Merge branch 'main' into cindex_inline

TApplencourt · web-flow · commit 91b9d79fb903 · 2025-10-10T11:42:06.000-05:00
diff --git a/clang-tools-extra/test/clang-tidy/checkers/cppcoreguidelines/pro-bounds-constant-array-index-c++03.cpp b/clang-tools-extra/test/clang-tidy/checkers/cppcoreguidelines/pro-bounds-constant-array-index-c++03.cpp
@@ -1,7 +1,5 @@
 // RUN: %check_clang_tidy -std=c++98-or-later %s cppcoreguidelines-pro-bounds-constant-array-index %t
 
-// Note: this test expects no diagnostics, but FileCheck cannot handle that,
-// hence the use of | count 0.
 template <int index> struct B {
   int get() {
     // The next line used to crash the check (in C++03 mode only).
diff --git a/libcxx/utils/ci/buildkite-pipeline.yml b/libcxx/utils/ci/buildkite-pipeline.yml
@@ -103,7 +103,7 @@ steps:
       queue: libcxx-builders
       os: aix
     <<: *common
-    skip: "Until https://github.com/llvm/llvm-project/issues/162516 has been resolved"
+    skip: "https://github.com/llvm/llvm-project/issues/162516"
 
   - label: AIX (64-bit)
     command: libcxx/utils/ci/run-buildbot aix
@@ -115,7 +115,7 @@ steps:
       queue: libcxx-builders
       os: aix
     <<: *common
-    skip: "Until https://github.com/llvm/llvm-project/issues/162516 has been resolved"
+    skip: "https://github.com/llvm/llvm-project/issues/162516"
 
 - group: ':freebsd: FreeBSD'
   steps:
diff --git a/llvm/include/llvm/IR/Instructions.h b/llvm/include/llvm/IR/Instructions.h
@@ -1715,7 +1715,7 @@ class SelectInst : public Instruction {
   static SelectInst *Create(Value *C, Value *S1, Value *S2,
                             const Twine &NameStr = "",
                             InsertPosition InsertBefore = nullptr,
-                            Instruction *MDFrom = nullptr) {
+                            const Instruction *MDFrom = nullptr) {
     SelectInst *Sel =
         new (AllocMarker) SelectInst(C, S1, S2, NameStr, InsertBefore);
     if (MDFrom)
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
@@ -880,11 +880,13 @@ Instruction *InstCombinerImpl::foldAddWithConstant(BinaryOperator &Add) {
   // zext(bool) + C -> bool ? C + 1 : C
   if (match(Op0, m_ZExt(m_Value(X))) &&
       X->getType()->getScalarSizeInBits() == 1)
-    return createSelectInst(X, InstCombiner::AddOne(Op1C), Op1);
+    return createSelectInstWithUnknownProfile(X, InstCombiner::AddOne(Op1C),
+                                              Op1);
   // sext(bool) + C -> bool ? C - 1 : C
   if (match(Op0, m_SExt(m_Value(X))) &&
       X->getType()->getScalarSizeInBits() == 1)
-    return createSelectInst(X, InstCombiner::SubOne(Op1C), Op1);
+    return createSelectInstWithUnknownProfile(X, InstCombiner::SubOne(Op1C),
+                                              Op1);
 
   // ~X + C --> (C-1) - X
   if (match(Op0, m_Not(m_Value(X)))) {
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h
@@ -471,15 +471,16 @@ class LLVM_LIBRARY_VISIBILITY InstCombinerImpl final
   Value *simplifyNonNullOperand(Value *V, bool HasDereferenceable,
                                 unsigned Depth = 0);
 
-  SelectInst *createSelectInst(Value *C, Value *S1, Value *S2,
-                               const Twine &NameStr = "",
-                               InsertPosition InsertBefore = nullptr,
-                               Instruction *MDFrom = nullptr) {
-    SelectInst *SI =
-        SelectInst::Create(C, S1, S2, NameStr, InsertBefore, MDFrom);
-    if (!MDFrom)
-      setExplicitlyUnknownBranchWeightsIfProfiled(*SI, F, DEBUG_TYPE);
-    return SI;
+  /// Create `select C, S1, S2`. Use only when the profile cannot be calculated
+  /// from existing profile metadata: if the Function has profiles, this will
+  /// set the profile of this select to "unknown".
+  SelectInst *
+  createSelectInstWithUnknownProfile(Value *C, Value *S1, Value *S2,
+                                     const Twine &NameStr = "",
+                                     InsertPosition InsertBefore = nullptr) {
+    auto *Sel = SelectInst::Create(C, S1, S2, NameStr, InsertBefore, nullptr);
+    setExplicitlyUnknownBranchWeightsIfProfiled(*Sel, F, DEBUG_TYPE);
+    return Sel;
   }
 
 public:
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp
@@ -1253,7 +1253,8 @@ Instruction *InstCombinerImpl::visitShl(BinaryOperator &I) {
     // shl (zext i1 X), C1 --> select (X, 1 << C1, 0)
     if (match(Op0, m_ZExt(m_Value(X))) && X->getType()->isIntOrIntVectorTy(1)) {
       auto *NewC = Builder.CreateShl(ConstantInt::get(Ty, 1), C1);
-      return createSelectInst(X, NewC, ConstantInt::getNullValue(Ty));
+      return createSelectInstWithUnknownProfile(X, NewC,
+                                                ConstantInt::getNullValue(Ty));
     }
   }
 
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -1737,7 +1737,7 @@ Instruction *InstCombinerImpl::foldBinopOfSextBoolToSelect(BinaryOperator &BO) {
   Constant *Zero = ConstantInt::getNullValue(BO.getType());
   Value *TVal = Builder.CreateBinOp(BO.getOpcode(), Ones, C);
   Value *FVal = Builder.CreateBinOp(BO.getOpcode(), Zero, C);
-  return createSelectInst(X, TVal, FVal);
+  return createSelectInstWithUnknownProfile(X, TVal, FVal);
 }
 
 static Value *simplifyOperationIntoSelectOperand(Instruction &I, SelectInst *SI,
diff --git a/mlir/include/mlir/Dialect/XeGPU/Transforms/Transforms.h b/mlir/include/mlir/Dialect/XeGPU/Transforms/Transforms.h
@@ -64,6 +64,10 @@ void populateXeGPUFoldAliasOpsPatterns(RewritePatternSet &patterns);
 
 /// Appends patterns for XeGPU SIMT distribution into `patterns`.
 void populateXeGPUSubgroupDistributePatterns(RewritePatternSet &patterns);
+/// Appends patterns for moving function body into gpu.warp_execute_on_lane0 op.
+void populateXeGPUMoveFuncBodyToWarpOpPatterns(RewritePatternSet &patterns);
+/// Appends patterns for XeGPU workgroup to subgroup distribution into
+/// `patterns`.
 void populateXeGPUWgToSgDistributePatterns(RewritePatternSet &patterns);
 
 /// Collect a set of patterns to unroll xegpu operations to a smaller shapes.
diff --git a/mlir/lib/Conversion/XeVMToLLVM/XeVMToLLVM.cpp b/mlir/lib/Conversion/XeVMToLLVM/XeVMToLLVM.cpp
@@ -714,6 +714,135 @@ class LLVMLoadStoreToOCLPattern : public OpConversionPattern<OpType> {
   }
 };
 
+//===----------------------------------------------------------------------===//
+// GPU index id operations
+//===----------------------------------------------------------------------===//
+/*
+// Launch Config ops
+//   dimidx - x, y, z - is fixed to i32
+//   return type is set by XeVM type converter
+// get_local_id
+xevm::WorkitemIdXOp;
+xevm::WorkitemIdYOp;
+xevm::WorkitemIdZOp;
+// get_local_size
+xevm::WorkgroupDimXOp;
+xevm::WorkgroupDimYOp;
+xevm::WorkgroupDimZOp;
+// get_group_id
+xevm::WorkgroupIdXOp;
+xevm::WorkgroupIdYOp;
+xevm::WorkgroupIdZOp;
+// get_num_groups
+xevm::GridDimXOp;
+xevm::GridDimYOp;
+xevm::GridDimZOp;
+// get_global_id : to be added if needed
+*/
+
+// Helpers to get the OpenCL function name and dimension argument for each op.
+static std::pair<StringRef, int64_t> getConfig(xevm::WorkitemIdXOp) {
+  return {"get_local_id", 0};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::WorkitemIdYOp) {
+  return {"get_local_id", 1};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::WorkitemIdZOp) {
+  return {"get_local_id", 2};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::WorkgroupDimXOp) {
+  return {"get_local_size", 0};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::WorkgroupDimYOp) {
+  return {"get_local_size", 1};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::WorkgroupDimZOp) {
+  return {"get_local_size", 2};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::WorkgroupIdXOp) {
+  return {"get_group_id", 0};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::WorkgroupIdYOp) {
+  return {"get_group_id", 1};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::WorkgroupIdZOp) {
+  return {"get_group_id", 2};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::GridDimXOp) {
+  return {"get_num_groups", 0};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::GridDimYOp) {
+  return {"get_num_groups", 1};
+}
+static std::pair<StringRef, int64_t> getConfig(xevm::GridDimZOp) {
+  return {"get_num_groups", 2};
+}
+/// Replace `xevm.*` with an `llvm.call` to the corresponding OpenCL func with
+/// a constant argument for the dimension - x, y or z.
+template <typename OpType>
+class LaunchConfigOpToOCLPattern : public OpConversionPattern<OpType> {
+  using OpConversionPattern<OpType>::OpConversionPattern;
+  LogicalResult
+  matchAndRewrite(OpType op, typename OpType::Adaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+    Location loc = op->getLoc();
+    auto [baseName, dim] = getConfig(op);
+    Type dimTy = rewriter.getI32Type();
+    Value dimVal = LLVM::ConstantOp::create(rewriter, loc, dimTy,
+                                            static_cast<int64_t>(dim));
+    std::string func = mangle(baseName, {dimTy}, {true});
+    Type resTy = op.getType();
+    auto call =
+        createDeviceFunctionCall(rewriter, func, resTy, {dimTy}, {dimVal}, {},
+                                 noUnwindWillReturnAttrs, op.getOperation());
+    constexpr auto noModRef = LLVM::ModRefInfo::NoModRef;
+    auto memAttr = rewriter.getAttr<LLVM::MemoryEffectsAttr>(
+        /*other=*/noModRef,
+        /*argMem=*/noModRef, /*inaccessibleMem=*/noModRef);
+    call.setMemoryEffectsAttr(memAttr);
+    rewriter.replaceOp(op, call);
+    return success();
+  }
+};
+
+/*
+// Subgroup ops
+// get_sub_group_local_id
+xevm::LaneIdOp;
+// get_sub_group_id
+xevm::SubgroupIdOp;
+// get_sub_group_size
+xevm::SubgroupSizeOp;
+// get_num_sub_groups : to be added if needed
+*/
+
+// Helpers to get the OpenCL function name for each op.
+static StringRef getConfig(xevm::LaneIdOp) { return "get_sub_group_local_id"; }
+static StringRef getConfig(xevm::SubgroupIdOp) { return "get_sub_group_id"; }
+static StringRef getConfig(xevm::SubgroupSizeOp) {
+  return "get_sub_group_size";
+}
+template <typename OpType>
+class SubgroupOpWorkitemOpToOCLPattern : public OpConversionPattern<OpType> {
+  using OpConversionPattern<OpType>::OpConversionPattern;
+  LogicalResult
+  matchAndRewrite(OpType op, typename OpType::Adaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+    std::string func = mangle(getConfig(op).str(), {});
+    Type resTy = op.getType();
+    auto call =
+        createDeviceFunctionCall(rewriter, func, resTy, {}, {}, {},
+                                 noUnwindWillReturnAttrs, op.getOperation());
+    constexpr auto noModRef = LLVM::ModRefInfo::NoModRef;
+    auto memAttr = rewriter.getAttr<LLVM::MemoryEffectsAttr>(
+        /*other=*/noModRef,
+        /*argMem=*/noModRef, /*inaccessibleMem=*/noModRef);
+    call.setMemoryEffectsAttr(memAttr);
+    rewriter.replaceOp(op, call);
+    return success();
+  }
+};
+
 //===----------------------------------------------------------------------===//
 // Pass Definition
 //===----------------------------------------------------------------------===//
@@ -775,7 +904,22 @@ void ::mlir::populateXeVMToLLVMConversionPatterns(ConversionTarget &target,
                LLVMLoadStoreToOCLPattern<LLVM::LoadOp>,
                LLVMLoadStoreToOCLPattern<LLVM::StoreOp>,
                BlockLoadStore1DToOCLPattern<BlockLoadOp>,
-               BlockLoadStore1DToOCLPattern<BlockStoreOp>>(
+               BlockLoadStore1DToOCLPattern<BlockStoreOp>,
+               LaunchConfigOpToOCLPattern<WorkitemIdXOp>,
+               LaunchConfigOpToOCLPattern<WorkitemIdYOp>,
+               LaunchConfigOpToOCLPattern<WorkitemIdZOp>,
+               LaunchConfigOpToOCLPattern<WorkgroupDimXOp>,
+               LaunchConfigOpToOCLPattern<WorkgroupDimYOp>,
+               LaunchConfigOpToOCLPattern<WorkgroupDimZOp>,
+               LaunchConfigOpToOCLPattern<WorkgroupIdXOp>,
+               LaunchConfigOpToOCLPattern<WorkgroupIdYOp>,
+               LaunchConfigOpToOCLPattern<WorkgroupIdZOp>,
+               LaunchConfigOpToOCLPattern<GridDimXOp>,
+               LaunchConfigOpToOCLPattern<GridDimYOp>,
+               LaunchConfigOpToOCLPattern<GridDimZOp>,
+               SubgroupOpWorkitemOpToOCLPattern<LaneIdOp>,
+               SubgroupOpWorkitemOpToOCLPattern<SubgroupIdOp>,
+               SubgroupOpWorkitemOpToOCLPattern<SubgroupSizeOp>>(
       patterns.getContext());
 }
 
diff --git a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUSubgroupDistribute.cpp b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUSubgroupDistribute.cpp
@@ -195,8 +195,7 @@ static bool requireTranspose(const xegpu::LayoutAttr layout,
 ///     }
 ///     return %0
 ///   }
-struct MoveFuncBodyToWarpExecuteOnLane0
-    : public OpRewritePattern<gpu::GPUFuncOp> {
+struct MoveFuncBodyToWarpOp : public OpRewritePattern<gpu::GPUFuncOp> {
   using OpRewritePattern<gpu::GPUFuncOp>::OpRewritePattern;
   LogicalResult matchAndRewrite(gpu::GPUFuncOp gpuFuncOp,
                                 PatternRewriter &rewriter) const override {
@@ -1447,6 +1446,11 @@ void xegpu::populateXeGPUSubgroupDistributePatterns(
       /*pattern benefit=*/highPatternBenefit);
 }
 
+void xegpu::populateXeGPUMoveFuncBodyToWarpOpPatterns(
+    RewritePatternSet &patterns) {
+  patterns.add<MoveFuncBodyToWarpOp>(patterns.getContext());
+}
+
 void XeGPUSubgroupDistributePass::runOnOperation() {
   // Step 1: Attach layouts to op operands.
   // TODO: Following assumptions are made:
@@ -1473,7 +1477,7 @@ void XeGPUSubgroupDistributePass::runOnOperation() {
   // gpu.warp_execute_on_lane_0 operation.
   {
     RewritePatternSet patterns(&getContext());
-    patterns.add<MoveFuncBodyToWarpExecuteOnLane0>(&getContext());
+    xegpu::populateXeGPUMoveFuncBodyToWarpOpPatterns(patterns);
 
     if (failed(applyPatternsGreedily(getOperation(), std::move(patterns)))) {
       signalPassFailure();
diff --git a/mlir/test/Conversion/XeVMToLLVM/xevm-to-llvm.mlir b/mlir/test/Conversion/XeVMToLLVM/xevm-to-llvm.mlir
diff --git a/mlir/test/Dialect/XeGPU/move-gpu-func-to-warp-op.mlir b/mlir/test/Dialect/XeGPU/move-gpu-func-to-warp-op.mlir
diff --git a/mlir/test/Dialect/XeGPU/subgroup-distribute-unit.mlir b/mlir/test/Dialect/XeGPU/subgroup-distribute-unit.mlir
diff --git a/mlir/test/lib/Dialect/XeGPU/TestXeGPUTransforms.cpp b/mlir/test/lib/Dialect/XeGPU/TestXeGPUTransforms.cpp

Original file line number	Diff line number	Diff line change
`@@ -1253,7 +1253,8 @@ Instruction *InstCombinerImpl::visitShl(BinaryOperator &I) {`
`1253`	`1253`	`// shl (zext i1 X), C1 --> select (X, 1 << C1, 0)`
`1254`	`1254`	`if (match(Op0, m_ZExt(m_Value(X))) && X->getType()->isIntOrIntVectorTy(1)) {`
`1255`	`1255`	`auto *NewC = Builder.CreateShl(ConstantInt::get(Ty, 1), C1);`
`1256`		`- return createSelectInst(X, NewC, ConstantInt::getNullValue(Ty));`
	`1256`	`+ return createSelectInstWithUnknownProfile(X, NewC,`
	`1257`	`+ ConstantInt::getNullValue(Ty));`
`1257`	`1258`	`}`
`1258`	`1259`	`}`
`1259`	`1260`
Original file line number	Diff line number	Diff line change
`@@ -1737,7 +1737,7 @@ Instruction *InstCombinerImpl::foldBinopOfSextBoolToSelect(BinaryOperator &BO) {`
`1737`	`1737`	`Constant *Zero = ConstantInt::getNullValue(BO.getType());`
`1738`	`1738`	`Value *TVal = Builder.CreateBinOp(BO.getOpcode(), Ones, C);`
`1739`	`1739`	`Value *FVal = Builder.CreateBinOp(BO.getOpcode(), Zero, C);`
`1740`		`- return createSelectInst(X, TVal, FVal);`
	`1740`	`+ return createSelectInstWithUnknownProfile(X, TVal, FVal);`
`1741`	`1741`	`}`
`1742`	`1742`
`1743`	`1743`	`static Value simplifyOperationIntoSelectOperand(Instruction &I, SelectInst SI,`