Originally posted by alexcortes
View Post
Announcement
Collapse
No announcement yet.
Skype 4.3 For Linux Released With Updated UI
Collapse
X
-
-
Originally posted by Akka View PostBut why are dmix better than pulseaudio. As I understands it they are both daemons?
so it's almost like a shared library
pa is a daemon that gets sound from the app either through alsa or by a shared library
shared library is recommended but basic
pa is also very slow since it uses "secret rabbit code" library for sample rate conversion
so the sound goes app (int's) -> alsa (converts to float) -> pa buffer then conversion -> "zero-copy" playback (where it gets converted back to int)
shared library shortens that path by trowing out one(or two, depends) copy
it's also at version 5 already, when you'd think version 1 would be stable
Comment
-
Originally posted by gens View Postdmix is a plugin for alsa
so it's almost like a shared library
pa is a daemon that gets sound from the app either through alsa or by a shared library
shared library is recommended but basic
pa is also very slow since it uses "secret rabbit code" library for sample rate conversion
so the sound goes app (int's) -> alsa (converts to float) -> pa buffer then conversion -> "zero-copy" playback (where it gets converted back to int)
shared library shortens that path by trowing out one(or two, depends) copy
it's also at version 5 already, when you'd think version 1 would be stable
Comment
-
Originally posted by atari314Udev, dbus, avahi, pulse audio, systemd...
Comment
-
Originally posted by jbernardo View PostPlease stop wasting oxygen. Thank you.
Originally posted by jbernardo View PostAll cancers ruining the linux OS for the users. 99.9% of issues I have with linux is one of these. At least pulseaudio can be removed easily as soon as it starts interfering with the sound.
Comment
-
Originally posted by jimbohale View PostWhat? You agree that we should have a sound system that sucks ass? The only solution is to put PulseAudio into the kernel (I don't mean that literally, I mean something functionally equivalent). That said, ANY modern system can handle copying less than a kilobyte of data and converting from float to and from int within a millisecond (assuming what he said is true), so that argument is moot.
You're 100% wrong. Most users don't have issues with it, it's only the users that mess with it and fuck it up when they don't know what they're doing. Use a real distribution that actually packaged it correctly and DON'T try to tinker with it when you don't know what you're doing and it will work fine. Because YOU have problems with it you want it removed. You're part of the problem, please stop expressed your unwanted opinion, thanks.
Comment
-
Originally posted by jimbohale View PostYou're 100% wrong. Most users don't have issues with it, it's only the users that mess with it and fuck it up when they don't know what they're doing. Use a real distribution that actually packaged it correctly and DON'T try to tinker with it when you don't know what you're doing and it will work fine. Because YOU have problems with it you want it removed. You're part of the problem, please stop expressed your unwanted opinion, thanks.
As for your "100% wrong", do you have any facts to enforce your theory? I can assure you that every time a friend of mine has sound problems on linux (Suse, ubuntu, arch, whatever) the easiest and fastest solution is to remove pulseaudio. Sometimes there is need for a quick configuration in alsa to change the order of the audio cards, but besides that the sound problems (lag, stuttering, etc.) disappear as by miracle just by removing pulseaudio.
Of course, for you playing a youtube video in firefox might be "mess with it and fuck it up".
And I still have to understand what is a real distribution. I guess "linux from scratch" doesn't count, but does arch count? Suse? Debian? Kubuntu? Gentoo?
I particularly love how you write "DON'T try to tinker with it". So now linux should be something like windows, that users aren't allowed to tinker with because it might break the unstable crap that has been added in the last couple of years?
You're part of the problem, please stop wasting oxygen. Thank you.
PS: there is an edit button. If you want to change milliseconds to nanoseconds, you don't need to quote a full post. But perhaps editing a post, when done by someone who doesn't know what he is doing, would be to "mess with it and fuck it up".
Comment
-
Originally posted by jbernardo View PostWhat an interesting declaration. Such an attitude. Have you left kindergarten yet?
As for your "100% wrong", do you have any facts to enforce your theory? I can assure you that every time a friend of mine has sound problems on linux (Suse, ubuntu, arch, whatever) the easiest and fastest solution is to remove pulseaudio. Sometimes there is need for a quick configuration in alsa to change the order of the audio cards, but besides that the sound problems (lag, stuttering, etc.) disappear as by miracle just by removing pulseaudio.
Of course, for you playing a youtube video in firefox might be "mess with it and fuck it up".
And I still have to understand what is a real distribution. I guess "linux from scratch" doesn't count, but does arch count? Suse? Debian? Kubuntu? Gentoo?
I particularly love how you write "DON'T try to tinker with it". So now linux should be something like windows, that users aren't allowed to tinker with because it might break the unstable crap that has been added in the last couple of years?
You're part of the problem, please stop wasting oxygen. Thank you.
PS: there is an edit button. If you want to change milliseconds to nanoseconds, you don't need to quote a full post. But perhaps editing a post, when done by someone who doesn't know what he is doing, would be to "mess with it and fuck it up".
The only audio issues that occurs to me, and apparently you too based on "youtube in firefox", is related to closed source crap software. Such as Flash player, Skype, and such. Maybe they mess with things.
Comment
-
Originally posted by jimbohale View PostWhat? You agree that we should have a sound system that sucks ass? The only solution is to put PulseAudio into the kernel (I don't mean that literally, I mean something functionally equivalent). That said, ANY modern system can handle copying less than a kilobyte of data and converting from float to and from int within a millisecond (assuming what he said is true), so that argument is moot.
plus a copy since you can't do it in place
i was wrong, it is converted in PA
so a copy from program then convert in PA
then comes the cpu intensive part that is maths (il ignore cache trashing since everything trashes cache these days)
per sample it goes, i guess, around 30-60 multiplications and adds
there comes the cpu usage since a multiply for a float is much slower then for an int (add is also slower ofc)
and there is no sse loop for that in "secret rabbit code" so it's all scalar (have not found any in pa either)
also no, gcc does not optimize it into "vectors" (llvm may, idk)
http://www.agner.org/optimize/instruction_tables.pdf if you are interested in how much overhead is that
the floating point in DSP... trend i guess is for another reason, and in my opinion bollocks
people compare a single precision float filter to a 16bit int filter, and ofc the float wins (not by much, nothing you can hear and most of the time even measure)
if they compared it to 32bit integers it would be a different story, one in which int's would win since they have 2^32 precision while floats have... it's complicated
and the whole "modern hardware is so fast" argument is also a load of balls
i have a... idk around 1GHz celeron laptop and 10% cpu usage on it really matters
also there is a good enough sound system in the kernel, OSSv4
also i started a userspace sound server a while ago that may or may not be finished some sunny day
and guess what, no configuration required since it will always play at the highest sampling rate of all the apps using it, dynamically, and the rest will be upsampled
all that with 16.16 bits fixed point integer calculations (so 32bit precision)
should be waaaaaaaay faster then fkin floats with the same, if not better, precision
another thing that wont be a problem is a graphical equalizer since it is practically a required step when resampling (earphone mode too kinda, room speaker placement, etc)
that is what all semi decent windows sound card drivers have been doing for years now, and kids yell "PA is advanced!"
http://www.dspguru.com/dsp/faqs/multirate/resampling is how it is done
and https://ccrma.stanford.edu/~jos/resample/resample.html for more precise theory
good about PA is that they reported bugs when making/maintaining it
and for the couple people with special needs that don't have the knowledge or patience to do it themselves
PS phoronix edit's last for ~30sec for me, even thou it says 5min
Comment
Comment