On 2023/12/21 1:08, Jeff Law wrote: > > > On 12/20/23 01:39, Wang Pengcheng wrote: >> From: wangpc >> >> The condition is RISCV_FUSE_ZEXTH, which is a mistake. >> >> gcc/ChangeLog: >> >> * config/riscv/riscv.cc (riscv_macro_fusion_pair_p): Fix condition. > Thanks! As soon as this patch finishes regression testing I'll push > it to the trunk. > > I'm curious, how did you find this? Did you see a case where fusible > ops weren't being kept together or did you find it just by code > examination? > > Jeff
Yeah, I just found it when I tried to understand the original fusion implementation commit. :-) And thanks a lot for formatting and merging the patch! I'm sorry that I almost forgot the process of contribution to GCC as the last time I did it is about two years ago.
