Announcement

Collapse
No announcement yet.

2013: A Good Year For Open-Source AMD?

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

  • #11
    I think he just hates me
    Last edited by bridgman; 24 May 2012, 02:18 PM.
    Test signature

    Comment


    • #12
      Originally posted by Nille View Post
      Q since when are you a AMD Hater. In the past you was the biggest AMD Fanboy in the forum and now you make a 180? turn.
      Q jumped on board the crazycheese train a while ago.

      Comment


      • #13
        Originally posted by bridgman View Post
        I think he just hates me
        Hows that possible? Canadian types are the most agreeable people on the planet. I could spit in your face and then you would certainly thank me for the human contact.

        Comment


        • #14
          Originally posted by smitty3268 View Post
          Q jumped on board the crazycheese train a while ago.
          What would you do in my shoes, when somebody offers to go opensource (1),
          clearly knows the shortcomings of development speed (which make the solution completely unattractive) (2),
          and denies ANY attempt to improve the situation or withdraws himself from that (3). Huh?

          How much different ideas have I thrown into Bridgeman camp? 1, 2, 3, 4? No, they are perfectly satisfied with the overall progress. At this rate R800 support will match catalyst in 2017, remember this.

          Mind you, I purchased two radeon cards for sake of opensource, the second dive was better, like going from completely unusable and crashing into slow as a fck. In two years. For same gen.
          Should I report that as bug? But the devs are already 101% busy, doing 1000x job which is perfect in Bridgeman's opinion. He thinks the formula is perfect, there is no need to correct anything, but I beg to differ.
          Remember, no external help is appreciated.

          Btw, first radeon (R100) is broken completely (KMS). I may upload errata soon, if someone cares (even for the sake of "radeon feature").

          Also, compare: http://openbenchmarking.org/result/1...DU3NzA&compare

          If Radeon driver Kickstarter project eventually gets rolling, that would be first ray of light.

          Where does my attitude towards gfx hardware and linux differs in general from that of Bridgeman? I don't consider Linux and GNU to be "construction yard of broken parts", I consider them to be an operating system.
          If the hardware is proprietary and works - fine, ok, usable.
          If it is opensource and broken and company is really in to fix it, taking everything from the community, maximizing the effort - even better. How to see it? Compare the development rate and attitude.
          If it is opensource and works - best case.

          The case with AMD? Opensource, broken and don't bother.

          The thing is, you DO NEED to pay to even test AMD driver because you need hardware. Multiply this by complete disinterest from their side and you have the idea.

          Why should I pay in this case? There are a lot of broken opensource projects, whose developers do not care about it(they even state it so). For me, such project belong to trash bin.

          Should I bother if help's not wanted? What J.C. said about throwing jewels in front of the pigs?

          Nvidia + Linux works for me as in "fine, ok, usable", so I have no option but to tolerate it. Burnt my fingers twice already, no thanks.

          Comment


          • #15
            Originally posted by Tgui View Post
            Hows that possible? Canadian types are the most agreeable people on the planet. I could spit in your face and then you would certainly thank me for the human contact.
            How many canadians do you need to screw the bulb in? Whole country?

            // Sorry, enough pointless talking.

            Comment


            • #16
              How much different ideas have I thrown into Bridgeman camp? 1, 2, 3, 4? No, they are perfectly satisfied with the overall progress. At this rate R800 support will match catalyst in 2017, remember this.
              The usual problem with OSS projects.
              Many guys have THE idea to bring the project forward, but usually those guys don't have the knowledge to implement it themself and the are angry, because their idea (beeing good or bad) isn't considered.

              Comment


              • #17
                Originally posted by Linuxhippy View Post
                The usual problem with OSS projects.
                Many guys have THE idea to bring the project forward, but usually those guys don't have the knowledge to implement it themself and the are angry, because their idea (beeing good or bad) isn't considered.
                No, no. Many guys have many different possibilities. Some have knowledge. Some can bug-test. Some can write review. Some can send cards. Some can purchase cards. Some can do marketing. Some can help rise the funds. Some can just cheer up. If the project is good and is liked by people, there is a ton of energy behind it.

                Right now, efforts to:

                report driver usage statistic back to hardware HQ
                improve the driver by means of optimization
                create ANY research on buyers behaviour in linux segment
                collect funds for the above
                create anything centralized

                have been blocked via chicken/egg problem shift or ignored.

                You offer to create all this on our own, aside from AMD, perhaps we should. I don't know. But this would mean they are definitely incapable of writing their own drivers (who cares if it will work).
                And we will also depend on them for hardware/documentation side, meaning they would control everything. Yet they don't step in. That's another thing, I don't understand - denial to innovate on linux side.

                Comment


                • #18
                  Originally posted by crazycheese View Post
                  What would you do in my shoes, when somebody offers to go opensource (1),
                  clearly knows the shortcomings of development speed (which make the solution completely unattractive) (2),
                  and denies ANY attempt to improve the situation or withdraws himself from that (3). Huh?
                  "go opensource" obviously means different things to different people. We said we would "support open source development efforts", ie provide documentation and/or sample code, provide a couple of developers to help with the work, and provide support to other developers by taking questions into our engineering groups on their behalf and providing answers where possible. You're obviously looking for something more, which is a fair request but not what we said we would do.

                  Originally posted by crazycheese View Post
                  How much different ideas have I thrown into Bridgeman camp? 1, 2, 3, 4? No, they are perfectly satisfied with the overall progress. At this rate R800 support will match catalyst in 2017, remember this.
                  Going from memory, I believe you provided two ideas.

                  1. An online registration form at amd.com where users could indicate whether they used open source or Catalyst drivers, with the hope being that responses would divert funds from Catalyst to open source. My response to this, which you ignored, was that we couldn't divert funds from the Catalyst Linux driver without hurting its usefulness for it's target market (3D workstation) and so the chances of it increasing funding for in-house open source development seemed very low (the risk of it *reducing* funding seemed higher).

                  2. One or more kickstarter-type projects aimed at specific improvements in the open source stack, whether driver-specific or generic. My response to this was essentially "seems like a good idea" but you seemed to be suggesting that AMD had to set those projects up and I never understood your reasoning for that.

                  Currently the proprietary drivers (all of them AFAIK) replace a lot of the open source graphics stack with proprietary interfaces. Until similar improvements are made in the generic open source framework, which *will* take a fair amount of time and effort, there will be gaps between open source and proprietary stacks. I am agitating internally for our proprietary driver folks to help where they can by pushing enhancements into the common framework where possible, and should know the outcome in early 2013.

                  Originally posted by crazycheese View Post
                  Mind you, I purchased two radeon cards for sake of opensource, the second dive was better, like going from completely unusable and crashing into slow as a fck. In two years. For same gen. Should I report that as bug?
                  If you are seeing "typical" performance deltas between open and Catalyst that is probably not worth reporting as a bug. If a specific app is behaving much slower than other comparable apps on the same driver that probably is worth reporting.

                  Originally posted by crazycheese View Post
                  But the devs are already 101% busy, doing 1000x job which is perfect in Bridgeman's opinion. He thinks the formula is perfect, there is no need to correct anything, but I beg to differ.
                  I don't know where you get this. What I have said is "this is the current plan, and we think it makes best use of the current budget we have to work with", which is *very* different from what you are saying.

                  Originally posted by crazycheese View Post
                  Remember, no external help is appreciated.
                  Huh ? Are you talking about external help as in developers & testers, or external help in the form of criticism and abuse ?

                  Originally posted by crazycheese View Post
                  Btw, first radeon (R100) is broken completely (KMS). I may upload errata soon, if someone cares (even for the sake of "radeon feature").
                  Probably worth filing (agd5f will know for sure) but only with complete system info. My understanding is that KMS needs a somewhat different set of system-specific quirks from UMS and that there probably are still some systems which have problems. agd5f and others are looking into the problems as they are identified, but I believe we are already at the point where most of the fixes have to be done "blind" because the problems can't be reproduced in house.

                  Originally posted by crazycheese View Post
                  Honestly, I'm not sure that raising frame rates from 113 fps to 400+ fps is the best thing for developers to be working on. I would rather see attention on the slowest apps (eg Warsow) first.

                  Originally posted by crazycheese View Post
                  If Radeon driver Kickstarter project eventually gets rolling, that would be first ray of light.

                  Where does my attitude towards gfx hardware and linux differs in general from that of Bridgeman? I don't consider Linux and GNU to be "construction yard of broken parts", I consider them to be an operating system.
                  Nor do I, does that mean we agree ? Where do you get these statements from anyways, they're certainly nothing I ever said.

                  Originally posted by crazycheese View Post
                  If the hardware is proprietary and works - fine, ok, usable.
                  If it is opensource and broken and company is really in to fix it, taking everything from the community, maximizing the effort - even better. How to see it? Compare the development rate and attitude.
                  I don't understand what you are saying here. Other than your specific suggestions, what do you think we are not taking from the community and where do you think we are not maximizing the effort ? Do you think the development rate is out of whack relative to number of developers ? Do you believe there is an attitude problem in the development community ?

                  Originally posted by crazycheese View Post
                  The case with AMD? Opensource, broken and don't bother.

                  The thing is, you DO NEED to pay to even test AMD driver because you need hardware. Multiply this by complete disinterest from their side and you have the idea.
                  Usual question -- by "complete disinterest" do you mean "bridgman didn't agree with your suggestions" or something more ?

                  Originally posted by crazycheese View Post
                  Why should I pay in this case? There are a lot of broken opensource projects, whose developers do not care about it(they even state it so). For me, such project belong to trash bin.
                  Have developers on the open source graphics drivers stated they do not care about it, or are you just talking about your own interpretation ?
                  Last edited by bridgman; 24 May 2012, 04:51 PM.
                  Test signature

                  Comment


                  • #19
                    Originally posted by crazycheese View Post
                    No, no. Many guys have many different possibilities. Some have knowledge. Some can bug-test. Some can write review. Some can send cards. Some can purchase cards. Some can do marketing. Some can help rise the funds. Some can just cheer up. If the project is good and is liked by people, there is a ton of energy behind it.

                    Right now, efforts to:

                    report driver usage statistic back to hardware HQ
                    improve the driver by means of optimization
                    create ANY research on buyers behaviour in linux segment
                    collect funds for the above
                    create anything centralized

                    have been blocked via chicken/egg problem shift or ignored.
                    Sorry, but none of these have been blocked or ignored. You're writing as if *you* tried to do something and *we* blocked it, which is not the case at all.

                    If you're saying "hey I posted in a forum that AMD should do all these things and one of their guys thought what they were doing today was a better use of the people they had" that's probably fair.
                    Test signature

                    Comment


                    • #20
                      Originally posted by bridgman View Post
                      "go opensource" obviously means different things to different people. We said we would "support open source development efforts", ie provide documentation and/or sample code, provide a couple of developers to help with the work, and provide support to other developers by taking questions into our engineering groups on their behalf and providing answers where possible. You're obviously looking for something more, which is a fair request but not what we said we would do.
                      I'm perfectly fine with that.

                      On the other hand, are there any ideas how to attract more developers to contribute the open graphics stack?
                      Of course I can only speak for myself and I consider the threshold to enter this "business" to be really high.
                      To be honest, I wouldn't even know where to start. Reading the specs? Start to understand the available code first?

                      Lowering this threshold by some introductions and tutorials would be great imho.
                      Not sure the current devs agree. After all someone would need to write them.

                      Those blog posts by Alex were worth a read.


                      Comment

                      Working...
                      X