[CIR][OpenMP] Add OpenMP-to-LLVM type conversion for CIR-to-LLVM lowering#190063
Merged
[CIR][OpenMP] Add OpenMP-to-LLVM type conversion for CIR-to-LLVM lowering#190063
Conversation
jsjodin
requested review from
andykaylor,
bcardosolopes,
lanza and
xlauko
as code owners
llvmbot
added
clang
Member
|
@llvm/pr-subscribers-clang @llvm/pr-subscribers-clangir Author: Jan Leyonberg (jsjodin) ChangesRegister OpenMP conversion legality and patterns in the CIR-to-LLVM pass so that OpenMP operations (e.g. omp.map.info, omp.target) have their CIR types properly converted to LLVM types during lowering. Without this, the conversion leaves behind unrealized_conversion_cast ops that cause translation to LLVM IR to fail. Also registers omp::PointerLikeType on cir::PointerType so that CIR pointers are accepted as operands in OpenMP map operations. Assised-by: Cursor / Claude Opus 4.6 Full diff: https://github.com/llvm/llvm-project/pull/190063.diff 4 Files Affected:
diff --git a/clang/lib/CIR/Dialect/OpenMP/RegisterOpenMPExtensions.cpp b/clang/lib/CIR/Dialect/OpenMP/RegisterOpenMPExtensions.cpp
index b5129202e66c4..3a66f93238808 100644
--- a/clang/lib/CIR/Dialect/OpenMP/RegisterOpenMPExtensions.cpp
+++ b/clang/lib/CIR/Dialect/OpenMP/RegisterOpenMPExtensions.cpp
@@ -11,8 +11,20 @@
//===----------------------------------------------------------------------===//
#include "clang/CIR/Dialect/OpenMP/RegisterOpenMPExtensions.h"
+#include "mlir/Dialect/OpenMP/OpenMPDialect.h"
#include "mlir/Dialect/OpenMP/OpenMPInterfaces.h"
#include "clang/CIR/Dialect/IR/CIRDialect.h"
+#include "clang/CIR/Dialect/IR/CIRTypes.h"
+
+namespace {
+struct OpenMPPointerLikeModel
+ : public mlir::omp::PointerLikeType::ExternalModel<OpenMPPointerLikeModel,
+ cir::PointerType> {
+ mlir::Type getElementType(mlir::Type pointer) const {
+ return mlir::cast<cir::PointerType>(pointer).getPointee();
+ }
+};
+} // namespace
namespace cir::omp {
@@ -20,6 +32,7 @@ void registerOpenMPExtensions(mlir::DialectRegistry ®istry) {
registry.addExtension(+[](mlir::MLIRContext *ctx, cir::CIRDialect *dialect) {
cir::FuncOp::attachInterface<
mlir::omp::DeclareTargetDefaultModel<cir::FuncOp>>(*ctx);
+ cir::PointerType::attachInterface<OpenMPPointerLikeModel>(*ctx);
});
}
diff --git a/clang/lib/CIR/Lowering/DirectToLLVM/CMakeLists.txt b/clang/lib/CIR/Lowering/DirectToLLVM/CMakeLists.txt
index 021397fee992b..00c6102f039f1 100644
--- a/clang/lib/CIR/Lowering/DirectToLLVM/CMakeLists.txt
+++ b/clang/lib/CIR/Lowering/DirectToLLVM/CMakeLists.txt
@@ -21,6 +21,7 @@ add_clang_library(clangCIRLoweringDirectToLLVM
MLIRCIRTargetLowering
MLIRBuiltinToLLVMIRTranslation
MLIRLLVMToLLVMIRTranslation
+ MLIROpenMPToLLVM
MLIROpenMPToLLVMIRTranslation
MLIROpenMPTransforms
MLIRIR
diff --git a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
index 149cd90b813ec..d8af7e22e0587 100644
--- a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
+++ b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
@@ -17,10 +17,12 @@
#include <optional>
#include "mlir/Conversion/LLVMCommon/TypeConverter.h"
+#include "mlir/Conversion/OpenMPToLLVM/ConvertOpenMPToLLVM.h"
#include "mlir/Dialect/DLTI/DLTI.h"
#include "mlir/Dialect/Func/IR/FuncOps.h"
#include "mlir/Dialect/LLVMIR/LLVMDialect.h"
#include "mlir/Dialect/LLVMIR/LLVMTypes.h"
+#include "mlir/Dialect/OpenMP/OpenMPDialect.h"
#include "mlir/Dialect/OpenMP/Transforms/Passes.h"
#include "mlir/Dialect/Ptr/IR/MemorySpaceInterfaces.h"
#include "mlir/IR/BuiltinAttributes.h"
@@ -3678,6 +3680,9 @@ void ConvertCIRToLLVMPass::runOnOperation() {
mlir::ConversionTarget target(getContext());
target.addLegalOp<mlir::ModuleOp>();
target.addLegalDialect<mlir::LLVM::LLVMDialect>();
+ mlir::configureOpenMPToLLVMConversionLegality(target, converter);
+ target.addLegalDialect<mlir::omp::OpenMPDialect>();
+ mlir::populateOpenMPToLLVMConversionPatterns(converter, patterns);
target.addIllegalDialect<mlir::BuiltinDialect, cir::CIRDialect,
mlir::func::FuncDialect>();
diff --git a/clang/test/CIR/Lowering/omp-target-map.cir b/clang/test/CIR/Lowering/omp-target-map.cir
new file mode 100644
index 0000000000000..b885cffc2bf48
--- /dev/null
+++ b/clang/test/CIR/Lowering/omp-target-map.cir
@@ -0,0 +1,30 @@
+// RUN: cir-opt %s --cir-to-llvm | FileCheck %s
+
+// Verify that OpenMP operations have their CIR types properly converted to
+// LLVM types during CIR-to-LLVM lowering. Without the OpenMP-to-LLVM
+// conversion patterns, this would fail with unrealized_conversion_cast errors.
+
+!s32i = !cir.int<s, 32>
+
+module {
+ // CHECK-LABEL: llvm.func @target_map_from
+ cir.func @target_map_from(%arg0 : !s32i) {
+ %0 = cir.alloca !s32i, !cir.ptr<!s32i>, ["x", init] {alignment = 4 : i64}
+ cir.store %arg0, %0 : !s32i, !cir.ptr<!s32i>
+
+ // CHECK: %[[ALLOCA:.*]] = llvm.alloca {{.*}} x i32
+ // CHECK: %[[MAP:.*]] = omp.map.info var_ptr(%[[ALLOCA]] : !llvm.ptr, i32) map_clauses(from) capture(ByRef) -> !llvm.ptr {name = "x"}
+ %1 = omp.map.info var_ptr(%0 : !cir.ptr<!s32i>, !s32i) map_clauses(from) capture(ByRef) -> !cir.ptr<!s32i> {name = "x"}
+
+ // CHECK: omp.target map_entries(%[[MAP]] -> %[[ARG:.*]] : !llvm.ptr)
+ omp.target map_entries(%1 -> %arg1 : !cir.ptr<!s32i>) {
+ // CHECK: %[[C10:.*]] = llvm.mlir.constant(10 : i32) : i32
+ // CHECK: llvm.store %[[C10]], %[[ARG]]
+ %c10 = cir.const #cir.int<10> : !s32i
+ cir.store %c10, %arg1 : !s32i, !cir.ptr<!s32i>
+ // CHECK: omp.terminator
+ omp.terminator
+ }
+ cir.return
+ }
+}
|
andykaylor
approved these changes
Apr 2, 2026
Contributor
andykaylor
left a comment
andykaylor
left a comment
There was a problem hiding this comment.
This looks good.
I don't see any equivalent code in the incubator. Was OpenMP just not working there?
Contributor
Author
I know that Luca Parigi was working on workshare loops here in PR #181841, and he was seeing the same issues, so likely this never worked in the incubator. |
jsjodin
force-pushed
the
jleyonberg/cir-omp-type-conversion
branch
from
7f12f72 to
f569f3c
Compare
…ring Register OpenMP conversion legality and patterns in the CIR-to-LLVM pass so that OpenMP operations (e.g. omp.map.info, omp.target) have their CIR types properly converted to LLVM types during lowering. Without this, the conversion leaves behind unrealized_conversion_cast ops that cause translation to LLVM IR to fail. Also registers omp::PointerLikeType on cir::PointerType so that CIR pointers are accepted as operands in OpenMP map operations. Assised-by: Cursor / Claude Opus 4.6
jsjodin
force-pushed
the
jleyonberg/cir-omp-type-conversion
branch
from
f569f3c to
bcad090
Compare
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/88/builds/22463 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/128/builds/12416 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/66/builds/29442 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/169/builds/21929 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/20/builds/15816 Here is the relevant piece of the build log for the reference |
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
4 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.
Register OpenMP conversion legality and patterns in the CIR-to-LLVM pass so that OpenMP operations (e.g. omp.map.info, omp.target) have their CIR types properly converted to LLVM types during lowering. Without this, the conversion leaves behind unrealized_conversion_cast ops that cause translation to LLVM IR to fail.
Also registers omp::PointerLikeType on cir::PointerType so that CIR pointers are accepted as operands in OpenMP map operations.
Assised-by: Cursor / Claude Opus 4.6