We'll be having our regular Toronto Area Python User's group meeting at Linux Caffe on Tuesday the 15th of September at 7:05pm.

Parallelization in Python: Code Samples, Experiences and Advocacy

Unlike functional languages (Haskel, Erlang), where parallelization is "baked into" the language, Python programmers often have to do a bit of work to get their code to run in parallel. We have a huge number of libraries that allow you to run your code concurrently, so this month we're going to collect code samples and descriptions of how you go about making your code run in parallel, and when each approach is appropriate.

So, to participate, spelunk through your codebases and try to find a few examples of parallel programming, whether it be threads, multiprocessing, database-based concurrency, file-base concurrency, grid systems, GPGPU code, continuations, green-threads, MPI, RPC, Twisted, asyncore, raw sockets/pipes or any of the dozens of other approaches. If possible, have a piece of code that shows "how it works in practice" and maybe consider how to answer these questions:

* How does it (the mechanism of parallelization) work (loosely)? * What benefits does it give you? * What problems does it introduce? * How does it scale (up/down)? * What level of granularity makes sense for it? * How does it handle communication between code? * How parallel can you get effectively? (10s of nodes? 10s of thousands? millions?) * Does it handle contention/locking/conflicts? * How stable/reliable is it?

We'll try to keep the discussion lively and interesting.

Venue details, maps and the like on the web-site:

http://www.pygta.org

We'll also likely have a discussion about putting together a series of 1/2 or 1-day classes (in a more formal classroom-like setting) for new Python users.

Have fun all, Mike