Working on measurements

build.bat

@@ -1,5 +1,5 @@
 @echo off
 
 pushd %~dp0
-clang main.cpp -O2 -mfma -mavx2 -Wall -Wno-unused-function -Wno-missing-braces -fno-exceptions -fdiagnostics-absolute-paths -I".." -g -o main.exe -Wl,user32.lib
+clang main.cpp -mfma -mavx2 -Wall -Wno-unused-function -Wno-missing-braces -fno-exceptions -fdiagnostics-absolute-paths -I".." -g -o main.exe -Wl,user32.lib
 popd

main.cpp

@@ -80,9 +80,9 @@
 
 // #include "obj_dump.cpp"
 #include "multimedia.cpp"
-#include "profile.cpp"
 #include "obj.cpp"
 #include "vec.cpp"
+#define PROFILE_SCOPE(x)
 
 struct Vertex {
   Vec3 pos;
@@ -281,7 +281,8 @@ F32 edge_function(Vec4 vecp0, Vec4 vecp1, Vec4 p) {
 #define S32x8 __m256i
 
 U64 filled_pixel_count;
-U64 filled_pixel_total_time;
+U64 filled_pixel_cycles;
+U64 triangle_count;
 #include "optimization_log.cpp"
 
 function
@@ -291,7 +292,7 @@ void draw_triangle_nearest(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 lig
                            Vec3 norm0, Vec3 norm1, Vec3 norm2) {
   if(src->pixels == 0) return;
 
-  PROFILE_SCOPE(draw_triangle);
+  U64 fill_pixels_begin = __rdtsc();
 
   F32 min_x1 = (F32)(min(p0.x, min(p1.x, p2.x)));
   F32 min_y1 = (F32)(min(p0.y, min(p1.y, p2.y)));
@@ -362,7 +363,6 @@ void draw_triangle_nearest(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 lig
   F32x8 p2_x = _mm256_set1_ps(p2.x);
   F32x8 p2_y = _mm256_set1_ps(p2.y);
 
-  U64 fill_pixels_begin = __rdtsc();
   for (S64 y = min_y; y < max_y; y++) {
     F32x8 Y = _mm256_set1_ps(y);
     for (S64 x8 = min_x; x8 < max_x; x8+=8) {
@@ -562,9 +562,8 @@ void draw_triangle_nearest(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 lig
     }
     destination += dst->x;
   }
-  U64 end_time = __rdtsc();
 
-  filled_pixel_total_time += end_time - fill_pixels_begin;
+  filled_pixel_cycles += __rdtsc() - fill_pixels_begin;
   filled_pixel_count  += (max_x - min_x)*(max_y - min_y);
 }
 
@@ -708,9 +707,11 @@ void draw_mesh(Render *r, String scene_name, Obj_Material *materials, Obj_Mesh *
       }
 
 
-      draw_triangle_nearest(&r->screen320, r->depth320, image, light_direction, in[0].pos, in[1].pos, in[2].pos, in[0].tex, in[1].tex, in[2].tex, in[0].norm, in[1].norm, in[2].norm);
+      triangle_count++;
+      draw_triangle_nearest_b(&r->screen320, r->depth320, image, light_direction, in[0].pos, in[1].pos, in[2].pos, in[0].tex, in[1].tex, in[2].tex, in[0].norm, in[1].norm, in[2].norm);
       if (in_count > 3) {
-        draw_triangle_nearest(&r->screen320, r->depth320, image, light_direction, in[0].pos, in[2].pos, in[3].pos, in[0].tex, in[2].tex, in[3].tex, in[0].norm, in[2].norm, in[3].norm);
+        triangle_count++;
+        draw_triangle_nearest_b(&r->screen320, r->depth320, image, light_direction, in[0].pos, in[2].pos, in[3].pos, in[0].tex, in[2].tex, in[3].tex, in[0].norm, in[2].norm, in[3].norm);
       }
     }
   }
@@ -1106,7 +1107,6 @@ main(int argc, char **argv) {
   thread_ctx.log_proc = windows_log;
   fprintf(global_file, "\n---------------------");
 
-
   os.window_size.x = 1280;
   os.window_size.y = 720;
   os.window_resizable = 1;
@@ -1209,23 +1209,14 @@ main(int argc, char **argv) {
 
     ui_end_frame(os.screen, &ui, &font);
     frame_data = string_fmt(os.frame_arena, "FPS:%f dt:%f frame:%u camera_pos: %f %f %f camera_yaw: %f %f"
-        "\nAvg_Time: %llu Time:%llu Count:%llu",
+        "\nCycle per pixel: %llu Cycles:%llu Pixels:%llu Triangles:%llu",
         os.fps, os.delta_time*1000, os.frame, r.camera_pos.x, r.camera_pos.y, r.camera_pos.z, r.camera_yaw.x, r.camera_yaw.y,
-        filled_pixel_total_time/filled_pixel_count, filled_pixel_total_time, filled_pixel_count);
+        filled_pixel_cycles/filled_pixel_count, filled_pixel_cycles, filled_pixel_count, triangle_count);
 
+    filled_pixel_count = 0;
+    filled_pixel_cycles = 0;
+    triangle_count = 0;
 
-    for(int i = 0; i < ProfileScopeName_Count; i++){
-      auto *scope = &profile_scopes[i];
-      if(scope->i == 0) continue;
-
-      U64 total = 0;
-      for(int i = 0; i < scope->i; i++){
-        total += scope->samples[i];
-      }
-
-      log_info("\n%s :: Total: %llu Hits: %llu, Avg: %llu", profile_scope_names[i], total, (U64)scope->i, total / scope->i);
-      scope->i = 0;
-    }
   }
 }
 

optimization_log.cpp

@@ -106,8 +106,9 @@ void draw_triangle_nearest_b(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
                            Vec4 p0,   Vec4 p1,   Vec4 p2,
                            Vec2 tex0, Vec2 tex1, Vec2 tex2,
                            Vec3 norm0, Vec3 norm1, Vec3 norm2) {
-  // if(os.frame > 10) PROFILE_BEGIN(draw_triangle);
+  if(src->pixels == 0) return;
-  // ZoneScopedN("draw_triangle");
+  U64 fill_pixels_begin = __rdtsc();
+
   F32 min_x1 = (F32)(min(p0.x, min(p1.x, p2.x)));
   F32 min_y1 = (F32)(min(p0.y, min(p1.y, p2.y)));
   F32 max_x1 = (F32)(max(p0.x, max(p1.x, p2.x)));
@@ -117,6 +118,9 @@ void draw_triangle_nearest_b(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
   S64 max_x = (S64)min((F32)dst->x, ceil(max_x1));
   S64 max_y = (S64)min((F32)dst->y, ceil(max_y1));
 
+  if (min_y >= max_y) return;
+  if (min_x >= max_x) return;
+
   F32 dy10 = (p1.y - p0.y);
   F32 dy21 = (p2.y - p1.y);
   F32 dy02 = (p0.y - p2.y);
@@ -135,7 +139,6 @@ void draw_triangle_nearest_b(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
 
   U32 *destination = dst->pixels + dst->x*min_y;
   F32 area = (p1.y - p0.y) * (p2.x - p0.x) - (p1.x - p0.x) * (p2.y - p0.y);
-  U64 fill_pixels_begin = __rdtsc();
   for (S64 y = min_y; y < max_y; y++) {
     F32 Cx0 = Cy0;
     F32 Cx1 = Cy1;
@@ -223,12 +226,10 @@ void draw_triangle_nearest_b(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
     Cy2 -= dx02;
     destination += dst->x;
   }
+  U64 end_time = __rdtsc();
 
-U64 end_time = __rdtsc();
+  filled_pixel_cycles += end_time - fill_pixels_begin;
-
-  filled_pixel_total_time += end_time - fill_pixels_begin;
   filled_pixel_count  += (max_x - min_x)*(max_y - min_y);
-     // if(os.frame > 10) PROFILE_END(draw_triangle);
 }
 
 
@@ -817,6 +818,7 @@ void draw_triangle_nearest_f(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
                            Vec2 tex0, Vec2 tex1, Vec2 tex2,
                            Vec3 norm0, Vec3 norm1, Vec3 norm2) {
   if(src->pixels == 0) return;
+  U64 fill_pixels_begin = __rdtsc();
 
   PROFILE_SCOPE(draw_triangle);
 
@@ -869,7 +871,6 @@ void draw_triangle_nearest_f(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
   F32 area = (p1.y - p0.y) * (p2.x - p0.x) - (p1.x - p0.x) * (p2.y - p0.y);
   Vec8 area8 = vec8(area);
 
-  U64 fill_pixels_begin = __rdtsc();
   for (S64 y = min_y; y < max_y; y++) {
     Vec8 Cx0 = vec8(Cy0);
     Vec8 Cx1 = vec8(Cy1);
@@ -1025,7 +1026,7 @@ void draw_triangle_nearest_f(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
   }
   U64 end_time = __rdtsc();
 
-  filled_pixel_total_time += end_time - fill_pixels_begin;
+  filled_pixel_cycles += end_time - fill_pixels_begin;
   filled_pixel_count      += (max_x - min_x)*(max_y - min_y);
 }
 
@@ -1035,6 +1036,7 @@ void draw_triangle_nearest_g(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
                            Vec2 tex0, Vec2 tex1, Vec2 tex2,
                            Vec3 norm0, Vec3 norm1, Vec3 norm2) {
   if(src->pixels == 0) return;
+  U64 fill_pixels_begin = __rdtsc();
 
   PROFILE_SCOPE(draw_triangle);
 
@@ -1109,7 +1111,6 @@ void draw_triangle_nearest_g(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
   F32 area = (p1.y - p0.y) * (p2.x - p0.x) - (p1.x - p0.x) * (p2.y - p0.y);
   F32x8 area8 = _mm256_set1_ps(area);
 
-  U64 fill_pixels_begin = __rdtsc();
   for (S64 y = min_y; y < max_y; y++) {
     F32x8 Cx0 = _mm256_set1_ps(Cy0);
     F32x8 Cx1 = _mm256_set1_ps(Cy1);
@@ -1311,8 +1312,6 @@ void draw_triangle_nearest_g(Bitmap* dst, F32 *depth_buffer, Bitmap *src, Vec3 l
     Cy2 -= dx02;
     destination += dst->x;
   }
-  U64 end_time = __rdtsc();
+  filled_pixel_cycles += __rdtsc() - fill_pixels_begin;
-
-  filled_pixel_total_time += end_time - fill_pixels_begin;
   filled_pixel_count      += (max_x - min_x)*(max_y - min_y);
 }

profile.cpp

@@ -1,54 +0,0 @@
-enum ProfileScopeName {
-  ProfileScopeName_draw_triangle,
-  ProfileScopeName_fill_triangle,
-  ProfileScopeName_draw_all_meshes,
-  ProfileScopeName_draw_mesh,
-  ProfileScopeName_draw_set_of_mesh_indices,
-  ProfileScopeName_main_loop,
-  ProfileScopeName_fill_triangle_after_depth_test,
-  ProfileScopeName_fill_triangle_inner,
-  ProfileScopeName_fill_triangle_outer,
-  ProfileScopeName_Count,
-};
-
-const char *profile_scope_names[] = {
-  "draw_triangle",
-  "fill_triangle",
-  "draw_all_meshes",
-  "draw_mesh",
-  "draw_set_of_mesh_indices",
-  "main_loop",
-  "fill_triangle_after_depth_test",
-  "fill_triangle_inner",
-  "fill_triangle_outer",
-};
-
-struct ProfileState {
-  U64 samples[5096*32];
-  S32 i;
-};
-
-global ProfileState profile_scopes[ProfileScopeName_Count];
-
-force_inline void
-profile_begin(ProfileScopeName name){
-  ProfileState *p = profile_scopes + name;
-  p->samples[p->i] = __rdtsc();
-}
-
-force_inline void
-profile_end(ProfileScopeName name){
-  ProfileState *p = profile_scopes + name;
-  p->samples[p->i] = __rdtsc() - p->samples[p->i];
-  p->i = (p->i + 1) % buff_cap(p->samples);
-}
-
-struct Profile_Scope{
-  ProfileScopeName n;
-  force_inline Profile_Scope(ProfileScopeName name){ profile_begin(name); n=name; }
-  force_inline ~Profile_Scope(){ profile_end(n); }
-};
-
-#define PROFILE_BEGIN(name) profile_begin(ProfileScopeName_##name)
-#define PROFILE_END(name) profile_end(ProfileScopeName_##name)
-#define PROFILE_SCOPE(name) Profile_Scope profile_scope_##__LINE__(ProfileScopeName_##name)