--- Comment #10 from paolo dot carlini at oracle dot com 2009-10-20 22:15
---
I meant info about __restrict__ of course.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41763
--- Comment #9 from paolo dot carlini at oracle dot com 2009-10-20 22:14
---
Fixed, anyway.
--
paolo dot carlini at oracle dot com changed:
What|Removed |Added
--- Comment #8 from paolo dot carlini at oracle dot com 2009-10-20 22:14
---
Ok, thanks Jakub. By the way, I was looking for some info about export, beyond
C99 and the GCC specifics, and found docs about the IBM compiler saying that in
case of pointers to const, it is still safe to use
--- Comment #7 from paolo at gcc dot gnu dot org 2009-10-20 21:21 ---
Subject: Bug 41763
Author: paolo
Date: Tue Oct 20 21:21:11 2009
New Revision: 153039
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=153039
Log:
2009-10-20 Paolo Carlini
PR libstdc++/41763
*
--- Comment #6 from jakub at gcc dot gnu dot org 2009-10-20 21:15 ---
Not compared with other arguments to be precise. It is fine to have
void foo (int *__restrict__ p, int *__restrict__ q)
{
int *r = p + 10;
while (p != r)
*p++ = *q++;
}
Although p is compared here, it is compa
--- Comment #5 from paolo dot carlini at oracle dot com 2009-10-20 18:12
---
Ok, thanks, now I have a better picture. As a matter of fact, in *most* of the
cases we should be *only* comparing pointers to the same object, I think this
is the only case guaranteed to behave sanely under th
--- Comment #4 from jakub at gcc dot gnu dot org 2009-10-20 17:48 ---
Well, make sure we don't use it when two different pointers point into the same
object. As in this example, both __b and __e are begin and end pointers within
the same object or one byte after the end of it. And __re
--- Comment #3 from paolo dot carlini at oracle dot com 2009-10-20 17:28
---
Thanks Gaby. Thus, if I understand the issue correctly, we have to be less
aggressive about __restrict__ and make sure we don't use it when we compare
pointers to different objects. I can work on the change.
--- Comment #2 from gdr at cs dot tamu dot edu 2009-10-20 16:42 ---
Subject: Re: valarray_array.h seems to overuse __restrict__
"paolo dot carlini at oracle dot com" writes:
| I think this line, in general all the uses of __restrict__ in valarray are
| *very* old... Let's CC Gaby, in
--- Comment #1 from paolo dot carlini at oracle dot com 2009-10-20 16:32
---
I think this line, in general all the uses of __restrict__ in valarray are
*very* old... Let's CC Gaby, in case he has some comments.
--
paolo dot carlini at oracle dot com changed:
What|Rem
10 matches
Mail list logo
