This took a while, but here it is. (Actually, I finished this about a
week ago, but had no Internet connection so I couldn't send it.)
The series starts with basically the same DAG appliedness test as
before, with the same known performance bug. Later on in the series,
the mechanism is changed to one that doesn't have the bug. I kept the
intermediate, slow state because the changesets read better that way,
and because the new mechanism is more complicated than the old so it
might be useful to be able to compare their output in case some bug
turns up further down the road.
To test the performance, I used a script (which I've unfortunately
misplaced) that (in the kernel repository) reset to one point 10000
commits in the past and one 5000 commits in the past, pushed a few
patches at each spot, and then created 100 applied and 100 unapplied
patches on top of upstream HEAD. This triggers the performance bug
with the first algorithm since we have unapplied commits very far from
HEAD.
Both algorithms are documented in the patches that introduce them.
--
Karl Hasselström, kha@xxxxxxxxxxx
www.treskal.com/kalle
-
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
