Comments on: Stream processing for the Masses? I don’t think so! http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/ A Blog on Parallel Programming and Concurrency by Michael Suess Wed, 09 Jul 2008 02:32:24 +0000 http://wordpress.org/?v=2.5 By: Rashed http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-37698 Rashed Fri, 28 Dec 2007 20:33:53 +0000 http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-37698 Oops! I thought this was a recent post!! My comments therefore are not relevant!! -rashed Oops! I thought this was a recent post!! My comments therefore are not relevant!!

-rashed

]]> By: Rashed http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-37695 Rashed Fri, 28 Dec 2007 20:20:58 +0000 http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-37695 It's interesting to see companies claiming as their own inventions and concepts developed decades ago. ATI is not first!! :) I would agree with Michael that his post is a bit too pessimistic especially while referring to GPGPU programming: the half empty as opposed to half-full. Kindly let me add that programmers have been doing GPGPU based programming for a while now and often using good abstractions therefore without need to be too much aware of internal details of GPUs. Also, this discussion will be incomplete without mention of CUDA (by NVIDIA) which provides a nice abstraction layer for GPGPU programmers. There are others already active in the field of GPU, multicore, and cell processor based programming (e.g. RapidMind) for providing a flat abstraction for the three kinds of compute hardware. Of course jury is still out as how these various threads of innovation would take shape in future! Cheers and Happy New Year! -rashed It’s interesting to see companies claiming as their own inventions and concepts developed decades ago. ATI is not first!! :)

I would agree with Michael that his post is a bit too pessimistic especially while referring to GPGPU programming: the half empty as opposed to half-full. Kindly let me add that programmers have been doing GPGPU based programming for a while now and often using good abstractions therefore without need to be too much aware of internal details of GPUs.

Also, this discussion will be incomplete without mention of CUDA (by NVIDIA) which provides a nice abstraction layer for GPGPU programmers. There are others already active in the field of GPU, multicore, and cell processor based programming (e.g. RapidMind) for providing a flat abstraction for the three kinds of compute hardware.

Of course jury is still out as how these various threads of innovation would take shape in future!

Cheers and Happy New Year!

-rashed

]]> By: Ian http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-10203 Ian Wed, 23 May 2007 19:08:39 +0000 http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-10203 It's nice that ATi have finally given people the opportunity to unleash some of the power in their graphics cards. But the real speed is currently lying dormant in the thousands of computers that don't have GPUs are can't be bothered to install the correct software. <a HREF="http://ianhickman.blogspot.com/2007/05/internet-as-stream-computer-part-1.html" rel="nofollow">Stream processing with AJAX</A> It’s nice that ATi have finally given people the opportunity to unleash some of the power in their graphics cards.

But the real speed is currently lying dormant in the thousands of computers that don’t have GPUs are can’t be bothered to install the correct software.

Stream processing with AJAX

]]> By: Vince http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-78 Vince Fri, 01 Dec 2006 09:14:28 +0000 http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-78 Very Nice introduction... I am trying to get started with GPGPU Very Nice introduction… I am trying to get started with GPGPU

]]> By: franjesus http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-55 franjesus Tue, 17 Oct 2006 15:06:11 +0000 http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-55 I forgot the <a href="http://www.ati.com/developer/siggraph06/dpvm_e.pdf">paper</a>. I forgot the paper.

]]> By: franjesus http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-54 franjesus Tue, 17 Oct 2006 15:04:48 +0000 http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-54 A short paper. Interesting stuff, I'm just wondering what will be nVidia's answer. Will they keep on the way of specific physics co-processor?? Also, the buy of ATi by AMD and AMD's push for HyperTransport-attached co-processors makes me hope for the day when big simulations won't be carried on clusters of thousands of computers, but on relatively cheap vector processing units, with the central CPU acting only as director and feeding instructions to the units. Let's hope memory and buses manage keep the pace of processors. A short paper.

Interesting stuff, I’m just wondering what will be nVidia’s answer. Will they keep on the way of specific physics co-processor??

Also, the buy of ATi by AMD and AMD’s push for HyperTransport-attached co-processors makes me hope for the day when big simulations won’t be carried on clusters of thousands of computers, but on relatively cheap vector processing units, with the central CPU acting only as director and feeding instructions to the units. Let’s hope memory and buses manage keep the pace of processors.

]]> By: PB http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-48 PB Tue, 03 Oct 2006 06:32:38 +0000 http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-48 The following SIGGRAPH talk from ATI might have some answers to issues brought up by Michael. http://www.atitech.com/developer/siggraph06/dpvm_sketch_siggraph.pdf - PB The following SIGGRAPH talk from ATI might have some answers to issues brought up by Michael.

http://www.atitech.com/developer/siggraph06/dpvm_sketch_siggraph.pdf

- PB

]]> By: Vivek http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-45 Vivek Wed, 27 Sep 2006 21:32:31 +0000 http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-45 Darn! When the NVidia GEForce first came out with the GPU concept, I predicted it would one day be used for non-graphics tasks! I wish I had a million $ for every time my predictions came true... Darn!
When the NVidia GEForce first came out with the GPU concept, I predicted it would one day be used for non-graphics tasks!

I wish I had a million $ for every time my predictions came true…

]]> By: Christopher http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-42 Christopher Fri, 22 Sep 2006 18:42:58 +0000 http://www.thinkingparallel.com/2006/09/22/stream-processing-for-the-masses-i-dont-think-so/#comment-42 Wow, EVERYONE is talking about streams this week. I even wrote my post about it yesterday (which made the front of dzone no less). ATI's announcement even got onto Digg. The PeakStream stuff looks pretty interesting, and the fact that they are funded by both Sequoia and KPCB means that streams will have a strong push in the market. Wow, EVERYONE is talking about streams this week. I even wrote my post about it yesterday (which made the front of dzone no less). ATI’s announcement even got onto Digg.

The PeakStream stuff looks pretty interesting, and the fact that they are funded by both Sequoia and KPCB means that streams will have a strong push in the market.

]]>