Re: [PATCH] Fix PR24574

[Martin Liška <mliska at suse dot cz>]

On 07/13/2016 04:15 PM, Richard Biener wrote:
> so not exactly a 50/50 prediction.  It is predicted by early return
> predictor it seems.  If I make the predictor not apply we predict
> it as 50/50 chance.  Hmm.  Honza?

Yes, following test-case eliminates the early return predictor:

int foo(int x, int *v)
{
  if (x != 0)
    x = x / 10;

   if (v)
     *v += 123;

   return x;
}

;;   basic block 2, loop depth 0, count 0, freq 10000, maybe hot
;;    prev block 0, next block 3, flags: (NEW, REACHABLE)
;;    pred:       ENTRY [100.0%]  (FALLTHRU,EXECUTABLE)
  if (x_5(D) != 0)
    goto <bb 3>;
  else
    goto <bb 4>;
;;    succ:       3 [50.0%]  (TRUE_VALUE,EXECUTABLE)
;;                4 [50.0%]  (FALSE_VALUE,EXECUTABLE)

;;   basic block 3, loop depth 0, count 0, freq 5000, maybe hot
;;    prev block 2, next block 4, flags: (NEW, REACHABLE)
;;    pred:       2 [50.0%]  (TRUE_VALUE,EXECUTABLE)
  x_6 = x_5(D) / 10;

and there's not predictor:

Predictions for bb 2
  no prediction heuristics: 50.0%
  combined heuristics: 50.0%

Martin