[HLSL][DXIL] implement sqrt intrinsic
#86560
Merged
Conversation
This file contains 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
completes llvm#86187 - fix hlsl_intrinsic to cover the correct cases - move to using `__builtin_elementwise_sqrt` - add lowering of `Intrinsic::sqrt` to dxilop 24.
farzonl
requested review from
bogner,
llvm-beanz,
bharadwajy,
python3kgae and
bob80905
llvmbot
added
clang
|
@llvm/pr-subscribers-hlsl @llvm/pr-subscribers-backend-x86 Author: Farzon Lotfi (farzonl) Changescompletes #86187
Full diff: https://github.com/llvm/llvm-project/pull/86560.diff 5 Files Affected:
diff --git a/clang/lib/Headers/hlsl/hlsl_intrinsics.h b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
index d9f27c0db57ce4..fcb64fde1b91f1 100644
--- a/clang/lib/Headers/hlsl/hlsl_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
@@ -1366,14 +1366,26 @@ float4 sin(float4);
/// \param Val The input value.
_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
-_HLSL_BUILTIN_ALIAS(__builtin_sqrtf16)
-half sqrt(half In);
-
-_HLSL_BUILTIN_ALIAS(__builtin_sqrtf)
-float sqrt(float In);
-
-_HLSL_BUILTIN_ALIAS(__builtin_sqrt)
-double sqrt(double In);
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_sqrt)
+half sqrt(half);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_sqrt)
+half2 sqrt(half2);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_sqrt)
+half3 sqrt(half3);
+_HLSL_16BIT_AVAILABILITY(shadermodel, 6.2)
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_sqrt)
+half4 sqrt(half4);
+
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_sqrt)
+float sqrt(float);
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_sqrt)
+float2 sqrt(float2);
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_sqrt)
+float3 sqrt(float3);
+_HLSL_BUILTIN_ALIAS(__builtin_elementwise_sqrt)
+float4 sqrt(float4);
//===----------------------------------------------------------------------===//
// trunc builtins
diff --git a/clang/test/CodeGenHLSL/builtins/sqrt.hlsl b/clang/test/CodeGenHLSL/builtins/sqrt.hlsl
index 2c2a09617cf86a..adbbf69a8e0685 100644
--- a/clang/test/CodeGenHLSL/builtins/sqrt.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/sqrt.hlsl
@@ -1,29 +1,53 @@
-// RUN: %clang_cc1 -std=hlsl2021 -finclude-default-header -x hlsl -triple \
-// RUN: dxil-pc-shadermodel6.2-library %s -fnative-half-type \
-// RUN: -emit-llvm -disable-llvm-passes -o - | FileCheck %s
+// RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
+// RUN: dxil-pc-shadermodel6.3-library %s -fnative-half-type \
+// RUN: -emit-llvm -disable-llvm-passes -o - | FileCheck %s \
+// RUN: --check-prefixes=CHECK,NATIVE_HALF
+// RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
+// RUN: dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
+// RUN: -o - | FileCheck %s --check-prefixes=CHECK,NO_HALF
-using hlsl::sqrt;
+// NATIVE_HALF: define noundef half @
+// NATIVE_HALF: %{{.*}} = call half @llvm.sqrt.f16(
+// NATIVE_HALF: ret half %{{.*}}
+// NO_HALF: define noundef float @"?test_sqrt_half@@YA$halff@$halff@@Z"(
+// NO_HALF: %{{.*}} = call float @llvm.sqrt.f32(
+// NO_HALF: ret float %{{.*}}
+half test_sqrt_half(half p0) { return sqrt(p0); }
+// NATIVE_HALF: define noundef <2 x half> @
+// NATIVE_HALF: %{{.*}} = call <2 x half> @llvm.sqrt.v2f16
+// NATIVE_HALF: ret <2 x half> %{{.*}}
+// NO_HALF: define noundef <2 x float> @
+// NO_HALF: %{{.*}} = call <2 x float> @llvm.sqrt.v2f32(
+// NO_HALF: ret <2 x float> %{{.*}}
+half2 test_sqrt_half2(half2 p0) { return sqrt(p0); }
+// NATIVE_HALF: define noundef <3 x half> @
+// NATIVE_HALF: %{{.*}} = call <3 x half> @llvm.sqrt.v3f16
+// NATIVE_HALF: ret <3 x half> %{{.*}}
+// NO_HALF: define noundef <3 x float> @
+// NO_HALF: %{{.*}} = call <3 x float> @llvm.sqrt.v3f32(
+// NO_HALF: ret <3 x float> %{{.*}}
+half3 test_sqrt_half3(half3 p0) { return sqrt(p0); }
+// NATIVE_HALF: define noundef <4 x half> @
+// NATIVE_HALF: %{{.*}} = call <4 x half> @llvm.sqrt.v4f16
+// NATIVE_HALF: ret <4 x half> %{{.*}}
+// NO_HALF: define noundef <4 x float> @
+// NO_HALF: %{{.*}} = call <4 x float> @llvm.sqrt.v4f32(
+// NO_HALF: ret <4 x float> %{{.*}}
+half4 test_sqrt_half4(half4 p0) { return sqrt(p0); }
-double sqrt_d(double x)
-{
- return sqrt(x);
-}
-
-// CHECK: define noundef double @"?sqrt_d@@YANN@Z"(
-// CHECK: call double @llvm.sqrt.f64(double %0)
-
-float sqrt_f(float x)
-{
- return sqrt(x);
-}
-
-// CHECK: define noundef float @"?sqrt_f@@YAMM@Z"(
-// CHECK: call float @llvm.sqrt.f32(float %0)
-
-half sqrt_h(half x)
-{
- return sqrt(x);
-}
-
-// CHECK: define noundef half @"?sqrt_h@@YA$f16@$f16@@Z"(
-// CHECK: call half @llvm.sqrt.f16(half %0)
+// CHECK: define noundef float @
+// CHECK: %{{.*}} = call float @llvm.sqrt.f32(
+// CHECK: ret float %{{.*}}
+float test_sqrt_float(float p0) { return sqrt(p0); }
+// CHECK: define noundef <2 x float> @
+// CHECK: %{{.*}} = call <2 x float> @llvm.sqrt.v2f32
+// CHECK: ret <2 x float> %{{.*}}
+float2 test_sqrt_float2(float2 p0) { return sqrt(p0); }
+// CHECK: define noundef <3 x float> @
+// CHECK: %{{.*}} = call <3 x float> @llvm.sqrt.v3f32
+// CHECK: ret <3 x float> %{{.*}}
+float3 test_sqrt_float3(float3 p0) { return sqrt(p0); }
+// CHECK: define noundef <4 x float> @
+// CHECK: %{{.*}} = call <4 x float> @llvm.sqrt.v4f32
+// CHECK: ret <4 x float> %{{.*}}
+float4 test_sqrt_float4(float4 p0) { return sqrt(p0); }
diff --git a/llvm/lib/Target/DirectX/DXIL.td b/llvm/lib/Target/DirectX/DXIL.td
index f7e69ebae15b6c..572d3323ebe75f 100644
--- a/llvm/lib/Target/DirectX/DXIL.td
+++ b/llvm/lib/Target/DirectX/DXIL.td
@@ -274,6 +274,10 @@ def Frac : DXILOpMapping<22, unary, int_dx_frac,
"Returns a fraction from 0 to 1 that represents the "
"decimal part of the input.",
[llvm_halforfloat_ty, LLVMMatchType<0>]>;
+def Sqrt : DXILOpMapping<24, unary, int_sqrt,
+ "Returns the square root of the specified floating-point"
+ "value, per component.",
+ [llvm_halforfloat_ty, LLVMMatchType<0>]>;
def RSqrt : DXILOpMapping<25, unary, int_dx_rsqrt,
"Returns the reciprocal of the square root of the specified value."
"rsqrt(x) = 1 / sqrt(x).",
diff --git a/llvm/test/CodeGen/DirectX/sqrt.ll b/llvm/test/CodeGen/DirectX/sqrt.ll
new file mode 100644
index 00000000000000..76a572efd20557
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/sqrt.ll
@@ -0,0 +1,20 @@
+; RUN: opt -S -dxil-op-lower < %s | FileCheck %s
+
+; Make sure dxil operation function calls for sqrt are generated for float and half.
+
+define noundef float @sqrt_float(float noundef %a) #0 {
+entry:
+; CHECK:call float @dx.op.unary.f32(i32 24, float %{{.*}})
+ %elt.sqrt = call float @llvm.sqrt.f32(float %a)
+ ret float %elt.sqrt
+}
+
+define noundef half @sqrt_half(half noundef %a) #0 {
+entry:
+; CHECK:call half @dx.op.unary.f16(i32 24, half %{{.*}})
+ %elt.sqrt = call half @llvm.sqrt.f16(half %a)
+ ret half %elt.sqrt
+}
+
+declare half @llvm.sqrt.f16(half)
+declare float @llvm.sqrt.f32(float)
diff --git a/llvm/test/CodeGen/DirectX/sqrt_error.ll b/llvm/test/CodeGen/DirectX/sqrt_error.ll
new file mode 100644
index 00000000000000..fffa2e19b80fa9
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/sqrt_error.ll
@@ -0,0 +1,10 @@
+; RUN: not opt -S -dxil-op-lower %s 2>&1 | FileCheck %s
+
+; DXIL operation sqrt does not support double overload type
+; CHECK: LLVM ERROR: Invalid Overload Type
+
+define noundef double @sqrt_double(double noundef %a) {
+entry:
+ %elt.sqrt = call double @llvm.sqrt.f64(double %a)
+ ret double %elt.sqrt
+}
|
llvm-beanz
approved these changes
Mar 25, 2024
|
✅ With the latest revision this PR passed the Python code formatter. |
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
python3kgae
approved these changes
Mar 25, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
completes #86187
__builtin_elementwise_sqrtIntrinsic::sqrtto dxilop 24.