https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117056
Bug ID: 117056
Summary: Assume vs vectorizer
Product: gcc
Version: 15.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: enhancement
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: pinskia at gcc dot gnu.org
Blocks: 53947
Target Milestone: ---
Take (not fully realistic example but mainly to show how assume statements can
get in the way):
```
static inline
int f(int a, int *c)
{
int d;
if (a > 0)
{
[[assume(*c == 0)]];
d = a;
}
else
{
[[assume(*c != 0)]];
d = -a;
}
return d;
}
void g(int *d, int l, int *e)
{
for(int i = 0; i < l; i++)
{
d[i] = f(d[i], &e[i]);
}
}
void g1(int *d, int l, int *e)
{
for(int i = 0; i < l; i++)
{
[[assume(e[i] == 0)]];
d[i] = d[i] + 1;
}
}
```
Right now g does not get vectorized as phiopt/ifcvt does not handle ASSUME
statements in a reasonible fashion.
g1 does get vectorized BUT the assume statement and the scalar induction
variable stays around and the induction variable selection is a bit odd.
Maybe this should be a few seperate bug reports or maybe this could be the
meta-bug around assume statements.
Anyways since assume is new the code quality to adding assume might be a few
years off though.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53947
[Bug 53947] [meta-bug] vectorizer missed-optimizations
