[MLIR] Provide bare pointer memref lowering option on gpu-to-nvvm pass
authorUday Bondhugula <uday@polymagelabs.com>
Sat, 17 Jun 2023 17:57:59 +0000 (23:27 +0530)
committerUday Bondhugula <uday@polymagelabs.com>
Sun, 18 Jun 2023 17:23:50 +0000 (22:53 +0530)
commit27ccf0f40771047f5e6a20122cd53157df30f593
tree105e53e413e945327baad437d75e68e7099f0689
parentb57e049fe5ccf48bb105b96020901d2912de9588
[MLIR] Provide bare pointer memref lowering option on gpu-to-nvvm pass

Provide the bare pointer memref lowering option on gpu-to-nvvm pass.
This is needed whenever we lower memrefs on the host function side and
the kernel calls on the host-side (gpu-to-llvm) with the bare ptr
convention. The GPU module side of the lowering should also "align" and
use the bare pointer convention.

Reviewed By: krzysz00

Differential Revision: https://reviews.llvm.org/D152480
mlir/include/mlir/Conversion/Passes.td
mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
mlir/test/Conversion/GPUToNVVM/memref.mlir [new file with mode: 0644]
mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32-bare-ptr.mlir [new file with mode: 0644]
mlir/test/Integration/GPU/CUDA/all-reduce-and.mlir