[HLSL][DXIL][SPIRV] QuadReadAcrossY intrinsic support#187440
Merged
[HLSL][DXIL][SPIRV] QuadReadAcrossY intrinsic support#187440
Conversation
llvmbot
added
backend:X86
clang:frontend
Member
|
@llvm/pr-subscribers-backend-spir-v @llvm/pr-subscribers-llvm-ir Author: Kai (kcloudy0717) ChangesThis PR adds QuadReadAcrossY intrinsic support in HLSL with codegen for both DirectX and SPIRV backends. Resolves #99176.
Patch is 22.96 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/187440.diff 16 Files Affected:
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 8e002c5b900aa..5a1ab7197b4a8 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -5276,6 +5276,12 @@ def HLSLQuadReadAcrossX : LangBuiltin<"HLSL_LANG"> {
let Prototype = "void(...)";
}
+def HLSLQuadReadAcrossY : LangBuiltin<"HLSL_LANG"> {
+ let Spellings = ["__builtin_hlsl_quad_read_across_y"];
+ let Attributes = [NoThrow, Const];
+ let Prototype = "void(...)";
+}
+
def HLSLClamp : LangBuiltin<"HLSL_LANG"> {
let Spellings = ["__builtin_hlsl_elementwise_clamp"];
let Attributes = [NoThrow, Const, CustomTypeChecking];
diff --git a/clang/lib/CodeGen/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
index 22293c5983d89..98edd479c02a5 100644
--- a/clang/lib/CodeGen/CGHLSLBuiltins.cpp
+++ b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
@@ -1402,6 +1402,13 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
&CGM.getModule(), ID, {OpExpr->getType()}),
ArrayRef{OpExpr}, "hlsl.quad.read.across.x");
}
+ case Builtin::BI__builtin_hlsl_quad_read_across_y: {
+ Value *OpExpr = EmitScalarExpr(E->getArg(0));
+ Intrinsic::ID ID = CGM.getHLSLRuntime().getQuadReadAcrossYIntrinsic();
+ return EmitRuntimeCall(Intrinsic::getOrInsertDeclaration(
+ &CGM.getModule(), ID, {OpExpr->getType()}),
+ ArrayRef{OpExpr}, "hlsl.quad.read.across.y");
+ }
case Builtin::BI__builtin_hlsl_elementwise_sign: {
auto *Arg0 = E->getArg(0);
Value *Op0 = EmitScalarExpr(Arg0);
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index d76cc18c9a259..f93ed28b5c787 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -158,6 +158,7 @@ class CGHLSLRuntime {
GENERATE_HLSL_INTRINSIC_FUNCTION(WaveGetLaneCount, wave_get_lane_count)
GENERATE_HLSL_INTRINSIC_FUNCTION(WaveReadLaneAt, wave_readlane)
GENERATE_HLSL_INTRINSIC_FUNCTION(QuadReadAcrossX, quad_read_across_x)
+ GENERATE_HLSL_INTRINSIC_FUNCTION(QuadReadAcrossY, quad_read_across_y)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitUHigh, firstbituhigh)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitSHigh, firstbitshigh)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitLow, firstbitlow)
diff --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index c2e4d74af6873..a1048d21b20e6 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -3604,6 +3604,105 @@ __attribute__((convergent)) double3 QuadReadAcrossX(double3);
_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_x)
__attribute__((convergent)) double4 QuadReadAcrossX(double4);
+//===----------------------------------------------------------------------===//
+// QuadReadAcrossY builtins
+//===----------------------------------------------------------------------===//
+
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half QuadReadAcrossY(half);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half2 QuadReadAcrossY(half2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half3 QuadReadAcrossY(half3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half4 QuadReadAcrossY(half4);
+
+#ifdef __HLSL_ENABLE_16_BIT
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t QuadReadAcrossY(int16_t);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t2 QuadReadAcrossY(int16_t2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t3 QuadReadAcrossY(int16_t3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t4 QuadReadAcrossY(int16_t4);
+
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t QuadReadAcrossY(uint16_t);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t2 QuadReadAcrossY(uint16_t2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t3 QuadReadAcrossY(uint16_t3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t4 QuadReadAcrossY(uint16_t4);
+#endif
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int QuadReadAcrossY(int);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int2 QuadReadAcrossY(int2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int3 QuadReadAcrossY(int3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int4 QuadReadAcrossY(int4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint QuadReadAcrossY(uint);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint2 QuadReadAcrossY(uint2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint3 QuadReadAcrossY(uint3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint4 QuadReadAcrossY(uint4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t QuadReadAcrossY(int64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t2 QuadReadAcrossY(int64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t3 QuadReadAcrossY(int64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t4 QuadReadAcrossY(int64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t QuadReadAcrossY(uint64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t2 QuadReadAcrossY(uint64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t3 QuadReadAcrossY(uint64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t4 QuadReadAcrossY(uint64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float QuadReadAcrossY(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float2 QuadReadAcrossY(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float3 QuadReadAcrossY(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float4 QuadReadAcrossY(float4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double QuadReadAcrossY(double);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double2 QuadReadAcrossY(double2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double3 QuadReadAcrossY(double3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double4 QuadReadAcrossY(double4);
+
//===----------------------------------------------------------------------===//
// sign builtins
//===----------------------------------------------------------------------===//
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index 0619295cd2fbb..badbf9416d44f 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -4236,7 +4236,8 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
TheCall->setType(ArgTyExpr);
break;
}
- case Builtin::BI__builtin_hlsl_quad_read_across_x: {
+ case Builtin::BI__builtin_hlsl_quad_read_across_x:
+ case Builtin::BI__builtin_hlsl_quad_read_across_y: {
if (SemaRef.checkArgCount(TheCall, 1))
return true;
diff --git a/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl b/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl
new file mode 100644
index 0000000000000..a050aa9cf1722
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl
@@ -0,0 +1,46 @@
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN: dxil-pc-shadermodel6.3-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN: FileCheck %s --check-prefixes=CHECK,CHECK-DXIL
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN: spirv-pc-vulkan-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN: FileCheck %s --check-prefixes=CHECK,CHECK-SPIRV
+
+// Test basic lowering to runtime function call.
+
+// CHECK-LABEL: test_int
+int test_int(int expr) {
+ // CHECK-SPIRV: %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.quad.read.across.y.i32([[TY]] %[[#]])
+ // CHECK-DXIL: %[[RET:.*]] = call [[TY:.*]] @llvm.dx.quad.read.across.y.i32([[TY]] %[[#]])
+ // CHECK: ret [[TY]] %[[RET]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.quad.read.across.y.i32([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare [[TY]] @llvm.spv.quad.read.across.y.i32([[TY]]) #[[#attr:]]
+
+// CHECK-LABEL: test_uint64_t
+uint64_t test_uint64_t(uint64_t expr) {
+ // CHECK-SPIRV: %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.quad.read.across.y.i64([[TY]] %[[#]])
+ // CHECK-DXIL: %[[RET:.*]] = call [[TY:.*]] @llvm.dx.quad.read.across.y.i64([[TY]] %[[#]])
+ // CHECK: ret [[TY]] %[[RET]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.quad.read.across.y.i64([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare [[TY]] @llvm.spv.quad.read.across.y.i64([[TY]]) #[[#attr:]]
+
+// Test basic lowering to runtime function call with array and float value.
+
+// CHECK-LABEL: test_floatv4
+float4 test_floatv4(float4 expr) {
+ // CHECK-SPIRV: %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn spir_func [[TY1:.*]] @llvm.spv.quad.read.across.y.v4f32([[TY1]] %[[#]]
+ // CHECK-DXIL: %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn [[TY1:.*]] @llvm.dx.quad.read.across.y.v4f32([[TY1]] %[[#]])
+ // CHECK: ret [[TY1]] %[[RET1]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY1]] @llvm.dx.quad.read.across.y.v4f32([[TY1]]) #[[#attr]]
+// CHECK-SPIRV: declare [[TY1]] @llvm.spv.quad.read.across.y.v4f32([[TY1]]) #[[#attr]]
+
+// CHECK: attributes #[[#attr]] = {{{.*}} convergent {{.*}}}
+
diff --git a/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl
new file mode 100644
index 0000000000000..2995895ff4c3d
--- /dev/null
+++ b/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl
@@ -0,0 +1,28 @@
+// RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.6-library %s -emit-llvm-only -disable-llvm-passes -verify
+
+int test_too_few_arg() {
+ return __builtin_hlsl_quad_read_across_y();
+ // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float2 test_too_many_arg(float2 p0) {
+ return __builtin_hlsl_quad_read_across_y(p0, p0);
+ // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+bool test_expr_bool_type_check(bool p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'bool'}}
+}
+
+bool2 test_expr_bool_vec_type_check(bool2 p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'bool2' (aka 'vector<bool, 2>')}}
+}
+
+struct S { float f; };
+
+S test_expr_struct_type_check(S p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'S' where a scalar or vector is required}}
+}
diff --git a/llvm/include/llvm/IR/IntrinsicsDirectX.td b/llvm/include/llvm/IR/IntrinsicsDirectX.td
index 3f7922382f090..8ee07b32c1eca 100644
--- a/llvm/include/llvm/IR/IntrinsicsDirectX.td
+++ b/llvm/include/llvm/IR/IntrinsicsDirectX.td
@@ -255,6 +255,7 @@ def int_dx_wave_prefix_usum : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatch
def int_dx_wave_prefix_product : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_wave_prefix_uproduct : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_quad_read_across_x : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+def int_dx_quad_read_across_y : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;
def int_dx_step : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty, LLVMMatchType<0>], [IntrNoMem]>;
def int_dx_splitdouble : DefaultAttrsIntrinsic<[llvm_anyint_ty, LLVMMatchType<0>],
diff --git a/llvm/include/llvm/IR/IntrinsicsSPIRV.td b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
index 5d467adb08c3d..6f030b7f2509f 100644
--- a/llvm/include/llvm/IR/IntrinsicsSPIRV.td
+++ b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
@@ -148,6 +148,7 @@ def int_spv_rsqrt : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty]
def int_spv_wave_prefix_sum : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_wave_prefix_product : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_quad_read_across_x : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+ def int_spv_quad_read_across_y : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;
def int_spv_radians : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty], [IntrNoMem]>;
def int_spv_group_memory_barrier_with_group_sync : ClangBuiltin<"__builtin_spirv_group_barrier">,
diff --git a/llvm/lib/Target/DirectX/DXIL.td b/llvm/lib/Target/DirectX/DXIL.td
index 20de198ef96d2..13fc147e60216 100644
--- a/llvm/lib/Target/DirectX/DXIL.td
+++ b/llvm/lib/Target/DirectX/DXIL.td
@@ -1214,6 +1214,10 @@ def QuadOp : DXILOp<123, quadOp> {
[
IntrinArgIndex<0>, IntrinArgI8<QuadOpKind_ReadAcrossX>
]>,
+ IntrinSelect<int_dx_quad_read_across_y,
+ [
+ IntrinArgIndex<0>, IntrinArgI8<QuadOpKind_ReadAcrossY>
+ ]>,
];
let arguments = [OverloadTy, Int8Ty];
diff --git a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
index fab8bea379f74..69b942a171992 100644
--- a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
+++ b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
@@ -107,6 +107,7 @@ static bool checkWaveOps(Intrinsic::ID IID) {
case Intrinsic::dx_wave_prefix_uproduct:
// Quad Op Variants
case Intrinsic::dx_quad_read_across_x:
+ case Intrinsic::dx_quad_read_across_y:
return true;
}
}
diff --git a/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp b/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
index 0c652b3bb29c0..830fcc9da862e 100644
--- a/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
+++ b/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
@@ -77,6 +77,7 @@ bool DirectXTTIImpl::isTargetIntrinsicTriviallyScalarizable(
case Intrinsic::dx_wave_prefix_usum:
case Intrinsic::dx_wave_prefix_uproduct:
case Intrinsic::dx_quad_read_across_x:
+ case Intrinsic::dx_quad_read_across_y:
case Intrinsic::dx_imad:
case Intrinsic::dx_umad:
case Intrinsic::dx_ddx_coarse:
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index eb30002d2e1a5..abad90891b4f5 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -4457,6 +4457,9 @@ bool SPIRVInstructionSelector::selectIntrinsic(Register ResVReg,
case Intrinsic::spv_quad_read_across_x: {
return selectQuadSwap(ResVReg, ResType, I, /*Direction*/ 0);
}
+ case Intrinsic::spv_quad_read_across_y: {
+ return selectQuadSwap(ResVReg, ResType, I, /*Direction*/ 1);
+ }
case Intrinsic::spv_step:
return selectExtInst(ResVReg, ResType, I, CL::step, GL::Step);
case Intrinsic::spv_radians:
diff --git a/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll b/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll
new file mode 100644
index 0000000000000..d90ac759aa23d
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll
@@ -0,0 +1,87 @@
+; RUN: opt -S -scalarizer -dxil-op-lower -mtriple=dxil-pc-shadermodel6.3-library < %s | FileCheck %s
+
+; Test that for scalar values, QuadReadAcrossY maps down to the DirectX op
+
+define noundef half @quad_read_across_y_half(half noundef %expr) {
+entry:
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr, i8 1)
+ %ret = call half @llvm.dx.quad.read.across.y.f16(half %expr)
+ ret half %ret
+}
+
+define noundef float @quad_read_across_y_float(float noundef %expr) {
+entry:
+; CHECK: call float @dx.op.quadOp.f32(i32 123, float %expr, i8 1)
+ %ret = call float @llvm.dx.quad.read.across.y.f32(float %expr)
+ ret float %ret
+}
+
+define noundef double @quad_read_across_y_double(double noundef %expr) {
+entry:
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr, i8 1)
+ %ret = call double @llvm.dx.quad.read.across.y.f64(double %expr)
+ ret double %ret
+}
+
+define noundef i16 @quad_read_across_y_i16(i16 noundef %expr) {
+entry:
+; CHECK: call i16 @dx.op.quadOp.i16(i32 123, i16 %expr, i8 1)
+ %ret = call i16 @llvm.dx.quad.read.across.y.i16(i16 %expr)
+ ret i16 %ret
+}
+
+define noundef i32 @quad_read_across_y_i32(i32 noundef %expr) {
+entry:
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr, i8 1)
+ %ret = call i32 @llvm.dx.quad.read.across.y.i32(i32 %expr)
+ ret i32 %ret
+}
+
+define noundef i64 @quad_read_across_y_i64(i64 noundef %expr) {
+entry:
+; CHECK: call i64 @dx.op.quadOp.i64(i32 123, i64 %expr, i8 1)
+ %ret = call i64 @llvm.dx.quad.read.across.y.i64(i64 %expr)
+ ret i64 %ret
+}
+
+declare half @llvm.dx.quad.read.across.y.f16(half)
+declare float @llvm.dx.quad.read.across.y.f32(float)
+declare double @llvm.dx.quad.read.across.y.f64(double)
+
+declare i16 @llvm.dx.quad.read.across.y.i16(i16)
+declare i32 @llvm.dx.quad.read.across.y.i32(i32)
+declare i64 @llvm.dx.quad.read.across.y.i64(i64)
+
+; Test that for vector values, QuadReadAcrossY scalarizes and maps down to the
+; DirectX op
+
+define noundef <2 x half> @quad_read_across_y_v2half(<2 x half> noundef %expr) {
+entry:
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr.i0, i8 1)
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr.i1, i8 1)
+ %ret = call <2 x half> @llvm.dx.quad.read.across.y.v2f16(<2 x half> %expr)
+ ret <2 x half> %ret
+}
+
+define noundef <3 x i32> @quad_read_across_y_v3i32(<3 x i32> noundef %expr) {
+entry:
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i0, i8 1)
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i1, i8 1)
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i2, i8 1)
+ %ret = call <3 x i32> @llvm.dx.quad.read.across.y.v3i32(<3 x i32> %expr)
+ ret <3 x i32> %ret
+}
+
+define noundef <4 x double> @quad_read_across_y_v4f64(<4 x double> noundef %expr) {
+entry:
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i0, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i1, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i2, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i3, i8 1)
+ ...
[truncated]
|
Member
|
@llvm/pr-subscribers-backend-directx Author: Kai (kcloudy0717) ChangesThis PR adds QuadReadAcrossY intrinsic support in HLSL with codegen for both DirectX and SPIRV backends. Resolves #99176.
Patch is 22.96 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/187440.diff 16 Files Affected:
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 8e002c5b900aa..5a1ab7197b4a8 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -5276,6 +5276,12 @@ def HLSLQuadReadAcrossX : LangBuiltin<"HLSL_LANG"> {
let Prototype = "void(...)";
}
+def HLSLQuadReadAcrossY : LangBuiltin<"HLSL_LANG"> {
+ let Spellings = ["__builtin_hlsl_quad_read_across_y"];
+ let Attributes = [NoThrow, Const];
+ let Prototype = "void(...)";
+}
+
def HLSLClamp : LangBuiltin<"HLSL_LANG"> {
let Spellings = ["__builtin_hlsl_elementwise_clamp"];
let Attributes = [NoThrow, Const, CustomTypeChecking];
diff --git a/clang/lib/CodeGen/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
index 22293c5983d89..98edd479c02a5 100644
--- a/clang/lib/CodeGen/CGHLSLBuiltins.cpp
+++ b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
@@ -1402,6 +1402,13 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
&CGM.getModule(), ID, {OpExpr->getType()}),
ArrayRef{OpExpr}, "hlsl.quad.read.across.x");
}
+ case Builtin::BI__builtin_hlsl_quad_read_across_y: {
+ Value *OpExpr = EmitScalarExpr(E->getArg(0));
+ Intrinsic::ID ID = CGM.getHLSLRuntime().getQuadReadAcrossYIntrinsic();
+ return EmitRuntimeCall(Intrinsic::getOrInsertDeclaration(
+ &CGM.getModule(), ID, {OpExpr->getType()}),
+ ArrayRef{OpExpr}, "hlsl.quad.read.across.y");
+ }
case Builtin::BI__builtin_hlsl_elementwise_sign: {
auto *Arg0 = E->getArg(0);
Value *Op0 = EmitScalarExpr(Arg0);
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index d76cc18c9a259..f93ed28b5c787 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -158,6 +158,7 @@ class CGHLSLRuntime {
GENERATE_HLSL_INTRINSIC_FUNCTION(WaveGetLaneCount, wave_get_lane_count)
GENERATE_HLSL_INTRINSIC_FUNCTION(WaveReadLaneAt, wave_readlane)
GENERATE_HLSL_INTRINSIC_FUNCTION(QuadReadAcrossX, quad_read_across_x)
+ GENERATE_HLSL_INTRINSIC_FUNCTION(QuadReadAcrossY, quad_read_across_y)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitUHigh, firstbituhigh)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitSHigh, firstbitshigh)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitLow, firstbitlow)
diff --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index c2e4d74af6873..a1048d21b20e6 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -3604,6 +3604,105 @@ __attribute__((convergent)) double3 QuadReadAcrossX(double3);
_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_x)
__attribute__((convergent)) double4 QuadReadAcrossX(double4);
+//===----------------------------------------------------------------------===//
+// QuadReadAcrossY builtins
+//===----------------------------------------------------------------------===//
+
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half QuadReadAcrossY(half);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half2 QuadReadAcrossY(half2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half3 QuadReadAcrossY(half3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half4 QuadReadAcrossY(half4);
+
+#ifdef __HLSL_ENABLE_16_BIT
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t QuadReadAcrossY(int16_t);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t2 QuadReadAcrossY(int16_t2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t3 QuadReadAcrossY(int16_t3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t4 QuadReadAcrossY(int16_t4);
+
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t QuadReadAcrossY(uint16_t);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t2 QuadReadAcrossY(uint16_t2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t3 QuadReadAcrossY(uint16_t3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t4 QuadReadAcrossY(uint16_t4);
+#endif
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int QuadReadAcrossY(int);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int2 QuadReadAcrossY(int2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int3 QuadReadAcrossY(int3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int4 QuadReadAcrossY(int4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint QuadReadAcrossY(uint);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint2 QuadReadAcrossY(uint2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint3 QuadReadAcrossY(uint3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint4 QuadReadAcrossY(uint4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t QuadReadAcrossY(int64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t2 QuadReadAcrossY(int64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t3 QuadReadAcrossY(int64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t4 QuadReadAcrossY(int64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t QuadReadAcrossY(uint64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t2 QuadReadAcrossY(uint64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t3 QuadReadAcrossY(uint64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t4 QuadReadAcrossY(uint64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float QuadReadAcrossY(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float2 QuadReadAcrossY(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float3 QuadReadAcrossY(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float4 QuadReadAcrossY(float4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double QuadReadAcrossY(double);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double2 QuadReadAcrossY(double2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double3 QuadReadAcrossY(double3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double4 QuadReadAcrossY(double4);
+
//===----------------------------------------------------------------------===//
// sign builtins
//===----------------------------------------------------------------------===//
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index 0619295cd2fbb..badbf9416d44f 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -4236,7 +4236,8 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
TheCall->setType(ArgTyExpr);
break;
}
- case Builtin::BI__builtin_hlsl_quad_read_across_x: {
+ case Builtin::BI__builtin_hlsl_quad_read_across_x:
+ case Builtin::BI__builtin_hlsl_quad_read_across_y: {
if (SemaRef.checkArgCount(TheCall, 1))
return true;
diff --git a/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl b/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl
new file mode 100644
index 0000000000000..a050aa9cf1722
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl
@@ -0,0 +1,46 @@
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN: dxil-pc-shadermodel6.3-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN: FileCheck %s --check-prefixes=CHECK,CHECK-DXIL
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN: spirv-pc-vulkan-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN: FileCheck %s --check-prefixes=CHECK,CHECK-SPIRV
+
+// Test basic lowering to runtime function call.
+
+// CHECK-LABEL: test_int
+int test_int(int expr) {
+ // CHECK-SPIRV: %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.quad.read.across.y.i32([[TY]] %[[#]])
+ // CHECK-DXIL: %[[RET:.*]] = call [[TY:.*]] @llvm.dx.quad.read.across.y.i32([[TY]] %[[#]])
+ // CHECK: ret [[TY]] %[[RET]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.quad.read.across.y.i32([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare [[TY]] @llvm.spv.quad.read.across.y.i32([[TY]]) #[[#attr:]]
+
+// CHECK-LABEL: test_uint64_t
+uint64_t test_uint64_t(uint64_t expr) {
+ // CHECK-SPIRV: %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.quad.read.across.y.i64([[TY]] %[[#]])
+ // CHECK-DXIL: %[[RET:.*]] = call [[TY:.*]] @llvm.dx.quad.read.across.y.i64([[TY]] %[[#]])
+ // CHECK: ret [[TY]] %[[RET]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.quad.read.across.y.i64([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare [[TY]] @llvm.spv.quad.read.across.y.i64([[TY]]) #[[#attr:]]
+
+// Test basic lowering to runtime function call with array and float value.
+
+// CHECK-LABEL: test_floatv4
+float4 test_floatv4(float4 expr) {
+ // CHECK-SPIRV: %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn spir_func [[TY1:.*]] @llvm.spv.quad.read.across.y.v4f32([[TY1]] %[[#]]
+ // CHECK-DXIL: %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn [[TY1:.*]] @llvm.dx.quad.read.across.y.v4f32([[TY1]] %[[#]])
+ // CHECK: ret [[TY1]] %[[RET1]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY1]] @llvm.dx.quad.read.across.y.v4f32([[TY1]]) #[[#attr]]
+// CHECK-SPIRV: declare [[TY1]] @llvm.spv.quad.read.across.y.v4f32([[TY1]]) #[[#attr]]
+
+// CHECK: attributes #[[#attr]] = {{{.*}} convergent {{.*}}}
+
diff --git a/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl
new file mode 100644
index 0000000000000..2995895ff4c3d
--- /dev/null
+++ b/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl
@@ -0,0 +1,28 @@
+// RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.6-library %s -emit-llvm-only -disable-llvm-passes -verify
+
+int test_too_few_arg() {
+ return __builtin_hlsl_quad_read_across_y();
+ // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float2 test_too_many_arg(float2 p0) {
+ return __builtin_hlsl_quad_read_across_y(p0, p0);
+ // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+bool test_expr_bool_type_check(bool p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'bool'}}
+}
+
+bool2 test_expr_bool_vec_type_check(bool2 p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'bool2' (aka 'vector<bool, 2>')}}
+}
+
+struct S { float f; };
+
+S test_expr_struct_type_check(S p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'S' where a scalar or vector is required}}
+}
diff --git a/llvm/include/llvm/IR/IntrinsicsDirectX.td b/llvm/include/llvm/IR/IntrinsicsDirectX.td
index 3f7922382f090..8ee07b32c1eca 100644
--- a/llvm/include/llvm/IR/IntrinsicsDirectX.td
+++ b/llvm/include/llvm/IR/IntrinsicsDirectX.td
@@ -255,6 +255,7 @@ def int_dx_wave_prefix_usum : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatch
def int_dx_wave_prefix_product : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_wave_prefix_uproduct : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_quad_read_across_x : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+def int_dx_quad_read_across_y : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;
def int_dx_step : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty, LLVMMatchType<0>], [IntrNoMem]>;
def int_dx_splitdouble : DefaultAttrsIntrinsic<[llvm_anyint_ty, LLVMMatchType<0>],
diff --git a/llvm/include/llvm/IR/IntrinsicsSPIRV.td b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
index 5d467adb08c3d..6f030b7f2509f 100644
--- a/llvm/include/llvm/IR/IntrinsicsSPIRV.td
+++ b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
@@ -148,6 +148,7 @@ def int_spv_rsqrt : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty]
def int_spv_wave_prefix_sum : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_wave_prefix_product : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_quad_read_across_x : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+ def int_spv_quad_read_across_y : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;
def int_spv_radians : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty], [IntrNoMem]>;
def int_spv_group_memory_barrier_with_group_sync : ClangBuiltin<"__builtin_spirv_group_barrier">,
diff --git a/llvm/lib/Target/DirectX/DXIL.td b/llvm/lib/Target/DirectX/DXIL.td
index 20de198ef96d2..13fc147e60216 100644
--- a/llvm/lib/Target/DirectX/DXIL.td
+++ b/llvm/lib/Target/DirectX/DXIL.td
@@ -1214,6 +1214,10 @@ def QuadOp : DXILOp<123, quadOp> {
[
IntrinArgIndex<0>, IntrinArgI8<QuadOpKind_ReadAcrossX>
]>,
+ IntrinSelect<int_dx_quad_read_across_y,
+ [
+ IntrinArgIndex<0>, IntrinArgI8<QuadOpKind_ReadAcrossY>
+ ]>,
];
let arguments = [OverloadTy, Int8Ty];
diff --git a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
index fab8bea379f74..69b942a171992 100644
--- a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
+++ b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
@@ -107,6 +107,7 @@ static bool checkWaveOps(Intrinsic::ID IID) {
case Intrinsic::dx_wave_prefix_uproduct:
// Quad Op Variants
case Intrinsic::dx_quad_read_across_x:
+ case Intrinsic::dx_quad_read_across_y:
return true;
}
}
diff --git a/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp b/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
index 0c652b3bb29c0..830fcc9da862e 100644
--- a/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
+++ b/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
@@ -77,6 +77,7 @@ bool DirectXTTIImpl::isTargetIntrinsicTriviallyScalarizable(
case Intrinsic::dx_wave_prefix_usum:
case Intrinsic::dx_wave_prefix_uproduct:
case Intrinsic::dx_quad_read_across_x:
+ case Intrinsic::dx_quad_read_across_y:
case Intrinsic::dx_imad:
case Intrinsic::dx_umad:
case Intrinsic::dx_ddx_coarse:
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index eb30002d2e1a5..abad90891b4f5 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -4457,6 +4457,9 @@ bool SPIRVInstructionSelector::selectIntrinsic(Register ResVReg,
case Intrinsic::spv_quad_read_across_x: {
return selectQuadSwap(ResVReg, ResType, I, /*Direction*/ 0);
}
+ case Intrinsic::spv_quad_read_across_y: {
+ return selectQuadSwap(ResVReg, ResType, I, /*Direction*/ 1);
+ }
case Intrinsic::spv_step:
return selectExtInst(ResVReg, ResType, I, CL::step, GL::Step);
case Intrinsic::spv_radians:
diff --git a/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll b/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll
new file mode 100644
index 0000000000000..d90ac759aa23d
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll
@@ -0,0 +1,87 @@
+; RUN: opt -S -scalarizer -dxil-op-lower -mtriple=dxil-pc-shadermodel6.3-library < %s | FileCheck %s
+
+; Test that for scalar values, QuadReadAcrossY maps down to the DirectX op
+
+define noundef half @quad_read_across_y_half(half noundef %expr) {
+entry:
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr, i8 1)
+ %ret = call half @llvm.dx.quad.read.across.y.f16(half %expr)
+ ret half %ret
+}
+
+define noundef float @quad_read_across_y_float(float noundef %expr) {
+entry:
+; CHECK: call float @dx.op.quadOp.f32(i32 123, float %expr, i8 1)
+ %ret = call float @llvm.dx.quad.read.across.y.f32(float %expr)
+ ret float %ret
+}
+
+define noundef double @quad_read_across_y_double(double noundef %expr) {
+entry:
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr, i8 1)
+ %ret = call double @llvm.dx.quad.read.across.y.f64(double %expr)
+ ret double %ret
+}
+
+define noundef i16 @quad_read_across_y_i16(i16 noundef %expr) {
+entry:
+; CHECK: call i16 @dx.op.quadOp.i16(i32 123, i16 %expr, i8 1)
+ %ret = call i16 @llvm.dx.quad.read.across.y.i16(i16 %expr)
+ ret i16 %ret
+}
+
+define noundef i32 @quad_read_across_y_i32(i32 noundef %expr) {
+entry:
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr, i8 1)
+ %ret = call i32 @llvm.dx.quad.read.across.y.i32(i32 %expr)
+ ret i32 %ret
+}
+
+define noundef i64 @quad_read_across_y_i64(i64 noundef %expr) {
+entry:
+; CHECK: call i64 @dx.op.quadOp.i64(i32 123, i64 %expr, i8 1)
+ %ret = call i64 @llvm.dx.quad.read.across.y.i64(i64 %expr)
+ ret i64 %ret
+}
+
+declare half @llvm.dx.quad.read.across.y.f16(half)
+declare float @llvm.dx.quad.read.across.y.f32(float)
+declare double @llvm.dx.quad.read.across.y.f64(double)
+
+declare i16 @llvm.dx.quad.read.across.y.i16(i16)
+declare i32 @llvm.dx.quad.read.across.y.i32(i32)
+declare i64 @llvm.dx.quad.read.across.y.i64(i64)
+
+; Test that for vector values, QuadReadAcrossY scalarizes and maps down to the
+; DirectX op
+
+define noundef <2 x half> @quad_read_across_y_v2half(<2 x half> noundef %expr) {
+entry:
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr.i0, i8 1)
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr.i1, i8 1)
+ %ret = call <2 x half> @llvm.dx.quad.read.across.y.v2f16(<2 x half> %expr)
+ ret <2 x half> %ret
+}
+
+define noundef <3 x i32> @quad_read_across_y_v3i32(<3 x i32> noundef %expr) {
+entry:
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i0, i8 1)
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i1, i8 1)
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i2, i8 1)
+ %ret = call <3 x i32> @llvm.dx.quad.read.across.y.v3i32(<3 x i32> %expr)
+ ret <3 x i32> %ret
+}
+
+define noundef <4 x double> @quad_read_across_y_v4f64(<4 x double> noundef %expr) {
+entry:
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i0, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i1, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i2, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i3, i8 1)
+ ...
[truncated]
|
Member
|
@llvm/pr-subscribers-hlsl Author: Kai (kcloudy0717) ChangesThis PR adds QuadReadAcrossY intrinsic support in HLSL with codegen for both DirectX and SPIRV backends. Resolves #99176.
Patch is 22.96 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/187440.diff 16 Files Affected:
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 8e002c5b900aa..5a1ab7197b4a8 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -5276,6 +5276,12 @@ def HLSLQuadReadAcrossX : LangBuiltin<"HLSL_LANG"> {
let Prototype = "void(...)";
}
+def HLSLQuadReadAcrossY : LangBuiltin<"HLSL_LANG"> {
+ let Spellings = ["__builtin_hlsl_quad_read_across_y"];
+ let Attributes = [NoThrow, Const];
+ let Prototype = "void(...)";
+}
+
def HLSLClamp : LangBuiltin<"HLSL_LANG"> {
let Spellings = ["__builtin_hlsl_elementwise_clamp"];
let Attributes = [NoThrow, Const, CustomTypeChecking];
diff --git a/clang/lib/CodeGen/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
index 22293c5983d89..98edd479c02a5 100644
--- a/clang/lib/CodeGen/CGHLSLBuiltins.cpp
+++ b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
@@ -1402,6 +1402,13 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
&CGM.getModule(), ID, {OpExpr->getType()}),
ArrayRef{OpExpr}, "hlsl.quad.read.across.x");
}
+ case Builtin::BI__builtin_hlsl_quad_read_across_y: {
+ Value *OpExpr = EmitScalarExpr(E->getArg(0));
+ Intrinsic::ID ID = CGM.getHLSLRuntime().getQuadReadAcrossYIntrinsic();
+ return EmitRuntimeCall(Intrinsic::getOrInsertDeclaration(
+ &CGM.getModule(), ID, {OpExpr->getType()}),
+ ArrayRef{OpExpr}, "hlsl.quad.read.across.y");
+ }
case Builtin::BI__builtin_hlsl_elementwise_sign: {
auto *Arg0 = E->getArg(0);
Value *Op0 = EmitScalarExpr(Arg0);
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index d76cc18c9a259..f93ed28b5c787 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -158,6 +158,7 @@ class CGHLSLRuntime {
GENERATE_HLSL_INTRINSIC_FUNCTION(WaveGetLaneCount, wave_get_lane_count)
GENERATE_HLSL_INTRINSIC_FUNCTION(WaveReadLaneAt, wave_readlane)
GENERATE_HLSL_INTRINSIC_FUNCTION(QuadReadAcrossX, quad_read_across_x)
+ GENERATE_HLSL_INTRINSIC_FUNCTION(QuadReadAcrossY, quad_read_across_y)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitUHigh, firstbituhigh)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitSHigh, firstbitshigh)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitLow, firstbitlow)
diff --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index c2e4d74af6873..a1048d21b20e6 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -3604,6 +3604,105 @@ __attribute__((convergent)) double3 QuadReadAcrossX(double3);
_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_x)
__attribute__((convergent)) double4 QuadReadAcrossX(double4);
+//===----------------------------------------------------------------------===//
+// QuadReadAcrossY builtins
+//===----------------------------------------------------------------------===//
+
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half QuadReadAcrossY(half);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half2 QuadReadAcrossY(half2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half3 QuadReadAcrossY(half3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half4 QuadReadAcrossY(half4);
+
+#ifdef __HLSL_ENABLE_16_BIT
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t QuadReadAcrossY(int16_t);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t2 QuadReadAcrossY(int16_t2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t3 QuadReadAcrossY(int16_t3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t4 QuadReadAcrossY(int16_t4);
+
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t QuadReadAcrossY(uint16_t);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t2 QuadReadAcrossY(uint16_t2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t3 QuadReadAcrossY(uint16_t3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t4 QuadReadAcrossY(uint16_t4);
+#endif
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int QuadReadAcrossY(int);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int2 QuadReadAcrossY(int2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int3 QuadReadAcrossY(int3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int4 QuadReadAcrossY(int4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint QuadReadAcrossY(uint);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint2 QuadReadAcrossY(uint2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint3 QuadReadAcrossY(uint3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint4 QuadReadAcrossY(uint4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t QuadReadAcrossY(int64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t2 QuadReadAcrossY(int64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t3 QuadReadAcrossY(int64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t4 QuadReadAcrossY(int64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t QuadReadAcrossY(uint64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t2 QuadReadAcrossY(uint64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t3 QuadReadAcrossY(uint64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t4 QuadReadAcrossY(uint64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float QuadReadAcrossY(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float2 QuadReadAcrossY(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float3 QuadReadAcrossY(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float4 QuadReadAcrossY(float4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double QuadReadAcrossY(double);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double2 QuadReadAcrossY(double2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double3 QuadReadAcrossY(double3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double4 QuadReadAcrossY(double4);
+
//===----------------------------------------------------------------------===//
// sign builtins
//===----------------------------------------------------------------------===//
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index 0619295cd2fbb..badbf9416d44f 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -4236,7 +4236,8 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
TheCall->setType(ArgTyExpr);
break;
}
- case Builtin::BI__builtin_hlsl_quad_read_across_x: {
+ case Builtin::BI__builtin_hlsl_quad_read_across_x:
+ case Builtin::BI__builtin_hlsl_quad_read_across_y: {
if (SemaRef.checkArgCount(TheCall, 1))
return true;
diff --git a/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl b/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl
new file mode 100644
index 0000000000000..a050aa9cf1722
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl
@@ -0,0 +1,46 @@
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN: dxil-pc-shadermodel6.3-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN: FileCheck %s --check-prefixes=CHECK,CHECK-DXIL
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN: spirv-pc-vulkan-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN: FileCheck %s --check-prefixes=CHECK,CHECK-SPIRV
+
+// Test basic lowering to runtime function call.
+
+// CHECK-LABEL: test_int
+int test_int(int expr) {
+ // CHECK-SPIRV: %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.quad.read.across.y.i32([[TY]] %[[#]])
+ // CHECK-DXIL: %[[RET:.*]] = call [[TY:.*]] @llvm.dx.quad.read.across.y.i32([[TY]] %[[#]])
+ // CHECK: ret [[TY]] %[[RET]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.quad.read.across.y.i32([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare [[TY]] @llvm.spv.quad.read.across.y.i32([[TY]]) #[[#attr:]]
+
+// CHECK-LABEL: test_uint64_t
+uint64_t test_uint64_t(uint64_t expr) {
+ // CHECK-SPIRV: %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.quad.read.across.y.i64([[TY]] %[[#]])
+ // CHECK-DXIL: %[[RET:.*]] = call [[TY:.*]] @llvm.dx.quad.read.across.y.i64([[TY]] %[[#]])
+ // CHECK: ret [[TY]] %[[RET]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.quad.read.across.y.i64([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare [[TY]] @llvm.spv.quad.read.across.y.i64([[TY]]) #[[#attr:]]
+
+// Test basic lowering to runtime function call with array and float value.
+
+// CHECK-LABEL: test_floatv4
+float4 test_floatv4(float4 expr) {
+ // CHECK-SPIRV: %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn spir_func [[TY1:.*]] @llvm.spv.quad.read.across.y.v4f32([[TY1]] %[[#]]
+ // CHECK-DXIL: %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn [[TY1:.*]] @llvm.dx.quad.read.across.y.v4f32([[TY1]] %[[#]])
+ // CHECK: ret [[TY1]] %[[RET1]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY1]] @llvm.dx.quad.read.across.y.v4f32([[TY1]]) #[[#attr]]
+// CHECK-SPIRV: declare [[TY1]] @llvm.spv.quad.read.across.y.v4f32([[TY1]]) #[[#attr]]
+
+// CHECK: attributes #[[#attr]] = {{{.*}} convergent {{.*}}}
+
diff --git a/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl
new file mode 100644
index 0000000000000..2995895ff4c3d
--- /dev/null
+++ b/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl
@@ -0,0 +1,28 @@
+// RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.6-library %s -emit-llvm-only -disable-llvm-passes -verify
+
+int test_too_few_arg() {
+ return __builtin_hlsl_quad_read_across_y();
+ // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float2 test_too_many_arg(float2 p0) {
+ return __builtin_hlsl_quad_read_across_y(p0, p0);
+ // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+bool test_expr_bool_type_check(bool p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'bool'}}
+}
+
+bool2 test_expr_bool_vec_type_check(bool2 p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'bool2' (aka 'vector<bool, 2>')}}
+}
+
+struct S { float f; };
+
+S test_expr_struct_type_check(S p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'S' where a scalar or vector is required}}
+}
diff --git a/llvm/include/llvm/IR/IntrinsicsDirectX.td b/llvm/include/llvm/IR/IntrinsicsDirectX.td
index 3f7922382f090..8ee07b32c1eca 100644
--- a/llvm/include/llvm/IR/IntrinsicsDirectX.td
+++ b/llvm/include/llvm/IR/IntrinsicsDirectX.td
@@ -255,6 +255,7 @@ def int_dx_wave_prefix_usum : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatch
def int_dx_wave_prefix_product : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_wave_prefix_uproduct : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_quad_read_across_x : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+def int_dx_quad_read_across_y : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;
def int_dx_step : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty, LLVMMatchType<0>], [IntrNoMem]>;
def int_dx_splitdouble : DefaultAttrsIntrinsic<[llvm_anyint_ty, LLVMMatchType<0>],
diff --git a/llvm/include/llvm/IR/IntrinsicsSPIRV.td b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
index 5d467adb08c3d..6f030b7f2509f 100644
--- a/llvm/include/llvm/IR/IntrinsicsSPIRV.td
+++ b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
@@ -148,6 +148,7 @@ def int_spv_rsqrt : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty]
def int_spv_wave_prefix_sum : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_wave_prefix_product : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_quad_read_across_x : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+ def int_spv_quad_read_across_y : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;
def int_spv_radians : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty], [IntrNoMem]>;
def int_spv_group_memory_barrier_with_group_sync : ClangBuiltin<"__builtin_spirv_group_barrier">,
diff --git a/llvm/lib/Target/DirectX/DXIL.td b/llvm/lib/Target/DirectX/DXIL.td
index 20de198ef96d2..13fc147e60216 100644
--- a/llvm/lib/Target/DirectX/DXIL.td
+++ b/llvm/lib/Target/DirectX/DXIL.td
@@ -1214,6 +1214,10 @@ def QuadOp : DXILOp<123, quadOp> {
[
IntrinArgIndex<0>, IntrinArgI8<QuadOpKind_ReadAcrossX>
]>,
+ IntrinSelect<int_dx_quad_read_across_y,
+ [
+ IntrinArgIndex<0>, IntrinArgI8<QuadOpKind_ReadAcrossY>
+ ]>,
];
let arguments = [OverloadTy, Int8Ty];
diff --git a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
index fab8bea379f74..69b942a171992 100644
--- a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
+++ b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
@@ -107,6 +107,7 @@ static bool checkWaveOps(Intrinsic::ID IID) {
case Intrinsic::dx_wave_prefix_uproduct:
// Quad Op Variants
case Intrinsic::dx_quad_read_across_x:
+ case Intrinsic::dx_quad_read_across_y:
return true;
}
}
diff --git a/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp b/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
index 0c652b3bb29c0..830fcc9da862e 100644
--- a/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
+++ b/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
@@ -77,6 +77,7 @@ bool DirectXTTIImpl::isTargetIntrinsicTriviallyScalarizable(
case Intrinsic::dx_wave_prefix_usum:
case Intrinsic::dx_wave_prefix_uproduct:
case Intrinsic::dx_quad_read_across_x:
+ case Intrinsic::dx_quad_read_across_y:
case Intrinsic::dx_imad:
case Intrinsic::dx_umad:
case Intrinsic::dx_ddx_coarse:
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index eb30002d2e1a5..abad90891b4f5 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -4457,6 +4457,9 @@ bool SPIRVInstructionSelector::selectIntrinsic(Register ResVReg,
case Intrinsic::spv_quad_read_across_x: {
return selectQuadSwap(ResVReg, ResType, I, /*Direction*/ 0);
}
+ case Intrinsic::spv_quad_read_across_y: {
+ return selectQuadSwap(ResVReg, ResType, I, /*Direction*/ 1);
+ }
case Intrinsic::spv_step:
return selectExtInst(ResVReg, ResType, I, CL::step, GL::Step);
case Intrinsic::spv_radians:
diff --git a/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll b/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll
new file mode 100644
index 0000000000000..d90ac759aa23d
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll
@@ -0,0 +1,87 @@
+; RUN: opt -S -scalarizer -dxil-op-lower -mtriple=dxil-pc-shadermodel6.3-library < %s | FileCheck %s
+
+; Test that for scalar values, QuadReadAcrossY maps down to the DirectX op
+
+define noundef half @quad_read_across_y_half(half noundef %expr) {
+entry:
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr, i8 1)
+ %ret = call half @llvm.dx.quad.read.across.y.f16(half %expr)
+ ret half %ret
+}
+
+define noundef float @quad_read_across_y_float(float noundef %expr) {
+entry:
+; CHECK: call float @dx.op.quadOp.f32(i32 123, float %expr, i8 1)
+ %ret = call float @llvm.dx.quad.read.across.y.f32(float %expr)
+ ret float %ret
+}
+
+define noundef double @quad_read_across_y_double(double noundef %expr) {
+entry:
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr, i8 1)
+ %ret = call double @llvm.dx.quad.read.across.y.f64(double %expr)
+ ret double %ret
+}
+
+define noundef i16 @quad_read_across_y_i16(i16 noundef %expr) {
+entry:
+; CHECK: call i16 @dx.op.quadOp.i16(i32 123, i16 %expr, i8 1)
+ %ret = call i16 @llvm.dx.quad.read.across.y.i16(i16 %expr)
+ ret i16 %ret
+}
+
+define noundef i32 @quad_read_across_y_i32(i32 noundef %expr) {
+entry:
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr, i8 1)
+ %ret = call i32 @llvm.dx.quad.read.across.y.i32(i32 %expr)
+ ret i32 %ret
+}
+
+define noundef i64 @quad_read_across_y_i64(i64 noundef %expr) {
+entry:
+; CHECK: call i64 @dx.op.quadOp.i64(i32 123, i64 %expr, i8 1)
+ %ret = call i64 @llvm.dx.quad.read.across.y.i64(i64 %expr)
+ ret i64 %ret
+}
+
+declare half @llvm.dx.quad.read.across.y.f16(half)
+declare float @llvm.dx.quad.read.across.y.f32(float)
+declare double @llvm.dx.quad.read.across.y.f64(double)
+
+declare i16 @llvm.dx.quad.read.across.y.i16(i16)
+declare i32 @llvm.dx.quad.read.across.y.i32(i32)
+declare i64 @llvm.dx.quad.read.across.y.i64(i64)
+
+; Test that for vector values, QuadReadAcrossY scalarizes and maps down to the
+; DirectX op
+
+define noundef <2 x half> @quad_read_across_y_v2half(<2 x half> noundef %expr) {
+entry:
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr.i0, i8 1)
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr.i1, i8 1)
+ %ret = call <2 x half> @llvm.dx.quad.read.across.y.v2f16(<2 x half> %expr)
+ ret <2 x half> %ret
+}
+
+define noundef <3 x i32> @quad_read_across_y_v3i32(<3 x i32> noundef %expr) {
+entry:
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i0, i8 1)
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i1, i8 1)
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i2, i8 1)
+ %ret = call <3 x i32> @llvm.dx.quad.read.across.y.v3i32(<3 x i32> %expr)
+ ret <3 x i32> %ret
+}
+
+define noundef <4 x double> @quad_read_across_y_v4f64(<4 x double> noundef %expr) {
+entry:
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i0, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i1, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i2, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i3, i8 1)
+ ...
[truncated]
|
Member
|
@llvm/pr-subscribers-clang-codegen Author: Kai (kcloudy0717) ChangesThis PR adds QuadReadAcrossY intrinsic support in HLSL with codegen for both DirectX and SPIRV backends. Resolves #99176.
Patch is 22.96 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/187440.diff 16 Files Affected:
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index 8e002c5b900aa..5a1ab7197b4a8 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -5276,6 +5276,12 @@ def HLSLQuadReadAcrossX : LangBuiltin<"HLSL_LANG"> {
let Prototype = "void(...)";
}
+def HLSLQuadReadAcrossY : LangBuiltin<"HLSL_LANG"> {
+ let Spellings = ["__builtin_hlsl_quad_read_across_y"];
+ let Attributes = [NoThrow, Const];
+ let Prototype = "void(...)";
+}
+
def HLSLClamp : LangBuiltin<"HLSL_LANG"> {
let Spellings = ["__builtin_hlsl_elementwise_clamp"];
let Attributes = [NoThrow, Const, CustomTypeChecking];
diff --git a/clang/lib/CodeGen/CGHLSLBuiltins.cpp b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
index 22293c5983d89..98edd479c02a5 100644
--- a/clang/lib/CodeGen/CGHLSLBuiltins.cpp
+++ b/clang/lib/CodeGen/CGHLSLBuiltins.cpp
@@ -1402,6 +1402,13 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
&CGM.getModule(), ID, {OpExpr->getType()}),
ArrayRef{OpExpr}, "hlsl.quad.read.across.x");
}
+ case Builtin::BI__builtin_hlsl_quad_read_across_y: {
+ Value *OpExpr = EmitScalarExpr(E->getArg(0));
+ Intrinsic::ID ID = CGM.getHLSLRuntime().getQuadReadAcrossYIntrinsic();
+ return EmitRuntimeCall(Intrinsic::getOrInsertDeclaration(
+ &CGM.getModule(), ID, {OpExpr->getType()}),
+ ArrayRef{OpExpr}, "hlsl.quad.read.across.y");
+ }
case Builtin::BI__builtin_hlsl_elementwise_sign: {
auto *Arg0 = E->getArg(0);
Value *Op0 = EmitScalarExpr(Arg0);
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index d76cc18c9a259..f93ed28b5c787 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -158,6 +158,7 @@ class CGHLSLRuntime {
GENERATE_HLSL_INTRINSIC_FUNCTION(WaveGetLaneCount, wave_get_lane_count)
GENERATE_HLSL_INTRINSIC_FUNCTION(WaveReadLaneAt, wave_readlane)
GENERATE_HLSL_INTRINSIC_FUNCTION(QuadReadAcrossX, quad_read_across_x)
+ GENERATE_HLSL_INTRINSIC_FUNCTION(QuadReadAcrossY, quad_read_across_y)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitUHigh, firstbituhigh)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitSHigh, firstbitshigh)
GENERATE_HLSL_INTRINSIC_FUNCTION(FirstBitLow, firstbitlow)
diff --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index c2e4d74af6873..a1048d21b20e6 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -3604,6 +3604,105 @@ __attribute__((convergent)) double3 QuadReadAcrossX(double3);
_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_x)
__attribute__((convergent)) double4 QuadReadAcrossX(double4);
+//===----------------------------------------------------------------------===//
+// QuadReadAcrossY builtins
+//===----------------------------------------------------------------------===//
+
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half QuadReadAcrossY(half);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half2 QuadReadAcrossY(half2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half3 QuadReadAcrossY(half3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) half4 QuadReadAcrossY(half4);
+
+#ifdef __HLSL_ENABLE_16_BIT
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t QuadReadAcrossY(int16_t);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t2 QuadReadAcrossY(int16_t2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t3 QuadReadAcrossY(int16_t3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int16_t4 QuadReadAcrossY(int16_t4);
+
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t QuadReadAcrossY(uint16_t);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t2 QuadReadAcrossY(uint16_t2);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t3 QuadReadAcrossY(uint16_t3);
+_HLSL_16BIT_AVAILABILITY_SHADERMODEL_DEFAULT()
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint16_t4 QuadReadAcrossY(uint16_t4);
+#endif
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int QuadReadAcrossY(int);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int2 QuadReadAcrossY(int2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int3 QuadReadAcrossY(int3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int4 QuadReadAcrossY(int4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint QuadReadAcrossY(uint);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint2 QuadReadAcrossY(uint2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint3 QuadReadAcrossY(uint3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint4 QuadReadAcrossY(uint4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t QuadReadAcrossY(int64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t2 QuadReadAcrossY(int64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t3 QuadReadAcrossY(int64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) int64_t4 QuadReadAcrossY(int64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t QuadReadAcrossY(uint64_t);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t2 QuadReadAcrossY(uint64_t2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t3 QuadReadAcrossY(uint64_t3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) uint64_t4 QuadReadAcrossY(uint64_t4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float QuadReadAcrossY(float);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float2 QuadReadAcrossY(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float3 QuadReadAcrossY(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) float4 QuadReadAcrossY(float4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double QuadReadAcrossY(double);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double2 QuadReadAcrossY(double2);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double3 QuadReadAcrossY(double3);
+_HLSL_BUILTIN_ALIAS(__builtin_hlsl_quad_read_across_y)
+__attribute__((convergent)) double4 QuadReadAcrossY(double4);
+
//===----------------------------------------------------------------------===//
// sign builtins
//===----------------------------------------------------------------------===//
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index 0619295cd2fbb..badbf9416d44f 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -4236,7 +4236,8 @@ bool SemaHLSL::CheckBuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
TheCall->setType(ArgTyExpr);
break;
}
- case Builtin::BI__builtin_hlsl_quad_read_across_x: {
+ case Builtin::BI__builtin_hlsl_quad_read_across_x:
+ case Builtin::BI__builtin_hlsl_quad_read_across_y: {
if (SemaRef.checkArgCount(TheCall, 1))
return true;
diff --git a/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl b/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl
new file mode 100644
index 0000000000000..a050aa9cf1722
--- /dev/null
+++ b/clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl
@@ -0,0 +1,46 @@
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN: dxil-pc-shadermodel6.3-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN: FileCheck %s --check-prefixes=CHECK,CHECK-DXIL
+// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -triple \
+// RUN: spirv-pc-vulkan-compute %s -emit-llvm -disable-llvm-passes -o - | \
+// RUN: FileCheck %s --check-prefixes=CHECK,CHECK-SPIRV
+
+// Test basic lowering to runtime function call.
+
+// CHECK-LABEL: test_int
+int test_int(int expr) {
+ // CHECK-SPIRV: %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.quad.read.across.y.i32([[TY]] %[[#]])
+ // CHECK-DXIL: %[[RET:.*]] = call [[TY:.*]] @llvm.dx.quad.read.across.y.i32([[TY]] %[[#]])
+ // CHECK: ret [[TY]] %[[RET]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.quad.read.across.y.i32([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare [[TY]] @llvm.spv.quad.read.across.y.i32([[TY]]) #[[#attr:]]
+
+// CHECK-LABEL: test_uint64_t
+uint64_t test_uint64_t(uint64_t expr) {
+ // CHECK-SPIRV: %[[RET:.*]] = call spir_func [[TY:.*]] @llvm.spv.quad.read.across.y.i64([[TY]] %[[#]])
+ // CHECK-DXIL: %[[RET:.*]] = call [[TY:.*]] @llvm.dx.quad.read.across.y.i64([[TY]] %[[#]])
+ // CHECK: ret [[TY]] %[[RET]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY]] @llvm.dx.quad.read.across.y.i64([[TY]]) #[[#attr:]]
+// CHECK-SPIRV: declare [[TY]] @llvm.spv.quad.read.across.y.i64([[TY]]) #[[#attr:]]
+
+// Test basic lowering to runtime function call with array and float value.
+
+// CHECK-LABEL: test_floatv4
+float4 test_floatv4(float4 expr) {
+ // CHECK-SPIRV: %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn spir_func [[TY1:.*]] @llvm.spv.quad.read.across.y.v4f32([[TY1]] %[[#]]
+ // CHECK-DXIL: %[[RET1:.*]] = call reassoc nnan ninf nsz arcp afn [[TY1:.*]] @llvm.dx.quad.read.across.y.v4f32([[TY1]] %[[#]])
+ // CHECK: ret [[TY1]] %[[RET1]]
+ return QuadReadAcrossY(expr);
+}
+
+// CHECK-DXIL: declare [[TY1]] @llvm.dx.quad.read.across.y.v4f32([[TY1]]) #[[#attr]]
+// CHECK-SPIRV: declare [[TY1]] @llvm.spv.quad.read.across.y.v4f32([[TY1]]) #[[#attr]]
+
+// CHECK: attributes #[[#attr]] = {{{.*}} convergent {{.*}}}
+
diff --git a/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl b/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl
new file mode 100644
index 0000000000000..2995895ff4c3d
--- /dev/null
+++ b/clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl
@@ -0,0 +1,28 @@
+// RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.6-library %s -emit-llvm-only -disable-llvm-passes -verify
+
+int test_too_few_arg() {
+ return __builtin_hlsl_quad_read_across_y();
+ // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
+}
+
+float2 test_too_many_arg(float2 p0) {
+ return __builtin_hlsl_quad_read_across_y(p0, p0);
+ // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+}
+
+bool test_expr_bool_type_check(bool p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'bool'}}
+}
+
+bool2 test_expr_bool_vec_type_check(bool2 p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'bool2' (aka 'vector<bool, 2>')}}
+}
+
+struct S { float f; };
+
+S test_expr_struct_type_check(S p0) {
+ return __builtin_hlsl_quad_read_across_y(p0);
+ // expected-error@-1 {{invalid operand of type 'S' where a scalar or vector is required}}
+}
diff --git a/llvm/include/llvm/IR/IntrinsicsDirectX.td b/llvm/include/llvm/IR/IntrinsicsDirectX.td
index 3f7922382f090..8ee07b32c1eca 100644
--- a/llvm/include/llvm/IR/IntrinsicsDirectX.td
+++ b/llvm/include/llvm/IR/IntrinsicsDirectX.td
@@ -255,6 +255,7 @@ def int_dx_wave_prefix_usum : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatch
def int_dx_wave_prefix_product : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_wave_prefix_uproduct : DefaultAttrsIntrinsic<[llvm_anyint_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_quad_read_across_x : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+def int_dx_quad_read_across_y : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_dx_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;
def int_dx_step : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty, LLVMMatchType<0>], [IntrNoMem]>;
def int_dx_splitdouble : DefaultAttrsIntrinsic<[llvm_anyint_ty, LLVMMatchType<0>],
diff --git a/llvm/include/llvm/IR/IntrinsicsSPIRV.td b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
index 5d467adb08c3d..6f030b7f2509f 100644
--- a/llvm/include/llvm/IR/IntrinsicsSPIRV.td
+++ b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
@@ -148,6 +148,7 @@ def int_spv_rsqrt : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty]
def int_spv_wave_prefix_sum : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_wave_prefix_product : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_quad_read_across_x : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
+ def int_spv_quad_read_across_y : DefaultAttrsIntrinsic<[llvm_any_ty], [LLVMMatchType<0>], [IntrConvergent, IntrNoMem]>;
def int_spv_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty], [IntrNoMem]>;
def int_spv_radians : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty], [IntrNoMem]>;
def int_spv_group_memory_barrier_with_group_sync : ClangBuiltin<"__builtin_spirv_group_barrier">,
diff --git a/llvm/lib/Target/DirectX/DXIL.td b/llvm/lib/Target/DirectX/DXIL.td
index 20de198ef96d2..13fc147e60216 100644
--- a/llvm/lib/Target/DirectX/DXIL.td
+++ b/llvm/lib/Target/DirectX/DXIL.td
@@ -1214,6 +1214,10 @@ def QuadOp : DXILOp<123, quadOp> {
[
IntrinArgIndex<0>, IntrinArgI8<QuadOpKind_ReadAcrossX>
]>,
+ IntrinSelect<int_dx_quad_read_across_y,
+ [
+ IntrinArgIndex<0>, IntrinArgI8<QuadOpKind_ReadAcrossY>
+ ]>,
];
let arguments = [OverloadTy, Int8Ty];
diff --git a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
index fab8bea379f74..69b942a171992 100644
--- a/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
+++ b/llvm/lib/Target/DirectX/DXILShaderFlags.cpp
@@ -107,6 +107,7 @@ static bool checkWaveOps(Intrinsic::ID IID) {
case Intrinsic::dx_wave_prefix_uproduct:
// Quad Op Variants
case Intrinsic::dx_quad_read_across_x:
+ case Intrinsic::dx_quad_read_across_y:
return true;
}
}
diff --git a/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp b/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
index 0c652b3bb29c0..830fcc9da862e 100644
--- a/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
+++ b/llvm/lib/Target/DirectX/DirectXTargetTransformInfo.cpp
@@ -77,6 +77,7 @@ bool DirectXTTIImpl::isTargetIntrinsicTriviallyScalarizable(
case Intrinsic::dx_wave_prefix_usum:
case Intrinsic::dx_wave_prefix_uproduct:
case Intrinsic::dx_quad_read_across_x:
+ case Intrinsic::dx_quad_read_across_y:
case Intrinsic::dx_imad:
case Intrinsic::dx_umad:
case Intrinsic::dx_ddx_coarse:
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index eb30002d2e1a5..abad90891b4f5 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -4457,6 +4457,9 @@ bool SPIRVInstructionSelector::selectIntrinsic(Register ResVReg,
case Intrinsic::spv_quad_read_across_x: {
return selectQuadSwap(ResVReg, ResType, I, /*Direction*/ 0);
}
+ case Intrinsic::spv_quad_read_across_y: {
+ return selectQuadSwap(ResVReg, ResType, I, /*Direction*/ 1);
+ }
case Intrinsic::spv_step:
return selectExtInst(ResVReg, ResType, I, CL::step, GL::Step);
case Intrinsic::spv_radians:
diff --git a/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll b/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll
new file mode 100644
index 0000000000000..d90ac759aa23d
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/QuadReadAcrossY.ll
@@ -0,0 +1,87 @@
+; RUN: opt -S -scalarizer -dxil-op-lower -mtriple=dxil-pc-shadermodel6.3-library < %s | FileCheck %s
+
+; Test that for scalar values, QuadReadAcrossY maps down to the DirectX op
+
+define noundef half @quad_read_across_y_half(half noundef %expr) {
+entry:
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr, i8 1)
+ %ret = call half @llvm.dx.quad.read.across.y.f16(half %expr)
+ ret half %ret
+}
+
+define noundef float @quad_read_across_y_float(float noundef %expr) {
+entry:
+; CHECK: call float @dx.op.quadOp.f32(i32 123, float %expr, i8 1)
+ %ret = call float @llvm.dx.quad.read.across.y.f32(float %expr)
+ ret float %ret
+}
+
+define noundef double @quad_read_across_y_double(double noundef %expr) {
+entry:
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr, i8 1)
+ %ret = call double @llvm.dx.quad.read.across.y.f64(double %expr)
+ ret double %ret
+}
+
+define noundef i16 @quad_read_across_y_i16(i16 noundef %expr) {
+entry:
+; CHECK: call i16 @dx.op.quadOp.i16(i32 123, i16 %expr, i8 1)
+ %ret = call i16 @llvm.dx.quad.read.across.y.i16(i16 %expr)
+ ret i16 %ret
+}
+
+define noundef i32 @quad_read_across_y_i32(i32 noundef %expr) {
+entry:
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr, i8 1)
+ %ret = call i32 @llvm.dx.quad.read.across.y.i32(i32 %expr)
+ ret i32 %ret
+}
+
+define noundef i64 @quad_read_across_y_i64(i64 noundef %expr) {
+entry:
+; CHECK: call i64 @dx.op.quadOp.i64(i32 123, i64 %expr, i8 1)
+ %ret = call i64 @llvm.dx.quad.read.across.y.i64(i64 %expr)
+ ret i64 %ret
+}
+
+declare half @llvm.dx.quad.read.across.y.f16(half)
+declare float @llvm.dx.quad.read.across.y.f32(float)
+declare double @llvm.dx.quad.read.across.y.f64(double)
+
+declare i16 @llvm.dx.quad.read.across.y.i16(i16)
+declare i32 @llvm.dx.quad.read.across.y.i32(i32)
+declare i64 @llvm.dx.quad.read.across.y.i64(i64)
+
+; Test that for vector values, QuadReadAcrossY scalarizes and maps down to the
+; DirectX op
+
+define noundef <2 x half> @quad_read_across_y_v2half(<2 x half> noundef %expr) {
+entry:
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr.i0, i8 1)
+; CHECK: call half @dx.op.quadOp.f16(i32 123, half %expr.i1, i8 1)
+ %ret = call <2 x half> @llvm.dx.quad.read.across.y.v2f16(<2 x half> %expr)
+ ret <2 x half> %ret
+}
+
+define noundef <3 x i32> @quad_read_across_y_v3i32(<3 x i32> noundef %expr) {
+entry:
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i0, i8 1)
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i1, i8 1)
+; CHECK: call i32 @dx.op.quadOp.i32(i32 123, i32 %expr.i2, i8 1)
+ %ret = call <3 x i32> @llvm.dx.quad.read.across.y.v3i32(<3 x i32> %expr)
+ ret <3 x i32> %ret
+}
+
+define noundef <4 x double> @quad_read_across_y_v4f64(<4 x double> noundef %expr) {
+entry:
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i0, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i1, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i2, i8 1)
+; CHECK: call double @dx.op.quadOp.f64(i32 123, double %expr.i3, i8 1)
+ ...
[truncated]
|
Contributor
Author
|
Pinging @bob80905 and @farzonl for review. The respective test PR can be found in llvm/offload-test-suite#993. |
kcloudy0717
force-pushed
the
kcloudy0717/QuadReadAcrossY
branch
from
31ae7cf to
1687ad7
Compare
Contributor
Author
|
@bob80905 This PR is good to go too. |
farzonl
reviewed
Mar 24, 2026
Comment on lines
+18
to
+19
| // CHECK-DXIL: declare [[TY]] @llvm.dx.quad.read.across.y.i32([[TY]]) #[[#attr:]] | ||
| // CHECK-SPIRV: declare [[TY]] @llvm.spv.quad.read.across.y.i32([[TY]]) #[[#attr:]] |
Member
There was a problem hiding this comment.
this is fine, but we do have a means of reducing the number of target specific check lines. See clang/test/CodeGenHLSL/builtins/dot.hlsl for an example:
// DXCHECK: %hlsl.dot = call i32 @llvm.[[ICF:dx]].sdot.v2i32(<2 x i32>
// SPVCHECK: %hlsl.dot = call i32 @llvm.[[ICF:spv]].sdot.v2i32(<2 x i32>
// CHECK: %hlsl.dot = call i32 @llvm.[[ICF]].sdot.v3i32(<3 x i32>
See how all successive checks just uses ICF after the first one?
Contributor
Author
There was a problem hiding this comment.
That's a neat trick! I didn't know about this, thanks for letting me know. I updated the codegen tests for the intrinsic, it now uses the ICF capture (as well as CC capture for spir_func calling convention) and it should have tests for all supported types including 16bit tests as well for the intrinsic. Could you give it a read?
Contributor
Author
There was a problem hiding this comment.
I'll apply the feedback to QuadReadAcrossX's test since those tests are not comprehensive as well. I'll open a separate PR for this and tag you for review.
farzonl
reviewed
Mar 24, 2026
| // Test basic lowering to runtime function call with array and float value. | ||
|
|
||
| // CHECK-LABEL: test_floatv4 | ||
| float4 test_floatv4(float4 expr) { |
Member
There was a problem hiding this comment.
We probably want some 16 bit testing aswell.
farzonl
reviewed
Mar 24, 2026
| @@ -0,0 +1,28 @@ | |||
| // RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.6-library %s -emit-llvm-only -disable-llvm-passes -verify | |||
Member
There was a problem hiding this comment.
Do we need -emit-llvm-only -disable-llvm-passes? This should just be sema checks we should never get to codegen. All that is needed is -verify
Contributor
Author
There was a problem hiding this comment.
No, this is probably unnecessary. The reason why it's here is because it was in majority of other error tests as well, I simply copied over. It should be removed now.
farzonl
approved these changes
Mar 24, 2026
Member
farzonl
left a comment
farzonl
left a comment
There was a problem hiding this comment.
Code LGTM. Tests need to be updated.
bob80905
approved these changes
Mar 25, 2026
Contributor
bob80905
left a comment
bob80905
left a comment
There was a problem hiding this comment.
LGTM, the unexpected passes are expected, now that we are adding support in clang.
Contributor
|
My This PR is the latest in HLSL area. Can I have any clue what might be wrong? |
Member
This isn’t the pr that changed things. |
Contributor
|
The issue resolves if I |
Contributor
I'm not sure why the I will investigate further |
Contributor
|
@Icohedron I think the destination directory must exist for |
Icohedron
added a commit
that referenced
this pull request
Mar 25, 2026
This PR should fix an issue reported by #187440 (comment) and #187610 (comment) where using `clang_generate_header` to generate a header into a directory that did not exist caused an error.
llvm-sync bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Mar 25, 2026
…(#188618) This PR should fix an issue reported by llvm/llvm-project#187440 (comment) and llvm/llvm-project#187610 (comment) where using `clang_generate_header` to generate a header into a directory that did not exist caused an error.
bogner
pushed a commit
to llvm/offload-test-suite
that referenced
this pull request
Mar 26, 2026
We can remove the XFAIL directives for Clang config for QuadReadAcrossY tests now that QuadReadAcrossY intrinsic has landed in LLVM: llvm/llvm-project#187440. I also appended `convergence` suffix for QuadReadAcrossY control flow test file. Co-authored-by: Finn Plummer <[email protected]>
ambergorzynski
pushed a commit
to ambergorzynski/llvm-project
that referenced
this pull request
Mar 27, 2026
This PR adds QuadReadAcrossY intrinsic support in HLSL with codegen for both DirectX and SPIRV backends. Resolves llvm#99176. - [x] Implement `QuadReadAcrossY` clang builtin, - [x] Link `QuadReadAcrossY` clang builtin with `hlsl_intrinsics.h` - [x] Add sema checks for `QuadReadAcrossY` to `CheckHLSLBuiltinFunctionCall` in `SemaChecking.cpp` - [x] Add codegen for `QuadReadAcrossY` to `EmitHLSLBuiltinExpr` in `CGBuiltin.cpp` - [x] Add codegen tests to `clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl` - [x] Add sema tests to `clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl` - [x] Create the `int_dx_QuadReadAcrossY` intrinsic in `IntrinsicsDirectX.td` - [x] Create the `DXILOpMapping` of `int_dx_QuadReadAcrossY` to `123` in `DXIL.td` - [x] Create the `QuadReadAcrossY.ll` and `QuadReadAcrossY_errors.ll` tests in `llvm/test/CodeGen/DirectX/` - [x] Create the `int_spv_QuadReadAcrossY` intrinsic in `IntrinsicsSPIRV.td` - [x] In SPIRVInstructionSelector.cpp create the `QuadReadAcrossY` lowering and map it to `int_spv_QuadReadAcrossY` in `SPIRVInstructionSelector::selectIntrinsic`. - [x] Create SPIR-V backend test case in `llvm/test/CodeGen/SPIRV/hlsl-intrinsics/QuadReadAcrossY.ll`
ambergorzynski
pushed a commit
to ambergorzynski/llvm-project
that referenced
this pull request
Mar 27, 2026
This PR should fix an issue reported by llvm#187440 (comment) and llvm#187610 (comment) where using `clang_generate_header` to generate a header into a directory that did not exist caused an error.
Aadarsh-Keshri
pushed a commit
to Aadarsh-Keshri/llvm-project
that referenced
this pull request
Mar 28, 2026
This PR adds QuadReadAcrossY intrinsic support in HLSL with codegen for both DirectX and SPIRV backends. Resolves llvm#99176. - [x] Implement `QuadReadAcrossY` clang builtin, - [x] Link `QuadReadAcrossY` clang builtin with `hlsl_intrinsics.h` - [x] Add sema checks for `QuadReadAcrossY` to `CheckHLSLBuiltinFunctionCall` in `SemaChecking.cpp` - [x] Add codegen for `QuadReadAcrossY` to `EmitHLSLBuiltinExpr` in `CGBuiltin.cpp` - [x] Add codegen tests to `clang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlsl` - [x] Add sema tests to `clang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlsl` - [x] Create the `int_dx_QuadReadAcrossY` intrinsic in `IntrinsicsDirectX.td` - [x] Create the `DXILOpMapping` of `int_dx_QuadReadAcrossY` to `123` in `DXIL.td` - [x] Create the `QuadReadAcrossY.ll` and `QuadReadAcrossY_errors.ll` tests in `llvm/test/CodeGen/DirectX/` - [x] Create the `int_spv_QuadReadAcrossY` intrinsic in `IntrinsicsSPIRV.td` - [x] In SPIRVInstructionSelector.cpp create the `QuadReadAcrossY` lowering and map it to `int_spv_QuadReadAcrossY` in `SPIRVInstructionSelector::selectIntrinsic`. - [x] Create SPIR-V backend test case in `llvm/test/CodeGen/SPIRV/hlsl-intrinsics/QuadReadAcrossY.ll`
Aadarsh-Keshri
pushed a commit
to Aadarsh-Keshri/llvm-project
that referenced
this pull request
Mar 28, 2026
This PR should fix an issue reported by llvm#187440 (comment) and llvm#187610 (comment) where using `clang_generate_header` to generate a header into a directory that did not exist caused an error.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR adds QuadReadAcrossY intrinsic support in HLSL with codegen for both DirectX and SPIRV backends. Resolves #99176.
QuadReadAcrossYclang builtin,QuadReadAcrossYclang builtin withhlsl_intrinsics.hQuadReadAcrossYtoCheckHLSLBuiltinFunctionCallinSemaChecking.cppQuadReadAcrossYtoEmitHLSLBuiltinExprinCGBuiltin.cppclang/test/CodeGenHLSL/builtins/QuadReadAcrossY.hlslclang/test/SemaHLSL/BuiltIns/QuadReadAcrossY-errors.hlslint_dx_QuadReadAcrossYintrinsic inIntrinsicsDirectX.tdDXILOpMappingofint_dx_QuadReadAcrossYto123inDXIL.tdQuadReadAcrossY.llandQuadReadAcrossY_errors.lltests inllvm/test/CodeGen/DirectX/int_spv_QuadReadAcrossYintrinsic inIntrinsicsSPIRV.tdQuadReadAcrossYlowering and map it toint_spv_QuadReadAcrossYinSPIRVInstructionSelector::selectIntrinsic.llvm/test/CodeGen/SPIRV/hlsl-intrinsics/QuadReadAcrossY.ll