https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55212
--- Comment #403 from Oleg Endo <olegendo at gcc dot gnu.org> --- (In reply to Kazumoto Kojima from comment #402) > I've opened PR 117182 for a wrong fldi0/1 issue. Looks target + RA issue. > > FYI, with the patches in PR 116932, PR 117111 and PR 117182 on top of > devel/sh-lra, c/c++ testsuite shows only one regression against trunk > -mno-lra: I've added the patch for PR 117182 (with a few alterations to the comments) to https://github.com/olegendo/gcc/tree/devel/sh-lra next to the other LRA patch from comment #277 ... and also rebased the branch on top of commit https://gcc.gnu.org/g:2390cbad85cbd122d4e58c94f7891d7c5fde49b3 This now includes the commit https://gcc.gnu.org/g:e7393cbb5f2cae50b42713e71984064073aa378a I'm wondering if that LRA patch could be doing anything for SH, since we've been having issues with scratch registers in patterns as well. > > FAIL: gcc.target/sh/hiconst.c scan-assembler-times mov\t#0 2 > There are a bunch of those target/sh micro-optimization tests that have been failing also without LRA. I'm trying to catch up with those gradually. It shouldn't be a show stopper. Similarly the issues with non-working GBR addressing modes in atomics ( PR 116932 ).