From: "ara.t.howard" Date: 2008-09-15T01:22:04+09:00 Subject: Re: how can i find lingering file descriptors? On Sep 14, 2008, at 1:48 AM, Brad Volz wrote: > n my case, I'm not doing the calculations in ruby code. I'm calling > an external program to fetch the data for me, so in both cases there > is an external process that can be scheduled by the operating system > onto whatever processor has cycles. Presumably, if that were not > the case, and I was doing a lot of cpu intensive work in Ruby > itself, the forkoff version would be faster by providing 'n' GIL > instances instead of just one. i think that's a correct assessment alright. time and time again i find that spawning a bunch of processes via threads is the easiest and most powerful method of achieving parallelism with ruby - and in general. i know it's not sexy, but being able to test a computation unit from the command line and then simply fan them out across cpus or even many nodes make incremental development super easy and let's us use all the tools we're accustomed to (ps, free, top, ssh, ruby) to manage the computational units. it's interesting that you're also following this pattern. cheers. a @ http://codeforpeople.com/ -- we can deny everything, except that we have the possibility of being better. simply reflect on that. h.h. the 14th dalai lama