https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124138
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Summary|[16 Regression] wrong code |[13/14/15 Regression] wrong
|with -mavx512f and |code with -mavx512f and
|__builtin_shufflevector() / |__builtin_shufflevector() /
|vpermilpd since r16-3677 |vpermilpd since r0-127405
Priority|P1 |P2
Target Milestone|16.0 |13.5
--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed on the trunk so far. As mentioned, while for V8DImode it is a [16
Regression], for V8DFmode it is wrong-code since GCC 4.9 which added the
AVX512F support, and we should fix it on the release branches too.