[Bug target/82858] New: __builtin_add_overflow() generates suboptimal code with unsigned types on x86

[lh_mouse at 126 dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82858

            Bug ID: 82858
           Summary: __builtin_add_overflow() generates suboptimal code
                    with unsigned types on x86
           Product: gcc
           Version: 7.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: lh_mouse at 126 dot com
  Target Milestone: ---

The following function:

```
unsigned saturated_add(unsigned a, unsigned b){
    unsigned c;
    if(__builtin_add_overflow(a, b, &c)){
        return -1;
    }
    return c;
}
```

, after being compiled with `gcc -O3 -Wall -Wextra -pedantic -pedantic-errors
-masm=intel`, results in:

```
saturated_add(unsigned int, unsigned int):
  add edi, esi
  jc .L3
  mov eax, edi
  ret
.L3:
  or eax, -1
  ret
```

This is suboptimal, as the branch can be simply eliminated as follows:

```
saturated_add(unsigned int, unsigned int):
  add edi, esi
  sbb eax, eax  // EAX = -1 if CF is set and 0 otherwise.
  or eax, edi
  ret
```