https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123870
Robin Dapp <rdapp at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rdapp at gcc dot gnu.org
--- Comment #5 from Robin Dapp <rdapp at gcc dot gnu.org> ---
I'm sorry that you're experiencing those issues with xtheadvector. It's
obvious that test coverage is much sparser than regular RVV and surely not
where we would like it to be. But there hasn't been a lot of commitment from
the vendor since the integration of the code and we might be starting to see
some bit rot by now.
That said, aren't riscv_vzext and riscv_vsext not available for xtheadvector
anyway? We shouldn't ICE of course but I think we don't anymore with GCC 16,
we just didn't backport everything I'm afraid. I'm not sure what the canonical
way to extend is, probably widening addition with zero?
And the same is probably true for the full-register moves (vmv2r etc.)?
So the bug is that GCC 14 emits it even though it is not available?
And we fixed that in 15 without backporting?
