[mlir] Flip accessors to prefixed form (NFC)Another mechanical sweep to keep diff small for flip to _Prefixed.
[mlir] Update accessors to prefixed form (NFC)Follow up from flipping dialects to both, flip accessor used to prefixedvariant ahead to flipping from _Both to _Prefixed. This just flips tothe acce
[mlir] Update accessors to prefixed form (NFC)Follow up from flipping dialects to both, flip accessor used to prefixedvariant ahead to flipping from _Both to _Prefixed. This just flips tothe accessors introduced in the preceding change which are just prefixedforms of the existing accessor changed from.Mechanical change using helper scripthttps://github.com/jpienaar/llvm-project/blob/main/clang-tools-extra/clang-tidy/misc/AddGetterCheck.cpp and clang-format.
show more ...
[mlir][nvgpu] fix MSVC warning regarding left shiftDifferential Revision: https://reviews.llvm.org/D128088
[mlir][nvgpu] fix missing build dependency for NVGPUTransformsFixes build failure caused by 51b925df941a66349deff2467203acc200de5e78
[mlir][nvgpu] shared memory access optimization passThis change adds a transformation and pass to the NvGPU dialect thatattempts to optimize reads/writes from a memref representing GPU sharedmem
[mlir][nvgpu] shared memory access optimization passThis change adds a transformation and pass to the NvGPU dialect thatattempts to optimize reads/writes from a memref representing GPU sharedmemory in order to avoid bank conflicts. Given a value representing ashared memory memref, it traverses all reads/writes within the parent opand, subject to suitable conditions, rewrites all last dimension indexvalues such that element locations in the final (col) dimension aregiven by`newColIdx = col % vecSize + perm[row](col/vecSize,row)`where `perm` is a permutation function indexed by `row` and `vecSize`is the vector access size in elements (currently assumes 128bitvectorized accesses, but this can be made a parameter). This specifictransformation can help optimize typical distributed & vectorized accessescommon to loading matrix multiplication operands to/from shared memory.Differential Revision: https://reviews.llvm.org/D127457