Saieiei wrote: Thanks @skatrak this makes sense. I’ll first clean up the mechanical issues you pointed out. For the representation question, my current implementation followed the existing task/taskloop in_reduction block-arg pattern, but I see why target is different from the device perspective since the mapped value is what the device body should naturally use. I’ll look into whether target in_reduction should instead be modeled through the map entry block argument and keep the extra reduction handling host-side only.
https://github.com/llvm/llvm-project/pull/199967 _______________________________________________ llvm-branch-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
