[DirectX][Scalarizer] Make sure TargetTransformInfoWrapperPass is setup - #216893
Open
farzonl wants to merge 1 commit into
Open
[DirectX][Scalarizer] Make sure TargetTransformInfoWrapperPass is setup#216893farzonl wants to merge 1 commit into
farzonl wants to merge 1 commit into
Conversation
fixes llvm#216888 In the process of testing the scalarization of HLSL long vectors I discoved that some intrinsics were crashing when used with llc but not when used with opt -scalarizer. This change makes sure that we have setup TargetTransformInfoWrapperPass so that we can call the DirectX versions of `isTargetIntrinsicWithScalarOpAtArg` and `isTargetIntrinsicWithOverloadTypeAtArg`.
|
@llvm/pr-subscribers-llvm-transforms @llvm/pr-subscribers-backend-directx Author: Farzon Lotfi (farzonl) Changesfixes #216888 In the process of testing the scalarization of HLSL long vectors I discoved that some intrinsics were crashing when used with llc but not when used with opt -scalarizer. This change makes sure that we have setup TargetTransformInfoWrapperPass so that we can call the DirectX versions of Patch is 54.47 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/216893.diff 77 Files Affected:
diff --git a/llvm/lib/Target/DirectX/DirectXTargetMachine.cpp b/llvm/lib/Target/DirectX/DirectXTargetMachine.cpp
index 1240e7cda597e..13d6b4f2ae81a 100644
--- a/llvm/lib/Target/DirectX/DirectXTargetMachine.cpp
+++ b/llvm/lib/Target/DirectX/DirectXTargetMachine.cpp
@@ -34,6 +34,7 @@
#include "DirectXSubtarget.h"
#include "DirectXTargetTransformInfo.h"
#include "TargetInfo/DirectXTargetInfo.h"
+#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/CodeGen/MachineModuleInfo.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/CodeGen/TargetPassConfig.h"
@@ -170,6 +171,7 @@ bool DirectXTargetMachine::addPassesToEmitFile(
CodeGenFileType FileType, bool DisableVerify,
MachineModuleInfoWrapperPass *MMIWP) {
TargetPassConfig *PassConfig = createPassConfig(PM);
+ PM.add(createTargetTransformInfoWrapperPass(getTargetIRAnalysis()));
PassConfig->addCodeGenPrepare();
switch (FileType) {
diff --git a/llvm/lib/Transforms/Scalar/Scalarizer.cpp b/llvm/lib/Transforms/Scalar/Scalarizer.cpp
index 006f4c13d045e..280a36ca02897 100644
--- a/llvm/lib/Transforms/Scalar/Scalarizer.cpp
+++ b/llvm/lib/Transforms/Scalar/Scalarizer.cpp
@@ -364,6 +364,7 @@ char ScalarizerLegacyPass::ID = 0;
INITIALIZE_PASS_BEGIN(ScalarizerLegacyPass, "scalarizer",
"Scalarize vector operations", false, false)
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
+INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass)
INITIALIZE_PASS_END(ScalarizerLegacyPass, "scalarizer",
"Scalarize vector operations", false, false)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/asdouble-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/asdouble-sm69.ll
new file mode 100644
index 0000000000000..7d31c31a87a6e
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/asdouble-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x double> @test_asdouble(
+; CHECK: call <17 x double> @llvm.dx.asdouble.v17i32
+define <17 x double> @test_asdouble(<17 x i32> %a, <17 x i32> %b) {
+ %result = call <17 x double> @llvm.dx.asdouble.v17i32(<17 x i32> %a, <17 x i32> %b)
+ ret <17 x double> %result
+}
+declare <17 x double> @llvm.dx.asdouble.v17i32(<17 x i32>, <17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/asdouble.ll b/llvm/test/CodeGen/DirectX/LongVector/asdouble.ll
new file mode 100644
index 0000000000000..84c20636975b1
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/asdouble.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x double> @test_asdouble(
+; CHECK-COUNT-17: call double @dx.op.makeDouble.f64(i32 101, i32 {{.*}}, i32 {{.*}})
+define <17 x double> @test_asdouble(<17 x i32> %a, <17 x i32> %b) {
+ %result = call <17 x double> @llvm.dx.asdouble.v17i32(<17 x i32> %a, <17 x i32> %b)
+ ret <17 x double> %result
+}
+declare <17 x double> @llvm.dx.asdouble.v17i32(<17 x i32>, <17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/ddx-coarse-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/ddx-coarse-sm69.ll
new file mode 100644
index 0000000000000..20419156a6634
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/ddx-coarse-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_ddx_coarse(
+; CHECK: call <17 x float> @llvm.dx.ddx.coarse.v17f32
+define <17 x float> @test_ddx_coarse(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.ddx.coarse.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.ddx.coarse.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/ddx-coarse.ll b/llvm/test/CodeGen/DirectX/LongVector/ddx-coarse.ll
new file mode 100644
index 0000000000000..0c8aa488a85db
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/ddx-coarse.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_ddx_coarse(
+; CHECK-COUNT-17: call float @dx.op.unary.f32(i32 83, float {{.*}})
+define <17 x float> @test_ddx_coarse(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.ddx.coarse.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.ddx.coarse.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/ddx-fine-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/ddx-fine-sm69.ll
new file mode 100644
index 0000000000000..c9444084ba33b
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/ddx-fine-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_ddx_fine(
+; CHECK: call <17 x float> @llvm.dx.ddx.fine.v17f32
+define <17 x float> @test_ddx_fine(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.ddx.fine.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.ddx.fine.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/ddx-fine.ll b/llvm/test/CodeGen/DirectX/LongVector/ddx-fine.ll
new file mode 100644
index 0000000000000..55aa92d918e40
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/ddx-fine.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_ddx_fine(
+; CHECK-COUNT-17: call float @dx.op.unary.f32(i32 85, float {{.*}})
+define <17 x float> @test_ddx_fine(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.ddx.fine.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.ddx.fine.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/ddy-coarse-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/ddy-coarse-sm69.ll
new file mode 100644
index 0000000000000..d092521ad008b
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/ddy-coarse-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_ddy_coarse(
+; CHECK: call <17 x float> @llvm.dx.ddy.coarse.v17f32
+define <17 x float> @test_ddy_coarse(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.ddy.coarse.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.ddy.coarse.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/ddy-coarse.ll b/llvm/test/CodeGen/DirectX/LongVector/ddy-coarse.ll
new file mode 100644
index 0000000000000..45ca4c052740b
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/ddy-coarse.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_ddy_coarse(
+; CHECK-COUNT-17: call float @dx.op.unary.f32(i32 84, float {{.*}})
+define <17 x float> @test_ddy_coarse(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.ddy.coarse.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.ddy.coarse.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/ddy-fine-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/ddy-fine-sm69.ll
new file mode 100644
index 0000000000000..4bba04ccc3b9d
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/ddy-fine-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_ddy_fine(
+; CHECK: call <17 x float> @llvm.dx.ddy.fine.v17f32
+define <17 x float> @test_ddy_fine(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.ddy.fine.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.ddy.fine.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/ddy-fine.ll b/llvm/test/CodeGen/DirectX/LongVector/ddy-fine.ll
new file mode 100644
index 0000000000000..8baf8297cbb86
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/ddy-fine.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_ddy_fine(
+; CHECK-COUNT-17: call float @dx.op.unary.f32(i32 86, float {{.*}})
+define <17 x float> @test_ddy_fine(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.ddy.fine.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.ddy.fine.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/firstbitlow-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/firstbitlow-sm69.ll
new file mode 100644
index 0000000000000..0fef613afb9ac
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/firstbitlow-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i32> @test_firstbitlow(
+; CHECK: call <17 x i32> @llvm.dx.firstbitlow.v17i32
+define <17 x i32> @test_firstbitlow(<17 x i32> %a) {
+ %result = call <17 x i32> @llvm.dx.firstbitlow.v17i32(<17 x i32> %a)
+ ret <17 x i32> %result
+}
+declare <17 x i32> @llvm.dx.firstbitlow.v17i32(<17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/firstbitlow.ll b/llvm/test/CodeGen/DirectX/LongVector/firstbitlow.ll
new file mode 100644
index 0000000000000..e4cce428612e0
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/firstbitlow.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i32> @test_firstbitlow(
+; CHECK-COUNT-17: call i32 @dx.op.unaryBits.i32(i32 32, i32 {{.*}})
+define <17 x i32> @test_firstbitlow(<17 x i32> %a) {
+ %result = call <17 x i32> @llvm.dx.firstbitlow.v17i32(<17 x i32> %a)
+ ret <17 x i32> %result
+}
+declare <17 x i32> @llvm.dx.firstbitlow.v17i32(<17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/firstbitshigh-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/firstbitshigh-sm69.ll
new file mode 100644
index 0000000000000..61a6a4ff308d3
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/firstbitshigh-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i32> @test_firstbitshigh(
+; CHECK: call <17 x i32> @llvm.dx.firstbitshigh.v17i32
+define <17 x i32> @test_firstbitshigh(<17 x i32> %a) {
+ %result = call <17 x i32> @llvm.dx.firstbitshigh.v17i32(<17 x i32> %a)
+ ret <17 x i32> %result
+}
+declare <17 x i32> @llvm.dx.firstbitshigh.v17i32(<17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/firstbitshigh.ll b/llvm/test/CodeGen/DirectX/LongVector/firstbitshigh.ll
new file mode 100644
index 0000000000000..24ed1f172f3d5
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/firstbitshigh.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i32> @test_firstbitshigh(
+; CHECK-COUNT-17: call i32 @dx.op.unaryBits.i32(i32 34, i32 {{.*}})
+define <17 x i32> @test_firstbitshigh(<17 x i32> %a) {
+ %result = call <17 x i32> @llvm.dx.firstbitshigh.v17i32(<17 x i32> %a)
+ ret <17 x i32> %result
+}
+declare <17 x i32> @llvm.dx.firstbitshigh.v17i32(<17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/firstbituhigh-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/firstbituhigh-sm69.ll
new file mode 100644
index 0000000000000..2dbad82686316
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/firstbituhigh-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i32> @test_firstbituhigh(
+; CHECK: call <17 x i32> @llvm.dx.firstbituhigh.v17i32
+define <17 x i32> @test_firstbituhigh(<17 x i32> %a) {
+ %result = call <17 x i32> @llvm.dx.firstbituhigh.v17i32(<17 x i32> %a)
+ ret <17 x i32> %result
+}
+declare <17 x i32> @llvm.dx.firstbituhigh.v17i32(<17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/firstbituhigh.ll b/llvm/test/CodeGen/DirectX/LongVector/firstbituhigh.ll
new file mode 100644
index 0000000000000..75b177d49b2a1
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/firstbituhigh.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i32> @test_firstbituhigh(
+; CHECK-COUNT-17: call i32 @dx.op.unaryBits.i32(i32 33, i32 {{.*}})
+define <17 x i32> @test_firstbituhigh(<17 x i32> %a) {
+ %result = call <17 x i32> @llvm.dx.firstbituhigh.v17i32(<17 x i32> %a)
+ ret <17 x i32> %result
+}
+declare <17 x i32> @llvm.dx.firstbituhigh.v17i32(<17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/frac-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/frac-sm69.ll
new file mode 100644
index 0000000000000..b8857c0db1e67
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/frac-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_frac(
+; CHECK: call <17 x float> @llvm.dx.frac.v17f32
+define <17 x float> @test_frac(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.frac.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.frac.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/frac.ll b/llvm/test/CodeGen/DirectX/LongVector/frac.ll
new file mode 100644
index 0000000000000..241303c8049fd
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/frac.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_frac(
+; CHECK-COUNT-17: call float @dx.op.unary.f32(i32 22, float {{.*}})
+define <17 x float> @test_frac(<17 x float> %a) {
+ %result = call <17 x float> @llvm.dx.frac.v17f32(<17 x float> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.frac.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/imad-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/imad-sm69.ll
new file mode 100644
index 0000000000000..267dd33adc6b0
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/imad-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i32> @test_imad(
+; CHECK: call <17 x i32> @llvm.dx.imad.v17i32
+define <17 x i32> @test_imad(<17 x i32> %a, <17 x i32> %b, <17 x i32> %c) {
+ %result = call <17 x i32> @llvm.dx.imad.v17i32(<17 x i32> %a, <17 x i32> %b, <17 x i32> %c)
+ ret <17 x i32> %result
+}
+declare <17 x i32> @llvm.dx.imad.v17i32(<17 x i32>, <17 x i32>, <17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/imad.ll b/llvm/test/CodeGen/DirectX/LongVector/imad.ll
new file mode 100644
index 0000000000000..0613133320f15
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/imad.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i32> @test_imad(
+; CHECK-COUNT-17: call i32 @dx.op.tertiary.i32(i32 48, i32 {{.*}}, i32 {{.*}}, i32 {{.*}})
+define <17 x i32> @test_imad(<17 x i32> %a, <17 x i32> %b, <17 x i32> %c) {
+ %result = call <17 x i32> @llvm.dx.imad.v17i32(<17 x i32> %a, <17 x i32> %b, <17 x i32> %c)
+ ret <17 x i32> %result
+}
+declare <17 x i32> @llvm.dx.imad.v17i32(<17 x i32>, <17 x i32>, <17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/isinf-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/isinf-sm69.ll
new file mode 100644
index 0000000000000..593dae2066da2
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/isinf-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i1> @test_isinf(
+; CHECK: call <17 x i1> @llvm.dx.isinf.v17f32
+define <17 x i1> @test_isinf(<17 x float> %a) {
+ %result = call <17 x i1> @llvm.dx.isinf.v17f32(<17 x float> %a)
+ ret <17 x i1> %result
+}
+declare <17 x i1> @llvm.dx.isinf.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/isinf.ll b/llvm/test/CodeGen/DirectX/LongVector/isinf.ll
new file mode 100644
index 0000000000000..c12b2774e5f54
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/isinf.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i1> @test_isinf(
+; CHECK-COUNT-17: call i1 @dx.op.isSpecialFloat.f32(i32 9, float {{.*}})
+define <17 x i1> @test_isinf(<17 x float> %a) {
+ %result = call <17 x i1> @llvm.dx.isinf.v17f32(<17 x float> %a)
+ ret <17 x i1> %result
+}
+declare <17 x i1> @llvm.dx.isinf.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/isnan-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/isnan-sm69.ll
new file mode 100644
index 0000000000000..003819d323d90
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/isnan-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i1> @test_isnan(
+; CHECK: call <17 x i1> @llvm.dx.isnan.v17f32
+define <17 x i1> @test_isnan(<17 x float> %a) {
+ %result = call <17 x i1> @llvm.dx.isnan.v17f32(<17 x float> %a)
+ ret <17 x i1> %result
+}
+declare <17 x i1> @llvm.dx.isnan.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/isnan.ll b/llvm/test/CodeGen/DirectX/LongVector/isnan.ll
new file mode 100644
index 0000000000000..91af7f2ac3dac
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/isnan.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i1> @test_isnan(
+; CHECK-COUNT-17: call i1 @dx.op.isSpecialFloat.f32(i32 8, float {{.*}})
+define <17 x i1> @test_isnan(<17 x float> %a) {
+ %result = call <17 x i1> @llvm.dx.isnan.v17f32(<17 x float> %a)
+ ret <17 x i1> %result
+}
+declare <17 x i1> @llvm.dx.isnan.v17f32(<17 x float>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/legacyf16tof32-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/legacyf16tof32-sm69.ll
new file mode 100644
index 0000000000000..d256f642e8d0c
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/legacyf16tof32-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_legacyf16tof32(
+; CHECK: call <17 x float> @llvm.dx.legacyf16tof32.v17i32
+define <17 x float> @test_legacyf16tof32(<17 x i32> %a) {
+ %result = call <17 x float> @llvm.dx.legacyf16tof32.v17i32(<17 x i32> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.legacyf16tof32.v17i32(<17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/legacyf16tof32.ll b/llvm/test/CodeGen/DirectX/LongVector/legacyf16tof32.ll
new file mode 100644
index 0000000000000..1360002cdb3e6
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/legacyf16tof32.ll
@@ -0,0 +1,9 @@
+; RUN: llc -mtriple=dxil-pc-shadermodel6.8-library -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x float> @test_legacyf16tof32(
+; CHECK-COUNT-17: call float @dx.op.legacyF16ToF32(i32 131, i32 {{.*}})
+define <17 x float> @test_legacyf16tof32(<17 x i32> %a) {
+ %result = call <17 x float> @llvm.dx.legacyf16tof32.v17i32(<17 x i32> %a)
+ ret <17 x float> %result
+}
+declare <17 x float> @llvm.dx.legacyf16tof32.v17i32(<17 x i32>)
diff --git a/llvm/test/CodeGen/DirectX/LongVector/legacyf32tof16-sm69.ll b/llvm/test/CodeGen/DirectX/LongVector/legacyf32tof16-sm69.ll
new file mode 100644
index 0000000000000..8dc5c3586ffc5
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/LongVector/legacyf32tof16-sm69.ll
@@ -0,0 +1,10 @@
+; XFAIL: *
+; RUN: llc -mtriple=dxil-pc-shadermodel6.9-library -stop-before=dxil-op-lower -o - %s | FileCheck %s
+
+; CHECK-LABEL: define <17 x i32> @test_legacyf32tof16(
+; CHECK: call <17 x i32> @llvm.dx.legacyf32tof16.v17f32
+define <17 x i32> @test_legacyf32tof16(<17 x float> %a) {
+ %result = call <17 x i32> @llvm.dx.legacyf32tof16.v17f32(<17 x float> %a)
+ ret <17 x i32> %result
+}
+declare <17 x...
[truncated]
|
Member
Author
|
@bogner I have a change to turn off the scalarizer for SM6.9, but doing it right now would break all codegen if you target sm 6.9. So to track that work instead of turning off the scalarizer now I am just going to add the tests as |
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
1 participant
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.
fixes #216888
In the process of testing the scalarization of HLSL long vectors I discoved that some intrinsics were crashing when used with llc but not when used with opt -scalarizer. This change makes sure that we have setup TargetTransformInfoWrapperPass so that we can call the DirectX versions of
isTargetIntrinsicWithScalarOpAtArgandisTargetIntrinsicWithOverloadTypeAtArg.Assisted by Copilot with GPT 5.6-Sol