From: Jeff King <peff@peff.net>
Date: Mon, 02 Apr 2012 20:51:12 GMT
Subject: Re: [PATCH 3/3] revision: insert unsorted, then sort in prepare_revision_walk()
Message-ID: <20120402205112.GA28824@sigill.intra.peff.net>
In-Reply-To: <20120402203728.GB26503@sigill.intra.peff.net>

On Mon, Apr 02, 2012 at 04:37:28PM -0400, Jeff King wrote:

> Has anyone looked seriously at a new index format that stores the
> redundant information in a more easily accessible way? It would increase
> our disk usage, but for something like linux-2.6, only by 10MB per
> 32-bit word. On most of my systems I would gladly spare some extra RAM
> for the disk cache if it meant I could avoid inflating a bunch of
> objects.

Actually, that is an over-statement of the size. That would be a
per-object piece of metadata. A per-commit piece like timestamp would be
only 1M per 32-bit word in linux-2.6 (about 1/4 million commits). Or put
another way, we could store timestamps and 20-byte parent sha1s in about
11M.

-Peff