[clang] [CIR] Upstream splat op for VectorType (PR #139827)
Amr Hesham via cfe-commits
cfe-commits at lists.llvm.org
Fri May 30 14:40:29 PDT 2025
https://github.com/AmrDeveloper updated https://github.com/llvm/llvm-project/pull/139827
>From eb000ef792f14fa0b59dddce1470ff98409dbe0e Mon Sep 17 00:00:00 2001
From: AmrDeveloper <amr96 at programmer.net>
Date: Tue, 13 May 2025 21:35:06 +0200
Subject: [PATCH 1/2] [CIR] Upstream splat op for VectorType
---
clang/include/clang/CIR/Dialect/IR/CIROps.td | 32 +++++++++
clang/lib/CIR/CodeGen/CIRGenExprScalar.cpp | 8 +++
.../CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp | 56 ++++++++++++++++
.../CIR/Lowering/DirectToLLVM/LowerToLLVM.h | 10 +++
clang/test/CIR/CodeGen/vector-ext.cpp | 63 ++++++++++++++++++
clang/test/CIR/CodeGen/vector.cpp | 65 ++++++++++++++++++-
clang/test/CIR/IR/vector.cir | 34 ++++++++++
7 files changed, 267 insertions(+), 1 deletion(-)
diff --git a/clang/include/clang/CIR/Dialect/IR/CIROps.td b/clang/include/clang/CIR/Dialect/IR/CIROps.td
index 07851610a2abd..40a4ebdd5167e 100644
--- a/clang/include/clang/CIR/Dialect/IR/CIROps.td
+++ b/clang/include/clang/CIR/Dialect/IR/CIROps.td
@@ -2190,4 +2190,36 @@ def VecShuffleDynamicOp : CIR_Op<"vec.shuffle.dynamic",
let hasVerifier = 1;
}
+//===----------------------------------------------------------------------===//
+// VecSplatOp
+//===----------------------------------------------------------------------===//
+
+def VecSplatOp : CIR_Op<"vec.splat", [Pure,
+ TypesMatchWith<"type of 'value' matches element type of 'result'", "result",
+ "value", "cast<VectorType>($_self).getElementType()">]> {
+
+ let summary = "Convert a scalar into a vector";
+ let description = [{
+ The `cir.vec.splat` operation creates a vector value from a scalar value.
+ All elements of the vector have the same value, that of the given scalar.
+
+ It's a separate operation from `cir.vec.create` because more
+ efficient LLVM IR can be generated for it, and because some optimization and
+ analysis passes can benefit from knowing that all elements of the vector
+ have the same value.
+
+ ```mlir
+ %value = cir.const #cir.int<3> : !s32i
+ %value_vec = cir.vec.splat %value : !s32i, !cir.vector<4 x !s32i>
+ ```
+ }];
+
+ let arguments = (ins CIR_AnyType:$value);
+ let results = (outs CIR_VectorType:$result);
+
+ let assemblyFormat = [{
+ $value `:` type($value) `,` qualified(type($result)) attr-dict
+ }];
+}
+
#endif // CLANG_CIR_DIALECT_IR_CIROPS_TD
diff --git a/clang/lib/CIR/CodeGen/CIRGenExprScalar.cpp b/clang/lib/CIR/CodeGen/CIRGenExprScalar.cpp
index 8448c164a5e58..dfe0e6e563f09 100644
--- a/clang/lib/CIR/CodeGen/CIRGenExprScalar.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenExprScalar.cpp
@@ -1685,6 +1685,14 @@ mlir::Value ScalarExprEmitter::VisitCastExpr(CastExpr *ce) {
cgf.convertType(destTy));
}
+ case CK_VectorSplat: {
+ // Create a vector object and fill all elements with the same scalar value.
+ assert(destTy->isVectorType() && "CK_VectorSplat to non-vector type");
+ return builder.create<cir::VecSplatOp>(
+ cgf.getLoc(subExpr->getSourceRange()), cgf.convertType(destTy),
+ Visit(subExpr));
+ }
+
default:
cgf.getCIRGenModule().errorNYI(subExpr->getSourceRange(),
"CastExpr: ", ce->getCastKindName());
diff --git a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
index b07e61638c3b4..3da94d47183d6 100644
--- a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
+++ b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
@@ -1708,6 +1708,7 @@ void ConvertCIRToLLVMPass::runOnOperation() {
CIRToLLVMVecExtractOpLowering,
CIRToLLVMVecInsertOpLowering,
CIRToLLVMVecCmpOpLowering,
+ CIRToLLVMVecSplatOpLowering,
CIRToLLVMVecShuffleDynamicOpLowering
// clang-format on
>(converter, patterns.getContext());
@@ -1862,6 +1863,61 @@ mlir::LogicalResult CIRToLLVMVecCmpOpLowering::matchAndRewrite(
return mlir::success();
}
+mlir::LogicalResult CIRToLLVMVecSplatOpLowering::matchAndRewrite(
+ cir::VecSplatOp op, OpAdaptor adaptor,
+ mlir::ConversionPatternRewriter &rewriter) const {
+ // Vector splat can be implemented with an `insertelement` and a
+ // `shufflevector`, which is better than an `insertelement` for each
+ // element in the vector. Start with an undef vector. Insert the value into
+ // the first element. Then use a `shufflevector` with a mask of all 0 to
+ // fill out the entire vector with that value.
+ const auto vecTy = mlir::cast<cir::VectorType>(op.getType());
+ const mlir::Type llvmTy = typeConverter->convertType(vecTy);
+ const mlir::Location loc = op.getLoc();
+ const mlir::Value poison = rewriter.create<mlir::LLVM::PoisonOp>(loc, llvmTy);
+
+ const mlir::Value elementValue = adaptor.getValue();
+ if (mlir::isa<mlir::LLVM::PoisonOp>(elementValue.getDefiningOp())) {
+ // If the splat value is poison, then we can just use poison value
+ // for the entire vector.
+ rewriter.replaceOp(op, poison);
+ return mlir::success();
+ }
+
+ if (auto constValue =
+ dyn_cast<mlir::LLVM::ConstantOp>(elementValue.getDefiningOp())) {
+ if (auto intAttr = dyn_cast<mlir::IntegerAttr>(constValue.getValue())) {
+ mlir::DenseIntElementsAttr denseVec = mlir::DenseIntElementsAttr::get(
+ mlir::cast<mlir::ShapedType>(llvmTy), intAttr.getValue());
+
+ const mlir::Value indexValue = rewriter.create<mlir::LLVM::ConstantOp>(
+ loc, denseVec.getType(), denseVec);
+ rewriter.replaceOp(op, indexValue);
+ return mlir::success();
+ }
+
+ if (auto fpAttr = dyn_cast<mlir::FloatAttr>(constValue.getValue())) {
+ mlir::DenseFPElementsAttr denseVec = mlir::DenseFPElementsAttr::get(
+ mlir::cast<mlir::ShapedType>(llvmTy), fpAttr.getValue());
+
+ const mlir::Value indexValue = rewriter.create<mlir::LLVM::ConstantOp>(
+ loc, denseVec.getType(), denseVec);
+ rewriter.replaceOp(op, indexValue);
+ return mlir::success();
+ }
+ }
+
+ const mlir::Value indexValue =
+ rewriter.create<mlir::LLVM::ConstantOp>(loc, rewriter.getI64Type(), 0);
+ const mlir::Value oneElement = rewriter.create<mlir::LLVM::InsertElementOp>(
+ loc, poison, elementValue, indexValue);
+ const SmallVector<int32_t> zeroValues(vecTy.getSize(), 0);
+ const mlir::Value shuffled = rewriter.create<mlir::LLVM::ShuffleVectorOp>(
+ loc, oneElement, poison, zeroValues);
+ rewriter.replaceOp(op, shuffled);
+ return mlir::success();
+}
+
mlir::LogicalResult CIRToLLVMVecShuffleDynamicOpLowering::matchAndRewrite(
cir::VecShuffleDynamicOp op, OpAdaptor adaptor,
mlir::ConversionPatternRewriter &rewriter) const {
diff --git a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.h b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.h
index 6b8862db2c8be..b2c8a9fc7eb49 100644
--- a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.h
+++ b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.h
@@ -352,6 +352,16 @@ class CIRToLLVMVecCmpOpLowering
mlir::ConversionPatternRewriter &) const override;
};
+class CIRToLLVMVecSplatOpLowering
+ : public mlir::OpConversionPattern<cir::VecSplatOp> {
+public:
+ using mlir::OpConversionPattern<cir::VecSplatOp>::OpConversionPattern;
+
+ mlir::LogicalResult
+ matchAndRewrite(cir::VecSplatOp op, OpAdaptor,
+ mlir::ConversionPatternRewriter &) const override;
+};
+
class CIRToLLVMVecShuffleDynamicOpLowering
: public mlir::OpConversionPattern<cir::VecShuffleDynamicOp> {
public:
diff --git a/clang/test/CIR/CodeGen/vector-ext.cpp b/clang/test/CIR/CodeGen/vector-ext.cpp
index 8a0479fc1d088..f434517165408 100644
--- a/clang/test/CIR/CodeGen/vector-ext.cpp
+++ b/clang/test/CIR/CodeGen/vector-ext.cpp
@@ -1091,3 +1091,66 @@ void foo17() {
// OGCG: %[[VEC_A:.*]] = alloca <2 x double>, align 16
// OGCG: %[[TMP:.*]] = load <2 x double>, ptr %[[VEC_A]], align 16
// OGCG: %[[RES:.*]]= fptoui <2 x double> %[[TMP]] to <2 x i16>
+
+void foo18() {
+ vi4 a = {1, 2, 3, 4};
+ vi4 shl = a << 3;
+
+ uvi4 b = {1u, 2u, 3u, 4u};
+ uvi4 shr = b >> 3u;
+}
+
+// CIR: %[[VEC_A:.*]] = cir.alloca !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>, ["a", init]
+// CIR: %[[SHL_RES:.*]] = cir.alloca !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>, ["shl", init]
+// CIR: %[[VEC_B:.*]] = cir.alloca !cir.vector<4 x !u32i>, !cir.ptr<!cir.vector<4 x !u32i>>, ["b", init]
+// CIR: %[[SHR_RES:.*]] = cir.alloca !cir.vector<4 x !u32i>, !cir.ptr<!cir.vector<4 x !u32i>>, ["shr", init]
+// CIR: %[[CONST_1:.*]] = cir.const #cir.int<1> : !s32i
+// CIR: %[[CONST_2:.*]] = cir.const #cir.int<2> : !s32i
+// CIR: %[[CONST_3:.*]] = cir.const #cir.int<3> : !s32i
+// CIR: %[[CONST_4:.*]] = cir.const #cir.int<4> : !s32i
+// CIR: %[[VEC_A_VAL:.*]] = cir.vec.create(%[[CONST_1]], %[[CONST_2]], %[[CONST_3]], %[[CONST_4]] :
+// CIR-SAME: !s32i, !s32i, !s32i, !s32i) : !cir.vector<4 x !s32i>
+// CIR: cir.store{{.*}} %[[VEC_A_VAL]], %[[VEC_A]] : !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>
+// CIR: %[[TMP_A:.*]] = cir.load{{.*}} %[[VEC_A]] : !cir.ptr<!cir.vector<4 x !s32i>>, !cir.vector<4 x !s32i>
+// CIR: %[[SH_AMOUNT:.*]] = cir.const #cir.int<3> : !s32i
+// CIR: %[[SPLAT_VEC:.*]] = cir.vec.splat %[[SH_AMOUNT]] : !s32i, !cir.vector<4 x !s32i>
+// CIR: %[[SHL:.*]] = cir.shift(left, %[[TMP_A]] : !cir.vector<4 x !s32i>, %[[SPLAT_VEC]] : !cir.vector<4 x !s32i>) -> !cir.vector<4 x !s32i>
+// CIR: cir.store{{.*}} %[[SHL]], %[[SHL_RES]] : !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>
+// CIR: %[[CONST_1:.*]] = cir.const #cir.int<1> : !u32i
+// CIR: %[[CONST_2:.*]] = cir.const #cir.int<2> : !u32i
+// CIR: %[[CONST_3:.*]] = cir.const #cir.int<3> : !u32i
+// CIR: %[[CONST_4:.*]] = cir.const #cir.int<4> : !u32i
+// CIR: %[[VEC_B_VAL:.*]] = cir.vec.create(%[[CONST_1]], %[[CONST_2]], %[[CONST_3]], %[[CONST_4]] :
+// CIR-SAME: !u32i, !u32i, !u32i, !u32i) : !cir.vector<4 x !u32i>
+// CIR: cir.store{{.*}} %[[VEC_B_VAL]], %[[VEC_B]] : !cir.vector<4 x !u32i>, !cir.ptr<!cir.vector<4 x !u32i>>
+// CIR: %[[TMP_B:.*]] = cir.load{{.*}} %[[VEC_B]] : !cir.ptr<!cir.vector<4 x !u32i>>, !cir.vector<4 x !u32i>
+// CIR: %[[SH_AMOUNT:.*]] = cir.const #cir.int<3> : !u32i
+// CIR: %[[SPLAT_VEC:.*]] = cir.vec.splat %[[SH_AMOUNT]] : !u32i, !cir.vector<4 x !u32i>
+// CIR: %[[SHR:.*]] = cir.shift(right, %[[TMP_B]] : !cir.vector<4 x !u32i>, %[[SPLAT_VEC]] : !cir.vector<4 x !u32i>) -> !cir.vector<4 x !u32i>
+// CIR: cir.store{{.*}} %[[SHR]], %[[SHR_RES]] : !cir.vector<4 x !u32i>, !cir.ptr<!cir.vector<4 x !u32i>>
+
+// LLVM: %[[VEC_A:.*]] = alloca <4 x i32>, i64 1, align 16
+// LLVM: %[[SHL_RES:.*]] = alloca <4 x i32>, i64 1, align 16
+// LLVM: %[[VEC_B:.*]] = alloca <4 x i32>, i64 1, align 16
+// LLVM: %[[SHR_RES:.*]] = alloca <4 x i32>, i64 1, align 16
+// LLVM: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_A]], align 16
+// LLVM: %[[TMP_A:.*]] = load <4 x i32>, ptr %[[VEC_A]], align 16
+// LLVM: %[[SHL:.*]] = shl <4 x i32> %[[TMP_A]], splat (i32 3)
+// LLVM: store <4 x i32> %[[SHL]], ptr %[[SHL_RES]], align 16
+// LLVM: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_B]], align 16
+// LLVM: %[[TMP_B:.*]] = load <4 x i32>, ptr %[[VEC_B]], align 16
+// LLVM: %[[SHR:.*]] = lshr <4 x i32> %[[TMP_B]], splat (i32 3)
+// LLVM: store <4 x i32> %[[SHR]], ptr %[[SHR_RES]], align 16
+
+// OGCG: %[[VEC_A:.*]] = alloca <4 x i32>, align 16
+// OGCG: %[[SHL_RES:.*]] = alloca <4 x i32>, align 16
+// OGCG: %[[VEC_B:.*]] = alloca <4 x i32>, align 16
+// OGCG: %[[SHR_RES:.*]] = alloca <4 x i32>, align 16
+// OGCG: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_A]], align 16
+// OGCG: %[[TMP_A:.*]] = load <4 x i32>, ptr %[[VEC_A]], align 16
+// OGCG: %[[SHL:.*]] = shl <4 x i32> %[[TMP_A]], splat (i32 3)
+// OGCG: store <4 x i32> %[[SHL]], ptr %[[SHL_RES]], align 16
+// OGCG: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_B]], align 16
+// OGCG: %[[TMP_B:.*]] = load <4 x i32>, ptr %[[VEC_B]], align 16
+// OGCG: %[[SHR:.*]] = lshr <4 x i32> %[[TMP_B]], splat (i32 3)
+// OGCG: store <4 x i32> %[[SHR]], ptr %[[SHR_RES]], align 16
\ No newline at end of file
diff --git a/clang/test/CIR/CodeGen/vector.cpp b/clang/test/CIR/CodeGen/vector.cpp
index 4c50f68a56162..b5ceae68e7994 100644
--- a/clang/test/CIR/CodeGen/vector.cpp
+++ b/clang/test/CIR/CodeGen/vector.cpp
@@ -1069,4 +1069,67 @@ void foo17() {
// OGCG: %[[VEC_A:.*]] = alloca <2 x double>, align 16
// OGCG: %[[TMP:.*]] = load <2 x double>, ptr %[[VEC_A]], align 16
-// OGCG: %[[RES:.*]]= fptoui <2 x double> %[[TMP]] to <2 x i16>
\ No newline at end of file
+// OGCG: %[[RES:.*]]= fptoui <2 x double> %[[TMP]] to <2 x i16>
+
+void foo18() {
+ vi4 a = {1, 2, 3, 4};
+ vi4 shl = a << 3;
+
+ uvi4 b = {1u, 2u, 3u, 4u};
+ uvi4 shr = b >> 3u;
+}
+
+// CIR: %[[VEC_A:.*]] = cir.alloca !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>, ["a", init]
+// CIR: %[[SHL_RES:.*]] = cir.alloca !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>, ["shl", init]
+// CIR: %[[VEC_B:.*]] = cir.alloca !cir.vector<4 x !u32i>, !cir.ptr<!cir.vector<4 x !u32i>>, ["b", init]
+// CIR: %[[SHR_RES:.*]] = cir.alloca !cir.vector<4 x !u32i>, !cir.ptr<!cir.vector<4 x !u32i>>, ["shr", init]
+// CIR: %[[CONST_1:.*]] = cir.const #cir.int<1> : !s32i
+// CIR: %[[CONST_2:.*]] = cir.const #cir.int<2> : !s32i
+// CIR: %[[CONST_3:.*]] = cir.const #cir.int<3> : !s32i
+// CIR: %[[CONST_4:.*]] = cir.const #cir.int<4> : !s32i
+// CIR: %[[VEC_A_VAL:.*]] = cir.vec.create(%[[CONST_1]], %[[CONST_2]], %[[CONST_3]], %[[CONST_4]] :
+// CIR-SAME: !s32i, !s32i, !s32i, !s32i) : !cir.vector<4 x !s32i>
+// CIR: cir.store{{.*}} %[[VEC_A_VAL]], %[[VEC_A]] : !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>
+// CIR: %[[TMP_A:.*]] = cir.load{{.*}} %[[VEC_A]] : !cir.ptr<!cir.vector<4 x !s32i>>, !cir.vector<4 x !s32i>
+// CIR: %[[SH_AMOUNT:.*]] = cir.const #cir.int<3> : !s32i
+// CIR: %[[SPLAT_VEC:.*]] = cir.vec.splat %[[SH_AMOUNT]] : !s32i, !cir.vector<4 x !s32i>
+// CIR: %[[SHL:.*]] = cir.shift(left, %[[TMP_A]] : !cir.vector<4 x !s32i>, %[[SPLAT_VEC]] : !cir.vector<4 x !s32i>) -> !cir.vector<4 x !s32i>
+// CIR: cir.store{{.*}} %[[SHL]], %[[SHL_RES]] : !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>
+// CIR: %[[CONST_1:.*]] = cir.const #cir.int<1> : !u32i
+// CIR: %[[CONST_2:.*]] = cir.const #cir.int<2> : !u32i
+// CIR: %[[CONST_3:.*]] = cir.const #cir.int<3> : !u32i
+// CIR: %[[CONST_4:.*]] = cir.const #cir.int<4> : !u32i
+// CIR: %[[VEC_B_VAL:.*]] = cir.vec.create(%[[CONST_1]], %[[CONST_2]], %[[CONST_3]], %[[CONST_4]] :
+// CIR-SAME: !u32i, !u32i, !u32i, !u32i) : !cir.vector<4 x !u32i>
+// CIR: cir.store{{.*}} %[[VEC_B_VAL]], %[[VEC_B]] : !cir.vector<4 x !u32i>, !cir.ptr<!cir.vector<4 x !u32i>>
+// CIR: %[[TMP_B:.*]] = cir.load{{.*}} %[[VEC_B]] : !cir.ptr<!cir.vector<4 x !u32i>>, !cir.vector<4 x !u32i>
+// CIR: %[[SH_AMOUNT:.*]] = cir.const #cir.int<3> : !u32i
+// CIR: %[[SPLAT_VEC:.*]] = cir.vec.splat %[[SH_AMOUNT]] : !u32i, !cir.vector<4 x !u32i>
+// CIR: %[[SHR:.*]] = cir.shift(right, %[[TMP_B]] : !cir.vector<4 x !u32i>, %[[SPLAT_VEC]] : !cir.vector<4 x !u32i>) -> !cir.vector<4 x !u32i>
+// CIR: cir.store{{.*}} %[[SHR]], %[[SHR_RES]] : !cir.vector<4 x !u32i>, !cir.ptr<!cir.vector<4 x !u32i>>
+
+// LLVM: %[[VEC_A:.*]] = alloca <4 x i32>, i64 1, align 16
+// LLVM: %[[SHL_RES:.*]] = alloca <4 x i32>, i64 1, align 16
+// LLVM: %[[VEC_B:.*]] = alloca <4 x i32>, i64 1, align 16
+// LLVM: %[[SHR_RES:.*]] = alloca <4 x i32>, i64 1, align 16
+// LLVM: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_A]], align 16
+// LLVM: %[[TMP_A:.*]] = load <4 x i32>, ptr %[[VEC_A]], align 16
+// LLVM: %[[SHL:.*]] = shl <4 x i32> %[[TMP_A]], splat (i32 3)
+// LLVM: store <4 x i32> %[[SHL]], ptr %[[SHL_RES]], align 16
+// LLVM: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_B]], align 16
+// LLVM: %[[TMP_B:.*]] = load <4 x i32>, ptr %[[VEC_B]], align 16
+// LLVM: %[[SHR:.*]] = lshr <4 x i32> %[[TMP_B]], splat (i32 3)
+// LLVM: store <4 x i32> %[[SHR]], ptr %[[SHR_RES]], align 16
+
+// OGCG: %[[VEC_A:.*]] = alloca <4 x i32>, align 16
+// OGCG: %[[SHL_RES:.*]] = alloca <4 x i32>, align 16
+// OGCG: %[[VEC_B:.*]] = alloca <4 x i32>, align 16
+// OGCG: %[[SHR_RES:.*]] = alloca <4 x i32>, align 16
+// OGCG: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_A]], align 16
+// OGCG: %[[TMP_A:.*]] = load <4 x i32>, ptr %[[VEC_A]], align 16
+// OGCG: %[[SHL:.*]] = shl <4 x i32> %[[TMP_A]], splat (i32 3)
+// OGCG: store <4 x i32> %[[SHL]], ptr %[[SHL_RES]], align 16
+// OGCG: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_B]], align 16
+// OGCG: %[[TMP_B:.*]] = load <4 x i32>, ptr %[[VEC_B]], align 16
+// OGCG: %[[SHR:.*]] = lshr <4 x i32> %[[TMP_B]], splat (i32 3)
+// OGCG: store <4 x i32> %[[SHR]], ptr %[[SHR_RES]], align 16
\ No newline at end of file
diff --git a/clang/test/CIR/IR/vector.cir b/clang/test/CIR/IR/vector.cir
index a455acf92ab6f..74b5d48588f36 100644
--- a/clang/test/CIR/IR/vector.cir
+++ b/clang/test/CIR/IR/vector.cir
@@ -187,4 +187,38 @@ cir.func @vector_shuffle_dynamic_test() {
// CHECK: cir.return
// CHECK: }
+cir.func @vector_splat_test() {
+ %0 = cir.alloca !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>, ["a", init]
+ %1 = cir.alloca !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>, ["shl", init]
+ %2 = cir.const #cir.int<1> : !s32i
+ %3 = cir.const #cir.int<2> : !s32i
+ %4 = cir.const #cir.int<3> : !s32i
+ %5 = cir.const #cir.int<4> : !s32i
+ %6 = cir.vec.create(%2, %3, %4, %5 : !s32i, !s32i, !s32i, !s32i) : !cir.vector<4 x !s32i>
+ cir.store %6, %0 : !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>
+ %7 = cir.load %0 : !cir.ptr<!cir.vector<4 x !s32i>>, !cir.vector<4 x !s32i>
+ %8 = cir.const #cir.int<3> : !s32i
+ %9 = cir.vec.splat %8 : !s32i, !cir.vector<4 x !s32i>
+ %10 = cir.shift(left, %7 : !cir.vector<4 x !s32i>, %9 : !cir.vector<4 x !s32i>) -> !cir.vector<4 x !s32i>
+ cir.store %10, %1 : !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>
+ cir.return
+}
+
+// CHECK: cir.func @vector_splat_test() {
+// CHECK: %0 = cir.alloca !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>, ["a", init]
+// CHECK: %1 = cir.alloca !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>, ["shl", init]
+// CHECK: %2 = cir.const #cir.int<1> : !s32i
+// CHECK: %3 = cir.const #cir.int<2> : !s32i
+// CHECK: %4 = cir.const #cir.int<3> : !s32i
+// CHECK: %5 = cir.const #cir.int<4> : !s32i
+// CHECK: %6 = cir.vec.create(%2, %3, %4, %5 : !s32i, !s32i, !s32i, !s32i) : !cir.vector<4 x !s32i>
+// CHECK: cir.store %6, %0 : !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>
+// CHECK: %7 = cir.load %0 : !cir.ptr<!cir.vector<4 x !s32i>>, !cir.vector<4 x !s32i>
+// CHECK: %8 = cir.const #cir.int<3> : !s32i
+// CHECK: %9 = cir.vec.splat %8 : !s32i, !cir.vector<4 x !s32i>
+// CHECK: %10 = cir.shift(left, %7 : !cir.vector<4 x !s32i>, %9 : !cir.vector<4 x !s32i>) -> !cir.vector<4 x !s32i>
+// CHECK: cir.store %10, %1 : !cir.vector<4 x !s32i>, !cir.ptr<!cir.vector<4 x !s32i>>
+// CHECK: cir.return
+// CHECK: }
+
}
>From 402ff5ab0610add5947131a56f9b878cada11302 Mon Sep 17 00:00:00 2001
From: AmrDeveloper <amr96 at programmer.net>
Date: Thu, 29 May 2025 18:05:18 +0200
Subject: [PATCH 2/2] Update test files for load and store with allign
---
clang/test/CIR/CodeGen/vector-ext.cpp | 3 ++-
clang/test/CIR/CodeGen/vector.cpp | 2 +-
2 files changed, 3 insertions(+), 2 deletions(-)
diff --git a/clang/test/CIR/CodeGen/vector-ext.cpp b/clang/test/CIR/CodeGen/vector-ext.cpp
index f434517165408..28d3dc0f42fca 100644
--- a/clang/test/CIR/CodeGen/vector-ext.cpp
+++ b/clang/test/CIR/CodeGen/vector-ext.cpp
@@ -990,6 +990,7 @@ void foo14() {
// OGCG: %[[TMP_B:.*]] = load <4 x float>, ptr %[[VEC_B]], align 16
// OGCG: %[[GE:.*]] = fcmp oge <4 x float> %[[TMP_A]], %[[TMP_B]]
// OGCG: %[[RES:.*]] = sext <4 x i1> %[[GE]] to <4 x i32>
+// OGCG: store <4 x i32> %[[RES]], ptr {{.*}}, align 16
void foo15() {
vi4 a;
@@ -1153,4 +1154,4 @@ void foo18() {
// OGCG: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_B]], align 16
// OGCG: %[[TMP_B:.*]] = load <4 x i32>, ptr %[[VEC_B]], align 16
// OGCG: %[[SHR:.*]] = lshr <4 x i32> %[[TMP_B]], splat (i32 3)
-// OGCG: store <4 x i32> %[[SHR]], ptr %[[SHR_RES]], align 16
\ No newline at end of file
+// OGCG: store <4 x i32> %[[SHR]], ptr %[[SHR_RES]], align 16
diff --git a/clang/test/CIR/CodeGen/vector.cpp b/clang/test/CIR/CodeGen/vector.cpp
index b5ceae68e7994..0bee7689e3d53 100644
--- a/clang/test/CIR/CodeGen/vector.cpp
+++ b/clang/test/CIR/CodeGen/vector.cpp
@@ -1132,4 +1132,4 @@ void foo18() {
// OGCG: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr %[[VEC_B]], align 16
// OGCG: %[[TMP_B:.*]] = load <4 x i32>, ptr %[[VEC_B]], align 16
// OGCG: %[[SHR:.*]] = lshr <4 x i32> %[[TMP_B]], splat (i32 3)
-// OGCG: store <4 x i32> %[[SHR]], ptr %[[SHR_RES]], align 16
\ No newline at end of file
+// OGCG: store <4 x i32> %[[SHR]], ptr %[[SHR_RES]], align 16
More information about the cfe-commits
mailing list