While we are talking about python 2.6 vs 2.7 in #1106
There is a few solutions available to establish benchmarking of python code:

• https://github.com/pydata/vbench -- somewhat weakly maintained, number of clones hanging around (I have one for numpy benchmarking, which noone cared about so no longer running : http://yarikoptic.github.io/numpy-vbench/)
• https://github.com/spacetelescope/asv/ -- seems to be the best bet for us. Adopted by scipy now and others. Would the best long run.

It would be nice if we had a set of benchmarks on

• simple parsing of sample log files for each filter, to detect when addition or a change of regexp or code change would incur significant performance hit
• initial start time to establish a simple demo jail on a sample file
• etc

This could help for many use-cases.
Would be cool if someone would research into either asv has any current integration with github... it would be ideal if we could set it (or buildbot) to trigger benchmark effect reporting for pull requests (dreaming now)