What is Digg?22 Comments
- YodaJones, on 11/20/2008, -2/+16Awesome news. Suck on that ISO.
- luchid, on 11/20/2008, -2/+12CUDA=Propietary, no interoperability.
OpenCL= Open, interoperable.
DX11 isn't even done, nor a worthy point of comparison for any of the competing technologies. - SuperSunny, on 11/20/2008, -0/+5Ironic, considering Apple uses Nvidia cards.
CUDA is for individual apps, not the whole system. OpenCL is for the system, and for individual apps. Plus, it's open. It maybe adopted into another OS. - Spuy767, on 11/20/2008, -0/+5When you read the above comment, you'll know that RobbieS doesn't really understand what he's talking about and just wants to bash apple even though this development could benefit all users of multi-core computers.
- SuperSunny, on 11/20/2008, -0/+4Future apps on Snow Leopard will use your GPU to accelerate the entire system and the running of each app. The GPU contains many more cores than a CPU, thus, splitting the workload evenly, and truly turning the computer into a powerful parallel processor.
Perhaps Final Cut rendering each cut at the same time, parallel. Things like that, applied to the WHOLE system. Beautiful
I know Nvidia is doing it with CUDA, but there are hardly any apps for CUDA, and it's not for the entire system. - RobbieS, on 11/20/2008, -1/+3Look, I'm not saying OpenCL can't succeed. I'm sure it could. But you have to consider all the factors of what's going. First, just throw out all these pre-conceived notions of it being "Open" at all. This isn't some OSS project. OpenCL is serving the business needs of the companies involved with it, primarily the main promoter of the standard, Apple. Apple didn't want to be tied down to a specific hardware vendor, which is certainly understandable on their part. But don't think OpenCL was made for the benefit of everyone, so rainbows and sunshine can rule the day. Hardly.
If you need proof, just look at what happened with OpenGL 3.0. If you really think these open standards groups really have anybody's best interest in hand besides their own, look at what happened to it. It's comically behind DX in terms of convenience and functionality (and I say this as a hardcore OGL fan). But even if I (and many others) am an OGL fan, I can't possibly argue it's _technical_ merits compared to DX. Nobody worth their engineering salt would.
That's my only argument, and geoken caught it. I'm not arguing against it's supposed 'spirit', which I also question. Rather, I argue against its technical worthiness. CUDA has more features, a more robust implementation, and supports OS X, Windows, and Linux. Is OpenCL really going to deliver those anytime soon? As a developer, why target OpenCL? I'm guessing it is going to run quite well on Apple with NVIDIA hardware, if for no other reason that Apple doesn't want to get embarassed. But what if you want to target other OSes within the next couple years? Or other hardware (ATI, maybe Larrabee)? I'd argue your better off just using those vendor specific implementations than OpenCL also. - geoken, on 11/20/2008, -1/+2DX11 builds are available on MS's site. You can go to youtube and watch dx10/dx11 comparison videos right now.
Also, stating that CUDA isn't interoperable doesn't really disprove the parents main point. He's basically arguing that OpenCL is a stripped down version of CUDA, and by extension is the main reason contributing to it's short development time. Complaining about CUDA's interoperability doesn't address his argument at all. - ORBAT, on 11/20/2008, -1/+2Wow, thanks for contributing.
- MtheoryX, on 11/20/2008, -0/+1Pushing it out faster does not necessarily mean it's more buggy.
The speed of pushing something through a standards body is not necessarily a reflection of the speed of development.
It's quite possible that the standards bodies are simply slow and inefficient in approvals and reviews. - MtheoryX, on 11/20/2008, -0/+1You know how it goes...if you don't have anything nice to say...
- Morghin, on 11/20/2008, -1/+2“If you go to some other larger standards bodies, it’s quite normal for a standard to take five years or more,” he explains. “That’s quite commonplace. You actually have to really push to get it down to eighteen months. Our record was [twelve] months, up to now; we’ve done this one in six.”
If this statement is even remotely accurate, then this is ***** insane. They spent loads of overtime on it though. How will this affect future processes? Will others feel the pressure to squeeze something out even faster (and more buggy) than we already have now? - pyrates, on 11/20/2008, -0/+1For those users of Ubuntu, I've submitted that Ubuntu adopt OpenCL. I'd really like to see them adopt this as well. Go vote it up! :)
http://brainstorm.ubuntu.com/idea/14344/ - gsnedders, on 11/20/2008, -0/+1The main delay in the W3C process, FWIW, is the requirement to have two completely interoperable implementations (non-experimental, public, shipping) to be published as a REC. This, for a large spec, requires a *****-load of test cases, and writing enough test cases is, in general, slow. Writing the spec is not the time consuming part: getting peer-review takes a fair while of just hanging around, and then there's what you get from implementation experience. That's what really takes the time.
I have my doubts about how interoperably implemented such a quickly written spec, presumably with next to no test cases, will be. - Zippo, on 11/20/2008, -2/+2Keep up the good work, guys.
- LeviTheSmith, on 11/20/2008, -2/+1*****
- wollsmosh, on 11/20/2008, -1/+0Don't really know what it all means, but it sounds like progress, and I like progress.
- scratt, on 11/20/2008, -6/+4Shame Khronos can't 'tout the speed' of their work on OpenGL3.0, draft or otherwise!
- lebondarken, on 11/20/2008, -4/+1
- RobbieS, on 11/20/2008, -14/+9That time isn't so impressive when you realize that OpenCL is basically a stripped down copy of CUDA. You think Apple and the standards body would have even been able to come up with any of these ideas, much less implement a spec in 6 months without all the groundwork CUDA did.
Wait until you see the spec. You'll just think..."Wait, this is just a stunted CUDA!" I suppose ATI support is important, and also to make sure that NVIDIA doesn't control that segment entirely. But this isn't being done for an open standard. This is being done because Apple was last to the compute bandwagon (CUDA, CTM/Stream/Brook++, DX 11 Compute Shader) and because Apple wanted a standard it could exert its influence on (who do you think screwed up OGL 3.0?). - inactive, on 11/20/2008, -7/+1There's a guy on my street who owns an Apple. He's kind of a kook, though.
- bbqsalad, on 11/20/2008, -9/+2corn on the cob
Browsing Digg on your phone just got easier with our enhancements to the 
© Digg Inc. 2009
— Content created and posted by Digg users is