public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug middle-end/30442] Expanded array initialization can use memset builtin function
[not found] <bug-30442-4@http.gcc.gnu.org/bugzilla/>
@ 2012-02-06 21:41 ` ubizjak at gmail dot com
2012-06-05 10:50 ` rguenth at gcc dot gnu.org
` (2 subsequent siblings)
3 siblings, 0 replies; 8+ messages in thread
From: ubizjak at gmail dot com @ 2012-02-06 21:41 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30442
Uros Bizjak <ubizjak at gmail dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed|2007-01-14 05:01:04 |2012-02-06
--- Comment #6 from Uros Bizjak <ubizjak at gmail dot com> 2012-02-06 21:40:50 UTC ---
Reconfirmed with 4.7.
^ permalink raw reply [flat|nested] 8+ messages in thread
* [Bug middle-end/30442] Expanded array initialization can use memset builtin function
[not found] <bug-30442-4@http.gcc.gnu.org/bugzilla/>
2012-02-06 21:41 ` [Bug middle-end/30442] Expanded array initialization can use memset builtin function ubizjak at gmail dot com
@ 2012-06-05 10:50 ` rguenth at gcc dot gnu.org
2012-06-05 12:38 ` rguenth at gcc dot gnu.org
2012-06-05 12:39 ` rguenth at gcc dot gnu.org
3 siblings, 0 replies; 8+ messages in thread
From: rguenth at gcc dot gnu.org @ 2012-06-05 10:50 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30442
--- Comment #7 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-06-05 10:50:22 UTC ---
The
long long test2(void)
{
long long a[32];
int i;
for (i = 0; i < 32; i++)
a[i] = 0;
return foo(a);
}
loop is transformed to memset at -O3. The unrolled version is not "re-rolled"
still, and basic-block vectorization does not catch it because of the
call in the basic-block. I'm trying to fix that.
^ permalink raw reply [flat|nested] 8+ messages in thread
* [Bug middle-end/30442] Expanded array initialization can use memset builtin function
[not found] <bug-30442-4@http.gcc.gnu.org/bugzilla/>
2012-02-06 21:41 ` [Bug middle-end/30442] Expanded array initialization can use memset builtin function ubizjak at gmail dot com
2012-06-05 10:50 ` rguenth at gcc dot gnu.org
@ 2012-06-05 12:38 ` rguenth at gcc dot gnu.org
2012-06-05 12:39 ` rguenth at gcc dot gnu.org
3 siblings, 0 replies; 8+ messages in thread
From: rguenth at gcc dot gnu.org @ 2012-06-05 12:38 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30442
--- Comment #8 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-06-05 12:38:30 UTC ---
Author: rguenth
Date: Tue Jun 5 12:38:26 2012
New Revision: 188235
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=188235
Log:
2012-06-05 Richard Guenther <rguenther@suse.de>
PR tree-optimization/30442
* tree-vect-data-refs.c (vect_analyze_data_refs): For basic-block
vectorization stop analysis at the first stmt we cannot compute
a data-reference for instead of giving up completely.
* gcc.dg/vect/bb-slp-30.c: New testcase.
Added:
trunk/gcc/testsuite/gcc.dg/vect/bb-slp-30.c
Modified:
trunk/gcc/ChangeLog
trunk/gcc/testsuite/ChangeLog
trunk/gcc/tree-vect-data-refs.c
^ permalink raw reply [flat|nested] 8+ messages in thread
* [Bug middle-end/30442] Expanded array initialization can use memset builtin function
[not found] <bug-30442-4@http.gcc.gnu.org/bugzilla/>
` (2 preceding siblings ...)
2012-06-05 12:38 ` rguenth at gcc dot gnu.org
@ 2012-06-05 12:39 ` rguenth at gcc dot gnu.org
3 siblings, 0 replies; 8+ messages in thread
From: rguenth at gcc dot gnu.org @ 2012-06-05 12:39 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30442
--- Comment #9 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-06-05 12:39:29 UTC ---
So, $summary is still true but we now at least vectorize the initialization.
^ permalink raw reply [flat|nested] 8+ messages in thread
* [Bug middle-end/30442] Expanded array initialization can use memset builtin function
2007-01-11 19:53 [Bug c/30442] New: Array " ubizjak at gmail dot com
` (2 preceding siblings ...)
2008-03-12 11:06 ` ubizjak at gmail dot com
@ 2008-03-13 6:51 ` irar at il dot ibm dot com
3 siblings, 0 replies; 8+ messages in thread
From: irar at il dot ibm dot com @ 2008-03-13 6:51 UTC (permalink / raw)
To: gcc-bugs
------- Comment #5 from irar at il dot ibm dot com 2008-03-13 06:51 -------
(In reply to comment #4)
> This still happens on mainline.
>
> I wonder if vectorizer infrastructure can be re-used here to detect unrolled
> and looped version of memset. In addition to loop that can be "vectorized", we
> have something resembling "vectorization" of straight code.
This code can be vectorized with basic block SLP, which is not implemented yet
(currently only SLP in loops is implemented).
> And looking at comment #3, the rewards from real-world code look really
> promising.
>
However, in comment #3 there is a loop:
: MALLOC(visited, int, d->tindex);
48796 16.2440 : for (i = 0; i < d->tindex; ++i)
94442 31.4394 : visited[i] = 0;
There was an effort to replace such loops with calls to builtins -
http://gcc.gnu.org/ml/gcc-patches/2007-07/msg00054.html.
Ira
--
irar at il dot ibm dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |irar at il dot ibm dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30442
^ permalink raw reply [flat|nested] 8+ messages in thread
* [Bug middle-end/30442] Expanded array initialization can use memset builtin function
2007-01-11 19:53 [Bug c/30442] New: Array " ubizjak at gmail dot com
2007-01-14 5:01 ` [Bug middle-end/30442] Expanded array " pinskia at gcc dot gnu dot org
2007-11-09 13:09 ` johan dot walles at gmail dot com
@ 2008-03-12 11:06 ` ubizjak at gmail dot com
2008-03-13 6:51 ` irar at il dot ibm dot com
3 siblings, 0 replies; 8+ messages in thread
From: ubizjak at gmail dot com @ 2008-03-12 11:06 UTC (permalink / raw)
To: gcc-bugs
------- Comment #4 from ubizjak at gmail dot com 2008-03-12 11:05 -------
This still happens on mainline.
I wonder if vectorizer infrastructure can be re-used here to detect unrolled
and looped version of memset. In addition to loop that can be "vectorized", we
have something resembling "vectorization" of straight code.
And looking at comment #3, the rewards from real-world code look really
promising.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30442
^ permalink raw reply [flat|nested] 8+ messages in thread
* [Bug middle-end/30442] Expanded array initialization can use memset builtin function
2007-01-11 19:53 [Bug c/30442] New: Array " ubizjak at gmail dot com
2007-01-14 5:01 ` [Bug middle-end/30442] Expanded array " pinskia at gcc dot gnu dot org
@ 2007-11-09 13:09 ` johan dot walles at gmail dot com
2008-03-12 11:06 ` ubizjak at gmail dot com
2008-03-13 6:51 ` irar at il dot ibm dot com
3 siblings, 0 replies; 8+ messages in thread
From: johan dot walles at gmail dot com @ 2007-11-09 13:09 UTC (permalink / raw)
To: gcc-bugs
------- Comment #3 from johan dot walles at gmail dot com 2007-11-09 13:09 -------
This optimization would have made grep 2.5.3 30% faster in a real-world test
case:
http://bugs.debian.org/450649
--
johan dot walles at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |johan dot walles at gmail
| |dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30442
^ permalink raw reply [flat|nested] 8+ messages in thread
* [Bug middle-end/30442] Expanded array initialization can use memset builtin function
2007-01-11 19:53 [Bug c/30442] New: Array " ubizjak at gmail dot com
@ 2007-01-14 5:01 ` pinskia at gcc dot gnu dot org
2007-11-09 13:09 ` johan dot walles at gmail dot com
` (2 subsequent siblings)
3 siblings, 0 replies; 8+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2007-01-14 5:01 UTC (permalink / raw)
To: gcc-bugs
------- Comment #2 from pinskia at gcc dot gnu dot org 2007-01-14 05:01 -------
This is a bit complex, we could use a loop reroller to roll this into a loop
and then transform that loop into memset for test1.
For test2 it is simple and a patch was presented at last year's gcc summit (and
I posted an older version of that patch a year or two ago).
--
pinskia at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
Severity|normal |enhancement
Status|UNCONFIRMED |NEW
Ever Confirmed|0 |1
GCC build triplet|i686-pc-linux-gnu |
GCC host triplet|i686-pc-linux-gnu |
GCC target triplet|i686-pc-linux-gnu |
Keywords| |missed-optimization
Last reconfirmed|0000-00-00 00:00:00 |2007-01-14 05:01:04
date| |
Summary|Array initialization can use|Expanded array
|memset builtin function |initialization can use
| |memset builtin function
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30442
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2012-06-05 12:39 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <bug-30442-4@http.gcc.gnu.org/bugzilla/>
2012-02-06 21:41 ` [Bug middle-end/30442] Expanded array initialization can use memset builtin function ubizjak at gmail dot com
2012-06-05 10:50 ` rguenth at gcc dot gnu.org
2012-06-05 12:38 ` rguenth at gcc dot gnu.org
2012-06-05 12:39 ` rguenth at gcc dot gnu.org
2007-01-11 19:53 [Bug c/30442] New: Array " ubizjak at gmail dot com
2007-01-14 5:01 ` [Bug middle-end/30442] Expanded array " pinskia at gcc dot gnu dot org
2007-11-09 13:09 ` johan dot walles at gmail dot com
2008-03-12 11:06 ` ubizjak at gmail dot com
2008-03-13 6:51 ` irar at il dot ibm dot com
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).