OpenE2K/gcc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

re PR tree-optimization/23821 (DOM and VRP creating harder to optimize code)

Browse Source 
2009-10-05  Richard Guenther  <rguenther@suse.de>

	PR tree-optimization/23821
	* tree-vrp.c (vrp_finalize): Do not perform copy propagation.
	* tree-ssa-dom.c (cprop_operand): Do not propagate copies into
	simple IV increments.

	* gcc.dg/torture/pr23821.c: New testcase.

From-SVN: r152449
This commit is contained in:
Richard Guenther 2009-10-05 13:18:09 +00:00 committed by Richard Biener
parent 916bd5f07f
commit e9d85fa6a7
5 changed files with 50 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
gcc/ChangeLog
View File

@ -1,3 +1,10 @@
2009-10-05 Richard Guenther <rguenther@suse.de>
PR tree-optimization/23821
* tree-vrp.c (vrp_finalize): Do not perform copy propagation.
* tree-ssa-dom.c (cprop_operand): Do not propagate copies into
simple IV increments.
2009-10-05 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
* config/arm/arm.c (arm_override_options): Really initialize

5
gcc/testsuite/ChangeLog
View File

@ -1,3 +1,8 @@
2009-10-05 Richard Guenther <rguenther@suse.de>
PR tree-optimization/23821
* gcc.dg/torture/pr23821.c: New testcase.
2009-10-05 Daniel Kraft <d@domob.eu>
PR fortran/41403

29
gcc/testsuite/gcc.dg/torture/pr23821.c Normal file
View File

@ -0,0 +1,29 @@
/* { dg-do compile } */
/* { dg-skip-if "" { *-*-* } { "-O0" } { "" } } */
/* At -O1 DOM threads a jump in a non-optimal way which leads to
the bogus propagation. */
/* { dg-skip-if "" { *-*-* } { "-O1" } { "" } } */
/* { dg-options "-fdump-tree-ivcanon-details" } */
static int a[199];
extern void abort (void);
int
main ()
{
int i, x;
for (i = 0; i < 199; i++)
{
x = a[i];
if (x != i)
abort ();
}
return 0;
}
/* Verify that we do not propagate the equivalence x == i into the
induction variable increment. */
/* { dg-final { scan-tree-dump "Added canonical iv" "ivcanon" } } */
/* { dg-final { cleanup-tree-dump "ivcanon" } } */

6
gcc/tree-ssa-dom.c
View File

@ -1998,6 +1998,12 @@ cprop_operand (gimple stmt, use_operand_p op_p)
if (loop_depth_of_name (val) > loop_depth_of_name (op))
return;
/* Do not propagate copies into simple IV increment statements.
See PR23821 for how this can disturb IV analysis. */
if (TREE_CODE (val) != INTEGER_CST
&& simple_iv_increment_p (stmt))
return;
/* Dump details. */
if (dump_file && (dump_flags & TDF_DETAILS))
{

5
gcc/tree-vrp.c
View File

@ -7237,7 +7237,7 @@ vrp_finalize (void)
}
/* We may have ended with ranges that have exactly one value. Those
values can be substituted as any other copy/const propagated
values can be substituted as any other const propagated
value using substitute_and_fold. */
single_val_range = XCNEWVEC (prop_value_t, num_ssa_names);
@ -7245,7 +7245,8 @@ vrp_finalize (void)
for (i = 0; i < num_ssa_names; i++)
if (vr_value[i]
&& vr_value[i]->type == VR_RANGE
&& vr_value[i]->min == vr_value[i]->max)
&& vr_value[i]->min == vr_value[i]->max
&& is_gimple_min_invariant (vr_value[i]->min))
{
single_val_range[i].value = vr_value[i]->min;
do_value_subst_p = true;