Add hlsl helper for reading other lanes via derivative swizzling

renderdoc/data/hlsl/quadswizzle.hlsl

@@ -0,0 +1,115 @@
+/******************************************************************************
+ * The MIT License (MIT)
+ *
+ * Copyright (c) 2025 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.
+ ******************************************************************************/
+
+float4 quadSwizzleHelper(float4 c0, uint quadLaneIndex, uint readIndex)
+{
+  bool quadX = (quadLaneIndex & 1u) != 0u;
+  bool quadY = (quadLaneIndex & 2u) != 0u;
+
+  bool readX = ((readIndex & 1u) != 0u);
+  bool readY = ((readIndex & 2u) != 0u);
+
+  float4 sign_x = 1.0f;
+  sign_x.x = quadX ? -1.0f : 1.0f;
+  sign_x.y = quadX ? -1.0f : 1.0f;
+  sign_x.z = quadX ? -1.0f : 1.0f;
+  sign_x.w = quadX ? -1.0f : 1.0f;
+
+  float4 sign_y = 1.0f;
+  sign_y.x = quadY ? -1.0f : 1.0f;
+  sign_y.y = quadY ? -1.0f : 1.0f;
+  sign_y.z = quadY ? -1.0f : 1.0f;
+  sign_y.w = quadY ? -1.0f : 1.0f;
+
+  float4 c1 = c0 + (sign_x * ddx_fine(c0));
+  float4 c2 = c0 + (sign_y * ddy_fine(c0));
+  float4 c3 = c2 + (sign_x * ddx_fine(c2));
+
+  if(quadLaneIndex == readIndex)
+    return c0;
+  else if(readY == quadY)
+    return c1;
+  else if(readX == quadX)
+    return c2;
+  else
+    return c3;
+}
+
+// helpers for smaller vector sizes
+float3 quadSwizzleHelper(float3 c0, uint quadLaneIndex, uint readIndex)
+{
+  return quadSwizzleHelper(float4(c0, 0.0f), quadLaneIndex, readIndex).xyz;
+}
+
+float2 quadSwizzleHelper(float2 c0, uint quadLaneIndex, uint readIndex)
+{
+  return quadSwizzleHelper(float4(c0, 0.0f, 0.0f), quadLaneIndex, readIndex).xy;
+}
+
+float quadSwizzleHelper(float c0, uint quadLaneIndex, uint readIndex)
+{
+  return quadSwizzleHelper(float4(c0, 0.0f, 0.0f, 0.0f), quadLaneIndex, readIndex).x;
+}
+
+// integer helpers. We expect to only use this for scalar uint and only for certain builtin inputs
+// that have small values and so can be cast to/from float accurately
+uint4 quadSwizzleHelper(uint4 c0, uint quadLaneIndex, uint readIndex)
+{
+  return uint4(quadSwizzleHelper(float4(c0), quadLaneIndex, readIndex));
+}
+
+uint3 quadSwizzleHelper(uint3 c0, uint quadLaneIndex, uint readIndex)
+{
+  return uint3(quadSwizzleHelper(float3(c0), quadLaneIndex, readIndex));
+}
+
+uint2 quadSwizzleHelper(uint2 c0, uint quadLaneIndex, uint readIndex)
+{
+  return uint2(quadSwizzleHelper(float2(c0), quadLaneIndex, readIndex));
+}
+
+uint quadSwizzleHelper(uint c0, uint quadLaneIndex, uint readIndex)
+{
+  return uint(quadSwizzleHelper(float(c0), quadLaneIndex, readIndex));
+}
+
+int4 quadSwizzleHelper(int4 c0, uint quadLaneIndex, uint readIndex)
+{
+  return uint4(quadSwizzleHelper(float4(c0), quadLaneIndex, readIndex));
+}
+
+int3 quadSwizzleHelper(int3 c0, uint quadLaneIndex, uint readIndex)
+{
+  return uint3(quadSwizzleHelper(float3(c0), quadLaneIndex, readIndex));
+}
+
+int2 quadSwizzleHelper(int2 c0, uint quadLaneIndex, uint readIndex)
+{
+  return uint2(quadSwizzleHelper(float2(c0), quadLaneIndex, readIndex));
+}
+
+int quadSwizzleHelper(int c0, uint quadLaneIndex, uint readIndex)
+{
+  return uint(quadSwizzleHelper(float(c0), quadLaneIndex, readIndex));
+}

renderdoc/data/renderdoc.rc

@@ -118,6 +118,7 @@ RESOURCE_shaderdebug_hlsl  TYPE_EMBED   "hlsl/shaderdebug.hlsl"
 RESOURCE_d3d12_pixelhistory_hlsl TYPE_EMBED   "hlsl/d3d12_pixelhistory.hlsl"
 RESOURCE_depth_copy_hlsl   TYPE_EMBED   "hlsl/depth_copy.hlsl"
 RESOURCE_raytracing_hlsl   TYPE_EMBED   "hlsl/raytracing.hlsl"
+RESOURCE_quadswizzle_hlsl  TYPE_EMBED   "hlsl/quadswizzle.hlsl"
 
 #ifdef RENDERDOC_BAKED_DXC_SHADERS
 

renderdoc/data/resource.h

@@ -21,6 +21,7 @@
 #define RESOURCE_d3d12_pixelhistory_hlsl  119
 #define RESOURCE_depth_copy_hlsl    120
 #define RESOURCE_raytracing_hlsl    121
+#define RESOURCE_quadswizzle_hlsl   131
 
 #define RESOURCE_fixedcol_0_dxbc    113
 #define RESOURCE_fixedcol_1_dxbc    114

renderdoc/renderdoc.vcxproj

@@ -718,6 +718,7 @@
     <None Include="data\hlsl\shaderdebug.hlsl" />
     <None Include="data\hlsl\depth_copy.hlsl" />
     <None Include="data\hlsl\raytracing.hlsl" />
+    <None Include="data\hlsl\quadswizzle.hlsl" />
   </ItemGroup>
   <!-- try to bake shaders needed in DXIL now, in case we won't be able to find a dxc at runtime -->
   <UsingTask TaskName="GetDXCExecutable" TaskFactory="CodeTaskFactory" AssemblyFile="$(MSBuildToolsPath)\Microsoft.Build.Tasks.Core.dll">

renderdoc/renderdoc.vcxproj.filters

@@ -1142,6 +1142,9 @@
     <None Include="data\hlsl\raytracing.hlsl">
       <Filter>Resources\hlsl</Filter>
     </None>
+    <None Include="data\hlsl\quadswizzle.hlsl">
+      <Filter>Resources\hlsl</Filter>
+    </None>
   </ItemGroup>
   <ItemGroup>
     <ResourceCompile Include="data\renderdoc.rc">