[RFC] new benchmark use cases

[Wouter van Heyst]

On Tue, Jul 18, 2006 at 01:49:04PM +0200, Jan Balster wrote:
> Hi,
> 
> some of you saw the logs of the performance
> meeting on #bzr yesterday.  I am one of the
> guys contracted to help with improving bzr's performance
> and helping Martin and Robert particularly.

For which we are grateful, please keep such meetings in #bzr so we can
read the backlog when we delurk. Oh, and welcome! :)


http://codespeak.net/bzr/perf_history/summary_bzr.dev.html is looking
good. Just not sure about the revisions growing to the right, it does
feel natural for a timeline, but it does mean scrolling when I want to
check the most recent timings. Love the graph though!

> One precondition to improve the performance is to have many
> good benchmarks first :) So one goal stated there was to implement
> more benchmarks based on use cases.
> 
> As far as i see, the following use cases are already benchmarked a bit:
> (/bzrlib/benchmarks/)
> 
>   * add
>   * status
>   * commit
>   * log
>   * log-specific-file
>   * checkout
> 
> More benchmarks might include:
> 
>   * create a bundle
>   * apply a bundle
>   * check
>   * branch
> 
> Can you think of more benchmarks or suggest refinements of existing ones?

As John mentioned, doing status/diff/log on a small set of files vs the
entire tree. Perhaps info to make sure we're not doing too much by
default (as is the case now, 6 minutes to get the size of revisions on
an sftp branch).

Wouter van Heyst