[Bug tree-optimization/99946] New: fail to exchange if conditions in terms of likely/unlikely probability

[Bug tree-optimization/99946] New: fail to exchange if conditions in terms of likely/unlikely probability

[jiangning.liu at amperecomputing dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99946

            Bug ID: 99946
           Summary: fail to exchange if conditions in terms of
                    likely/unlikely probability
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jiangning.liu at amperecomputing dot com
  Target Milestone: ---

For this simple case,

$ cat test_cond.c 
#define likely(x)       __builtin_expect((x),1)
#define unlikely(x)     __builtin_expect((x),0)

extern void g(void);

int a, b;
void f(void)
{
  if (likely(a>0))
    if (unlikely(b>0))
      g();
}

We expect gcc compiler can exchange the if conditions to be like below,

  if (unlikely(b>0))
    if (likely(a>0))
      g();

This way, performance can be improved due to saving the comparison for a>0.

At the moment, gcc generate code as below,

.LFB0:
        .cfi_startproc
        movl    a(%rip), %edx
        testl   %edx, %edx
        jle     .L1
        movl    b(%rip), %eax
        testl   %eax, %eax
        jg      .L4
.L1:
        ret

[Bug tree-optimization/99946] fail to exchange if conditions in terms of likely/unlikely probability

[jiangning.liu at amperecomputing dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99946

--- Comment #1 from Jiangning Liu <jiangning.liu at amperecomputing dot com> ---
Is there any gcc pass that can deal with this simple optimization?

[Bug tree-optimization/99946] fail to exchange if conditions in terms of likely/unlikely probability

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99946

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
   Last reconfirmed|                            |2021-04-07
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
There's no pass doing all of this.  There's ifcombine that could end up
turning it into

 tem = likely(a>0) & unlikely(b>0);
 if (tem)
   g ();

and then RTL expansion might expand it to branchy code again.

But of course __builtin_expect is long gone then and we'll lose the
likelyness of the individual tests.

Still ifcombine has analysis logic as to "collecting" a complex predicate,
what's then missing is to "re-associate" the tests according to probability.

Which means implementing it in ifcombine should be feasible.