15 Jan
2010
15 Jan
'10
5:03 a.m.
hi...
i am working on a c++0x DSP library.
variadic templates prove to be a nice way to handle
the massive function inlining required to build efficient samplebased
processing graphs.
the idioms i am currently using for the containers require gcc-4.5
though, so this is still a bit of a show-stopper :)
its still in a pretty early state. but you can already see where its
heading.
if somebody is interested:
http://hochstrom.endofinternet.org/trac/ttsoot/wiki
--
torben Hohn
15 Jan
15 Jan
5:19 a.m.
Torben,
this looks very interesting. Quite a lot of C++ complexity, though.
I'll keep following it to see where you are heading.
Regards
Victor
On 15 Jan 2010, at 09:03, torbenh wrote:
hi...
i am working on a c++0x DSP library.
variadic templates prove to be a nice way to handle
the massive function inlining required to build efficient samplebased
processing graphs.
the idioms i am currently using for the containers require gcc-4.5
though, so this is still a bit of a show-stopper :)
its still in a pretty early state. but you can already see where its
heading.
if somebody is interested:
http://hochstrom.endofinternet.org/trac/ttsoot/wiki
--
torben Hohn
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-dev
5:49 a.m.
Le 15 janv. 2010 à 10:03, torbenh a écrit :
hi...
i am working on a c++0x DSP library.
variadic templates prove to be a nice way to handle
the massive function inlining required to build efficient samplebased
processing graphs.
the idioms i am currently using for the containers require gcc-4.5
though, so this is still a bit of a show-stopper :)
its still in a pretty early state. but you can already see where its
heading.
if somebody is interested:
http://hochstrom.endofinternet.org/trac/ttsoot/wiki
--
torben Hohn
How would this approach compare to the FAUST approach?
Stephane
17 Jan
17 Jan
12:59 a.m.
On Fri, Jan 15, 2010 at 10:49:21AM +0100, Stéphane Letz wrote:
Le 15 janv. 2010 à 10:03, torbenh a écrit :
its more or less equivalent to faust scalar code.
the containers provide similar composition ops to the faust ones.
but since my primitives are classes i can provide richer primitives
than faust.
i dont really see a way to automate what vector mode is doing.
but it should be possible to define a set of blockbased containers.
so that:
Chain< Blocked< A,B,C >, Blocked< D,E,F > >::process_block()
would expand to 2 loops.
hi...
i am working on a c++0x DSP library.
variadic templates prove to be a nice way to handle
the massive function inlining required to build efficient samplebased
processing graphs.
the idioms i am currently using for the containers require gcc-4.5
though, so this is still a bit of a show-stopper :)
its still in a pretty early state. but you can already see where its
heading.
if somebody is interested:
http://hochstrom.endofinternet.org/trac/ttsoot/wiki
--
torben Hohn
How would this approach compare to the FAUST approach?
Stephane
--
torben Hohn
17 Feb
17 Feb
6:33 a.m.
New subject: xazz - yet *another* DSP library - or language? (was: Re: ttsoot - yet another DSP library :))
On 17 Jan 2010, at 04:59, torbenh wrote:
On Fri, Jan 15, 2010 at 10:49:21AM +0100, Stéphane
Letz wrote:
torben, Stéphane, LAD:
This exchange got me wondering... how "Faust-like" could a language, embedded
directly in C++ using template metaprograming techniques, be made to be?
Could it be made to accept e.g.:
(expr0:(expr1,expr2,expr3):expr4)
In place of e.g.:
Serial(Serial(expr0,Parallel(Parallel(expr1,expr2),expr3),expr4))?
Wondering turned to coding and I ended up with Xazz:
/*===========
begin sine.cpp
===========*/
#include <unistd.h>
#include "xazz/base.hpp"
#include "xazz/math.hpp"
#include "xazz/jack.hpp"
int main( void )
{
XAZZ_IMPORT_BASE
XAZZ_IMPORT_MATH
#define _ XAZZ_CONST
const auto samplerate = _(44100.0f);
const auto osc440 = _(440.0f)
Le 15 janv. 2010 à 10:03, torbenh a écrit :
its more or less equivalent to faust scalar code.
the containers provide similar composition ops to the faust ones.
hi...
i am working on a c++0x DSP library.
variadic templates prove to be a nice way to handle
the massive function inlining required to build efficient samplebased
processing graphs.
the idioms i am currently using for the containers require gcc-4.5
though, so this is still a bit of a show-stopper :)
its still in a pretty early state. but you can already see where its
heading.
if somebody is interested:
http://hochstrom.endofinternet.org/trac/ttsoot/wiki
--
torben Hohn
How would this approach compare to the FAUST approach? (((twopi,samplerate) > div) > mul)
((add <= ( pi>lt, wire, (wire, twopi) > sub ) > select) & wire )
sin_;
#undef _
xazz::Jack<decltype(osc440)> MyJack(osc440, "osc440");
while( true ) { sleep(10); }
return 0;
}
/*===========
end sine.cpp
===========*/
That's a Jack app that outputs a 440Hz sine wave.
Notes:
1. C++ obviously doesn't have the same operators as Faust so Xazz uses:
">" serial composition (Faust's ":")
"," parallel composition (Faust's ",")
"<=" split composition (Faust's "<:")
">=" join composition (Faust's ":>" )
"&" recursive composition (Faust's "~")
"wire" identity (Faust's "_")
"cut" cut (Faust's "!")
2. It still needs c++0x but not so close to the bleeding edge. g++ 4.4 required.
3. It looks like we're composing the Xazz program at runtime (with consequent loss of
efficiency) but the "meat" of osc440 is what type it is, not what data it
contains. The compiler works that out at compile time and -- thanks to c++0x auto -- we
never have to write it down ourselves.
4. This is a V0.0 "proof of concept" release. Its LGPL.
5. Ummm.... thats it. Headers and a couple more examples here:
http://www.sjenkins.pwp.blueyonder.co.uk/xazz/xazz_0_0.tar.gz
Simon Jenkins
Bristol, UK
5452
days inactive
5485
days old
linux-audio-dev@lists.linuxaudio.org
4 comments
4 participants
participants (4)
-
Simon Jenkins -
Stéphane Letz -
torbenh -
Victor Lazzarini