wsdiff could stand to be seriously sped up by doing its comparisons in
parallel:
6510912 wsdiff should make use of more than 1 CPU
so I've put something together that does just that. It fires up one thread
for each online CPU on the machine, and each thread pulls from a work queue
until its done. I need someone to review the changes:
http://www.comfychair.org/~duvall/webrev-wsdiff-mt/
if you can claim experience with Python threading, that'd be even better.
My threading-fu is weak, and my feel for how it's supposed to work in
Python is pretty limited.
Also, any comments on design would be appreciated. A couple of notes:
- I'm not sure if the way I choose the number of threads is appropriate,
or if there's a better way.
- I feel like all the queue-termination code is hacky, but I don't know
of a better way to do it, short of requiring Python 2.5 to bring in
new Queue module APIs.
- For some reason, hitting ^C causes the worker threads to throw
exceptions where I don't expect them, but otherwise I don't see any
deleterious effects. MainThread is the only thread that actually sees
SIGINT, as I would expect; it's just that the other threads don't
terminate as cleanly as I would expect. I can't reproduce that in a
reduced version of the program.
- Yes, there are a bunch of debugging print statements that will go away.
Thanks,
Danek
_______________________________________________
tools-discuss mailing list
[email protected]
