[PD] readsf~ and writesf~, locking, blocking and real-time priority

Matt Barber brbrofsvl at gmail.com
Wed Oct 7 10:11:46 CEST 2015 

I think Ableton allows you to choose whether you want to stream from disk
or load to RAM.

Keeping data in RAM depends on what you're trying to play. If you need to
playback something like an hour-long 8-channel file @ 96k, pretty much your
only practical option is disk streaming (but you might be looking at a
dedicated playback software rather than [readsf~] in that case).

On Wed, Oct 7, 2015 at 2:59 AM, Robert Esler <robert at urbanstew.org> wrote:

> Matt,
> I think on a Desktop OS keeping the data in RAM wouldn’t be much of an
> issue.  I’m sure software like Ableton’s Live doesn’t stream audio from
> disk.   But on something where RAM is scarce like a mobile phone or single
> board CPU the OS may not provide enough memory and fail.  Especially with
> libpd becoming more ubiquitous, a real-time safe streaming option may be
> valuable.
>
> But you’re right, the designer has a choice as to how data is accessed and
> should know the potential consequences.  Education is still key.
>
> -Rob
>
> On Oct 6, 2015, at 11:15 PM, Matt Barber <brbrofsvl at gmail.com> wrote:
>
> ​Out of curiosity, can you give a test case where my idiom above (which
> keeps the file loaded as long as you need it) would be too onerous or just
> impossible to use? I think I can think of a few, like some kind of patch
> that follows an improvisation and chooses files to stream immediately in
> response to something the performer did. In those cases, though, I'd be
> very reluctant to trust the disk drive to serve multiple files immediately.
> SSDs might perform better and a ramdisk better still, but in the general
> case I think I would do everything I could do design my composition/patch
> to be consistent with preloading any files I need immediately.
>
> On Tue, Oct 6, 2015 at 4:41 PM, Jonathan Wilkes via Pd-list <
> pd-list at lists.iem.at> wrote:
>
>> I'm not asking when you would implement this.  We were talking about
>> determinism, after all. :)
>>
>> But I guess I'm confusing a program's deterministic behavior-- e.g.,
>> output
>> at predictable times-- with a message that has a value derived from
>> machine- or OS-specific behavior.
>>
>> -Jonathan
>>
>>
>>
>> On Tuesday, October 6, 2015 3:32 PM, Miller Puckette <msp at ucsd.edu>
>> wrote:
>>
>>
>> I don't think the design constrains it out - I just haven't done it, since
>> it's rather more than a minor project.
>>
>> cheers
>> M
>>
>> On Tue, Oct 06, 2015 at 07:27:51PM +0000, Jonathan Wilkes wrote:
>> > Hm... still, I can't get that value from within a patch because of
>> your design constraints, right?
>> >
>> >
>> >      On Tuesday, October 6, 2015 2:20 PM, Miller Puckette <msp at ucsd.edu>
>> wrote:
>> >
>> >
>> >  So in that case you'd really want a different metric, perhaps 'what
>> was the
>> > minimum fill count while the file was playing'.
>> >
>> > cheers
>> > M
>> >
>> > On Tue, Oct 06, 2015 at 05:40:27PM +0000, Jonathan Wilkes via Pd-list
>> wrote:
>> > > But as Matt's student I want to be able to measure the time it takes
>> to open the file, to make a more informed decision when I design my patch.
>> I wantthe pure data when I'm only opening a few soundfiles, and then Iwant
>> the pure data when I try to open lots of files.  Pd already gives
>> me[realtime] which I could use to create non-deterministic patches.
>> Obviouslyits author realized that the ability to measure time outweighs the
>> risk of doingthat.  One would assume that same tradeoff to be equally
>> important, if notmoreso, for the few instances of object behavior which
>> [realtime] cannotmeasure.
>> > > The other students are now rolling their eyes.  I think they're on to
>> me.
>> > > -Jonathan
>> > >
>> > >
>> > >
>> > >      On Tuesday, October 6, 2015 12:11 PM, Miller Puckette <
>> msp at ucsd.edu> wrote:
>> > >
>> > >
>> > >  A worthy question.
>> > >
>> > > If you want the soundfile to start exactly when you specify it (I
>> think this
>> > > should normally be the case :) then it's beside the point exactly
>> when the
>> > > computer could have coughed it up - it only matters that it be there
>> by the
>> > > desired time.
>> > >
>> > > If you want the soundfile to play "whenever the computer can manage
>> it" - and
>> > > the sooner the better - well, then a "ready" message would be
>> useful.  I
>> > > imagine one could shave off 1/5 second or so, but it would be
>> inconsistent.
>> > > Perhaps this is useful in some cases but I don't think it would be
>> often -
>> > > and the downside is that it wouldn't be deterministic (a fundamental
>> design
>> > > principle of Pd).
>> > >
>> > > cheers
>> > > Miller
>> > >
>> > >
>> > > On Tue, Oct 06, 2015 at 03:58:38PM +0000, Jonathan Wilkes wrote:
>> > > > If I were one of Matt's students, I'd ask why this "Pure Data
>> readsf~ business" won't just tell me when it has actually opened the file.
>> Why does thecomputer get to know when it's ready, but we students have to
>> guess bylistening for glitches?
>> > > > -Jonathan
>> > > >
>> > > >
>> > > >
>> > > >      On Tuesday, October 6, 2015 1:02 AM, Matt Barber <
>> brbrofsvl at gmail.com> wrote:
>> > > >
>> > > >
>> > > >
>> > > > 4) Has anyone ever “broken” these objects or experienced glitching?
>> > > >
>> > > > ​Once in 2005 we were having awful trouble streaming through Pd but
>> we were never sure whether it was [readsf~] per se, a very slow disk, or
>> xruns in ALSA/JACK, and we had only one performance laptop available. My
>> best guess is that it was an ALSA/JACK problem, since other software had a
>> few issues with glitching just on realtime audio processing.
>> > > > pthread_mutex_lock() ... This might be a good time for a PSA for
>> interested newcomers to Pd, though, if any happen to be following this
>> thread (ahem). Having taught Pd for some 10 years now, one bad habit I've
>> seen nearly every student fall into is failing to preload the file before
>> playing, trying to do the initial read and the playing at the same logical
>> time. Usually there isn't a problem, but once in a while a taxed system
>> that is already streaming several files can glitch hard on a new stream.
>> I've attached a generic [readsf~] idiom that has been useful for first-year
>> students when they want to jump in and get Pd to play some sound files with
>> a GUI after they've fooled around with the control examples and
>> oscillators. This is before we get into event triggering, so the clunky
>> multiple play/stop buttons is edited out later on; the main thing is how to
>> keep the file open at all times. This turns out to be even more important
>> for rehearsal than for performance, when you need to be able to jump around
>> at will. ... pthread_mutex_unlock()
>> > > >
>> > > >
>> > > > On Tue, Oct 6, 2015 at 12:15 AM, Miller Puckette <msp at ucsd.edu>
>> wrote:
>> > > >
>> > > > I think you can get away with sharing a lock between two
>> high-prioroty
>> > > > processes as long as neither one holds the lock for more than a
>> small
>> > > > amount of time and if the OS can be counted on to give control to a
>> real-time
>> > > > process quickly once it becomes runnable (i.e., if it's blocked on
>> a lock,
>> > > > once that lock is released).
>> > > >
>> > > > The situation I don't know about is this:  if Pd's main thread
>> failed to get
>> > > > the lock, so that control (presumably) passed back to the other
>> thread that
>> > > > had the lock, how much time can pass before the other thread blocks
>> on
>> > > > something so that control (again presumably) gets passed back to
>> the main
>> > > > thread?
>> > > >
>> > > > But anyway, since neither thread holds onto the lock for more than
>> a few
>> > > > lines of C code (with no system calls) it's probably blue-moon rare
>> that the
>> > > > scheduler interrupts one thread right in the middle of a critical
>> section and
>> > > > passes control to the other one that then blocks.  So this is
>> essentially
>> > > > untested.
>> > > >
>> > > > Threads can never be used confidently in a real-time situation.
>> But I don't
>> > > > see any reasonable way without them to implement readsf~/writesf~,
>> so there
>> > > > we are...
>> > > >
>> > > > cheers
>> > > > Miller
>> > > >
>> > > > P.S. one can issue non-blocking reads/writes, but there's also
>> "open" which
>> > > > is much more likely to hiccup than "read", and I don't know of any
>> async open
>> > > > call in any OS.
>> > > >
>> > > >
>> > > > On Tue, Oct 06, 2015 at 03:10:31AM +0000, Jonathan Wilkes via
>> Pd-list wrote:
>> > > > > 1) One thing I noticed is that the article you cited seems to
>> focus
>> > > > > on tasks not critical to the computation/delivery of audio
>> samples.For example, if your program were blocking or locking in order to
>> do a GUIupdate.  But here, the data must arrive in time to compute the next
>> block.  If ittakes too long to read the next portion of the sound file,
>> then you're going to geta glitch.
>> > > > > But I'm not sure I really grasp how locking works, nor really the
>> whole file i/oprocess in general.
>> > > > >
>> > > > > Here's a naive question: why can't you just tell the OS to treat
>> the file asif it were a non-blocking socket, add the fd to Pd's event loop
>> with
>> > > > > sys_addpollfn, and then receive the incoming data to the relevant
>> function?(Warning: some or all of the above may technically be gibberish...)
>> > > > >
>> > > > > -Jonathan
>> > > > >
>> > > > >
>> > > > >
>> > > > >
>> > > > >
>> > > > >
>> > > > >
>> > > > >
>> > > > >      On Monday, October 5, 2015 10:01 PM, Robert Esler <
>> robert at urbanstew.org> wrote:
>> > > > >
>> > > > >
>> > > > >  I’m trying to understand why readsf~ and writesf~ work so well.
>> > > > >
>> > > > > I’m particularly referencing Ross Bencina’s article: http://www.rossbencina.com/code/real-time-audio-programming-101-time-waits-for-nothing
>> and his subsequent paper,
>> http://www.rossbencina.com/static/writings/File_IO_ACMC2014_Bencina.pdf
>> > > > >
>> > > > > If you are not into asynchronous message passing and lock-free
>> queueing then I’ll summarize the articles briefly:
>> > > > >
>> > > > > When engaging in file I/O (e.g reading from or writing to an
>> audio file) do not use locks or blocking. He goes on to say that this can
>> lead to priority inversion, unbound execution time and “scheduler paranoia”.
>> > > > >
>> > > > > This is all absolutely true in my experience in the audio jungle.
>> > > > >
>> > > > > Pd’s async file I/O objects (readsf~ and writesf~) use both locks
>> and blocking via a mutex and the pthread_cond_signal and pthread_cond_init
>> functions.  Look at the source code file d_soundfile.c for more details.
>> The gist of it is that these objects have two threads.  One parent thread
>> that sends the data to the dsp scheduler, and a child thread that grabs the
>> data from the file, and subsequently the child signals the parent when it
>> has more data.
>> > > > >
>> > > > > Based on Bencina’s paper, readsf~ and writesf~ could (should?)
>> glitch and may not be real-time safe.
>> > > > >
>> > > > > My questions are:
>> > > > >
>> > > > > 1) Have I completely misunderstood d_soundfile.c and it is
>> actually entirely safe. If so, why is it safe?
>> > > > >
>> > > > > 2) Why doesn’t Pd glitch more often when using these objects?
>> > > > >
>> > > > > 3) Does Pd need lock-free message queueing for such inter-thread
>> communication?
>> > > > >
>> > > > > 4) Has anyone ever “broken” these objects or experienced
>> glitching?
>> > > > >
>> > > > > Thanks for the extra brain power.
>> > > > > -R
>> > > > >
>> > > > >
>> > > > >
>> > > > > _______________________________________________
>> > > > > Pd-list at lists.iem.at mailing list
>> > > > > UNSUBSCRIBE and account-management ->
>> http://lists.puredata.info/listinfo/pd-list
>> > > > >
>> > > > >
>> > > > >
>> > > >
>> > > > > _______________________________________________
>> > > > > Pd-list at lists.iem.at mailing list
>> > > > > UNSUBSCRIBE and account-management ->
>> http://lists.puredata.info/listinfo/pd-list
>> > > >
>> > > >
>> > > > _______________________________________________
>> > > > Pd-list at lists.iem.at mailing list
>> > > > UNSUBSCRIBE and account-management ->
>> http://lists.puredata.info/listinfo/pd-list
>> > > >
>> > > >
>> > > >
>> > > >
>> > > >
>> > >
>> > >
>> >
>> > > _______________________________________________
>> > > Pd-list at lists.iem.at mailing list
>> > > UNSUBSCRIBE and account-management ->
>> http://lists.puredata.info/listinfo/pd-list
>> >
>> >
>> >
>>
>>
>>
>> _______________________________________________
>> Pd-list at lists.iem.at mailing list
>> UNSUBSCRIBE and account-management ->
>> http://lists.puredata.info/listinfo/pd-list
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20151007/3e6f7b5c/attachment-0001.html>

More information about the Pd-list mailing list