This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
regression in ra ?
- From: Tommy Vercetti <vercetti at zlew dot org>
- To: gcc at gcc dot gnu dot org
- Date: Sat, 12 Feb 2005 13:23:07 +0100
- Subject: regression in ra ?
- Gj-laptop: yes
- Organization: ZLEFF
Hi
I've found small issue in ra probably. Maybe there's bug filled out for it
already, but I can't find it.
For simple loop like that:
for( unsigned int i=0;i<synapsy;i++ ) {
wagi[i] = 0;
}
on p3 I get:
.L40:
xorl %ecx, %ecx
movl $0, (%eax,%edx,8)
movl %ecx, 4(%eax,%edx,8)
incl %edx
cmpl %edx, %ebx
jne .L40
question is why is ecx zeroed (xored) on every iteration ?
Full code is at:
http://viewcvs.pointblue.com.pl/index.cgi/*checkout*/gj/neurony/neuron.cpp
lines 43-45.
Thanks.
--
Vercetti