https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97417
--- Comment #14 from Jim Wilson <wilson at gcc dot gnu.org> --- Actually, for the SImode to DImode conversion, zero extending is unlikely to be correct in that case. The rv64 'w' instructions require the input to be sign-extended from 32-bits to 64-bits, so a sign-extend is probably required. There will be a similar consideration for rv128 when we get to that. So maybe we should only handle QImode and HImode here. Or maybe we make the choice of zero-extend or sign-extend depend on the mode, and use zero-extend for smaller than SImode and sign-extend for SImode and larger. For qimode, char is unsigned by default, so zero extension is likely the right choice. For himode, it isn't clear which is best, but the arm port does a zero extend. Also, the Huawei code size proposal says that zero exnteded byte and short loads are more important than sign extended byte and short load, so that is more evidence that zero extend is more useful in those cases.
