This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug rtl-optimization/48156] [4.6/4.7 Regression] wrong code with -fcrossjumping
- From: "jakub at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Fri, 18 Mar 2011 10:50:48 +0000
- Subject: [Bug rtl-optimization/48156] [4.6/4.7 Regression] wrong code with -fcrossjumping
- Auto-submitted: auto-generated
- References: <bug-48156-4@http.gcc.gnu.org/bugzilla/>
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48156
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-03-18 10:50:29 UTC ---
The problem is not in %rdx, but %rsi. It was originally initialized from %rbx
right before the call, but now is the %rsi assignment moved before the testing
of the j && equals (i, j) condition, which is wrong, because %rsi is call
clobbered and thus potentionally clobbered by the call, and more importantly
even set right before the call to equals function.