Hotfix/texture2d gather (#876)

* First pass test to see if GatherRed works.

* Add support for generating R_Float32 textures.

* Set default texture format.

* * Alter the texture2d-gather to work with a R_Float32 texture
* Add support for scalar Texture2d types with GatherXXX in stdlib

* Remove some left over commented out test code from texture2d-gather.hlsl

6 files changed, 146 insertions, 20 deletions

diff --git a/source/slang/core.meta.slang b/source/slang/core.meta.slang
index 0c2b70e10..f3e57b90b 100644
--- a/source/slang/core.meta.slang
+++ b/source/slang/core.meta.slang
@@ -1001,8 +1001,12 @@ for (int tt = 0; tt < kBaseTextureTypeCount; ++tt)
             static const struct {
                 char const* genericPrefix;
                 char const* elementType;
+                char const* outputType;
             } kGatherExtensionCases[] = {
-                { "__generic<T, let N : int>", "vector<T,N>" },
+                { "__generic<T, let N : int>", "vector<T,N>", "vector<T, 4>" },
+                { "", "float", "vector<float, 4>" },
+                { "", "int" , "vector<int, 4>"},
+                { "", "uint", "vector<uint, 4>"},
 
                 // TODO: need a case here for scalars `T`, but also
                 // need to ensure that case doesn't accidentally match
@@ -1022,7 +1026,6 @@ for (int tt = 0; tt < kBaseTextureTypeCount; ++tt)
                 sb << "<" << cc.elementType << " >";
                 sb << "\n{\n";
 
-
                 // `Gather`
                 // (tricky because it returns a 4-vector of the element type
                 // of the texture components...)
@@ -1046,27 +1049,29 @@ for (int tt = 0; tt < kBaseTextureTypeCount; ++tt)
                     auto componentIndex = kk.componentIndex;
                     auto componentName = kk.componentName;
 
+                    auto outputType = cc.outputType;
+
                     EMIT_LINE_DIRECTIVE();
 
                     sb << "__target_intrinsic(glsl, \"textureGather($p, $2, " << componentIndex << ")\")\n";
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location);\n";
 
                     EMIT_LINE_DIRECTIVE();
                     sb << "__target_intrinsic(glsl, \"textureGatherOffset($p, $2, $3, " << componentIndex << ")\")\n";
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location, ";
                     sb << "constexpr int" << kBaseTextureTypes[tt].coordCount << " offset);\n";
 
                     EMIT_LINE_DIRECTIVE();
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location, ";
                     sb << "constexpr int" << kBaseTextureTypes[tt].coordCount << " offset, ";
                     sb << "out uint status);\n";
 
                     EMIT_LINE_DIRECTIVE();
                     sb << "__target_intrinsic(glsl, \"textureGatherOffsets($p, $2, int" << kBaseTextureTypes[tt].coordCount << "[]($3, $4, $5, $6), " << componentIndex << ")\")\n";
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location, ";
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset1, ";
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset2, ";
@@ -1074,7 +1079,7 @@ for (int tt = 0; tt < kBaseTextureTypeCount; ++tt)
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset4);\n";
 
                     EMIT_LINE_DIRECTIVE();
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location, ";
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset1, ";
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset2, ";
diff --git a/source/slang/core.meta.slang.h b/source/slang/core.meta.slang.h
index 25feb0005..2ae45a01c 100644
--- a/source/slang/core.meta.slang.h
+++ b/source/slang/core.meta.slang.h
@@ -1019,8 +1019,12 @@ for (int tt = 0; tt < kBaseTextureTypeCount; ++tt)
             static const struct {
                 char const* genericPrefix;
                 char const* elementType;
+                char const* outputType;
             } kGatherExtensionCases[] = {
-                { "__generic<T, let N : int>", "vector<T,N>" },
+                { "__generic<T, let N : int>", "vector<T,N>", "vector<T, 4>" },
+                { "", "float", "vector<float, 4>" },
+                { "", "int" , "vector<int, 4>"},
+                { "", "uint", "vector<uint, 4>"},
 
                 // TODO: need a case here for scalars `T`, but also
                 // need to ensure that case doesn't accidentally match
@@ -1040,7 +1044,6 @@ for (int tt = 0; tt < kBaseTextureTypeCount; ++tt)
                 sb << "<" << cc.elementType << " >";
                 sb << "\n{\n";
 
-
                 // `Gather`
                 // (tricky because it returns a 4-vector of the element type
                 // of the texture components...)
@@ -1064,27 +1067,29 @@ for (int tt = 0; tt < kBaseTextureTypeCount; ++tt)
                     auto componentIndex = kk.componentIndex;
                     auto componentName = kk.componentName;
 
+                    auto outputType = cc.outputType;
+
                     EMIT_LINE_DIRECTIVE();
 
                     sb << "__target_intrinsic(glsl, \"textureGather($p, $2, " << componentIndex << ")\")\n";
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location);\n";
 
                     EMIT_LINE_DIRECTIVE();
                     sb << "__target_intrinsic(glsl, \"textureGatherOffset($p, $2, $3, " << componentIndex << ")\")\n";
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location, ";
                     sb << "constexpr int" << kBaseTextureTypes[tt].coordCount << " offset);\n";
 
                     EMIT_LINE_DIRECTIVE();
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location, ";
                     sb << "constexpr int" << kBaseTextureTypes[tt].coordCount << " offset, ";
                     sb << "out uint status);\n";
 
                     EMIT_LINE_DIRECTIVE();
                     sb << "__target_intrinsic(glsl, \"textureGatherOffsets($p, $2, int" << kBaseTextureTypes[tt].coordCount << "[]($3, $4, $5, $6), " << componentIndex << ")\")\n";
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location, ";
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset1, ";
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset2, ";
@@ -1092,7 +1097,7 @@ for (int tt = 0; tt < kBaseTextureTypeCount; ++tt)
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset4);\n";
 
                     EMIT_LINE_DIRECTIVE();
-                    sb << "vector<T, 4> Gather" << componentName << "(SamplerState s, ";
+                    sb << outputType << " Gather" << componentName << "(SamplerState s, ";
                     sb << "float" << kBaseTextureTypes[tt].coordCount << " location, ";
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset1, ";
                     sb << "int" << kBaseTextureTypes[tt].coordCount << " offset2, ";
@@ -1197,7 +1202,7 @@ for (auto op : binaryOps)
         sb << "__intrinsic_op(" << int(op.opCode) << ") matrix<" << resultType << ",N,M> operator" << op.opName << "(" << leftQual << "matrix<" << leftType << ",N,M> left, " << rightType << " right);\n";
     }
 }
-SLANG_RAW("#line 1182 \"core.meta.slang\"")
+SLANG_RAW("#line 1187 \"core.meta.slang\"")
 SLANG_RAW("\n")
 SLANG_RAW("\n")
 SLANG_RAW("// Operators to apply to `enum` types\n")
diff --git a/tests/bugs/texture2d-gather.hlsl b/tests/bugs/texture2d-gather.hlsl
new file mode 100644
index 000000000..3886472bd
--- /dev/null
+++ b/tests/bugs/texture2d-gather.hlsl
@@ -0,0 +1,44 @@
+//TEST(smoke):COMPARE_HLSL_RENDER:
+//TEST_INPUT: Texture2D(size=16, content=chessboard, format=R_Float32):dxbinding(0),glbinding(0)
+
+Texture2D<float> g_texture : register(t0);
+SamplerState g_sampler : register(s0);
+
+cbuffer Uniforms
+{
+	float4x4 modelViewProjection;
+}
+
+struct AssembledVertex
+{
+	float3	position;
+	float3	color;
+};
+
+// Vertex  Shader
+struct VertexStageInput
+{
+	AssembledVertex assembledVertex	: A;
+};
+
+struct VertexStageOutput
+{
+	float4          color           : COLOR; 
+	float4			position		: SV_Position;
+};
+
+VertexStageOutput vertexMain(VertexStageInput input) 
+{
+    VertexStageOutput output;
+    
+    output.position = mul(modelViewProjection, float4(input.assembledVertex.position, 1.0));
+    output.color = float4(input.assembledVertex.color, 1.0f);
+    
+    return output;
+}
+
+// Fragment Shader
+float4 fragmentMain(VertexStageOutput input) : SV_Target
+{
+    return g_texture.GatherRed(g_sampler, input.color.xy);
+}
\ No newline at end of file
diff --git a/tools/render-test/shader-input-layout.cpp b/tools/render-test/shader-input-layout.cpp
index 5a2021b01..6d51a1980 100644
--- a/tools/render-test/shader-input-layout.cpp
+++ b/tools/render-test/shader-input-layout.cpp
@@ -208,10 +208,25 @@ namespace renderer_test
                                 }
                                 else if(word == "format")
                                 {
+                                    Format format = Format::Unknown;
+
                                     parser.Read("=");
                                     auto formatWord = parser.ReadWord();
                                     if(formatWord == "R_UInt32")
-                                        entry.bufferDesc.format = Format::R_UInt32;
+                                    {
+                                        format = Format::R_UInt32;
+                                    }
+                                    else if (formatWord == "R_Float32")
+                                    {
+                                        format = Format::R_Float32;
+                                    }
+                                    else if (formatWord == "RGBA_Unorm_UInt8")
+                                    {
+                                        format = Format::RGBA_Unorm_UInt8;
+                                    }
+
+                                    entry.textureDesc.format = format;
+                                    entry.bufferDesc.format = format;
                                 }
                                 if (parser.LookAhead(","))
                                     parser.Read(",");
@@ -265,15 +280,64 @@ namespace renderer_test
                 }
             }
         }
+    }
+
+    void generateTextureData(TextureData& output, const InputTextureDesc& desc)
+    {
+        switch (desc.format)
+        {
+            case Format::RGBA_Unorm_UInt8:
+            {
+                generateTextureDataRGB8(output, desc);
+                break;
+            }
+            case Format::R_Float32:
+            {
+                TextureData work;
+                generateTextureDataRGB8(work, desc);
 
+                output.textureSize = work.textureSize;
+                output.mipLevels = work.mipLevels;
+                output.arraySize = work.arraySize;
 
+                List<List<unsigned int>>& dstBuffer = output.dataBuffer;
+
+                int numMips = int(work.dataBuffer.Count());
+                dstBuffer.SetSize(numMips);
+
+                for (int i = 0; i < numMips; ++i)
+                {
+                    const int numPixels = int(work.dataBuffer[i].Count());
+                    const unsigned int* srcPixels = work.dataBuffer[i].Buffer();
+
+                    dstBuffer[i].SetSize(numPixels);
+
+                    float* dstPixels = (float*)dstBuffer[i].Buffer();
+
+                    for (int j = 0; j < numPixels; ++j)
+                    {
+                        // Copy out red
+                        const unsigned int srcPixel = srcPixels[j];
+                        const float value = (srcPixel & 0xff) * 1.0f / 255;
+                        dstPixels[j] = value;
+                    }
+                }
+                break;
+            }
+            default:
+            {
+                SLANG_ASSERT(!"Unhandled format");
+                break;
+            }
+        }
     }
-    void generateTextureData(TextureData & output, const InputTextureDesc & inputDesc)
+
+    void generateTextureDataRGB8(TextureData& output, const InputTextureDesc& inputDesc)
     {
         int arrLen = inputDesc.arrayLength;
         if (arrLen == 0)
             arrLen = 1;
-        List<List<unsigned int>> & dataBuffer = output.dataBuffer;
+        List<List<unsigned int>>& dataBuffer = output.dataBuffer;
         int arraySize = arrLen;
         if (inputDesc.isCube)
             arraySize *= 6;
diff --git a/tools/render-test/shader-input-layout.h b/tools/render-test/shader-input-layout.h
index 1044a6eea..d9188fadd 100644
--- a/tools/render-test/shader-input-layout.h
+++ b/tools/render-test/shader-input-layout.h
@@ -27,6 +27,9 @@ struct InputTextureDesc
     bool isDepthTexture = false;
     bool isRWTexture = false;
     int size = 4;
+
+    Format format = Format::RGBA_Unorm_UInt8;            
+
     InputTextureContent content = InputTextureContent::One;
 };
 
@@ -80,7 +83,9 @@ public:
     void Parse(const char * source);
 };
 
-void generateTextureData(TextureData & output, const InputTextureDesc & desc);
+void generateTextureDataRGB8(TextureData& output, const InputTextureDesc& desc);
+void generateTextureData(TextureData& output, const InputTextureDesc& desc);
+
 
 } // namespace render_test
 
diff --git a/tools/render-test/shader-renderer-util.cpp b/tools/render-test/shader-renderer-util.cpp
index 0d0f8a3a5..3dc717c5b 100644
--- a/tools/render-test/shader-renderer-util.cpp
+++ b/tools/render-test/shader-renderer-util.cpp
@@ -28,7 +28,10 @@ void BindingStateImpl::apply(Renderer* renderer, PipelineType pipelineType)
     TextureResource::Desc textureResourceDesc;
     textureResourceDesc.init(Resource::Type::Unknown);
 
-    textureResourceDesc.format = Format::RGBA_Unorm_UInt8;
+    // Default to RGBA_Unorm_UInt8
+    const Format format = (inputDesc.format == Format::Unknown) ? Format::RGBA_Unorm_UInt8 : inputDesc.format;
+
+    textureResourceDesc.format = format;
     textureResourceDesc.numMipLevels = texData.mipLevels;
     textureResourceDesc.arraySize = inputDesc.arrayLength;
     textureResourceDesc.bindFlags = bindFlags;