 The Sony/Toshiba/IBM Cell Broadband Engine (CellBE) is a very powerful and capable processor able to crunch a significant number of floating point operations far surpassing the capabilities of a typical CPU. The Floating Operations per second (FLOPs) numbers on the CellBE is astounding, but theoretically that is just peak performance and you'd have to tweak your code for a good long time to even come close to that, unless you're the atypical guy who counts a simple multiply accumulate (MAC) as a main kernel in their app. Learn how it all happened at: http://www.research.ibm.com/cell/
|
 Brook+ is a stream programming language for GPGPUs offered by AMD. The API which exploits AMD's compute abstraction layer (CAL) is an extension of Brook. Brook+ runtime provides support for both the GPU and CPU backends. Here is a link to a quick overview: http://ati.amd.com/technology/streamcomputing/AMD-Brookplus.pdf
|
 CUDA from Nvidia allows programmers to utilize GPUs from their video cards for GPGPUs. The language, similar to that of a parallel programming language encourages programmers to think in wide programming. Exploiting the underlying hardware using Single Instruction Multiple Data (SIMD) is the best course of action. For more information check out: http://www.nvidia.com/object/cuda_home.html
|
|
Answers to my favorite FAQs |
|
I am not sure how you stumbled on to this website, but while you're here, let me tell you how much search engines such as google, yahoo, and others saved countless hours trying to fight the good fight against bad behaving applications and drivers. In light of that I would like to return the favor and post FAQs that have helped me in past. These are going to be fairly random, since I don't always get stuck on the same topic ;). But hopefully there is something useful for you here. Enjoy! |
|
