re PR tree-optimization/52943 (likely wrong code bug caused by predictive commoning)
authorRichard Guenther <rguenther@suse.de>
Thu, 12 Apr 2012 11:38:47 +0000 (11:38 +0000)
committerRichard Biener <rguenth@gcc.gnu.org>
Thu, 12 Apr 2012 11:38:47 +0000 (11:38 +0000)
commit5f1fab5819256a0a30779d8d0c2fffc3ed2ee9c2
tree309556804f1e89c251beb3a6c838fd3141487079
parente5b332cdb894baa96b7dc83b69140b3106fd5153
re PR tree-optimization/52943 (likely wrong code bug caused by predictive commoning)

2012-04-12  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/52943
* tree-chrec.h (chrec_is_positive): Remove.
* tree-scalar-evolution.c (chrec_is_positive): Move ...
* tree-data-ref.c (chrec_is_positive): ... here.  Make static.
Return false for a constant zero instead of negative.
(analyze_siv_subscript_cst_affine): Handle zero difference
in the initial condition explicitely.

* gcc.dg/torture/pr52943.c: New testcase.

From-SVN: r186374
gcc/ChangeLog
gcc/testsuite/ChangeLog
gcc/testsuite/gcc.dg/torture/pr52943.c [new file with mode: 0644]
gcc/tree-chrec.h
gcc/tree-data-ref.c
gcc/tree-scalar-evolution.c