[Stackless] Newbie Stackless Question

[James Turner]

Hi there,

>     I'm working on a new project, and have decided that Stackless is
> probably best for my needs, since ideally I wish to keep memory and CPU use
> to a minimum. Apparently microthreads will do the memory bit; I have no idea
> if its more CPU intensive or not. This is not URGENT, its just a strong
> preference.

In my experiences using uthreads, CPU *and* memory are extremely well 
conserved--there is little overhead switching between microthreads 
because they don't need a stack context switch.  However, I've found
sometimes you have to play tricks with your program to make I/O work
properly--the lib will sometimes exit all threads if they happen
to be all waiting on socket I/O or something at the same time, 
thinking that there is no more work left to do.

>     Here's my question: All of the examples in the uthread homepage show the
> creation of the threads all at once, then a final 'uthread.run()' to
> actually do everything. Almost all of my threads will be loops, and will be
> long-running. Further, I won't know how many to create when I start the
> application.
> 
>     Is there a way to add a thread in the middle of an application's run,
> after we've already done a uthread.run()?

I don't believe so.  You of course have runAndContinue(), but you still
can't add more after you've already launched your microthreads.

This is just from my experience--there are *many* people on this list
that are more knowledgeable than I and may be able to correct me 
here.  If so, please do, because I to would love to be able to do
something similar to thread.start_new_thread(func,(args,)) using
uthreads.

With normal threads, I'd never recommend this, but it may be 
suitable in your case:  If there's a reasonable maximum number of 
threads you need, run that amount.  You say 'dozens of microthreads'; 
well you can easily have hundreds or more, and each one won't 
schedule up until it has finished wait()ing, or poll()ing, or
whatever--you can still hum along efficient with resources with
even hundreds of microthreads waiting to do your bidding.

Hope some of this helps.

--James
_______________________________________________
Stackless mailing list
Stackless at starship.python.net
http://starship.python.net/mailman/listinfo/stackless