Announcement

Collapse
No announcement yet.

Radeon On Linux 3.8: Minor For Now, New Code Coming

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

  • Radeon On Linux 3.8: Minor For Now, New Code Coming

    Phoronix: Radeon On Linux 3.8: Minor For Now, New Code Coming

    Alex Deucher of AMD emailed in a pull request on Friday afternoon for the Radeon graphics driver that will end up as part of the DRM pull for the Linux 3.8 kernel...

    http://www.phoronix.com/vr.php?view=MTI0NTg

  • #2
    As far as the minor highlights of this pull request, there's support for ELD registers for passing audio information to sound hardware
    This mean TrueHD/DTS-HD MA/multichannel LPCM bitstreaming support?

    Comment


    • #3
      Originally posted by RussianNeuroMancer View Post
      This mean TrueHD/DTS-HD MA/multichannel LPCM bitstreaming support?
      Yes, that's exactly what it means. But we still need to add the matching support on the alsa side.

      Comment


      • #4
        BS as ever

        i'm damn sure this news has nothing to do with realworld performance... still the opensource amd graphics will lag miles behind the propriatary catalyst. i'm running propriatry catalyst on e-350 on windows 7 and the very latest open source amd graphics (kernel 3.6 + mesa 9 + drm 2.40) on linux, in a dual boot system. open source is so damn bad, shoddy and inefficient that it's no more an option at all....

        Comment


        • #5
          @mammath:

          Just keep this pseudo knowledge for yourself. You did not understand a single word of what this commit brings for OSS using people.

          @Deathsimple:
          Thank you very much for getting these registers out and for staying on track. Thank you!

          Comment


          • #6
            fritsch, perhaps you did not get the point. very sad.

            i know the latest development will bring much difference to radeon open source graphhics. no doubt about it. but when compared to catalyst (feature, power usage and performance) it'll still lag miles behind.
            fyi, i've using using amd, intel and nvidia platforms for quite a long. and i know the pace of amd's open source endeavors, catalyst always takes lead in big leaps. imo, amd open source is just a hype and false good will seeking effort.

            intel = best in open source graphics
            nvidia = no opensource efforts, but the propriatary drivers are very good on linux platform
            amd = shoddy open source graphics + shoddy implementation of catalyst

            only fools would use open source amd graphics drivers....

            please consider drivers, firmware, and other opensource/propriatary software as tools, not as a cult or religion. performance, stability, power usage and usability matters most to the users where amd fails in both open source and proprietary. amd doesn't give a damn to the open source but anyways its propriatry graphics work better than amd's. hope you got my point.
            Last edited by manmath; 12-08-2012, 05:14 AM. Reason: typo

            Comment


            • #7
              I did not say you are wrong. I only said, this is the wrong position to make such comments.

              Deadsimple is doing a great job in bringing things forward. Now he came up with the Registers for Bitstream Audio, which means DTS-HD, TrueHD, LPCM for all AMD linux hardware via HDMI. This is something we are really happy about. As those registers are known now - they can be used within ALSA. And as you see those can be used to even make it useful with fglrx.

              The other things you mention are true, yeah no question. But it does not help, to bring staff into bad mood, by always seeing the bad points, when there is actually some progress.

              I am doing xvba development since over a year now. And yes, what you say is right, there is a whole lot to be improved on the OSS side and on the closed side. But I see lately a good relationship to AMD, things change and this shall be honored.

              For this reason don't turn such good news into bad news, as it is a big process.

              Comment


              • #8
                Originally posted by fritsch View Post
                But I see lately a good relationship to AMD, things change and this shall be honored.
                Thanks for the devs who are working really hard to make something good. But, my comment was not mean for them, it's for AMD. It does put efforts on OSS but the result is so minismal that it can be ignored. For example what's the use of open source drivers for certain chips if they are ~40% feature complete and ~20% performing after 3 years of release of those particular devices. Does it really make any sense (on the part of AMD, not the devs)?

                AMD, please compare the case with Intel and nVidia. But i doubt if you really care the meagre linux/unix userbase.

                Comment


                • #9
                  Originally posted by manmath View Post
                  intel = best in open source graphics
                  I don't really agree. How lang has it taken for the GMA X4500 to implement vaapi? What about the poulsbo drivers?

                  And it's probably my fault for running bleeding edge, but...
                  Code:
                  [29894.653601] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
                  [29900.736978] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
                  [29912.424243] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
                  [29923.485522] [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
                  Code:
                  [30088.599017] INFO: task X:20131 blocked for more than 120 seconds.
                  [30088.599017] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
                  [30088.599019] X               D
                  [30088.599020]  ffff880800680460 
                  [30088.599021]     0 20131  20127 0x00400000
                  [30088.599022]  ffff88080969b628
                  [30088.599023]  0000000000000086
                  [30088.599024]  ffff880754248000
                  [30088.599024]  ffff88080969bfd8
                  [30088.599025]  ffff88080969bfd8
                  [30088.599026]  ffff88080969bfd8
                  [30088.599026]  ffff88080b529800
                  [30088.599027]  ffff880754248000
                  [30088.599028]  0000000000000000
                  [30088.599028]  ffff8808007dfae0
                  [30088.599029]  0000000000000400
                  [30088.599029]  0000000000001262
                  [30088.599030] Call Trace:
                  [30088.599039]  [<ffffffff81473da9>] schedule+0x29/0x70
                  [30088.599059]  [<ffffffffa0472bc5>] intel_crtc_wait_for_pending_flips+0x75/0xd0 [i915]
                  [30088.599063]  [<ffffffff810687b0>] ? abort_exclusive_wait+0xb0/0xb0
                  [30088.599071]  [<ffffffffa0475c67>] ironlake_crtc_disable+0x87/0x890 [i915]
                  [30088.599075]  [<ffffffff8100bd33>] ? __switch_to_xtra+0x163/0x1a0
                  [30088.599082]  [<ffffffffa047a064>] intel_crtc_disable+0x34/0x130 [i915]
                  [30088.599090]  [<ffffffffa047ccc2>] intel_set_mode+0x262/0xa50 [i915]
                  [30088.599092]  [<ffffffff81473795>] ? __schedule+0x435/0x920
                  [30088.599095]  [<ffffffff81026065>] ? native_smp_send_reschedule+0x45/0x60
                  [30088.599098]  [<ffffffff81076a94>] ? resched_task+0x64/0x70
                  [30088.599101]  [<ffffffff8107fd39>] ? check_preempt_wakeup+0x1c9/0x270
                  [30088.599104]  [<ffffffff81077665>] ? check_preempt_curr+0x85/0xa0
                  [30088.599106]  [<ffffffff8107484a>] ? ttwu_stat+0xaa/0x120
                  [30088.599117]  [<ffffffff81250f01>] ? vsnprintf+0x461/0x600
                  [30088.599120]  [<ffffffff81251144>] ? snprintf+0x34/0x40
                  [30088.599126]  [<ffffffffa047dc06>] intel_crtc_set_config+0x756/0x9b0 [i915]
                  [30088.599132]  [<ffffffffa0312de2>] drm_fb_helper_set_par+0x82/0x100 [drm_kms_helper]
                  [30088.599134]  [<ffffffff81077655>] ? check_preempt_curr+0x75/0xa0
                  [30088.599138]  [<ffffffff8128cbe5>] fb_set_var+0x1a5/0x470
                  [30088.599140]  [<ffffffff81001829>] ? __switch_to+0x3d9/0x4e0
                  [30088.599144]  [<ffffffff81299421>] fbcon_blank+0x1e1/0x2e0
                  [30088.599148]  [<ffffffff812f1006>] do_unblank_screen+0xa6/0x1d0
                  [30088.599151]  [<ffffffff812e66b5>] complete_change_console+0x65/0xf0
                  [30088.599154]  [<ffffffff812e7902>] vt_ioctl+0x11c2/0x12c0
                  [30088.599161]  [<ffffffffa02c4022>] ? drm_ioctl+0xf2/0x570 [drm]
                  [30088.599163]  [<ffffffff812dc048>] tty_ioctl+0x258/0xb40
                  [30088.599166]  [<ffffffff814785c4>] ? __do_page_fault+0x2d4/0x590
                  [30088.599169]  [<ffffffff81179937>] do_vfs_ioctl+0x97/0x530
                  [30088.599171]  [<ffffffff81179e61>] sys_ioctl+0x91/0xb0
                  [30088.599174]  [<ffffffff8147c7dd>] system_call_fastpath+0x1a/0x1f

                  Comment


                  • #10
                    Originally posted by manmath View Post
                    please consider drivers, firmware, and other opensource/propriatary software as tools, not as a cult or religion. performance, stability, power usage and usability matters most to the users where amd fails in both open source and proprietary. amd doesn't give a damn to the open source but anyways its propriatry graphics work better than amd's. hope you got my point.
                    Except that OSS drivers, even at this point, are better than the closed blob in some regards. Consider their rock-solid stability, perfect integration with the system, light weight and standards compliance. There are plenty of people for whom these are more important than raw power.

                    Originally posted by ChrisXY View Post
                    I don't really agree. How lang has it taken for the GMA X4500 to implement vaapi? What about the poulsbo drivers?
                    I don't think it's all that relevant now, as X4500 is running with VA-API just fine at the moment. And Poulsbo is not Intel's, it's the work of Imagination Technologies.

                    Comment


                    • #11
                      Originally posted by GreatEmerald View Post
                      Except that OSS drivers, even at this point, are better than the closed blob in some regards. Consider their rock-solid stability, perfect integration with the system, light weight and standards compliance. There are plenty of people for whom these are more important than raw power.
                      Yes, and that is why I use them, even though I consider myself to very much be a gamer. The performance and power management could be improved, but other than that I find it to be a far superior experience to any proprietary blob driver (the PM is not even that much of a problem for me when using the power profiles). The maintenance burden is just that much lower. On a laptop I am perfectly fine and may even prefer an Intel GPU but on the desktop I still want the performance benefits of a dedicated chip, which can even be felt using the FOSS drivers. I am happy with my AMD card and I am continually grateful for the work that is still going in to better support it.

                      Comment


                      • #12
                        The only thing I'd like is better power management, my card is running pretty hot (but I never compared with the proprietary drivers, so there might be not much left to do).

                        Comment


                        • #13
                          Originally posted by manmath View Post
                          Thanks for the devs who are working really hard to make something good. But, my comment was not mean for them, it's for AMD. It does put efforts on OSS but the result is so minismal that it can be ignored. For example what's the use of open source drivers for certain chips if they are ~40% feature complete and ~20% performing after 3 years of release of those particular devices. Does it really make any sense (on the part of AMD, not the devs)?

                          AMD, please compare the case with Intel and nVidia. But i doubt if you really care the meagre linux/unix userbase.
                          What does AMD owe you and the rest of the FOSS Community? Not a damn thing. Intel has a garbage GPU solution, but hey! It's open source. Intel has tens of billions to toss at projects while AMD is working getting back in the black [which they will this upcoming 3rd quarter] yet you whine about AMD not opening up their proprietary stack for Intel to covet?

                          Grow up. Businesses have to make money.

                          Wake me when Linux creates a stable ABI and has desktop penetration above a pimple. All the server penetration is being driven by AMD, Intel, IBM, Oracle whose data clients could careless about advanced Audio and Video performance for gaming.

                          If AMD hits its stride like they are positioning for 2013 you'll see a much improved Catalyst and FOSS will get a lot of freebie coding, but keep whining they don't pull their weight. I'm sure they're more interested in Valve as those guys making their games scream on Windows, OS X and Linux for the Catalyst driver means more GPGPU sales and more jobs for AMD to have a future against Intel.

                          Comment


                          • #14
                            Originally posted by Marc Driftmeyer View Post
                            Wake me when Linux creates a stable ABI and has desktop penetration above a pimple.
                            Absolutely right! Agree 100%. But a stable abi from Linux and a sizeable desktop penetration is just a pipe dream. Intel knows it yet releases better drivers. Users expect something similar from AMD. Most (include me) are aware of AMD's opensource efforts. No denying. But the reality is it is not substantial yet to make OSS AMD drivers a compelling reason. Well, I'll keep waiting for stable/better drivers from AMD (doing so for last couple of years)!

                            Comment


                            • #15
                              Originally posted by pankkake View Post
                              The only thing I'd like is better power management, my card is running pretty hot (but I never compared with the proprietary drivers, so there might be not much left to do).
                              You can try smth like that with opensource AMD drivers:
                              echo profile > /sys/class/drm/card0/device/power_method
                              echo low > /sys/class/drm/card0/device/power_profile

                              This will downclock 1st GPU to lowest frequency it can do. It's enough for desktop effects, etc. But 3D games will suffer a lot.3
                              If you have more than 1 GPU, replace card0 with desired target.

                              To get full clocks again, you can use
                              echo high > /sys/class/drm/card0/device/power_profile
                              This forces GPU to run at full speed. Desirable for 3D games, etc.

                              There is also dynpm method instead of profile (echo dynpm > /sys/class/drm/card0/device/power_method) which meant to switch frequencies automatically, based on load. But it's seriously borked at the moment. For me it only toggles to medium power state (which is 4 times higher freq than low so a lot of heat for nothing) and can cause screen to blink a bit on reclocking. Also GPU load detection isn't very great and sometimes GPU does not switches to highest clock even under serious load. Hence less FPS in games, etc. In fact I fail to understand why AMD can't spend some time reworking this code. It's probably not the hardest part but one of most annoying issues so far.
                              Last edited by 0xBADCODE; 12-09-2012, 11:33 AM.

                              Comment

                              Working...
                              X