Added legalizer and instruction selector for @llvm.lround.* and @llvm.llround.* intrinsic #126473
Conversation
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
…vm.llround.* intrinsic --added test files for @llvm.lround.* and @llvm.llround.* intrinsic --MachineBasicBlock.h file was just copied from the github (no changes made)
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be notified. If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers. If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
Member
|
@llvm/pr-subscribers-backend-spir-v Author: Aadesh P S (AadeshPS) Changes…vm.llround.* intrinsic --added test files for @llvm.lround.* and @llvm.llround.* intrinsic --MachineBasicBlock.h file was just copied from the github (no changes made) Full diff: https://github.com/llvm/llvm-project/pull/126473.diff 5 Files Affected:
diff --git a/llvm/include/llvm/CodeGen/MachineBasicBlock.h b/llvm/include/llvm/CodeGen/MachineBasicBlock.h
index 0b803a97247422c..e8e5a385c931ac1 100644
--- a/llvm/include/llvm/CodeGen/MachineBasicBlock.h
+++ b/llvm/include/llvm/CodeGen/MachineBasicBlock.h
@@ -1460,4 +1460,4 @@ inline auto instructionsWithoutDebug(IterT It, IterT End,
} // end namespace llvm
-#endif // LLVM_CODEGEN_MACHINEBASICBLOCK_H
+#endif // LLVM_CODEGEN_MACHINEBASICBLOCK_H
\ No newline at end of file
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index e7d8fe5bd8015b1..303887d16fa6c8c 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -284,7 +284,12 @@ class SPIRVInstructionSelector : public InstructionSelector {
GL::GLSLExtInst GLInst) const;
bool selectExtInst(Register ResVReg, const SPIRVType *ResType,
MachineInstr &I, const ExtInstList &ExtInsts) const;
-
+ bool selectExtInstForLRound(Register ResVReg, const SPIRVType *ResType,
+ MachineInstr &I, CL::OpenCLExtInst CLInst,
+ GL::GLSLExtInst GLInst) const;
+ bool selectExtInstForLRound(Register ResVReg, const SPIRVType *ResType,
+ MachineInstr &I, const ExtInstList &ExtInsts) const;
+
bool selectLog10(Register ResVReg, const SPIRVType *ResType,
MachineInstr &I) const;
@@ -625,6 +630,22 @@ bool SPIRVInstructionSelector::spvSelect(Register ResVReg,
case TargetOpcode::G_UCMP:
return selectSUCmp(ResVReg, ResType, I, false);
+ case TargetOpcode::G_LROUND:
+ case TargetOpcode::G_LLROUND:{
+ Register Result = MRI->createVirtualRegister(MRI -> getRegClass(ResVReg), "hello");
+ MRI->setRegClass(Result,&SPIRV::iIDRegClass);
+ GR.assignSPIRVTypeToVReg(GR.getSPIRVTypeForVReg(I.getOperand(1).getReg()), Result, *(I.getParent() -> getParent()));
+ bool Rounded = selectExtInstForLRound(Result, GR.getSPIRVTypeForVReg(Result), I, CL::round, GL::Round);
+ if(Rounded){
+ MachineBasicBlock &BB = *I.getParent();
+ MachineFunction &MF = *BB.getParent();
+ auto MIB = BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRV::OpConvertFToS))
+ .addDef(ResVReg)
+ .addUse(GR.getSPIRVTypeID(ResType))
+ .addUse(Result);
+ return MIB.constrainAllUses(TII, TRI, RBI);
+ }
+ }
case TargetOpcode::G_STRICT_FMA:
case TargetOpcode::G_FMA:
return selectExtInst(ResVReg, ResType, I, CL::fma, GL::Fma);
@@ -956,6 +977,47 @@ bool SPIRVInstructionSelector::selectExtInst(Register ResVReg,
return false;
}
+bool SPIRVInstructionSelector::selectExtInstForLRound(Register ResVReg,
+ const SPIRVType *ResType,
+ MachineInstr &I,
+ CL::OpenCLExtInst CLInst,
+ GL::GLSLExtInst GLInst) const {
+ ExtInstList ExtInsts = {{SPIRV::InstructionSet::OpenCL_std, CLInst},
+ {SPIRV::InstructionSet::GLSL_std_450, GLInst}};
+ return selectExtInstForLRound(ResVReg, ResType, I, ExtInsts);
+}
+
+bool SPIRVInstructionSelector::selectExtInstForLRound(Register ResVReg,
+ const SPIRVType *ResType,
+ MachineInstr &I,
+ const ExtInstList &Insts) const {
+
+ for (const auto &Ex : Insts) {
+ SPIRV::InstructionSet::InstructionSet Set = Ex.first;
+ uint32_t Opcode = Ex.second;
+ if (STI.canUseExtInstSet(Set)) {
+ MachineBasicBlock &BB = *I.getParent();
+ auto MIB = BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRV::OpExtInst))
+ .addDef(ResVReg)
+ .addUse(GR.getSPIRVTypeID(ResType))
+ .addImm(static_cast<uint32_t>(Set))
+ .addImm(Opcode);
+ const unsigned NumOps = I.getNumOperands();
+ unsigned Index = 1;
+ if (Index < NumOps &&
+ I.getOperand(Index).getType() ==
+ MachineOperand::MachineOperandType::MO_IntrinsicID)
+ Index = 2;
+ for (; Index < NumOps; ++Index)
+ MIB.add(I.getOperand(Index));
+ MIB.constrainAllUses(TII, TRI, RBI);
+ return true;
+ }
+ }
+ return false;
+}
+
+
bool SPIRVInstructionSelector::selectOpWithSrcs(Register ResVReg,
const SPIRVType *ResType,
MachineInstr &I,
diff --git a/llvm/lib/Target/SPIRV/SPIRVLegalizerInfo.cpp b/llvm/lib/Target/SPIRV/SPIRVLegalizerInfo.cpp
index fa5e0a80576d01b..e9b73df0bb12dce 100644
--- a/llvm/lib/Target/SPIRV/SPIRVLegalizerInfo.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVLegalizerInfo.cpp
@@ -305,6 +305,10 @@ SPIRVLegalizerInfo::SPIRVLegalizerInfo(const SPIRVSubtarget &ST) {
{G_UADDO, G_SADDO, G_USUBO, G_SSUBO, G_UMULO, G_SMULO})
.alwaysLegal();
+ getActionDefinitionsBuilder({G_LROUND, G_LLROUND})
+ .legalForCartesianProduct(allFloatScalarsAndVectors, allIntScalarsAndVectors);
+
+
// FP conversions.
getActionDefinitionsBuilder({G_FPTRUNC, G_FPEXT})
.legalForCartesianProduct(allFloatScalarsAndVectors);
diff --git a/llvm/test/CodeGen/SPIRV/llvm-intrinsics/llround.ll b/llvm/test/CodeGen/SPIRV/llvm-intrinsics/llround.ll
new file mode 100644
index 000000000000000..7794a6ed46b0b48
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/llvm-intrinsics/llround.ll
@@ -0,0 +1,108 @@
+; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+
+
+; CHECK: [[opencl:%[0-9]+]] = OpExtInstImport "OpenCL.std"
+; CHECK-DAG: [[f32:%[0-9]+]] = OpTypeFloat 32
+; CHECK-DAG: [[i32:%[0-9]+]] = OpTypeInt 32 0
+; CHECK-DAG: [[f64:%[0-9]+]] = OpTypeFloat 64
+; CHECK-DAG: [[i64:%[0-9]+]] = OpTypeInt 64 0
+; CHECK-DAG: [[vecf32:%[0-9]+]] = OpTypeVector [[f32]]
+; CHECK-DAG: [[veci32:%[0-9]+]] = OpTypeVector [[i32]]
+; CHECK-DAG: [[vecf64:%[0-9]+]] = OpTypeVector [[f64]]
+; CHECK-DAG: [[veci64:%[0-9]+]] = OpTypeVector [[i64]]
+
+; CHECK: [[rounded_i32_f32:%[0-9]+]] = OpExtInst [[f32]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[i32]] [[rounded_i32_f32]]
+; CHECK: [[rounded_i32_f64:%[0-9]+]] = OpExtInst [[f64]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[i32]] [[rounded_i32_f64]]
+; CHECK: [[rounded_i64_f32:%[0-9]+]] = OpExtInst [[f32]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[i64]] [[rounded_i64_f32]]
+; CHECK: [[rounded_i64_f64:%[0-9]+]] = OpExtInst [[f64]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[i64]] [[rounded_i64_f64]]
+; CHECK: [[rounded_v4i32_f32:%[0-9]+]] = OpExtInst [[vecf32]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[veci32]] [[rounded_v4i32_f32]]
+; CHECK: [[rounded_v4i32_f64:%[0-9]+]] = OpExtInst [[vecf64]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[veci32]] [[rounded_v4i32_f64]]
+; CHECK: [[rounded_v4i64_f32:%[0-9]+]] = OpExtInst [[vecf32]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[veci64]] [[rounded_v4i64_f32]]
+; CHECK: [[rounded_v4i64_f64:%[0-9]+]] = OpExtInst [[vecf64]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[veci64]] [[rounded_v4i64_f64]]
+
+
+target triple = "spir64-unknown-unknown"
+
+define spir_func i32 @test_llround_i32_f32(float %arg0) {
+entry:
+
+ %0 = call i32 @llvm.llround.i32.f32(float %arg0)
+ ret i32 %0
+}
+
+define spir_func i32 @test_llround_i32_f64(double %arg0) {
+entry:
+
+
+ %0 = call i32 @llvm.llround.i32.f64(double %arg0)
+ ret i32 %0
+}
+
+define spir_func i64 @test_llround_i64_f32(float %arg0) {
+entry:
+
+
+ %0 = call i64 @llvm.llround.i64.f32(float %arg0)
+ ret i64 %0
+}
+
+define spir_func i64 @test_llround_i64_f64(double %arg0) {
+entry:
+
+
+ %0 = call i64 @llvm.llround.i64.f64(double %arg0)
+ ret i64 %0
+}
+
+define spir_func <4 x i32> @test_llround_v4i32_f32(<4 x float> %arg0) {
+entry:
+
+
+ %0 = call <4 x i32> @llvm.llround.v4i32.f32(<4 x float> %arg0)
+ ret <4 x i32> %0
+}
+
+
+define spir_func <4 x i32> @test_llround_v4i32_f64(<4 x double> %arg0) {
+entry:
+
+
+ %0 = call <4 x i32> @llvm.llround.v4i32.f64(<4 x double> %arg0)
+ ret <4 x i32> %0
+}
+
+define spir_func <4 x i64> @test_llround_v4i64_f32(<4 x float> %arg0) {
+entry:
+
+
+ %0 = call <4 x i64> @llvm.llround.v4i64.f32(<4 x float> %arg0)
+ ret <4 x i64> %0
+}
+
+
+define spir_func <4 x i64> @test_llround_v4i64_f64(<4 x double> %arg0) {
+entry:
+
+ %0 = call <4 x i64> @llvm.llround.v4i64.f64(<4 x double> %arg0)
+ ret <4 x i64> %0
+}
+
+declare i32 @llvm.llround.i32.f32(float)
+declare i32 @llvm.llround.i32.f64(double)
+declare i64 @llvm.llround.i64.f32(float)
+declare i64 @llvm.llround.i64.f64(double)
+
+declare <4 x i32> @llvm.llround.v4i32.f32(<4 x float>)
+declare <4 x i32> @llvm.llround.v4i32.f64(<4 x double>)
+declare <4 x i64> @llvm.llround.v4i64.f32(<4 x float>)
+declare <4 x i64> @llvm.llround.v4i64.f64(<4 x double>)
\ No newline at end of file
diff --git a/llvm/test/CodeGen/SPIRV/llvm-intrinsics/lround.ll b/llvm/test/CodeGen/SPIRV/llvm-intrinsics/lround.ll
new file mode 100644
index 000000000000000..34060c62c5ed2b2
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/llvm-intrinsics/lround.ll
@@ -0,0 +1,108 @@
+; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+
+
+; CHECK: [[opencl:%[0-9]+]] = OpExtInstImport "OpenCL.std"
+; CHECK-DAG: [[f32:%[0-9]+]] = OpTypeFloat 32
+; CHECK-DAG: [[i32:%[0-9]+]] = OpTypeInt 32 0
+; CHECK-DAG: [[f64:%[0-9]+]] = OpTypeFloat 64
+; CHECK-DAG: [[i64:%[0-9]+]] = OpTypeInt 64 0
+; CHECK-DAG: [[vecf32:%[0-9]+]] = OpTypeVector [[f32]]
+; CHECK-DAG: [[veci32:%[0-9]+]] = OpTypeVector [[i32]]
+; CHECK-DAG: [[vecf64:%[0-9]+]] = OpTypeVector [[f64]]
+; CHECK-DAG: [[veci64:%[0-9]+]] = OpTypeVector [[i64]]
+
+; CHECK: [[rounded_i32_f32:%[0-9]+]] = OpExtInst [[f32]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[i32]] [[rounded_i32_f32]]
+; CHECK: [[rounded_i32_f64:%[0-9]+]] = OpExtInst [[f64]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[i32]] [[rounded_i32_f64]]
+; CHECK: [[rounded_i64_f32:%[0-9]+]] = OpExtInst [[f32]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[i64]] [[rounded_i64_f32]]
+; CHECK: [[rounded_i64_f64:%[0-9]+]] = OpExtInst [[f64]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[i64]] [[rounded_i64_f64]]
+; CHECK: [[rounded_v4i32_f32:%[0-9]+]] = OpExtInst [[vecf32]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[veci32]] [[rounded_v4i32_f32]]
+; CHECK: [[rounded_v4i32_f64:%[0-9]+]] = OpExtInst [[vecf64]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[veci32]] [[rounded_v4i32_f64]]
+; CHECK: [[rounded_v4i64_f32:%[0-9]+]] = OpExtInst [[vecf32]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[veci64]] [[rounded_v4i64_f32]]
+; CHECK: [[rounded_v4i64_f64:%[0-9]+]] = OpExtInst [[vecf64]] [[opencl]] round %[[#]]
+; CHECK-NEXT: %[[#]] = OpConvertFToS [[veci64]] [[rounded_v4i64_f64]]
+
+
+target triple = "spir64-unknown-unknown"
+
+define spir_func i32 @test_lround_i32_f32(float %arg0) {
+entry:
+
+ %0 = call i32 @llvm.lround.i32.f32(float %arg0)
+ ret i32 %0
+}
+
+define spir_func i32 @test_lround_i32_f64(double %arg0) {
+entry:
+
+
+ %0 = call i32 @llvm.lround.i32.f64(double %arg0)
+ ret i32 %0
+}
+
+define spir_func i64 @test_lround_i64_f32(float %arg0) {
+entry:
+
+
+ %0 = call i64 @llvm.lround.i64.f32(float %arg0)
+ ret i64 %0
+}
+
+define spir_func i64 @test_lround_i64_f64(double %arg0) {
+entry:
+
+
+ %0 = call i64 @llvm.lround.i64.f64(double %arg0)
+ ret i64 %0
+}
+
+define spir_func <4 x i32> @test_lround_v4i32_f32(<4 x float> %arg0) {
+entry:
+
+
+ %0 = call <4 x i32> @llvm.lround.v4i32.f32(<4 x float> %arg0)
+ ret <4 x i32> %0
+}
+
+
+define spir_func <4 x i32> @test_lround_v4i32_f64(<4 x double> %arg0) {
+entry:
+
+
+ %0 = call <4 x i32> @llvm.lround.v4i32.f64(<4 x double> %arg0)
+ ret <4 x i32> %0
+}
+
+define spir_func <4 x i64> @test_lround_v4i64_f32(<4 x float> %arg0) {
+entry:
+
+
+ %0 = call <4 x i64> @llvm.lround.v4i64.f32(<4 x float> %arg0)
+ ret <4 x i64> %0
+}
+
+
+define spir_func <4 x i64> @test_lround_v4i64_f64(<4 x double> %arg0) {
+entry:
+
+ %0 = call <4 x i64> @llvm.lround.v4i64.f64(<4 x double> %arg0)
+ ret <4 x i64> %0
+}
+
+declare i32 @llvm.lround.i32.f32(float)
+declare i32 @llvm.lround.i32.f64(double)
+declare i64 @llvm.lround.i64.f32(float)
+declare i64 @llvm.lround.i64.f64(double)
+
+declare <4 x i32> @llvm.lround.v4i32.f32(<4 x float>)
+declare <4 x i32> @llvm.lround.v4i32.f64(<4 x double>)
+declare <4 x i64> @llvm.lround.v4i64.f32(<4 x float>)
+declare <4 x i64> @llvm.lround.v4i64.f64(<4 x double>)
\ No newline at end of file
|
AadeshPS
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
Added legalizer and instruction selector for @llvm.lround.* and @llvm.llround.* intrinsic
Added test files for @llvm.lround.* and @llvm.llround.* intrinsic
MachineBasicBlock.h file was just copied from the github (no changes made)