[Mlir-commits] [mlir] [MLIR][NVVM][NVGPU] Combine prefetch and prefetch.tensormap (PR #153134)

[Jordan Rupprecht]

rupprecht wrote:

This PR introduced a `LLVM ERROR: Cannot select: intrinsic %llvm.nvvm.prefetch.tensormap` failure in some tests, e.g. [gemm_f32_f16_f16_128x128x128.mlir](https://github.com/llvm/llvm-project/blob/main/mlir/test/Integration/GPU/CUDA/sm90/gemm_f32_f16_f16_128x128x128.mlir). (See also the comment in https://github.com/llvm/llvm-project/pull/156830#issue-3382764077). @abhilash1910, were you looking at this already?

https://github.com/llvm/llvm-project/pull/153134