[CIR][X86] Add support for cpuid/cpuidex #173197
Open
+108 −1
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
rturrado requested review from andykaylor, bcardosolopes, lanza and xlauko as code owners 
llvmbot added clang 
Member
| @llvm/pr-subscribers-clang Author: Roberto Turrado Camblor (rturrado) ChangesAdds support for the Part of 167765. Full diff: https://github.com/llvm/llvm-project/pull/173197.diff 3 Files Affected:
diff --git a/clang/include/clang/CIR/Dialect/IR/CIROps.td b/clang/include/clang/CIR/Dialect/IR/CIROps.td index 7e7424fd71878..c1292ac6ce241 100644 --- a/clang/include/clang/CIR/Dialect/IR/CIROps.td +++ b/clang/include/clang/CIR/Dialect/IR/CIROps.td @@ -5762,4 +5762,31 @@ def CIR_BlockAddressOp : CIR_Op<"block_address", [Pure]> { }]; } +//===----------------------------------------------------------------------===// +// CpuIdOp +//===----------------------------------------------------------------------===// + +def CIR_CpuIdOp : CIR_Op<"cpuid"> { + let summary = "Get information about the CPU"; + let description = [{ + The `cir.cpuid` operation takes a base pointer to an array of 4 integers, a + function ID and a sub-function ID. The array of 4 integers is filled with + different information about the processor. + + Example: + + ```mlir + cir.cpuid %basePtr, %funcId, %subFuncId + : (!cir.ptr<!cir.array<4 x !s32i>>, !s32i, !s32i) -> + ``` + }]; + + let arguments = + (ins Arg<CIR_PtrToArray, "array address", [MemWrite]>:$basePtr, + CIR_SInt32:$funcId, CIR_SInt32:$subFuncId); + // TODO: remove once we can return an optional mlir::Value from + // emitX86BuiltinExpr + let results = (outs CIR_VectorType:$result); +} + #endif // CLANG_CIR_DIALECT_IR_CIROPS_TD diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp b/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp index 1c87e945de846..c58391ecc5b42 100644 --- a/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp @@ -25,6 +25,7 @@ #include "clang/CIR/Dialect/IR/CIRTypes.h" #include "clang/CIR/MissingFeatures.h" #include "llvm/ADT/Sequence.h" +#include "llvm/IR/InlineAsm.h" #include "llvm/Support/ErrorHandling.h" #include <string> @@ -1835,7 +1836,15 @@ CIRGenFunction::emitX86BuiltinExpr(unsigned builtinID, const CallExpr *expr) { case X86::BI__builtin_ia32_cvtneps2bf16_256_mask: case X86::BI__builtin_ia32_cvtneps2bf16_512_mask: case X86::BI__cpuid: - case X86::BI__cpuidex: + case X86::BI__cpuidex: { + mlir::Type i32Ty = builder.getSInt32Ty(); + mlir::Value subFuncId = builtinID == X86::BI__cpuidex + ? ops[2] + : builder.getConstInt(loc, sInt32Ty, 0); + cir::CpuIdOp::create(builder, loc, i32Ty, /*basePtr=*/ops[0], + /*funcId=*/ops[1], /*subFuncId=*/subFuncId); + return mlir::Value{}; + } case X86::BI__emul: case X86::BI__emulu: case X86::BI__mulh: diff --git a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp index eb0a219f18618..24924085ea9ea 100644 --- a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp +++ b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp @@ -4193,6 +4193,76 @@ mlir::LogicalResult CIRToLLVMAwaitOpLowering::matchAndRewrite( return mlir::failure(); } +mlir::LogicalResult CIRToLLVMCpuIdOpLowering::matchAndRewrite( + cir::CpuIdOp op, OpAdaptor adaptor, + mlir::ConversionPatternRewriter &rewriter) const { + mlir::Type i32Ty = rewriter.getI32Type(); + mlir::Type i64Ty = rewriter.getI64Type(); + mlir::Type i32PtrTy = mlir::LLVM::LLVMPointerType::get(i32Ty.getContext(), 0); + + mlir::Type cpuidRetTy = mlir::LLVM::LLVMStructType::getLiteral( + rewriter.getContext(), {i32Ty, i32Ty, i32Ty, i32Ty}); + + mlir::Value funcId = adaptor.getFuncId(); + mlir::Value subFuncId = adaptor.getSubFuncId(); + mlir::StringAttr opNameAttr = op->getAttrOfType<mlir::StringAttr>("name"); + if (!opNameAttr) + return mlir::failure(); + if (opNameAttr.getValue() == "cpuid") + subFuncId = mlir::LLVM::ConstantOp::create(rewriter, op.getLoc(), i32Ty, 0); + std::vector operands{funcId, subFuncId}; + + StringRef asmString, constraints; + mlir::ModuleOp moduleOp = op->getParentOfType<mlir::ModuleOp>(); + mlir::StringAttr tripleAttr = + moduleOp->getAttrOfType<mlir::StringAttr>("llvm.target_triple"); + if (!tripleAttr) + return mlir::failure(); + llvm::Triple triple(tripleAttr.getValue().str()); + if (triple.getArch() == llvm::Triple::x86) { + asmString = "cpuid"; + constraints = "={ax},={bx},={cx},={dx},{ax},{cx}"; + } else { + // x86-64 uses %rbx as the base register, so preserve it. + asmString = "xchgq %rbx, ${1:q}\n" + "cpuid\n" + "xchgq %rbx, ${1:q}"; + constraints = "={ax},=r,={cx},={dx},0,2"; + } + + mlir::Value inlineAsm = + mlir::LLVM::InlineAsmOp::create( + rewriter, op.getLoc(), cpuidRetTy, mlir::ValueRange(operands), + rewriter.getStringAttr(asmString), + rewriter.getStringAttr(constraints), + /*has_side_effects=*/mlir::UnitAttr{}, + /*is_align_stack=*/mlir::UnitAttr{}, + /*tail_call_kind=*/mlir::LLVM::TailCallKindAttr{}, + /*asm_dialect=*/mlir::LLVM::AsmDialectAttr{}, + /*operand_attrs=*/mlir::ArrayAttr{}) + .getResult(0); + + mlir::Value basePtr = adaptor.getBasePtr(); + + mlir::DataLayout layout(op->getParentOfType<mlir::ModuleOp>()); + unsigned alignment = layout.getTypeABIAlignment(i32Ty); + for (unsigned i = 0; i < 4; i++) { + mlir::Value extracted = + mlir::LLVM::ExtractValueOp::create(rewriter, op.getLoc(), inlineAsm, i) + .getResult(); + mlir::Value index = mlir::LLVM::ConstantOp::create( + rewriter, op.getLoc(), i64Ty, rewriter.getI64IntegerAttr(i)); + llvm::SmallVector<mlir::Value, 1> gepIndices = {index}; + mlir::Value storePtr = mlir::LLVM::GEPOp::create( + rewriter, op.getLoc(), i32PtrTy, i32Ty, basePtr, + gepIndices, mlir::LLVM::GEPNoWrapFlags::none) + .getResult(); + mlir::LLVM::StoreOp::create(rewriter, op.getLoc(), extracted, storePtr, + alignment); + } + return mlir::success(); +} + std::unique_ptr<mlir::Pass> createConvertCIRToLLVMPass() { return std::make_unique<ConvertCIRToLLVMPass>(); } |
Member
| @llvm/pr-subscribers-clangir Author: Roberto Turrado Camblor (rturrado) ChangesAdds support for the Part of 167765. Full diff: https://github.com/llvm/llvm-project/pull/173197.diff 3 Files Affected:
diff --git a/clang/include/clang/CIR/Dialect/IR/CIROps.td b/clang/include/clang/CIR/Dialect/IR/CIROps.td index 7e7424fd71878..c1292ac6ce241 100644 --- a/clang/include/clang/CIR/Dialect/IR/CIROps.td +++ b/clang/include/clang/CIR/Dialect/IR/CIROps.td @@ -5762,4 +5762,31 @@ def CIR_BlockAddressOp : CIR_Op<"block_address", [Pure]> { }]; } +//===----------------------------------------------------------------------===// +// CpuIdOp +//===----------------------------------------------------------------------===// + +def CIR_CpuIdOp : CIR_Op<"cpuid"> { + let summary = "Get information about the CPU"; + let description = [{ + The `cir.cpuid` operation takes a base pointer to an array of 4 integers, a + function ID and a sub-function ID. The array of 4 integers is filled with + different information about the processor. + + Example: + + ```mlir + cir.cpuid %basePtr, %funcId, %subFuncId + : (!cir.ptr<!cir.array<4 x !s32i>>, !s32i, !s32i) -> + ``` + }]; + + let arguments = + (ins Arg<CIR_PtrToArray, "array address", [MemWrite]>:$basePtr, + CIR_SInt32:$funcId, CIR_SInt32:$subFuncId); + // TODO: remove once we can return an optional mlir::Value from + // emitX86BuiltinExpr + let results = (outs CIR_VectorType:$result); +} + #endif // CLANG_CIR_DIALECT_IR_CIROPS_TD diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp b/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp index 1c87e945de846..c58391ecc5b42 100644 --- a/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp +++ b/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp @@ -25,6 +25,7 @@ #include "clang/CIR/Dialect/IR/CIRTypes.h" #include "clang/CIR/MissingFeatures.h" #include "llvm/ADT/Sequence.h" +#include "llvm/IR/InlineAsm.h" #include "llvm/Support/ErrorHandling.h" #include <string> @@ -1835,7 +1836,15 @@ CIRGenFunction::emitX86BuiltinExpr(unsigned builtinID, const CallExpr *expr) { case X86::BI__builtin_ia32_cvtneps2bf16_256_mask: case X86::BI__builtin_ia32_cvtneps2bf16_512_mask: case X86::BI__cpuid: - case X86::BI__cpuidex: + case X86::BI__cpuidex: { + mlir::Type i32Ty = builder.getSInt32Ty(); + mlir::Value subFuncId = builtinID == X86::BI__cpuidex + ? ops[2] + : builder.getConstInt(loc, sInt32Ty, 0); + cir::CpuIdOp::create(builder, loc, i32Ty, /*basePtr=*/ops[0], + /*funcId=*/ops[1], /*subFuncId=*/subFuncId); + return mlir::Value{}; + } case X86::BI__emul: case X86::BI__emulu: case X86::BI__mulh: diff --git a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp index eb0a219f18618..24924085ea9ea 100644 --- a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp +++ b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp @@ -4193,6 +4193,76 @@ mlir::LogicalResult CIRToLLVMAwaitOpLowering::matchAndRewrite( return mlir::failure(); } +mlir::LogicalResult CIRToLLVMCpuIdOpLowering::matchAndRewrite( + cir::CpuIdOp op, OpAdaptor adaptor, + mlir::ConversionPatternRewriter &rewriter) const { + mlir::Type i32Ty = rewriter.getI32Type(); + mlir::Type i64Ty = rewriter.getI64Type(); + mlir::Type i32PtrTy = mlir::LLVM::LLVMPointerType::get(i32Ty.getContext(), 0); + + mlir::Type cpuidRetTy = mlir::LLVM::LLVMStructType::getLiteral( + rewriter.getContext(), {i32Ty, i32Ty, i32Ty, i32Ty}); + + mlir::Value funcId = adaptor.getFuncId(); + mlir::Value subFuncId = adaptor.getSubFuncId(); + mlir::StringAttr opNameAttr = op->getAttrOfType<mlir::StringAttr>("name"); + if (!opNameAttr) + return mlir::failure(); + if (opNameAttr.getValue() == "cpuid") + subFuncId = mlir::LLVM::ConstantOp::create(rewriter, op.getLoc(), i32Ty, 0); + std::vector operands{funcId, subFuncId}; + + StringRef asmString, constraints; + mlir::ModuleOp moduleOp = op->getParentOfType<mlir::ModuleOp>(); + mlir::StringAttr tripleAttr = + moduleOp->getAttrOfType<mlir::StringAttr>("llvm.target_triple"); + if (!tripleAttr) + return mlir::failure(); + llvm::Triple triple(tripleAttr.getValue().str()); + if (triple.getArch() == llvm::Triple::x86) { + asmString = "cpuid"; + constraints = "={ax},={bx},={cx},={dx},{ax},{cx}"; + } else { + // x86-64 uses %rbx as the base register, so preserve it. + asmString = "xchgq %rbx, ${1:q}\n" + "cpuid\n" + "xchgq %rbx, ${1:q}"; + constraints = "={ax},=r,={cx},={dx},0,2"; + } + + mlir::Value inlineAsm = + mlir::LLVM::InlineAsmOp::create( + rewriter, op.getLoc(), cpuidRetTy, mlir::ValueRange(operands), + rewriter.getStringAttr(asmString), + rewriter.getStringAttr(constraints), + /*has_side_effects=*/mlir::UnitAttr{}, + /*is_align_stack=*/mlir::UnitAttr{}, + /*tail_call_kind=*/mlir::LLVM::TailCallKindAttr{}, + /*asm_dialect=*/mlir::LLVM::AsmDialectAttr{}, + /*operand_attrs=*/mlir::ArrayAttr{}) + .getResult(0); + + mlir::Value basePtr = adaptor.getBasePtr(); + + mlir::DataLayout layout(op->getParentOfType<mlir::ModuleOp>()); + unsigned alignment = layout.getTypeABIAlignment(i32Ty); + for (unsigned i = 0; i < 4; i++) { + mlir::Value extracted = + mlir::LLVM::ExtractValueOp::create(rewriter, op.getLoc(), inlineAsm, i) + .getResult(); + mlir::Value index = mlir::LLVM::ConstantOp::create( + rewriter, op.getLoc(), i64Ty, rewriter.getI64IntegerAttr(i)); + llvm::SmallVector<mlir::Value, 1> gepIndices = {index}; + mlir::Value storePtr = mlir::LLVM::GEPOp::create( + rewriter, op.getLoc(), i32PtrTy, i32Ty, basePtr, + gepIndices, mlir::LLVM::GEPNoWrapFlags::none) + .getResult(); + mlir::LLVM::StoreOp::create(rewriter, op.getLoc(), extracted, storePtr, + alignment); + } + return mlir::success(); +} + std::unique_ptr<mlir::Pass> createConvertCIRToLLVMPass() { return std::make_unique<ConvertCIRToLLVMPass>(); } |
Contributor Author
| I haven't added any tests yet. I have seen there are currently 2 types of tests for
Thanks! |
rturrado force-pushed the 167765_cpuid branch from 1bd67a4 to 111be7a Compare 
🐧 Linux x64 Test Results
✅ The build succeeded and all tests passed. |
rturrado force-pushed the 167765_cpuid branch 2 times, most recently from 980272e to ebcbcc3 Compare 
rturrado force-pushed the 167765_cpuid branch from ebcbcc3 to 412f88f Compare 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.
Adds support for the
__builtin_ia32_cpuidand__builtin_ia32_cpuidexX86 builtins.Part of 167765.