[PD] [pdconv16_r] Expanding abstractions & Compiling Vanilla Patches As Objects (Gen~?)

Pierre Guillot guillotpierre6 at gmail.com
Tue Nov 1 18:54:25 CET 2016


@Alewandre Sorry, I just realize that I misunderstood the first mail. When
you spoke about [inlet~], I thought about the inlets of the objects and not
the inlets of the abstractions. My approach is for for externals so perhaps
that should be another subject for the round table?

It seems that Enzien audio reinterprets the pd patches and doesn't use the
pd code. If this framework acts like gen, it probably generates an
pre-optimized C code from a patch (and then the compiler can compile and
perform the optimizations). Perhaps I'm wrong but I think that there is a
reason why Enzien audio doesn't use the pd code and why gen uses its own
objects and not the native dsp objects: they had to rewrite all the code of
the objects. I mean, if we want to convert a patch to a perform method (we
simplify the approach by limiting the patch to the dsp objects only),  the
gen-like object have to know the textual form (or an equivalent) of the
perform methods of all the objects. But for the moment, we can only
retrieve the addresses of the perform methods and even if we implement a
mechanism to retrieve these methods as text, we still have to rethink their
syntaxes (because they are surely not usable and not optimized for the new
context). Perhaps my explanations are not really clear and perhaps I should
keep that for the irl discussion... I just think that such application
amounts to recreate a brand new software. It would be great but it would
require a lot of work. Considering the existing projects like libpd, FAUST
or even Enzien audio, does the prize justify the cost? Sorry if it seems
that I want to nip the project in the bud :)

2016-11-01 17:41 GMT+01:00 Andy Farnell <padawan12 at obiwannabe.co.uk>:

> Its an idea I've long supported, for perhaps 10 years.
>
> But there are some subtle dangers to openness.
>
> Given the very small size of Pd patch code in relation
> to even the simplest compiled external we might agree a
> principle, a standard, an unwritten law if you like....
>
> that any pd abstraction converted by a [gen~] like mechanism
> into a compiled external, carries with it the dataflow source
> as a data chunk. That means the dataflow source is always distributed
> with it. Further the whole process is reversible with a
> "convert-external-to-abstraction" process.
>
> What we would have then is a kind of great "folding editor/compiler"
> for DSP development where two views of code are possible.
> This would be powerful in teaching C as well as data-flow with
> the same tool.
>
> Andy
>
> On Tue, Nov 01, 2016 at 02:12:08PM -0200, Alexandre Torres Porres wrote:
> > > More generally, it would be great if abstractions could do
> > > anything a compiled object could do.
> >
> > Exactly ;)
> >
> > And again, let me add, there are things like the heavy compiler,
> > https://enzienaudio.com where you can compile pd patches into optimized
> code
> >
> > how does that work? Wouldn't that be something like the "gen~" idea I
> > brought up? How hard would it be to have a compiler for a patch to be
> > turned into a coded object?
> >
> > if abstractions could do anything a compiled object could do including
> > being optimized and efficient, that would be amazing...
> >
> > cheers
> >
> > 2016-11-01 13:56 GMT-02:00 Alex Norman <x37v.alex at gmail.com>:
> >
> > > Miller did seem open to a control outlet on the inlet~ object. This was
> > > when we were discussing the clone object and how you have to pass
> messages
> > > to the first control inlet, if you have one, instead of just the first
> > > inlet always, to control the cloning operations. More generally, it
> would
> > > be great if abstractions could do anything a compiled object could do.
> > > Alex
> > >
> > > On November 1, 2016 8:47:11 AM PDT, Alexandre Torres Porres <
> > > porres at gmail.com> wrote:
> > >
> > >> 2016-11-01 8:42 GMT-02:00 Pierre Guillot <guillotpierre6 at gmail.com>:
> > >>
> > >>> Hi Alexandre,
> > >>>
> > >>> > I wonder if a thing like libpd could work as turning a vanilla
> patch
> > >>> into a
> > >>> > compiled object to be used inside pd... that'd be something like
> gen~
> > >>> in
> > >>> > max/msp.
> > >>>
> > >>> Can you be more specific ? For the moment, I think it would be
> > >>> equivalent to use an abstraction or the object [pd~] (libpd loads
> > >>> dynamically a patch so I guess that the execution of the patch
> cannot be
> > >>> optimized and except if the patch has been be somehow included
> inside the
> > >>> binary, you'll have to share the patch with the object). For me, the
> main
> > >>> advantage of gen~ is that it generates code that can be used inside
> an
> > >>> application but libpd already offers this feature. So what would be
> the
> > >>> advantage?
> > >>>
> > >>
> > >>
> > >> Well, I thought the code could be optimized somehow, which I believe
> is
> > >> something gen~ does, and that could be an advantage... but I really
> know
> > >> nothing and now it seems that is not possible.
> > >>
> > >>
> > >> > A - being able to retrieve control data from [inlet~]
> > >>>
> > >>> I did it in the Cicm Wrapper but it was pretty tricky. If you use the
> > >>> object [hoa.process~], you can send messages via a signal inlet for
> > >>> example. I'm not very proud of this because I had to hack a bit the
> inlet
> > >>> class. Now, I don't know if I must remove this feature or keep it...
> > >>> Perhaps somebody could tell/remind us if there is a reason why signal
> > >>> inlets can't receive messages.
> > >>>
> > >>
> > >> cool, there's also a [route~] object from zexy which could be
> embedded in
> > >> inlet~
> > >>
> > >>
> > >> > B - being able to know if a signal is connected to [inlet~]
> > >>>
> > >>> I also did it in the Cicm Wrapper, perhaps this feature could be
> > >>> included in the "m_pd.h" interface because for the moment you need to
> > >>> include "g_canvas.h" and "m_imp.h". Anyway, if you want a simple
> code that
> > >>> shows how to do it, I have an example
> > >>> <https://github.com/pierreguillot/pd-dummy/blob/
> master/src/connected_tilde.c>
> > >>> in my dummy library.
> > >>>
> > >>
> > >> awesome, it's be great to have something like this in vanilla in
> order to
> > >> improve the design of abstractions ;)
> > >>
> > >> cheers
> > >>
> > >> ------------------------------
> > >>
> > >> Pd-list at lists.iem.at mailing list
> > >> UNSUBSCRIBE and account-management -> https://lists.puredata.info/
> listinfo/pd-list
> > >>
> > >>
>
> > _______________________________________________
> > Pd-list at lists.iem.at mailing list
> > UNSUBSCRIBE and account-management -> https://lists.puredata.info/
> listinfo/pd-list
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20161101/d71c3a29/attachment.html>


More information about the Pd-list mailing list