In my WIP pdlua scripting library https://github.com/sebshader/lscoreI use 1 clock from pdlua for timing but then use my own heap/priority-queue-based scheduler as a wrapper for it.So if I need a new delay/clock I add it to my own scheduling queue, and then the priority queue pops the (next time - current time) for the underlying t_clock that pdlua uses every time a delay occurs.(and then the clocks are used for delays within coroutine yields so you can write music)In my case I also need a lot more manipulations of the scheduling queues though (my next project is to have 1 scheduling queue peraccelerando/ritardando group so a tree of descendants of an ancestor coroutine that sprouted them can accelerando/ritardando together) -Seb
-----Original Message-----
Date: Sun, 6 Jun 2021 18:52:17 -0700 From: Miller Puckette msp@ucsd.edu To: Iain Duncan iainduncanlists@gmail.com Cc: pd-dev pd-dev@lists.iem.at Subject: Re: [PD-dev] Creating delayed functions in Pd external? Message-ID: 20210607015217.4fb2ym5rl27d7olu@ucsd.edu Content-Type: text/plain; charset=us-ascii
Well, Pd's own scheduler for clocks keeps them in a linked list and chases down it to set and unset them. So if you did anything smart for the scheme callbacks Pd would see to it that it's still linear time in the number of active callbacks.
cheers M
On Sun, Jun 06, 2021 at 06:22:41PM -0700, Iain Duncan wrote:
Thanks Miller. That brings up one more question.. ;-)
In Scheme for Max, the way I do it is I keep a hash-table in C going with keys that are created on Scheme delay calls, so that one can cancel clocks by fetching them from the hashtable. In Max, there's a cross-platform hash-table implementation that I'm using. Is there something similar for Pd, or if not, is there an approach you would recommend for keeping a key-value store in C for the clocks by a symbolic key? I guess a good question might be if this is even necessary, given there can't be *that* many clocks scheduled for the future. And association list type thing might do just as well.
thanks for the help, it's going well, now that I can finally work on it! ought to have a first alpha folks can play with in the next couple of weeks. iain
On Sun, Jun 6, 2021 at 5:36 PM Miller Puckette msp@ucsd.edu wrote:
clock() is the only mechanism - for repeats, the easiest thing is often to just re-use a clock() and re-set it each time it goes off (as in the metro obejct). You can indeed create clock obejcts on the fly - that's what pipe does. But you'll want to keep track of them so you can cancel them if the owning object goes away.
cheers M
On Sun, Jun 06, 2021 at 05:16:09PM -0700, Iain Duncan wrote:
Ah fantastic, thanks. I was looking in pipe and not seeing it, but was probably just lost in other new details and not seeing the forest for the trees. :-)
Couple of follow ups:
- is there a separate facility for making a repeated callback (ie not
one-shot), or does one just do both with clock?
- is it safe to make clocks as we need them (ie during a method call, not
necessarily at object instantiation time), or is this the kind of thing where for real time use one needs to make a clock pool and a pool manager and all that?
thanks! iain
On Sun, Jun 6, 2021 at 5:12 PM Miller Puckette msp@ucsd.edu wrote:
Yep, clock_delay() . Simples example is in Pd's "delay" object,
x_time.c
cheers Miller
On Sun, Jun 06, 2021 at 04:21:46PM -0700, Iain Duncan wrote:
Hi folks, I'm hoping someone can point me in the right direction
here.
I'm
porting Scheme for Max to pure data and I'm stuck figuring out how
to get
delayed functions going. In Max, the SDK has a facility to make
register
a
callback to executed at some point in the future, a few different
ways.
Is
there a Pd equivalent, and if so, could anyone point me at resources
or
code for it? I basically just need to be able to have a callback
fire
off
at the right time with one argument, which can be void pointer to the
rest
of the stuff i want to get.
thanks! iain
Pd-dev mailing list Pd-dev@lists.iem.at
https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.puredata.info_lis...
--
--
Seb's way is more work than mine but can achieve much higher performance. (Incidentally, it's the approack Barry Vercoe took in his synthetic performer of 1984 - the 4X gave him, I think, 256 "clocks" and he just used one of them to wake him up every millisecond.
cheers Miller On Mon, Jun 07, 2021 at 09:09:37PM +0000, Sebastian Shader via Pd-dev wrote:
In my WIP pdlua scripting library??https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_sebshader_ls... use 1 clock from pdlua for timing but then use my own heap/priority-queue-based scheduler as a wrapper for it.So if I need a new delay/clock I add it to my own scheduling queue, and then the priority queue pops the (next time - current time) for the??underlying t_clock that pdlua uses every time a delay occurs.(and then the clocks are used for delays within coroutine yields so you can write music)In my case I also need a lot more manipulations of the scheduling queues though (my next project is to have 1 scheduling queue peraccelerando/ritardando group so a tree of descendants of an ancestor coroutine that sprouted them can accelerando/ritardando together) -Seb
-----Original Message-----
Date: Sun, 6 Jun 2021 18:52:17 -0700 From: Miller Puckette msp@ucsd.edu To: Iain Duncan iainduncanlists@gmail.com Cc: pd-dev pd-dev@lists.iem.at Subject: Re: [PD-dev] Creating delayed functions in Pd external? Message-ID: 20210607015217.4fb2ym5rl27d7olu@ucsd.edu Content-Type: text/plain; charset=us-ascii
Well, Pd's own scheduler for clocks keeps them in a linked list and chases down it to set and unset them.?? So if you did anything smart for the scheme callbacks Pd would see to it that it's still linear time in the number of active callbacks.
cheers M
On Sun, Jun 06, 2021 at 06:22:41PM -0700, Iain Duncan wrote:
Thanks Miller. That brings up one more question.. ;-)
In Scheme for Max, the way I do it is I keep a hash-table in C going with keys that are created on Scheme delay calls, so that one can cancel clocks by fetching them from the hashtable. In Max, there's a cross-platform hash-table implementation that I'm using. Is there something similar for Pd, or if not, is there an approach you would recommend for keeping a key-value store in C for the clocks by a symbolic key? I guess a good question might be if this is even necessary, given there can't be *that* many clocks scheduled for the future. And association list type thing might do just as well.
thanks for the help, it's going well, now that I can finally work on it! ought to have a first alpha folks can play with in the next couple of weeks. iain
On Sun, Jun 6, 2021 at 5:36 PM Miller Puckette msp@ucsd.edu wrote:
clock() is the only mechanism - for repeats, the easiest thing is often to just re-use a clock() and re-set it each time it goes off (as in the metro obejct).?? You can indeed create clock obejcts on the fly - that's what pipe does.?? But you'll want to keep track of them so you can cancel them if the owning object goes away.
cheers M
On Sun, Jun 06, 2021 at 05:16:09PM -0700, Iain Duncan wrote:
Ah fantastic, thanks. I was looking in pipe and not seeing it, but was probably just lost in other new details and not seeing the forest for the trees. :-)
Couple of follow ups:
- is there a separate facility for making a repeated callback (ie not
one-shot), or does one just do both with clock?
- is it safe to make clocks as we need them (ie during a method call, not
necessarily at object instantiation time), or is this the kind of thing where for real time use one needs to make a clock pool and a pool manager and all that?
thanks! iain
On Sun, Jun 6, 2021 at 5:12 PM Miller Puckette msp@ucsd.edu wrote:
Yep, clock_delay() .?? Simples example is in Pd's "delay" object,
x_time.c
cheers Miller
On Sun, Jun 06, 2021 at 04:21:46PM -0700, Iain Duncan wrote:
Hi folks, I'm hoping someone can point me in the right direction
here.
I'm
porting Scheme for Max to pure data and I'm stuck figuring out how
to get
delayed functions going. In Max, the SDK has a facility to make
register
a
callback to executed at some point in the future, a few different
ways.
Is
there a Pd equivalent, and if so, could anyone point me at resources
or
code for it??? I basically just need to be able to have a callback
fire
off
at the right time with one argument, which can be void pointer to the
rest
of the stuff i want to get.
thanks! iain
Pd-dev mailing list Pd-dev@lists.iem.at
https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.puredata.info_lis...
--
--
--
Subject: Digest Footer
Pd-dev mailing list Pd-dev@lists.iem.at https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.puredata.info_lis...
End of Pd-dev Digest, Vol 194, Issue 4
Pd-dev mailing list Pd-dev@lists.iem.at https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.puredata.info_lis...
Thanks Seb and Miller. Good suggestion, I've played with both approaches in the Max version actually. I'll start simple, but thanks for sharing your code Seb, I may move to that later (or add it as another option).
thank again! iain
On Mon, Jun 7, 2021 at 3:12 PM Miller Puckette via Pd-dev < pd-dev@lists.iem.at> wrote:
Seb's way is more work than mine but can achieve much higher performance. (Incidentally, it's the approack Barry Vercoe took in his synthetic performer of 1984 - the 4X gave him, I think, 256 "clocks" and he just used one of them to wake him up every millisecond.
cheers Miller On Mon, Jun 07, 2021 at 09:09:37PM +0000, Sebastian Shader via Pd-dev wrote:
In my WIP pdlua scripting library??
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_sebshader_ls... use 1 clock from pdlua for timing but then use my own heap/priority-queue-based scheduler as a wrapper for it.So if I need a new delay/clock I add it to my own scheduling queue, and then the priority queue pops the (next time - current time) for the??underlying t_clock that pdlua uses every time a delay occurs.(and then the clocks are used for delays within coroutine yields so you can write music)In my case I also need a lot more manipulations of the scheduling queues though (my next project is to have 1 scheduling queue peraccelerando/ritardando group so a tree of descendants of an ancestor coroutine that sprouted them can accelerando/ritardando together)
-Seb
-----Original Message-----
Date: Sun, 6 Jun 2021 18:52:17 -0700 From: Miller Puckette msp@ucsd.edu To: Iain Duncan iainduncanlists@gmail.com Cc: pd-dev pd-dev@lists.iem.at Subject: Re: [PD-dev] Creating delayed functions in Pd external? Message-ID: 20210607015217.4fb2ym5rl27d7olu@ucsd.edu Content-Type: text/plain; charset=us-ascii
Well, Pd's own scheduler for clocks keeps them in a linked list and
chases
down it to set and unset them.?? So if you did anything smart for the scheme callbacks Pd would see to it that it's still linear time in the number of active callbacks.
cheers M
On Sun, Jun 06, 2021 at 06:22:41PM -0700, Iain Duncan wrote:
Thanks Miller. That brings up one more question.. ;-)
In Scheme for Max, the way I do it is I keep a hash-table in C going with keys that are created on Scheme delay calls, so that one can
cancel
clocks by fetching them from the hashtable. In Max, there's a cross-platform hash-table implementation that I'm using. Is there
something
similar for Pd, or if not, is there an approach you would recommend for keeping a key-value store in C for the clocks by a symbolic key? I
guess a
good question might be if this is even necessary, given there can't be *that* many clocks scheduled for the future. And association list type thing might do just as well.
thanks for the help, it's going well, now that I can finally work on
it!
ought to have a first alpha folks can play with in the next couple of
weeks.
iain
On Sun, Jun 6, 2021 at 5:36 PM Miller Puckette msp@ucsd.edu wrote:
clock() is the only mechanism - for repeats, the easiest thing is
often
to just re-use a clock() and re-set it each time it goes off (as in the metro obejct).?? You can indeed create clock obejcts on the fly - that's what pipe does.?? But you'll want to keep track of them so
you can
cancel them if the owning object goes away.
cheers M
On Sun, Jun 06, 2021 at 05:16:09PM -0700, Iain Duncan wrote:
Ah fantastic, thanks. I was looking in pipe and not seeing it, but
was
probably just lost in other new details and not seeing the forest
for the
trees. :-)
Couple of follow ups:
- is there a separate facility for making a repeated callback (ie
not
one-shot), or does one just do both with clock?
- is it safe to make clocks as we need them (ie during a method
call, not
necessarily at object instantiation time), or is this the kind of
thing
where for real time use one needs to make a clock pool and a pool
manager
and all that?
thanks! iain
On Sun, Jun 6, 2021 at 5:12 PM Miller Puckette msp@ucsd.edu
wrote:
Yep, clock_delay() .?? Simples example is in Pd's "delay" object,
x_time.c
cheers Miller
On Sun, Jun 06, 2021 at 04:21:46PM -0700, Iain Duncan wrote: > Hi folks, I'm hoping someone can point me in the right
direction
here.
I'm > porting Scheme for Max to pure data and I'm stuck figuring out
how
to get
> delayed functions going. In Max, the SDK has a facility to make
register
a > callback to executed at some point in the future, a few
different
ways.
Is > there a Pd equivalent, and if so, could anyone point me at
resources
or
> code for it??? I basically just need to be able to have a
callback
fire
off > at the right time with one argument, which can be void pointer
to the
rest > of the stuff i want to get. > > thanks! > iain
> _______________________________________________ > Pd-dev mailing list > Pd-dev@lists.iem.at >
https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.puredata.info_lis...
--
--
--
Subject: Digest Footer
Pd-dev mailing list Pd-dev@lists.iem.at
https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.puredata.info_lis...
End of Pd-dev Digest, Vol 194, Issue 4
Pd-dev mailing list Pd-dev@lists.iem.at
https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.puredata.info_lis...
--
Pd-dev mailing list Pd-dev@lists.iem.at https://lists.puredata.info/listinfo/pd-dev
-
Iain Duncan -
Miller Puckette -
Sebastian Shader