https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79177
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org
--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The inline asm input constraints don't make sense, if you have __m256 var, you
should use "x" constraint (i.e. SSE regs) for it rather than "r" (general
regs).
That said, the testcase compiles just fine with -mavx or -mavx2, the error is
only if you compile without -mavx, but then it is appropriate, without AVX
enabled __m256 is just a generic vector, that certainly can't be put into a SSE
register.
