[Stackless] The Future of Stackless

Carlos Eduardo carlosedp at gmail.com
Fri Dec 28 20:51:47 CET 2007


Some time ago, I thought about creating a module that derives from thr processing package(from cheesestop). What it does, is mimic the threading module but using multiple processes instead. The IPC is done over sockets or pipes,depending on platform.

In this module, we will have one process per core and one stackless scheduler for each. Each process will have a manager tasklet that acts as an IO dispatcher and a tasklet manager. It sends the newly created tasklets to the other processes and manages them. The channel communication is done with one channel in each side and some kind of Ipc to bridge them. When a tasklet sends something using a channel, it gets blocked as usual but its data is sent to the other process and awaits in the new channel.

This is a rough idea and I don't know exactly where to start and how to do some stuff. Do you guys seem this idea feasible?

I will give a shot when the holyday is over...

Carlos

-----Original Message-----
From: Andrew Dalke <dalke at dalkescientific.com>
Sent: quarta-feira, 26 de dezembro de 2007 20:08
To: Stackless mailing list <stackless at stackless.com>
Subject: Re: [Stackless] The Future of Stackless

On Dec 25, 2007, at 7:32 PM, Andrew Francis wrote:
> Okay, I read the FAQ and have heard some of these
> arguments before. However these approaches strike me
> as a "boil the ocean" solution. Perhaps a simpler
> solution will please 80% of the users whom wish to
> exploit multiple CPUs.

Feel free to work on it.

> I guess my point is that it would be desirable to have
> some conceptual model that allows us to analyse issues
> and steadily improve Stackless Python's power and
> performance.

My point is that the implementation, based as it is in CPython, will  
not be able to support those improvements without a huge amount of  
work.  Too much of it assumes the GIL.  If you're going to do all  
that work, perhaps it's more interesting to work on something like PyPy.

				Andrew
				dalke at dalkescientific.com



_______________________________________________
Stackless mailing list
Stackless at stackless.com
http://www.stackless.com/mailman/listinfo/stackless





More information about the Stackless mailing list