On Wed, 30 May 2018, Jan Hubicka wrote:
> >
> > This makes tracer not explode with -fno-guess-branch-probabilities.
> > I've settled with find_best_successor/predecessor not returning
> > anything if _any_ edge in the interesting direction doesn't have
> > ->count () initialized (rather than ignoring such edges).
> >
> > Honza - I suppose it is on purpose that functions like
> > .to_frequency () do not ICE for uninitialized counters?
> > It at least looks like "previous" behavior was more sane
> > for tracer in the counts/frequencies that were exposed.
> >
> > Bootstrapped and tested on x86_64-unknown-linux-gnu.
> >
> > Honza, does this look OK to you?
> >
> > tracer going wild on this testcase exposes the CFG cleanup
> > scalability issue I've posted the following RFC for:
> > https://gcc.gnu.org/ml/gcc-patches/2018-05/msg01644.html
> >
> > Thanks,
> > Richard.
> >
> > 2018-05-30 Richard Biener <rguent...@suse.de>
> >
> > PR tree-optimization/85964
> > * tracer.c (better_p): Drop initialized count check, we only
> > call the function with initialized counts now.
> > (find_best_successor): Do find a best edge if one
> > has uninitialized count.
> > (find_best_predecessor): Likewise. Do BB frequency check only
> > if count is initialized.
>
> >
> > Index: gcc/tracer.c
> > ===================================================================
> > --- gcc/tracer.c (revision 260896)
> > +++ gcc/tracer.c (working copy)
> > @@ -132,8 +132,7 @@ count_insns (basic_block bb)
> > static bool
> > better_p (const_edge e1, const_edge e2)
> > {
> > - if (e1->count ().initialized_p () && e2->count ().initialized_p ()
> > - && ((e1->count () > e2->count ()) || (e1->count () < e2->count ())))
> > + if ((e1->count () > e2->count ()) || (e1->count () < e2->count ()))
> > return e1->count () > e2->count ();
> > /* This is needed to avoid changes in the decision after
> > CFG is modified. */
> > @@ -152,12 +151,15 @@ find_best_successor (basic_block bb)
> > edge_iterator ei;
> >
> > FOR_EACH_EDGE (e, ei, bb->succs)
> > - if (!best || better_p (e, best))
> > - best = e;
> > + {
> > + if (!e->count ().initialized_p ())
> > + return NULL;
> > + if (!best || better_p (e, best))
> > + best = e;
> > + }
> > if (!best || ignore_bb_p (best->dest))
> > return NULL;
> > - if (best->probability.initialized_p ()
> > - && best->probability.to_reg_br_prob_base () <= probability_cutoff)
> > + if (best->probability.to_reg_br_prob_base () <= probability_cutoff)
>
> Technically we could accept when one edge has large known probability and
> other unknown,
> but in practice it won't matter because w/o profile guessing tracer is
> useless anyway.
So the above hunk requires the fix below.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
2018-06-04 Richard Biener <rguent...@suse.de>
PR tree-optimization/86038
* tracer.c (find_best_successor): Check probability for
being initialized, bail out if not.
* gcc.dg/pr86038.c: New testcase.
Index: gcc/tracer.c
===================================================================
--- gcc/tracer.c (revision 261136)
+++ gcc/tracer.c (working copy)
@@ -159,7 +159,8 @@ find_best_successor (basic_block bb)
}
if (!best || ignore_bb_p (best->dest))
return NULL;
- if (best->probability.to_reg_br_prob_base () <= probability_cutoff)
+ if (!best->probability.initialized_p ()
+ || best->probability.to_reg_br_prob_base () <= probability_cutoff)
return NULL;
return best;
}
Index: gcc/testsuite/gcc.dg/pr86038.c
===================================================================
--- gcc/testsuite/gcc.dg/pr86038.c (nonexistent)
+++ gcc/testsuite/gcc.dg/pr86038.c (working copy)
@@ -0,0 +1,12 @@
+/* { dg-do compile } */
+/* { dg-require-effective-target pthread } */
+/* { dg-options "-O2 -ftracer -ftree-parallelize-loops=2 -fno-tree-scev-cprop
--param parloops-schedule=dynamic" } */
+
+int
+sd (int lw)
+{
+ while (lw < 1)
+ ++lw;
+
+ return lw;
+}
