https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101260
--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> --- I've reproduced it with -O1 -march=z14 and it started with r10-7093-g5dc1390b41db5c1765e25fd21dad1a930a015aac So, I think it is much more likely some RA issue or RTL optimization issue and the r12-145 change just makes it latent.
