-
Notifications
You must be signed in to change notification settings - Fork 12.7k
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.
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
[DXIL] Add radians intrinsic (#110616) #113546
[DXIL] Add radians intrinsic (#110616) #113546
Conversation
Created using spr 1.3.4
@llvm/pr-subscribers-backend-directx Author: None (agozillon) Changesmakes progress on #99151 Changes
Related PRsFull diff: https://github.com/llvm/llvm-project/pull/113546.diff 3 Files Affected:
diff --git a/llvm/include/llvm/IR/IntrinsicsDirectX.td b/llvm/include/llvm/IR/IntrinsicsDirectX.td
index 1adfcdc9a1ed2f..f2b9e286ebb476 100644
--- a/llvm/include/llvm/IR/IntrinsicsDirectX.td
+++ b/llvm/include/llvm/IR/IntrinsicsDirectX.td
@@ -86,4 +86,5 @@ def int_dx_rsqrt : DefaultAttrsIntrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>]
def int_dx_wave_is_first_lane : DefaultAttrsIntrinsic<[llvm_i1_ty], [], [IntrConvergent]>;
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_radians : DefaultAttrsIntrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem]>;
}
diff --git a/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp b/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
index 39e73a11d9d7ee..0bcd03c7fad38d 100644
--- a/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
+++ b/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
@@ -64,6 +64,7 @@ static bool isIntrinsicExpansion(Function &F) {
case Intrinsic::dx_udot:
case Intrinsic::dx_sign:
case Intrinsic::dx_step:
+ case Intrinsic::dx_radians:
return true;
}
return false;
@@ -442,6 +443,14 @@ static Value *expandStepIntrinsic(CallInst *Orig) {
return Builder.CreateSelect(Cond, Zero, One);
}
+static Value *expandRadiansIntrinsic(CallInst *Orig) {
+ Value *X = Orig->getOperand(0);
+ Type *Ty = X->getType();
+ IRBuilder<> Builder(Orig);
+ Value *PiOver180 = ConstantFP::get(Ty, llvm::numbers::pi / 180.0);
+ return Builder.CreateFMul(X, PiOver180);
+}
+
static Intrinsic::ID getMaxForClamp(Type *ElemTy,
Intrinsic::ID ClampIntrinsic) {
if (ClampIntrinsic == Intrinsic::dx_uclamp)
@@ -561,6 +570,9 @@ static bool expandIntrinsic(Function &F, CallInst *Orig) {
break;
case Intrinsic::dx_step:
Result = expandStepIntrinsic(Orig);
+ case Intrinsic::dx_radians:
+ Result = expandRadiansIntrinsic(Orig);
+ break;
}
if (Result) {
Orig->replaceAllUsesWith(Result);
diff --git a/llvm/test/CodeGen/DirectX/radians.ll b/llvm/test/CodeGen/DirectX/radians.ll
new file mode 100644
index 00000000000000..73ec013775c3e9
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/radians.ll
@@ -0,0 +1,79 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -dxil-intrinsic-expansion -scalarizer -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
+
+declare half @llvm.dx.radians.f16(half)
+declare float @llvm.dx.radians.f32(float)
+
+declare <4 x half> @llvm.dx.radians.v4f16(<4 x half>)
+declare <4 x float> @llvm.dx.radians.v4f32(<4 x float>)
+
+define noundef half @radians_half(half noundef %a) {
+; CHECK-LABEL: define noundef half @radians_half(
+; CHECK-SAME: half noundef [[A:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[TMP0:%.*]] = fmul half [[A]], 0xH2478
+; CHECK-NEXT: ret half [[TMP0]]
+;
+entry:
+ %elt.radians = call half @llvm.dx.radians.f16(half %a)
+ ret half %elt.radians
+}
+
+define noundef float @radians_float(float noundef %a) {
+; CHECK-LABEL: define noundef float @radians_float(
+; CHECK-SAME: float noundef [[A:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[TMP0:%.*]] = fmul float [[A]], 0x3F91DF46A0000000
+; CHECK-NEXT: ret float [[TMP0]]
+;
+entry:
+ %elt.radians = call float @llvm.dx.radians.f32(float %a)
+ ret float %elt.radians
+}
+
+define noundef <4 x half> @radians_half_vector(<4 x half> noundef %a) {
+; CHECK-LABEL: define noundef <4 x half> @radians_half_vector(
+; CHECK-SAME: <4 x half> noundef [[A:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK: [[ee0:%.*]] = extractelement <4 x half> [[A]], i64 0
+; CHECK: [[ie0:%.*]] = fmul half [[ee0]], 0xH2478
+; CHECK: [[ee1:%.*]] = extractelement <4 x half> [[A]], i64 1
+; CHECK: [[ie1:%.*]] = fmul half [[ee1]], 0xH2478
+; CHECK: [[ee2:%.*]] = extractelement <4 x half> [[A]], i64 2
+; CHECK: [[ie2:%.*]] = fmul half [[ee2]], 0xH2478
+; CHECK: [[ee3:%.*]] = extractelement <4 x half> [[A]], i64 3
+; CHECK: [[ie3:%.*]] = fmul half [[ee3]], 0xH2478
+; CHECK: [[TMP0:%.*]] = insertelement <4 x half> poison, half [[ie0]], i64 0
+; CHECK: [[TMP1:%.*]] = insertelement <4 x half> %[[TMP0]], half [[ie1]], i64 1
+; CHECK: [[TMP2:%.*]] = insertelement <4 x half> %[[TMP1]], half [[ie2]], i64 2
+; CHECK: [[TMP3:%.*]] = insertelement <4 x half> %[[TMP2]], half [[ie3]], i64 3
+; CHECK: ret <4 x half> [[TMP3]]
+;
+entry:
+ %elt.radians = call <4 x half> @llvm.dx.radians.v4f16(<4 x half> %a)
+ ret <4 x half> %elt.radians
+}
+
+define noundef <4 x float> @radians_float_vector(<4 x float> noundef %a) {
+; CHECK-LABEL: define noundef <4 x float> @radians_float_vector(
+; CHECK-SAME: <4 x float> noundef [[A:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK: [[ee0:%.*]] = extractelement <4 x float> [[A]], i64 0
+; CHECK: [[ie0:%.*]] = fmul float [[ee0]], 0x3F91DF46A0000000
+; CHECK: [[ee1:%.*]] = extractelement <4 x float> [[A]], i64 1
+; CHECK: [[ie1:%.*]] = fmul float [[ee1]], 0x3F91DF46A0000000
+; CHECK: [[ee2:%.*]] = extractelement <4 x float> [[A]], i64 2
+; CHECK: [[ie2:%.*]] = fmul float [[ee2]], 0x3F91DF46A0000000
+; CHECK: [[ee3:%.*]] = extractelement <4 x float> [[A]], i64 3
+; CHECK: [[ie3:%.*]] = fmul float [[ee3]], 0x3F91DF46A0000000
+; CHECK: [[TMP0:%.*]] = insertelement <4 x float> poison, float [[ie0]], i64 0
+; CHECK: [[TMP1:%.*]] = insertelement <4 x float> %[[TMP0]], float [[ie1]], i64 1
+; CHECK: [[TMP2:%.*]] = insertelement <4 x float> %[[TMP1]], float [[ie2]], i64 2
+; CHECK: [[TMP3:%.*]] = insertelement <4 x float> %[[TMP2]], float [[ie3]], i64 3
+; CHECK: ret <4 x float> [[TMP3]]
+;
+entry:
+ %elt.radians = call <4 x float> @llvm.dx.radians.v4f32(<4 x float> %a)
+ ret <4 x float> %elt.radians
+}
+
|
@llvm/pr-subscribers-llvm-ir Author: None (agozillon) Changesmakes progress on #99151 Changes
Related PRsFull diff: https://github.com/llvm/llvm-project/pull/113546.diff 3 Files Affected:
diff --git a/llvm/include/llvm/IR/IntrinsicsDirectX.td b/llvm/include/llvm/IR/IntrinsicsDirectX.td
index 1adfcdc9a1ed2f..f2b9e286ebb476 100644
--- a/llvm/include/llvm/IR/IntrinsicsDirectX.td
+++ b/llvm/include/llvm/IR/IntrinsicsDirectX.td
@@ -86,4 +86,5 @@ def int_dx_rsqrt : DefaultAttrsIntrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>]
def int_dx_wave_is_first_lane : DefaultAttrsIntrinsic<[llvm_i1_ty], [], [IntrConvergent]>;
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_radians : DefaultAttrsIntrinsic<[llvm_anyfloat_ty], [LLVMMatchType<0>], [IntrNoMem]>;
}
diff --git a/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp b/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
index 39e73a11d9d7ee..0bcd03c7fad38d 100644
--- a/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
+++ b/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
@@ -64,6 +64,7 @@ static bool isIntrinsicExpansion(Function &F) {
case Intrinsic::dx_udot:
case Intrinsic::dx_sign:
case Intrinsic::dx_step:
+ case Intrinsic::dx_radians:
return true;
}
return false;
@@ -442,6 +443,14 @@ static Value *expandStepIntrinsic(CallInst *Orig) {
return Builder.CreateSelect(Cond, Zero, One);
}
+static Value *expandRadiansIntrinsic(CallInst *Orig) {
+ Value *X = Orig->getOperand(0);
+ Type *Ty = X->getType();
+ IRBuilder<> Builder(Orig);
+ Value *PiOver180 = ConstantFP::get(Ty, llvm::numbers::pi / 180.0);
+ return Builder.CreateFMul(X, PiOver180);
+}
+
static Intrinsic::ID getMaxForClamp(Type *ElemTy,
Intrinsic::ID ClampIntrinsic) {
if (ClampIntrinsic == Intrinsic::dx_uclamp)
@@ -561,6 +570,9 @@ static bool expandIntrinsic(Function &F, CallInst *Orig) {
break;
case Intrinsic::dx_step:
Result = expandStepIntrinsic(Orig);
+ case Intrinsic::dx_radians:
+ Result = expandRadiansIntrinsic(Orig);
+ break;
}
if (Result) {
Orig->replaceAllUsesWith(Result);
diff --git a/llvm/test/CodeGen/DirectX/radians.ll b/llvm/test/CodeGen/DirectX/radians.ll
new file mode 100644
index 00000000000000..73ec013775c3e9
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/radians.ll
@@ -0,0 +1,79 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -dxil-intrinsic-expansion -scalarizer -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
+
+declare half @llvm.dx.radians.f16(half)
+declare float @llvm.dx.radians.f32(float)
+
+declare <4 x half> @llvm.dx.radians.v4f16(<4 x half>)
+declare <4 x float> @llvm.dx.radians.v4f32(<4 x float>)
+
+define noundef half @radians_half(half noundef %a) {
+; CHECK-LABEL: define noundef half @radians_half(
+; CHECK-SAME: half noundef [[A:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[TMP0:%.*]] = fmul half [[A]], 0xH2478
+; CHECK-NEXT: ret half [[TMP0]]
+;
+entry:
+ %elt.radians = call half @llvm.dx.radians.f16(half %a)
+ ret half %elt.radians
+}
+
+define noundef float @radians_float(float noundef %a) {
+; CHECK-LABEL: define noundef float @radians_float(
+; CHECK-SAME: float noundef [[A:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK-NEXT: [[TMP0:%.*]] = fmul float [[A]], 0x3F91DF46A0000000
+; CHECK-NEXT: ret float [[TMP0]]
+;
+entry:
+ %elt.radians = call float @llvm.dx.radians.f32(float %a)
+ ret float %elt.radians
+}
+
+define noundef <4 x half> @radians_half_vector(<4 x half> noundef %a) {
+; CHECK-LABEL: define noundef <4 x half> @radians_half_vector(
+; CHECK-SAME: <4 x half> noundef [[A:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK: [[ee0:%.*]] = extractelement <4 x half> [[A]], i64 0
+; CHECK: [[ie0:%.*]] = fmul half [[ee0]], 0xH2478
+; CHECK: [[ee1:%.*]] = extractelement <4 x half> [[A]], i64 1
+; CHECK: [[ie1:%.*]] = fmul half [[ee1]], 0xH2478
+; CHECK: [[ee2:%.*]] = extractelement <4 x half> [[A]], i64 2
+; CHECK: [[ie2:%.*]] = fmul half [[ee2]], 0xH2478
+; CHECK: [[ee3:%.*]] = extractelement <4 x half> [[A]], i64 3
+; CHECK: [[ie3:%.*]] = fmul half [[ee3]], 0xH2478
+; CHECK: [[TMP0:%.*]] = insertelement <4 x half> poison, half [[ie0]], i64 0
+; CHECK: [[TMP1:%.*]] = insertelement <4 x half> %[[TMP0]], half [[ie1]], i64 1
+; CHECK: [[TMP2:%.*]] = insertelement <4 x half> %[[TMP1]], half [[ie2]], i64 2
+; CHECK: [[TMP3:%.*]] = insertelement <4 x half> %[[TMP2]], half [[ie3]], i64 3
+; CHECK: ret <4 x half> [[TMP3]]
+;
+entry:
+ %elt.radians = call <4 x half> @llvm.dx.radians.v4f16(<4 x half> %a)
+ ret <4 x half> %elt.radians
+}
+
+define noundef <4 x float> @radians_float_vector(<4 x float> noundef %a) {
+; CHECK-LABEL: define noundef <4 x float> @radians_float_vector(
+; CHECK-SAME: <4 x float> noundef [[A:%.*]]) {
+; CHECK-NEXT: [[ENTRY:.*:]]
+; CHECK: [[ee0:%.*]] = extractelement <4 x float> [[A]], i64 0
+; CHECK: [[ie0:%.*]] = fmul float [[ee0]], 0x3F91DF46A0000000
+; CHECK: [[ee1:%.*]] = extractelement <4 x float> [[A]], i64 1
+; CHECK: [[ie1:%.*]] = fmul float [[ee1]], 0x3F91DF46A0000000
+; CHECK: [[ee2:%.*]] = extractelement <4 x float> [[A]], i64 2
+; CHECK: [[ie2:%.*]] = fmul float [[ee2]], 0x3F91DF46A0000000
+; CHECK: [[ee3:%.*]] = extractelement <4 x float> [[A]], i64 3
+; CHECK: [[ie3:%.*]] = fmul float [[ee3]], 0x3F91DF46A0000000
+; CHECK: [[TMP0:%.*]] = insertelement <4 x float> poison, float [[ie0]], i64 0
+; CHECK: [[TMP1:%.*]] = insertelement <4 x float> %[[TMP0]], float [[ie1]], i64 1
+; CHECK: [[TMP2:%.*]] = insertelement <4 x float> %[[TMP1]], float [[ie2]], i64 2
+; CHECK: [[TMP3:%.*]] = insertelement <4 x float> %[[TMP2]], float [[ie3]], i64 3
+; CHECK: ret <4 x float> [[TMP3]]
+;
+entry:
+ %elt.radians = call <4 x float> @llvm.dx.radians.v4f32(<4 x float> %a)
+ ret <4 x float> %elt.radians
+}
+
|
Closing as it was an unintentional PR generation via SPR (someone elses PR that's already landed) |
makes progress on #99151
Changes
Related PRs
#110802
#110800