https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126042
Bug ID: 126042
Summary: -O2 -fno-trapping-math makes worse code than with
trapping math for `<=>`
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: enhancement
Priority: P3
Component: tree-optimization
Assignee: pinskia at gcc dot gnu.org
Reporter: pinskia at gcc dot gnu.org
Blocks: 54192
Target Milestone: ---
Take:
```
#include <compare>
bool f5 (double i, double j)
{
auto c = i <=> j; return c >= 0;
}
```
At -O2 -fno-trapping-math we get in phiopt4:
```
_6 = i_1(D) > j_2(D);
_5 = i_1(D) u>= j_2(D);
_4 = _5 & _6;
```
That is:
```
int f(double i, double j)
{
int t = i > j;
int t1 = !__builtin_isless (i,j);
return t & t1;
}
``
Which is only optimized during reassociation to `i > j`. But really should be
done in match too.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54192
[Bug 54192] -fno-trapping-math by default?
