On Mon, 2026-06-29 at 17:39 -0400, Vladimir Makarov wrote: > On 6/28/26 8:45 AM, Oleg Endo wrote: > > Hi all, > > > > This patch has been sitting in my SH-LRA conversion stash for a while. I > > can't reproduce the associated issues without the patch anymore on the > > latest master version (or haven't tried hard enough). There have been > > several changes and fixes to LRA since the patch was written, so I'm not > > sure if these changes are still needed at all? I'd like to avoid adding > > more clutter to the backend if possible. > > > > In essence, what this patch does is rejecting invalid hard-regs for memory > > address registers when using LRA. It seems the behavior of reload and LRA > > used to differ in that regard. > > > > Any feedback highly appreciated. Thanks a lot! > > I am not a maintainer so can not approve the patch but it looks ok to me > as it accepts pseudo-registers which exist in LRA until its very end. >
Thanks for looking at it! Committed to master. Best regards, Oleg Endo
