[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [PATCH] Improve single byte read stream performance

From: Stefan Fuhrmann <stefanfuhrmann_at_alice-dsl.de>
Date: Mon, 15 Mar 2010 19:47:16 +0100

On Sunday 14 March 2010 22:53:59 Philip Martin wrote:
> Johan Corveleyn <jcorvel_at_gmail.com> writes:
> > On Tue, Mar 9, 2010 at 8:04 PM, Julian Foad <julian.foad_at_wandisco.com>
wrote:
> >> On Sun, 2010-03-07, Stefan Fuhrman wrote:
> >>> Speed up input stream processing in config parser and
> >>> others that read single bytes from a stream.
> >>>
> >>> * subversion/libsvn_subr/subst.c
> >>> (translated_stream_read): Add an optimized code path
> >>> for single byte read requests.
> >>>
> >>> Patch by: Stefan Fuhrmann <stefanfuhrmann{_AT_}alice-dsl.de>
> >>
> >> Committed in r921057.
> >
> > Any chance this could be backported to 1.6.x?
> >
> > This seems like a nice performance improvement with low impact on the
> > rest of the code. Would be nice if this could be put out there as
> > quickly as possible (i.e. not having to wait for 1.7) for users to
> > enjoy ...
>
> Can you quantify the improvement? I don't suppose a GUI like
> TortoiseSVN reads config files very often. Any improvement is lost in
> the noise when I run the command line client on Linux. It doesn't
> change the number of system calls reported by 'strace -c'.

Actually, TSVN reads the config file quite often: thread pools
make it difficult to reuse svn_client_ctx_t instances, so they
get created rather frequently. But this affects only a small
part of TSVN's functionality. I found the "bottleneck" when
running the TSVN /trunk code against an ra_local repository
performing many proplist requests.

The improvement only reduces the overhead of reading
buffered data. Therefore, the number of system calls will
not change.

Backporting the patch to 1.6.x is not necessary from the
TSVN point of view. If someone could verify that this actually
improves authz file handling, many users would benefit.
*That* might then be a strong argument for the backport.

-- Stefan^2.
Received on 2010-03-15 19:48:05 CET

This is an archived mail posted to the Subversion Dev mailing list.