Re: [PATCH] index-pack: avoid excessive re-reading of pack directory

From: Jeff King <peff@peff.net>
Date: Jun 9, 2015, 17:41 UTC
Message-ID: <20150609174106.GB32034@peff.net>
In-Reply-To: <20150609172436.GA32034@peff.net>

On Tue, Jun 09, 2015 at 01:24:36PM -0400, Jeff King wrote:

Show 9 quoted lines

> I tested this on my system, and confirmed that for a "git clone
> --no-local --bare git.git":
> 
>   1. It cuts the number of openat/getdents/close syscalls by several
>      orders of magnitude.
> 
>   2. The overall time drops from ~11.4s to ~10.5s. I suppose if I timed
>      only the `index-pack` process, it would be even higher (as a
>      percentage improvement).

Just for fun, I did a "git pack-objects --all --stdout" from linux.git, and then timed "git index-pack --stdin" on it in an empty repo. With a configured alternate pointing to another empty repo, just to make it more unfair. And then I stored it all on a ramdisk to emphasize the cost of the syscalls versus hitting the disk. The numbers I got were:

  [before]
  real    2m13.093s
  user    3m31.884s
  sys     0m55.208s

  [after]
  real    1m40.389s
  user    3m10.776s
  sys     0m26.012s

That's sort of a ridiculous test, but it does show that this was having some impact even on "normal" systems without insane syscall latencies.

-Peff