https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112325

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu.org
            Summary|Missed vectorization after  |Missed vectorization of
                   |cunrolli                    |reduction after unrolling

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
We're detecting the integer reduction:

t.c:21:18: note:   Starting SLP discovery for
t.c:21:18: note:     _236 = _230 * _235;
t.c:21:18: note:     _223 = _217 * _222;
t.c:21:18: note:     _210 = _204 * _209;
t.c:21:18: note:     _197 = _191 * _196;
t.c:21:18: note:     _184 = _178 * _183;
t.c:21:18: note:     _171 = _165 * _170;
t.c:21:18: note:     _158 = _152 * _157;
t.c:21:18: note:     _145 = _139 * _144;
t.c:21:18: note:     _132 = _126 * _131;
t.c:21:18: note:     _119 = _113 * _118;
t.c:21:18: note:     _106 = _100 * _105;
t.c:21:18: note:     _93 = _87 * _92;
t.c:21:18: note:     _80 = _74 * _79;
t.c:21:18: note:     _67 = _61 * _66;
t.c:21:18: note:     sumi.1_42 = (unsigned int) sumi_96;
t.c:21:18: note:     _41 = _27 * _40;
t.c:21:18: note:     _54 = _48 * _53;

but we're not clever enough to split out

t.c:21:18: note:     sumi.1_42 = (unsigned int) sumi_96;

it's possibly easy enough to do.

Reply via email to