Announcement

Collapse
No announcement yet.

Pinos Is For Linux Video What PulseAudio Is For Audio

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

  • #31
    Ignoring the idiots who bash PA for the extra features, bug fixes in ALSA, convenience and obviously easier lives to all the programmers out there for the cost of 50ms of delay (oh no, the crime of waiting an extra 50ms for that Miley Cirus song!!), I see this as being fairly good.

    One capture laptop, sending multiple streams to other servers which can then do stuff all over standardised APIs which will be supported by a great majority of the webcam software, plugging and switching cams on-the-go without so much fiddling on the active recording software.

    I do see a slight problem with the fact that at higher resolutions some webcams capture at lower framerates, meaning that you can't capture at higher rates but lower resolutions for other apps.. Obviously processing video is a CPU intensive task and needs to be done in the best possible way, but as we have no idea on how it will be approached yet (I haven't seen anything on it) then you cannot possibly criticise it for anything yet apart from wanting to add a standard and ease use.

    Comment


    • #32
      Originally posted by DMJC View Post
      This is probably going to end up bloated and slow like Totem is vs gnome-mplayer. Honestly I think that the correct way to handle a video pipelining system, should be to implement quartz composer/corevideo via gnustep/opal. The Apple video pipeline is the best one that exists. It's not going to suddenly be beaten by Linux developers trying to reinvent the wheel.
      Where do you think many inspirations came from? Lets remember quartz composer have features specifically made for Apple OS X. The source code of Pinos is available right now for testing purpose so try before making early conclusions and come with constructive feedback to improve Pino.

      Comment


      • #33
        Originally posted by DMJC View Post
        This is probably going to end up bloated and slow like Totem is vs gnome-mplayer. Honestly I think that the correct way to handle a video pipelining system, should be to implement quartz composer/corevideo via gnustep/opal. The Apple video pipeline is the best one that exists. It's not going to suddenly be beaten by Linux developers trying to reinvent the wheel.
        Ok, i just played this (http://downloads.4ksamples.com/%5B21...mples.com).mp4) video in totem using the vaapi backend. Even allowing for the extracopy per frame, it's still only using at most 20% cpu i5-4210u.

        Comment


        • #34
          Originally posted by psychoticmeow View Post

          Oh no! Using CPU cycles to enable functionality! When will this insanity ever end?
          These are probably the people that buy a high end CPU and GPU only to run twm and conky so they can post screenshots of their amazing hardware doing absolutely nothing.

          Comment


          • #35
            Originally posted by duby229 View Post
            That's exactly what PAs problem. They try to brute force what shouldn't be done and ended up with an audio server with 50ms to 2s of latency and no possible way to predict where at in the middle it might be. The only reason people don't see it is because of gigantic buffers allocated by the -playback- software to try and hide it.
            From what I remember, PulseAudio, by design, tries to go for as much latency as possible, unless the app specifically requests a lower value. It's not possible to flat out say "latency is bad", because it's simply a trade-off. If you want to go for minimal latency, you will incur maximum energy consumption. If you want to minimize energy consumption, you have to ramp up latency.

            Seeing as most casual PC and laptop users have little need for ultra-low latency, but would greatly appreciate low energy consumption, I think PA has made the best choice it could.

            Comment


            • #36
              Well I have to say I totally hate unnecessary power and cpu time consumption and that is the reason I don't use pulseaudio on my machines. If I ever need a separate sink to play the audio (e.g. of a specific application) I usually set it up on asoundrc.
              I do screencast too, with ffmpeg, so no problems on that too.
              That said I probably won't use or follow pinos because I don't feel it's actually needed (pulseaudio may have its reasons to exist on the other hand, just saying)

              Comment


              • #37
                Originally posted by tinko View Post


                I can't help to notice the apparent contradiction of you saying that you have no need for pulse audio on one hand but on the other hand you speak of all the pain one has to go through without it. If I misunderstood you here, you can clear that up of course.

                I definitely remember the frustration before pulseaudio, especially with software outside of the standard repo. I remember having no audio after using audio in the flash player or skype and having to restart after using certain apps that messed up the sound...
                My system was working perfectly with ALSA+dmix until, one day, about six months ago, my system pseudo-spontaneously (I assume it was due to a package update) decided that the unconnected HDMI audio lines (HDMI-to-DVI cable) in my nVidia card's HDMI port were default rather than my Intel HDA onboard analog audio out.

                Given that I didn't have time to re-teach myself how to spend hours debugging a custom asoundrc to get "default asoundrc except with a different default audio device", I restored the default Lubuntu PulseAudio configuration and I've been living with the 4-10% CPU hit (while audio is playing, Athlon II X2 270) and having to set PULSE_LATENCY_MSEC=60 to fix crackling in almost every closed-source game I play.

                (In some rare instances, the only way to get working audio in an ALSA app is to launch the app via pasuspender... especially old 1.2.x-era versions of Wine needed because modern 32-bit versions broke compatibility with some 16-bit Windows apps like the game BrickLayer. I think, for Audacious Media Player, I had to switch from the native Pulse output to the ALSA->libalsa->Pulse output to avoid stutter-crackle problems.)

                Hopefully, when my more pressing concerns are taken care of, I'll be able to make time to go back to dmix.

                As for frustration, I remember that, but it stopped years ago, back around when PulseAudio had just come out in its initial horrendously buggy state. I've had years of Everything Just Works? since then.

                Comment


                • #38
                  Originally posted by FuturePilot View Post

                  These are probably the people that buy a high end CPU and GPU only to run twm and conky so they can post screenshots of their amazing hardware doing absolutely nothing.
                  Not necessarily. Some of us want to run emulators on a near-silent system because we have to sleep near it but want to leave it on. That means setting an upper limit on TDP (eg. 65W CPU + nVidia GeForce GTX750) and then picking the highest per-core speed available (even if that means sacrificing number of cores).

                  ...also, I run AMD chips even when Intel is beating them because there have been proof-of-concept Intel Management Engine rootkits and not all BIOSes properly disable AMT when not in use. (AMD is only just starting to add an ARM TrustZone core, so you can still contact them and ask for a list of chips that have TrustZone so you can avoid them.)

                  Comment


                  • #39
                    Oh man, this is real nice. I just recently had the issue where I couldn't use the webcam to chat at the same time as using ZoneMinder to use it as a security camera. It's not that you'd want both specifically at the same time, but the chat won't be able to use the camera if it's already in use by ZoneMinder. And if it could, then ZoneMinder would suddenly get black frames as output and start all alarms.
                    Unfortunately it will probably be a long, long time until this hits CentOS. Maybe shorter for openSUSE 42 though, hmm.

                    Comment


                    • #40
                      Originally posted by ssokolow View Post
                      Given that I didn't have time to re-teach myself how to spend hours debugging a custom asoundrc to get "default asoundrc except with a different default audio device", I restored the default Lubuntu PulseAudio configuration and I've been living with the 4-10% CPU hit (while audio is playing, Athlon II X2 270) and having to set PULSE_LATENCY_MSEC=60 to fix crackling in almost every closed-source game I play.
                      Something seems off here. On Arch Linux with the default PulseAudio config it uses less than 1% of the CPU when playing audio.

                      Comment

                      Working...
                      X