[Bug rtl-optimization/98212] New: X86 unoptimal code for float equallity comparison followed by jump

[denis.campredon at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98212

            Bug ID: 98212
           Summary: X86 unoptimal code for float equallity comparison
                    followed by jump
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: denis.campredon at gmail dot com
  Target Milestone: ---

For f1 code an unnecessary `comiss` instruction is used, the parity flag is
still set after the `jp` instruction.
For f2, I'm not sure if it's the optimal way to do that.

The same problems appear for `float`, `double` and `long double`

------
void f();

void f1(float a, float b) {
    if (a != b)
        f();
}

void f2(float a, float b) {
    if (a == b)
        f();
}

------
f1(float, float):
        ucomiss xmm0, xmm1
        jp      .L4
        comiss  xmm0, xmm1
        jne     .L4
        ret
.L4:
        jmp     f()

f2(float, float):
        ucomiss xmm0, xmm1
        jnp     .L11
.L7:
        ret
.L11:
        jne     .L7
        jmp     f()
------