Merge branch 'master' into coerce-constraint

Author: csyonghe

source/slang/slang-emit-hlsl.cpp

@@ -1252,23 +1252,30 @@ void HLSLSourceEmitter::emitSimpleValueImpl(IRInst* inst)
             {
             case IRConstant::FloatKind::Nan:
                 {
-                    m_writer->emit("(0.0 / 0.0)");
+                    m_writer->emit("(0.0f / 0.0f)");
                     return;
                 }
             case IRConstant::FloatKind::PositiveInfinity:
                 {
-                    m_writer->emit("(1.0 / 0.0)");
+                    m_writer->emit("(1.0f / 0.0f)");
                     return;
                 }
             case IRConstant::FloatKind::NegativeInfinity:
                 {
-                    m_writer->emit("(-1.0 / 0.0)");
+                    m_writer->emit("(-1.0f / 0.0f)");
                     return;
                 }
             default:
-                break;
+                {
+                    m_writer->emit(constantInst->value.floatVal);
+                    // Add 'f' suffix for 32-bit float literals to ensure DXC treats them as float
+                    if (constantInst->getDataType()->getOp() == kIROp_FloatType)
+                    {
+                        m_writer->emit("f");
+                    }
+                    return;
+                }
             }
-            break;
         }
 
     default:

source/slang/slang-ir-lower-binding-query.cpp

@@ -288,7 +288,118 @@ struct BindingQueryLoweringContext : public WorkListPass
     //
     OpaqueValueInfo computeOpaqueValueInfo(IRInst* opaqueValue)
     {
-        if (auto globalParam = as<IRGlobalParam>(opaqueValue))
+        if (auto getElement = as<IRGetElement>(opaqueValue))
+        {
+            IRInst* baseInst = getElement->getBase();
+            IRInst* indexInst = getElement->getIndex();
+
+            IRInst* elementType = getElement->getDataType();
+
+            // TODO(JS): This a hack to make this work for arrays of resource type.
+            // It won't work in the general case as it stands because we would need
+            // to propogate layout kind types needed at usage sites.
+            // Without knowing the resource kind that is being processed it's not possible
+            // to accumulate the calculation.
+            //
+            // So presumably we need to request a binding query for a specific resource kind.
+            // We could do this by making the type of the binding query hold the type.
+
+            // We need to add instructions which will work out the binding for the base
+            OpaqueValueInfo baseInfo = findOrComputeOpaqueValueInfo(baseInst);
+
+            // If we couldn't find it we are done
+            if (baseInfo.registerIndex == nullptr || baseInfo.registerSpace == nullptr)
+            {
+                return baseInfo;
+            }
+
+
+            LayoutResourceKind kind = LayoutResourceKind::None;
+            Index stride = 1;
+
+            if (auto resourceType = as<IRResourceType>(elementType))
+            {
+                const auto shape = resourceType->getShape();
+
+                switch (shape)
+                {
+                case SLANG_TEXTURE_1D:
+                case SLANG_TEXTURE_2D:
+                case SLANG_TEXTURE_3D:
+                case SLANG_TEXTURE_CUBE:
+                case SLANG_STRUCTURED_BUFFER:
+                case SLANG_BYTE_ADDRESS_BUFFER:
+                case SLANG_TEXTURE_BUFFER:
+                    {
+                        const auto access = resourceType->getAccess();
+                        bool isReadOnly = (access == SLANG_RESOURCE_ACCESS_READ);
+
+                        kind = isReadOnly ? LayoutResourceKind::ShaderResource
+                                          : LayoutResourceKind::UnorderedAccess;
+                        break;
+                    }
+                default:
+                    break;
+                }
+            }
+            else if (as<IRSamplerStateTypeBase>(elementType))
+            {
+                kind = LayoutResourceKind::SamplerState;
+            }
+            else if (as<IRConstantBufferType>(elementType))
+            {
+                kind = LayoutResourceKind::ConstantBuffer;
+            }
+
+            if (kind == LayoutResourceKind::None)
+            {
+                // Can't determine the kind
+                return OpaqueValueInfo();
+            }
+
+            // If the element type has type layout we can try and use that
+            if (auto layoutDecoration = elementType->findDecoration<IRLayoutDecoration>())
+            {
+                // We have to calculate
+                if (auto elementTypeLayout = as<IRTypeLayout>(layoutDecoration->getLayout()))
+                {
+                    IRTypeSizeAttr* sizeAttr = elementTypeLayout->findSizeAttr(kind);
+                    sizeAttr = sizeAttr ? sizeAttr
+                                        : elementTypeLayout->findSizeAttr(
+                                              LayoutResourceKind::DescriptorTableSlot);
+
+                    if (!sizeAttr)
+                    {
+                        // Couldn't work it out
+                        return OpaqueValueInfo();
+                    }
+
+                    // TODO(JS): Perhaps we have to do something else if not finite?
+                    stride = sizeAttr->getFiniteSize();
+                }
+            }
+
+            SLANG_UNUSED(indexInst);
+
+            // Okay we need to create an instruction which is
+            // base + stride * index
+
+            IRBuilder builder(module);
+
+            builder.setInsertBefore(opaqueValue);
+
+            auto calcRegisterInst = builder.emitAdd(
+                indexType,
+                builder.emitMul(indexType, builder.getIntValue(indexType, stride), indexInst),
+                baseInfo.registerIndex);
+
+            OpaqueValueInfo finalInfo;
+            finalInfo.registerIndex = calcRegisterInst;
+            finalInfo.registerSpace = baseInfo.registerSpace;
+
+            return finalInfo;
+        }
+        else if (auto globalParam = as<IRGlobalParam>(opaqueValue))
         {
             // The simple/base case is when we have a global shader
             // parameter that has layout information attached.

source/slang/slang-ir-specialize-resources.cpp

@@ -1359,6 +1359,10 @@ bool isIllegalGLSLParameterType(IRType* type)
         return true;
     if (as<IRDynamicResourceType>(type))
         return true;
+    if (as<IRHLSLInputPatchType>(type))
+        return true;
+    if (as<IRHLSLOutputPatchType>(type))
+        return true;
     return false;
 }
 

source/slang/slang-ir.cpp

@@ -5279,6 +5279,11 @@ IRInst* IRBuilder::emitElementAddress(IRInst* basePtr, IRInst* index)
         SLANG_ASSERT(as<IRIntLit>(index));
         type = (IRType*)tupleType->getOperand(getIntVal(index));
     }
+    else if (auto hlslInputPatchType = as<IRHLSLInputPatchType>(valueType))
+    {
+        type = hlslInputPatchType->getElementType();
+    }
+
     SLANG_RELEASE_ASSERT(type);
     auto inst = createInst<IRGetElementPtr>(
         this,

tests/expected-failure-github.txt

@@ -1,9 +1,7 @@
-tests/language-feature/spirv-asm/imageoperands-warning.slang (vk)
 tests/language-feature/saturated-cooperation/simple.slang (vk)
 tests/language-feature/saturated-cooperation/fuse3.slang (vk)
 tests/language-feature/saturated-cooperation/fuse-product.slang (vk)
 tests/language-feature/saturated-cooperation/fuse.slang (vk)
-tests/bugs/byte-address-buffer-interlocked-add-f32.slang (vk)
 tests/render/render0.hlsl (mtl)
 tests/render/multiple-stage-io-locations.slang (mtl)
 tests/render/nointerpolation.hlsl (mtl)

tests/gpu-feature/texture/query/footprint/nv-shader-texture-footprint-array.slang

@@ -0,0 +1,76 @@
+//DISABLE_TEST:SIMPLE:-target dxil-assembly -entry fragmentMain -profile sm_6_5 -stage fragment -DNV_SHADER_EXTN_SLOT=u0 
+
+//TEST:SIMPLE(filecheck=SPIRV):-target spirv-assembly -entry fragmentMain -stage fragment
+//TEST:SIMPLE(filecheck=DXIL):-target dxil-assembly -entry fragmentMain -profile sm_6_5 -stage fragment -DNV_SHADER_EXTN_SLOT=u0 
+//TEST:SIMPLE(filecheck=HLSL):-target hlsl -entry fragmentMain -stage fragment
+
+//DISABLED_TEST:SIMPLE:-target spirv-assembly -entry fragmentMain -stage fragment
+//DISABLED_TEST:SIMPLE:-target dxil-assembly -entry fragmentMain -stage fragment
+//DISABLED_TEST:SIMPLE:-target hlsl -entry fragmentMain -stage fragment
+
+uniform Texture2D textures[] : register(t2, space10);
+uniform SamplerState sampler;
+uniform RWStructuredBuffer<uint> outputBuffer;
+
+static Texture2D _getBindlessTexture2d(uint texIdx)
+{
+    return textures[NonUniformResourceIndex(texIdx)];
+}
+
+void accumulate(inout uint r, uint u)
+{
+    r = r ^ u;
+}
+
+void accumulate(inout uint r, bool b)
+{
+    accumulate(r, uint(b));
+}
+
+void accumulate(inout uint r, uint2 u)
+{
+    accumulate(r, u.x);
+    accumulate(r, u.y);
+}
+
+void accumulate(inout uint r, uint3 u)
+{
+    accumulate(r, u.x);
+    accumulate(r, u.y);
+    accumulate(r, u.z);
+}
+
+void accumulate(inout uint r, TextureFootprint2D f)
+{
+    accumulate(r, f.anchor);
+    accumulate(r, f.offset);
+    accumulate(r, f.mask);
+    accumulate(r, f.lod);
+    accumulate(r, f.granularity);
+    accumulate(r, f.isSingleLevel);
+}
+
+cbuffer Uniforms
+{
+    uniform float2 coords;
+    uniform uint granularity;
+};
+
+void fragmentMain(
+    float v : VARYING)
+{
+    uint index = uint(v);
+    uint r = 0;
+
+    accumulate(r, _getBindlessTexture2d(index).queryFootprintCoarse(granularity, sampler, coords));
+
+// SPIRV: Extension  "SPV_NV_shader_image_footprint"
+// SPIRV: ImageSampleFootprintNV
+
+// DXIL: struct struct.NvShaderExtnStruct
+
+// HLSL: NvFootprintCoarse
+
+    outputBuffer[index] = r;
+}
+

tests/hlsl/float-literal-suffix.slang

@@ -0,0 +1,10 @@
+//TEST:SIMPLE(filecheck=HLSL):-target hlsl -profile ps_6_6 -entry fragmentMain
+//TEST:SIMPLE(filecheck=DXIL):-target dxil -profile ps_6_6 -entry fragmentMain
+
+float4 fragmentMain(float2 uv : TEXCOORD) : SV_Target
+{
+    //HLSL:, ddx({{.*}}1.5f)
+    //DXIL: = call float @dx.op.unary.f32(i32 {{.*}} ; DerivCoarseX(value)
+    float val = 1.5;
+    return float4(1.0, ddx(val), 0.0, 1.0);
+}

tests/spirv/tessellation-patch-as-argument.slang

@@ -0,0 +1,100 @@
+//TEST:SIMPLE(filecheck=HULL):   -target spirv -stage hull   -entry hullMain
+//TEST:SIMPLE(filecheck=DOMAIN): -target spirv -stage domain -entry domainMain
+
+// Testing if `InputPatch` and `OutputPatch` can be used as function arguments.
+// `[ForceInline]` can be used to workaround but it should work without it.
+
+// HULL: OpEntryPoint TessellationControl %hullMain
+// HULL: = OpVariable %{{[a-zA-Z_0-9]*}} Input
+
+// DOMAIN: OpEntryPoint TessellationEvaluation %domainMain
+// DOMAIN: = OpVariable %{{[a-zA-Z_0-9]*}} Output
+
+struct VS_OUT
+{
+    float3 position : POSITION;
+};
+
+struct HS_OUT
+{
+    float3 position : POSITION;
+};
+
+struct HSC_OUT
+{
+    float EdgeTessFactor[4] : SV_TessFactor;
+    float InsideTessFactor[2] : SV_InsideTessFactor;
+};
+
+struct DS_OUT
+{
+    float4 position : SV_Position;
+};
+
+
+VS_OUT GetInputPatch(InputPatch<VS_OUT, 4> patch, int index)
+{
+    return patch[index];
+}
+
+// Hull Shader (HS)
+[domain("quad")]
+[partitioning("integer")]
+[outputtopology("triangle_cw")]
+[outputcontrolpoints(4)]
+[patchconstantfunc("constants")]
+HS_OUT hullMain(InputPatch<VS_OUT, 4> patch, uint i : SV_OutputControlPointID)
+{
+    HS_OUT o;
+    o.position = patch[i].position;
+    return o;
+}
+
+HSC_OUT constants(InputPatch<VS_OUT, 4> patch)
+{
+    float3 p0 = GetInputPatch(patch, 0).position;
+    float3 p1 = patch[1].position;
+    float3 p2 = patch[2].position;
+    float3 p3 = patch[3].position;
+
+    HSC_OUT o;
+    o.EdgeTessFactor[0] = dot(p0, p1);
+    o.EdgeTessFactor[1] = dot(p0, p3);
+    o.EdgeTessFactor[2] = dot(p2, p3);
+    o.EdgeTessFactor[3] = dot(p1, p2);
+    o.InsideTessFactor[0] = lerp(o.EdgeTessFactor[1], o.EdgeTessFactor[3], 0.5);
+    o.InsideTessFactor[1] = lerp(o.EdgeTessFactor[0], o.EdgeTessFactor[2], 0.5);
+    return o;
+}
+
+HS_OUT GetOutputPatch(const OutputPatch<HS_OUT, 4> patch, int index)
+{
+    return patch[index];
+}
+
+[domain("quad")]
+DS_OUT domainMain(
+    float2 uv : SV_DomainLocation,      // Tessellated coordinates (u, v)
+    const OutputPatch<HS_OUT, 4> patch, // Control points from the hull shader
+    const HSC_OUT patchConstants        // Patch constants calculated by the hull shader
+)
+{
+    DS_OUT o;
+
+    // Interpolate the position of the tessellated point within the patch
+    float3 p0 = GetOutputPatch(patch, 0).position;
+    float3 p1 = patch[1].position;
+    float3 p2 = patch[2].position;
+    float3 p3 = patch[3].position;
+
+    // Bilinear interpolation of the position in the quad
+    float3 interpolatedPosition =
+          p0 * (1 - uv.x) * (1 - uv.y)
+        + p1 * uv.x * (1 - uv.y)
+        + p3 * uv.x * uv.y
+        + p2 * (1 - uv.x) * uv.y;
+
+    // Output final position in clip space
+    o.position = float4(interpolatedPosition, 1.0);
+    return o;
+}