cuda: Add more formats for texture read/write (#7012)

* WiP: Add more formats for texture reads

* fix test

* format code

* add  float2/float4 versions for 1D and 3D as well

* fixed review comment

* fix review comments

---------

Co-authored-by: slangbot <186143334+slangbot@users.noreply.github.com>
Co-authored-by: Ellie Hermaszewska <ellieh@nvidia.com>

1 files changed, 77 insertions, 37 deletions

diff --git a/tests/compute/texture-subscript-cuda.slang b/tests/compute/texture-subscript-cuda.slang
index e64f42b19..7c4a2cc78 100644
--- a/tests/compute/texture-subscript-cuda.slang
+++ b/tests/compute/texture-subscript-cuda.slang
@@ -5,14 +5,24 @@
 Texture1D<float> cudaT1D;
 //TEST_INPUT: Texture2D(size=8, content = one):name cudaT2D
 Texture2D<float> cudaT2D;
+//TEST_INPUT: Texture2D(size=8, content = one):name cudaT2D_f2
+Texture2D<float2> cudaT2D_f2;
+//TEST_INPUT: Texture2D(size=8, content = one):name cudaT2D_f4
+Texture2D<float4> cudaT2D_f4;
 //TEST_INPUT: Texture3D(size=8, content = one):name cudaT3D
 Texture3D<float> cudaT3D;
+//TEST_INPUT: Texture3D(size=8, content = one):name cudaT3D_f2
+Texture3D<float2> cudaT3D_f2;
+//TEST_INPUT: Texture3D(size=8, content = one):name cudaT3D_f4
+Texture3D<float4> cudaT3D_f4;
 //TEST_INPUT: TextureCube(size=16, content = one):name cudaTCube
 TextureCube<float> cudaTCube;
 //TEST_INPUT: Texture2D(size=16, content = one, arrayLength=3):name cudaT2DArray
 Texture2DArray<float> cudaT2DArray;
-//TEST_INPUT: TextureCube(size=16, content = one, arrayLength=1):name cudaTCubeArray
-TextureCubeArray<float> cudaTCubeArray;
+//TEST_INPUT: Texture2D(size=16, content = one, arrayLength=3):name cudaT2DArray_f2
+Texture2DArray<float2> cudaT2DArray_f2;
+//TEST_INPUT: Texture2D(size=16, content = one, arrayLength=3):name cudaT2DArray_f4
+Texture2DArray<float4> cudaT2DArray_f4;
 
 //TEST_INPUT: ubuffer(data=[0 0 0 0 0 0 0], stride=4):out,name cudaOutputBuffer
 RWStructuredBuffer<float> cudaOutputBuffer;
@@ -23,39 +33,69 @@ void computeMain(int3 dispatchThreadID : SV_DispatchThreadID)
 {
     int idx = dispatchThreadID.x;
     
-	switch (idx)
-	{
-	case 1:
-		{
-			int var = 0;
-			float result = cudaT1D[0];
-			// This is not supported in PTX.
-			//cudaOutputBuffer[idx] = result;
-		}	
-	break;
-	
-	case 2:
-		{
-			int2 var = int2(1, 2);
-			float result = cudaT2D[var];
-			cudaOutputBuffer[idx] = result;
-		}
-	break;
-
-	case 3:
-		{
-			int3 var = int3(1, 1, 1);
-			float result = cudaT3D[var];
-			cudaOutputBuffer[idx] = result;
-		}
-	break;
-
-	case 4:
-		{
-			int3 var = int3(0, 0, 1);
-			float result = cudaT2DArray[var];
-			cudaOutputBuffer[idx] = result;
-		}
-	break;
-	}
+    switch (idx)
+    {
+    case 1:
+        {
+            int var = 0;
+            float result = cudaT1D[0];
+            // This is not supported in PTX.
+            //cudaOutputBuffer[idx] = result;
+        }    
+    break;
+    
+    case 2:
+        {
+            int2 var = int2(1, 2);
+            float result = cudaT2D[var];
+            cudaOutputBuffer[idx] = result;
+            
+            float2 result2 = cudaT2D_f2[var];
+            cudaOutputBuffer[idx] += result2.x;
+            cudaOutputBuffer[idx] += result2.y;
+
+            float4 result4 = cudaT2D_f4[var];
+            cudaOutputBuffer[idx] += result4.x;
+            cudaOutputBuffer[idx] += result4.y;
+            cudaOutputBuffer[idx] += result4.z;
+            cudaOutputBuffer[idx] += result4.w;
+        }
+    break;
+
+    case 3:
+        {
+            int3 var = int3(1, 1, 1);
+            float result = cudaT3D[var];
+            cudaOutputBuffer[idx] = result;
+            
+            float2 result2 = cudaT3D_f2[var];
+            cudaOutputBuffer[idx] += result2.x;
+            cudaOutputBuffer[idx] += result2.y;
+
+            float4 result4 = cudaT3D_f4[var];
+            cudaOutputBuffer[idx] += result4.x;
+            cudaOutputBuffer[idx] += result4.y;
+            cudaOutputBuffer[idx] += result4.z;
+            cudaOutputBuffer[idx] += result4.w;
+        }
+    break;
+
+    case 4:
+        {
+            int3 var = int3(0, 0, 1);
+            float result = cudaT2DArray[var];
+            cudaOutputBuffer[idx] = result;
+            
+            float2 result2 = cudaT2DArray_f2[var];
+            cudaOutputBuffer[idx] += result2.x;
+            cudaOutputBuffer[idx] += result2.y;
+
+            float4 result4 = cudaT2DArray_f4[var];
+            cudaOutputBuffer[idx] += result4.x;
+            cudaOutputBuffer[idx] += result4.y;
+            cudaOutputBuffer[idx] += result4.z;
+            cudaOutputBuffer[idx] += result4.w;
+        }
+    break;
+    }
 } 
\ No newline at end of file