[cairo] Changes in cairo-perf

[Chris Wilson]

On Sat, 2009-03-07 at 01:21 -0300, André Tupinambá wrote:
> Why was this code changed? Could the old feature came back?

Because previously it would set up every test before deciding in the run
method that the test was to be excluded. As that is quite slow and
happened to tripping over a bug in the kernel (at the time), I disabled
it.

Whilst I sympathise with you about the desirability to filter tests on
sources - I'd much rather have a more explicit mechanism (and consistent
test naming as well). So I would welcome a patch to control the various
aspects of the tests, such as which sources to run, sizes, and subsets.
As well as more tests and an estimate of the (per-test) synchronization
cost ;-)
-ickle