Announcement

Collapse
No announcement yet.

AMD's Paying For Some Open-Source OpenCL Love

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Originally posted by Kano View Post
    Amd has got huge performance problems as the speed/core is not good. But compared to intel they have got the faster gpu. Amd officially dislikes cpu only benchmarks - they hate sysmark as their cpus win absolutely nothing against Intel. So they NEED software that can use the gpu. They do nothing else than Nvidia, they even helped to produce CUDA apps because those do not run on Amd gpus. In order to win some benchmarks AMD needs OpenCL, so they need an app that really speeds up using it. Paying a developer to create what they need is somehow logical, but the idea comes a bit late
    which is odd now that the BD with its x2 integer SIMD units onboard.... but then...(below)

    Originally posted by leeenux View Post
    Yeah, provided you're using this logic:

    take a single-threaded benchmarks where AMD ties Intel, normalize to 8 core vs 4 core(because hyperthreading is magically much less like a real core in this particular scenario than it normally is when Intel fans are bragging about it), then exclaim "well, it took AMD twice as many cores just to tie Intel, so they're half as fast per core".

    Although Llano and Trinity should certainly stomp anything Intel has when properly utilizing GPU compute, but when AMD benchmarks 8 x faster than Intel in Gimp3, everyone will start counting shaders as cores; "OMG, AMD needed 120x more cores to beat Intel by a mere 800%!!!! FAIL!!!!!"
    you dont like benchmarks leeenux, only REAL applications matter after all, right , your new round here i take it...


    so OK lets take a REAL WORLD APP code that almost every one uses in some way today weather they know it or not at some point ,
    such as the video streaming veetle, livestream,USTREAM tv, Justin.tv, and many others.

    that app code being the multi threaded x264 ,you know that one single app that is intensely integer SSE*, SIMD,AVX optimised for x86 and some ARM NEON CPU/SOC Today.

    how come even after L
    oren merrit , Dark shikari and others got remote access they only managed an average 2% to 10% improvement in parts the BD x2 integer SIMD should have walked it but didn't.

    OC you might make the valid claim that Michael cant be bothered to ever find the time to git pull a current x264 with LOT's of AVX and other long standing speed improvements and include/integrate it into his Phoronix Test Suite so that old 2010 version makes AMD look even worse , and you may have a point, so get Michael to do something about it, but that still doesnt make AMD CPUs faster...

    only better optimised internal AMD HW microcode etc for x264 and other common integer SIMD apps will make a real difference but that work
    doesn't happen to date.

    and notice how i didn't once refer to using GPU compute, or the ATI gfx SAD instructions
    etc to make x264 faster, as if they were any good or easy to do for this every day FOSS application code base
    used everywhere, then someone would have already submitted a patch to the x264 devs IRC channel,
    and the x264 devs have tried to get some Gfx devs to actually contribute, not one single gfx or OpenCL coder has stepped up
    or completed even a single re-factor of any single routine .and submitted a patch to the x264 devs to date.
    Last edited by popper; 30 November 2011, 07:53 PM.

    Comment


    • #22
      Nice to see some GEGL and GIMP love.

      Originally posted by crazycheese View Post
      GIMP is FLOSS software, please.
      I don't see anywhere in his post where he says that it isn't. He doesn't appear to misuse any of the various terms used to describe free, Free, open, or Open software, either (since he doesn't appear to have used any of those, or other terms). He did say that it runs on Windows, but that doesn't preclude it from running elsewhere and it didn't appear that he was implying that. As such, I don't see what your problem is.

      P.S.: Michael, sure GIMP misses their release deadlines sometimes (or often, even), but IIRC it's because they don't have many people working on it (and maybe because of internal disputes, too; can't exactly remember off hand).

      Comment


      • #23
        Originally posted by Nobu View Post
        I don't see anywhere in his post where he says that it isn't. He doesn't appear to misuse any of the various terms used to describe free, Free, open, or Open software, either (since he doesn't appear to have used any of those, or other terms). He did say that it runs on Windows, but that doesn't preclude it from running elsewhere and it didn't appear that he was implying that. As such, I don't see what your problem is.
        My "problem", is that in his post, he essentually made a point that only statistical majority of userbase counts towards determining which OS platform is deemed to dominate the application market.
        Which is BS, since this is opensource software and runs anywhere. And it should continue to run anywhere. So his point about windows user base is utter BS.

        While I used GIMP on windows to correct (and recover due to wrong ISO setting on camera) an array of ~2000 12Mpix photos from my brothers marriage, that was only because they had windows preinstalled on laptop. This does not make me prefer windows or ask developers to stop all, except windows gimp branch development. In fact, I'd be really happy to wipe that windows clean - but they got *used* to that BS. And windows, including all its developements and software base has become to what it is, only because of hideous preinstallation agreements. By issueing the claim to improve only software segments which have more major userbase, he essentually uses same approach as microsoft. People use what they are used to, which forces software developers support only this, which forces people again to use this. Circle closed. Thats not a problem, except they all are enclosed in consuming human wastes.

        Or as hitler said: When a lie is said multiple times, it becomes truth. No?
        Last edited by crazycheese; 01 December 2011, 06:53 AM.

        Comment


        • #24
          He wasn't suggesting Gimp developers stop development for other platforms, he was providing one reason why AMD might be doing what they are. It's not a strong argument; nobody can really know how many users of Gimp there are on Windows. But considering there are so many Windows users (regardless of whether it's because they got accustomed to using it and never learned or switched to another OS, or they just prefer it), it's fairly safe to assume that there are a lot of people who use Gimp who also use Windows. And since it's a fairly popular program, I wouldn't doubt there are at least as many on Windows as there are on other operating systems. To say that just because it runs anywhere there's no way there could be more users on Windows is just as much BS as to say that because some program is closed source software there's no way there could be more users on Linux or other operating systems (in general).

          There are three polls on gimptalk, but they're pretty old and not many people voted, so their results aren't really useful in generating a conclusion. All you can really say from those results is that the people on that forum mostly use Windows and Gimp, but even that is a stretch because I doubt that most of the people in the forums even took the poll.

          As to your other point, businesses will do what will make money for themselves. They may make charity every now and then, if they're doing well, but otherwise it's mostly important for them to be profitable (so they can continue to make products, their customers can continue to buy their products, and they can continue to make money...the cycle continues). It is even more important when it is a traded company, as they have a responsibility to be profitable for their stakeholders. This may indeed be a charitable move by AMD, but more than likely it also brings them financial benefits in one way or another. If you want it another way, you'll need to go to a socialist system, or some non-free market system, and I'd rather stay away from both (as one stifles growth {too much competition}, while the other stifles innovation {not enough competition}).

          Comment


          • #25
            it is about linux, and free software developers

            You guys come up with some strange notions, it has nothing to do with windows.

            This is clearly related to the fact that AMD have finally started putting more effort into their linux stuff. Such as the fact that the opencl runtime is now integrated into catalyst releases and not dependent on downloading a separate sdk, and they've also just released their profiling tool for gnu/linux - and as GPL2 software (previously only a visual studio plugin). They are also attempting to build a new community platform, and any marketing/promotion like this is a pre-cursor to that.

            And it's also obviously related to them embracing opencl as their only supported platform for GPU development - which is really their only choice as CAL would never reach critical mass.

            Having a show-case project which demonstrates the utility and potential of the api is an obvious move. Although TBH the gimp is a weird choice since most of it is fast enough on a desktop cpu anyway, and/or it's performance is limited by internal design and not the hardware.

            Comment


            • #26
              Originally posted by notzed View Post
              You guys come up with some strange notions, it has nothing to do with windows.

              This is clearly related to the fact that AMD have finally started putting more effort into their linux stuff. Such as the fact that the opencl runtime is now integrated into catalyst releases and not dependent on downloading a separate sdk, and they've also just released their profiling tool for gnu/linux - and as GPL2 software (previously only a visual studio plugin). They are also attempting to build a new community platform, and any marketing/promotion like this is a pre-cursor to that.

              And it's also obviously related to them embracing opencl as their only supported platform for GPU development - which is really their only choice as CAL would never reach critical mass.

              Having a show-case project which demonstrates the utility and potential of the api is an obvious move. Although TBH the gimp is a weird choice since most of it is fast enough on a desktop cpu anyway, and/or it's performance is limited by internal design and not the hardware.
              of all the key words in this whole thread you elected to highlight you choose "windows", how quaint.

              in relation to your "it's also obviously related to them embracing opencl as their only supported platform for GPU development", ill state it again so you don't totally miss the point this time.....

              OVDecode.so Never actually appeared in the 32 bit linux lib/x86 dir, did that change all of a sudden today as regards the OpenCL OpenVideo driver for video decode, the answer would appear to still be that even a full 12 months later they don't care enough about a OpenCL video decode library to actually compile and distribute it in any form.... so much for them embracing opencl

              while its nice that AMD are giving student developer Victor (Oliveira) A stipend form of salary to keep working on this code as im sure he is happy about that for his independent efforts, where is even a minuscule payment and hardware for
              the likes of Veerappan and his VP8+OpenCL acceleration?, in case you missed it notzed, Veerappan is the ONLY developer that has so far managed to actually provide a working prototype OpenCL video decode source and he continues to work on that as time allows

              are you going to deny that OpenCL video decode is very important to the masses (both x86 and ARM NEON gl,cl .etc) ?, and given that fact it does not seem like AMD are in anyway really committed to embracing opencl ASAP for anything other than some large PR sound bytes now and again.
              Last edited by popper; 04 December 2011, 01:03 AM.

              Comment


              • #27
                Originally posted by popper View Post
                while its nice that AMD are giving student developer Victor (Oliveira) A stipend form of salary to keep working on this code as im sure he is happy about that for his independent efforts, where is even a minuscule payment and hardware for
                the likes of Veerappan and his VP8+OpenCL acceleration?, in case you missed it notzed, Veerappan is the ONLY developer that has so far managed to actually provide a working prototype OpenCL video decode source and he continues to work on that as time allows
                Thanks for the recognition, popper.

                While my initial prototype was done as my master's thesis project, the last few months of development (Aug - Nov) were actually paid for by a mobile handset developer who I'd prefer not to name. That contract is up (the new code was still kept open source on my github repo), and I've settled into a position with a pharmaceutical research software company. I'll still keep poking around at the VP8/CL code as time allows, and hopefully one of these days I'll get the code accepted into the upstream repository so it doesn't bit rot from neglect.

                In the process, I've also come up with a few algorithmic changes that would allow a VP8 decoder to utilize as many CPU cores as a machine would realistically have in the loop filtering stage (one of the most intensive parts), so the multi-threaded CPU decoder could eventually benefit from the OpenCL work as well.

                Comment


                • #28
                  Is possible obtain more informations about that initiative?
                  For example which is the minimum OpenCL that will be supported?
                  Also, from what i read on the article here, but even on different article on the web, the focus is only on the Gpu, where are present performance problem for move data from the central memory on the video memory (situation that will be resolved once this transfer will be no more necessary, Amd BD2 and i don't known if is possible using Ivy Bridge) while using OpenCL is possible use even multi-core Cpu,
                  Gimp is a single core application so will be an huge benefit even if was possible only use multi-core Cpu.
                  So my questions is, after the work is done (march), how easy will be extend it so that it can be applied at the multicore CPU?
                  On this way will be possible avoid the need to be tied to proprietary drivers without lost much on performance (at least respect the modernl GPU included inside at the Cpu).

                  Comment


                  • #29
                    Originally posted by crazycheese View Post
                    DX11 is not better, it is closed proprietary library/API, officially pushed by company, widely known for Embrace, Extend, Extinguish and other non-ethical tactics, to the standard in computer graphics.
                    Regardless of ethics, it's technologically superior...

                    There are some "open hardware" CPU designs out there... But I'd wager you're using a CPU created by Intel or AMD, which only exist because of those company's closed source proprietary software they use to design their chips.. You can pull the ethics card all you want, and I don't disagree here with what you think of these bad ethics.. What I'm talking about though has nothing to do with ethics, it has to do with the design of the technology itself.. The core technology behind OpenGL is inferior to DX11, and the only reason you believe it's superior is because you're overplaying the ethics card.

                    I keep hearing why OpenGL is superior because it's supposedly so open.. But let's not forget that a lot of technologies in OpenGL requires patent licensing in order for you, or your users, to be able to use it.. From what I've seen, OpenGL is a *LONG* way away from what a lot of people claim it is, and the OpenGL consortium really doesn't seem to care. At least with DirectX, you can use all of the technologies (S3TC, floating point textures, etc.) without having to pay a dime to license any patents. Microsoft has already licensed it for you, or they're the patent holder themselves.

                    Comment


                    • #30
                      At least with DirectX, you can use all of the technologies (S3TC, floating point textures, etc.) without having to pay a dime to license any patents. Microsoft has already licensed it for you, or they're the patent holder themselves.
                      Yes, that will surely apply to the DX11 gallium state tracker

                      Comment

                      Working...
                      X