Hey all,
Ive currently been asked to write a report on the differences between the different programming methods for GPU cards
so far I have CUDA, CTM, Brook, Brook+, Rapidmind and PeakStream
I know that CUDA works on linux, winXP
question 1) Does brook work on linux, how about CTM, brook+ and rapidmind and peakstream
2) how easily available are the SDK’s for CTM, brook, brook+, rapidmind and peakstream?
3) how easily available is sample code or code that people have put up on the web for CUDA,CTM,brook, brook+, rapidmind and peakstream comparitively
4) which method gives greatest access to the GPU’s functionality for optimisation purposes
5) what are the advantages/disadvantages of using any particular method
6) other than CUDA, which languages are closest to c
7) which methods implement IEEE?
8) What other differences are there that I can put in my report?
You’ll get a much more useful response if you go do a few hours of reading on the Web first, then ask any specific CUDA related questions you have here. This isn’t just an issue of wasting people’s time: overly broad questions tend to get you vague answers.
(Here’s one tip, though. When reading through the first few chapters of the programming guide, pay special attention to the discussion of shared memory and blocks. This separates CUDA from some other GPGPU technologies.)
Wow, I saw their web site a while ago, and now it’s gone! Well, since Google bought it, that sounds reasonable to me…but how about PhysX which is acquired by NVIDIA? I remember I saw some rumor said it’s going to be ported on CUDA, but nothing comes out since then. Any news? (sorry I know it’s off topic…just curious about that…)
Yes the PhysX team is now nicely integrated in NVIDIA and doing a ton with CUDA. So it’s been nice having an even stronger user group internally. Sorry I can’t report on when you’ll see products with this.
Will physx be a seperate software product, or will it be a free download (for generating extra hardware sales). I read Physx was ported to CUDA in a month, I must say that sounds impressive. External Image
Ah cool. Someone I know had bought a laptop with Physx card, for some simulation purposes. I told him, he should be using CUDA instead for speed, since there were 2 8800M’s in it, but now he can have the best of both worlds :)