This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [patch] malloc per-thread cache ready for review
- From: Wilco Dijkstra <Wilco dot Dijkstra at arm dot com>
- To: Siddhesh Poyarekar <siddhesh at gotplt dot org>, Markus Trippelsdorf <markus at trippelsdorf dot de>, Carlos O'Donell <carlos at redhat dot com>
- Cc: "dj at redhat dot com" <dj at redhat dot com>, nd <nd at arm dot com>, "libc-alpha at sourceware dot org" <libc-alpha at sourceware dot org>
- Date: Thu, 2 Feb 2017 11:53:08 +0000
- Subject: Re: [patch] malloc per-thread cache ready for review
- Authentication-results: sourceware.org; auth=none
- Authentication-results: spf=none (sender IP is ) smtp.mailfrom=Wilco dot Dijkstra at arm dot com;
- Nodisclaimer: True
- Spamdiagnosticmetadata: NSPM
- Spamdiagnosticoutput: 1:99
Siddhesh Poyarekar wrote:
> DJ's patch does not only impact multi-threaded workloads positively,
> (although I assume it will likely improve those workloads by a great
> deal) it shortens the hot path quite considerably and possibly (a
> hand-waving guess) also has a positive on icache usage because in my
> testing (multiple runs on multiple systems because I couldn't believe it
> at first) I saw a significant improvement in SPEC2006 numbers, often
> more than the amount of time spent in malloc in those tests.
I tried the new malloc (the patch as posted with several fixes from DJ's branch),
however I don't see any improvement on SPEC. That's odd since several subtests
spend ~10% of the time in malloc, and even with a single thread the locking is very
expensive, so bypassing it should give a good speedup. And yes having written
several fast/realtime mallocs in a few hundred lines, I also don't get why the GLIBC
one needs over 5000 lines...
Or is there some other config trick needed to turn it on? (I see the new fast path
in the binaries so it is built correctly).
Wilco