[PATCH] Increase performance on Windows by up to 2x

[Bryan O'Sullivan]

On Thu, Mar 26, 2009 at 6:14 PM, Matt Mackall <mpm at selenic.com> wrote:

>
> Well it shouldn't. We should have one read to grab the index and one
> read to grab the relevant chunk of the log.


There's at least one additional 4-byte read to get version info, followed by
a seek.

Python uses regular stdio buffered I/O for file objects. I assume that a
win32 read amounts to a system call, so it's naturally more expensive to do
a few of those interleaved with a seek or two than one bigger one that
obviates the later need for a seek.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://selenic.com/pipermail/mercurial-devel/attachments/20090327/345e5006/attachment.htm