[FoRK] [rkoch@rkoch.org: [Amygdala-development] The Cell programming model]

Eugen Leitl eugen
Thu Jun 30 07:22:56 PDT 2005

I definitely need one. Looks like fun, and comparatively easy.

----- Forwarded message from Rüdiger Koch <rkoch at rkoch.org> -----

From: Rüdiger Koch <rkoch at rkoch.org>
Date: Thu, 30 Jun 2005 06:29:34 -0700 (PDT)
To: amygdala-development at lists.sourceforge.net
Subject: [Amygdala-development] The Cell programming model

There is a description of how the interface of the SPUs will look like
- it's an adaption from a speech the author gave at Linuxtag a few days

In short it looks as if there is going to be 3 different programming
models / levels of abstraction one can choose from:

1. a virtual filesystem similar to procfs
2. a library abstraction that works by explicitly creating some sort of
SPU thread - that code will look similar to the pthread interface
3. gcc autovec which is not done yet and will be available later

I guess here it comes to that tradeoff flexibility vs performance if
one uses the specialized interfaces 1 or 2. If that's required we
either get plenty of #ifdef into the code or we use configure to choose
from 2 different engines. If we can use 3. we can probably treat one
Cell as an 8-way system with a vector unit per CPU.

At the end there is also a link to a presentation by Sony:

The section "Programming Models" starts at slide 24.


SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
Amygdala-development mailing list
Amygdala-development at lists.sourceforge.net

----- End forwarded message -----
Eugen* Leitl <a href="http://leitl.org">leitl</a>
ICBM: 48.07100, 11.36820            http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE

More information about the FoRK mailing list