Remove 2x QR instruction from the SSS pass

ScriptableRenderPipeline/HDRenderPipeline/HDRP/Camera/HDCamera.cs

         public Vector4[] frustumPlaneEquations;
         public Camera camera;
         public uint taaFrameIndex;
+        public Vector2 taaFrameRotation;
         public Vector4 viewParam;
         public PostProcessRenderContext postprocessRenderContext;
 
 
             const uint taaFrameCount = 8;
             taaFrameIndex = taaEnabled ? (uint)Time.renderedFrameCount % taaFrameCount : 0;
+            taaFrameRotation = new Vector2(Mathf.Sin(taaFrameIndex * (0.5f * Mathf.PI)),
+                                           Mathf.Cos(taaFrameIndex * (0.5f * Mathf.PI)));
 
             viewMatrix = gpuView;
             projMatrix = gpuProj;
             cmd.SetGlobalMatrix(HDShaderIDs._PrevViewProjMatrix, prevViewProjMatrix);
             cmd.SetGlobalVectorArray(HDShaderIDs._FrustumPlanes, frustumPlaneEquations);
             cmd.SetGlobalInt(HDShaderIDs._TaaFrameIndex, (int)taaFrameIndex);
+            cmd.SetGlobalVector(HDShaderIDs._TaaFrameRotation, taaFrameRotation);
         }
 
         // TODO: We should set all the value below globally and not let it under the control of Unity,

ScriptableRenderPipeline/HDRenderPipeline/HDRP/HDStringConstants.cs

         public static readonly int _ScreenSize = Shader.PropertyToID("_ScreenSize");
         public static readonly int _PrevViewProjMatrix = Shader.PropertyToID("_PrevViewProjMatrix");
         public static readonly int _FrustumPlanes = Shader.PropertyToID("_FrustumPlanes");
-        public static readonly int _TaaFrameIndex = Shader.PropertyToID("_TaaFrameIndex");
+        public static readonly int _TaaFrameIndex    = Shader.PropertyToID("_TaaFrameIndex");
+        public static readonly int _TaaFrameRotation = Shader.PropertyToID("_TaaFrameRotation");
 
         public static readonly int _DepthTexture                   = Shader.PropertyToID("_DepthTexture");
         public static readonly int _CameraColorTexture             = Shader.PropertyToID("_CameraColorTexture");

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

 #define SSS_USE_LDS_CACHE     1
 #define SSS_TAA_INTEGRATION   1 // Smoother results at the cost of a tiny amount of flickering in under-sampled areas
 #define SSS_ENABLE_NEAR_FIELD 0 // Greatly increases the number of samples. Comes at a high cost.
-#define SSS_SAMPLE_TEST_HTILE 0 // Potential optimization. YMMV.
 #define SSS_USE_TANGENT_PLANE 0 // Improves the accuracy of the approximation(0 -> 1st order). High cost. Does not work with back-facing normals.
 #define SSS_CLAMP_ARTIFACT    0 // Reduces bleeding. Use with SSS_USE_TANGENT_PLANE.
 #define SSS_DEBUG_LOD         0
                     float3 tangentX, float3 tangentY, float4x4 projMatrix,
                     inout float3 totalIrradiance, inout float3 totalWeight)
 {
-    float  r   = _FilterKernels[profileID][i][iR];
+    float r   = _FilterKernels[profileID][i][iR];
-    float  phi = SampleDiskFibonacci(i, n).y;
+    float phi = SampleDiskFibonacci(i, n).y;
-    phi += VanDerCorputBase2(_TaaFrameIndex % 4) * TWO_PI;
-#endif
+    float sinPsi = _TaaFrameRotation.x;
+    float cosPsi = _TaaFrameRotation.y;
+
+    // The angle 'psi' is loop-invariant. All the trigonometry is done at compile time.
+    // cos(a + b) = cos(a) * cos(b) - sin(a) * sin(b)
+    // sin(a + b) = sin(a) * cos(b) + cos(a) * sin(b)
+    float cosSum = cos(phi) * cosPsi - sin(phi) * sinPsi;
+    float sinSum = sin(phi) * cosPsi + cos(phi) * sinPsi;
+
+    float2 vec = r * float2(cosSum, sinSum);
+#else
+#endif
 
     // Compute the screen-space position and the squared distance (in mm) in the image plane.
     int2 position; float xy2;

ScriptableRenderPipeline/HDRenderPipeline/HDRP/ShaderVariables.hlsl

     int unity_StereoEyeIndex;
 #endif
 
-    float3 unity_ShadowColor;
-    uint _TaaFrameIndex;      // [0, 7]
+    float4 unity_ShadowColor;
+    float2 _TaaFrameRotation; // {x = sin(_TaaFrameIndex * PI/2), y = cos(_TaaFrameIndex * PI/2), z = unused}
+    uint   _TaaFrameIndex;    // [0, 7]
+    uint   _Align128;
     // Volumetric lighting. Should be a struct in 'UnityPerFrame'.
     // Unfortunately, structures inside constant buffers are not supported by Unity.
     float3 _GlobalFog_Scattering;