[hipl-dev] a small study on arch performance

From: Miika Komu <miika@xxxxxx>
To: hipl-dev@xxxxxxxxxxxxx
Date: Wed, 18 May 2005 16:32:56 +0300 (EEST)

I conducted a small expirement during the weekend to understand better the
performance difference with tla and bazaar on large trees. As an example,
I had an linux-2.6 tree where I mean modified a single file, and then made
a commit. I repeated this two times for a specific arch configuration. I
got the following figures out:

tla 1.2 with revlibs:
        2 min 00 secs (first commit)
        0 min 59 secs (second commit)

tla 1.3.1 without revlibs:
        5 min 56 secs
        2 min 17 secs

tla 1.3.1 with revlibs:
        1 min 00 secs
        1 min 06 secs

baz 1.3.2 without revlibs:
        2 min 00 secs
        1 min 23 secs

baz 1.3.2 with revlibs:
        2 min 05 secs
        0 min 53 secs

One should notice that since I repeated the commits only twice, the
figures above are not statistically sound. Despite of this shortcoming, I
would draw a few quick conclusions about the latest versions of the
sofware:

* tla and baz performance with revlibs is roughly equal
* baz is somewhat faster than tla without revlibs

As a result, at least I am currently using bazaar instead of tla. It works
almost exactly the same way as tla, but you just have to replace "tla"
command with "baz". If you are happy with tla, there is no need to switch
to bazaar. Bazaar has the deb and rpm packages here:

http://bazaar.canonical.com/download.html

-- 
Miika Komu              miika@xxxxxx          http://www.iki.fi/miika/

[hipl-dev] a small study on arch performance

Other related posts: