From: David Willmore <davidwillmore@gmail.com>
Date: Wed, 20 Apr 2005 16:10:19 GMT
Subject: Re: [PATCH] write-tree performance problems
Message-ID: <1cddf4cc05042009101b151139@mail.gmail.com>
In-Reply-To: <Pine.LNX.4.58.0504200833580.6467@ppc970.osdl.org>

On 4/20/05, Linus Torvalds <torvalds@osdl.org> wrote:
> It really _shouldn't_ be faster. It still does the compression, and throws
> the end result away.

Am I misunderstanding or is the proglem that doing:
<file with unknown status> -> compress -> sha1 -> compare with existing hash

is expensive?

What about doing:
<file it's supposed to be equal to> -> uncompress -> compare with
unknown status file

It's more file I/O, but the uncompress is much cheaper than the compress.

On a second issue, what's the format of the main 'index' file?  Is it:
<pathspec> <sha1hash>
<pathspec> <sha1hash> 
?
If so, that's not going to compress well.  A file like:
<pathspec1>
<pathspec2>

<sha1hash1>
<sha1hash2>

Will compress better.

Stop me if I'm way off base--I'm just following the mailing list, I
haven't tried out the code.

Cheers,
David