https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114684
Bug ID: 114684
Summary: `-(int)(unsigned:1)signed:1` could just be
`(int)signed:1`
Product: gcc
Version: 14.0
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
Target Milestone: ---
Take:
```
struct t{
signed s1 : 1;
unsigned u1 : 1;
} b = {-1};
int f()
{
struct t tt = b;
return tt.s1;
}
int f1()
{
struct t tt = b;
tt.u1 = tt.s1;
int ttt = tt.u1;
return -ttt;
}
```
These 2 functions should produce the same code. (For aarch64 it almost does,
there is an extra sign extend in f). I noticed this while fixing PR 114666. It
is also related to PR 101955. Note this only works for 1bit signed values.
