https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117722
--- Comment #4 from JuzheZhong <juzhe.zhong at rivai dot ai> --- (In reply to Robin Dapp from comment #3) > First, pixel_sad_4x4 is not very hot, 8x8 and 16x16 are. > > Second, we are vectorizing this, but with -mno-vector-strict-align. > > IMHO we don't need to synthesize an usad pattern. Why ? We tried implement ssad pattern, the codegen is same as Clang and better than native codegen with -mno-vector-strict-align. And not all RISC-V board support misalign load store.
