New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Slowdown in integer loops #15975
Comments
From sschoeling@linet-services.deCreated by sschoeling@linet-services.deThe following snippet loses about 20% speed between 5.25.9 and 5.26.0RC1 use integer; 5.25.9: 3.120s For some reason the slowdown is less without use integer: 5.25.9: 3.200s Bisect on the same repo and config.sh as this perlbug says: commit f49e846 S_do_op_dump_bar(): don't print TRANS op_pv field My recent commit v5.25.9-32-gabd07ec made dump.c display the op_pv The simple solution is to only display the address but not contents Perl Info
|
From @jkeenanOn Fri, 12 May 2017 12:45:00 GMT, sschoeling@linet-services.de wrote:
Focusing first on the variant without 'use integer;' ... While I get a slowdown from 5.24.1 to HEAD of blead, the slowdown is not as great as yours and I don't find evidence that the slowdown is caused (solely) by the commit you identified. Please see attachments. Thank you very much. -- |
From @jkeenanperl-5.24.1 real 0m2.773s $ git clean -dfx real 0m2.790s $ git clean -dfx real 0m2.843s $ git clean -dfx real 0m2.854s $ git clean -dfx real 0m2.817s |
The RT System itself - Status changed from 'new' to 'open' |
From @jkeenanOn Sat, 13 May 2017 01:55:59 GMT, jkeenan wrote:
In addition, these results may be very platform-specific. Please find attached results from running the same program in a FreeBSD-10.3 VM. There was significant performance improvement between 5.20.3 and 5.24.1 and a slight (probably not significant) improvement between 5.24.1 and HEAD. Thank you very much. |
From @jkeenan# FreeBSD-10.3-RELEASE # vendor perl $ git checkout v5.24.1 $ ./perl -Ilib -v | head -2 | tail -1 |
From @jkeenanOn Sat, 13 May 2017 02:11:42 GMT, jkeenan wrote:
Correction: slight decline in performance between 5.24.1 and HEAD. -- |
From @iabynOn Fri, May 12, 2017 at 07:11:43PM -0700, James E Keenan via RT wrote:
I think you are just seeing noise. That commit is a small change to debugging code which is never called by I ran your program under 'perf stat -r 10' and captured the total CPU with use integer: 5.24.0 11,154,909,863 cycles without use integer: 5.24.0 11,574,608,582 cycles For me, the headline figure would be that without 'use integer', RC1 is -- |
From s.schoeling@linet-services.deOn Sat, 13 May 2017 11:01:41 -0700
May very well be, yes. Normally I'd take more time to investigate, but since it's RC already, I went for report first. This machine has a comparatively old gcc and that has bitten me before. -- |
From @xsawyerxOn 05/15/2017 10:29 AM, Sven Schöling wrote:
No worries. Thank you for reporting! I believe we can close this ticket, then? |
From s.schoeling@linet-services.de
From my point of view: yes. -- |
@xsawyerx - Status changed from 'open' to 'rejected' |
Migrated from rt.perl.org#131291 (status was 'rejected')
Searchable as RT131291$
The text was updated successfully, but these errors were encountered: