Skip to content

[HLSL] Implement ddx/ddy_fine intrinsics #168874

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[HLSL] Implement ddx/ddy_fine intrinsics #168874

wants to merge 1 commit into from

Conversation

@Alexander-Johnston
Copy link
Contributor

@Alexander-Johnston Alexander-Johnston commented Nov 20, 2025

Implements the HLSL ddx_fine and ddy_fine intrinsics.
For the SPIRV backend the intrinsics are ensured to be unavailable in opencl (as they require fragment execution stage).

Closes #99098
Closes #99101

@llvmbot llvmbot added backend:X86 clang:frontend Language frontend issues, e.g. anything involving "Sema" clang:headers Headers provided by Clang, e.g. for intrinsics clang:codegen IR generation bugs: mangling, exceptions, etc. backend:DirectX HLSL HLSL Language Support backend:SPIR-V llvm:ir labels Nov 20, 2025
@llvmbot
Copy link
Member

llvmbot commented Nov 20, 2025 •
edited
Loading

@llvm/pr-subscribers-backend-directx

@llvm/pr-subscribers-clang-codegen

Author: Alexander Johnston (Alexander-Johnston)

Changes

Implements the HLSL ddx_fine and ddy_fine intrinsics.
For the SPIRV backend the intrinsics are ensured to be unavailable in opencl (as they require fragment execution stage).

Closes #99098
Closes #99101

Patch is 39.46 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/168874.diff

25 Files Affected:

  • (modified) clang/include/clang/Basic/Builtins.td (+12)
  • (modified) clang/lib/CodeGen/CGHLSLBuiltins.cpp (+18)
  • (modified) clang/lib/CodeGen/CGHLSLRuntime.h (+2)
  • (modified) clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h (+68)
  • (modified) clang/lib/Sema/SemaHLSL.cpp (+3-1)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl (+86)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl (+86)
  • (added) clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl (+22)
  • (added) clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl (+22)
  • (modified) llvm/include/llvm/IR/IntrinsicsDirectX.td (+2)
  • (modified) llvm/include/llvm/IR/IntrinsicsSPIRV.td (+2)
  • (modified) llvm/lib/Target/DirectX/DXIL.td (+18)
  • (modified) llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp (+2)
  • (modified) llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp (+4)
  • (modified) llvm/lib/Target/SPIRV/SPIRVModuleAnalysis.cpp (+3-1)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine.ll (+40)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine.ll (+40)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddx_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddy_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddx_fine-error.ll (+12)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddy_fine-error.ll (+12)
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 502382a069856..64de76a01f335 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -5265,6 +5265,18 @@ def HLSLDdyCoarse : LangBuiltin<"HLSL_LANG"> {
   let Prototype = "void(...)";
 }
 
+def HLSLDdxFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddx_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
+def HLSLDdyFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddy_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
 // Builtins for XRay.
 def XRayCustomEvent : Builtin {
   let Spellings = ["__xray_customevent"];
diff --git a/clang/lib/CodeGen/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
index 12d9a98915ce3..bfa9a53b7804e 100644
--- a/clang/lib/CodeGen/CGHLSLBuiltins.cpp
+++ b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
@@ -942,6 +942,24 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
                                    ArrayRef<Value *>{Op0}, nullptr,
                                    "hlsl.ddy.coarse");
   }
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddx_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdxFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddx.fine");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddy_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdyFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddy.fine");
+  }
   case Builtin::BI__builtin_get_spirv_spec_constant_bool:
   case Builtin::BI__builtin_get_spirv_spec_constant_short:
   case Builtin::BI__builtin_get_spirv_spec_constant_ushort:
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index c883282a8d9c8..2ad39c91280a0 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -171,6 +171,8 @@ class CGHLSLRuntime {
   GENERATE_HLSL_INTRINSIC_FUNCTION(GetDimensionsX, resource_getdimensions_x)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdxCoarse, ddx_coarse)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdyCoarse, ddy_coarse)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdxFine, ddx_fine)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdyFine, ddy_fine)
 
   //===----------------------------------------------------------------------===//
   // End of reserved area for HLSL intrinsic getters.
diff --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index 38b95ee90736a..f58150ed61106 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -3014,5 +3014,73 @@ float3 ddy_coarse(float3);
 _HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_coarse)
 float4 ddy_coarse(float4);
 
+//===----------------------------------------------------------------------===//
+// ddx_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddx_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space x-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half ddx_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half2 ddx_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half3 ddx_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half4 ddx_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float ddx_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float2 ddx_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float3 ddx_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float4 ddx_fine(float4);
+
+//===----------------------------------------------------------------------===//
+// ddy_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddy_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space y-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half ddy_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half2 ddy_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half3 ddy_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half4 ddy_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float ddy_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float2 ddy_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float3 ddy_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float4 ddy_fine(float4);
+
 } // namespace hlsl
 #endif //_HLSL_HLSL_ALIAS_INTRINSICS_H_
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index c5666941fd36a..cc2939314b0eb 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -3211,7 +3211,9 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
   case Builtin::BI__builtin_hlsl_elementwise_rsqrt:
   case Builtin::BI__builtin_hlsl_elementwise_frac:
   case Builtin::BI__builtin_hlsl_elementwise_ddx_coarse:
-  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse: {
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse:
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine:
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
     if (SemaRef.checkArgCount(TheCall, 1))
       return true;
     if (CheckAllArgTypesAreCorrect(&SemaRef, TheCall,
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
new file mode 100644
index 0000000000000..69f7ab3c6ce62
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
new file mode 100644
index 0000000000000..2630260abcb43
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.dx.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.spv.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddx.fine
+half2 test_f16_ddx_fine2(half2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.dx.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.spv.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddx.fine
+half3 test_f16_ddx_fine3(half3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.dx.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.spv.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddx.fine
+half4 test_f16_ddx_fine4(half4 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.dx.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.spv.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddx.fine
+float2 test_f32_ddx_fine2(float2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.dx.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.spv.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddx.fine
+float3 test_f32_ddx_fine3(float3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.dx.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.spv.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddx.fine
+float4 test_f32_ddx_fine4(float4 val) {
+    return ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
new file mode 100644
index 0000000000000..00630721ceb66
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
new file mode 100644
index 0000000000000..7e32ee29e767d
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.dx.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.spv.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddy.fine
+half2 test_f16_ddy_fine2(half2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.dx.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.spv.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddy.fine
+half3 test_f16_ddy_fine3(half3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.dx.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.spv.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddy.fine
+half4 test_f16_ddy_fine4(half4 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.dx.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.spv.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddy.fine
+float2 test_f32_ddy_fine2(float2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.dx.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.spv.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddy.fine
+float3 test_f32_ddy_fine3(float3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.dx.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.spv.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddy.fine
+float4 test_f32_ddy_fine4(float4 val) {
+    return ddy_fine(val);
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
new file mode 100644
index 0000000000000..71196943c322a
--- /dev/null
+++ b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
@@ -0,0 +1,22 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.6-library %s -fnative-half-type -verify
+// RUN: %clang_cc1 -triple spirv-unknown-vulkan1.3-library %s -fnative-half-type -verify
+
+float no_arg() {
+  return __builtin_hlsl_elementwise_ddx_fine();
+  // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float too_many_args(float val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val, val);
+  // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+float test_integer_scalar_input(int val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'int')}}
+}
+
+double test_double_scalar_input(double val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'double')}}
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl
new file mode 100644
index 0000000000000..e17d5f36832b2
--- /dev/null...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Nov 20, 2025

@llvm/pr-subscribers-backend-spir-v

Author: Alexander Johnston (Alexander-Johnston)

Changes

Implements the HLSL ddx_fine and ddy_fine intrinsics.
For the SPIRV backend the intrinsics are ensured to be unavailable in opencl (as they require fragment execution stage).

Closes #99098
Closes #99101

Patch is 39.46 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/168874.diff

25 Files Affected:

  • (modified) clang/include/clang/Basic/Builtins.td (+12)
  • (modified) clang/lib/CodeGen/CGHLSLBuiltins.cpp (+18)
  • (modified) clang/lib/CodeGen/CGHLSLRuntime.h (+2)
  • (modified) clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h (+68)
  • (modified) clang/lib/Sema/SemaHLSL.cpp (+3-1)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl (+86)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl (+86)
  • (added) clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl (+22)
  • (added) clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl (+22)
  • (modified) llvm/include/llvm/IR/IntrinsicsDirectX.td (+2)
  • (modified) llvm/include/llvm/IR/IntrinsicsSPIRV.td (+2)
  • (modified) llvm/lib/Target/DirectX/DXIL.td (+18)
  • (modified) llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp (+2)
  • (modified) llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp (+4)
  • (modified) llvm/lib/Target/SPIRV/SPIRVModuleAnalysis.cpp (+3-1)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine.ll (+40)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine.ll (+40)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddx_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddy_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddx_fine-error.ll (+12)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddy_fine-error.ll (+12)
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 502382a069856..64de76a01f335 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -5265,6 +5265,18 @@ def HLSLDdyCoarse : LangBuiltin<"HLSL_LANG"> {
   let Prototype = "void(...)";
 }
 
+def HLSLDdxFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddx_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
+def HLSLDdyFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddy_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
 // Builtins for XRay.
 def XRayCustomEvent : Builtin {
   let Spellings = ["__xray_customevent"];
diff --git a/clang/lib/CodeGen/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
index 12d9a98915ce3..bfa9a53b7804e 100644
--- a/clang/lib/CodeGen/CGHLSLBuiltins.cpp
+++ b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
@@ -942,6 +942,24 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
                                    ArrayRef<Value *>{Op0}, nullptr,
                                    "hlsl.ddy.coarse");
   }
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddx_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdxFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddx.fine");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddy_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdyFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddy.fine");
+  }
   case Builtin::BI__builtin_get_spirv_spec_constant_bool:
   case Builtin::BI__builtin_get_spirv_spec_constant_short:
   case Builtin::BI__builtin_get_spirv_spec_constant_ushort:
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index c883282a8d9c8..2ad39c91280a0 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -171,6 +171,8 @@ class CGHLSLRuntime {
   GENERATE_HLSL_INTRINSIC_FUNCTION(GetDimensionsX, resource_getdimensions_x)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdxCoarse, ddx_coarse)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdyCoarse, ddy_coarse)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdxFine, ddx_fine)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdyFine, ddy_fine)
 
   //===----------------------------------------------------------------------===//
   // End of reserved area for HLSL intrinsic getters.
diff --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index 38b95ee90736a..f58150ed61106 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -3014,5 +3014,73 @@ float3 ddy_coarse(float3);
 _HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_coarse)
 float4 ddy_coarse(float4);
 
+//===----------------------------------------------------------------------===//
+// ddx_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddx_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space x-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half ddx_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half2 ddx_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half3 ddx_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half4 ddx_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float ddx_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float2 ddx_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float3 ddx_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float4 ddx_fine(float4);
+
+//===----------------------------------------------------------------------===//
+// ddy_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddy_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space y-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half ddy_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half2 ddy_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half3 ddy_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half4 ddy_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float ddy_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float2 ddy_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float3 ddy_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float4 ddy_fine(float4);
+
 } // namespace hlsl
 #endif //_HLSL_HLSL_ALIAS_INTRINSICS_H_
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index c5666941fd36a..cc2939314b0eb 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -3211,7 +3211,9 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
   case Builtin::BI__builtin_hlsl_elementwise_rsqrt:
   case Builtin::BI__builtin_hlsl_elementwise_frac:
   case Builtin::BI__builtin_hlsl_elementwise_ddx_coarse:
-  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse: {
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse:
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine:
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
     if (SemaRef.checkArgCount(TheCall, 1))
       return true;
     if (CheckAllArgTypesAreCorrect(&SemaRef, TheCall,
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
new file mode 100644
index 0000000000000..69f7ab3c6ce62
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
new file mode 100644
index 0000000000000..2630260abcb43
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.dx.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.spv.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddx.fine
+half2 test_f16_ddx_fine2(half2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.dx.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.spv.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddx.fine
+half3 test_f16_ddx_fine3(half3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.dx.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.spv.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddx.fine
+half4 test_f16_ddx_fine4(half4 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.dx.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.spv.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddx.fine
+float2 test_f32_ddx_fine2(float2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.dx.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.spv.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddx.fine
+float3 test_f32_ddx_fine3(float3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.dx.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.spv.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddx.fine
+float4 test_f32_ddx_fine4(float4 val) {
+    return ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
new file mode 100644
index 0000000000000..00630721ceb66
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
new file mode 100644
index 0000000000000..7e32ee29e767d
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.dx.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.spv.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddy.fine
+half2 test_f16_ddy_fine2(half2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.dx.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.spv.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddy.fine
+half3 test_f16_ddy_fine3(half3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.dx.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.spv.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddy.fine
+half4 test_f16_ddy_fine4(half4 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.dx.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.spv.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddy.fine
+float2 test_f32_ddy_fine2(float2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.dx.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.spv.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddy.fine
+float3 test_f32_ddy_fine3(float3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.dx.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.spv.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddy.fine
+float4 test_f32_ddy_fine4(float4 val) {
+    return ddy_fine(val);
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
new file mode 100644
index 0000000000000..71196943c322a
--- /dev/null
+++ b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
@@ -0,0 +1,22 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.6-library %s -fnative-half-type -verify
+// RUN: %clang_cc1 -triple spirv-unknown-vulkan1.3-library %s -fnative-half-type -verify
+
+float no_arg() {
+  return __builtin_hlsl_elementwise_ddx_fine();
+  // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float too_many_args(float val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val, val);
+  // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+float test_integer_scalar_input(int val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'int')}}
+}
+
+double test_double_scalar_input(double val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'double')}}
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl
new file mode 100644
index 0000000000000..e17d5f36832b2
--- /dev/null...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Nov 20, 2025

@llvm/pr-subscribers-backend-x86

Author: Alexander Johnston (Alexander-Johnston)

Changes

Implements the HLSL ddx_fine and ddy_fine intrinsics.
For the SPIRV backend the intrinsics are ensured to be unavailable in opencl (as they require fragment execution stage).

Closes #99098
Closes #99101

Patch is 39.46 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/168874.diff

25 Files Affected:

  • (modified) clang/include/clang/Basic/Builtins.td (+12)
  • (modified) clang/lib/CodeGen/CGHLSLBuiltins.cpp (+18)
  • (modified) clang/lib/CodeGen/CGHLSLRuntime.h (+2)
  • (modified) clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h (+68)
  • (modified) clang/lib/Sema/SemaHLSL.cpp (+3-1)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl (+86)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl (+86)
  • (added) clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl (+22)
  • (added) clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl (+22)
  • (modified) llvm/include/llvm/IR/IntrinsicsDirectX.td (+2)
  • (modified) llvm/include/llvm/IR/IntrinsicsSPIRV.td (+2)
  • (modified) llvm/lib/Target/DirectX/DXIL.td (+18)
  • (modified) llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp (+2)
  • (modified) llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp (+4)
  • (modified) llvm/lib/Target/SPIRV/SPIRVModuleAnalysis.cpp (+3-1)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine.ll (+40)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine.ll (+40)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddx_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddy_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddx_fine-error.ll (+12)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddy_fine-error.ll (+12)
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 502382a069856..64de76a01f335 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -5265,6 +5265,18 @@ def HLSLDdyCoarse : LangBuiltin<"HLSL_LANG"> {
   let Prototype = "void(...)";
 }
 
+def HLSLDdxFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddx_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
+def HLSLDdyFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddy_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
 // Builtins for XRay.
 def XRayCustomEvent : Builtin {
   let Spellings = ["__xray_customevent"];
diff --git a/clang/lib/CodeGen/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
index 12d9a98915ce3..bfa9a53b7804e 100644
--- a/clang/lib/CodeGen/CGHLSLBuiltins.cpp
+++ b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
@@ -942,6 +942,24 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
                                    ArrayRef<Value *>{Op0}, nullptr,
                                    "hlsl.ddy.coarse");
   }
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddx_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdxFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddx.fine");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddy_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdyFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddy.fine");
+  }
   case Builtin::BI__builtin_get_spirv_spec_constant_bool:
   case Builtin::BI__builtin_get_spirv_spec_constant_short:
   case Builtin::BI__builtin_get_spirv_spec_constant_ushort:
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index c883282a8d9c8..2ad39c91280a0 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -171,6 +171,8 @@ class CGHLSLRuntime {
   GENERATE_HLSL_INTRINSIC_FUNCTION(GetDimensionsX, resource_getdimensions_x)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdxCoarse, ddx_coarse)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdyCoarse, ddy_coarse)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdxFine, ddx_fine)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdyFine, ddy_fine)
 
   //===----------------------------------------------------------------------===//
   // End of reserved area for HLSL intrinsic getters.
diff --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index 38b95ee90736a..f58150ed61106 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -3014,5 +3014,73 @@ float3 ddy_coarse(float3);
 _HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_coarse)
 float4 ddy_coarse(float4);
 
+//===----------------------------------------------------------------------===//
+// ddx_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddx_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space x-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half ddx_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half2 ddx_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half3 ddx_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half4 ddx_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float ddx_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float2 ddx_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float3 ddx_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float4 ddx_fine(float4);
+
+//===----------------------------------------------------------------------===//
+// ddy_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddy_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space y-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half ddy_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half2 ddy_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half3 ddy_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half4 ddy_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float ddy_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float2 ddy_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float3 ddy_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float4 ddy_fine(float4);
+
 } // namespace hlsl
 #endif //_HLSL_HLSL_ALIAS_INTRINSICS_H_
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index c5666941fd36a..cc2939314b0eb 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -3211,7 +3211,9 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
   case Builtin::BI__builtin_hlsl_elementwise_rsqrt:
   case Builtin::BI__builtin_hlsl_elementwise_frac:
   case Builtin::BI__builtin_hlsl_elementwise_ddx_coarse:
-  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse: {
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse:
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine:
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
     if (SemaRef.checkArgCount(TheCall, 1))
       return true;
     if (CheckAllArgTypesAreCorrect(&SemaRef, TheCall,
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
new file mode 100644
index 0000000000000..69f7ab3c6ce62
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
new file mode 100644
index 0000000000000..2630260abcb43
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.dx.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.spv.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddx.fine
+half2 test_f16_ddx_fine2(half2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.dx.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.spv.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddx.fine
+half3 test_f16_ddx_fine3(half3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.dx.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.spv.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddx.fine
+half4 test_f16_ddx_fine4(half4 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.dx.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.spv.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddx.fine
+float2 test_f32_ddx_fine2(float2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.dx.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.spv.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddx.fine
+float3 test_f32_ddx_fine3(float3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.dx.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.spv.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddx.fine
+float4 test_f32_ddx_fine4(float4 val) {
+    return ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
new file mode 100644
index 0000000000000..00630721ceb66
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
new file mode 100644
index 0000000000000..7e32ee29e767d
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.dx.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.spv.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddy.fine
+half2 test_f16_ddy_fine2(half2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.dx.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.spv.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddy.fine
+half3 test_f16_ddy_fine3(half3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.dx.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.spv.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddy.fine
+half4 test_f16_ddy_fine4(half4 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.dx.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.spv.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddy.fine
+float2 test_f32_ddy_fine2(float2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.dx.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.spv.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddy.fine
+float3 test_f32_ddy_fine3(float3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.dx.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.spv.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddy.fine
+float4 test_f32_ddy_fine4(float4 val) {
+    return ddy_fine(val);
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
new file mode 100644
index 0000000000000..71196943c322a
--- /dev/null
+++ b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
@@ -0,0 +1,22 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.6-library %s -fnative-half-type -verify
+// RUN: %clang_cc1 -triple spirv-unknown-vulkan1.3-library %s -fnative-half-type -verify
+
+float no_arg() {
+  return __builtin_hlsl_elementwise_ddx_fine();
+  // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float too_many_args(float val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val, val);
+  // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+float test_integer_scalar_input(int val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'int')}}
+}
+
+double test_double_scalar_input(double val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'double')}}
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl
new file mode 100644
index 0000000000000..e17d5f36832b2
--- /dev/null...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Nov 20, 2025

@llvm/pr-subscribers-llvm-ir

Author: Alexander Johnston (Alexander-Johnston)

Changes

Implements the HLSL ddx_fine and ddy_fine intrinsics.
For the SPIRV backend the intrinsics are ensured to be unavailable in opencl (as they require fragment execution stage).

Closes #99098
Closes #99101

Patch is 39.46 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/168874.diff

25 Files Affected:

  • (modified) clang/include/clang/Basic/Builtins.td (+12)
  • (modified) clang/lib/CodeGen/CGHLSLBuiltins.cpp (+18)
  • (modified) clang/lib/CodeGen/CGHLSLRuntime.h (+2)
  • (modified) clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h (+68)
  • (modified) clang/lib/Sema/SemaHLSL.cpp (+3-1)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl (+86)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl (+86)
  • (added) clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl (+22)
  • (added) clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl (+22)
  • (modified) llvm/include/llvm/IR/IntrinsicsDirectX.td (+2)
  • (modified) llvm/include/llvm/IR/IntrinsicsSPIRV.td (+2)
  • (modified) llvm/lib/Target/DirectX/DXIL.td (+18)
  • (modified) llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp (+2)
  • (modified) llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp (+4)
  • (modified) llvm/lib/Target/SPIRV/SPIRVModuleAnalysis.cpp (+3-1)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine.ll (+40)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine.ll (+40)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddx_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddy_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddx_fine-error.ll (+12)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddy_fine-error.ll (+12)
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 502382a069856..64de76a01f335 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -5265,6 +5265,18 @@ def HLSLDdyCoarse : LangBuiltin<"HLSL_LANG"> {
   let Prototype = "void(...)";
 }
 
+def HLSLDdxFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddx_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
+def HLSLDdyFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddy_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
 // Builtins for XRay.
 def XRayCustomEvent : Builtin {
   let Spellings = ["__xray_customevent"];
diff --git a/clang/lib/CodeGen/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
index 12d9a98915ce3..bfa9a53b7804e 100644
--- a/clang/lib/CodeGen/CGHLSLBuiltins.cpp
+++ b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
@@ -942,6 +942,24 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
                                    ArrayRef<Value *>{Op0}, nullptr,
                                    "hlsl.ddy.coarse");
   }
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddx_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdxFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddx.fine");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddy_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdyFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddy.fine");
+  }
   case Builtin::BI__builtin_get_spirv_spec_constant_bool:
   case Builtin::BI__builtin_get_spirv_spec_constant_short:
   case Builtin::BI__builtin_get_spirv_spec_constant_ushort:
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index c883282a8d9c8..2ad39c91280a0 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -171,6 +171,8 @@ class CGHLSLRuntime {
   GENERATE_HLSL_INTRINSIC_FUNCTION(GetDimensionsX, resource_getdimensions_x)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdxCoarse, ddx_coarse)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdyCoarse, ddy_coarse)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdxFine, ddx_fine)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdyFine, ddy_fine)
 
   //===----------------------------------------------------------------------===//
   // End of reserved area for HLSL intrinsic getters.
diff --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index 38b95ee90736a..f58150ed61106 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -3014,5 +3014,73 @@ float3 ddy_coarse(float3);
 _HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_coarse)
 float4 ddy_coarse(float4);
 
+//===----------------------------------------------------------------------===//
+// ddx_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddx_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space x-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half ddx_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half2 ddx_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half3 ddx_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half4 ddx_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float ddx_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float2 ddx_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float3 ddx_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float4 ddx_fine(float4);
+
+//===----------------------------------------------------------------------===//
+// ddy_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddy_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space y-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half ddy_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half2 ddy_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half3 ddy_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half4 ddy_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float ddy_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float2 ddy_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float3 ddy_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float4 ddy_fine(float4);
+
 } // namespace hlsl
 #endif //_HLSL_HLSL_ALIAS_INTRINSICS_H_
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index c5666941fd36a..cc2939314b0eb 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -3211,7 +3211,9 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
   case Builtin::BI__builtin_hlsl_elementwise_rsqrt:
   case Builtin::BI__builtin_hlsl_elementwise_frac:
   case Builtin::BI__builtin_hlsl_elementwise_ddx_coarse:
-  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse: {
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse:
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine:
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
     if (SemaRef.checkArgCount(TheCall, 1))
       return true;
     if (CheckAllArgTypesAreCorrect(&SemaRef, TheCall,
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
new file mode 100644
index 0000000000000..69f7ab3c6ce62
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
new file mode 100644
index 0000000000000..2630260abcb43
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.dx.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.spv.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddx.fine
+half2 test_f16_ddx_fine2(half2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.dx.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.spv.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddx.fine
+half3 test_f16_ddx_fine3(half3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.dx.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.spv.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddx.fine
+half4 test_f16_ddx_fine4(half4 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.dx.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.spv.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddx.fine
+float2 test_f32_ddx_fine2(float2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.dx.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.spv.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddx.fine
+float3 test_f32_ddx_fine3(float3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.dx.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.spv.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddx.fine
+float4 test_f32_ddx_fine4(float4 val) {
+    return ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
new file mode 100644
index 0000000000000..00630721ceb66
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
new file mode 100644
index 0000000000000..7e32ee29e767d
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.dx.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.spv.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddy.fine
+half2 test_f16_ddy_fine2(half2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.dx.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.spv.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddy.fine
+half3 test_f16_ddy_fine3(half3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.dx.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.spv.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddy.fine
+half4 test_f16_ddy_fine4(half4 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.dx.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.spv.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddy.fine
+float2 test_f32_ddy_fine2(float2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.dx.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.spv.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddy.fine
+float3 test_f32_ddy_fine3(float3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.dx.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.spv.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddy.fine
+float4 test_f32_ddy_fine4(float4 val) {
+    return ddy_fine(val);
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
new file mode 100644
index 0000000000000..71196943c322a
--- /dev/null
+++ b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
@@ -0,0 +1,22 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.6-library %s -fnative-half-type -verify
+// RUN: %clang_cc1 -triple spirv-unknown-vulkan1.3-library %s -fnative-half-type -verify
+
+float no_arg() {
+  return __builtin_hlsl_elementwise_ddx_fine();
+  // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float too_many_args(float val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val, val);
+  // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+float test_integer_scalar_input(int val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'int')}}
+}
+
+double test_double_scalar_input(double val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'double')}}
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl
new file mode 100644
index 0000000000000..e17d5f36832b2
--- /dev/null...
[truncated]

@llvmbot
Copy link
Member

llvmbot commented Nov 20, 2025

@llvm/pr-subscribers-hlsl

Author: Alexander Johnston (Alexander-Johnston)

Changes

Implements the HLSL ddx_fine and ddy_fine intrinsics.
For the SPIRV backend the intrinsics are ensured to be unavailable in opencl (as they require fragment execution stage).

Closes #99098
Closes #99101

Patch is 39.46 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/168874.diff

25 Files Affected:

  • (modified) clang/include/clang/Basic/Builtins.td (+12)
  • (modified) clang/lib/CodeGen/CGHLSLBuiltins.cpp (+18)
  • (modified) clang/lib/CodeGen/CGHLSLRuntime.h (+2)
  • (modified) clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h (+68)
  • (modified) clang/lib/Sema/SemaHLSL.cpp (+3-1)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl (+86)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl (+26)
  • (added) clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl (+86)
  • (added) clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl (+22)
  • (added) clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl (+22)
  • (modified) llvm/include/llvm/IR/IntrinsicsDirectX.td (+2)
  • (modified) llvm/include/llvm/IR/IntrinsicsSPIRV.td (+2)
  • (modified) llvm/lib/Target/DirectX/DXIL.td (+18)
  • (modified) llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp (+2)
  • (modified) llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp (+4)
  • (modified) llvm/lib/Target/SPIRV/SPIRVModuleAnalysis.cpp (+3-1)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddx_fine.ll (+40)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine-errors.ll (+29)
  • (added) llvm/test/CodeGen/DirectX/ddy_fine.ll (+40)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddx_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/ddy_fine.ll (+47)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddx_fine-error.ll (+12)
  • (added) llvm/test/CodeGen/SPIRV/opencl/ddy_fine-error.ll (+12)
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 502382a069856..64de76a01f335 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -5265,6 +5265,18 @@ def HLSLDdyCoarse : LangBuiltin<"HLSL_LANG"> {
   let Prototype = "void(...)";
 }
 
+def HLSLDdxFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddx_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
+def HLSLDdyFine : LangBuiltin<"HLSL_LANG"> {
+  let Spellings = ["__builtin_hlsl_elementwise_ddy_fine"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
 // Builtins for XRay.
 def XRayCustomEvent : Builtin {
   let Spellings = ["__xray_customevent"];
diff --git a/clang/lib/CodeGen/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
index 12d9a98915ce3..bfa9a53b7804e 100644
--- a/clang/lib/CodeGen/CGHLSLBuiltins.cpp
+++ b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
@@ -942,6 +942,24 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
                                    ArrayRef<Value *>{Op0}, nullptr,
                                    "hlsl.ddy.coarse");
   }
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddx_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdxFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddx.fine");
+  }
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
+    Value *Op0 = EmitScalarExpr(E->getArg(0));
+    if (!E->getArg(0)->getType()->hasFloatingRepresentation())
+      llvm_unreachable("ddy_fine operand must have a float representation");
+    Intrinsic::ID ID = CGM.getHLSLRuntime().getDdyFineIntrinsic();
+    return Builder.CreateIntrinsic(/*ReturnType=*/Op0->getType(), ID,
+                                   ArrayRef<Value *>{Op0}, nullptr,
+                                   "hlsl.ddy.fine");
+  }
   case Builtin::BI__builtin_get_spirv_spec_constant_bool:
   case Builtin::BI__builtin_get_spirv_spec_constant_short:
   case Builtin::BI__builtin_get_spirv_spec_constant_ushort:
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index c883282a8d9c8..2ad39c91280a0 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -171,6 +171,8 @@ class CGHLSLRuntime {
   GENERATE_HLSL_INTRINSIC_FUNCTION(GetDimensionsX, resource_getdimensions_x)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdxCoarse, ddx_coarse)
   GENERATE_HLSL_INTRINSIC_FUNCTION(DdyCoarse, ddy_coarse)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdxFine, ddx_fine)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(DdyFine, ddy_fine)
 
   //===----------------------------------------------------------------------===//
   // End of reserved area for HLSL intrinsic getters.
diff --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index 38b95ee90736a..f58150ed61106 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -3014,5 +3014,73 @@ float3 ddy_coarse(float3);
 _HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_coarse)
 float4 ddy_coarse(float4);
 
+//===----------------------------------------------------------------------===//
+// ddx_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddx_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space x-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half ddx_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half2 ddx_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half3 ddx_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+half4 ddx_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float ddx_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float2 ddx_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float3 ddx_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddx_fine)
+float4 ddx_fine(float4);
+
+//===----------------------------------------------------------------------===//
+// ddy_fine builtin
+//===----------------------------------------------------------------------===//
+
+/// \fn T ddy_fine(T value)
+/// \brief Computes a high precision partial derivative with respect to the
+/// screen-space y-coordinate.
+/// \param value The input value.
+///
+/// The return value is a floating point scalar or vector containing the high
+/// prevision partial derivative of the input value.
+
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half ddy_fine(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half2 ddy_fine(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half3 ddy_fine(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+half4 ddy_fine(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float ddy_fine(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float2 ddy_fine(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float3 ddy_fine(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_ddy_fine)
+float4 ddy_fine(float4);
+
 } // namespace hlsl
 #endif //_HLSL_HLSL_ALIAS_INTRINSICS_H_
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index c5666941fd36a..cc2939314b0eb 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -3211,7 +3211,9 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
   case Builtin::BI__builtin_hlsl_elementwise_rsqrt:
   case Builtin::BI__builtin_hlsl_elementwise_frac:
   case Builtin::BI__builtin_hlsl_elementwise_ddx_coarse:
-  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse: {
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_coarse:
+  case Builtin::BI__builtin_hlsl_elementwise_ddx_fine:
+  case Builtin::BI__builtin_hlsl_elementwise_ddy_fine: {
     if (SemaRef.checkArgCount(TheCall, 1))
       return true;
     if (CheckAllArgTypesAreCorrect(&SemaRef, TheCall,
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
new file mode 100644
index 0000000000000..69f7ab3c6ce62
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return __builtin_hlsl_elementwise_ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
new file mode 100644
index 0000000000000..2630260abcb43
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddx-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddx_fineDh
+// CHECK: %hlsl.ddx.fine = call {{.*}} half @llvm.dx.ddx.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddx_fineDh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} half @llvm.spv.ddx.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddx.fine
+half test_f16_ddx_fine(half val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.dx.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddx_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x half> @llvm.spv.ddx.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddx.fine
+half2 test_f16_ddx_fine2(half2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.dx.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddx_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x half> @llvm.spv.ddx.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddx.fine
+half3 test_f16_ddx_fine3(half3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.dx.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddx_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x half> @llvm.spv.ddx.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddx.fine
+half4 test_f16_ddx_fine4(half4 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddx_finef
+// CHECK: %hlsl.ddx.fine = call {{.*}} float @llvm.dx.ddx.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddx_finef
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} float @llvm.spv.ddx.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddx.fine
+float test_f32_ddx_fine(float val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.dx.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddx_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <2 x float> @llvm.spv.ddx.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddx.fine
+float2 test_f32_ddx_fine2(float2 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.dx.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddx_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <3 x float> @llvm.spv.ddx.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddx.fine
+float3 test_f32_ddx_fine3(float3 val) {
+    return ddx_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.dx.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddx.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddx_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddx.fine = call {{.*}} <4 x float> @llvm.spv.ddx.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddx.fine
+float4 test_f32_ddx_fine4(float4 val) {
+    return ddx_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
new file mode 100644
index 0000000000000..00630721ceb66
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine-builtin.hlsl
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return __builtin_hlsl_elementwise_ddy_fine(val);
+}
diff --git a/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
new file mode 100644
index 0000000000000..7e32ee29e767d
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/ddy-fine.hlsl
@@ -0,0 +1,86 @@
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple dxil-pc-shadermodel6.3-library %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK
+// RUN: %clang_cc1 -finclude-default-header  -x hlsl  -triple spirv-pc-vulkan-compute  %s \
+// RUN:  -emit-llvm -disable-llvm-passes -fnative-half-type -o - | \
+// RUN:  FileCheck %s --check-prefixes=CHECK-SPIRV
+
+// CHECK-LABEL: half @_Z17test_f16_ddy_fineDh
+// CHECK: %hlsl.ddy.fine = call {{.*}} half @llvm.dx.ddy.fine.f16(half %{{.*}})
+// CHECK: ret half %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: half @_Z17test_f16_ddy_fineDh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} half @llvm.spv.ddy.fine.f16(half %{{.*}})
+// CHECK-SPIRV: ret half %hlsl.ddy.fine
+half test_f16_ddy_fine(half val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.dx.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK: ret <2 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x half> @_Z18test_f16_ddy_fine2Dv2_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x half> @llvm.spv.ddy.fine.v2f16(<2 x half> %{{.*}})
+// CHECK-SPIRV: ret <2 x half> %hlsl.ddy.fine
+half2 test_f16_ddy_fine2(half2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.dx.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK: ret <3 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x half> @_Z18test_f16_ddy_fine3Dv3_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x half> @llvm.spv.ddy.fine.v3f16(<3 x half> %{{.*}})
+// CHECK-SPIRV: ret <3 x half> %hlsl.ddy.fine
+half3 test_f16_ddy_fine3(half3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.dx.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK: ret <4 x half> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x half> @_Z18test_f16_ddy_fine4Dv4_Dh
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x half> @llvm.spv.ddy.fine.v4f16(<4 x half> %{{.*}})
+// CHECK-SPIRV: ret <4 x half> %hlsl.ddy.fine
+half4 test_f16_ddy_fine4(half4 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: float @_Z17test_f32_ddy_finef
+// CHECK: %hlsl.ddy.fine = call {{.*}} float @llvm.dx.ddy.fine.f32(float %{{.*}})
+// CHECK: ret float %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: float @_Z17test_f32_ddy_finef
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} float @llvm.spv.ddy.fine.f32(float %{{.*}})
+// CHECK-SPIRV: ret float %hlsl.ddy.fine
+float test_f32_ddy_fine(float val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.dx.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK: ret <2 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <2 x float> @_Z18test_f32_ddy_fine2Dv2_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <2 x float> @llvm.spv.ddy.fine.v2f32(<2 x float> %{{.*}})
+// CHECK-SPIRV: ret <2 x float> %hlsl.ddy.fine
+float2 test_f32_ddy_fine2(float2 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.dx.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK: ret <3 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <3 x float> @_Z18test_f32_ddy_fine3Dv3_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <3 x float> @llvm.spv.ddy.fine.v3f32(<3 x float> %{{.*}})
+// CHECK-SPIRV: ret <3 x float> %hlsl.ddy.fine
+float3 test_f32_ddy_fine3(float3 val) {
+    return ddy_fine(val);
+}
+
+// CHECK-LABEL: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.dx.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK: ret <4 x float> %hlsl.ddy.fine
+// CHECK-LABEL-SPIRV: <4 x float> @_Z18test_f32_ddy_fine4Dv4_f
+// CHECK-SPIRV: %hlsl.ddy.fine = call {{.*}} <4 x float> @llvm.spv.ddy.fine.v4f32(<4 x float> %{{.*}})
+// CHECK-SPIRV: ret <4 x float> %hlsl.ddy.fine
+float4 test_f32_ddy_fine4(float4 val) {
+    return ddy_fine(val);
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
new file mode 100644
index 0000000000000..71196943c322a
--- /dev/null
+++ b/clang/test/SemaHLSL/BuiltIns/ddx-fine-errors.hlsl
@@ -0,0 +1,22 @@
+// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.6-library %s -fnative-half-type -verify
+// RUN: %clang_cc1 -triple spirv-unknown-vulkan1.3-library %s -fnative-half-type -verify
+
+float no_arg() {
+  return __builtin_hlsl_elementwise_ddx_fine();
+  // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float too_many_args(float val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val, val);
+  // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+float test_integer_scalar_input(int val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'int')}}
+}
+
+double test_double_scalar_input(double val) {
+  return __builtin_hlsl_elementwise_ddx_fine(val);
+  // expected-error@-1 {{1st argument must be a scalar or vector of 16 or 32 bit floating-point types (was 'double')}}
+}
diff --git a/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/ddy-fine-errors.hlsl
new file mode 100644
index 0000000000000..e17d5f36832b2
--- /dev/null...
[truncated]

@github-actions
Copy link

github-actions bot commented Nov 20, 2025 •
edited
Loading

🐧 Linux x64 Test Results

  • 192864 tests passed
  • 6188 tests skipped

@Alexander-Johnston
[HLSL] Implement ddx/ddy_fine intrinsics
7d46602 
Implements the HLSL ddx_fine and ddy_fine intrinsics together as the
implementations are nearly identical.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

backend:DirectX backend:SPIR-V backend:X86 clang:codegen IR generation bugs: mangling, exceptions, etc. clang:frontend Language frontend issues, e.g. anything involving "Sema" clang:headers Headers provided by Clang, e.g. for intrinsics HLSL HLSL Language Support llvm:ir

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Implement the ddy_fine HLSL Function Implement the ddx_fine HLSL Function

2 participants

@Alexander-Johnston @llvmbot