http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #31 from Steven Bosscher 2012-08-08
06:28:16 UTC ---
Author: steven
Date: Wed Aug 8 06:28:10 2012
New Revision: 190222
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=190222
Log:
PR middle-end/54146
* ifcvt.c: Include
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #31 from Steven Bosscher 2012-08-08
06:28:16 UTC ---
Author: steven
Date: Wed Aug 8 06:28:10 2012
New Revision: 190222
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=190222
Log:
PR middle-end/54146
* ifcvt.c: Include
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #30 from Steven Bosscher 2012-08-07
22:36:30 UTC ---
> Created attachment 27957 [details]
With the attachment, time spent in rewrite_into_loop_closed_ssa is almost 0
(and that includes the time in the verifier). Compile time for the
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #29 from Steven Bosscher 2012-08-07
22:28:11 UTC ---
Created attachment 27957
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27957
Do not traverse sibling loops
The idea here is to note that for a nested loop we know for sure th
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #28 from Steven Bosscher 2012-08-07
19:58:00 UTC ---
To illustrate the rewrite_into_closed_loop_ssa problem, consider this test
case:
extern void use1 (int);
extern void use2 (int);
extern int confuse_loop (void);
void
foo (void)
{
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
Richard Guenther changed:
What|Removed |Added
CC||jamborm at gcc dot gnu.org
--- Comment
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #26 from Steven Bosscher 2012-08-06
22:58:22 UTC ---
(In reply to comment #25)
> 185939 is the number of basic blocks that end up in livein. That is a bitmap,
> so most time is spent in traversing bitmap linked lists.
Oh, and this do
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #25 from Steven Bosscher 2012-08-06
22:42:12 UTC ---
(In reply to comment #21)
> The tree loop optimizers spend 285s out of 1360s total compile time (with my
> flatten hack and ifcvt patch applied) in compute_global_livein. That's 21%
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #24 from Steven Bosscher 2012-08-06
20:55:37 UTC ---
(In reply to comment #23)
> Created attachment 27953 [details]
Needs this extra bit:
diff -u ira.c ira.c
--- ira.c (working copy)
+++ ira.c (working copy)
@@ -3539,7 +
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #23 from Steven Bosscher 2012-08-06
20:22:02 UTC ---
Created attachment 27953
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27953
Be memory friendlier in build_insn_chain
My first ever reload patch! :-)
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #22 from Steven Bosscher 2012-08-06
19:36:35 UTC ---
IRA/reload spends a rather significant amount of time here:
FOR_EACH_BB_REVERSE (bb)
{
bitmap_iterator bi;
rtx insn;
CLEAR_REG_SET (live_relevant_regs);
-
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #21 from Steven Bosscher 2012-08-06
15:36:04 UTC ---
(In reply to comment #20)
Yay, it's always nice to be right the first time when diagnosing a problem.
The tree loop optimizers spend 285s out of 1360s total compile time (with my
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #20 from stevenb.gcc at gmail dot com 2012-08-06 09:09:02 UTC ---
On Mon, Aug 6, 2012 at 10:45 AM, rguenth at gcc dot gnu.org
wrote:
> Ick, I suppose similar issues exist on the tree level for passes that
> think that memory / compile
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #19 from Richard Guenther 2012-08-06
08:45:12 UTC ---
(In reply to comment #17)
> (In reply to comment #14)
> > if-conversion : 177.26 (but due to loop_optimizer_init)
>
> Hmm, this is not loop_optimizer_init. All time i
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #18 from Steven Bosscher 2012-08-05
21:13:26 UTC ---
Created attachment 27946
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27946
Speed up ifcvt.c:check_cond_move_block
This speeds up the pre-reload if-conversion passes by usin
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #17 from Steven Bosscher 2012-08-05
18:48:55 UTC ---
(In reply to comment #14)
> if-conversion : 177.26 (but due to loop_optimizer_init)
Hmm, this is not loop_optimizer_init. All time is spent in the two memset calls
in
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
Steven Bosscher changed:
What|Removed |Added
Attachment #27917|0 |1
is obsolete|
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #15 from Michael Matz 2012-08-03 14:43:14
UTC ---
Author: matz
Date: Fri Aug 3 14:43:09 2012
New Revision: 190126
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=190126
Log:
PR tree-optimization/54146
* cfgexpand.c (ad
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #14 from Steven Bosscher 2012-08-03
08:59:29 UTC ---
Created attachment 27930
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27930
Do not inline_merge_summary if called via flatten_function
As I noted in comment #12, the flatten
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
--- Comment #13 from rguenther at suse dot de
2012-08-02 10:35:56 UTC ---
On Thu, 2 Aug 2012, steven at gcc dot gnu.org wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
>
> Steven Bosscher changed:
>
>What|Removed
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54146
Steven Bosscher changed:
What|Removed |Added
Component|middle-end |tree-optimization
Summary|Ve
21 matches
Mail list logo
