[clang] [CIR] Support void calls in cir.call_llvm_intrinsic lowering (PR #199516)

[via cfe-commits]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-clangir

Author: Chaitanya (skc7)

<details>
<summary>Changes</summary>

`LowerToLLVM` module now skips type conversion when the op has zero results and selects the void or single-result `CallIntrinsicOp::create` overload to support void return type builtins lowering.

---
Full diff: https://github.com/llvm/llvm-project/pull/199516.diff


1 Files Affected:

- (modified) clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp (+14-6) 


``````````diff
diff --git a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
index c4e98e299dfc1..8940272d27ffe 100644
--- a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
+++ b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
@@ -358,8 +358,12 @@ createCallLLVMIntrinsicOp(mlir::ConversionPatternRewriter &rewriter,
                           mlir::Type resultTy, mlir::ValueRange operands) {
   auto intrinsicNameAttr =
       mlir::StringAttr::get(rewriter.getContext(), intrinsicName);
-  return mlir::LLVM::CallIntrinsicOp::create(rewriter, loc, resultTy,
-                                             intrinsicNameAttr, operands);
+  // CallIntrinsicOp has distinct void / single-result create overloads.
+  if (resultTy)
+    return mlir::LLVM::CallIntrinsicOp::create(rewriter, loc, resultTy,
+                                               intrinsicNameAttr, operands);
+  return mlir::LLVM::CallIntrinsicOp::create(rewriter, loc, intrinsicNameAttr,
+                                             operands);
 }
 
 static mlir::LLVM::CallIntrinsicOp replaceOpWithCallLLVMIntrinsicOp(
@@ -375,10 +379,14 @@ static mlir::LLVM::CallIntrinsicOp replaceOpWithCallLLVMIntrinsicOp(
 mlir::LogicalResult CIRToLLVMLLVMIntrinsicCallOpLowering::matchAndRewrite(
     cir::LLVMIntrinsicCallOp op, OpAdaptor adaptor,
     mlir::ConversionPatternRewriter &rewriter) const {
-  mlir::Type llvmResTy =
-      getTypeConverter()->convertType(op->getResultTypes()[0]);
-  if (!llvmResTy)
-    return op.emitError("expected LLVM result type");
+  // Result is Optional on the op, so void intrinsics have zero
+  // results; leave llvmResTy null in that case.
+  mlir::Type llvmResTy;
+  if (op->getNumResults() != 0) {
+    llvmResTy = getTypeConverter()->convertType(op->getResultTypes()[0]);
+    if (!llvmResTy)
+      return op.emitError("expected LLVM result type");
+  }
   StringRef name = op.getIntrinsicName();
 
   // Some LLVM intrinsics require ElementType attribute to be attached to

``````````

</details>


https://github.com/llvm/llvm-project/pull/199516