[Bug middle-end/43340] miscompiled code at -O2

2010-03-12 Thread rguenth at gcc dot gnu dot org
--- Comment #9 from rguenth at gcc dot gnu dot org 2010-03-12 14:20 --- (In reply to comment #8) > looks like this rather is a bug in CP2K. fi_mat and fj_mat can alias... So, > I'll close this bug, but it is worthwhile to note that -fno-strict-aliasing > did > not 'fix' this problem.

[Bug middle-end/43340] miscompiled code at -O2

2010-03-12 Thread jv244 at cam dot ac dot uk
--- Comment #8 from jv244 at cam dot ac dot uk 2010-03-12 12:35 --- looks like this rather is a bug in CP2K. fi_mat and fj_mat can alias... So, I'll close this bug, but it is worthwhile to note that -fno-strict-aliasing did not 'fix' this problem. -- jv244 at cam dot ac dot uk chang

[Bug middle-end/43340] miscompiled code at -O2

2010-03-12 Thread jv244 at cam dot ac dot uk
--- Comment #7 from jv244 at cam dot ac dot uk 2010-03-12 12:01 --- > ? You can also try -fno-ipa-cp and/or -fno-strict-overflow -O2 -fno-inline -fno-tree-vrp -fno-tree-pre -fno-strict-aliasing -fno-ipa-cp BUG -O2 -fno-inline -fno-tree-vrp -fno-tree-pre -fno-strict-aliasing -fno-ipa-cp

[Bug middle-end/43340] miscompiled code at -O2

2010-03-12 Thread rguenth at gcc dot gnu dot org
--- Comment #6 from rguenth at gcc dot gnu dot org 2010-03-12 11:46 --- -O2 -fno-inline -fno-tree-vrp -fno-tree-pre -fno-strict-aliasing ? You can also try -fno-ipa-cp and/or -fno-strict-overflow -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43340

[Bug middle-end/43340] miscompiled code at -O2

2010-03-12 Thread jv244 at cam dot ac dot uk
--- Comment #5 from jv244 at cam dot ac dot uk 2010-03-12 11:41 --- (In reply to comment #4) > Candidates to try for -O1 vs. -O2 are -f[no-]tree-vrp, -f[no-]tree-pre, > -f[no-]strict-aliasing. You can also rule out inlining effects by > -fno-inline. no luck :-( -O1 OK -O2 BUG -O2 -fno

[Bug middle-end/43340] miscompiled code at -O2

2010-03-12 Thread rguenth at gcc dot gnu dot org
--- Comment #4 from rguenth at gcc dot gnu dot org 2010-03-12 11:20 --- Candidates to try for -O1 vs. -O2 are -f[no-]tree-vrp, -f[no-]tree-pre, -f[no-]strict-aliasing. You can also rule out inlining effects by -fno-inline. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43340

[Bug middle-end/43340] miscompiled code at -O2

2010-03-12 Thread jv244 at cam dot ac dot uk
--- Comment #3 from jv244 at cam dot ac dot uk 2010-03-12 09:00 --- (In reply to comment #2) > What kind of dependencies? all derived types, defined in various modules, each with further dependencies, and the objects that are module variables. > The routine just calls one another rou

[Bug middle-end/43340] miscompiled code at -O2

2010-03-12 Thread jakub at gcc dot gnu dot org
--- Comment #2 from jakub at gcc dot gnu dot org 2010-03-12 08:40 --- What kind of dependencies? The routine just calls one another routinem so all you need is find out what arguments it is called with, write a small wrapper that calls it with those arguments, and write a dummy rotint t

[Bug middle-end/43340] miscompiled code at -O2

2010-03-12 Thread jv244 at cam dot ac dot uk
--- Comment #1 from jv244 at cam dot ac dot uk 2010-03-12 08:14 --- Created an attachment (id=20092) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20092&action=view) result of -fdump-tree-all results of -fdump-tree-all at -O1 and -O2 -- http://gcc.gnu.org/bugzilla/show_bug.cg