On Tue, Oct 5, 2021 at 6:08 PM Jeff Law <jeffreya...@gmail.com> wrote:
> OK. I think there may still be some concern if the latch is marked as a > joiner. I think we should always reject those before the loop optimizers > run as the joiner's clone would introduce a second latch. I think that can > be handled in a follow-up refinement. I've been mumbling, to myself mostly, that the loop threading restrictions should be owned by the loop experts. There's not much I can contribute here, except testcases and patches that I *think* are what y'all want. That being said, I will follow-up with the suggestions made by Richi, because I've already done the work, and have cobbled up the relevant gimple FE tests ;-). > > > > > > > Co-authored-by: Jeff Law <jeffrea...@gmail.com> > > I'm not sure I deserve co-authorship :-) You're doing all the work, I'm just > reporting the regressions. Heh. Heh. Ok. I honestly thought about putting everyone's name here. Aldy
