https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85919
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |crazylht at gmail dot com
--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Reading back through the mail thread I think a reasonable simplification on the
x86 backend side would be to only expose mask_gather_load with AVX512VL,
side-stepping the issue of mixed AVX2 + AVX512 style masking with
-mno-avx512lvl.
Re-using the define-insns is probably OK, but the mode-iterators for the
mask_gather_load/store define_expands should be modeled to have the number
of output and offset lanes the same (and accept both AVX2 and AVX512 style
masks).
https://gcc.gnu.org/pipermail/gcc-patches/2021-August/577656.html
had the last actual patch posted.
