https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110232
Bug ID: 110232
Summary: failure to constant fold VIEW_CONVERT_EXPR<vector(16)
<signed-boolean:1>>(65534)
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: rguenth at gcc dot gnu.org
Target Milestone: ---
With
int a[4096];
void foo ()
{
for (int i = 1; i < 4095; ++i)
a[i] = 42;
}
and -O3 -march=znver4 --param vect-partial-vector-usage=2 -fno-vect-cost-model
which triggers peeling for alignment and a fully masked loop we end up with
<bb 2> [local count: 10737416]:
_17 = VIEW_CONVERT_EXPR<vector(16) <signed-boolean:1>>(65534);
<bb 3> [local count: 75161909]:
# i_8 = PHI <i_5(5), 1(2)>
# ivtmp_2 = PHI <ivtmp_1(5), 4094(2)>
# vectp_a.4_6 = PHI <vectp_a.4_11(5), &a(2)>
# ivtmp_14 = PHI <ivtmp_15(5), 4095(2)>
# loop_mask_12 = PHI <_18(5), _17(2)>
.MASK_STORE (vectp_a.4_6, 512B, loop_mask_12, { 42, 42, 42, 42, 42, 42, 42,
42, 42, 42, 42, 42, 42, 42, 42, 42 });
which isn't constant folded until we expand to RTL.