Open cl and grand central- are they showing in betas yet?

in macOS edited January 2014
Can someone that has access to the latest betas of snowleopard tell us if these two technologies are available yet? If so, are they allowing the dual gpu MacBook pros to use both gpus at the same time?


  • Reply 1 of 7
    s.metcalfs.metcalf Posts: 901member
    I don't think Apple have ever said they plan to allow that, so I'm sure the answer is no. It would be good to get rapid switching between the two at least, so you don't have to log out to change. Maybe it will happen but I wonder what the heat and power drain will be like using both GPUs. Careful heat and speed (throttling) of the GPUs could mean it's possible but I wouldn't get your hopes up about Apple enabling this. They have already underclocked the discreet GPU in the MacBook Pros so if heat wasn't an issue you'd think they'd at least use the factory clock rates.

    Originally Posted by iansilv View Post

    Can someone that has access to the latest betas of snowleopard tell us if these two technologies are available yet? If so, are they allowing the dual gpu MacBook pros to use both gpus at the same time?

  • Reply 2 of 7
    iansilviansilv Posts: 283member
    Ok but are the betas allowing people to test grand central and open cl?
  • Reply 3 of 7
    You can already download and install OpenCL implementations on a whole number of OSs:

    The "open" in the name is not a marketing gimic, Apple submitted it as a standard and handed off control of it to the Khronos group (the same organization that controls OpenGL).
  • Reply 4 of 7
    iansilviansilv Posts: 283member
    Umm... Heh ok. Can someon just tell me if in the betas of snowleopard that are currently released to developers, if opencl and grand central are in there so the user can see their full speed benefits right now. Can someone clarify this direct question please?
  • Reply 5 of 7
    mr. hmr. h Posts: 4,797member
    It's not a matter of users seeing the benefits of openCL and grand central. Those are features for developers; users will only see the benefits if the programs they use are updated to make use of these technologies.

    All the developer previews of 10.6 have had openCL and grand central, as they are the main features of 10.6 from a developer's perspective. The point of developer previews is to allow developers time to get coding with new APIs and making sure their apps are ready when the new OS arrives. The added benefit to Apple is they get a load of beta-testers too.
  • Reply 6 of 7
    karl kuehnkarl kuehn Posts: 756member
    And durring the WWDC preview wan't it stated that the new version of Mail used GrandCentralDispatch to coordinate the number of active threads, so it could easily add or remove them as needed. This would clearly demonstrate that the system resources were there.

    But I think that the biggest problem here is that people don't understand what either GCD or OpenCL do, and so have this vague notion that they will magically make everything better. So:

    Both a resources that developers can use. They don't magically kick in, nor are they going to be of much use to most programs. There will be slight general bumps in speed as I am sure that a few OS-level things will use them, but that bump is unlikely to be generally noticed.

    OpenCL will probably only ever see use in the most high-performance code. Some games might use it to help with their physics calculations, but most games are already limited by the graphics card, so trying to use the graphics card resources for the physics is not a good idea there. Where OpenCL will really shine is in scientific simulations. For very limited calculations (oddly what simulations often do) modern graphics cards can process more than supercomputers of a few years ago could. It might be that some video codecs will also fall in this category, but there you are also limited by the amount of data that you can push through, so I question how much you are going to get out of that.

    GrandCentralDispatch is a little more generally applicable, but even there some programs will benefit from using it (ei: being rewritten to use it) where others won't. The basic difference is that some tasks are easily divided into parallel sections, whereas others need to be done sequentially. GCD (and the blocks extension to C that it requires) does two things: provides a language level concept of a block of work, and then provides a OS-interface to a method of getting them to run. Threads are nasty things to program, and people usually trip up and get it wrong. GCD provides a uniform way of letting Apple's code handle most of the difficult parts of figuring out what resources are available and creating and destroying threads while the programmer concentrates on what they do. Really nice for programs that use threads, but useless for everything else.
  • Reply 7 of 7
    MarvinMarvin Posts: 14,436moderator
    The OpenCL drivers had to be approved for both parties as it needs to work the same on all platforms, being an open standard. Developers can code and test in CUDA though and porting to OpenCL would be straight-forward.

    The first demos have already arrived from the GPU manufacturers. Here is what is noted as the first end-user demo of OpenCL:

    NVidia had an OpenCL demo, which was basically one of the demos in the CUDA SDK.

    They may not have used Snow Leopard for these demos though.

    I think Apple probably have something planned for Quicktime X regarding hardware encoding. The GPU transcoders focus on ipod/youtube conversion and Quicktime X has this. They won't do this in standard Quicktime (Pro) as the codecs would all have to be updated.

    Typically, this will offer about 3 times faster than CPU encoding. I'd expect the 9400M to be 3 times faster than a Core 2 Duo. Apple's encoder is very slow right now though so it will probably be a 6x speedup and the 9400M is being marketed as showing a 6x speedup using Badaboom on Windows.
Sign In or Register to comment.