perf: increase thread group size for several comp shaders

Author: hoffstadt

extensions/pl_renderer_ext.c

@@ -3020,11 +3020,11 @@ pl_renderer_render_view(plView* ptView, plCamera* ptCamera, plCamera* ptCullCame
         uJumpSteps = 1;
 
     const plDispatch tDispach = {
-        .uGroupCountX     = (uint32_t)ceilf(tDimensions.x / 8.0f),
-        .uGroupCountY     = (uint32_t)ceilf(tDimensions.y / 8.0f),
+        .uGroupCountX     = (uint32_t)ceilf(tDimensions.x / 32.0f),
+        .uGroupCountY     = (uint32_t)ceilf(tDimensions.y / 32.0f),
         .uGroupCountZ     = 1,
-        .uThreadPerGroupX = 8,
-        .uThreadPerGroupY = 8,
+        .uThreadPerGroupX = 32,
+        .uThreadPerGroupY = 32,
         .uThreadPerGroupZ = 1
     };
 
@@ -3210,11 +3210,11 @@ pl_renderer_render_view(plView* ptView, plCamera* ptCamera, plCamera* ptCullCame
     plTexture* ptFinalTexture = gptGfx->get_texture(gptData->ptDevice, ptView->tFinalTexture);
 
     plDispatch tTonemapDispatch = {
-        .uGroupCountX     = (uint32_t)(ceilf(ptFinalTexture->tDesc.tDimensions.x / 8.0f)),
-        .uGroupCountY     = (uint32_t)(ceilf(ptFinalTexture->tDesc.tDimensions.y / 8.0f)),
+        .uGroupCountX     = (uint32_t)(ceilf(ptFinalTexture->tDesc.tDimensions.x / 32.0f)),
+        .uGroupCountY     = (uint32_t)(ceilf(ptFinalTexture->tDesc.tDimensions.y / 32.0f)),
         .uGroupCountZ     = 1,
-        .uThreadPerGroupX = 8,
-        .uThreadPerGroupY = 8,
+        .uThreadPerGroupX = 32,
+        .uThreadPerGroupY = 32,
         .uThreadPerGroupZ = 1
     };
     gptGfx->dispatch(ptPostEncoder, 1, &tTonemapDispatch);

shaders/jumpfloodalgo.comp

@@ -51,12 +51,12 @@ jump(vec3 minSeed, ivec2 current, ivec2 offset)
 // [SECTION] entry
 //-----------------------------------------------------------------------------
 
-layout (local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+layout (local_size_x = 32, local_size_y = 32, local_size_z = 1) in;
 
 void main() 
 {
-    const int iXCoord = int(gl_WorkGroupID.x * 8 + gl_LocalInvocationID.x);
-    const int iYCoord = int(gl_WorkGroupID.y * 8 + gl_LocalInvocationID.y);
+    const int iXCoord = int(gl_WorkGroupID.x * 32 + gl_LocalInvocationID.x);
+    const int iYCoord = int(gl_WorkGroupID.y * 32 + gl_LocalInvocationID.y);
 
     if(iXCoord >= tDynamicData.tJumpDistance.x || iYCoord >= tDynamicData.tJumpDistance.y)
         return;

shaders/tonemap.comp

@@ -88,14 +88,14 @@ layout(set = 3, binding = 0) uniform PL_DYNAMIC_DATA
 // [SECTION] entry
 //-----------------------------------------------------------------------------
 
-layout (local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+layout (local_size_x = 32, local_size_y = 32, local_size_z = 1) in;
 
 void
 main() 
 {
 
-    const int iXCoord = int(gl_WorkGroupID.x * 8 + gl_LocalInvocationID.x);
-    const int iYCoord = int(gl_WorkGroupID.y * 8 + gl_LocalInvocationID.y);
+    const int iXCoord = int(gl_WorkGroupID.x * 32 + gl_LocalInvocationID.x);
+    const int iYCoord = int(gl_WorkGroupID.y * 32 + gl_LocalInvocationID.y);
 
     const ivec2 tImageSize = imageSize(inputImage);