https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114978
--- Comment #31 from Chen Chen <chz0808 at gmail dot com> --- (In reply to Xi Ruoyao from comment #29) > For 15 r15-7525 is intended for this issue. But I don't know if it's a good > idea to backport it, as it's only a workaround, not a proper fix. > > Could someone try the diff in PR 115842 comment 6 (one time just on the > trunk, another with r15-7525 reverted)? I did a quick test and found that the diff in PR 115842 comment 6 does not work on loongarch64 with gcc14.2.0. >From my tests the regression issue on 548.exchange2_r was actually fixed long before the patch r15-7525 (the workaround you mentioned). Using the same compiling parameters"-O3 -march=la464 -flto", on Debian sid loong64 system I got a score of 9.90 with gcc15 snapshot 20250105, a score of 13.6 with gcc15 snapshot 20250112, and a score of 13.3 with gcc15 snapshot 20250119-20250209. The effect of r15-7525 on 548.exchange2_r is minimal, from a score of 13.3 --—> 13.6 with gcc15 snapshot 20250216 and later (I tested until snapshot 20250302). chenglulu should be able to find which patch is responsible for the score change between gcc15 snapshots 20250105 and 20250112.
