https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104658
Bug ID: 104658
Summary: Inefficient vectorization using mask CTORs
Product: gcc
Version: 12.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: rguenth at gcc dot gnu.org
Target Milestone: ---
Originally observed as part of PR101636 - we are sometimes mixing mask &
non-mask
vector defs resulting in external defs for SLP nodes of mask type which are not
code-generated efficiently, resulting in
<signed-boolean:1> _135 = _49 ? -1 : 0;
_144 = { _135, .... }
and quite awful bit-insert code. That's not expected cost wise and the
appropriate fix is to the bool pattern recog which should have prepared the
IL to avoid the situation.
