Re: Huge performance bottleneck reading packs

From: Jeff King <peff@peff.net>
Date: Oct 12, 2016, 23:47 UTC
Message-ID: <20161012234753.tbqhuc7qdyklpfzv@sigill.intra.peff.net>
In-Reply-To: <20161012231807.syockv2emrsjf55r@sigill.intra.peff.net>

On Wed, Oct 12, 2016 at 07:18:07PM -0400, Jeff King wrote:

> Also, is it possible to make the repository in question available? I
> might be able to reproduce based on your description, but it would save
> time if I could directly run gdb on your example.

I tried this by making a bunch of packs in linux.git (my standard "this is pretty big" repo), like so:

  for i in $(seq 1000); do
    git rev-list --objects HEAD~$((i+1))..HEAD~$i |
    git pack-objects --delta-base-offset .git/objects/pack/pack
  done

and then doing a 25,000-object fetch from upstream (no significance to the number; that's just how far behind upstream I happened to be).

However, I didn't notice any regression. In fact, it was much _slower_ than v1.9.0, because that older version didn't have threaded index-pack.

If you can't share the repo directly, can you tell us more about your fetch? How many objects are in your repository? How many objects are fetched? How many refs are there on the remote side?

-Peff