[mlir] make the bitwidth of device side index computations configurable
authorTobias Gysi <tobias.gysi@inf.ethz.ch>
Mon, 22 Jun 2020 09:37:01 +0000 (11:37 +0200)
committerAlex Zinenko <zinenko@google.com>
Mon, 22 Jun 2020 09:43:37 +0000 (11:43 +0200)
commitd10b1a38a7dfb994623f27f263b67f5fc76e08cc
tree30d5b0b8cc402e35c3145f971e05c13cd84baa5d
parente935a540ea29d5de297595801aed1fb70fabfbf6
[mlir] make the bitwidth of device side index computations configurable

The patch makes the index type lowering of the GPU to NVVM/ROCDL
conversion configurable. It introduces a pass option that controls the
bitwidth used when lowering index computations.

Differential Revision: https://reviews.llvm.org/D80285
mlir/include/mlir/Conversion/GPUToNVVM/GPUToNVVMPass.h
mlir/include/mlir/Conversion/GPUToROCDL/GPUToROCDLPass.h
mlir/include/mlir/Conversion/Passes.td
mlir/include/mlir/Conversion/StandardToLLVM/ConvertStandardToLLVM.h
mlir/include/mlir/Conversion/StandardToLLVM/ConvertStandardToLLVMPass.h
mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
mlir/lib/Conversion/GPUToROCDL/LowerGpuOpsToROCDLOps.cpp
mlir/lib/Conversion/StandardToLLVM/StandardToLLVM.cpp
mlir/test/Conversion/GPUToNVVM/gpu-to-nvvm.mlir
mlir/test/Conversion/GPUToROCDL/gpu-to-rocdl.mlir