[Mlir-commits] [mlir] [AMDGPU] Implement gpu.subgroup_reduce with DPP intrinsics on AMD GPUs (PR #133204)

[llvmlistbot at llvm.org]

stefankoncarevic wrote:

Are there plans to ensure compatibility with RDNA architectures specifically Navi3 and Navi4 in this implementation? 
If so, it's important to note that bcast_15 and bcast_31 are not supported for Navi3 and Navi4. Instead, you might want to consider using permlane as an alternative for these architectures.

https://github.com/llvm/llvm-project/pull/133204