faster log benchmark results

Ian Clatworthy ian.clatworthy at internode.on.net
Tue Jan 20 05:51:18 GMT 2009


As promised, attached are the benchmarking results comparing
bzr.dev vs my-faster-log patch. These figures now hold regardless
of the default log format for a user: both --long and --short
are equally covered.

Here's a brief explanation of what the labels mean:

* mainline = --short
* merges = --long
* delta = -v
* recent/tip/mainline/initial/early = -r[-10../-1/10/1/..10]
* incremental = --limit/-l 5

I feel the incremental benchmarks are quite important
because they show how fast users see *something*, even if
they aren't running with that option specified.

Given the size of the emacs history (> 100k revisions), I
think many of these figures are now approaching acceptable,
particularly for users with log_format=short configured in
their bazaar.conf file.

The main one that stands out to me still is LogMergeRevision,
i.e. showing a single revision off the mainline. (For the Emacs
repository used, the magic revno tested was 77537.1.3262.)
The bzr-revnocache plugin I hacked together yesterday greatly
improves this but it's still slower than I expect. Note
that bzr-revnocache wasn't used running the benchmark above.

The other big issue remains log -v on the whole branch.
That has been excluded from the above because it takes
hours on mega-repositories like this. The work going into
the next repo format should speed that use case up.
(I'll verify our progress towards that end soon.)

In summary, I think the patch is a nice step forward in
terms of log performance (and internals for that matter!).

Ian C.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: log-patch-emacs-results.txt
Url: https://lists.ubuntu.com/archives/bazaar/attachments/20090120/5808d650/attachment.txt 


More information about the bazaar mailing list