[Openmp-commits] [openmp] [OpenMP][libomptarget][RFC] extend libomptarget with mechanism to execute fill memory on the target (PR #73801)

Johannes Doerfert via Openmp-commits openmp-commits at lists.llvm.org
Thu Nov 30 09:51:13 PST 2023


jdoerfert wrote:

I see. I'm not unhappy with this, so if you clean it up we can merge it. I also believe a kernel library is going to be super useful, so once we have that, and performance is good, we can rip this out again. Since I won't do the kernel library right away, I think, we can totally keep this for now. If you want to look into the library approach, I'm happy with that too. Maybe measure the memset speed first. Nice thing would be that the kernel would work for any size, not just multiples of 4.


https://github.com/llvm/llvm-project/pull/73801


More information about the Openmp-commits mailing list