Announcement

Collapse
No announcement yet.

Reasons For Losing Motivation In Wayland

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

  • #31
    Originally posted by Delgarde View Post
    Unit testing is *not* a magic bullet. It's certainly a very useful tool for developers, but having automated tests isn't going magically transform your development process such that bugs and breaks never occur. There's plenty of crap code out there with 100% test coverage...
    Unit testing will make sure that a specific function works and will continue to work. And if you, for every single bug that passes your testing suite creates new test to cover that bug you can make sure that you will never get the same bug again.
    With a huge testing suite you should encounter far less regressions than a untested project. However for bugs in new code the test doesn't help very much.

    Comment


    • #32
      Originally posted by brosis View Post
      I had a Xi 1554 with Mobility X1900 and it drove 3D desktop just fine in low profile.
      Good for you. For RadeonHD 4250 users (and that GPU is very common in laptops), it is not so rosy.

      Comment


      • #33
        Originally posted by valeriodean View Post
        Buy an Intel card. Their work as Open Source Development Centre deserve it.
        Yes, Intel drivers work quite well.

        About nvidia, there is always a middle finger available for them and for their ancient, cronic and stupid way of thinking.
        Wayland is the future, with or without nvidia.
        Withtout proprietary drivers, Wayland is a lost cause. Nvidia and even AMD open source drivers are still no match for their proprietary counterparts and they won't be for a very long time. There are just too many AMD and Nvidia cards out there so without the proprietary drivers, people will just not use Wayland.

        Comment


        • #34
          Originally posted by DanL View Post
          Good for you. For RadeonHD 4250 users (and that GPU is very common in laptops), it is not so rosy.
          Well, I have Mobility 3870 waiting for blind VGA BIOS reflash as replacement. I said, I "had", the X1900 needs reballing and that has nothing to do with opensource driver.
          So, 3870 is 75Watt class, similar to Mobile HD4670.

          The 4250 is a piece of joke,.. how much, 15 Watts? Check your cooling system, replace thermal pads with 5 or 7kW ones; above and below the card; replace thermal compound.
          It can be that the card is actually dying from temperature anyway, even on standard low profile; but due to catalyst having control over thermal sensors, it manages to constantly throttle the performance down because it detects card being in critical situation even on low clocks.
          You can clean the memory chips with rubber, gpu/cpu die with isopropyl and polish the cooling interface with abrasive milk and sponge.

          Originally posted by Temar View Post
          Withtout proprietary drivers, Wayland is a lost cause. Nvidia and even AMD open source drivers are still no match for their proprietary counterparts and they won't be for a very long time. There are just too many AMD and Nvidia cards out there so without the proprietary drivers, people will just not use Wayland.
          AMD radeon is a match for proprietary catalyst, we just need the damn dynpm and we are happy. Then, you can really be free to use what you want.
          Last edited by brosis; 06-12-2013, 05:24 PM.

          Comment


          • #35
            Originally posted by chris200x9 View Post
            Personally I'm losing interest because it only works on the open drivers. This wouldn't be so much of a problem if they had PM functionality since the performance is already good enough in the open drivers, but without PM the open drivers are pretty much useless for everything except starting house fires.
            That's the main reason I use proprietary video drivers (fglrx), because I trust them somewhat better to have proper clock frequency and fan control. While Linux has been my main OS that I use for almost everything, I have a half decent video card (HD 5870 which was right up there at the time of purchase) for playing games in Windows that I don't want ruined. (can't afford to upgrade that any time soon). I know that card works well enough with open source drivers now (it didn't at all when I bought it) but I still don't want to take a chance on cooling controls.

            I've gone through a few Nvidia cards in the past and I wonder if using "nv" (and buggy proprietary Nvidia with known fan control problems at one time... I used to have to set it manually for a while) contributed to their short life.

            Comment


            • #36
              Originally posted by brosis View Post
              The 4250 is a piece of joke,.. how much, 15 Watts? Check your cooling system, replace thermal pads with 5 or 7kW ones; above and below the card; replace thermal compound.
              Damn it, you are so a FOSS zealot that you are not able to read properly and try to make up things, only that you don't have to damage the image that you have made up of the free drivers in your mind. I have not said the the GPU is overheating, I said that the system is overheating. Read: The CPU alone runs with 15C higher temperature on radeon than on Catalyst. And no, there is no throttling at all. And yes, I know how to maintain cooling systems (man, I probably have build and repaired more PC systems than you have ever seen in your live).

              Maybe I have to put this in simple words for you: This driver problem. No hardware problem. Hardware worky worky OK.
              Up to my ignore list.

              Comment


              • #37
                Originally posted by brosis View Post
                AMD radeon is a match for proprietary catalyst,...
                No, they are not, even if you put aside things like power management, OpenGL, OpenCL or UVD support. The open source drivers are not even a match when it comes to basic 3D performance. Every single Phoronix comparison confirms this: http://www.phoronix.com/scan.php?pag...pril2013&num=1

                When looking at power management, UVD, OpenGL and OpenCL support it gets even worse. Sure, many of these features are available on some level, but they are not en par with the proprietary drivers. So don't lie to yourself because it won't change anything. Without the support of proprietary drivers, Wayland will simply be ignored by the users.

                Comment


                • #38
                  Originally posted by Temar View Post
                  Yes, Intel drivers work quite well.



                  Withtout proprietary drivers, Wayland is a lost cause. Nvidia and even AMD open source drivers are still no match for their proprietary counterparts and they won't be for a very long time. There are just too many AMD and Nvidia cards out there so without the proprietary drivers, people will just not use Wayland.
                  The history love repeat itself.
                  Why nvida started to develop driver for X instead of say "there are too many windows os out there, people will just not use linux"?
                  The reason that push nvidia that time will push again when "the reason" will switch to wayland.

                  Comment


                  • #39
                    Originally posted by Pajn View Post
                    Unit testing will make sure that a specific function works and will continue to work. And if you, for every single bug that passes your testing suite creates new test to cover that bug you can make sure that you will never get the same bug again.
                    Strictly speaking, no - it makes sure that that function will continue to satisfy the constraints of the test. And that assumes that the test is a) testing the right thing, and b) testing it correctly. I'm a fan of unit testing, but I've seen too many badly-written tests for me to believe that it will solve all problems. More than once, I've fixed user-reported bugs, only to see tests break because they were testing the broken behaviour. And I've introduced regressions that were uncaught despite high test coverage, because the tests weren't really testing anything beyond that the code didn't crash.

                    They're a very good tool for encouraging clean design and for detecting regressions, but only if the tests themselves are good. If they're not, mostly all they give you is false confidence.

                    Comment


                    • #40
                      Originally posted by Pajn View Post
                      Unit testing will make sure that a specific function works and will continue to work. And if you, for every single bug that passes your testing suite creates new test to cover that bug you can make sure that you will never get the same bug again.
                      With a huge testing suite you should encounter far less regressions than a untested project. However for bugs in new code the test doesn't help very much.
                      First, bugs can be within the integration of components. Unit testing tests components. That's why it's not a magic bullet.

                      And about new code, if you design before implementing, you can make tests before implementing, making sure most of the possible bugs in that component are out of the way.

                      Comment


                      • #41
                        Originally posted by Vim_User View Post
                        Damn it, you are so a FOSS zealot that you are not able to read properly and try to make up things, only that you don't have to damage the image that you have made up of the free drivers in your mind. I have not said the the GPU is overheating, I said that the system is overheating. Read: The CPU alone runs with 15C higher temperature on radeon than on Catalyst. And no, there is no throttling at all. And yes, I know how to maintain cooling systems (man, I probably have build and repaired more PC systems than you have ever seen in your live).

                        Maybe I have to put this in simple words for you: This driver problem. No hardware problem. Hardware worky worky OK.
                        Up to my ignore list.
                        You are DanL?... I was talking to DanL.

                        From you I asked please the bug report. But instead you got aggressive. Fine, you can't fix your own notebook so, then ask someone to do it.
                        Last edited by brosis; 06-12-2013, 06:37 PM.

                        Comment


                        • #42
                          Originally posted by Temar View Post
                          No, they are not, even if you put aside things like power management, OpenGL, OpenCL or UVD support. The open source drivers are not even a match when it comes to basic 3D performance. Every single Phoronix comparison confirms this: http://www.phoronix.com/scan.php?pag...pril2013&num=1

                          When looking at power management, UVD, OpenGL and OpenCL support it gets even worse. Sure, many of these features are available on some level, but they are not en par with the proprietary drivers. So don't lie to yourself because it won't change anything. Without the support of proprietary drivers, Wayland will simply be ignored by the users.
                          3D: http://openbenchmarking.org/result/1...500792103#r-14 70-110% of catalyst.
                          PM: yes, we miss dynpm, no one denies that
                          OpenCL: I don't use it, really; but people say it gets near 60% of fglrx already?
                          UVD: AMD recently released UVD support for HD4xxx and up via VDPAU

                          I switched to AMD directly from Intel drivers, so I don't want fglrx right from start, and the open drivers work pretty flawless.
                          Hell, HD5850+opensource on my desktop machine is good enough for all games, including WINE.

                          Even Darxus, the person complaining in the article, got his HD6850 + opensource, switching from nvidia + nvidia and he used it to test Wayland. So Wayland runs perfectly without proprietary drivers. If you set proprietary driver as requirement to Wayland, then its your own opinion.
                          Last edited by brosis; 06-12-2013, 06:30 PM.

                          Comment


                          • #43
                            Perhaps related...

                            Darxus is a Ubuntu user.

                            I'm not saying that has anything to do with this, but it does seem somewhat coincidental that he's publicly complaining about Wayland right after Mir is announced and it's obvious his distro isn't going to pursue using it.

                            Comment


                            • #44
                              Originally posted by smitty3268 View Post
                              Darxus is a Ubuntu user.

                              I'm not saying that has anything to do with this, but it does seem somewhat coincidental that he's publicly complaining about Wayland right after Mir is announced and it's obvious his distro isn't going to pursue using it.
                              Yes, he is, but he is pretty sane (and biker - bikers are awesome ). This is what I found, I can say I am completely with him on Ubuntu; but the whole case of three unfixed, seemingly unrelated bugs is more complex, because those bugs break in XWayland, and upon being fixed expose even deeper bugs.

                              I wanted to post in "Wayland situation" that "its a display server, a huge thing, anchored deep, that never goes without bugs; and those get in our face first (after developers) and we have to deal with them, inevitably(as in reporting)"; but I decided against, because I seemed logical to me (systemd, pulseaudio, even unity).

                              Comment


                              • #45
                                Originally posted by brosis View Post
                                Hi MuPuF! I have suggested prior that you guys use kickstarter-ish scheme for this: realistic goals - money - donations - mercenary work.
                                My vision is that there is no need to posses 100 developers, there is need for good organization where personal interests (I am *not* talking about altruism here) will drive the development. And 100 developers will come.
                                Yes, you already told me your vision. The thing is, paying the devs wouldn't make them work faster, trust me. As for the mercenary work, it won't work as it is almost impossible to predict how long it will take to implement a feature (unless it is a menial task) and I also don't want people to be eager about pushing something just because they want/need the money.

                                Originally posted by brosis View Post
                                For example, you could start to use phoronix mass media, to post monthly refresher on:
                                - accomplished goals (essentially changelog)
                                Yeah, that's the point of the TiNDC. No-one seems interested in writing one anymore because there isn't much to report anyway and phoronix tend to relay all the big news anyway.

                                Originally posted by brosis View Post
                                - ongoing efforts, requiring attention (for contributors)
                                See the status matrix: http://nouveau.freedesktop.org/wiki/FeatureMatrix/

                                Originally posted by brosis View Post
                                - required efforts, requiring known specific donation (for supporters and mercenary developers)
                                - hardware requirements, for those interested in hardware donation
                                Whenever someone has a donation to make, please propose on IRC.

                                Originally posted by brosis View Post
                                - testing USB images, for people interested in ability to run your tests on their nvidia hardware, effortlessly and without interaction; and then submit data to you.
                                The thing is that we have almost no hardware tests even though we are writing more and more of them (not sure you appreciate how much work this represents).
                                Making USB images are useless as long as we don't have enough tests. If someone is interested in writing more, be my guest, this is a very nice way to learn about the hardware!

                                Comment

                                Working...
                                X