1.13 vs 1.14rc1 performance summary

John Arbash Meinel john at arbash-meinel.com
Tue Apr 14 14:11:01 BST 2009

Hash: SHA1

Ian Clatworthy wrote:
> See the attached. There's clearly a bit of noise creeping
> in now and then so I ran the benchmark twice. All runs
> are using 1.9 format.
> In summary:
> * send and log are faster
> * branch is probably faster (but branching outside a shared
>   repo is still terribly, terribly slow on large projects)
> * nothing else (in the scope of the benchmark) has obviously
>   regressed.
> Ian C.

Branch:shared	11.0	6.1	24.4	9.4 bzr branch $... fix

I'm really happy to see this, I'm guessing it is the "iter_files_bytes"

Add:commit	5.3	1.1	1.9	2.5 bzr commit -m "testing add"

Somethingvery strange going on here, given that the former sped up by
almost 4x, but the latter slowed down.

Change:commit	0.9	1.0	1.9	7.6 bzr commit -m ...
Tag:tag		0.2	5.7	0.2	0.5 bzr tag BETA-1

Something to be aware of. I don't know if this is genuine or not. 1.9 =>
7.6 is pretty serious, as is 0.2 => 5.7.

Bundle:send	23.9	1.2	7.1	2.6 bzr send -o zzzBundle.path

23.9s => 1.2.. I know there were a couple of improvements here,do you
think these are genuine? I guess emacs has a *very* long mainline, so it
might be possible. I don't think emacs has a much larger number of commit

In a very strange twist, with the second run:
Branch:shared	11.9	7.7	9.5	16.1 bzr branch $... fix

Suddenly we go *up*...

So I guess we just need to take these with a bit of skepticism. I wonder
if there would be a way to clean up the results a bit. Maybe calling
'sync' between actions?

Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


More information about the bazaar mailing list