diff options
Diffstat (limited to 'main.c')
| -rw-r--r-- | main.c | 384 |
1 files changed, 384 insertions, 0 deletions
@@ -0,0 +1,384 @@ +#include <Windows.h> +#include <emmintrin.h> +#include <smmintrin.h> + +int _fltused; + +BOOL g_is_open = FALSE; +LONG g_width = 640; +LONG g_height = 480; +void* g_pixels = NULL; +BITMAPINFO bmp_info; +UINT64 avg = 0U; +UINT64 avg_fs = 0U; +UINT64 avg_simd = 0U; +UINT64 frame_count = 0U; +__m128i g_zero; +__m128i g_one; + +typedef struct aos_tri { + UINT32 p1[2]; + UINT32 p2[2]; + UINT32 p3[2]; +} aos_tri; + +typedef struct soa_tri { + UINT32 x[3]; + UINT32 y[3]; +} soa_tri; + +aos_tri aos_triangles[] = { + { + .p1 = { 128, 384 }, + .p2 = { 320, 96 }, + .p3 = { 512, 384 }, + }, +}; +soa_tri soa_triangles[] = { + { + .x = { 128, 320, 512 }, + .y = { 384, 96, 384 }, + }, +}; + +void RenderFullscren(void) { + DWORD64 start = __rdtsc(); + + UINT8* pixels = (UINT8*)g_pixels; + aos_tri t = aos_triangles[0]; + UINT32 max_x = max(t.p1[0], max(t.p2[0], t.p3[0])); + UINT32 min_x = min(t.p1[0], min(t.p2[0], t.p3[0])); + UINT32 max_y = max(t.p1[1], max(t.p2[1], t.p3[1])); + UINT32 min_y = min(t.p1[1], min(t.p2[1], t.p3[1])); + + INT32 v1[2] = { t.p2[0] - t.p1[0], t.p2[1] - t.p1[1] }; + INT32 v2[2] = { t.p3[0] - t.p2[0], t.p3[1] - t.p2[1] }; + INT32 v3[2] = { t.p1[0] - t.p3[0], t.p1[1] - t.p3[1] }; + + for (INT32 i = 0; i < g_width * g_height; i++) { + UINT32 x = i % g_width; + UINT32 y = i / g_width; + if (min_x <= x && x <= max_x && + min_y <= y && y <= max_y) { + + INT32 t1[2] = { x - t.p1[0], y - t.p1[1] }; + INT32 t2[2] = { x - t.p2[0], y - t.p2[1] }; + INT32 t3[2] = { x - t.p3[0], y - t.p3[1] }; + + INT32 d_t1 = v1[0] * t1[1] - t1[0] * v1[1]; + INT32 d_t2 = v2[0] * t2[1] - t2[0] * v2[1]; + INT32 d_t3 = v3[0] * t3[1] - t3[0] * v3[1]; + + if (d_t1 > 0.f && d_t2 > 0.f && d_t3 > 0.f) { + pixels[i * 4] = 255; + } + } + } + + DWORD64 end = __rdtsc(); + DWORD64 diff = end - start; + + avg_fs = ((avg_fs * frame_count) + diff) / (frame_count + 1); + + char buffer[1024]; + wsprintfA(buffer, "fullscreen scan %u cycles per triangle\n", diff); + OutputDebugStringA(buffer); + wsprintfA(buffer, "fullscreen scan %u avg cycles per triangle\n", avg_fs); + OutputDebugStringA(buffer); +} + +void RenderTriangle(void) { + DWORD64 start = __rdtsc(); + + UINT8* pixels = (UINT8*)g_pixels; + aos_tri t = aos_triangles[0]; + UINT32 max_x = max(t.p1[0], max(t.p2[0], t.p3[0])); + UINT32 min_x = min(t.p1[0], min(t.p2[0], t.p3[0])); + UINT32 max_y = max(t.p1[1], max(t.p2[1], t.p3[1])); + UINT32 min_y = min(t.p1[1], min(t.p2[1], t.p3[1])); + + INT32 v1[2] = { t.p2[0] - t.p1[0], t.p2[1] - t.p1[1] }; + INT32 v2[2] = { t.p3[0] - t.p2[0], t.p3[1] - t.p2[1] }; + INT32 v3[2] = { t.p1[0] - t.p3[0], t.p1[1] - t.p3[1] }; + + for (UINT32 y = min_y; y < max_y + 1; y++) { + for (UINT32 x = min_x; x < max_x + 1; x++) { + + INT32 t1[2] = { x - t.p1[0], y - t.p1[1] }; + INT32 t2[2] = { x - t.p2[0], y - t.p2[1] }; + INT32 t3[2] = { x - t.p3[0], y - t.p3[1] }; + + INT32 d_t1 = v1[0] * t1[1] - t1[0] * v1[1]; + INT32 d_t2 = v2[0] * t2[1] - t2[0] * v2[1]; + INT32 d_t3 = v3[0] * t3[1] - t3[0] * v3[1]; + + if (d_t1 > 0.f && d_t2 > 0.f && d_t3 > 0.f) { + pixels[((y * g_width) + x) * 4 + 1] = 255; + } + } + } + + DWORD64 end = __rdtsc(); + DWORD64 diff = end - start; + + // avg = (avg + diff) / 2; + avg = ((avg * frame_count) + diff) / (frame_count + 1); + + char buffer[1024]; + wsprintfA(buffer, "%u cycles per triangle\n", diff); + OutputDebugStringA(buffer); + wsprintfA(buffer, "%u avg cycles per triangle\n", avg); + OutputDebugStringA(buffer); +} + +void RenderTriangleSIMD(void) { + DWORD64 start = __rdtsc(); + + UINT64* pixels = (UINT64*)g_pixels; + soa_tri t = soa_triangles[0]; + UINT32 max_x = max(t.x[0], max(t.x[1], t.x[2])); + UINT32 min_x = min(t.x[0], min(t.x[1], t.x[2])); + UINT32 max_y = max(t.y[0], max(t.y[1], t.y[2])); + UINT32 min_y = min(t.y[0], min(t.y[1], t.y[2])); + + max_x += max_x % 2; + min_x -= min_x % 2; + max_y += max_y % 2; + min_y -= min_y % 2; + + __m128i p1_x = _mm_set1_epi32(t.x[0]); + __m128i p1_y = _mm_set1_epi32(t.y[0]); + __m128i p2_x = _mm_set1_epi32(t.x[1]); + __m128i p2_y = _mm_set1_epi32(t.y[1]); + __m128i p3_x = _mm_set1_epi32(t.x[2]); + __m128i p3_y = _mm_set1_epi32(t.y[2]); + + __m128i v1_x = _mm_set1_epi32(t.x[1] - t.x[0]); + __m128i v1_y = _mm_set1_epi32(t.y[1] - t.y[0]); + __m128i v2_x = _mm_set1_epi32(t.x[2] - t.x[1]); + __m128i v2_y = _mm_set1_epi32(t.y[2] - t.y[1]); + __m128i v3_x = _mm_set1_epi32(t.x[0] - t.x[2]); + __m128i v3_y = _mm_set1_epi32(t.y[0] - t.y[2]); + + // min and max must be multiples of 2 + for (UINT32 y = min_y; y < max_y + 1; y+=2) { + UINT32 row_1 = y * g_width; + UINT32 row_2 = (y + 1) * g_width; + for (UINT32 x = min_x; x < max_x + 1; x+=2) { + // work on 2x2 + // _________ + // | | | + // | 1 | 2 | + // |___|___| + // | | | + // | 3 | 4 | + // |___|___| + // 1 + // { + // INT32 t1[2] = { x - tt.p1[0], y - tt.p1[1] }; + // INT32 t2[2] = { x - tt.p2[0], y - tt.p2[1] }; + // INT32 t3[2] = { x - tt.p3[0], y - tt.p3[1] }; + + // INT32 dt_t1 = v1[0] * t1[1] - t1[0] * v1[1]; + // INT32 dt_t2 = v2[0] * t2[1] - t2[0] * v2[1]; + // INT32 dt_t3 = v3[0] * t3[1] - t3[0] * v3[1]; + + // if (dt_t1 > 0.f && dt_t2 > 0.f && dt_t3 > 0.f) { + // pixels[((y * g_width) + x) * 4 + 2] = 255; + // } + // } + // // 2 + // { + // INT32 t1[2] = { x + 1 - t.p1[0], y - t.p1[1] }; + // INT32 t2[2] = { x + 1 - t.p2[0], y - t.p2[1] }; + // INT32 t3[2] = { x + 1 - t.p3[0], y - t.p3[1] }; + + // INT32 d_t1 = v1[0] * t1[1] - t1[0] * v1[1]; + // INT32 d_t2 = v2[0] * t2[1] - t2[0] * v2[1]; + // INT32 d_t3 = v3[0] * t3[1] - t3[0] * v3[1]; + + // if (d_t1 > 0.f && d_t2 > 0.f && d_t3 > 0.f) { + // pixels[((y * g_width) + x + 1) * 4 + 2] = 255; + // } + // } + // // 3 + // { + // INT32 t1[2] = { x - t.p1[0], y + 1 - t.p1[1] }; + // INT32 t2[2] = { x - t.p2[0], y + 1 - t.p2[1] }; + // INT32 t3[2] = { x - t.p3[0], y + 1 - t.p3[1] }; + + // INT32 d_t1 = v1[0] * t1[1] - t1[0] * v1[1]; + // INT32 d_t2 = v2[0] * t2[1] - t2[0] * v2[1]; + // INT32 d_t3 = v3[0] * t3[1] - t3[0] * v3[1]; + + // if (d_t1 > 0.f && d_t2 > 0.f && d_t3 > 0.f) { + // pixels[(((y + 1) * g_width) + x) * 4 + 2] = 255; + // } + // } + // // 4 + // { + // INT32 t1[2] = { x + 1 - t.p1[0], y + 1 - t.p1[1] }; + // INT32 t2[2] = { x + 1 - t.p2[0], y + 1 - t.p2[1] }; + // INT32 t3[2] = { x + 1 - t.p3[0], y + 1 - t.p3[1] }; + + // INT32 d_t1 = v1[0] * t1[1] - t1[0] * v1[1]; + // INT32 d_t2 = v2[0] * t2[1] - t2[0] * v2[1]; + // INT32 d_t3 = v3[0] * t3[1] - t3[0] * v3[1]; + + // if (d_t1 > 0.f && d_t2 > 0.f && d_t3 > 0.f) { + // pixels[(((y + 1) * g_width) + x + 1) * 4 + 2] = 255; + // } + // } + + __m128i vx = _mm_set_epi32(x + 1, x, x + 1, x); + __m128i vy = _mm_set_epi32(y + 1, y + 1, y, y); + + __m128i p_t1x = _mm_sub_epi32(vx, p1_x); + __m128i p_t1y = _mm_sub_epi32(vy, p1_y); + __m128i p_t2x = _mm_sub_epi32(vx, p2_x); + __m128i p_t2y = _mm_sub_epi32(vy, p2_y); + __m128i p_t3x = _mm_sub_epi32(vx, p3_x); + __m128i p_t3y = _mm_sub_epi32(vy, p3_y); + + __m128i d_t1_1 = _mm_mullo_epi32(v1_x, p_t1y); + __m128i d_t1_2 = _mm_mullo_epi32(v1_y, p_t1x); + __m128i d_t2_1 = _mm_mullo_epi32(v2_x, p_t2y); + __m128i d_t2_2 = _mm_mullo_epi32(v2_y, p_t2x); + __m128i d_t3_1 = _mm_mullo_epi32(v3_x, p_t3y); + __m128i d_t3_2 = _mm_mullo_epi32(v3_y, p_t3x); + + __m128i d_t1 = _mm_sub_epi32(d_t1_1, d_t1_2); + __m128i d_t2 = _mm_sub_epi32(d_t2_1, d_t2_2); + __m128i d_t3 = _mm_sub_epi32(d_t3_1, d_t3_2); + + __m128i t1_gtz = _mm_cmpgt_epi32(d_t1, g_zero); + __m128i t2_gtz = _mm_cmpgt_epi32(d_t2, g_zero); + __m128i t1_and_t2 = _mm_and_si128(t1_gtz, t2_gtz); + __m128i t3_gtz = _mm_cmpgt_epi32(d_t3, g_zero); + + __m128i all_gtz = _mm_and_si128(t1_and_t2, t3_gtz); + + pixels[(row_1 + x) / 2] = _mm_extract_epi64(all_gtz, 0); + pixels[(row_2 + x) / 2] = _mm_extract_epi64(all_gtz, 1); + } + } + + DWORD64 end = __rdtsc(); + DWORD64 diff = end - start; + + // avg_simd = (avg_simd + diff) / 2; + avg_simd = ((avg_simd * frame_count) + diff) / (frame_count + 1); + + char buffer[1024]; + wsprintfA(buffer, "SIMD %u cycles per triangle\n", diff); + OutputDebugStringA(buffer); + wsprintfA(buffer, "SIMD %u avg cycles per triangle\n", avg_simd); + OutputDebugStringA(buffer); +} + +void ResizeBackBuffer() { + if (g_pixels) { + VirtualFree(g_pixels, 0, MEM_RELEASE); + } + + bmp_info.bmiHeader.biSize = sizeof(BITMAPINFOHEADER); + bmp_info.bmiHeader.biWidth = g_width; + bmp_info.bmiHeader.biHeight = -g_height; + bmp_info.bmiHeader.biPlanes = 1; + bmp_info.bmiHeader.biBitCount = 32; + bmp_info.bmiHeader.biCompression = BI_RGB; + + g_pixels = VirtualAlloc(NULL, g_width * g_height * 4, MEM_COMMIT, PAGE_READWRITE); +} + +LRESULT CALLBACK WndProc(HWND window, UINT msg, WPARAM wparam, LPARAM lparam) { + switch(msg) { + case WM_CREATE: { + g_is_open = TRUE; + + RECT rect; + GetWindowRect(window, &rect); + + g_width = rect.right - rect.left; + g_height = rect.bottom - rect.top; + + ResizeBackBuffer(); + break; + } + case WM_SIZE: { + g_width = LOWORD(lparam); + g_height = HIWORD(lparam); + + ResizeBackBuffer(); + break; + } + case WM_CLOSE: { + g_is_open = FALSE; + PostQuitMessage(0); + break; + } + case WM_DESTROY: { + DestroyWindow(window); + break; + } + case WM_PAINT: { + RenderFullscren(); + RenderTriangle(); + RenderTriangleSIMD(); + frame_count++; + + PAINTSTRUCT paint; + HDC hdc = BeginPaint(window, &paint); + + StretchDIBits( + hdc, + 0, 0, g_width, g_height, + 0, 0, g_width, g_height, + g_pixels, + &bmp_info, + DIB_RGB_COLORS, + SRCCOPY + ); + + EndPaint(window, &paint); + break; + } + default: + break; + } + + return DefWindowProcA(window, msg, wparam, lparam); +} + +int mainCRTStartup(void) { + const char* class_name = "sftw_rst_class"; + WNDCLASS wnd_class = { + .style = CS_HREDRAW | CS_VREDRAW, + .lpfnWndProc = &WndProc, + .hInstance = NULL, + .lpszClassName = class_name, + }; + RegisterClass(&wnd_class); + + HWND wnd = CreateWindowA(class_name, "Software Rasterizer", WS_OVERLAPPEDWINDOW, 10, 10, g_width, g_height, NULL, NULL, NULL, NULL); + + ShowWindow(wnd, SW_SHOW); + + g_zero = _mm_set1_epi32(-1); + g_one = _mm_set1_epi32(255); + + while(g_is_open) { + MSG msg; + while(PeekMessage(&msg, NULL, 0, 0, PM_REMOVE)) { + TranslateMessage(&msg); + DispatchMessage(&msg); + } + + RenderFullscren(); + RenderTriangle(); + RenderTriangleSIMD(); + frame_count++; + } + + return 0; +} |