Do not use the 32x32 macrotile order

All we care about is the optimal layout inside each 8x8 tile

ScriptableRenderPipeline/HDRenderPipeline/HDRP/Material/SubsurfaceScattering/SubsurfaceScattering.compute

 #pragma kernel SubsurfaceScattering
 
 [numthreads(GROUP_SIZE_2D, 1, 1)]
-void SubsurfaceScattering(uint3 reorderedGroupId : SV_GroupID,
-                          uint  groupThreadId    : SV_GroupThreadID)
+void SubsurfaceScattering(uint2 groupId       : SV_GroupID,
+                          uint  groupThreadId : SV_GroupThreadID)
-
-    // We dispatch 4x swizzled 16x16 groups per a 32x32 macrotile.
-    // Therefore, we need to reorder. TODO: macrotile order.
-    uint2 groupQuad = DeinterleaveQuad(reorderedGroupId.x);
-    uint2 groupId   = uint2(reorderedGroupId.y * 2 + groupQuad.x, reorderedGroupId.z * 2 + groupQuad.y);
 
     // Arrange threads in the Morton order to optimally match the memory layout of GCN tiles.
     uint2 groupCoord  = DecodeMorton2D(groupThreadId);

ScriptableRenderPipeline/HDRenderPipeline/HDRP/Material/SubsurfaceScattering/SubsurfaceScatteringManager.cs

                         cmd.SetComputeTextureParam(m_SubsurfaceScatteringCS, m_SubsurfaceScatteringKernel, HDShaderIDs._CameraFilteringBuffer, m_CameraFilteringBuffer);
 
                         // Perform the SSS filtering pass which fills 'm_CameraFilteringBufferRT'.
-                        // We dispatch 4x swizzled 16x16 groups per a 32x32 macro tile.
-                        cmd.DispatchCompute(m_SubsurfaceScatteringCS, m_SubsurfaceScatteringKernel, 4, ((int)hdCamera.screenSize.x + 31) / 32, ((int)hdCamera.screenSize.y + 31) / 32);
+                        cmd.DispatchCompute(m_SubsurfaceScatteringCS, m_SubsurfaceScatteringKernel, ((int)hdCamera.screenSize.x + 15) / 16, ((int)hdCamera.screenSize.y + 15) / 16, 1);
 
                         cmd.SetGlobalTexture(HDShaderIDs._IrradianceSource, m_CameraFilteringBuffer);  // Cannot set a RT on a material
 
                         cmd.SetComputeTextureParam(m_SubsurfaceScatteringCS, m_SubsurfaceScatteringKernel, HDShaderIDs._CameraColorTexture, colorBufferRT);
 
                         // Perform the SSS filtering pass which performs an in-place update of 'colorBuffer'.
-                        // We dispatch 4x swizzled 16x16 groups per a 32x32 macro tile.
-                        cmd.DispatchCompute(m_SubsurfaceScatteringCS, m_SubsurfaceScatteringKernel, 4, ((int)hdCamera.screenSize.x + 31) / 32, ((int)hdCamera.screenSize.y + 31) / 32);
+                        cmd.DispatchCompute(m_SubsurfaceScatteringCS, m_SubsurfaceScatteringKernel, ((int)hdCamera.screenSize.x + 15) / 16, ((int)hdCamera.screenSize.y + 15) / 16, 1);
                     }
                 }
                 else