https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79079
--- Comment #3 from Kito Cheng <npickito at gmail dot com> --- Hi Jakub: Thank your review, your are right, I might change all gen_lowpart to convert_modes to deal with all other TRULY_NOOP_TRUNCATION == false case, in RISC-V and MIPS64 only TRULY_NOOP_TRUNCATION(64, 32) is false so only one modification is enough but it's still potential bug here. I will update the patch, re-run testsuite and post to gcc-patch mailing list, thanks again :)
