[Mlir-commits] [mlir] [mlir][GPUToNVVM] enable fallback to generic LLVM lowering for math dialect in convert-gpu-to-nvvm pass (PR #165728)

[Yang Bai]

yangtetris wrote:

> I'd need to check, I know there are some things that have to be improved, but I think for the `gpu-to-nvvm` case almost everything is there.

Hi @fabianmcg , I'm trying to replace the usage of `convert-gpu-to-nvvm` with `dynamic convert-to-llvm + ConvertToLLVMAttrInterface` in my project. However, I can't find any way to set the `LowerToLLVMOption::useBarePtrCallConv` field, which is needed for creating the type converter. Could we enhance the `convert-to-llvm` pass to support `useBarePtrCallConv`? 


https://github.com/llvm/llvm-project/pull/165728