags) +{ + VMA_ASSERT(m_AllAllocations || m_pBlockVector->CalcAllocationCount() == m_AllocationCount); + + const size_t blockCount = m_pBlockVector->GetBlockCount(); + if(blockCount == 0 || maxBytesToMove == 0 || maxAllocationsToMove == 0) + { + return VK_SUCCESS; + } + + PreprocessMetadata(); + + // Sort blocks in order from most destination. + + m_BlockInfos.resize(blockCount); + for(size_t i = 0; i < blockCount; ++i) + { + m_BlockInfos[i].origBlockIndex = i; + } + + VMA_SORT(m_BlockInfos.begin(), m_BlockInfos.end(), [this](const BlockInfo& lhs, const BlockInfo& rhs) -> bool { + return m_pBlockVector->GetBlock(lhs.origBlockIndex)->m_pMetadata->GetSumFreeSize() < + m_pBlockVector->GetBlock(rhs.origBlockIndex)->m_pMetadata->GetSumFreeSize(); + }); + + // THE MAIN ALGORITHM + + FreeSpaceDatabase freeSpaceDb; + + size_t dstBlockInfoIndex = 0; + size_t dstOrigBlockIndex = m_BlockInfos[dstBlockInfoIndex].origBlockIndex; + VmaDeviceMemoryBlock* pDstBlock = m_pBlockVector->GetBlock(dstOrigBlockIndex); + VmaBlockMetadata_Generic* pDstMetadata = (VmaBlockMetadata_Generic*)pDstBlock->m_pMetadata; + VkDeviceSize dstBlockSize = pDstMetadata->GetSize(); + VkDeviceSize dstOffset = 0; + + bool end = false; + for(size_t srcBlockInfoIndex = 0; !end && srcBlockInfoIndex < blockCount; ++srcBlockInfoIndex) + { + const size_t srcOrigBlockIndex = m_BlockInfos[srcBlockInfoIndex].origBlockIndex; + VmaDeviceMemoryBlock* const pSrcBlock = m_pBlockVector->GetBlock(srcOrigBlockIndex); + VmaBlockMetadata_Generic* const pSrcMetadata = (VmaBlockMetadata_Generic*)pSrcBlock->m_pMetadata; + for(VmaSuballocationList::iterator srcSuballocIt = pSrcMetadata->m_Suballocations.begin(); + !end && srcSuballocIt != pSrcMetadata->m_Suballocations.end(); ) + { + VmaAllocation_T* const pAlloc = srcSuballocIt->hAllocation; + const VkDeviceSize srcAllocAlignment = pAlloc->GetAlignment(); + const VkDeviceSize srcAllocSize = srcSuballocIt->size; + if(m_AllocationsMoved == maxAllocationsToMove || + m_BytesMoved + srcAllocSize > maxBytesToMove) + { + end = true; + break; + } + const VkDeviceSize srcAllocOffset = srcSuballocIt->offset; + + VmaDefragmentationMove move = {}; + // Try to place it in one of free spaces from the database. + size_t freeSpaceInfoIndex; + VkDeviceSize dstAllocOffset; + if(freeSpaceDb.Fetch(srcAllocAlignment, srcAllocSize, + freeSpaceInfoIndex, dstAllocOffset)) + { + size_t freeSpaceOrigBlockIndex = m_BlockInfos[freeSpaceInfoIndex].origBlockIndex; + VmaDeviceMemoryBlock* pFreeSpaceBlock = m_pBlockVector->GetBlock(freeSpaceOrigBlockIndex); + VmaBlockMetadata_Generic* pFreeSpaceMetadata = (VmaBlockMetadata_Generic*)pFreeSpaceBlock->m_pMetadata; + + // Same block + if(freeSpaceInfoIndex == srcBlockInfoIndex) + { + VMA_ASSERT(dstAllocOffset <= srcAllocOffset); + + // MOVE OPTION 1: Move the allocation inside the same block by decreasing offset. + + VmaSuballocation suballoc = *srcSuballocIt; + suballoc.offset = dstAllocOffset; + suballoc.hAllocation->ChangeOffset(dstAllocOffset); + m_BytesMoved += srcAllocSize; + ++m_AllocationsMoved; + + VmaSuballocationList::iterator nextSuballocIt = srcSuballocIt; + ++nextSuballocIt; + pSrcMetadata->m_Suballocations.erase(srcSuballocIt); + srcSuballocIt = nextSuballocIt; + + InsertSuballoc(pFreeSpaceMetadata, suballoc); + + move.srcBlockIndex = srcOrigBlockIndex; + move.dstBlockIndex = freeSpaceOrigBlockIndex; + move.srcOffset = srcAllocOffset; + move.dstOffset = dstAllocOffset; + move.size = srcAllocSize; + + moves.push_back(move); + } + // Different block + else + { + // MOVE OPTION 2: Move the allocation to a different block. + + VMA_ASSERT(freeSpaceInfoIndex < srcBlockInfoIndex); + + VmaSuballocation suballoc = *srcSuballocIt; + suballoc.offset = dstAllocOffset; + suballoc.hAllocation->ChangeBlockAllocation(m_hAllocator, pFreeSpaceBlock, dstAllocOffset); + m_BytesMoved += srcAllocSize; + ++m_AllocationsMoved; + + VmaSuballocationList::iterator nextSuballocIt = srcSuballocIt; + ++nextSuballocIt; + pSrcMetadata->m_Suballocations.erase(srcSuballocIt); + srcSuballocIt = nextSuballocIt; + + InsertSuballoc(pFreeSpaceMetadata, suballoc); + + move.srcBlockIndex = srcOrigBlockIndex; + move.dstBlockIndex = freeSpaceOrigBlockIndex; + move.srcOffset = srcAllocOffset; + move.dstOffset = dstAllocOffset; + move.size = srcAllocSize; + + moves.push_back(move); + } + } + else + { + dstAllocOffset = VmaAlignUp(dstOffset, srcAllocAlignment); + + // If the allocation doesn't fit before the end of dstBlock, forward to next block. + while(dstBlockInfoIndex < srcBlockInfoIndex && + dstAllocOffset + srcAllocSize > dstBlockSize) + { + // But before that, register remaining free space at the end of dst block. + freeSpaceDb.Register(dstBlockInfoIndex, dstOffset, dstBlockSize - dstOffset); + + ++dstBlockInfoIndex; + dstOrigBlockIndex = m_BlockInfos[dstBlockInfoIndex].origBlockIndex; + pDstBlock = m_pBlockVector->GetBlock(dstOrigBlockIndex); + pDstMetadata = (VmaBlockMetadata_Generic*)pDstBlock->m_pMetadata; + dstBlockSize = pDstMetadata->GetSize(); + dstOffset = 0; + dstAllocOffset = 0; + } + + // Same block + if(dstBlockInfoIndex == srcBlockInfoIndex) + { + VMA_ASSERT(dstAllocOffset <= srcAllocOffset); + + const bool overlap = dstAllocOffset + srcAllocSize > srcAllocOffset; + + bool skipOver = overlap; + if(overlap && m_OverlappingMoveSupported && dstAllocOffset < srcAllocOffset) + { + // If destination and source place overlap, skip if it would move it + // by only < 1/64 of its size. + skipOver = (srcAllocOffset - dstAllocOffset) * 64 < srcAllocSize; + } + + if(skipOver) + { + freeSpaceDb.Register(dstBlockInfoIndex, dstOffset, srcAllocOffset - dstOffset); + + dstOffset = srcAllocOffset + srcAllocSize; + ++srcSuballocIt; + } + // MOVE OPTION 1: Move the allocation inside the same block by decreasing offset. + else + { + srcSuballocIt->offset = dstAllocOffset; + srcSuballocIt->hAllocation->ChangeOffset(dstAllocOffset); + dstOffset = dstAllocOffset + srcAllocSize; + m_BytesMoved += srcAllocSize; + ++m_AllocationsMoved; + ++srcSuballocIt; + + move.srcBlockIndex = srcOrigBlockIndex; + move.dstBlockIndex = dstOrigBlockIndex; + move.srcOffset = srcAllocOffset; + move.dstOffset = dstAllocOffset; + move.size = srcAllocSize; + + moves.push_back(move); + } + } + // Different block + else + { + // MOVE OPTION 2: Move the allocation to a different block. + + VMA_ASSERT(dstBlockInfoIndex < srcBlockInfoIndex); + VMA_ASSERT(dstAllocOffset + srcAllocSize <= dstBlockSize); + + VmaSuballocation suballoc = *srcSuballocIt; + suballoc.offset = dstAllocOffset; + suballoc.hAllocation->ChangeBlockAllocation(m_hAllocator, pDstBlock, dstAllocOffset); + dstOffset = dstAllocOffset + srcAllocSize; + m_BytesMoved += srcAllocSize; + ++m_AllocationsMoved; + + VmaSuballocationList::iterator nextSuballocIt = srcSuballocIt; + ++nextSuballocIt; + pSrcMetadata->m_Suballocations.erase(srcSuballocIt); + srcSuballocIt = nextSuballocIt; + + pDstMetadata->m_Suballocations.push_back(suballoc); + + move.srcBlockIndex = srcOrigBlockIndex; + move.dstBlockIndex = dstOrigBlockIndex; + move.srcOffset = srcAllocOffset; + move.dstOffset = dstAllocOffset; + move.size = srcAllocSize; + + moves.push_back(move); + } + } + } + } + + m_BlockInfos.clear(); + + PostprocessMetadata(); + + return VK_SUCCESS; +} + +void VmaDefragmentationAlgorithm_Fast::PreprocessMetadata() +{ + const size_t blockCount = m_pBlockVector->GetBlockCount(); + for(size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + { + VmaBlockMetadata_Generic* const pMetadata = + (VmaBlockMetadata_Generic*)m_pBlockVector->GetBlock(blockIndex)->m_pMetadata; + pMetadata->m_FreeCount = 0; + pMetadata->m_SumFreeSize = pMetadata->GetSize(); + pMetadata->m_FreeSuballocationsBySize.clear(); + for(VmaSuballocationList::iterator it = pMetadata->m_Suballocations.begin(); + it != pMetadata->m_Suballocations.end(); ) + { + if(it->type == VMA_SUBALLOCATION_TYPE_FREE) + { + VmaSuballocationList::iterator nextIt = it; + ++nextIt; + pMetadata->m_Suballocations.erase(it); + it = nextIt; + } + else + { + ++it; + } + } + } +} + +void VmaDefragmentationAlgorithm_Fast::PostprocessMetadata() +{ + const size_t blockCount = m_pBlockVector->GetBlockCount(); + for(size_t blockIndex = 0; blockIndex < blockCount; ++blockIndex) + { + VmaBlockMetadata_Generic* const pMetadata = + (VmaBlockMetadata_Generic*)m_pBlockVector->GetBlock(blockIndex)->m_pMetadata; + const VkDeviceSize blockSize = pMetadata->GetSize(); + + // No allocations in this block - entire area is free. + if(pMetadata->m_Suballocations.empty()) + { + pMetadata->m_FreeCount = 1; + //pMetadata->m_SumFreeSize is already set to blockSize. + VmaSuballocation suballoc = { + 0, // offset + blockSize, // size + VMA_NULL, // hAllocation + VMA_SUBALLOCATION_TYPE_FREE }; + pMetadata->m_Suballocations.push_back(suballoc); + pMetadata->RegisterFreeSuballocation(pMetadata->m_Suballocations.begin()); + } + // There are some allocations in this block. + else + { + VkDeviceSize offset = 0; + VmaSuballocationList::iterator it; + for(it = pMetadata->m_Suballocations.begin(); + it != pMetadata->m_Suballocations.end(); + ++it) + { + VMA_ASSERT(it->type != VMA_SUBALLOCATION_TYPE_FREE); + VMA_ASSERT(it->offset >= offset); + + // Need to insert preceding free space. + if(it->offset > offset) + { + ++pMetadata->m_FreeCount; + const VkDeviceSize freeSize = it->offset - offset; + VmaSuballocation suballoc = { + offset, // offset + freeSize, // size + VMA_NULL, // hAllocation + VMA_SUBALLOCATION_TYPE_FREE }; + VmaSuballocationList::iterator precedingFreeIt = pMetadata->m_Suballocations.insert(it, suballoc); + if(freeSize >= VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) + { + pMetadata->m_FreeSuballocationsBySize.push_back(precedingFreeIt); + } + } + + pMetadata->m_SumFreeSize -= it->size; + offset = it->offset + it->size; + } + + // Need to insert trailing free space. + if(offset < blockSize) + { + ++pMetadata->m_FreeCount; + const VkDeviceSize freeSize = blockSize - offset; + VmaSuballocation suballoc = { + offset, // offset + freeSize, // size + VMA_NULL, // hAllocation + VMA_SUBALLOCATION_TYPE_FREE }; + VMA_ASSERT(it == pMetadata->m_Suballocations.end()); + VmaSuballocationList::iterator trailingFreeIt = pMetadata->m_Suballocations.insert(it, suballoc); + if(freeSize > VMA_MIN_FREE_SUBALLOCATION_SIZE_TO_REGISTER) + { + pMetadata->m_FreeSuballocationsBySize.push_back(trailingFreeIt); + } + } + + VMA_SORT( + pMetadata->m_FreeSuballocationsBySize.begin(), + pMetadata->m_FreeSuballocationsBySize.end(), + VmaSuballocationItemSizeLess()); + } + + VMA_HEAVY_ASSERT(pMetadata->Validate()); + } +} + +void VmaDefragmentationAlgorithm_Fast::InsertSuballoc(VmaBlockMetadata_Generic* pMetadata, const VmaSuballocation& suballoc) +{ + // TODO: Optimize somehow. Remember iterator instead of searching for it linearly. + VmaSuballocationList::iterator it = pMetadata->m_Suballocations.begin(); + while(it != pMetadata->m_Suballocations.end()) + { + if(it->offset < suballoc.offset) + { + ++it; + } + } + pMetadata->m_Suballocations.insert(it, suballoc); +} + +//////////////////////////////////////////////////////////////////////////////// +// VmaBlockVectorDefragmentationContext + +VmaBlockVectorDefragmentationContext::VmaBlockVectorDefragmentationContext( + VmaAllocator hAllocator, + VmaPool hCustomPool, + VmaBlockVector* pBlockVector, + uint32_t currFrameIndex) : + res(VK_SUCCESS), + mutexLocked(false), + blockContexts(VmaStlAllocator(hAllocator->GetAllocationCallbacks())), + defragmentationMoves(VmaStlAllocator(hAllocator->GetAllocationCallbacks())), + defragmentationMovesProcessed(0), + defragmentationMovesCommitted(0), + hasDefragmentationPlan(0), + m_hAllocator(hAllocator), + m_hCustomPool(hCustomPool), + m_pBlockVector(pBlockVector), + m_CurrFrameIndex(currFrameIndex), + m_pAlgorithm(VMA_NULL), + m_Allocations(VmaStlAllocator(hAllocator->GetAllocationCallbacks())), + m_AllAllocations(false) +{ +} + +VmaBlockVectorDefragmentationContext::~VmaBlockVectorDefragmentationContext() +{ + vma_delete(m_hAllocator, m_pAlgorithm); +} + +void VmaBlockVectorDefragmentationContext::AddAllocation(VmaAllocation hAlloc, VkBool32* pChanged) +{ + AllocInfo info = { hAlloc, pChanged }; + m_Allocations.push_back(info); +} + +void VmaBlockVectorDefragmentationContext::Begin(bool overlappingMoveSupported, VmaDefragmentationFlags flags) +{ + const bool allAllocations = m_AllAllocations || + m_Allocations.size() == m_pBlockVector->CalcAllocationCount(); + + /******************************** + HERE IS THE CHOICE OF DEFRAGMENTATION ALGORITHM. + ********************************/ + + /* + Fast algorithm is supported only when certain criteria are met: + - VMA_DEBUG_MARGIN is 0. + - All allocations in this block vector are moveable. + - There is no possibility of image/buffer granularity conflict. + - The defragmentation is not incremental + */ + if(VMA_DEBUG_MARGIN == 0 && + allAllocations && + !m_pBlockVector->IsBufferImageGranularityConflictPossible() && + !(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL)) + { + m_pAlgorithm = vma_new(m_hAllocator, VmaDefragmentationAlgorithm_Fast)( + m_hAllocator, m_pBlockVector, m_CurrFrameIndex, overlappingMoveSupported); + } + else + { + m_pAlgorithm = vma_new(m_hAllocator, VmaDefragmentationAlgorithm_Generic)( + m_hAllocator, m_pBlockVector, m_CurrFrameIndex, overlappingMoveSupported); + } + + if(allAllocations) + { + m_pAlgorithm->AddAll(); + } + else + { + for(size_t i = 0, count = m_Allocations.size(); i < count; ++i) + { + m_pAlgorithm->AddAllocation(m_Allocations[i].hAlloc, m_Allocations[i].pChanged); + } + } +} + +//////////////////////////////////////////////////////////////////////////////// +// VmaDefragmentationContext + +VmaDefragmentationContext_T::VmaDefragmentationContext_T( + VmaAllocator hAllocator, + uint32_t currFrameIndex, + uint32_t flags, + VmaDefragmentationStats* pStats) : + m_hAllocator(hAllocator), + m_CurrFrameIndex(currFrameIndex), + m_Flags(flags), + m_pStats(pStats), + m_CustomPoolContexts(VmaStlAllocator(hAllocator->GetAllocationCallbacks())) +{ + memset(m_DefaultPoolContexts, 0, sizeof(m_DefaultPoolContexts)); +} + +VmaDefragmentationContext_T::~VmaDefragmentationContext_T() +{ + for(size_t i = m_CustomPoolContexts.size(); i--; ) + { + VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_CustomPoolContexts[i]; + pBlockVectorCtx->GetBlockVector()->DefragmentationEnd(pBlockVectorCtx, m_Flags, m_pStats); + vma_delete(m_hAllocator, pBlockVectorCtx); + } + for(size_t i = m_hAllocator->m_MemProps.memoryTypeCount; i--; ) + { + VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_DefaultPoolContexts[i]; + if(pBlockVectorCtx) + { + pBlockVectorCtx->GetBlockVector()->DefragmentationEnd(pBlockVectorCtx, m_Flags, m_pStats); + vma_delete(m_hAllocator, pBlockVectorCtx); + } + } +} + +void VmaDefragmentationContext_T::AddPools(uint32_t poolCount, const VmaPool* pPools) +{ + for(uint32_t poolIndex = 0; poolIndex < poolCount; ++poolIndex) + { + VmaPool pool = pPools[poolIndex]; + VMA_ASSERT(pool); + // Pools with algorithm other than default are not defragmented. + if(pool->m_BlockVector.GetAlgorithm() == 0) + { + VmaBlockVectorDefragmentationContext* pBlockVectorDefragCtx = VMA_NULL; + + for(size_t i = m_CustomPoolContexts.size(); i--; ) + { + if(m_CustomPoolContexts[i]->GetCustomPool() == pool) + { + pBlockVectorDefragCtx = m_CustomPoolContexts[i]; + break; + } + } + + if(!pBlockVectorDefragCtx) + { + pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( + m_hAllocator, + pool, + &pool->m_BlockVector, + m_CurrFrameIndex); + m_CustomPoolContexts.push_back(pBlockVectorDefragCtx); + } + + pBlockVectorDefragCtx->AddAll(); + } + } +} + +void VmaDefragmentationContext_T::AddAllocations( + uint32_t allocationCount, + const VmaAllocation* pAllocations, + VkBool32* pAllocationsChanged) +{ + // Dispatch pAllocations among defragmentators. Create them when necessary. + for(uint32_t allocIndex = 0; allocIndex < allocationCount; ++allocIndex) + { + const VmaAllocation hAlloc = pAllocations[allocIndex]; + VMA_ASSERT(hAlloc); + // DedicatedAlloc cannot be defragmented. + if((hAlloc->GetType() == VmaAllocation_T::ALLOCATION_TYPE_BLOCK) && + // Lost allocation cannot be defragmented. + (hAlloc->GetLastUseFrameIndex() != VMA_FRAME_INDEX_LOST)) + { + VmaBlockVectorDefragmentationContext* pBlockVectorDefragCtx = VMA_NULL; + + const VmaPool hAllocPool = hAlloc->GetBlock()->GetParentPool(); + // This allocation belongs to custom pool. + if(hAllocPool != VK_NULL_HANDLE) + { + // Pools with algorithm other than default are not defragmented. + if(hAllocPool->m_BlockVector.GetAlgorithm() == 0) + { + for(size_t i = m_CustomPoolContexts.size(); i--; ) + { + if(m_CustomPoolContexts[i]->GetCustomPool() == hAllocPool) + { + pBlockVectorDefragCtx = m_CustomPoolContexts[i]; + break; + } + } + if(!pBlockVectorDefragCtx) + { + pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( + m_hAllocator, + hAllocPool, + &hAllocPool->m_BlockVector, + m_CurrFrameIndex); + m_CustomPoolContexts.push_back(pBlockVectorDefragCtx); + } + } + } + // This allocation belongs to default pool. + else + { + const uint32_t memTypeIndex = hAlloc->GetMemoryTypeIndex(); + pBlockVectorDefragCtx = m_DefaultPoolContexts[memTypeIndex]; + if(!pBlockVectorDefragCtx) + { + pBlockVectorDefragCtx = vma_new(m_hAllocator, VmaBlockVectorDefragmentationContext)( + m_hAllocator, + VMA_NULL, // hCustomPool + m_hAllocator->m_pBlockVectors[memTypeIndex], + m_CurrFrameIndex); + m_DefaultPoolContexts[memTypeIndex] = pBlockVectorDefragCtx; + } + } + + if(pBlockVectorDefragCtx) + { + VkBool32* const pChanged = (pAllocationsChanged != VMA_NULL) ? + &pAllocationsChanged[allocIndex] : VMA_NULL; + pBlockVectorDefragCtx->AddAllocation(hAlloc, pChanged); + } + } + } +} + +VkResult VmaDefragmentationContext_T::Defragment( + VkDeviceSize maxCpuBytesToMove, uint32_t maxCpuAllocationsToMove, + VkDeviceSize maxGpuBytesToMove, uint32_t maxGpuAllocationsToMove, + VkCommandBuffer commandBuffer, VmaDefragmentationStats* pStats, VmaDefragmentationFlags flags) +{ + if(pStats) + { + memset(pStats, 0, sizeof(VmaDefragmentationStats)); + } + + if(flags & VMA_DEFRAGMENTATION_FLAG_INCREMENTAL) + { + // For incremental defragmetnations, we just earmark how much we can move + // The real meat is in the defragmentation steps + m_MaxCpuBytesToMove = maxCpuBytesToMove; + m_MaxCpuAllocationsToMove = maxCpuAllocationsToMove; + + m_MaxGpuBytesToMove = maxGpuBytesToMove; + m_MaxGpuAllocationsToMove = maxGpuAllocationsToMove; + + if(m_MaxCpuBytesToMove == 0 && m_MaxCpuAllocationsToMove == 0 && + m_MaxGpuBytesToMove == 0 && m_MaxGpuAllocationsToMove == 0) + return VK_SUCCESS; + + return VK_NOT_READY; + } + + if(commandBuffer == VK_NULL_HANDLE) + { + maxGpuBytesToMove = 0; + maxGpuAllocationsToMove = 0; + } + + VkResult res = VK_SUCCESS; + + // Process default pools. + for(uint32_t memTypeIndex = 0; + memTypeIndex < m_hAllocator->GetMemoryTypeCount() && res >= VK_SUCCESS; + ++memTypeIndex) + { + VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_DefaultPoolContexts[memTypeIndex]; + if(pBlockVectorCtx) + { + VMA_ASSERT(pBlockVectorCtx->GetBlockVector()); + pBlockVectorCtx->GetBlockVector()->Defragment( + pBlockVectorCtx, + pStats, flags, + maxCpuBytesToMove, maxCpuAllocationsToMove, + maxGpuBytesToMove, maxGpuAllocationsToMove, + commandBuffer); + if(pBlockVectorCtx->res != VK_SUCCESS) + { + res = pBlockVectorCtx->res; + } + } + } + + // Process custom pools. + for(size_t customCtxIndex = 0, customCtxCount = m_CustomPoolContexts.size(); + customCtxIndex < customCtxCount && res >= VK_SUCCESS; + ++customCtxIndex) + { + VmaBlockVectorDefragmentationContext* pBlockVectorCtx = m_CustomPoolContexts[customCtxIndex]; + VMA_ASSERT(pBlockVectorCtx && pBlockVectorCtx->GetBlockVector()); + pBlockVectorCtx->GetBlockVector()->Defragment( + pBlockVectorCtx, + pStats, flags, + maxCpuBytesToMove, maxCpuAllocationsToMove, + maxGpuBytesToMove, maxGpuAllocationsToMove, + commandBuffer); + if(pBlockVectorCtx->res != VK_SUCCESS) + { + res = pBlockVectorCtx->res; + } + } + + return res; +} + +VkResult VmaDefragmentationContext_T::DefragmentPassBegin(VmaDefragmentationPassInfo* pInfo) +{ + VmaDefragmentationPassMoveInfo* pCurrentMove = pInfo->pMoves; + uint32_t movesLeft = pInfo->moveCount; + + // Process default pools. + for(uint32_t memTypeIndex = 0; + memTypeIndex < m_hAllocator->GetMemoryTypeCount(); + ++memTypeIndex) + { + VmaBlockVectorDefragmentationContext *pBlockVectorCtx = m_DefaultPoolContexts[memTypeIndex]; + if(pBlockVectorCtx) + { + VMA_ASSERT(pBlockVectorCtx->GetBlockVector()); + + if(!pBlockVectorCtx->hasDefragmentationPlan) + { + pBlockVectorCtx->GetBlockVector()->Defragment( + pBlockVectorCtx, + m_pStats, m_Flags, + m_MaxCpuBytesToMove, m_MaxCpuAllocationsToMove, + m_MaxGpuBytesToMove, m_MaxGpuAllocationsToMove, + VK_NULL_HANDLE); + + if(pBlockVectorCtx->res < VK_SUCCESS) + continue; + + pBlockVectorCtx->hasDefragmentationPlan = true; + } + + const uint32_t processed = pBlockVectorCtx->GetBlockVector()->ProcessDefragmentations( + pBlockVectorCtx, + pCurrentMove, movesLeft); + + movesLeft -= processed; + pCurrentMove += processed; + } + } + + // Process custom pools. + for(size_t customCtxIndex = 0, customCtxCount = m_CustomPoolContexts.size(); + customCtxIndex < customCtxCount; + ++customCtxIndex) + { + VmaBlockVectorDefragmentationContext *pBlockVectorCtx = m_CustomPoolContexts[customCtxIndex]; + VMA_ASSERT(pBlockVectorCtx && pBlockVectorCtx->GetBlockVector()); + + if(!pBlockVectorCtx->hasDefragmentationPlan) + { + pBlockVectorCtx->GetBlockVector()->Defragment( + pBlockVectorCtx, + m_pStats, m_Flags, + m_MaxCpuBytesToMove, m_MaxCpuAllocationsToMove, + m_MaxGpuBytesToMove, m_MaxGpuAllocationsToMove, + VK_NULL_HANDLE); + + if(pBlockVectorCtx->res < VK_SUCCESS) + continue; + + pBlockVectorCtx->hasDefragmentationPlan = true; + } + + const uint32_t processed = pBlockVectorCtx->GetBlockVector()->ProcessDefragmentations( + pBlockVectorCtx, + pCurrentMove, movesLeft); + + movesLeft -= processed; + pCurrentMove += processed; + } + + pInfo->moveCount = pInfo->moveCount - movesLeft; + + return VK_SUCCESS; +} +VkResult VmaDefragmentationContext_T::DefragmentPassEnd() +{ + VkResult res = VK_SUCCESS; + + // Process default pools. + for(uint32_t memTypeIndex = 0; + memTypeIndex < m_hAllocator->GetMemoryTypeCount(); + ++memTypeIndex) + { + VmaBlockVectorDefragmentationContext *pBlockVectorCtx = m_DefaultPoolContexts[memTypeIndex]; + if(pBlockVectorCtx) + { + VMA_ASSERT(pBlockVectorCtx->GetBlockVector()); + + if(!pBlockVectorCtx->hasDefragmentationPlan) + { + res = VK_NOT_READY; + continue; + } + + pBlockVectorCtx->GetBlockVector()->CommitDefragmentations( + pBlockVectorCtx, m_pStats); + + if(pBlockVectorCtx->defragmentationMoves.size() != pBlockVectorCtx->defragmentationMovesCommitted) + res = VK_NOT_READY; + } + } + + // Process custom pools. + for(size_t customCtxIndex = 0, customCtxCount = m_CustomPoolContexts.size(); + customCtxIndex < customCtxCount; + ++customCtxIndex) + { + VmaBlockVectorDefragmentationContext *pBlockVectorCtx = m_CustomPoolContexts[customCtxIndex]; + VMA_ASSERT(pBlockVectorCtx && pBlockVectorCtx->GetBlockVector()); + + if(!pBlockVectorCtx->hasDefragmentationPlan) + { + res = VK_NOT_READY; + continue; + } + + pBlockVectorCtx->GetBlockVector()->CommitDefragmentations( + pBlockVectorCtx, m_pStats); + + if(pBlockVectorCtx->defragmentationMoves.size() != pBlockVectorCtx->defragmentationMovesCommitted) + res = VK_NOT_READY; + } + + return res; +} + +//////////////////////////////////////////////////////////////////////////////// +// VmaRecorder + +#if VMA_RECORDING_ENABLED + +VmaRecorder::VmaRecorder() : + m_UseMutex(true), + m_Flags(0), + m_File(VMA_NULL), + m_RecordingStartTime(std::chrono::high_resolution_clock::now()) +{ +} + +VkResult VmaRecorder::Init(const VmaRecordSettings& settings, bool useMutex) +{ + m_UseMutex = useMutex; + m_Flags = settings.flags; + +#if defined(_WIN32) + // Open file for writing. + errno_t err = fopen_s(&m_File, settings.pFilePath, "wb"); + + if(err != 0) + { + return VK_ERROR_INITIALIZATION_FAILED; + } +#else + // Open file for writing. + m_File = fopen(settings.pFilePath, "wb"); + + if(m_File == 0) + { + return VK_ERROR_INITIALIZATION_FAILED; + } +#endif + + // Write header. + fprintf(m_File, "%s\n", "Vulkan Memory Allocator,Calls recording"); + fprintf(m_File, "%s\n", "1,8"); + + return VK_SUCCESS; +} + +VmaRecorder::~VmaRecorder() +{ + if(m_File != VMA_NULL) + { + fclose(m_File); + } +} + +void VmaRecorder::RecordCreateAllocator(uint32_t frameIndex) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaCreateAllocator\n", callParams.threadId, callParams.time, frameIndex); + Flush(); +} + +void VmaRecorder::RecordDestroyAllocator(uint32_t frameIndex) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaDestroyAllocator\n", callParams.threadId, callParams.time, frameIndex); + Flush(); +} + +void VmaRecorder::RecordCreatePool(uint32_t frameIndex, const VmaPoolCreateInfo& createInfo, VmaPool pool) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaCreatePool,%u,%u,%llu,%llu,%llu,%u,%p\n", callParams.threadId, callParams.time, frameIndex, + createInfo.memoryTypeIndex, + createInfo.flags, + createInfo.blockSize, + (uint64_t)createInfo.minBlockCount, + (uint64_t)createInfo.maxBlockCount, + createInfo.frameInUseCount, + pool); + Flush(); +} + +void VmaRecorder::RecordDestroyPool(uint32_t frameIndex, VmaPool pool) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaDestroyPool,%p\n", callParams.threadId, callParams.time, frameIndex, + pool); + Flush(); +} + +void VmaRecorder::RecordAllocateMemory(uint32_t frameIndex, + const VkMemoryRequirements& vkMemReq, + const VmaAllocationCreateInfo& createInfo, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + UserDataString userDataStr(createInfo.flags, createInfo.pUserData); + fprintf(m_File, "%u,%.3f,%u,vmaAllocateMemory,%llu,%llu,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, + vkMemReq.size, + vkMemReq.alignment, + vkMemReq.memoryTypeBits, + createInfo.flags, + createInfo.usage, + createInfo.requiredFlags, + createInfo.preferredFlags, + createInfo.memoryTypeBits, + createInfo.pool, + allocation, + userDataStr.GetString()); + Flush(); +} + +void VmaRecorder::RecordAllocateMemoryPages(uint32_t frameIndex, + const VkMemoryRequirements& vkMemReq, + const VmaAllocationCreateInfo& createInfo, + uint64_t allocationCount, + const VmaAllocation* pAllocations) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + UserDataString userDataStr(createInfo.flags, createInfo.pUserData); + fprintf(m_File, "%u,%.3f,%u,vmaAllocateMemoryPages,%llu,%llu,%u,%u,%u,%u,%u,%u,%p,", callParams.threadId, callParams.time, frameIndex, + vkMemReq.size, + vkMemReq.alignment, + vkMemReq.memoryTypeBits, + createInfo.flags, + createInfo.usage, + createInfo.requiredFlags, + createInfo.preferredFlags, + createInfo.memoryTypeBits, + createInfo.pool); + PrintPointerList(allocationCount, pAllocations); + fprintf(m_File, ",%s\n", userDataStr.GetString()); + Flush(); +} + +void VmaRecorder::RecordAllocateMemoryForBuffer(uint32_t frameIndex, + const VkMemoryRequirements& vkMemReq, + bool requiresDedicatedAllocation, + bool prefersDedicatedAllocation, + const VmaAllocationCreateInfo& createInfo, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + UserDataString userDataStr(createInfo.flags, createInfo.pUserData); + fprintf(m_File, "%u,%.3f,%u,vmaAllocateMemoryForBuffer,%llu,%llu,%u,%u,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, + vkMemReq.size, + vkMemReq.alignment, + vkMemReq.memoryTypeBits, + requiresDedicatedAllocation ? 1 : 0, + prefersDedicatedAllocation ? 1 : 0, + createInfo.flags, + createInfo.usage, + createInfo.requiredFlags, + createInfo.preferredFlags, + createInfo.memoryTypeBits, + createInfo.pool, + allocation, + userDataStr.GetString()); + Flush(); +} + +void VmaRecorder::RecordAllocateMemoryForImage(uint32_t frameIndex, + const VkMemoryRequirements& vkMemReq, + bool requiresDedicatedAllocation, + bool prefersDedicatedAllocation, + const VmaAllocationCreateInfo& createInfo, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + UserDataString userDataStr(createInfo.flags, createInfo.pUserData); + fprintf(m_File, "%u,%.3f,%u,vmaAllocateMemoryForImage,%llu,%llu,%u,%u,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, + vkMemReq.size, + vkMemReq.alignment, + vkMemReq.memoryTypeBits, + requiresDedicatedAllocation ? 1 : 0, + prefersDedicatedAllocation ? 1 : 0, + createInfo.flags, + createInfo.usage, + createInfo.requiredFlags, + createInfo.preferredFlags, + createInfo.memoryTypeBits, + createInfo.pool, + allocation, + userDataStr.GetString()); + Flush(); +} + +void VmaRecorder::RecordFreeMemory(uint32_t frameIndex, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaFreeMemory,%p\n", callParams.threadId, callParams.time, frameIndex, + allocation); + Flush(); +} + +void VmaRecorder::RecordFreeMemoryPages(uint32_t frameIndex, + uint64_t allocationCount, + const VmaAllocation* pAllocations) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaFreeMemoryPages,", callParams.threadId, callParams.time, frameIndex); + PrintPointerList(allocationCount, pAllocations); + fprintf(m_File, "\n"); + Flush(); +} + +void VmaRecorder::RecordSetAllocationUserData(uint32_t frameIndex, + VmaAllocation allocation, + const void* pUserData) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + UserDataString userDataStr( + allocation->IsUserDataString() ? VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT : 0, + pUserData); + fprintf(m_File, "%u,%.3f,%u,vmaSetAllocationUserData,%p,%s\n", callParams.threadId, callParams.time, frameIndex, + allocation, + userDataStr.GetString()); + Flush(); +} + +void VmaRecorder::RecordCreateLostAllocation(uint32_t frameIndex, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaCreateLostAllocation,%p\n", callParams.threadId, callParams.time, frameIndex, + allocation); + Flush(); +} + +void VmaRecorder::RecordMapMemory(uint32_t frameIndex, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaMapMemory,%p\n", callParams.threadId, callParams.time, frameIndex, + allocation); + Flush(); +} + +void VmaRecorder::RecordUnmapMemory(uint32_t frameIndex, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaUnmapMemory,%p\n", callParams.threadId, callParams.time, frameIndex, + allocation); + Flush(); +} + +void VmaRecorder::RecordFlushAllocation(uint32_t frameIndex, + VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaFlushAllocation,%p,%llu,%llu\n", callParams.threadId, callParams.time, frameIndex, + allocation, + offset, + size); + Flush(); +} + +void VmaRecorder::RecordInvalidateAllocation(uint32_t frameIndex, + VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaInvalidateAllocation,%p,%llu,%llu\n", callParams.threadId, callParams.time, frameIndex, + allocation, + offset, + size); + Flush(); +} + +void VmaRecorder::RecordCreateBuffer(uint32_t frameIndex, + const VkBufferCreateInfo& bufCreateInfo, + const VmaAllocationCreateInfo& allocCreateInfo, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + UserDataString userDataStr(allocCreateInfo.flags, allocCreateInfo.pUserData); + fprintf(m_File, "%u,%.3f,%u,vmaCreateBuffer,%u,%llu,%u,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, + bufCreateInfo.flags, + bufCreateInfo.size, + bufCreateInfo.usage, + bufCreateInfo.sharingMode, + allocCreateInfo.flags, + allocCreateInfo.usage, + allocCreateInfo.requiredFlags, + allocCreateInfo.preferredFlags, + allocCreateInfo.memoryTypeBits, + allocCreateInfo.pool, + allocation, + userDataStr.GetString()); + Flush(); +} + +void VmaRecorder::RecordCreateImage(uint32_t frameIndex, + const VkImageCreateInfo& imageCreateInfo, + const VmaAllocationCreateInfo& allocCreateInfo, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + UserDataString userDataStr(allocCreateInfo.flags, allocCreateInfo.pUserData); + fprintf(m_File, "%u,%.3f,%u,vmaCreateImage,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%u,%p,%p,%s\n", callParams.threadId, callParams.time, frameIndex, + imageCreateInfo.flags, + imageCreateInfo.imageType, + imageCreateInfo.format, + imageCreateInfo.extent.width, + imageCreateInfo.extent.height, + imageCreateInfo.extent.depth, + imageCreateInfo.mipLevels, + imageCreateInfo.arrayLayers, + imageCreateInfo.samples, + imageCreateInfo.tiling, + imageCreateInfo.usage, + imageCreateInfo.sharingMode, + imageCreateInfo.initialLayout, + allocCreateInfo.flags, + allocCreateInfo.usage, + allocCreateInfo.requiredFlags, + allocCreateInfo.preferredFlags, + allocCreateInfo.memoryTypeBits, + allocCreateInfo.pool, + allocation, + userDataStr.GetString()); + Flush(); +} + +void VmaRecorder::RecordDestroyBuffer(uint32_t frameIndex, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaDestroyBuffer,%p\n", callParams.threadId, callParams.time, frameIndex, + allocation); + Flush(); +} + +void VmaRecorder::RecordDestroyImage(uint32_t frameIndex, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaDestroyImage,%p\n", callParams.threadId, callParams.time, frameIndex, + allocation); + Flush(); +} + +void VmaRecorder::RecordTouchAllocation(uint32_t frameIndex, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaTouchAllocation,%p\n", callParams.threadId, callParams.time, frameIndex, + allocation); + Flush(); +} + +void VmaRecorder::RecordGetAllocationInfo(uint32_t frameIndex, + VmaAllocation allocation) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaGetAllocationInfo,%p\n", callParams.threadId, callParams.time, frameIndex, + allocation); + Flush(); +} + +void VmaRecorder::RecordMakePoolAllocationsLost(uint32_t frameIndex, + VmaPool pool) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaMakePoolAllocationsLost,%p\n", callParams.threadId, callParams.time, frameIndex, + pool); + Flush(); +} + +void VmaRecorder::RecordDefragmentationBegin(uint32_t frameIndex, + const VmaDefragmentationInfo2& info, + VmaDefragmentationContext ctx) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaDefragmentationBegin,%u,", callParams.threadId, callParams.time, frameIndex, + info.flags); + PrintPointerList(info.allocationCount, info.pAllocations); + fprintf(m_File, ","); + PrintPointerList(info.poolCount, info.pPools); + fprintf(m_File, ",%llu,%u,%llu,%u,%p,%p\n", + info.maxCpuBytesToMove, + info.maxCpuAllocationsToMove, + info.maxGpuBytesToMove, + info.maxGpuAllocationsToMove, + info.commandBuffer, + ctx); + Flush(); +} + +void VmaRecorder::RecordDefragmentationEnd(uint32_t frameIndex, + VmaDefragmentationContext ctx) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaDefragmentationEnd,%p\n", callParams.threadId, callParams.time, frameIndex, + ctx); + Flush(); +} + +void VmaRecorder::RecordSetPoolName(uint32_t frameIndex, + VmaPool pool, + const char* name) +{ + CallParams callParams; + GetBasicParams(callParams); + + VmaMutexLock lock(m_FileMutex, m_UseMutex); + fprintf(m_File, "%u,%.3f,%u,vmaSetPoolName,%p,%s\n", callParams.threadId, callParams.time, frameIndex, + pool, name != VMA_NULL ? name : ""); + Flush(); +} + +VmaRecorder::UserDataString::UserDataString(VmaAllocationCreateFlags allocFlags, const void* pUserData) +{ + if(pUserData != VMA_NULL) + { + if((allocFlags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0) + { + m_Str = (const char*)pUserData; + } + else + { + // If VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT is not specified, convert the string's memory address to a string and store it. + snprintf(m_PtrStr, 17, "%p", pUserData); + m_Str = m_PtrStr; + } + } + else + { + m_Str = ""; + } +} + +void VmaRecorder::WriteConfiguration( + const VkPhysicalDeviceProperties& devProps, + const VkPhysicalDeviceMemoryProperties& memProps, + uint32_t vulkanApiVersion, + bool dedicatedAllocationExtensionEnabled, + bool bindMemory2ExtensionEnabled, + bool memoryBudgetExtensionEnabled, + bool deviceCoherentMemoryExtensionEnabled) +{ + fprintf(m_File, "Config,Begin\n"); + + fprintf(m_File, "VulkanApiVersion,%u,%u\n", VK_VERSION_MAJOR(vulkanApiVersion), VK_VERSION_MINOR(vulkanApiVersion)); + + fprintf(m_File, "PhysicalDevice,apiVersion,%u\n", devProps.apiVersion); + fprintf(m_File, "PhysicalDevice,driverVersion,%u\n", devProps.driverVersion); + fprintf(m_File, "PhysicalDevice,vendorID,%u\n", devProps.vendorID); + fprintf(m_File, "PhysicalDevice,deviceID,%u\n", devProps.deviceID); + fprintf(m_File, "PhysicalDevice,deviceType,%u\n", devProps.deviceType); + fprintf(m_File, "PhysicalDevice,deviceName,%s\n", devProps.deviceName); + + fprintf(m_File, "PhysicalDeviceLimits,maxMemoryAllocationCount,%u\n", devProps.limits.maxMemoryAllocationCount); + fprintf(m_File, "PhysicalDeviceLimits,bufferImageGranularity,%llu\n", devProps.limits.bufferImageGranularity); + fprintf(m_File, "PhysicalDeviceLimits,nonCoherentAtomSize,%llu\n", devProps.limits.nonCoherentAtomSize); + + fprintf(m_File, "PhysicalDeviceMemory,HeapCount,%u\n", memProps.memoryHeapCount); + for(uint32_t i = 0; i < memProps.memoryHeapCount; ++i) + { + fprintf(m_File, "PhysicalDeviceMemory,Heap,%u,size,%llu\n", i, memProps.memoryHeaps[i].size); + fprintf(m_File, "PhysicalDeviceMemory,Heap,%u,flags,%u\n", i, memProps.memoryHeaps[i].flags); + } + fprintf(m_File, "PhysicalDeviceMemory,TypeCount,%u\n", memProps.memoryTypeCount); + for(uint32_t i = 0; i < memProps.memoryTypeCount; ++i) + { + fprintf(m_File, "PhysicalDeviceMemory,Type,%u,heapIndex,%u\n", i, memProps.memoryTypes[i].heapIndex); + fprintf(m_File, "PhysicalDeviceMemory,Type,%u,propertyFlags,%u\n", i, memProps.memoryTypes[i].propertyFlags); + } + + fprintf(m_File, "Extension,VK_KHR_dedicated_allocation,%u\n", dedicatedAllocationExtensionEnabled ? 1 : 0); + fprintf(m_File, "Extension,VK_KHR_bind_memory2,%u\n", bindMemory2ExtensionEnabled ? 1 : 0); + fprintf(m_File, "Extension,VK_EXT_memory_budget,%u\n", memoryBudgetExtensionEnabled ? 1 : 0); + fprintf(m_File, "Extension,VK_AMD_device_coherent_memory,%u\n", deviceCoherentMemoryExtensionEnabled ? 1 : 0); + + fprintf(m_File, "Macro,VMA_DEBUG_ALWAYS_DEDICATED_MEMORY,%u\n", VMA_DEBUG_ALWAYS_DEDICATED_MEMORY ? 1 : 0); + fprintf(m_File, "Macro,VMA_DEBUG_ALIGNMENT,%llu\n", (VkDeviceSize)VMA_DEBUG_ALIGNMENT); + fprintf(m_File, "Macro,VMA_DEBUG_MARGIN,%llu\n", (VkDeviceSize)VMA_DEBUG_MARGIN); + fprintf(m_File, "Macro,VMA_DEBUG_INITIALIZE_ALLOCATIONS,%u\n", VMA_DEBUG_INITIALIZE_ALLOCATIONS ? 1 : 0); + fprintf(m_File, "Macro,VMA_DEBUG_DETECT_CORRUPTION,%u\n", VMA_DEBUG_DETECT_CORRUPTION ? 1 : 0); + fprintf(m_File, "Macro,VMA_DEBUG_GLOBAL_MUTEX,%u\n", VMA_DEBUG_GLOBAL_MUTEX ? 1 : 0); + fprintf(m_File, "Macro,VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY,%llu\n", (VkDeviceSize)VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY); + fprintf(m_File, "Macro,VMA_SMALL_HEAP_MAX_SIZE,%llu\n", (VkDeviceSize)VMA_SMALL_HEAP_MAX_SIZE); + fprintf(m_File, "Macro,VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE,%llu\n", (VkDeviceSize)VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE); + + fprintf(m_File, "Config,End\n"); +} + +void VmaRecorder::GetBasicParams(CallParams& outParams) +{ + #if defined(_WIN32) + outParams.threadId = GetCurrentThreadId(); + #else + // Use C++11 features to get thread id and convert it to uint32_t. + // There is room for optimization since sstream is quite slow. + // Is there a better way to convert std::this_thread::get_id() to uint32_t? + std::thread::id thread_id = std::this_thread::get_id(); + std::stringstream thread_id_to_string_converter; + thread_id_to_string_converter << thread_id; + std::string thread_id_as_string = thread_id_to_string_converter.str(); + outParams.threadId = static_cast(std::stoi(thread_id_as_string.c_str())); + #endif + + auto current_time = std::chrono::high_resolution_clock::now(); + + outParams.time = std::chrono::duration(current_time - m_RecordingStartTime).count(); +} + +void VmaRecorder::PrintPointerList(uint64_t count, const VmaAllocation* pItems) +{ + if(count) + { + fprintf(m_File, "%p", pItems[0]); + for(uint64_t i = 1; i < count; ++i) + { + fprintf(m_File, " %p", pItems[i]); + } + } +} + +void VmaRecorder::Flush() +{ + if((m_Flags & VMA_RECORD_FLUSH_AFTER_CALL_BIT) != 0) + { + fflush(m_File); + } +} + +#endif // #if VMA_RECORDING_ENABLED + +//////////////////////////////////////////////////////////////////////////////// +// VmaAllocationObjectAllocator + +VmaAllocationObjectAllocator::VmaAllocationObjectAllocator(const VkAllocationCallbacks* pAllocationCallbacks) : + m_Allocator(pAllocationCallbacks, 1024) +{ +} + +template VmaAllocation VmaAllocationObjectAllocator::Allocate(Types... args) +{ + VmaMutexLock mutexLock(m_Mutex); + return m_Allocator.Alloc(std::forward(args)...); +} + +void VmaAllocationObjectAllocator::Free(VmaAllocation hAlloc) +{ + VmaMutexLock mutexLock(m_Mutex); + m_Allocator.Free(hAlloc); +} + +//////////////////////////////////////////////////////////////////////////////// +// VmaAllocator_T + +VmaAllocator_T::VmaAllocator_T(const VmaAllocatorCreateInfo* pCreateInfo) : + m_UseMutex((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_EXTERNALLY_SYNCHRONIZED_BIT) == 0), + m_VulkanApiVersion(pCreateInfo->vulkanApiVersion != 0 ? pCreateInfo->vulkanApiVersion : VK_API_VERSION_1_0), + m_UseKhrDedicatedAllocation((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT) != 0), + m_UseKhrBindMemory2((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_KHR_BIND_MEMORY2_BIT) != 0), + m_UseExtMemoryBudget((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_EXT_MEMORY_BUDGET_BIT) != 0), + m_UseAmdDeviceCoherentMemory((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_AMD_DEVICE_COHERENT_MEMORY_BIT) != 0), + m_UseKhrBufferDeviceAddress((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT) != 0), + m_hDevice(pCreateInfo->device), + m_hInstance(pCreateInfo->instance), + m_AllocationCallbacksSpecified(pCreateInfo->pAllocationCallbacks != VMA_NULL), + m_AllocationCallbacks(pCreateInfo->pAllocationCallbacks ? + *pCreateInfo->pAllocationCallbacks : VmaEmptyAllocationCallbacks), + m_AllocationObjectAllocator(&m_AllocationCallbacks), + m_HeapSizeLimitMask(0), + m_PreferredLargeHeapBlockSize(0), + m_PhysicalDevice(pCreateInfo->physicalDevice), + m_CurrentFrameIndex(0), + m_GpuDefragmentationMemoryTypeBits(UINT32_MAX), + m_Pools(VmaStlAllocator(GetAllocationCallbacks())), + m_NextPoolId(0), + m_GlobalMemoryTypeBits(UINT32_MAX) +#if VMA_RECORDING_ENABLED + ,m_pRecorder(VMA_NULL) +#endif +{ + if(m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) + { + m_UseKhrDedicatedAllocation = false; + m_UseKhrBindMemory2 = false; + } + + if(VMA_DEBUG_DETECT_CORRUPTION) + { + // Needs to be multiply of uint32_t size because we are going to write VMA_CORRUPTION_DETECTION_MAGIC_VALUE to it. + VMA_ASSERT(VMA_DEBUG_MARGIN % sizeof(uint32_t) == 0); + } + + VMA_ASSERT(pCreateInfo->physicalDevice && pCreateInfo->device && pCreateInfo->instance); + + if(m_VulkanApiVersion < VK_MAKE_VERSION(1, 1, 0)) + { +#if !(VMA_DEDICATED_ALLOCATION) + if((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT) != 0) + { + VMA_ASSERT(0 && "VMA_ALLOCATOR_CREATE_KHR_DEDICATED_ALLOCATION_BIT set but required extensions are disabled by preprocessor macros."); + } +#endif +#if !(VMA_BIND_MEMORY2) + if((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_KHR_BIND_MEMORY2_BIT) != 0) + { + VMA_ASSERT(0 && "VMA_ALLOCATOR_CREATE_KHR_BIND_MEMORY2_BIT set but required extension is disabled by preprocessor macros."); + } +#endif + } +#if !(VMA_MEMORY_BUDGET) + if((pCreateInfo->flags & VMA_ALLOCATOR_CREATE_EXT_MEMORY_BUDGET_BIT) != 0) + { + VMA_ASSERT(0 && "VMA_ALLOCATOR_CREATE_EXT_MEMORY_BUDGET_BIT set but required extension is disabled by preprocessor macros."); + } +#endif +#if !(VMA_BUFFER_DEVICE_ADDRESS) + if(m_UseKhrBufferDeviceAddress) + { + VMA_ASSERT(0 && "VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT is set but required extension or Vulkan 1.2 is not available in your Vulkan header or its support in VMA has been disabled by a preprocessor macro."); + } +#endif +#if VMA_VULKAN_VERSION < 1002000 + if(m_VulkanApiVersion >= VK_MAKE_VERSION(1, 2, 0)) + { + VMA_ASSERT(0 && "vulkanApiVersion >= VK_API_VERSION_1_2 but required Vulkan version is disabled by preprocessor macros."); + } +#endif +#if VMA_VULKAN_VERSION < 1001000 + if(m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) + { + VMA_ASSERT(0 && "vulkanApiVersion >= VK_API_VERSION_1_1 but required Vulkan version is disabled by preprocessor macros."); + } +#endif + + memset(&m_DeviceMemoryCallbacks, 0 ,sizeof(m_DeviceMemoryCallbacks)); + memset(&m_PhysicalDeviceProperties, 0, sizeof(m_PhysicalDeviceProperties)); + memset(&m_MemProps, 0, sizeof(m_MemProps)); + + memset(&m_pBlockVectors, 0, sizeof(m_pBlockVectors)); + memset(&m_pDedicatedAllocations, 0, sizeof(m_pDedicatedAllocations)); + memset(&m_VulkanFunctions, 0, sizeof(m_VulkanFunctions)); + + if(pCreateInfo->pDeviceMemoryCallbacks != VMA_NULL) + { + m_DeviceMemoryCallbacks.pUserData = pCreateInfo->pDeviceMemoryCallbacks->pUserData; + m_DeviceMemoryCallbacks.pfnAllocate = pCreateInfo->pDeviceMemoryCallbacks->pfnAllocate; + m_DeviceMemoryCallbacks.pfnFree = pCreateInfo->pDeviceMemoryCallbacks->pfnFree; + } + + ImportVulkanFunctions(pCreateInfo->pVulkanFunctions); + + (*m_VulkanFunctions.vkGetPhysicalDeviceProperties)(m_PhysicalDevice, &m_PhysicalDeviceProperties); + (*m_VulkanFunctions.vkGetPhysicalDeviceMemoryProperties)(m_PhysicalDevice, &m_MemProps); + + VMA_ASSERT(VmaIsPow2(VMA_DEBUG_ALIGNMENT)); + VMA_ASSERT(VmaIsPow2(VMA_DEBUG_MIN_BUFFER_IMAGE_GRANULARITY)); + VMA_ASSERT(VmaIsPow2(m_PhysicalDeviceProperties.limits.bufferImageGranularity)); + VMA_ASSERT(VmaIsPow2(m_PhysicalDeviceProperties.limits.nonCoherentAtomSize)); + + m_PreferredLargeHeapBlockSize = (pCreateInfo->preferredLargeHeapBlockSize != 0) ? + pCreateInfo->preferredLargeHeapBlockSize : static_cast(VMA_DEFAULT_LARGE_HEAP_BLOCK_SIZE); + + m_GlobalMemoryTypeBits = CalculateGlobalMemoryTypeBits(); + + if(pCreateInfo->pHeapSizeLimit != VMA_NULL) + { + for(uint32_t heapIndex = 0; heapIndex < GetMemoryHeapCount(); ++heapIndex) + { + const VkDeviceSize limit = pCreateInfo->pHeapSizeLimit[heapIndex]; + if(limit != VK_WHOLE_SIZE) + { + m_HeapSizeLimitMask |= 1u << heapIndex; + if(limit < m_MemProps.memoryHeaps[heapIndex].size) + { + m_MemProps.memoryHeaps[heapIndex].size = limit; + } + } + } + } + + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + const VkDeviceSize preferredBlockSize = CalcPreferredBlockSize(memTypeIndex); + + m_pBlockVectors[memTypeIndex] = vma_new(this, VmaBlockVector)( + this, + VK_NULL_HANDLE, // hParentPool + memTypeIndex, + preferredBlockSize, + 0, + SIZE_MAX, + GetBufferImageGranularity(), + pCreateInfo->frameInUseCount, + false, // explicitBlockSize + false); // linearAlgorithm + // No need to call m_pBlockVectors[memTypeIndex][blockVectorTypeIndex]->CreateMinBlocks here, + // becase minBlockCount is 0. + m_pDedicatedAllocations[memTypeIndex] = vma_new(this, AllocationVectorType)(VmaStlAllocator(GetAllocationCallbacks())); + + } +} + +VkResult VmaAllocator_T::Init(const VmaAllocatorCreateInfo* pCreateInfo) +{ + VkResult res = VK_SUCCESS; + + if(pCreateInfo->pRecordSettings != VMA_NULL && + !VmaStrIsEmpty(pCreateInfo->pRecordSettings->pFilePath)) + { +#if VMA_RECORDING_ENABLED + m_pRecorder = vma_new(this, VmaRecorder)(); + res = m_pRecorder->Init(*pCreateInfo->pRecordSettings, m_UseMutex); + if(res != VK_SUCCESS) + { + return res; + } + m_pRecorder->WriteConfiguration( + m_PhysicalDeviceProperties, + m_MemProps, + m_VulkanApiVersion, + m_UseKhrDedicatedAllocation, + m_UseKhrBindMemory2, + m_UseExtMemoryBudget, + m_UseAmdDeviceCoherentMemory); + m_pRecorder->RecordCreateAllocator(GetCurrentFrameIndex()); +#else + VMA_ASSERT(0 && "VmaAllocatorCreateInfo::pRecordSettings used, but not supported due to VMA_RECORDING_ENABLED not defined to 1."); + return VK_ERROR_FEATURE_NOT_PRESENT; +#endif + } + +#if VMA_MEMORY_BUDGET + if(m_UseExtMemoryBudget) + { + UpdateVulkanBudget(); + } +#endif // #if VMA_MEMORY_BUDGET + + return res; +} + +VmaAllocator_T::~VmaAllocator_T() +{ +#if VMA_RECORDING_ENABLED + if(m_pRecorder != VMA_NULL) + { + m_pRecorder->RecordDestroyAllocator(GetCurrentFrameIndex()); + vma_delete(this, m_pRecorder); + } +#endif + + VMA_ASSERT(m_Pools.empty()); + + for(size_t i = GetMemoryTypeCount(); i--; ) + { + if(m_pDedicatedAllocations[i] != VMA_NULL && !m_pDedicatedAllocations[i]->empty()) + { + VMA_ASSERT(0 && "Unfreed dedicated allocations found."); + } + + vma_delete(this, m_pDedicatedAllocations[i]); + vma_delete(this, m_pBlockVectors[i]); + } +} + +void VmaAllocator_T::ImportVulkanFunctions(const VmaVulkanFunctions* pVulkanFunctions) +{ +#if VMA_STATIC_VULKAN_FUNCTIONS == 1 + ImportVulkanFunctions_Static(); +#endif + + if(pVulkanFunctions != VMA_NULL) + { + ImportVulkanFunctions_Custom(pVulkanFunctions); + } + +#if VMA_DYNAMIC_VULKAN_FUNCTIONS == 1 + ImportVulkanFunctions_Dynamic(); +#endif + + ValidateVulkanFunctions(); +} + +#if VMA_STATIC_VULKAN_FUNCTIONS == 1 + +void VmaAllocator_T::ImportVulkanFunctions_Static() +{ + // Vulkan 1.0 + m_VulkanFunctions.vkGetPhysicalDeviceProperties = (PFN_vkGetPhysicalDeviceProperties)vkGetPhysicalDeviceProperties; + m_VulkanFunctions.vkGetPhysicalDeviceMemoryProperties = (PFN_vkGetPhysicalDeviceMemoryProperties)vkGetPhysicalDeviceMemoryProperties; + m_VulkanFunctions.vkAllocateMemory = (PFN_vkAllocateMemory)vkAllocateMemory; + m_VulkanFunctions.vkFreeMemory = (PFN_vkFreeMemory)vkFreeMemory; + m_VulkanFunctions.vkMapMemory = (PFN_vkMapMemory)vkMapMemory; + m_VulkanFunctions.vkUnmapMemory = (PFN_vkUnmapMemory)vkUnmapMemory; + m_VulkanFunctions.vkFlushMappedMemoryRanges = (PFN_vkFlushMappedMemoryRanges)vkFlushMappedMemoryRanges; + m_VulkanFunctions.vkInvalidateMappedMemoryRanges = (PFN_vkInvalidateMappedMemoryRanges)vkInvalidateMappedMemoryRanges; + m_VulkanFunctions.vkBindBufferMemory = (PFN_vkBindBufferMemory)vkBindBufferMemory; + m_VulkanFunctions.vkBindImageMemory = (PFN_vkBindImageMemory)vkBindImageMemory; + m_VulkanFunctions.vkGetBufferMemoryRequirements = (PFN_vkGetBufferMemoryRequirements)vkGetBufferMemoryRequirements; + m_VulkanFunctions.vkGetImageMemoryRequirements = (PFN_vkGetImageMemoryRequirements)vkGetImageMemoryRequirements; + m_VulkanFunctions.vkCreateBuffer = (PFN_vkCreateBuffer)vkCreateBuffer; + m_VulkanFunctions.vkDestroyBuffer = (PFN_vkDestroyBuffer)vkDestroyBuffer; + m_VulkanFunctions.vkCreateImage = (PFN_vkCreateImage)vkCreateImage; + m_VulkanFunctions.vkDestroyImage = (PFN_vkDestroyImage)vkDestroyImage; + m_VulkanFunctions.vkCmdCopyBuffer = (PFN_vkCmdCopyBuffer)vkCmdCopyBuffer; + + // Vulkan 1.1 +#if VMA_VULKAN_VERSION >= 1001000 + if(m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) + { + m_VulkanFunctions.vkGetBufferMemoryRequirements2KHR = (PFN_vkGetBufferMemoryRequirements2)vkGetBufferMemoryRequirements2; + m_VulkanFunctions.vkGetImageMemoryRequirements2KHR = (PFN_vkGetImageMemoryRequirements2)vkGetImageMemoryRequirements2; + m_VulkanFunctions.vkBindBufferMemory2KHR = (PFN_vkBindBufferMemory2)vkBindBufferMemory2; + m_VulkanFunctions.vkBindImageMemory2KHR = (PFN_vkBindImageMemory2)vkBindImageMemory2; + m_VulkanFunctions.vkGetPhysicalDeviceMemoryProperties2KHR = (PFN_vkGetPhysicalDeviceMemoryProperties2)vkGetPhysicalDeviceMemoryProperties2; + } +#endif +} + +#endif // #if VMA_STATIC_VULKAN_FUNCTIONS == 1 + +void VmaAllocator_T::ImportVulkanFunctions_Custom(const VmaVulkanFunctions* pVulkanFunctions) +{ + VMA_ASSERT(pVulkanFunctions != VMA_NULL); + +#define VMA_COPY_IF_NOT_NULL(funcName) \ + if(pVulkanFunctions->funcName != VMA_NULL) m_VulkanFunctions.funcName = pVulkanFunctions->funcName; + + VMA_COPY_IF_NOT_NULL(vkGetPhysicalDeviceProperties); + VMA_COPY_IF_NOT_NULL(vkGetPhysicalDeviceMemoryProperties); + VMA_COPY_IF_NOT_NULL(vkAllocateMemory); + VMA_COPY_IF_NOT_NULL(vkFreeMemory); + VMA_COPY_IF_NOT_NULL(vkMapMemory); + VMA_COPY_IF_NOT_NULL(vkUnmapMemory); + VMA_COPY_IF_NOT_NULL(vkFlushMappedMemoryRanges); + VMA_COPY_IF_NOT_NULL(vkInvalidateMappedMemoryRanges); + VMA_COPY_IF_NOT_NULL(vkBindBufferMemory); + VMA_COPY_IF_NOT_NULL(vkBindImageMemory); + VMA_COPY_IF_NOT_NULL(vkGetBufferMemoryRequirements); + VMA_COPY_IF_NOT_NULL(vkGetImageMemoryRequirements); + VMA_COPY_IF_NOT_NULL(vkCreateBuffer); + VMA_COPY_IF_NOT_NULL(vkDestroyBuffer); + VMA_COPY_IF_NOT_NULL(vkCreateImage); + VMA_COPY_IF_NOT_NULL(vkDestroyImage); + VMA_COPY_IF_NOT_NULL(vkCmdCopyBuffer); + +#if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + VMA_COPY_IF_NOT_NULL(vkGetBufferMemoryRequirements2KHR); + VMA_COPY_IF_NOT_NULL(vkGetImageMemoryRequirements2KHR); +#endif + +#if VMA_BIND_MEMORY2 || VMA_VULKAN_VERSION >= 1001000 + VMA_COPY_IF_NOT_NULL(vkBindBufferMemory2KHR); + VMA_COPY_IF_NOT_NULL(vkBindImageMemory2KHR); +#endif + +#if VMA_MEMORY_BUDGET + VMA_COPY_IF_NOT_NULL(vkGetPhysicalDeviceMemoryProperties2KHR); +#endif + +#undef VMA_COPY_IF_NOT_NULL +} + +#if VMA_DYNAMIC_VULKAN_FUNCTIONS == 1 + +void VmaAllocator_T::ImportVulkanFunctions_Dynamic() +{ +#define VMA_FETCH_INSTANCE_FUNC(memberName, functionPointerType, functionNameString) \ + if(m_VulkanFunctions.memberName == VMA_NULL) \ + m_VulkanFunctions.memberName = \ + (functionPointerType)vkGetInstanceProcAddr(m_hInstance, functionNameString); +#define VMA_FETCH_DEVICE_FUNC(memberName, functionPointerType, functionNameString) \ + if(m_VulkanFunctions.memberName == VMA_NULL) \ + m_VulkanFunctions.memberName = \ + (functionPointerType)vkGetDeviceProcAddr(m_hDevice, functionNameString); + + VMA_FETCH_INSTANCE_FUNC(vkGetPhysicalDeviceProperties, PFN_vkGetPhysicalDeviceProperties, "vkGetPhysicalDeviceProperties"); + VMA_FETCH_INSTANCE_FUNC(vkGetPhysicalDeviceMemoryProperties, PFN_vkGetPhysicalDeviceMemoryProperties, "vkGetPhysicalDeviceMemoryProperties"); + VMA_FETCH_DEVICE_FUNC(vkAllocateMemory, PFN_vkAllocateMemory, "vkAllocateMemory"); + VMA_FETCH_DEVICE_FUNC(vkFreeMemory, PFN_vkFreeMemory, "vkFreeMemory"); + VMA_FETCH_DEVICE_FUNC(vkMapMemory, PFN_vkMapMemory, "vkMapMemory"); + VMA_FETCH_DEVICE_FUNC(vkUnmapMemory, PFN_vkUnmapMemory, "vkUnmapMemory"); + VMA_FETCH_DEVICE_FUNC(vkFlushMappedMemoryRanges, PFN_vkFlushMappedMemoryRanges, "vkFlushMappedMemoryRanges"); + VMA_FETCH_DEVICE_FUNC(vkInvalidateMappedMemoryRanges, PFN_vkInvalidateMappedMemoryRanges, "vkInvalidateMappedMemoryRanges"); + VMA_FETCH_DEVICE_FUNC(vkBindBufferMemory, PFN_vkBindBufferMemory, "vkBindBufferMemory"); + VMA_FETCH_DEVICE_FUNC(vkBindImageMemory, PFN_vkBindImageMemory, "vkBindImageMemory"); + VMA_FETCH_DEVICE_FUNC(vkGetBufferMemoryRequirements, PFN_vkGetBufferMemoryRequirements, "vkGetBufferMemoryRequirements"); + VMA_FETCH_DEVICE_FUNC(vkGetImageMemoryRequirements, PFN_vkGetImageMemoryRequirements, "vkGetImageMemoryRequirements"); + VMA_FETCH_DEVICE_FUNC(vkCreateBuffer, PFN_vkCreateBuffer, "vkCreateBuffer"); + VMA_FETCH_DEVICE_FUNC(vkDestroyBuffer, PFN_vkDestroyBuffer, "vkDestroyBuffer"); + VMA_FETCH_DEVICE_FUNC(vkCreateImage, PFN_vkCreateImage, "vkCreateImage"); + VMA_FETCH_DEVICE_FUNC(vkDestroyImage, PFN_vkDestroyImage, "vkDestroyImage"); + VMA_FETCH_DEVICE_FUNC(vkCmdCopyBuffer, PFN_vkCmdCopyBuffer, "vkCmdCopyBuffer"); + +#if VMA_VULKAN_VERSION >= 1001000 + if(m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) + { + VMA_FETCH_DEVICE_FUNC(vkGetBufferMemoryRequirements2KHR, PFN_vkGetBufferMemoryRequirements2, "vkGetBufferMemoryRequirements2"); + VMA_FETCH_DEVICE_FUNC(vkGetImageMemoryRequirements2KHR, PFN_vkGetImageMemoryRequirements2, "vkGetImageMemoryRequirements2"); + VMA_FETCH_DEVICE_FUNC(vkBindBufferMemory2KHR, PFN_vkBindBufferMemory2, "vkBindBufferMemory2"); + VMA_FETCH_DEVICE_FUNC(vkBindImageMemory2KHR, PFN_vkBindImageMemory2, "vkBindImageMemory2"); + VMA_FETCH_INSTANCE_FUNC(vkGetPhysicalDeviceMemoryProperties2KHR, PFN_vkGetPhysicalDeviceMemoryProperties2, "vkGetPhysicalDeviceMemoryProperties2"); + } +#endif + +#if VMA_DEDICATED_ALLOCATION + if(m_UseKhrDedicatedAllocation) + { + VMA_FETCH_DEVICE_FUNC(vkGetBufferMemoryRequirements2KHR, PFN_vkGetBufferMemoryRequirements2KHR, "vkGetBufferMemoryRequirements2KHR"); + VMA_FETCH_DEVICE_FUNC(vkGetImageMemoryRequirements2KHR, PFN_vkGetImageMemoryRequirements2KHR, "vkGetImageMemoryRequirements2KHR"); + } +#endif + +#if VMA_BIND_MEMORY2 + if(m_UseKhrBindMemory2) + { + VMA_FETCH_DEVICE_FUNC(vkBindBufferMemory2KHR, PFN_vkBindBufferMemory2KHR, "vkBindBufferMemory2KHR"); + VMA_FETCH_DEVICE_FUNC(vkBindImageMemory2KHR, PFN_vkBindImageMemory2KHR, "vkBindImageMemory2KHR"); + } +#endif // #if VMA_BIND_MEMORY2 + +#if VMA_MEMORY_BUDGET + if(m_UseExtMemoryBudget) + { + VMA_FETCH_INSTANCE_FUNC(vkGetPhysicalDeviceMemoryProperties2KHR, PFN_vkGetPhysicalDeviceMemoryProperties2KHR, "vkGetPhysicalDeviceMemoryProperties2KHR"); + } +#endif // #if VMA_MEMORY_BUDGET + +#undef VMA_FETCH_DEVICE_FUNC +#undef VMA_FETCH_INSTANCE_FUNC +} + +#endif // #if VMA_DYNAMIC_VULKAN_FUNCTIONS == 1 + +void VmaAllocator_T::ValidateVulkanFunctions() +{ + VMA_ASSERT(m_VulkanFunctions.vkGetPhysicalDeviceProperties != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkGetPhysicalDeviceMemoryProperties != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkAllocateMemory != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkFreeMemory != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkMapMemory != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkUnmapMemory != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkFlushMappedMemoryRanges != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkInvalidateMappedMemoryRanges != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkBindBufferMemory != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkBindImageMemory != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkGetBufferMemoryRequirements != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkGetImageMemoryRequirements != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkCreateBuffer != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkDestroyBuffer != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkCreateImage != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkDestroyImage != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkCmdCopyBuffer != VMA_NULL); + +#if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + if(m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0) || m_UseKhrDedicatedAllocation) + { + VMA_ASSERT(m_VulkanFunctions.vkGetBufferMemoryRequirements2KHR != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkGetImageMemoryRequirements2KHR != VMA_NULL); + } +#endif + +#if VMA_BIND_MEMORY2 || VMA_VULKAN_VERSION >= 1001000 + if(m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0) || m_UseKhrBindMemory2) + { + VMA_ASSERT(m_VulkanFunctions.vkBindBufferMemory2KHR != VMA_NULL); + VMA_ASSERT(m_VulkanFunctions.vkBindImageMemory2KHR != VMA_NULL); + } +#endif + +#if VMA_MEMORY_BUDGET || VMA_VULKAN_VERSION >= 1001000 + if(m_UseExtMemoryBudget || m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) + { + VMA_ASSERT(m_VulkanFunctions.vkGetPhysicalDeviceMemoryProperties2KHR != VMA_NULL); + } +#endif +} + +VkDeviceSize VmaAllocator_T::CalcPreferredBlockSize(uint32_t memTypeIndex) +{ + const uint32_t heapIndex = MemoryTypeIndexToHeapIndex(memTypeIndex); + const VkDeviceSize heapSize = m_MemProps.memoryHeaps[heapIndex].size; + const bool isSmallHeap = heapSize <= VMA_SMALL_HEAP_MAX_SIZE; + return VmaAlignUp(isSmallHeap ? (heapSize / 8) : m_PreferredLargeHeapBlockSize, (VkDeviceSize)32); +} + +VkResult VmaAllocator_T::AllocateMemoryOfType( + VkDeviceSize size, + VkDeviceSize alignment, + bool dedicatedAllocation, + VkBuffer dedicatedBuffer, + VkBufferUsageFlags dedicatedBufferUsage, + VkImage dedicatedImage, + const VmaAllocationCreateInfo& createInfo, + uint32_t memTypeIndex, + VmaSuballocationType suballocType, + size_t allocationCount, + VmaAllocation* pAllocations) +{ + VMA_ASSERT(pAllocations != VMA_NULL); + VMA_DEBUG_LOG(" AllocateMemory: MemoryTypeIndex=%u, AllocationCount=%zu, Size=%llu", memTypeIndex, allocationCount, size); + + VmaAllocationCreateInfo finalCreateInfo = createInfo; + + // If memory type is not HOST_VISIBLE, disable MAPPED. + if((finalCreateInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0 && + (m_MemProps.memoryTypes[memTypeIndex].propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) == 0) + { + finalCreateInfo.flags &= ~VMA_ALLOCATION_CREATE_MAPPED_BIT; + } + // If memory is lazily allocated, it should be always dedicated. + if(finalCreateInfo.usage == VMA_MEMORY_USAGE_GPU_LAZILY_ALLOCATED) + { + finalCreateInfo.flags |= VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT; + } + + VmaBlockVector* const blockVector = m_pBlockVectors[memTypeIndex]; + VMA_ASSERT(blockVector); + + const VkDeviceSize preferredBlockSize = blockVector->GetPreferredBlockSize(); + bool preferDedicatedMemory = + VMA_DEBUG_ALWAYS_DEDICATED_MEMORY || + dedicatedAllocation || + // Heuristics: Allocate dedicated memory if requested size if greater than half of preferred block size. + size > preferredBlockSize / 2; + + if(preferDedicatedMemory && + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) == 0 && + finalCreateInfo.pool == VK_NULL_HANDLE) + { + finalCreateInfo.flags |= VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT; + } + + if((finalCreateInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0) + { + if((finalCreateInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) + { + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + else + { + return AllocateDedicatedMemory( + size, + suballocType, + memTypeIndex, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT) != 0, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0, + finalCreateInfo.pUserData, + dedicatedBuffer, + dedicatedBufferUsage, + dedicatedImage, + allocationCount, + pAllocations); + } + } + else + { + VkResult res = blockVector->Allocate( + m_CurrentFrameIndex.load(), + size, + alignment, + finalCreateInfo, + suballocType, + allocationCount, + pAllocations); + if(res == VK_SUCCESS) + { + return res; + } + + // 5. Try dedicated memory. + if((finalCreateInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) + { + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + else + { + res = AllocateDedicatedMemory( + size, + suballocType, + memTypeIndex, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT) != 0, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0, + (finalCreateInfo.flags & VMA_ALLOCATION_CREATE_USER_DATA_COPY_STRING_BIT) != 0, + finalCreateInfo.pUserData, + dedicatedBuffer, + dedicatedBufferUsage, + dedicatedImage, + allocationCount, + pAllocations); + if(res == VK_SUCCESS) + { + // Succeeded: AllocateDedicatedMemory function already filld pMemory, nothing more to do here. + VMA_DEBUG_LOG(" Allocated as DedicatedMemory"); + return VK_SUCCESS; + } + else + { + // Everything failed: Return error code. + VMA_DEBUG_LOG(" vkAllocateMemory FAILED"); + return res; + } + } + } +} + +VkResult VmaAllocator_T::AllocateDedicatedMemory( + VkDeviceSize size, + VmaSuballocationType suballocType, + uint32_t memTypeIndex, + bool withinBudget, + bool map, + bool isUserDataString, + void* pUserData, + VkBuffer dedicatedBuffer, + VkBufferUsageFlags dedicatedBufferUsage, + VkImage dedicatedImage, + size_t allocationCount, + VmaAllocation* pAllocations) +{ + VMA_ASSERT(allocationCount > 0 && pAllocations); + + if(withinBudget) + { + const uint32_t heapIndex = MemoryTypeIndexToHeapIndex(memTypeIndex); + VmaBudget heapBudget = {}; + GetBudget(&heapBudget, heapIndex, 1); + if(heapBudget.usage + size * allocationCount > heapBudget.budget) + { + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + } + + VkMemoryAllocateInfo allocInfo = { VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO }; + allocInfo.memoryTypeIndex = memTypeIndex; + allocInfo.allocationSize = size; + +#if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + VkMemoryDedicatedAllocateInfoKHR dedicatedAllocInfo = { VK_STRUCTURE_TYPE_MEMORY_DEDICATED_ALLOCATE_INFO_KHR }; + if(m_UseKhrDedicatedAllocation || m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) + { + if(dedicatedBuffer != VK_NULL_HANDLE) + { + VMA_ASSERT(dedicatedImage == VK_NULL_HANDLE); + dedicatedAllocInfo.buffer = dedicatedBuffer; + VmaPnextChainPushFront(&allocInfo, &dedicatedAllocInfo); + } + else if(dedicatedImage != VK_NULL_HANDLE) + { + dedicatedAllocInfo.image = dedicatedImage; + VmaPnextChainPushFront(&allocInfo, &dedicatedAllocInfo); + } + } +#endif // #if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + +#if VMA_BUFFER_DEVICE_ADDRESS + VkMemoryAllocateFlagsInfoKHR allocFlagsInfo = { VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_FLAGS_INFO_KHR }; + if(m_UseKhrBufferDeviceAddress) + { + bool canContainBufferWithDeviceAddress = true; + if(dedicatedBuffer != VK_NULL_HANDLE) + { + canContainBufferWithDeviceAddress = dedicatedBufferUsage == UINT32_MAX || // Usage flags unknown + (dedicatedBufferUsage & VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT_EXT) != 0; + } + else if(dedicatedImage != VK_NULL_HANDLE) + { + canContainBufferWithDeviceAddress = false; + } + if(canContainBufferWithDeviceAddress) + { + allocFlagsInfo.flags = VK_MEMORY_ALLOCATE_DEVICE_ADDRESS_BIT_KHR; + VmaPnextChainPushFront(&allocInfo, &allocFlagsInfo); + } + } +#endif // #if VMA_BUFFER_DEVICE_ADDRESS + + size_t allocIndex; + VkResult res = VK_SUCCESS; + for(allocIndex = 0; allocIndex < allocationCount; ++allocIndex) + { + res = AllocateDedicatedMemoryPage( + size, + suballocType, + memTypeIndex, + allocInfo, + map, + isUserDataString, + pUserData, + pAllocations + allocIndex); + if(res != VK_SUCCESS) + { + break; + } + } + + if(res == VK_SUCCESS) + { + // Register them in m_pDedicatedAllocations. + { + VmaMutexLockWrite lock(m_DedicatedAllocationsMutex[memTypeIndex], m_UseMutex); + AllocationVectorType* pDedicatedAllocations = m_pDedicatedAllocations[memTypeIndex]; + VMA_ASSERT(pDedicatedAllocations); + for(allocIndex = 0; allocIndex < allocationCount; ++allocIndex) + { + VmaVectorInsertSorted(*pDedicatedAllocations, pAllocations[allocIndex]); + } + } + + VMA_DEBUG_LOG(" Allocated DedicatedMemory Count=%zu, MemoryTypeIndex=#%u", allocationCount, memTypeIndex); + } + else + { + // Free all already created allocations. + while(allocIndex--) + { + VmaAllocation currAlloc = pAllocations[allocIndex]; + VkDeviceMemory hMemory = currAlloc->GetMemory(); + + /* + There is no need to call this, because Vulkan spec allows to skip vkUnmapMemory + before vkFreeMemory. + + if(currAlloc->GetMappedData() != VMA_NULL) + { + (*m_VulkanFunctions.vkUnmapMemory)(m_hDevice, hMemory); + } + */ + + FreeVulkanMemory(memTypeIndex, currAlloc->GetSize(), hMemory); + m_Budget.RemoveAllocation(MemoryTypeIndexToHeapIndex(memTypeIndex), currAlloc->GetSize()); + currAlloc->SetUserData(this, VMA_NULL); + m_AllocationObjectAllocator.Free(currAlloc); + } + + memset(pAllocations, 0, sizeof(VmaAllocation) * allocationCount); + } + + return res; +} + +VkResult VmaAllocator_T::AllocateDedicatedMemoryPage( + VkDeviceSize size, + VmaSuballocationType suballocType, + uint32_t memTypeIndex, + const VkMemoryAllocateInfo& allocInfo, + bool map, + bool isUserDataString, + void* pUserData, + VmaAllocation* pAllocation) +{ + VkDeviceMemory hMemory = VK_NULL_HANDLE; + VkResult res = AllocateVulkanMemory(&allocInfo, &hMemory); + if(res < 0) + { + VMA_DEBUG_LOG(" vkAllocateMemory FAILED"); + return res; + } + + void* pMappedData = VMA_NULL; + if(map) + { + res = (*m_VulkanFunctions.vkMapMemory)( + m_hDevice, + hMemory, + 0, + VK_WHOLE_SIZE, + 0, + &pMappedData); + if(res < 0) + { + VMA_DEBUG_LOG(" vkMapMemory FAILED"); + FreeVulkanMemory(memTypeIndex, size, hMemory); + return res; + } + } + + *pAllocation = m_AllocationObjectAllocator.Allocate(m_CurrentFrameIndex.load(), isUserDataString); + (*pAllocation)->InitDedicatedAllocation(memTypeIndex, hMemory, suballocType, pMappedData, size); + (*pAllocation)->SetUserData(this, pUserData); + m_Budget.AddAllocation(MemoryTypeIndexToHeapIndex(memTypeIndex), size); + if(VMA_DEBUG_INITIALIZE_ALLOCATIONS) + { + FillAllocation(*pAllocation, VMA_ALLOCATION_FILL_PATTERN_CREATED); + } + + return VK_SUCCESS; +} + +void VmaAllocator_T::GetBufferMemoryRequirements( + VkBuffer hBuffer, + VkMemoryRequirements& memReq, + bool& requiresDedicatedAllocation, + bool& prefersDedicatedAllocation) const +{ +#if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + if(m_UseKhrDedicatedAllocation || m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) + { + VkBufferMemoryRequirementsInfo2KHR memReqInfo = { VK_STRUCTURE_TYPE_BUFFER_MEMORY_REQUIREMENTS_INFO_2_KHR }; + memReqInfo.buffer = hBuffer; + + VkMemoryDedicatedRequirementsKHR memDedicatedReq = { VK_STRUCTURE_TYPE_MEMORY_DEDICATED_REQUIREMENTS_KHR }; + + VkMemoryRequirements2KHR memReq2 = { VK_STRUCTURE_TYPE_MEMORY_REQUIREMENTS_2_KHR }; + VmaPnextChainPushFront(&memReq2, &memDedicatedReq); + + (*m_VulkanFunctions.vkGetBufferMemoryRequirements2KHR)(m_hDevice, &memReqInfo, &memReq2); + + memReq = memReq2.memoryRequirements; + requiresDedicatedAllocation = (memDedicatedReq.requiresDedicatedAllocation != VK_FALSE); + prefersDedicatedAllocation = (memDedicatedReq.prefersDedicatedAllocation != VK_FALSE); + } + else +#endif // #if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + { + (*m_VulkanFunctions.vkGetBufferMemoryRequirements)(m_hDevice, hBuffer, &memReq); + requiresDedicatedAllocation = false; + prefersDedicatedAllocation = false; + } +} + +void VmaAllocator_T::GetImageMemoryRequirements( + VkImage hImage, + VkMemoryRequirements& memReq, + bool& requiresDedicatedAllocation, + bool& prefersDedicatedAllocation) const +{ +#if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + if(m_UseKhrDedicatedAllocation || m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) + { + VkImageMemoryRequirementsInfo2KHR memReqInfo = { VK_STRUCTURE_TYPE_IMAGE_MEMORY_REQUIREMENTS_INFO_2_KHR }; + memReqInfo.image = hImage; + + VkMemoryDedicatedRequirementsKHR memDedicatedReq = { VK_STRUCTURE_TYPE_MEMORY_DEDICATED_REQUIREMENTS_KHR }; + + VkMemoryRequirements2KHR memReq2 = { VK_STRUCTURE_TYPE_MEMORY_REQUIREMENTS_2_KHR }; + VmaPnextChainPushFront(&memReq2, &memDedicatedReq); + + (*m_VulkanFunctions.vkGetImageMemoryRequirements2KHR)(m_hDevice, &memReqInfo, &memReq2); + + memReq = memReq2.memoryRequirements; + requiresDedicatedAllocation = (memDedicatedReq.requiresDedicatedAllocation != VK_FALSE); + prefersDedicatedAllocation = (memDedicatedReq.prefersDedicatedAllocation != VK_FALSE); + } + else +#endif // #if VMA_DEDICATED_ALLOCATION || VMA_VULKAN_VERSION >= 1001000 + { + (*m_VulkanFunctions.vkGetImageMemoryRequirements)(m_hDevice, hImage, &memReq); + requiresDedicatedAllocation = false; + prefersDedicatedAllocation = false; + } +} + +VkResult VmaAllocator_T::AllocateMemory( + const VkMemoryRequirements& vkMemReq, + bool requiresDedicatedAllocation, + bool prefersDedicatedAllocation, + VkBuffer dedicatedBuffer, + VkBufferUsageFlags dedicatedBufferUsage, + VkImage dedicatedImage, + const VmaAllocationCreateInfo& createInfo, + VmaSuballocationType suballocType, + size_t allocationCount, + VmaAllocation* pAllocations) +{ + memset(pAllocations, 0, sizeof(VmaAllocation) * allocationCount); + + VMA_ASSERT(VmaIsPow2(vkMemReq.alignment)); + + if(vkMemReq.size == 0) + { + return VK_ERROR_VALIDATION_FAILED_EXT; + } + if((createInfo.flags & VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT) != 0 && + (createInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) + { + VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT together with VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT makes no sense."); + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + if((createInfo.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0 && + (createInfo.flags & VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT) != 0) + { + VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_MAPPED_BIT together with VMA_ALLOCATION_CREATE_CAN_BECOME_LOST_BIT is invalid."); + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + if(requiresDedicatedAllocation) + { + if((createInfo.flags & VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT) != 0) + { + VMA_ASSERT(0 && "VMA_ALLOCATION_CREATE_NEVER_ALLOCATE_BIT specified while dedicated allocation is required."); + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + if(createInfo.pool != VK_NULL_HANDLE) + { + VMA_ASSERT(0 && "Pool specified while dedicated allocation is required."); + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + } + if((createInfo.pool != VK_NULL_HANDLE) && + ((createInfo.flags & (VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT)) != 0)) + { + VMA_ASSERT(0 && "Specifying VMA_ALLOCATION_CREATE_DEDICATED_MEMORY_BIT when pool != null is invalid."); + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + + if(createInfo.pool != VK_NULL_HANDLE) + { + const VkDeviceSize alignmentForPool = VMA_MAX( + vkMemReq.alignment, + GetMemoryTypeMinAlignment(createInfo.pool->m_BlockVector.GetMemoryTypeIndex())); + + VmaAllocationCreateInfo createInfoForPool = createInfo; + // If memory type is not HOST_VISIBLE, disable MAPPED. + if((createInfoForPool.flags & VMA_ALLOCATION_CREATE_MAPPED_BIT) != 0 && + (m_MemProps.memoryTypes[createInfo.pool->m_BlockVector.GetMemoryTypeIndex()].propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) == 0) + { + createInfoForPool.flags &= ~VMA_ALLOCATION_CREATE_MAPPED_BIT; + } + + return createInfo.pool->m_BlockVector.Allocate( + m_CurrentFrameIndex.load(), + vkMemReq.size, + alignmentForPool, + createInfoForPool, + suballocType, + allocationCount, + pAllocations); + } + else + { + // Bit mask of memory Vulkan types acceptable for this allocation. + uint32_t memoryTypeBits = vkMemReq.memoryTypeBits; + uint32_t memTypeIndex = UINT32_MAX; + VkResult res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfo, &memTypeIndex); + if(res == VK_SUCCESS) + { + VkDeviceSize alignmentForMemType = VMA_MAX( + vkMemReq.alignment, + GetMemoryTypeMinAlignment(memTypeIndex)); + + res = AllocateMemoryOfType( + vkMemReq.size, + alignmentForMemType, + requiresDedicatedAllocation || prefersDedicatedAllocation, + dedicatedBuffer, + dedicatedBufferUsage, + dedicatedImage, + createInfo, + memTypeIndex, + suballocType, + allocationCount, + pAllocations); + // Succeeded on first try. + if(res == VK_SUCCESS) + { + return res; + } + // Allocation from this memory type failed. Try other compatible memory types. + else + { + for(;;) + { + // Remove old memTypeIndex from list of possibilities. + memoryTypeBits &= ~(1u << memTypeIndex); + // Find alternative memTypeIndex. + res = vmaFindMemoryTypeIndex(this, memoryTypeBits, &createInfo, &memTypeIndex); + if(res == VK_SUCCESS) + { + alignmentForMemType = VMA_MAX( + vkMemReq.alignment, + GetMemoryTypeMinAlignment(memTypeIndex)); + + res = AllocateMemoryOfType( + vkMemReq.size, + alignmentForMemType, + requiresDedicatedAllocation || prefersDedicatedAllocation, + dedicatedBuffer, + dedicatedBufferUsage, + dedicatedImage, + createInfo, + memTypeIndex, + suballocType, + allocationCount, + pAllocations); + // Allocation from this alternative memory type succeeded. + if(res == VK_SUCCESS) + { + return res; + } + // else: Allocation from this memory type failed. Try next one - next loop iteration. + } + // No other matching memory type index could be found. + else + { + // Not returning res, which is VK_ERROR_FEATURE_NOT_PRESENT, because we already failed to allocate once. + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + } + } + } + // Can't find any single memory type maching requirements. res is VK_ERROR_FEATURE_NOT_PRESENT. + else + return res; + } +} + +void VmaAllocator_T::FreeMemory( + size_t allocationCount, + const VmaAllocation* pAllocations) +{ + VMA_ASSERT(pAllocations); + + for(size_t allocIndex = allocationCount; allocIndex--; ) + { + VmaAllocation allocation = pAllocations[allocIndex]; + + if(allocation != VK_NULL_HANDLE) + { + if(TouchAllocation(allocation)) + { + if(VMA_DEBUG_INITIALIZE_ALLOCATIONS) + { + FillAllocation(allocation, VMA_ALLOCATION_FILL_PATTERN_DESTROYED); + } + + switch(allocation->GetType()) + { + case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: + { + VmaBlockVector* pBlockVector = VMA_NULL; + VmaPool hPool = allocation->GetBlock()->GetParentPool(); + if(hPool != VK_NULL_HANDLE) + { + pBlockVector = &hPool->m_BlockVector; + } + else + { + const uint32_t memTypeIndex = allocation->GetMemoryTypeIndex(); + pBlockVector = m_pBlockVectors[memTypeIndex]; + } + pBlockVector->Free(allocation); + } + break; + case VmaAllocation_T::ALLOCATION_TYPE_DEDICATED: + FreeDedicatedMemory(allocation); + break; + default: + VMA_ASSERT(0); + } + } + + // Do this regardless of whether the allocation is lost. Lost allocations still account to Budget.AllocationBytes. + m_Budget.RemoveAllocation(MemoryTypeIndexToHeapIndex(allocation->GetMemoryTypeIndex()), allocation->GetSize()); + allocation->SetUserData(this, VMA_NULL); + m_AllocationObjectAllocator.Free(allocation); + } + } +} + +VkResult VmaAllocator_T::ResizeAllocation( + const VmaAllocation alloc, + VkDeviceSize newSize) +{ + // This function is deprecated and so it does nothing. It's left for backward compatibility. + if(newSize == 0 || alloc->GetLastUseFrameIndex() == VMA_FRAME_INDEX_LOST) + { + return VK_ERROR_VALIDATION_FAILED_EXT; + } + if(newSize == alloc->GetSize()) + { + return VK_SUCCESS; + } + return VK_ERROR_OUT_OF_POOL_MEMORY; +} + +void VmaAllocator_T::CalculateStats(VmaStats* pStats) +{ + // Initialize. + InitStatInfo(pStats->total); + for(size_t i = 0; i < VK_MAX_MEMORY_TYPES; ++i) + InitStatInfo(pStats->memoryType[i]); + for(size_t i = 0; i < VK_MAX_MEMORY_HEAPS; ++i) + InitStatInfo(pStats->memoryHeap[i]); + + // Process default pools. + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + VmaBlockVector* const pBlockVector = m_pBlockVectors[memTypeIndex]; + VMA_ASSERT(pBlockVector); + pBlockVector->AddStats(pStats); + } + + // Process custom pools. + { + VmaMutexLockRead lock(m_PoolsMutex, m_UseMutex); + for(size_t poolIndex = 0, poolCount = m_Pools.size(); poolIndex < poolCount; ++poolIndex) + { + m_Pools[poolIndex]->m_BlockVector.AddStats(pStats); + } + } + + // Process dedicated allocations. + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + const uint32_t memHeapIndex = MemoryTypeIndexToHeapIndex(memTypeIndex); + VmaMutexLockRead dedicatedAllocationsLock(m_DedicatedAllocationsMutex[memTypeIndex], m_UseMutex); + AllocationVectorType* const pDedicatedAllocVector = m_pDedicatedAllocations[memTypeIndex]; + VMA_ASSERT(pDedicatedAllocVector); + for(size_t allocIndex = 0, allocCount = pDedicatedAllocVector->size(); allocIndex < allocCount; ++allocIndex) + { + VmaStatInfo allocationStatInfo; + (*pDedicatedAllocVector)[allocIndex]->DedicatedAllocCalcStatsInfo(allocationStatInfo); + VmaAddStatInfo(pStats->total, allocationStatInfo); + VmaAddStatInfo(pStats->memoryType[memTypeIndex], allocationStatInfo); + VmaAddStatInfo(pStats->memoryHeap[memHeapIndex], allocationStatInfo); + } + } + + // Postprocess. + VmaPostprocessCalcStatInfo(pStats->total); + for(size_t i = 0; i < GetMemoryTypeCount(); ++i) + VmaPostprocessCalcStatInfo(pStats->memoryType[i]); + for(size_t i = 0; i < GetMemoryHeapCount(); ++i) + VmaPostprocessCalcStatInfo(pStats->memoryHeap[i]); +} + +void VmaAllocator_T::GetBudget(VmaBudget* outBudget, uint32_t firstHeap, uint32_t heapCount) +{ +#if VMA_MEMORY_BUDGET + if(m_UseExtMemoryBudget) + { + if(m_Budget.m_OperationsSinceBudgetFetch < 30) + { + VmaMutexLockRead lockRead(m_Budget.m_BudgetMutex, m_UseMutex); + for(uint32_t i = 0; i < heapCount; ++i, ++outBudget) + { + const uint32_t heapIndex = firstHeap + i; + + outBudget->blockBytes = m_Budget.m_BlockBytes[heapIndex]; + outBudget->allocationBytes = m_Budget.m_AllocationBytes[heapIndex]; + + if(m_Budget.m_VulkanUsage[heapIndex] + outBudget->blockBytes > m_Budget.m_BlockBytesAtBudgetFetch[heapIndex]) + { + outBudget->usage = m_Budget.m_VulkanUsage[heapIndex] + + outBudget->blockBytes - m_Budget.m_BlockBytesAtBudgetFetch[heapIndex]; + } + else + { + outBudget->usage = 0; + } + + // Have to take MIN with heap size because explicit HeapSizeLimit is included in it. + outBudget->budget = VMA_MIN( + m_Budget.m_VulkanBudget[heapIndex], m_MemProps.memoryHeaps[heapIndex].size); + } + } + else + { + UpdateVulkanBudget(); // Outside of mutex lock + GetBudget(outBudget, firstHeap, heapCount); // Recursion + } + } + else +#endif + { + for(uint32_t i = 0; i < heapCount; ++i, ++outBudget) + { + const uint32_t heapIndex = firstHeap + i; + + outBudget->blockBytes = m_Budget.m_BlockBytes[heapIndex]; + outBudget->allocationBytes = m_Budget.m_AllocationBytes[heapIndex]; + + outBudget->usage = outBudget->blockBytes; + outBudget->budget = m_MemProps.memoryHeaps[heapIndex].size * 8 / 10; // 80% heuristics. + } + } +} + +static const uint32_t VMA_VENDOR_ID_AMD = 4098; + +VkResult VmaAllocator_T::DefragmentationBegin( + const VmaDefragmentationInfo2& info, + VmaDefragmentationStats* pStats, + VmaDefragmentationContext* pContext) +{ + if(info.pAllocationsChanged != VMA_NULL) + { + memset(info.pAllocationsChanged, 0, info.allocationCount * sizeof(VkBool32)); + } + + *pContext = vma_new(this, VmaDefragmentationContext_T)( + this, m_CurrentFrameIndex.load(), info.flags, pStats); + + (*pContext)->AddPools(info.poolCount, info.pPools); + (*pContext)->AddAllocations( + info.allocationCount, info.pAllocations, info.pAllocationsChanged); + + VkResult res = (*pContext)->Defragment( + info.maxCpuBytesToMove, info.maxCpuAllocationsToMove, + info.maxGpuBytesToMove, info.maxGpuAllocationsToMove, + info.commandBuffer, pStats, info.flags); + + if(res != VK_NOT_READY) + { + vma_delete(this, *pContext); + *pContext = VMA_NULL; + } + + return res; +} + +VkResult VmaAllocator_T::DefragmentationEnd( + VmaDefragmentationContext context) +{ + vma_delete(this, context); + return VK_SUCCESS; +} + +VkResult VmaAllocator_T::DefragmentationPassBegin( + VmaDefragmentationPassInfo* pInfo, + VmaDefragmentationContext context) +{ + return context->DefragmentPassBegin(pInfo); +} +VkResult VmaAllocator_T::DefragmentationPassEnd( + VmaDefragmentationContext context) +{ + return context->DefragmentPassEnd(); + +} + +void VmaAllocator_T::GetAllocationInfo(VmaAllocation hAllocation, VmaAllocationInfo* pAllocationInfo) +{ + if(hAllocation->CanBecomeLost()) + { + /* + Warning: This is a carefully designed algorithm. + Do not modify unless you really know what you're doing :) + */ + const uint32_t localCurrFrameIndex = m_CurrentFrameIndex.load(); + uint32_t localLastUseFrameIndex = hAllocation->GetLastUseFrameIndex(); + for(;;) + { + if(localLastUseFrameIndex == VMA_FRAME_INDEX_LOST) + { + pAllocationInfo->memoryType = UINT32_MAX; + pAllocationInfo->deviceMemory = VK_NULL_HANDLE; + pAllocationInfo->offset = 0; + pAllocationInfo->size = hAllocation->GetSize(); + pAllocationInfo->pMappedData = VMA_NULL; + pAllocationInfo->pUserData = hAllocation->GetUserData(); + return; + } + else if(localLastUseFrameIndex == localCurrFrameIndex) + { + pAllocationInfo->memoryType = hAllocation->GetMemoryTypeIndex(); + pAllocationInfo->deviceMemory = hAllocation->GetMemory(); + pAllocationInfo->offset = hAllocation->GetOffset(); + pAllocationInfo->size = hAllocation->GetSize(); + pAllocationInfo->pMappedData = VMA_NULL; + pAllocationInfo->pUserData = hAllocation->GetUserData(); + return; + } + else // Last use time earlier than current time. + { + if(hAllocation->CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, localCurrFrameIndex)) + { + localLastUseFrameIndex = localCurrFrameIndex; + } + } + } + } + else + { +#if VMA_STATS_STRING_ENABLED + uint32_t localCurrFrameIndex = m_CurrentFrameIndex.load(); + uint32_t localLastUseFrameIndex = hAllocation->GetLastUseFrameIndex(); + for(;;) + { + VMA_ASSERT(localLastUseFrameIndex != VMA_FRAME_INDEX_LOST); + if(localLastUseFrameIndex == localCurrFrameIndex) + { + break; + } + else // Last use time earlier than current time. + { + if(hAllocation->CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, localCurrFrameIndex)) + { + localLastUseFrameIndex = localCurrFrameIndex; + } + } + } +#endif + + pAllocationInfo->memoryType = hAllocation->GetMemoryTypeIndex(); + pAllocationInfo->deviceMemory = hAllocation->GetMemory(); + pAllocationInfo->offset = hAllocation->GetOffset(); + pAllocationInfo->size = hAllocation->GetSize(); + pAllocationInfo->pMappedData = hAllocation->GetMappedData(); + pAllocationInfo->pUserData = hAllocation->GetUserData(); + } +} + +bool VmaAllocator_T::TouchAllocation(VmaAllocation hAllocation) +{ + // This is a stripped-down version of VmaAllocator_T::GetAllocationInfo. + if(hAllocation->CanBecomeLost()) + { + uint32_t localCurrFrameIndex = m_CurrentFrameIndex.load(); + uint32_t localLastUseFrameIndex = hAllocation->GetLastUseFrameIndex(); + for(;;) + { + if(localLastUseFrameIndex == VMA_FRAME_INDEX_LOST) + { + return false; + } + else if(localLastUseFrameIndex == localCurrFrameIndex) + { + return true; + } + else // Last use time earlier than current time. + { + if(hAllocation->CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, localCurrFrameIndex)) + { + localLastUseFrameIndex = localCurrFrameIndex; + } + } + } + } + else + { +#if VMA_STATS_STRING_ENABLED + uint32_t localCurrFrameIndex = m_CurrentFrameIndex.load(); + uint32_t localLastUseFrameIndex = hAllocation->GetLastUseFrameIndex(); + for(;;) + { + VMA_ASSERT(localLastUseFrameIndex != VMA_FRAME_INDEX_LOST); + if(localLastUseFrameIndex == localCurrFrameIndex) + { + break; + } + else // Last use time earlier than current time. + { + if(hAllocation->CompareExchangeLastUseFrameIndex(localLastUseFrameIndex, localCurrFrameIndex)) + { + localLastUseFrameIndex = localCurrFrameIndex; + } + } + } +#endif + + return true; + } +} + +VkResult VmaAllocator_T::CreatePool(const VmaPoolCreateInfo* pCreateInfo, VmaPool* pPool) +{ + VMA_DEBUG_LOG(" CreatePool: MemoryTypeIndex=%u, flags=%u", pCreateInfo->memoryTypeIndex, pCreateInfo->flags); + + VmaPoolCreateInfo newCreateInfo = *pCreateInfo; + + if(newCreateInfo.maxBlockCount == 0) + { + newCreateInfo.maxBlockCount = SIZE_MAX; + } + if(newCreateInfo.minBlockCount > newCreateInfo.maxBlockCount) + { + return VK_ERROR_INITIALIZATION_FAILED; + } + // Memory type index out of range or forbidden. + if(pCreateInfo->memoryTypeIndex >= GetMemoryTypeCount() || + ((1u << pCreateInfo->memoryTypeIndex) & m_GlobalMemoryTypeBits) == 0) + { + return VK_ERROR_FEATURE_NOT_PRESENT; + } + + const VkDeviceSize preferredBlockSize = CalcPreferredBlockSize(newCreateInfo.memoryTypeIndex); + + *pPool = vma_new(this, VmaPool_T)(this, newCreateInfo, preferredBlockSize); + + VkResult res = (*pPool)->m_BlockVector.CreateMinBlocks(); + if(res != VK_SUCCESS) + { + vma_delete(this, *pPool); + *pPool = VMA_NULL; + return res; + } + + // Add to m_Pools. + { + VmaMutexLockWrite lock(m_PoolsMutex, m_UseMutex); + (*pPool)->SetId(m_NextPoolId++); + VmaVectorInsertSorted(m_Pools, *pPool); + } + + return VK_SUCCESS; +} + +void VmaAllocator_T::DestroyPool(VmaPool pool) +{ + // Remove from m_Pools. + { + VmaMutexLockWrite lock(m_PoolsMutex, m_UseMutex); + bool success = VmaVectorRemoveSorted(m_Pools, pool); + VMA_ASSERT(success && "Pool not found in Allocator."); + } + + vma_delete(this, pool); +} + +void VmaAllocator_T::GetPoolStats(VmaPool pool, VmaPoolStats* pPoolStats) +{ + pool->m_BlockVector.GetPoolStats(pPoolStats); +} + +void VmaAllocator_T::SetCurrentFrameIndex(uint32_t frameIndex) +{ + m_CurrentFrameIndex.store(frameIndex); + +#if VMA_MEMORY_BUDGET + if(m_UseExtMemoryBudget) + { + UpdateVulkanBudget(); + } +#endif // #if VMA_MEMORY_BUDGET +} + +void VmaAllocator_T::MakePoolAllocationsLost( + VmaPool hPool, + size_t* pLostAllocationCount) +{ + hPool->m_BlockVector.MakePoolAllocationsLost( + m_CurrentFrameIndex.load(), + pLostAllocationCount); +} + +VkResult VmaAllocator_T::CheckPoolCorruption(VmaPool hPool) +{ + return hPool->m_BlockVector.CheckCorruption(); +} + +VkResult VmaAllocator_T::CheckCorruption(uint32_t memoryTypeBits) +{ + VkResult finalRes = VK_ERROR_FEATURE_NOT_PRESENT; + + // Process default pools. + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + if(((1u << memTypeIndex) & memoryTypeBits) != 0) + { + VmaBlockVector* const pBlockVector = m_pBlockVectors[memTypeIndex]; + VMA_ASSERT(pBlockVector); + VkResult localRes = pBlockVector->CheckCorruption(); + switch(localRes) + { + case VK_ERROR_FEATURE_NOT_PRESENT: + break; + case VK_SUCCESS: + finalRes = VK_SUCCESS; + break; + default: + return localRes; + } + } + } + + // Process custom pools. + { + VmaMutexLockRead lock(m_PoolsMutex, m_UseMutex); + for(size_t poolIndex = 0, poolCount = m_Pools.size(); poolIndex < poolCount; ++poolIndex) + { + if(((1u << m_Pools[poolIndex]->m_BlockVector.GetMemoryTypeIndex()) & memoryTypeBits) != 0) + { + VkResult localRes = m_Pools[poolIndex]->m_BlockVector.CheckCorruption(); + switch(localRes) + { + case VK_ERROR_FEATURE_NOT_PRESENT: + break; + case VK_SUCCESS: + finalRes = VK_SUCCESS; + break; + default: + return localRes; + } + } + } + } + + return finalRes; +} + +void VmaAllocator_T::CreateLostAllocation(VmaAllocation* pAllocation) +{ + *pAllocation = m_AllocationObjectAllocator.Allocate(VMA_FRAME_INDEX_LOST, false); + (*pAllocation)->InitLost(); +} + +VkResult VmaAllocator_T::AllocateVulkanMemory(const VkMemoryAllocateInfo* pAllocateInfo, VkDeviceMemory* pMemory) +{ + const uint32_t heapIndex = MemoryTypeIndexToHeapIndex(pAllocateInfo->memoryTypeIndex); + + // HeapSizeLimit is in effect for this heap. + if((m_HeapSizeLimitMask & (1u << heapIndex)) != 0) + { + const VkDeviceSize heapSize = m_MemProps.memoryHeaps[heapIndex].size; + VkDeviceSize blockBytes = m_Budget.m_BlockBytes[heapIndex]; + for(;;) + { + const VkDeviceSize blockBytesAfterAllocation = blockBytes + pAllocateInfo->allocationSize; + if(blockBytesAfterAllocation > heapSize) + { + return VK_ERROR_OUT_OF_DEVICE_MEMORY; + } + if(m_Budget.m_BlockBytes[heapIndex].compare_exchange_strong(blockBytes, blockBytesAfterAllocation)) + { + break; + } + } + } + else + { + m_Budget.m_BlockBytes[heapIndex] += pAllocateInfo->allocationSize; + } + + // VULKAN CALL vkAllocateMemory. + VkResult res = (*m_VulkanFunctions.vkAllocateMemory)(m_hDevice, pAllocateInfo, GetAllocationCallbacks(), pMemory); + + if(res == VK_SUCCESS) + { +#if VMA_MEMORY_BUDGET + ++m_Budget.m_OperationsSinceBudgetFetch; +#endif + + // Informative callback. + if(m_DeviceMemoryCallbacks.pfnAllocate != VMA_NULL) + { + (*m_DeviceMemoryCallbacks.pfnAllocate)(this, pAllocateInfo->memoryTypeIndex, *pMemory, pAllocateInfo->allocationSize, m_DeviceMemoryCallbacks.pUserData); + } + } + else + { + m_Budget.m_BlockBytes[heapIndex] -= pAllocateInfo->allocationSize; + } + + return res; +} + +void VmaAllocator_T::FreeVulkanMemory(uint32_t memoryType, VkDeviceSize size, VkDeviceMemory hMemory) +{ + // Informative callback. + if(m_DeviceMemoryCallbacks.pfnFree != VMA_NULL) + { + (*m_DeviceMemoryCallbacks.pfnFree)(this, memoryType, hMemory, size, m_DeviceMemoryCallbacks.pUserData); + } + + // VULKAN CALL vkFreeMemory. + (*m_VulkanFunctions.vkFreeMemory)(m_hDevice, hMemory, GetAllocationCallbacks()); + + m_Budget.m_BlockBytes[MemoryTypeIndexToHeapIndex(memoryType)] -= size; +} + +VkResult VmaAllocator_T::BindVulkanBuffer( + VkDeviceMemory memory, + VkDeviceSize memoryOffset, + VkBuffer buffer, + const void* pNext) +{ + if(pNext != VMA_NULL) + { +#if VMA_VULKAN_VERSION >= 1001000 || VMA_BIND_MEMORY2 + if((m_UseKhrBindMemory2 || m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) && + m_VulkanFunctions.vkBindBufferMemory2KHR != VMA_NULL) + { + VkBindBufferMemoryInfoKHR bindBufferMemoryInfo = { VK_STRUCTURE_TYPE_BIND_BUFFER_MEMORY_INFO_KHR }; + bindBufferMemoryInfo.pNext = pNext; + bindBufferMemoryInfo.buffer = buffer; + bindBufferMemoryInfo.memory = memory; + bindBufferMemoryInfo.memoryOffset = memoryOffset; + return (*m_VulkanFunctions.vkBindBufferMemory2KHR)(m_hDevice, 1, &bindBufferMemoryInfo); + } + else +#endif // #if VMA_VULKAN_VERSION >= 1001000 || VMA_BIND_MEMORY2 + { + return VK_ERROR_EXTENSION_NOT_PRESENT; + } + } + else + { + return (*m_VulkanFunctions.vkBindBufferMemory)(m_hDevice, buffer, memory, memoryOffset); + } +} + +VkResult VmaAllocator_T::BindVulkanImage( + VkDeviceMemory memory, + VkDeviceSize memoryOffset, + VkImage image, + const void* pNext) +{ + if(pNext != VMA_NULL) + { +#if VMA_VULKAN_VERSION >= 1001000 || VMA_BIND_MEMORY2 + if((m_UseKhrBindMemory2 || m_VulkanApiVersion >= VK_MAKE_VERSION(1, 1, 0)) && + m_VulkanFunctions.vkBindImageMemory2KHR != VMA_NULL) + { + VkBindImageMemoryInfoKHR bindBufferMemoryInfo = { VK_STRUCTURE_TYPE_BIND_IMAGE_MEMORY_INFO_KHR }; + bindBufferMemoryInfo.pNext = pNext; + bindBufferMemoryInfo.image = image; + bindBufferMemoryInfo.memory = memory; + bindBufferMemoryInfo.memoryOffset = memoryOffset; + return (*m_VulkanFunctions.vkBindImageMemory2KHR)(m_hDevice, 1, &bindBufferMemoryInfo); + } + else +#endif // #if VMA_BIND_MEMORY2 + { + return VK_ERROR_EXTENSION_NOT_PRESENT; + } + } + else + { + return (*m_VulkanFunctions.vkBindImageMemory)(m_hDevice, image, memory, memoryOffset); + } +} + +VkResult VmaAllocator_T::Map(VmaAllocation hAllocation, void** ppData) +{ + if(hAllocation->CanBecomeLost()) + { + return VK_ERROR_MEMORY_MAP_FAILED; + } + + switch(hAllocation->GetType()) + { + case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: + { + VmaDeviceMemoryBlock* const pBlock = hAllocation->GetBlock(); + char *pBytes = VMA_NULL; + VkResult res = pBlock->Map(this, 1, (void**)&pBytes); + if(res == VK_SUCCESS) + { + *ppData = pBytes + (ptrdiff_t)hAllocation->GetOffset(); + hAllocation->BlockAllocMap(); + } + return res; + } + case VmaAllocation_T::ALLOCATION_TYPE_DEDICATED: + return hAllocation->DedicatedAllocMap(this, ppData); + default: + VMA_ASSERT(0); + return VK_ERROR_MEMORY_MAP_FAILED; + } +} + +void VmaAllocator_T::Unmap(VmaAllocation hAllocation) +{ + switch(hAllocation->GetType()) + { + case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: + { + VmaDeviceMemoryBlock* const pBlock = hAllocation->GetBlock(); + hAllocation->BlockAllocUnmap(); + pBlock->Unmap(this, 1); + } + break; + case VmaAllocation_T::ALLOCATION_TYPE_DEDICATED: + hAllocation->DedicatedAllocUnmap(this); + break; + default: + VMA_ASSERT(0); + } +} + +VkResult VmaAllocator_T::BindBufferMemory( + VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkBuffer hBuffer, + const void* pNext) +{ + VkResult res = VK_SUCCESS; + switch(hAllocation->GetType()) + { + case VmaAllocation_T::ALLOCATION_TYPE_DEDICATED: + res = BindVulkanBuffer(hAllocation->GetMemory(), allocationLocalOffset, hBuffer, pNext); + break; + case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: + { + VmaDeviceMemoryBlock* const pBlock = hAllocation->GetBlock(); + VMA_ASSERT(pBlock && "Binding buffer to allocation that doesn't belong to any block. Is the allocation lost?"); + res = pBlock->BindBufferMemory(this, hAllocation, allocationLocalOffset, hBuffer, pNext); + break; + } + default: + VMA_ASSERT(0); + } + return res; +} + +VkResult VmaAllocator_T::BindImageMemory( + VmaAllocation hAllocation, + VkDeviceSize allocationLocalOffset, + VkImage hImage, + const void* pNext) +{ + VkResult res = VK_SUCCESS; + switch(hAllocation->GetType()) + { + case VmaAllocation_T::ALLOCATION_TYPE_DEDICATED: + res = BindVulkanImage(hAllocation->GetMemory(), allocationLocalOffset, hImage, pNext); + break; + case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: + { + VmaDeviceMemoryBlock* pBlock = hAllocation->GetBlock(); + VMA_ASSERT(pBlock && "Binding image to allocation that doesn't belong to any block. Is the allocation lost?"); + res = pBlock->BindImageMemory(this, hAllocation, allocationLocalOffset, hImage, pNext); + break; + } + default: + VMA_ASSERT(0); + } + return res; +} + +VkResult VmaAllocator_T::FlushOrInvalidateAllocation( + VmaAllocation hAllocation, + VkDeviceSize offset, VkDeviceSize size, + VMA_CACHE_OPERATION op) +{ + VkResult res = VK_SUCCESS; + + VkMappedMemoryRange memRange = {}; + if(GetFlushOrInvalidateRange(hAllocation, offset, size, memRange)) + { + switch(op) + { + case VMA_CACHE_FLUSH: + res = (*GetVulkanFunctions().vkFlushMappedMemoryRanges)(m_hDevice, 1, &memRange); + break; + case VMA_CACHE_INVALIDATE: + res = (*GetVulkanFunctions().vkInvalidateMappedMemoryRanges)(m_hDevice, 1, &memRange); + break; + default: + VMA_ASSERT(0); + } + } + // else: Just ignore this call. + return res; +} + +VkResult VmaAllocator_T::FlushOrInvalidateAllocations( + uint32_t allocationCount, + const VmaAllocation* allocations, + const VkDeviceSize* offsets, const VkDeviceSize* sizes, + VMA_CACHE_OPERATION op) +{ + typedef VmaStlAllocator RangeAllocator; + typedef VmaSmallVector RangeVector; + RangeVector ranges = RangeVector(RangeAllocator(GetAllocationCallbacks())); + + for(uint32_t allocIndex = 0; allocIndex < allocationCount; ++allocIndex) + { + const VmaAllocation alloc = allocations[allocIndex]; + const VkDeviceSize offset = offsets != VMA_NULL ? offsets[allocIndex] : 0; + const VkDeviceSize size = sizes != VMA_NULL ? sizes[allocIndex] : VK_WHOLE_SIZE; + VkMappedMemoryRange newRange; + if(GetFlushOrInvalidateRange(alloc, offset, size, newRange)) + { + ranges.push_back(newRange); + } + } + + VkResult res = VK_SUCCESS; + if(!ranges.empty()) + { + switch(op) + { + case VMA_CACHE_FLUSH: + res = (*GetVulkanFunctions().vkFlushMappedMemoryRanges)(m_hDevice, (uint32_t)ranges.size(), ranges.data()); + break; + case VMA_CACHE_INVALIDATE: + res = (*GetVulkanFunctions().vkInvalidateMappedMemoryRanges)(m_hDevice, (uint32_t)ranges.size(), ranges.data()); + break; + default: + VMA_ASSERT(0); + } + } + // else: Just ignore this call. + return res; +} + +void VmaAllocator_T::FreeDedicatedMemory(const VmaAllocation allocation) +{ + VMA_ASSERT(allocation && allocation->GetType() == VmaAllocation_T::ALLOCATION_TYPE_DEDICATED); + + const uint32_t memTypeIndex = allocation->GetMemoryTypeIndex(); + { + VmaMutexLockWrite lock(m_DedicatedAllocationsMutex[memTypeIndex], m_UseMutex); + AllocationVectorType* const pDedicatedAllocations = m_pDedicatedAllocations[memTypeIndex]; + VMA_ASSERT(pDedicatedAllocations); + bool success = VmaVectorRemoveSorted(*pDedicatedAllocations, allocation); + VMA_ASSERT(success); + } + + VkDeviceMemory hMemory = allocation->GetMemory(); + + /* + There is no need to call this, because Vulkan spec allows to skip vkUnmapMemory + before vkFreeMemory. + + if(allocation->GetMappedData() != VMA_NULL) + { + (*m_VulkanFunctions.vkUnmapMemory)(m_hDevice, hMemory); + } + */ + + FreeVulkanMemory(memTypeIndex, allocation->GetSize(), hMemory); + + VMA_DEBUG_LOG(" Freed DedicatedMemory MemoryTypeIndex=%u", memTypeIndex); +} + +uint32_t VmaAllocator_T::CalculateGpuDefragmentationMemoryTypeBits() const +{ + VkBufferCreateInfo dummyBufCreateInfo; + VmaFillGpuDefragmentationBufferCreateInfo(dummyBufCreateInfo); + + uint32_t memoryTypeBits = 0; + + // Create buffer. + VkBuffer buf = VK_NULL_HANDLE; + VkResult res = (*GetVulkanFunctions().vkCreateBuffer)( + m_hDevice, &dummyBufCreateInfo, GetAllocationCallbacks(), &buf); + if(res == VK_SUCCESS) + { + // Query for supported memory types. + VkMemoryRequirements memReq; + (*GetVulkanFunctions().vkGetBufferMemoryRequirements)(m_hDevice, buf, &memReq); + memoryTypeBits = memReq.memoryTypeBits; + + // Destroy buffer. + (*GetVulkanFunctions().vkDestroyBuffer)(m_hDevice, buf, GetAllocationCallbacks()); + } + + return memoryTypeBits; +} + +uint32_t VmaAllocator_T::CalculateGlobalMemoryTypeBits() const +{ + // Make sure memory information is already fetched. + VMA_ASSERT(GetMemoryTypeCount() > 0); + + uint32_t memoryTypeBits = UINT32_MAX; + + if(!m_UseAmdDeviceCoherentMemory) + { + // Exclude memory types that have VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD. + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + if((m_MemProps.memoryTypes[memTypeIndex].propertyFlags & VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD_COPY) != 0) + { + memoryTypeBits &= ~(1u << memTypeIndex); + } + } + } + + return memoryTypeBits; +} + +bool VmaAllocator_T::GetFlushOrInvalidateRange( + VmaAllocation allocation, + VkDeviceSize offset, VkDeviceSize size, + VkMappedMemoryRange& outRange) const +{ + const uint32_t memTypeIndex = allocation->GetMemoryTypeIndex(); + if(size > 0 && IsMemoryTypeNonCoherent(memTypeIndex)) + { + const VkDeviceSize nonCoherentAtomSize = m_PhysicalDeviceProperties.limits.nonCoherentAtomSize; + const VkDeviceSize allocationSize = allocation->GetSize(); + VMA_ASSERT(offset <= allocationSize); + + outRange.sType = VK_STRUCTURE_TYPE_MAPPED_MEMORY_RANGE; + outRange.pNext = VMA_NULL; + outRange.memory = allocation->GetMemory(); + + switch(allocation->GetType()) + { + case VmaAllocation_T::ALLOCATION_TYPE_DEDICATED: + outRange.offset = VmaAlignDown(offset, nonCoherentAtomSize); + if(size == VK_WHOLE_SIZE) + { + outRange.size = allocationSize - outRange.offset; + } + else + { + VMA_ASSERT(offset + size <= allocationSize); + outRange.size = VMA_MIN( + VmaAlignUp(size + (offset - outRange.offset), nonCoherentAtomSize), + allocationSize - outRange.offset); + } + break; + case VmaAllocation_T::ALLOCATION_TYPE_BLOCK: + { + // 1. Still within this allocation. + outRange.offset = VmaAlignDown(offset, nonCoherentAtomSize); + if(size == VK_WHOLE_SIZE) + { + size = allocationSize - offset; + } + else + { + VMA_ASSERT(offset + size <= allocationSize); + } + outRange.size = VmaAlignUp(size + (offset - outRange.offset), nonCoherentAtomSize); + + // 2. Adjust to whole block. + const VkDeviceSize allocationOffset = allocation->GetOffset(); + VMA_ASSERT(allocationOffset % nonCoherentAtomSize == 0); + const VkDeviceSize blockSize = allocation->GetBlock()->m_pMetadata->GetSize(); + outRange.offset += allocationOffset; + outRange.size = VMA_MIN(outRange.size, blockSize - outRange.offset); + + break; + } + default: + VMA_ASSERT(0); + } + return true; + } + return false; +} + +#if VMA_MEMORY_BUDGET + +void VmaAllocator_T::UpdateVulkanBudget() +{ + VMA_ASSERT(m_UseExtMemoryBudget); + + VkPhysicalDeviceMemoryProperties2KHR memProps = { VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_PROPERTIES_2_KHR }; + + VkPhysicalDeviceMemoryBudgetPropertiesEXT budgetProps = { VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_BUDGET_PROPERTIES_EXT }; + VmaPnextChainPushFront(&memProps, &budgetProps); + + GetVulkanFunctions().vkGetPhysicalDeviceMemoryProperties2KHR(m_PhysicalDevice, &memProps); + + { + VmaMutexLockWrite lockWrite(m_Budget.m_BudgetMutex, m_UseMutex); + + for(uint32_t heapIndex = 0; heapIndex < GetMemoryHeapCount(); ++heapIndex) + { + m_Budget.m_VulkanUsage[heapIndex] = budgetProps.heapUsage[heapIndex]; + m_Budget.m_VulkanBudget[heapIndex] = budgetProps.heapBudget[heapIndex]; + m_Budget.m_BlockBytesAtBudgetFetch[heapIndex] = m_Budget.m_BlockBytes[heapIndex].load(); + + // Some bugged drivers return the budget incorrectly, e.g. 0 or much bigger than heap size. + if(m_Budget.m_VulkanBudget[heapIndex] == 0) + { + m_Budget.m_VulkanBudget[heapIndex] = m_MemProps.memoryHeaps[heapIndex].size * 8 / 10; // 80% heuristics. + } + else if(m_Budget.m_VulkanBudget[heapIndex] > m_MemProps.memoryHeaps[heapIndex].size) + { + m_Budget.m_VulkanBudget[heapIndex] = m_MemProps.memoryHeaps[heapIndex].size; + } + if(m_Budget.m_VulkanUsage[heapIndex] == 0 && m_Budget.m_BlockBytesAtBudgetFetch[heapIndex] > 0) + { + m_Budget.m_VulkanUsage[heapIndex] = m_Budget.m_BlockBytesAtBudgetFetch[heapIndex]; + } + } + m_Budget.m_OperationsSinceBudgetFetch = 0; + } +} + +#endif // #if VMA_MEMORY_BUDGET + +void VmaAllocator_T::FillAllocation(const VmaAllocation hAllocation, uint8_t pattern) +{ + if(VMA_DEBUG_INITIALIZE_ALLOCATIONS && + !hAllocation->CanBecomeLost() && + (m_MemProps.memoryTypes[hAllocation->GetMemoryTypeIndex()].propertyFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) != 0) + { + void* pData = VMA_NULL; + VkResult res = Map(hAllocation, &pData); + if(res == VK_SUCCESS) + { + memset(pData, (int)pattern, (size_t)hAllocation->GetSize()); + FlushOrInvalidateAllocation(hAllocation, 0, VK_WHOLE_SIZE, VMA_CACHE_FLUSH); + Unmap(hAllocation); + } + else + { + VMA_ASSERT(0 && "VMA_DEBUG_INITIALIZE_ALLOCATIONS is enabled, but couldn't map memory to fill allocation."); + } + } +} + +uint32_t VmaAllocator_T::GetGpuDefragmentationMemoryTypeBits() +{ + uint32_t memoryTypeBits = m_GpuDefragmentationMemoryTypeBits.load(); + if(memoryTypeBits == UINT32_MAX) + { + memoryTypeBits = CalculateGpuDefragmentationMemoryTypeBits(); + m_GpuDefragmentationMemoryTypeBits.store(memoryTypeBits); + } + return memoryTypeBits; +} + +#if VMA_STATS_STRING_ENABLED + +void VmaAllocator_T::PrintDetailedMap(VmaJsonWriter& json) +{ + bool dedicatedAllocationsStarted = false; + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + VmaMutexLockRead dedicatedAllocationsLock(m_DedicatedAllocationsMutex[memTypeIndex], m_UseMutex); + AllocationVectorType* const pDedicatedAllocVector = m_pDedicatedAllocations[memTypeIndex]; + VMA_ASSERT(pDedicatedAllocVector); + if(pDedicatedAllocVector->empty() == false) + { + if(dedicatedAllocationsStarted == false) + { + dedicatedAllocationsStarted = true; + json.WriteString("DedicatedAllocations"); + json.BeginObject(); + } + + json.BeginString("Type "); + json.ContinueString(memTypeIndex); + json.EndString(); + + json.BeginArray(); + + for(size_t i = 0; i < pDedicatedAllocVector->size(); ++i) + { + json.BeginObject(true); + const VmaAllocation hAlloc = (*pDedicatedAllocVector)[i]; + hAlloc->PrintParameters(json); + json.EndObject(); + } + + json.EndArray(); + } + } + if(dedicatedAllocationsStarted) + { + json.EndObject(); + } + + { + bool allocationsStarted = false; + for(uint32_t memTypeIndex = 0; memTypeIndex < GetMemoryTypeCount(); ++memTypeIndex) + { + if(m_pBlockVectors[memTypeIndex]->IsEmpty() == false) + { + if(allocationsStarted == false) + { + allocationsStarted = true; + json.WriteString("DefaultPools"); + json.BeginObject(); + } + + json.BeginString("Type "); + json.ContinueString(memTypeIndex); + json.EndString(); + + m_pBlockVectors[memTypeIndex]->PrintDetailedMap(json); + } + } + if(allocationsStarted) + { + json.EndObject(); + } + } + + // Custom pools + { + VmaMutexLockRead lock(m_PoolsMutex, m_UseMutex); + const size_t poolCount = m_Pools.size(); + if(poolCount > 0) + { + json.WriteString("Pools"); + json.BeginObject(); + for(size_t poolIndex = 0; poolIndex < poolCount; ++poolIndex) + { + json.BeginString(); + json.ContinueString(m_Pools[poolIndex]->GetId()); + json.EndString(); + + m_Pools[poolIndex]->m_BlockVector.PrintDetailedMap(json); + } + json.EndObject(); + } + } +} + +#endif // #if VMA_STATS_STRING_ENABLED + +//////////////////////////////////////////////////////////////////////////////// +// Public interface + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateAllocator( + const VmaAllocatorCreateInfo* pCreateInfo, + VmaAllocator* pAllocator) +{ + VMA_ASSERT(pCreateInfo && pAllocator); + VMA_ASSERT(pCreateInfo->vulkanApiVersion == 0 || + (VK_VERSION_MAJOR(pCreateInfo->vulkanApiVersion) == 1 && VK_VERSION_MINOR(pCreateInfo->vulkanApiVersion) <= 2)); + VMA_DEBUG_LOG("vmaCreateAllocator"); + *pAllocator = vma_new(pCreateInfo->pAllocationCallbacks, VmaAllocator_T)(pCreateInfo); + return (*pAllocator)->Init(pCreateInfo); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyAllocator( + VmaAllocator allocator) +{ + if(allocator != VK_NULL_HANDLE) + { + VMA_DEBUG_LOG("vmaDestroyAllocator"); + VkAllocationCallbacks allocationCallbacks = allocator->m_AllocationCallbacks; + vma_delete(&allocationCallbacks, allocator); + } +} + +VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocatorInfo(VmaAllocator allocator, VmaAllocatorInfo* pAllocatorInfo) +{ + VMA_ASSERT(allocator && pAllocatorInfo); + pAllocatorInfo->instance = allocator->m_hInstance; + pAllocatorInfo->physicalDevice = allocator->GetPhysicalDevice(); + pAllocatorInfo->device = allocator->m_hDevice; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaGetPhysicalDeviceProperties( + VmaAllocator allocator, + const VkPhysicalDeviceProperties **ppPhysicalDeviceProperties) +{ + VMA_ASSERT(allocator && ppPhysicalDeviceProperties); + *ppPhysicalDeviceProperties = &allocator->m_PhysicalDeviceProperties; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaGetMemoryProperties( + VmaAllocator allocator, + const VkPhysicalDeviceMemoryProperties** ppPhysicalDeviceMemoryProperties) +{ + VMA_ASSERT(allocator && ppPhysicalDeviceMemoryProperties); + *ppPhysicalDeviceMemoryProperties = &allocator->m_MemProps; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaGetMemoryTypeProperties( + VmaAllocator allocator, + uint32_t memoryTypeIndex, + VkMemoryPropertyFlags* pFlags) +{ + VMA_ASSERT(allocator && pFlags); + VMA_ASSERT(memoryTypeIndex < allocator->GetMemoryTypeCount()); + *pFlags = allocator->m_MemProps.memoryTypes[memoryTypeIndex].propertyFlags; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaSetCurrentFrameIndex( + VmaAllocator allocator, + uint32_t frameIndex) +{ + VMA_ASSERT(allocator); + VMA_ASSERT(frameIndex != VMA_FRAME_INDEX_LOST); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + allocator->SetCurrentFrameIndex(frameIndex); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaCalculateStats( + VmaAllocator allocator, + VmaStats* pStats) +{ + VMA_ASSERT(allocator && pStats); + VMA_DEBUG_GLOBAL_MUTEX_LOCK + allocator->CalculateStats(pStats); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaGetBudget( + VmaAllocator allocator, + VmaBudget* pBudget) +{ + VMA_ASSERT(allocator && pBudget); + VMA_DEBUG_GLOBAL_MUTEX_LOCK + allocator->GetBudget(pBudget, 0, allocator->GetMemoryHeapCount()); +} + +#if VMA_STATS_STRING_ENABLED + +VMA_CALL_PRE void VMA_CALL_POST vmaBuildStatsString( + VmaAllocator allocator, + char** ppStatsString, + VkBool32 detailedMap) +{ + VMA_ASSERT(allocator && ppStatsString); + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + VmaStringBuilder sb(allocator); + { + VmaJsonWriter json(allocator->GetAllocationCallbacks(), sb); + json.BeginObject(); + + VmaBudget budget[VK_MAX_MEMORY_HEAPS]; + allocator->GetBudget(budget, 0, allocator->GetMemoryHeapCount()); + + VmaStats stats; + allocator->CalculateStats(&stats); + + json.WriteString("Total"); + VmaPrintStatInfo(json, stats.total); + + for(uint32_t heapIndex = 0; heapIndex < allocator->GetMemoryHeapCount(); ++heapIndex) + { + json.BeginString("Heap "); + json.ContinueString(heapIndex); + json.EndString(); + json.BeginObject(); + + json.WriteString("Size"); + json.WriteNumber(allocator->m_MemProps.memoryHeaps[heapIndex].size); + + json.WriteString("Flags"); + json.BeginArray(true); + if((allocator->m_MemProps.memoryHeaps[heapIndex].flags & VK_MEMORY_HEAP_DEVICE_LOCAL_BIT) != 0) + { + json.WriteString("DEVICE_LOCAL"); + } + json.EndArray(); + + json.WriteString("Budget"); + json.BeginObject(); + { + json.WriteString("BlockBytes"); + json.WriteNumber(budget[heapIndex].blockBytes); + json.WriteString("AllocationBytes"); + json.WriteNumber(budget[heapIndex].allocationBytes); + json.WriteString("Usage"); + json.WriteNumber(budget[heapIndex].usage); + json.WriteString("Budget"); + json.WriteNumber(budget[heapIndex].budget); + } + json.EndObject(); + + if(stats.memoryHeap[heapIndex].blockCount > 0) + { + json.WriteString("Stats"); + VmaPrintStatInfo(json, stats.memoryHeap[heapIndex]); + } + + for(uint32_t typeIndex = 0; typeIndex < allocator->GetMemoryTypeCount(); ++typeIndex) + { + if(allocator->MemoryTypeIndexToHeapIndex(typeIndex) == heapIndex) + { + json.BeginString("Type "); + json.ContinueString(typeIndex); + json.EndString(); + + json.BeginObject(); + + json.WriteString("Flags"); + json.BeginArray(true); + VkMemoryPropertyFlags flags = allocator->m_MemProps.memoryTypes[typeIndex].propertyFlags; + if((flags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT) != 0) + { + json.WriteString("DEVICE_LOCAL"); + } + if((flags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) != 0) + { + json.WriteString("HOST_VISIBLE"); + } + if((flags & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) != 0) + { + json.WriteString("HOST_COHERENT"); + } + if((flags & VK_MEMORY_PROPERTY_HOST_CACHED_BIT) != 0) + { + json.WriteString("HOST_CACHED"); + } + if((flags & VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT) != 0) + { + json.WriteString("LAZILY_ALLOCATED"); + } + if((flags & VK_MEMORY_PROPERTY_PROTECTED_BIT) != 0) + { + json.WriteString(" PROTECTED"); + } + if((flags & VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD_COPY) != 0) + { + json.WriteString(" DEVICE_COHERENT"); + } + if((flags & VK_MEMORY_PROPERTY_DEVICE_UNCACHED_BIT_AMD_COPY) != 0) + { + json.WriteString(" DEVICE_UNCACHED"); + } + json.EndArray(); + + if(stats.memoryType[typeIndex].blockCount > 0) + { + json.WriteString("Stats"); + VmaPrintStatInfo(json, stats.memoryType[typeIndex]); + } + + json.EndObject(); + } + } + + json.EndObject(); + } + if(detailedMap == VK_TRUE) + { + allocator->PrintDetailedMap(json); + } + + json.EndObject(); + } + + const size_t len = sb.GetLength(); + char* const pChars = vma_new_array(allocator, char, len + 1); + if(len > 0) + { + memcpy(pChars, sb.GetData(), len); + } + pChars[len] = '\0'; + *ppStatsString = pChars; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaFreeStatsString( + VmaAllocator allocator, + char* pStatsString) +{ + if(pStatsString != VMA_NULL) + { + VMA_ASSERT(allocator); + size_t len = strlen(pStatsString); + vma_delete_array(allocator, pStatsString, len + 1); + } +} + +#endif // #if VMA_STATS_STRING_ENABLED + +/* +This function is not protected by any mutex because it just reads immutable data. +*/ +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndex( + VmaAllocator allocator, + uint32_t memoryTypeBits, + const VmaAllocationCreateInfo* pAllocationCreateInfo, + uint32_t* pMemoryTypeIndex) +{ + VMA_ASSERT(allocator != VK_NULL_HANDLE); + VMA_ASSERT(pAllocationCreateInfo != VMA_NULL); + VMA_ASSERT(pMemoryTypeIndex != VMA_NULL); + + memoryTypeBits &= allocator->GetGlobalMemoryTypeBits(); + + if(pAllocationCreateInfo->memoryTypeBits != 0) + { + memoryTypeBits &= pAllocationCreateInfo->memoryTypeBits; + } + + uint32_t requiredFlags = pAllocationCreateInfo->requiredFlags; + uint32_t preferredFlags = pAllocationCreateInfo->preferredFlags; + uint32_t notPreferredFlags = 0; + + // Convert usage to requiredFlags and preferredFlags. + switch(pAllocationCreateInfo->usage) + { + case VMA_MEMORY_USAGE_UNKNOWN: + break; + case VMA_MEMORY_USAGE_GPU_ONLY: + if(!allocator->IsIntegratedGpu() || (preferredFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) == 0) + { + preferredFlags |= VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT; + } + break; + case VMA_MEMORY_USAGE_CPU_ONLY: + requiredFlags |= VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT; + break; + case VMA_MEMORY_USAGE_CPU_TO_GPU: + requiredFlags |= VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; + if(!allocator->IsIntegratedGpu() || (preferredFlags & VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT) == 0) + { + preferredFlags |= VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT; + } + break; + case VMA_MEMORY_USAGE_GPU_TO_CPU: + requiredFlags |= VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT; + preferredFlags |= VK_MEMORY_PROPERTY_HOST_CACHED_BIT; + break; + case VMA_MEMORY_USAGE_CPU_COPY: + notPreferredFlags |= VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT; + break; + case VMA_MEMORY_USAGE_GPU_LAZILY_ALLOCATED: + requiredFlags |= VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT; + break; + default: + VMA_ASSERT(0); + break; + } + + // Avoid DEVICE_COHERENT unless explicitly requested. + if(((pAllocationCreateInfo->requiredFlags | pAllocationCreateInfo->preferredFlags) & + (VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD_COPY | VK_MEMORY_PROPERTY_DEVICE_UNCACHED_BIT_AMD_COPY)) == 0) + { + notPreferredFlags |= VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD_COPY; + } + + *pMemoryTypeIndex = UINT32_MAX; + uint32_t minCost = UINT32_MAX; + for(uint32_t memTypeIndex = 0, memTypeBit = 1; + memTypeIndex < allocator->GetMemoryTypeCount(); + ++memTypeIndex, memTypeBit <<= 1) + { + // This memory type is acceptable according to memoryTypeBits bitmask. + if((memTypeBit & memoryTypeBits) != 0) + { + const VkMemoryPropertyFlags currFlags = + allocator->m_MemProps.memoryTypes[memTypeIndex].propertyFlags; + // This memory type contains requiredFlags. + if((requiredFlags & ~currFlags) == 0) + { + // Calculate cost as number of bits from preferredFlags not present in this memory type. + uint32_t currCost = VmaCountBitsSet(preferredFlags & ~currFlags) + + VmaCountBitsSet(currFlags & notPreferredFlags); + // Remember memory type with lowest cost. + if(currCost < minCost) + { + *pMemoryTypeIndex = memTypeIndex; + if(currCost == 0) + { + return VK_SUCCESS; + } + minCost = currCost; + } + } + } + } + return (*pMemoryTypeIndex != UINT32_MAX) ? VK_SUCCESS : VK_ERROR_FEATURE_NOT_PRESENT; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndexForBufferInfo( + VmaAllocator allocator, + const VkBufferCreateInfo* pBufferCreateInfo, + const VmaAllocationCreateInfo* pAllocationCreateInfo, + uint32_t* pMemoryTypeIndex) +{ + VMA_ASSERT(allocator != VK_NULL_HANDLE); + VMA_ASSERT(pBufferCreateInfo != VMA_NULL); + VMA_ASSERT(pAllocationCreateInfo != VMA_NULL); + VMA_ASSERT(pMemoryTypeIndex != VMA_NULL); + + const VkDevice hDev = allocator->m_hDevice; + VkBuffer hBuffer = VK_NULL_HANDLE; + VkResult res = allocator->GetVulkanFunctions().vkCreateBuffer( + hDev, pBufferCreateInfo, allocator->GetAllocationCallbacks(), &hBuffer); + if(res == VK_SUCCESS) + { + VkMemoryRequirements memReq = {}; + allocator->GetVulkanFunctions().vkGetBufferMemoryRequirements( + hDev, hBuffer, &memReq); + + res = vmaFindMemoryTypeIndex( + allocator, + memReq.memoryTypeBits, + pAllocationCreateInfo, + pMemoryTypeIndex); + + allocator->GetVulkanFunctions().vkDestroyBuffer( + hDev, hBuffer, allocator->GetAllocationCallbacks()); + } + return res; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFindMemoryTypeIndexForImageInfo( + VmaAllocator allocator, + const VkImageCreateInfo* pImageCreateInfo, + const VmaAllocationCreateInfo* pAllocationCreateInfo, + uint32_t* pMemoryTypeIndex) +{ + VMA_ASSERT(allocator != VK_NULL_HANDLE); + VMA_ASSERT(pImageCreateInfo != VMA_NULL); + VMA_ASSERT(pAllocationCreateInfo != VMA_NULL); + VMA_ASSERT(pMemoryTypeIndex != VMA_NULL); + + const VkDevice hDev = allocator->m_hDevice; + VkImage hImage = VK_NULL_HANDLE; + VkResult res = allocator->GetVulkanFunctions().vkCreateImage( + hDev, pImageCreateInfo, allocator->GetAllocationCallbacks(), &hImage); + if(res == VK_SUCCESS) + { + VkMemoryRequirements memReq = {}; + allocator->GetVulkanFunctions().vkGetImageMemoryRequirements( + hDev, hImage, &memReq); + + res = vmaFindMemoryTypeIndex( + allocator, + memReq.memoryTypeBits, + pAllocationCreateInfo, + pMemoryTypeIndex); + + allocator->GetVulkanFunctions().vkDestroyImage( + hDev, hImage, allocator->GetAllocationCallbacks()); + } + return res; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreatePool( + VmaAllocator allocator, + const VmaPoolCreateInfo* pCreateInfo, + VmaPool* pPool) +{ + VMA_ASSERT(allocator && pCreateInfo && pPool); + + VMA_DEBUG_LOG("vmaCreatePool"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + VkResult res = allocator->CreatePool(pCreateInfo, pPool); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordCreatePool(allocator->GetCurrentFrameIndex(), *pCreateInfo, *pPool); + } +#endif + + return res; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyPool( + VmaAllocator allocator, + VmaPool pool) +{ + VMA_ASSERT(allocator); + + if(pool == VK_NULL_HANDLE) + { + return; + } + + VMA_DEBUG_LOG("vmaDestroyPool"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordDestroyPool(allocator->GetCurrentFrameIndex(), pool); + } +#endif + + allocator->DestroyPool(pool); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaGetPoolStats( + VmaAllocator allocator, + VmaPool pool, + VmaPoolStats* pPoolStats) +{ + VMA_ASSERT(allocator && pool && pPoolStats); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + allocator->GetPoolStats(pool, pPoolStats); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaMakePoolAllocationsLost( + VmaAllocator allocator, + VmaPool pool, + size_t* pLostAllocationCount) +{ + VMA_ASSERT(allocator && pool); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordMakePoolAllocationsLost(allocator->GetCurrentFrameIndex(), pool); + } +#endif + + allocator->MakePoolAllocationsLost(pool, pLostAllocationCount); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckPoolCorruption(VmaAllocator allocator, VmaPool pool) +{ + VMA_ASSERT(allocator && pool); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + VMA_DEBUG_LOG("vmaCheckPoolCorruption"); + + return allocator->CheckPoolCorruption(pool); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaGetPoolName( + VmaAllocator allocator, + VmaPool pool, + const char** ppName) +{ + VMA_ASSERT(allocator && pool && ppName); + + VMA_DEBUG_LOG("vmaGetPoolName"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + *ppName = pool->GetName(); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaSetPoolName( + VmaAllocator allocator, + VmaPool pool, + const char* pName) +{ + VMA_ASSERT(allocator && pool); + + VMA_DEBUG_LOG("vmaSetPoolName"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + pool->SetName(pName); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordSetPoolName(allocator->GetCurrentFrameIndex(), pool, pName); + } +#endif +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemory( + VmaAllocator allocator, + const VkMemoryRequirements* pVkMemoryRequirements, + const VmaAllocationCreateInfo* pCreateInfo, + VmaAllocation* pAllocation, + VmaAllocationInfo* pAllocationInfo) +{ + VMA_ASSERT(allocator && pVkMemoryRequirements && pCreateInfo && pAllocation); + + VMA_DEBUG_LOG("vmaAllocateMemory"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + VkResult result = allocator->AllocateMemory( + *pVkMemoryRequirements, + false, // requiresDedicatedAllocation + false, // prefersDedicatedAllocation + VK_NULL_HANDLE, // dedicatedBuffer + UINT32_MAX, // dedicatedBufferUsage + VK_NULL_HANDLE, // dedicatedImage + *pCreateInfo, + VMA_SUBALLOCATION_TYPE_UNKNOWN, + 1, // allocationCount + pAllocation); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordAllocateMemory( + allocator->GetCurrentFrameIndex(), + *pVkMemoryRequirements, + *pCreateInfo, + *pAllocation); + } +#endif + + if(pAllocationInfo != VMA_NULL && result == VK_SUCCESS) + { + allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); + } + + return result; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryPages( + VmaAllocator allocator, + const VkMemoryRequirements* pVkMemoryRequirements, + const VmaAllocationCreateInfo* pCreateInfo, + size_t allocationCount, + VmaAllocation* pAllocations, + VmaAllocationInfo* pAllocationInfo) +{ + if(allocationCount == 0) + { + return VK_SUCCESS; + } + + VMA_ASSERT(allocator && pVkMemoryRequirements && pCreateInfo && pAllocations); + + VMA_DEBUG_LOG("vmaAllocateMemoryPages"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + VkResult result = allocator->AllocateMemory( + *pVkMemoryRequirements, + false, // requiresDedicatedAllocation + false, // prefersDedicatedAllocation + VK_NULL_HANDLE, // dedicatedBuffer + UINT32_MAX, // dedicatedBufferUsage + VK_NULL_HANDLE, // dedicatedImage + *pCreateInfo, + VMA_SUBALLOCATION_TYPE_UNKNOWN, + allocationCount, + pAllocations); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordAllocateMemoryPages( + allocator->GetCurrentFrameIndex(), + *pVkMemoryRequirements, + *pCreateInfo, + (uint64_t)allocationCount, + pAllocations); + } +#endif + + if(pAllocationInfo != VMA_NULL && result == VK_SUCCESS) + { + for(size_t i = 0; i < allocationCount; ++i) + { + allocator->GetAllocationInfo(pAllocations[i], pAllocationInfo + i); + } + } + + return result; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryForBuffer( + VmaAllocator allocator, + VkBuffer buffer, + const VmaAllocationCreateInfo* pCreateInfo, + VmaAllocation* pAllocation, + VmaAllocationInfo* pAllocationInfo) +{ + VMA_ASSERT(allocator && buffer != VK_NULL_HANDLE && pCreateInfo && pAllocation); + + VMA_DEBUG_LOG("vmaAllocateMemoryForBuffer"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + VkMemoryRequirements vkMemReq = {}; + bool requiresDedicatedAllocation = false; + bool prefersDedicatedAllocation = false; + allocator->GetBufferMemoryRequirements(buffer, vkMemReq, + requiresDedicatedAllocation, + prefersDedicatedAllocation); + + VkResult result = allocator->AllocateMemory( + vkMemReq, + requiresDedicatedAllocation, + prefersDedicatedAllocation, + buffer, // dedicatedBuffer + UINT32_MAX, // dedicatedBufferUsage + VK_NULL_HANDLE, // dedicatedImage + *pCreateInfo, + VMA_SUBALLOCATION_TYPE_BUFFER, + 1, // allocationCount + pAllocation); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordAllocateMemoryForBuffer( + allocator->GetCurrentFrameIndex(), + vkMemReq, + requiresDedicatedAllocation, + prefersDedicatedAllocation, + *pCreateInfo, + *pAllocation); + } +#endif + + if(pAllocationInfo && result == VK_SUCCESS) + { + allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); + } + + return result; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaAllocateMemoryForImage( + VmaAllocator allocator, + VkImage image, + const VmaAllocationCreateInfo* pCreateInfo, + VmaAllocation* pAllocation, + VmaAllocationInfo* pAllocationInfo) +{ + VMA_ASSERT(allocator && image != VK_NULL_HANDLE && pCreateInfo && pAllocation); + + VMA_DEBUG_LOG("vmaAllocateMemoryForImage"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + VkMemoryRequirements vkMemReq = {}; + bool requiresDedicatedAllocation = false; + bool prefersDedicatedAllocation = false; + allocator->GetImageMemoryRequirements(image, vkMemReq, + requiresDedicatedAllocation, prefersDedicatedAllocation); + + VkResult result = allocator->AllocateMemory( + vkMemReq, + requiresDedicatedAllocation, + prefersDedicatedAllocation, + VK_NULL_HANDLE, // dedicatedBuffer + UINT32_MAX, // dedicatedBufferUsage + image, // dedicatedImage + *pCreateInfo, + VMA_SUBALLOCATION_TYPE_IMAGE_UNKNOWN, + 1, // allocationCount + pAllocation); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordAllocateMemoryForImage( + allocator->GetCurrentFrameIndex(), + vkMemReq, + requiresDedicatedAllocation, + prefersDedicatedAllocation, + *pCreateInfo, + *pAllocation); + } +#endif + + if(pAllocationInfo && result == VK_SUCCESS) + { + allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); + } + + return result; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaFreeMemory( + VmaAllocator allocator, + VmaAllocation allocation) +{ + VMA_ASSERT(allocator); + + if(allocation == VK_NULL_HANDLE) + { + return; + } + + VMA_DEBUG_LOG("vmaFreeMemory"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordFreeMemory( + allocator->GetCurrentFrameIndex(), + allocation); + } +#endif + + allocator->FreeMemory( + 1, // allocationCount + &allocation); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaFreeMemoryPages( + VmaAllocator allocator, + size_t allocationCount, + const VmaAllocation* pAllocations) +{ + if(allocationCount == 0) + { + return; + } + + VMA_ASSERT(allocator); + + VMA_DEBUG_LOG("vmaFreeMemoryPages"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordFreeMemoryPages( + allocator->GetCurrentFrameIndex(), + (uint64_t)allocationCount, + pAllocations); + } +#endif + + allocator->FreeMemory(allocationCount, pAllocations); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaResizeAllocation( + VmaAllocator allocator, + VmaAllocation allocation, + VkDeviceSize newSize) +{ + VMA_ASSERT(allocator && allocation); + + VMA_DEBUG_LOG("vmaResizeAllocation"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + return allocator->ResizeAllocation(allocation, newSize); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaGetAllocationInfo( + VmaAllocator allocator, + VmaAllocation allocation, + VmaAllocationInfo* pAllocationInfo) +{ + VMA_ASSERT(allocator && allocation && pAllocationInfo); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordGetAllocationInfo( + allocator->GetCurrentFrameIndex(), + allocation); + } +#endif + + allocator->GetAllocationInfo(allocation, pAllocationInfo); +} + +VMA_CALL_PRE VkBool32 VMA_CALL_POST vmaTouchAllocation( + VmaAllocator allocator, + VmaAllocation allocation) +{ + VMA_ASSERT(allocator && allocation); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordTouchAllocation( + allocator->GetCurrentFrameIndex(), + allocation); + } +#endif + + return allocator->TouchAllocation(allocation); +} + +VMA_CALL_PRE void VMA_CALL_POST vmaSetAllocationUserData( + VmaAllocator allocator, + VmaAllocation allocation, + void* pUserData) +{ + VMA_ASSERT(allocator && allocation); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + allocation->SetUserData(allocator, pUserData); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordSetAllocationUserData( + allocator->GetCurrentFrameIndex(), + allocation, + pUserData); + } +#endif +} + +VMA_CALL_PRE void VMA_CALL_POST vmaCreateLostAllocation( + VmaAllocator allocator, + VmaAllocation* pAllocation) +{ + VMA_ASSERT(allocator && pAllocation); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK; + + allocator->CreateLostAllocation(pAllocation); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordCreateLostAllocation( + allocator->GetCurrentFrameIndex(), + *pAllocation); + } +#endif +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaMapMemory( + VmaAllocator allocator, + VmaAllocation allocation, + void** ppData) +{ + VMA_ASSERT(allocator && allocation && ppData); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + VkResult res = allocator->Map(allocation, ppData); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordMapMemory( + allocator->GetCurrentFrameIndex(), + allocation); + } +#endif + + return res; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaUnmapMemory( + VmaAllocator allocator, + VmaAllocation allocation) +{ + VMA_ASSERT(allocator && allocation); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordUnmapMemory( + allocator->GetCurrentFrameIndex(), + allocation); + } +#endif + + allocator->Unmap(allocation); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFlushAllocation(VmaAllocator allocator, VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size) +{ + VMA_ASSERT(allocator && allocation); + + VMA_DEBUG_LOG("vmaFlushAllocation"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + const VkResult res = allocator->FlushOrInvalidateAllocation(allocation, offset, size, VMA_CACHE_FLUSH); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordFlushAllocation( + allocator->GetCurrentFrameIndex(), + allocation, offset, size); + } +#endif + + return res; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaInvalidateAllocation(VmaAllocator allocator, VmaAllocation allocation, VkDeviceSize offset, VkDeviceSize size) +{ + VMA_ASSERT(allocator && allocation); + + VMA_DEBUG_LOG("vmaInvalidateAllocation"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + const VkResult res = allocator->FlushOrInvalidateAllocation(allocation, offset, size, VMA_CACHE_INVALIDATE); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordInvalidateAllocation( + allocator->GetCurrentFrameIndex(), + allocation, offset, size); + } +#endif + + return res; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaFlushAllocations( + VmaAllocator allocator, + uint32_t allocationCount, + const VmaAllocation* allocations, + const VkDeviceSize* offsets, + const VkDeviceSize* sizes) +{ + VMA_ASSERT(allocator); + + if(allocationCount == 0) + { + return VK_SUCCESS; + } + + VMA_ASSERT(allocations); + + VMA_DEBUG_LOG("vmaFlushAllocations"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + const VkResult res = allocator->FlushOrInvalidateAllocations(allocationCount, allocations, offsets, sizes, VMA_CACHE_FLUSH); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + //TODO + } +#endif + + return res; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaInvalidateAllocations( + VmaAllocator allocator, + uint32_t allocationCount, + const VmaAllocation* allocations, + const VkDeviceSize* offsets, + const VkDeviceSize* sizes) +{ + VMA_ASSERT(allocator); + + if(allocationCount == 0) + { + return VK_SUCCESS; + } + + VMA_ASSERT(allocations); + + VMA_DEBUG_LOG("vmaInvalidateAllocations"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + const VkResult res = allocator->FlushOrInvalidateAllocations(allocationCount, allocations, offsets, sizes, VMA_CACHE_INVALIDATE); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + //TODO + } +#endif + + return res; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCheckCorruption(VmaAllocator allocator, uint32_t memoryTypeBits) +{ + VMA_ASSERT(allocator); + + VMA_DEBUG_LOG("vmaCheckCorruption"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + return allocator->CheckCorruption(memoryTypeBits); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragment( + VmaAllocator allocator, + const VmaAllocation* pAllocations, + size_t allocationCount, + VkBool32* pAllocationsChanged, + const VmaDefragmentationInfo *pDefragmentationInfo, + VmaDefragmentationStats* pDefragmentationStats) +{ + // Deprecated interface, reimplemented using new one. + + VmaDefragmentationInfo2 info2 = {}; + info2.allocationCount = (uint32_t)allocationCount; + info2.pAllocations = pAllocations; + info2.pAllocationsChanged = pAllocationsChanged; + if(pDefragmentationInfo != VMA_NULL) + { + info2.maxCpuAllocationsToMove = pDefragmentationInfo->maxAllocationsToMove; + info2.maxCpuBytesToMove = pDefragmentationInfo->maxBytesToMove; + } + else + { + info2.maxCpuAllocationsToMove = UINT32_MAX; + info2.maxCpuBytesToMove = VK_WHOLE_SIZE; + } + // info2.flags, maxGpuAllocationsToMove, maxGpuBytesToMove, commandBuffer deliberately left zero. + + VmaDefragmentationContext ctx; + VkResult res = vmaDefragmentationBegin(allocator, &info2, pDefragmentationStats, &ctx); + if(res == VK_NOT_READY) + { + res = vmaDefragmentationEnd( allocator, ctx); + } + return res; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragmentationBegin( + VmaAllocator allocator, + const VmaDefragmentationInfo2* pInfo, + VmaDefragmentationStats* pStats, + VmaDefragmentationContext *pContext) +{ + VMA_ASSERT(allocator && pInfo && pContext); + + // Degenerate case: Nothing to defragment. + if(pInfo->allocationCount == 0 && pInfo->poolCount == 0) + { + return VK_SUCCESS; + } + + VMA_ASSERT(pInfo->allocationCount == 0 || pInfo->pAllocations != VMA_NULL); + VMA_ASSERT(pInfo->poolCount == 0 || pInfo->pPools != VMA_NULL); + VMA_HEAVY_ASSERT(VmaValidatePointerArray(pInfo->allocationCount, pInfo->pAllocations)); + VMA_HEAVY_ASSERT(VmaValidatePointerArray(pInfo->poolCount, pInfo->pPools)); + + VMA_DEBUG_LOG("vmaDefragmentationBegin"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + VkResult res = allocator->DefragmentationBegin(*pInfo, pStats, pContext); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordDefragmentationBegin( + allocator->GetCurrentFrameIndex(), *pInfo, *pContext); + } +#endif + + return res; +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaDefragmentationEnd( + VmaAllocator allocator, + VmaDefragmentationContext context) +{ + VMA_ASSERT(allocator); + + VMA_DEBUG_LOG("vmaDefragmentationEnd"); + + if(context != VK_NULL_HANDLE) + { + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordDefragmentationEnd( + allocator->GetCurrentFrameIndex(), context); + } +#endif + + return allocator->DefragmentationEnd(context); + } + else + { + return VK_SUCCESS; + } +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBeginDefragmentationPass( + VmaAllocator allocator, + VmaDefragmentationContext context, + VmaDefragmentationPassInfo* pInfo + ) +{ + VMA_ASSERT(allocator); + VMA_ASSERT(pInfo); + + VMA_DEBUG_LOG("vmaBeginDefragmentationPass"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + if(context == VK_NULL_HANDLE) + { + pInfo->moveCount = 0; + return VK_SUCCESS; + } + + return allocator->DefragmentationPassBegin(pInfo, context); +} +VMA_CALL_PRE VkResult VMA_CALL_POST vmaEndDefragmentationPass( + VmaAllocator allocator, + VmaDefragmentationContext context) +{ + VMA_ASSERT(allocator); + + VMA_DEBUG_LOG("vmaEndDefragmentationPass"); + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + if(context == VK_NULL_HANDLE) + return VK_SUCCESS; + + return allocator->DefragmentationPassEnd(context); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindBufferMemory( + VmaAllocator allocator, + VmaAllocation allocation, + VkBuffer buffer) +{ + VMA_ASSERT(allocator && allocation && buffer); + + VMA_DEBUG_LOG("vmaBindBufferMemory"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + return allocator->BindBufferMemory(allocation, 0, buffer, VMA_NULL); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindBufferMemory2( + VmaAllocator allocator, + VmaAllocation allocation, + VkDeviceSize allocationLocalOffset, + VkBuffer buffer, + const void* pNext) +{ + VMA_ASSERT(allocator && allocation && buffer); + + VMA_DEBUG_LOG("vmaBindBufferMemory2"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + return allocator->BindBufferMemory(allocation, allocationLocalOffset, buffer, pNext); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindImageMemory( + VmaAllocator allocator, + VmaAllocation allocation, + VkImage image) +{ + VMA_ASSERT(allocator && allocation && image); + + VMA_DEBUG_LOG("vmaBindImageMemory"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + return allocator->BindImageMemory(allocation, 0, image, VMA_NULL); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaBindImageMemory2( + VmaAllocator allocator, + VmaAllocation allocation, + VkDeviceSize allocationLocalOffset, + VkImage image, + const void* pNext) +{ + VMA_ASSERT(allocator && allocation && image); + + VMA_DEBUG_LOG("vmaBindImageMemory2"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + return allocator->BindImageMemory(allocation, allocationLocalOffset, image, pNext); +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateBuffer( + VmaAllocator allocator, + const VkBufferCreateInfo* pBufferCreateInfo, + const VmaAllocationCreateInfo* pAllocationCreateInfo, + VkBuffer* pBuffer, + VmaAllocation* pAllocation, + VmaAllocationInfo* pAllocationInfo) +{ + VMA_ASSERT(allocator && pBufferCreateInfo && pAllocationCreateInfo && pBuffer && pAllocation); + + if(pBufferCreateInfo->size == 0) + { + return VK_ERROR_VALIDATION_FAILED_EXT; + } + if((pBufferCreateInfo->usage & VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT_COPY) != 0 && + !allocator->m_UseKhrBufferDeviceAddress) + { + VMA_ASSERT(0 && "Creating a buffer with VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT is not valid if VMA_ALLOCATOR_CREATE_BUFFER_DEVICE_ADDRESS_BIT was not used."); + return VK_ERROR_VALIDATION_FAILED_EXT; + } + + VMA_DEBUG_LOG("vmaCreateBuffer"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + *pBuffer = VK_NULL_HANDLE; + *pAllocation = VK_NULL_HANDLE; + + // 1. Create VkBuffer. + VkResult res = (*allocator->GetVulkanFunctions().vkCreateBuffer)( + allocator->m_hDevice, + pBufferCreateInfo, + allocator->GetAllocationCallbacks(), + pBuffer); + if(res >= 0) + { + // 2. vkGetBufferMemoryRequirements. + VkMemoryRequirements vkMemReq = {}; + bool requiresDedicatedAllocation = false; + bool prefersDedicatedAllocation = false; + allocator->GetBufferMemoryRequirements(*pBuffer, vkMemReq, + requiresDedicatedAllocation, prefersDedicatedAllocation); + + // 3. Allocate memory using allocator. + res = allocator->AllocateMemory( + vkMemReq, + requiresDedicatedAllocation, + prefersDedicatedAllocation, + *pBuffer, // dedicatedBuffer + pBufferCreateInfo->usage, // dedicatedBufferUsage + VK_NULL_HANDLE, // dedicatedImage + *pAllocationCreateInfo, + VMA_SUBALLOCATION_TYPE_BUFFER, + 1, // allocationCount + pAllocation); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordCreateBuffer( + allocator->GetCurrentFrameIndex(), + *pBufferCreateInfo, + *pAllocationCreateInfo, + *pAllocation); + } +#endif + + if(res >= 0) + { + // 3. Bind buffer with memory. + if((pAllocationCreateInfo->flags & VMA_ALLOCATION_CREATE_DONT_BIND_BIT) == 0) + { + res = allocator->BindBufferMemory(*pAllocation, 0, *pBuffer, VMA_NULL); + } + if(res >= 0) + { + // All steps succeeded. + #if VMA_STATS_STRING_ENABLED + (*pAllocation)->InitBufferImageUsage(pBufferCreateInfo->usage); + #endif + if(pAllocationInfo != VMA_NULL) + { + allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); + } + + return VK_SUCCESS; + } + allocator->FreeMemory( + 1, // allocationCount + pAllocation); + *pAllocation = VK_NULL_HANDLE; + (*allocator->GetVulkanFunctions().vkDestroyBuffer)(allocator->m_hDevice, *pBuffer, allocator->GetAllocationCallbacks()); + *pBuffer = VK_NULL_HANDLE; + return res; + } + (*allocator->GetVulkanFunctions().vkDestroyBuffer)(allocator->m_hDevice, *pBuffer, allocator->GetAllocationCallbacks()); + *pBuffer = VK_NULL_HANDLE; + return res; + } + return res; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyBuffer( + VmaAllocator allocator, + VkBuffer buffer, + VmaAllocation allocation) +{ + VMA_ASSERT(allocator); + + if(buffer == VK_NULL_HANDLE && allocation == VK_NULL_HANDLE) + { + return; + } + + VMA_DEBUG_LOG("vmaDestroyBuffer"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordDestroyBuffer( + allocator->GetCurrentFrameIndex(), + allocation); + } +#endif + + if(buffer != VK_NULL_HANDLE) + { + (*allocator->GetVulkanFunctions().vkDestroyBuffer)(allocator->m_hDevice, buffer, allocator->GetAllocationCallbacks()); + } + + if(allocation != VK_NULL_HANDLE) + { + allocator->FreeMemory( + 1, // allocationCount + &allocation); + } +} + +VMA_CALL_PRE VkResult VMA_CALL_POST vmaCreateImage( + VmaAllocator allocator, + const VkImageCreateInfo* pImageCreateInfo, + const VmaAllocationCreateInfo* pAllocationCreateInfo, + VkImage* pImage, + VmaAllocation* pAllocation, + VmaAllocationInfo* pAllocationInfo) +{ + VMA_ASSERT(allocator && pImageCreateInfo && pAllocationCreateInfo && pImage && pAllocation); + + if(pImageCreateInfo->extent.width == 0 || + pImageCreateInfo->extent.height == 0 || + pImageCreateInfo->extent.depth == 0 || + pImageCreateInfo->mipLevels == 0 || + pImageCreateInfo->arrayLayers == 0) + { + return VK_ERROR_VALIDATION_FAILED_EXT; + } + + VMA_DEBUG_LOG("vmaCreateImage"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + + *pImage = VK_NULL_HANDLE; + *pAllocation = VK_NULL_HANDLE; + + // 1. Create VkImage. + VkResult res = (*allocator->GetVulkanFunctions().vkCreateImage)( + allocator->m_hDevice, + pImageCreateInfo, + allocator->GetAllocationCallbacks(), + pImage); + if(res >= 0) + { + VmaSuballocationType suballocType = pImageCreateInfo->tiling == VK_IMAGE_TILING_OPTIMAL ? + VMA_SUBALLOCATION_TYPE_IMAGE_OPTIMAL : + VMA_SUBALLOCATION_TYPE_IMAGE_LINEAR; + + // 2. Allocate memory using allocator. + VkMemoryRequirements vkMemReq = {}; + bool requiresDedicatedAllocation = false; + bool prefersDedicatedAllocation = false; + allocator->GetImageMemoryRequirements(*pImage, vkMemReq, + requiresDedicatedAllocation, prefersDedicatedAllocation); + + res = allocator->AllocateMemory( + vkMemReq, + requiresDedicatedAllocation, + prefersDedicatedAllocation, + VK_NULL_HANDLE, // dedicatedBuffer + UINT32_MAX, // dedicatedBufferUsage + *pImage, // dedicatedImage + *pAllocationCreateInfo, + suballocType, + 1, // allocationCount + pAllocation); + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordCreateImage( + allocator->GetCurrentFrameIndex(), + *pImageCreateInfo, + *pAllocationCreateInfo, + *pAllocation); + } +#endif + + if(res >= 0) + { + // 3. Bind image with memory. + if((pAllocationCreateInfo->flags & VMA_ALLOCATION_CREATE_DONT_BIND_BIT) == 0) + { + res = allocator->BindImageMemory(*pAllocation, 0, *pImage, VMA_NULL); + } + if(res >= 0) + { + // All steps succeeded. + #if VMA_STATS_STRING_ENABLED + (*pAllocation)->InitBufferImageUsage(pImageCreateInfo->usage); + #endif + if(pAllocationInfo != VMA_NULL) + { + allocator->GetAllocationInfo(*pAllocation, pAllocationInfo); + } + + return VK_SUCCESS; + } + allocator->FreeMemory( + 1, // allocationCount + pAllocation); + *pAllocation = VK_NULL_HANDLE; + (*allocator->GetVulkanFunctions().vkDestroyImage)(allocator->m_hDevice, *pImage, allocator->GetAllocationCallbacks()); + *pImage = VK_NULL_HANDLE; + return res; + } + (*allocator->GetVulkanFunctions().vkDestroyImage)(allocator->m_hDevice, *pImage, allocator->GetAllocationCallbacks()); + *pImage = VK_NULL_HANDLE; + return res; + } + return res; +} + +VMA_CALL_PRE void VMA_CALL_POST vmaDestroyImage( + VmaAllocator allocator, + VkImage image, + VmaAllocation allocation) +{ + VMA_ASSERT(allocator); + + if(image == VK_NULL_HANDLE && allocation == VK_NULL_HANDLE) + { + return; + } + + VMA_DEBUG_LOG("vmaDestroyImage"); + + VMA_DEBUG_GLOBAL_MUTEX_LOCK + +#if VMA_RECORDING_ENABLED + if(allocator->GetRecorder() != VMA_NULL) + { + allocator->GetRecorder()->RecordDestroyImage( + allocator->GetCurrentFrameIndex(), + allocation); + } +#endif + + if(image != VK_NULL_HANDLE) + { + (*allocator->GetVulkanFunctions().vkDestroyImage)(allocator->m_hDevice, image, allocator->GetAllocationCallbacks()); + } + if(allocation != VK_NULL_HANDLE) + { + allocator->FreeMemory( + 1, // allocationCount + &allocation); + } +} + +#endif // #ifdef VMA_IMPLEMENTATION -- cgit v1.3