I'm about to backport one patch (a fix for PR48189) to 4.6 branch.
Bootstrapped/regtested on x86_64-linux, acked by Jakub on IRC.
Will commit soon.
ChangeLog:
2013-04-04 Marek Polacek <pola...@redhat.com>
Backported from mainline
2013-01-09 Steven Bosscher <ste...@gcc.gnu.org>
Jakub Jelinek <ja...@redhat.com>
PR tree-optimization/48189
* predict.c (predict_loops): If max is 0, don't call compare_tree_int.
If nitercst is 0, don't predict the exit edge.
testsuite/ChangeLog:
2013-04-04 Marek Polacek <pola...@redhat.com>
Backported from mainline
2013-01-09 Jakub Jelinek <ja...@redhat.com>
PR tree-optimization/48189
* gcc.dg/pr48189.c: New test.
--- gcc/predict.c.mp 2013-04-04 11:59:59.816219085 +0200
+++ gcc/predict.c 2013-04-04 12:00:13.418263854 +0200
@@ -1,6 +1,6 @@
/* Branch prediction routines for the GNU compiler.
Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2007, 2008, 2009, 2010
- Free Software Foundation, Inc.
+ 2011, 2012, 2013 Free Software Foundation, Inc.
This file is part of GCC.
@@ -966,7 +966,8 @@ predict_loops (void)
if (TREE_CODE (niter) == INTEGER_CST)
{
if (host_integerp (niter, 1)
- && compare_tree_int (niter, max-1) == -1)
+ && max
+ && compare_tree_int (niter, max - 1) == -1)
nitercst = tree_low_cst (niter, 1) + 1;
else
nitercst = max;
@@ -988,6 +989,11 @@ predict_loops (void)
else
continue;
+ /* If the prediction for number of iterations is zero, do not
+ predict the exit edges. */
+ if (nitercst == 0)
+ continue;
+
probability = ((REG_BR_PROB_BASE + nitercst / 2) / nitercst);
predict_edge (ex, predictor, probability);
}
--- gcc/testsuite/gcc.dg/pr48189.c.mp 2013-04-04 11:59:45.581173690 +0200
+++ gcc/testsuite/gcc.dg/pr48189.c 2013-04-04 11:59:36.446144550 +0200
@@ -0,0 +1,13 @@
+/* PR tree-optimization/48189 */
+/* { dg-do compile } */
+/* { dg-options "-O --param max-predicted-iterations=0" } */
+
+struct S { int s[8]; };
+
+void
+foo (int *x, struct S *y)
+{
+ int i;
+ for (i = 0; y[i].s[i]; i++)
+ *x++ = y[i].s[i];
+}
Marek
