[Bug 236233] Python multiprocessing is 4x slower with 11.2 compared to 11.0 kernel

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Wed Mar 6 10:09:16 UTC 2019


https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=236233

Alex Richardson <arichardson at FreeBSD.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|New                         |Open
            Summary|Python multiprocessing is   |Python multiprocessing is
                   |20x slower with 11.2        |4x slower with 11.2
                   |compared to 11.0 kernel     |compared to 11.0 kernel

--- Comment #2 from Alex Richardson <arichardson at FreeBSD.org> ---
I finished bisecting and the following commit is causing the slowdown:

6dd025b40ee6870bea6ba670f30dcf684edc3f6c is the first bad commit
commit 6dd025b40ee6870bea6ba670f30dcf684edc3f6c
Author: kib <kib at FreeBSD.org>
Date:   Sat Feb 17 18:00:01 2018 +0000

    MFC r328083,328096,328116,328119,328120,328128,328135,328153,328157,
    328166,328177,328199,328202,328205,328468,328470,328624,328625,328627,
    328628,329214,329297,329365:

    Meltdown mitigation by PTI, PCID optimization of PTI, and kernel use of
IBRS
    for some mitigations of Spectre.

    Tested by:      emaste, Arshan Khanifar <arshankhanifar at gmail.com>
    Discussed with: jkim
    Sponsored by:   The FreeBSD Foundation



I also checked that ibrs was off so this seems to be caused by something else
in this massive commit.
$ sysctl hw.ibrs_active
hw.ibrs_active: 0
$ sysctl hw.ibrs_disable
hw.ibrs_disable: 1


The time it took to run the test command was very consistent:
Working commit: r313908+bc19f9718e7(stable/11)  -> 26.8seconds
Known bad commit: a190b45c053(stable/11) -> 121.12s

1st step: fd638422ccfbbe7a11e751ed5a7a237f2dc8d004 -> 133.14s
2nd step: 4ebdf0a463e767672045047ec82c75bf545e9a7c -> 25.48s
3rd step: 483c180d2dece828b1800896e642bcb806075c1f -> 133.49s
4th step: e38b1b844d045fb5dcf4db6f2795f7797022ca17 -> 24.97s
5th step (141 to go): 97ca83c5b618a025fdac8e62db23afa7abd198c8 -> 25.27s
6th step (70 to go): cdf8c6144ce326759989ebcbc0d9b5195bd4f811 -> 133.64s
7th step (35 to go): e2f594b309b8fd60b6ab39982b9ba0f6da798f38 -> 134.01s
8th step (17 to go): 195e563d5c9edb667d92e2480a07080ae6ba40ac -> 24.80s
9th step (8 to go): 61669078139eef2cb35756674ab6e925e97ab5ba -> 134.34s
10th step (4 to go): af7d202e1dcdeda020ee1fdb96e2b1eb4dbbad39 -> 133.04s
11th step (1 to go): 6dd025b40ee6870bea6ba670f30dcf684edc3f6c -> 133.92s
12th step (0 to go): ebe022b0c9f14dc22edbe0308795d42c9f93fd39 -> 24.86s

-- 
You are receiving this mail because:
You are the assignee for the bug.


More information about the freebsd-bugs mailing list