[#58149] [ruby-trunk - Feature #9076][Open] New one-argument block syntax: &. — "asterite (Ary Borenszweig)" <ary@...>

23 messages 2013/11/04

[#58176] [ruby-trunk - Bug #9082][Open] popen3 hangs when stderr gets lots of output — "rosenfeld (Rodrigo Rosenfeld Rosas)" <rr.rosas@...>

15 messages 2013/11/05

[#58207] [ruby-trunk - Bug #9089][Open] rb_fix2uint no longer raises a RangeError when given negative values — "NoKarma (Arthur Schreiber)" <schreiber.arthur@...>

9 messages 2013/11/06

[#58243] [ruby-trunk - Feature #9098][Open] Indent heredoc against the left margin by default when "indented closing identifier" is turned on. — "sikachu (Prem Sichanugrist)" <s@...>

24 messages 2013/11/09

[#58306] [ruby-trunk - Bug #9106][Open] 'gem install' doesn't copy .so files of ext libs — "tagomoris (Satoshi TAGOMORI)" <tagomoris@...>

15 messages 2013/11/13

[#58324] [ruby-trunk - Feature #9108][Open] Hash sub-selections — "wardrop (Tom Wardrop)" <tom@...>

28 messages 2013/11/14

[#58342] [ruby-trunk - Feature #9112][Open] Make module lookup more dynamic (Including modules into a module after it has already been included) — "PragTob (Tobias Pfeiffer)" <pragtob@...>

16 messages 2013/11/14

[#58350] [ruby-trunk - Feature #9113][Open] Ship Ruby for Linux with jemalloc out-of-the-box — "sam.saffron (Sam Saffron)" <sam.saffron@...>

59 messages 2013/11/15

[#58374] [ruby-trunk - Bug #9115][Open] Logger traps all exceptions; breaks Timeout — "cphoenix (Chris Phoenix)" <cphoenix@...>

10 messages 2013/11/16

[#58375] [ruby-trunk - Feature #9116][Open] String#rsplit missing — "artagnon (Ramkumar Ramachandra)" <artagnon@...>

12 messages 2013/11/16

[#58396] [ruby-trunk - Bug #9121][Open] [PATCH] Remove rbtree implementation of SortedSet due to performance regression — "xshay (Xavier Shay)" <contact@...>

15 messages 2013/11/18

[#58404] [ruby-trunk - Feature #9123][Open] Make Numeric#nonzero? behavior consistent with Numeric#zero? — "sferik (Erik Michaels-Ober)" <sferik@...>

40 messages 2013/11/18

[#58411] [ruby-trunk - Bug #9124][Open] TestSocket errors in test-all on Arch 64-bit — "jonforums (Jon Forums)" <redmine@...>

14 messages 2013/11/18

[#58438] [ruby-trunk - Bug #9129][Open] Regression in support for IPv6 literals in URIs with Net::HTTP — "kallistec (Daniel DeLeo)" <dan@...>

11 messages 2013/11/19

[#58545] [ruby-trunk - Feature #9145][Open] Queue#pop(true) return nil if empty instead of raising ThreadError — "jsc (Justin Collins)" <redmine@...>

9 messages 2013/11/24

[#58653] [ruby-trunk - Bug #9170][Open] Math.sqrt returns different types when mathn is included; breaks various gems - this bug can be reproduced in Ruby 1.8 as well — "kranzky (Jason Hutchens)" <JasonHutchens@...>

7 messages 2013/11/28

[ruby-core:58476] [ruby-trunk - Feature #9113] Ship Ruby for Linux with jemalloc out-of-the-box

From: "kosaki (Motohiro KOSAKI)" <kosaki.motohiro@...>
Date: 2013-11-22 02:40:11 UTC
List: ruby-core #58476
Issue #9113 has been updated by kosaki (Motohiro KOSAKI).


@duerst It is not correct. I and glibc folks are working on several improvement about malloc. Moreover, each allocator has different pros/cons. jemalloc can retrieve some workload better and glibc allocator can retrieve some other workload. There is no single perfect allocator. That's our difficulty. 

@sam.saffron The Facebook's page you pointed out is out of date. It compare glibc 2.5 vs jemalloc 2.1.0. But latest are glibc 2.18 and jemalloc 3.4.1. And, glibc malloc and jemalloc shares a lot of basic design. So, this documentation is completely useless. If
you have several workload which glibc doesn't work well, please make and share benchmark instead of rumor. Then, we can improve several bottlenecks.
----------------------------------------
Feature #9113: Ship Ruby for Linux with jemalloc out-of-the-box
https://bugs.ruby-lang.org/issues/9113#change-43069

Author: sam.saffron (Sam Saffron)
Status: Rejected
Priority: Normal
Assignee: 
Category: build
Target version: 


libc's malloc is a problem, it fragments badly meaning forks share less memory and is slow compared to tcmalloc or jemalloc. 

both jemalloc and tcmalloc are heavily battle tested and stable. 

2 years ago redis picked up the jemalloc dependency see: http://oldblog.antirez.com/post/everything-about-redis-24.html 

To quote antirez:
``
But an allocator is a serious thing. Since we introduced the specially encoded data types Redis started suffering from fragmentation. We tried different things to fix the problem, but basically the Linux default allocator in glibc sucks really, really hard. 
``

--- 

I recently benched Discourse with tcmalloc / jemalloc and default and noticed 2 very important thing: 

median request time reduce by up to 10% (under both)
PSS (proportional share size) is reduced by 10% under jemalloc and 8% under tcmalloc.

We can always use LD_PRELOAD to yank these in, but my concern is that standard distributions are using a far from optimal memory allocator. It would be awesome if the build, out-of-the-box, just checked if it was on Linux  (eg: https://github.com/antirez/redis/blob/unstable/src/Makefile#L30-L34 ) and then used jemalloc instead. 


-- 
http://bugs.ruby-lang.org/

In This Thread