mlir::x86vector::avx2 Namespace Reference

Helpers extracted from: More...

Namespaces
	inline_asm
	intrin

Classes
struct	TransposeLoweringOptions
	Structure to control the behavior of specialized AVX2 transpose lowering. More...
struct	LoweringOptions
	Options for controlling specialized AVX2 lowerings. More...

Functions
void	transpose4x8xf32 (ImplicitLocOpBuilder &ib, MutableArrayRef< Value > vs)
	Generic lowerings may either use intrin or inline_asm depending on needs. More...
void	transpose8x8xf32 (ImplicitLocOpBuilder &ib, MutableArrayRef< Value > vs)
	8x8xf32-specific AVX2 transpose lowering. More...
void	populateSpecializedTransposeLoweringPatterns (RewritePatternSet &patterns, LoweringOptions options=LoweringOptions(), int benefit=10)
	Insert specialized transpose lowering patterns. More...

Detailed Description

Helpers extracted from:

• clang/lib/Headers/avxintrin.h
• clang/test/CodeGen/X86/avx-builtins.c
• clang/test/CodeGen/X86/avx2-builtins.c
• clang/test/CodeGen/X86/avx-shuffle-builtins.c as well as the Intel Intrinsics Guide (https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html) make it easier to just implement known good lowerings. All intrinsics correspond 1-1 to the Intel definition.

Function Documentation

populateSpecializedTransposeLoweringPatterns()

void mlir::x86vector::avx2::populateSpecializedTransposeLoweringPatterns	(	RewritePatternSet &	patterns,
		LoweringOptions	options = LoweringOptions(),
		int	benefit = 10
	)

Insert specialized transpose lowering patterns.

Definition at line 287 of file AVXTranspose.cpp.

References options, and mlir::patterns.

transpose4x8xf32()

void mlir::x86vector::avx2::transpose4x8xf32	(	ImplicitLocOpBuilder &	ib,
		MutableArrayRef< Value >	vs
	)

Generic lowerings may either use intrin or inline_asm depending on needs.

AVX2 4x8xf32-specific transpose lowering using a "C intrinsics" model.

4x8xf32-specific AVX2 transpose lowering.

Definition at line 114 of file AVXTranspose.cpp.

References mlir::get(), mlir::Builder::getContext(), mlir::x86vector::avx2::intrin::mm256Permute2f128Ps(), mlir::x86vector::avx2::intrin::mm256ShufflePs(), mlir::x86vector::avx2::intrin::mm256UnpackHiPs(), and mlir::x86vector::avx2::intrin::mm256UnpackLoPs().

Referenced by TransposeOpLowering::matchAndRewrite().

transpose8x8xf32()

void mlir::x86vector::avx2::transpose8x8xf32	(	ImplicitLocOpBuilder &	ib,
		MutableArrayRef< Value >	vs
	)

8x8xf32-specific AVX2 transpose lowering.

AVX2 8x8xf32-specific transpose lowering using a "C intrinsics" model.

Definition at line 139 of file AVXTranspose.cpp.

References mlir::get(), mlir::Builder::getContext(), mlir::x86vector::avx2::inline_asm::mm256BlendPsAsm(), mlir::x86vector::avx2::intrin::mm256Permute2f128Ps(), mlir::x86vector::avx2::intrin::mm256ShufflePs(), mlir::x86vector::avx2::intrin::mm256UnpackHiPs(), and mlir::x86vector::avx2::intrin::mm256UnpackLoPs().

Referenced by TransposeOpLowering::matchAndRewrite().