See http://www.editgrid.com/user/jaalto/vc-test for a spreadsheet comparing
git/hg/bzr

Note in particular the script:
  http://www.editgrid.com/user/jaalto/vc-test

The above is an artificial test basically doing a giant record on the Linux
kernel source (they admit that this is rather artificial, but it's a start), and
also a get.

We'd probably choke on the first test given issue80, but we should at least have
the infrastructure in place.  We'll get there one day!

See also Ian's http://projects.haskell.org/camp/benchmarking/get.shtml

Agreed -- it is good to have measurements even if they show failure or bad
performance.  That way you can tell if your changes make it stop failing or
perform better!  :-)

Jason pointed this out to me:
http://ldn.linuxfoundation.org/article/dvcs-round-one-system-rule-them-all-part-3

Funny, I thought I saw this a while ago, must have just failed to react.
 I've requested his benchmarking scripts.