https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120352
Bug ID: 120352
Summary: scalar epiloque not needed for early break when exit
block is invariant
Product: gcc
Version: 16.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: tnfchris at gcc dot gnu.org
Blocks: 53947, 115130
Target Milestone: ---
The following sequence
#define N 4
int a[N] = {0,0,0,1};
int b[N] = {0,0,0,1};
__attribute__((noipa, noinline))
int foo ()
{
for (int i = 0; i < N; i++)
{
if (a[i] > b[i])
return 1;
}
return 0;
}
on AArch64 compiles to:
ldr q31, [x2, #:lo12:.LANCHOR0]
ldr q30, [x1, 16]
cmgt v30.4s, v31.4s, v30.4s
umaxp v30.4s, v30.4s, v30.4s
fmov x3, d30
cbnz x3, .L8
ret
.L8:
ldr w0, [x2, #:lo12:.LANCHOR0]
ldr w3, [x1, 16]
cmp w3, w0
blt .L6
ldr w0, [x1, 4]
ldr w2, [x1, 20]
cmp w2, w0
blt .L6
ldr w0, [x1, 8]
ldr w2, [x1, 24]
cmp w2, w0
blt .L6
ldr w0, [x1, 28]
ldr w2, [x1, 12]
cmp w2, w0
cset w0, gt
ret
.L6:
mov w0, 1
ret
However since the body of the break contains only loop invariant instructions
and the remainder of the loop is empty we don't need the scalar epiloque as the
IV value is irrelevant.
This sequence is somewhat common as it's a `contains` check
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53947
[Bug 53947] [meta-bug] vectorizer missed-optimizations
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115130
[Bug 115130] [meta-bug] early break vectorization
