https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125700
Bug ID: 125700
Summary: `(x == MAX_VALUE) ? y : MIN<x, y>` -> MIN<x,y>
Product: gcc
Version: 16.0
Status: UNCONFIRMED
Keywords: easyhack, missed-optimization
Severity: enhancement
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: pinskia at gcc dot gnu.org
Target Milestone: ---
Take:
```
unsigned fumin(unsigned x, unsigned y)
{
if (x == -1u) return y;
return x < y ? x : y;
}
signed fsmin(signed x, signed y)
{
if (x == __INT_MAX__) return y;
return x < y ? x : y;
}
unsigned fumax(unsigned x, unsigned y)
{
if (x == 0) return y;
return x > y ? x : y;
}
signed fsmax(signed x, signed y)
{
if (x == -__INT_MAX__-1) return y;
return x > y ? x : y;
}
```
The comparison against the min/max value should be removed and these just
become min/max of x and y.
This is the generic version of what was mentioned in
https://github.com/llvm/llvm-project/issues/202576 . Mentioning umin/umax and
smin/smax.
