In my project the audio content (text to be spoken by synthetic voices) is
expected to change dynamically which means the live audio output capability
of Csound is planned to be used. Multichannel output (currently surround
5.1) will be used. The different voices will be placed at different
positions in a virtual room or are even moving in this room. The pulse
responses of the virtual room are/will be provided for the different
positions of the static voices or have to be calculated for the moving
voices during the performance. The voices have to be folded with the proper
pulse response during the performance. This does not impose a problem as
long as only few voices are active simultaneously. However, in case of a
large number voices being active at the same time either as a chorus or as a
cacophony I expect limitation by the CPU speed using a single core. Are
there any multicore capabilities implemented in Csound (e.g. distributing
the processing of different instruments to different cores, in this case the
convolution)?  



--
View this message in context: http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800.html
Sent from the Csound - General mailing list archive at Nabble.com.

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

You can read a recent paper John fittch wrote about how csound uses multithreading. But that is very "low level". Unlike supercollider's supernova csound has no way for user to control multithreading, besodes just understanding its concepts, and the -j4 compiler flag. There are many theories and many different opinions even within csound when and if multithreading is a good idea and for convolution i think parallelism would come in good use. Besides convolition, csound would handle formant synthesis very easily on modern machine. I just live coded 2 days ago with 4-5 voices fof plus many samples and poscil based synths and the computer didnt even sweat. So my point is for the optimization is that its better to do it the last because the things that slow your processor down are many times something much more simple that using multiple cores.

Just a question still. How many voices do you want for cacophony? Because yes if you are talking about numbers in thousands then yes its going to make things slow depending on buffer size and kr/sr value. Chorus/delay/pitch shift fx's could deliver convincing resault for lesser cpu consumption. But dont know for sure.

Thanks for the answers. Well I could imagine something like 20 to 50 voices. The problem I see may be the convolution of so many voices with different pulse response function in order to get automatically for voices placed closer to the sweet spot a larger ratio (and different time structure) in the volumes for the direct signal to the reflections that for voices far away.

If you have several instances of the same convolution instrument, and no global variables connecting them, using -j N 
should work very well to distribute these
among N cores. If you use higher ksmps
values, it could speed up processing 
considerably.


Victor Lazzarini
Dean of Arts, Celtic Studies, and Philosophy
Maynooth University
Ireland

> On 3 Jan 2016, at 10:12, MarieCurie  wrote:
> 
> In my project the audio content (text to be spoken by synthetic voices) is
> expected to change dynamically which means the live audio output capability
> of Csound is planned to be used. Multichannel output (currently surround
> 5.1) will be used. The different voices will be placed at different
> positions in a virtual room or are even moving in this room. The pulse
> responses of the virtual room are/will be provided for the different
> positions of the static voices or have to be calculated for the moving
> voices during the performance. The voices have to be folded with the proper
> pulse response during the performance. This does not impose a problem as
> long as only few voices are active simultaneously. However, in case of a
> large number voices being active at the same time either as a chorus or as a
> cacophony I expect limitation by the CPU speed using a single core. Are
> there any multicore capabilities implemented in Csound (e.g. distributing
> the processing of different instruments to different cores, in this case the
> convolution)?  
> 
> 
> 
> --
> View this message in context: http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800.html
> Sent from the Csound - General mailing list archive at Nabble.com.
> 
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>        https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

Thanks for the answers. Well I could imagine something like 20 to 50 voices. The problem I see may be the convolution of so many voices with different pulse response function in order to get automatically for voices placed closer to the sweet spot a larger ratio (and different time structure) in the volumes for the direct signal to the reflections that for voices far away.

Karin

 

On 3 Jan 2016, at 12:37, hlolli [via Csound] <[hidden email]> wrote:

Just a question still. How many voices do you want for cacophony? Because yes if you are talking about numbers in thousands then yes its going to make things slow depending on buffer size and kr/sr value. Chorus/delay/pitch shift fx's could deliver convincing resault for lesser cpu consumption. But dont know for sure.

On Jan 3, 2016 12:30 PM, "Hlöðver Sigurðsson" <<a href="x-msg://39/user/SendEmail.jtp?type=node&amp;node=5745808&amp;i=0" target="_top" rel="nofollow" link="external" class="">[hidden email]> wrote:

You can read a recent paper John fittch wrote about how csound uses multithreading. But that is very "low level". Unlike supercollider's supernova csound has no way for user to control multithreading, besodes just understanding its concepts, and the -j4 compiler flag. There are many theories and many different opinions even within csound when and if multithreading is a good idea and for convolution i think parallelism would come in good use. Besides convolition, csound would handle formant synthesis very easily on modern machine. I just live coded 2 days ago with 4-5 voices fof plus many samples and poscil based synths and the computer didnt even sweat. So my point is for the optimization is that its better to do it the last because the things that slow your processor down are many times something much more simple that using multiple cores.

On Jan 3, 2016 11:13 AM, "MarieCurie" <<a href="x-msg://39/user/SendEmail.jtp?type=node&amp;node=5745808&amp;i=1" target="_top" rel="nofollow" link="external" class="">[hidden email]> wrote:

In my project the audio content (text to be spoken by synthetic voices) is
expected to change dynamically which means the live audio output capability
of Csound is planned to be used. Multichannel output (currently surround
5.1) will be used. The different voices will be placed at different
positions in a virtual room or are even moving in this room. The pulse
responses of the virtual room are/will be provided for the different
positions of the static voices or have to be calculated for the moving
voices during the performance. The voices have to be folded with the proper
pulse response during the performance. This does not impose a problem as
long as only few voices are active simultaneously. However, in case of a
large number voices being active at the same time either as a chorus or as a
cacophony I expect limitation by the CPU speed using a single core. Are
there any multicore capabilities implemented in Csound (e.g. distributing
the processing of different instruments to different cores, in this case the
convolution)?



--
View this message in context: http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800.html
Sent from the Csound - General mailing list archive at Nabble.com.

Csound mailing list
<a href="x-msg://39/user/SendEmail.jtp?type=node&amp;node=5745808&amp;i=2" target="_top" rel="nofollow" link="external" class="">[hidden email]
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

Csound mailing list <a href="x-msg://39/user/SendEmail.jtp?type=node&amp;node=5745808&amp;i=3" target="_top" rel="nofollow" link="external" class="">[hidden email] https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to https://github.com/csound/csound/issues Discussions of bugs and features can be posted here

---

If you reply to this email, your message will be added to the discussion below:

http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745808.html

To unsubscribe from Does Csound have multicore capabilities?, click here.
NAML

---

View this message in context: Re: [Csnd] Does Csound have multicore capabilities?
Sent from the Csound - General mailing list archive at Nabble.com.
Csound mailing list Csound@listserv.heanet.ie https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to https://github.com/csound/csound/issues Discussions of bugs and features can be posted here

thanks, I’m just getting started with Csound, I’m doing some simple examples using CsoundQt, like using ftconv with impulse response for surround 5.1 which I've generated the for the virtual room by another program. This really works nicely. The dry mono voice is placed by ftconv in surround at the right position and gets reverberation at the same time according the room parameters and the position at which the impulse response is  simulated.

You can prevent normalisation (rescaling) by setting a negative GEN number (-1 in this case). 
========================
Dr Victor Lazzarini
Dean of Arts, Celtic Studies and Philosophy,
Maynooth University,
Maynooth, Co Kildare, Ireland
Tel: 00 353 7086936
Fax: 00 353 1 7086952 

> On 3 Jan 2016, at 12:27, MarieCurie  wrote:
> 
> thanks, I’m just getting started with Csound, I’m doing some simple examples using CsoundQt, like using ftconv with impulse response for surround 5.1 which I've generated the for the virtual room by another program. This really works nicely. The dry mono voice is placed by ftconv in surround at the right position and gets reverberation at the same time according the room parameters and the position at which the impulse response is  simulated.
> 
> Btw. FTCONV works GEN01 (instead of GEN52 which is advised in the FLOSS manual for multichannel use) also well for surround when using 
> 
> giSurround	ftgen	0, 0, iftlen, 1, Sfile2, 0, 0, 0
> 
> not as in the example for Stereo in CsoundQt, where it is done independently for the different channels. This yields the wrong  reverberation because the correlation between the channels gets lost due to the arbitrary normalisation.
> 
> 
>> On 3 Jan 2016, at 13:06, Victor Lazzarini [via Csound] <[hidden email]> wrote:
>> 
>> That should not be a problem. You can
>> place your impulse responses in different
>> tables and use ftconv with the correct 
>> function table by passing the table 
>> number as a parameter to the instrument.
>> 
>> Victor Lazzarini
>> Dean of Arts, Celtic Studies, and Philosophy
>> Maynooth University
>> Ireland
>> 
>> On 3 Jan 2016, at 11:53, MarieCurie <[hidden email]> wrote:
>> 
>>> Thanks for the answers. Well I could imagine something like 20 to 50 voices. The problem I see may be the convolution of so many voices with different pulse response function in order to get automatically for voices placed closer to the sweet spot a larger ratio (and different time structure) in the volumes for the direct signal to the reflections that for voices far away.
>>> 
>>> Karin
>>>  
>>>> On 3 Jan 2016, at 12:37, hlolli [via Csound] <[hidden email]> wrote:
>>>> 
>>>> Just a question still. How many voices do you want for cacophony? Because yes if you are talking about numbers in thousands then yes its going to make things slow depending on buffer size and kr/sr value. Chorus/delay/pitch shift fx's could deliver convincing resault for lesser cpu consumption. But dont know for sure.
>>>> 
>>>> On Jan 3, 2016 12:30 PM, "Hlöðver Sigurðsson" <x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=0" target="_top" rel="nofollow" link="external" class="">[hidden email]> wrote:
>>>> You can read a recent paper John fittch wrote about how csound uses multithreading. But that is very "low level". Unlike supercollider's supernova csound has no way for user to control multithreading, besodes just understanding its concepts, and the -j4 compiler flag. There are many theories and many different opinions even within csound when and if multithreading is a good idea and for convolution i think parallelism would come in good use. Besides convolition, csound would handle formant synthesis very easily on modern machine. I just live coded 2 days ago with 4-5 voices fof plus many samples and poscil based synths and the computer didnt even sweat. So my point is for the optimization is that its better to do it the last because the things that slow your processor down are many times something much more simple that using multiple cores.
>>>> 
>>>> On Jan 3, 2016 11:13 AM, "MarieCurie" <x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=1" target="_top" rel="nofollow" link="external" class="">[hidden email]> wrote:
>>>> In my project the audio content (text to be spoken by synthetic voices) is
>>>> expected to change dynamically which means the live audio output capability
>>>> of Csound is planned to be used. Multichannel output (currently surround
>>>> 5.1) will be used. The different voices will be placed at different
>>>> positions in a virtual room or are even moving in this room. The pulse
>>>> responses of the virtual room are/will be provided for the different
>>>> positions of the static voices or have to be calculated for the moving
>>>> voices during the performance. The voices have to be folded with the proper
>>>> pulse response during the performance. This does not impose a problem as
>>>> long as only few voices are active simultaneously. However, in case of a
>>>> large number voices being active at the same time either as a chorus or as a
>>>> cacophony I expect limitation by the CPU speed using a single core. Are
>>>> there any multicore capabilities implemented in Csound (e.g. distributing
>>>> the processing of different instruments to different cores, in this case the
>>>> convolution)?
>>>> 
>>>> 
>>>> 
>>>> --
>>>> View this message in context: http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800.html
>>>> Sent from the Csound - General mailing list archive at Nabble.com.
>>>> 
>>>> Csound mailing list
>>>> x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=2" target="_top" rel="nofollow" link="external" class="">[hidden email]
>>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>>>> Send bugs reports to
>>>>         https://github.com/csound/csound/issues
>>>> Discussions of bugs and features can be posted here
>>>> Csound mailing list x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=3" target="_top" rel="nofollow" link="external" class="">[hidden email] https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to https://github.com/csound/csound/issues Discussions of bugs and features can be posted here 
>>>> 
>>>> If you reply to this email, your message will be added to the discussion below:
>>>> http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745808.html
>>>> To unsubscribe from Does Csound have multicore capabilities?, click here.
>>>> NAML
>>> 
>>> 
>>> View this message in context: Re: [Csnd] Does Csound have multicore capabilities?
>>> Sent from the Csound - General mailing list archive at Nabble.com.
>>> Csound mailing list [hidden email] https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to https://github.com/csound/csound/issues Discussions of bugs and features can be posted here
>> Csound mailing list [hidden email] https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to https://github.com/csound/csound/issues Discussions of bugs and features can be posted here 
>> 
>> If you reply to this email, your message will be added to the discussion below:
>> http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745813.html
>> To unsubscribe from Does Csound have multicore capabilities?, click here.
>> NAML
> 
> cheers, 
> 
> Karin
> 
> View this message in context: Re: [Csnd] Does Csound have multicore capabilities?
> Sent from the Csound - General mailing list archive at Nabble.com.
> Csound mailing list Csound@listserv.heanet.ie https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to https://github.com/csound/csound/issues Discussions of bugs and features can be posted here

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

Running multiple threads in Csound works pretty well for low granularity processing, e.g. many similar instruments or many large similar opcodes, with reasonably large ksmps of 100 or more. Multithreading overhead is high and finer grained pieces will run slower not faster. I should think that convolution is a good case for this but you will have to experiment.

If you use C++ you can use std::thread or OpenMP in your own code. You could write your instruments as opcodes, your Csound instrs would just call your opcodes to encapsulate all their processing.

Regards,
Mike

well, I just learnt about Csound shortly before Xmas when I was discussing
some problems with my project with a sound engineer from the TV. For the
framework which I used so far which is primitive compared to Csound (and not
supported since long) but had the capability of being accessed via script
which is needed for my project. For this software I’ve written C code which
does the simulation of a virtual room producing multichannel (surruound 5.1)
output. This code allows for different multichannel panning schemes (similar
to those in EXAMPLE 05B1 and eqn. 3 from V. Pulkki's article for the stereo
case) and calculates the reverberation to a desired order of reflections
depending on the dimensions of the virtual room, the reflectivity and the
positions of the source and the sweet spot. The path of the source can be
steered by a string for a valid arithmetic expression (standard arithmetics
including the standard C-functions) such that the source can be moved almost
freely in (x,y) coordinates of the virtual room. This code yields
information on 

- the angle (or position) of the source, 
- the distance dependent volume, 
- the position dependent time delay between direct and reflected signals,
- the position dependent volume ratio between direct and reflected signals,
- the informations on the room contained in the reflections,
- the information on the movement (Doppler effect and beats which differ for
the different channels because of different relative velocities w.r.t. the
different walls). 

As far as I understand by not he multichannel packages VBAP and Ambisonic 
address the first two points only. As far as I've understood Puckki's
article and the source code of VBAP in Csound it does the panning of the
phantom sources only between neighbouring loudspeakers which means you are
not able to put a phantom source at any position. This should be possible
with Ambisonic however it also does not include item 3-6 from the list
above. I certainly will have a look to this code.

An example of a source @440Hz moving with a velocity of 4m/sec in a room
with (x,y) 20m x 50m (speed of sound 343,2 m/sec) for 5.1 channels
(L-R-C-LFE-SL-SR) is shown below. One nicely sees the source passing the
sweet spot until it reaches the wall behind the listener then it slowly
moves until it reaches the right back corner, where it stops. (For
visualisation I've used Praat)
 




-----
cheers,

Karin
--
View this message in context: http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745829.html
Sent from the Csound - General mailing list archive at Nabble.com.

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

hi karin -

thanks for your feedback on the FLOSS manual example!

if you can write me your corrections to the example (including its 
number), i am happy to correct this.  or i could give you write access, 
in case you want to do it on your own (maybe adding a multi-channel 
example).

you can write me also off-list, if you prefer.

cheers -
	joachim


On 03/01/16 13:27, MarieCurie wrote:
> thanks, I’m just getting started with Csound, I’m doing some simple
> examples using CsoundQt, like using ftconv with impulse response for
> surround 5.1 which I've generated the for the virtual room by another
> program. This really works nicely. The dry mono voice is placed by
> ftconv in surround at the right position and gets reverberation at the
> same time according the room parameters and the position at which the
> impulse response is  simulated.
>
> Btw. FTCONV works GEN01 (instead of GEN52 which is advised in the FLOSS
> manual for multichannel use) also well for surround when using
>
> giSurround ftgen 0, 0, iftlen, 1, Sfile2, 0, 0, 0
>
> not as in the example for Stereo in CsoundQt, where it is done
> independently for the different channels. This yields the wrong
>   reverberation because the correlation between the channels gets lost
> due to the arbitrary normalisation.
>
>
>> On 3 Jan 2016, at 13:06, Victor Lazzarini [via Csound] <[hidden email]
>> > wrote:
>>
>> That should not be a problem. You can
>> place your impulse responses in different
>> tables and use ftconv with the correct
>> function table by passing the table
>> number as a parameter to the instrument.
>>
>> Victor Lazzarini
>> Dean of Arts, Celtic Studies, and Philosophy
>> Maynooth University
>> Ireland
>>
>> On 3 Jan 2016, at 11:53, MarieCurie <> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745813&i=0" target="_top"
>> rel="nofollow" link="external" class="">[hidden email]> wrote:
>>
>>> Thanks for the answers. Well I could imagine something like 20 to 50
>>> voices. The problem I see may be the convolution of so many voices
>>> with different pulse response function in order to get automatically
>>> for voices placed closer to the sweet spot a larger ratio (and
>>> different time structure) in the volumes for the direct signal to the
>>> reflections that for voices far away.
>>>
>>> Karin
>>>
>>>> On 3 Jan 2016, at 12:37, hlolli [via Csound] <>>> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745810&i=0"
>>>> target="_top" rel="nofollow" link="external" class="">[hidden
>>>> email]> wrote:
>>>>
>>>> Just a question still. How many voices do you want for cacophony?
>>>> Because yes if you are talking about numbers in thousands then yes
>>>> its going to make things slow depending on buffer size and kr/sr
>>>> value. Chorus/delay/pitch shift fx's could deliver convincing
>>>> resault for lesser cpu consumption. But dont know for sure.
>>>>
>>>> On Jan 3, 2016 12:30 PM, "Hlöðver Sigurðsson" <>>> class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=0"
>>>> target="_top" rel="nofollow" link="external" class="">[hidden
>>>> email]> wrote:
>>>>
>>>>     You can read a recent paper John fittch wrote about how csound
>>>>     uses multithreading. But that is very "low level". Unlike
>>>>     supercollider's supernova csound has no way for user to control
>>>>     multithreading, besodes just understanding its concepts, and the
>>>>     -j4 compiler flag. There are many theories and many different
>>>>     opinions even within csound when and if multithreading is a good
>>>>     idea and for convolution i think parallelism would come in good
>>>>     use. Besides convolition, csound would handle formant synthesis
>>>>     very easily on modern machine. I just live coded 2 days ago with
>>>>     4-5 voices fof plus many samples and poscil based synths and the
>>>>     computer didnt even sweat. So my point is for the optimization
>>>>     is that its better to do it the last because the things that
>>>>     slow your processor down are many times something much more
>>>>     simple that using multiple cores.
>>>>
>>>>     On Jan 3, 2016 11:13 AM, "MarieCurie" <>>>     class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=1"
>>>>     target="_top" rel="nofollow" link="external" class="">[hidden
>>>>     email]> wrote:
>>>>
>>>>         In my project the audio content (text to be spoken by
>>>>         synthetic voices) is
>>>>         expected to change dynamically which means the live audio
>>>>         output capability
>>>>         of Csound is planned to be used. Multichannel output
>>>>         (currently surround
>>>>         5.1) will be used. The different voices will be placed at
>>>>         different
>>>>         positions in a virtual room or are even moving in this room.
>>>>         The pulse
>>>>         responses of the virtual room are/will be provided for the
>>>>         different
>>>>         positions of the static voices or have to be calculated for
>>>>         the moving
>>>>         voices during the performance. The voices have to be folded
>>>>         with the proper
>>>>         pulse response during the performance. This does not impose
>>>>         a problem as
>>>>         long as only few voices are active simultaneously. However,
>>>>         in case of a
>>>>         large number voices being active at the same time either as
>>>>         a chorus or as a
>>>>         cacophony I expect limitation by the CPU speed using a
>>>>         single core. Are
>>>>         there any multicore capabilities implemented in Csound (e.g.
>>>>         distributing
>>>>         the processing of different instruments to different cores,
>>>>         in this case the
>>>>         convolution)?
>>>>
>>>>
>>>>
>>>>         --
>>>>         View this message in context:
>>>>         http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800.html
>>>>         Sent from the Csound - General mailing list archive at
>>>>         Nabble.com .
>>>>
>>>>         Csound mailing list
>>>>         >>>         class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=2"
>>>>         target="_top" rel="nofollow" link="external"
>>>>         class="">[hidden email]
>>>>         https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>>>>         Send bugs reports to
>>>>         https://github.com/csound/csound/issues
>>>>         Discussions of bugs and features can be posted here
>>>>
>>>> Csound mailing list >>> class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=3"
>>>> target="_top" rel="nofollow" link="external" class="">[hidden email]
>>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>>>> https://github.com/csound/csound/issues Discussions of bugs and
>>>> features can be posted here
>>>>
>>>> ------------------------------------------------------------------------
>>>> If you reply to this email, your message will be added to the
>>>> discussion below:
>>>> http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745808.html
>>>>
>>>> To unsubscribe from Does Csound have multicore capabilities?, click
>>>> here.
>>>> NAML
>>>> 
>>>>
>>>
>>>
>>> ------------------------------------------------------------------------
>>> View this message in context: Re: [Csnd] Does Csound have multicore
>>> capabilities?
>>> 
>>> Sent from the Csound - General mailing list archive
>>>  at
>>> Nabble.com .
>>> Csound mailing list >> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745813&i=1"
>>> target="_top" rel="nofollow" link="external" class="">[hidden email]
>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>>> https://github.com/csound/csound/issues Discussions of bugs and
>>> features can be posted here
>> Csound mailing list > href="x-msg://43/user/SendEmail.jtp?type=node&node=5745813&i=2" target="_top"
>> rel="nofollow" link="external" class="">[hidden email]
>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>> https://github.com/csound/csound/issues Discussions of bugs and
>> features can be posted here
>>
>> ------------------------------------------------------------------------
>> If you reply to this email, your message will be added to the
>> discussion below:
>> http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745813.html
>>
>> To unsubscribe from Does Csound have multicore capabilities?, click here.
>> NAML
>> 
>>
>
> cheers,
>
> Karin
>
> ------------------------------------------------------------------------
> View this message in context: Re: [Csnd] Does Csound have multicore
> capabilities?
> 
> Sent from the Csound - General mailing list archive
>  at
> Nabble.com.
> Csound mailing list Csound@listserv.heanet.ie
> 
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
> https://github.com/csound/csound/issues Discussions of bugs and features
> can be posted here

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

hi Joachim,


sorry, the ftconv example in the FLOSS manual looks fine. I mixed it with the example which comes with CsoundQt (Examples>Miscellaneous>Universal Convolution) which is buggy.

cheers,

Karin

> On 3 Jan 2016, at 19:49, joachim heintz  wrote:
> 
> hi karin -
> 
> thanks for your feedback on the FLOSS manual example!
> 
> if you can write me your corrections to the example (including its number), i am happy to correct this.  or i could give you write access, in case you want to do it on your own (maybe adding a multi-channel example).
> 
> you can write me also off-list, if you prefer.
> 
> cheers -
> 	joachim
> 
> 
> On 03/01/16 13:27, MarieCurie wrote:
>> thanks, I’m just getting started with Csound, I’m doing some simple
>> examples using CsoundQt, like using ftconv with impulse response for
>> surround 5.1 which I've generated the for the virtual room by another
>> program. This really works nicely. The dry mono voice is placed by
>> ftconv in surround at the right position and gets reverberation at the
>> same time according the room parameters and the position at which the
>> impulse response is  simulated.
>> 
>> Btw. FTCONV works GEN01 (instead of GEN52 which is advised in the FLOSS
>> manual for multichannel use) also well for surround when using
>> 
>> giSurround ftgen 0, 0, iftlen, 1, Sfile2, 0, 0, 0
>> 
>> not as in the example for Stereo in CsoundQt, where it is done
>> independently for the different channels. This yields the wrong
>>  reverberation because the correlation between the channels gets lost
>> due to the arbitrary normalisation.
>> 
>> 
>>> On 3 Jan 2016, at 13:06, Victor Lazzarini [via Csound] <[hidden email]
>>> > wrote:
>>> 
>>> That should not be a problem. You can
>>> place your impulse responses in different
>>> tables and use ftconv with the correct
>>> function table by passing the table
>>> number as a parameter to the instrument.
>>> 
>>> Victor Lazzarini
>>> Dean of Arts, Celtic Studies, and Philosophy
>>> Maynooth University
>>> Ireland
>>> 
>>> On 3 Jan 2016, at 11:53, MarieCurie <>> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745813&i=0" target="_top"
>>> rel="nofollow" link="external" class="">[hidden email]> wrote:
>>> 
>>>> Thanks for the answers. Well I could imagine something like 20 to 50
>>>> voices. The problem I see may be the convolution of so many voices
>>>> with different pulse response function in order to get automatically
>>>> for voices placed closer to the sweet spot a larger ratio (and
>>>> different time structure) in the volumes for the direct signal to the
>>>> reflections that for voices far away.
>>>> 
>>>> Karin
>>>> 
>>>>> On 3 Jan 2016, at 12:37, hlolli [via Csound] <>>>> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745810&i=0"
>>>>> target="_top" rel="nofollow" link="external" class="">[hidden
>>>>> email]> wrote:
>>>>> 
>>>>> Just a question still. How many voices do you want for cacophony?
>>>>> Because yes if you are talking about numbers in thousands then yes
>>>>> its going to make things slow depending on buffer size and kr/sr
>>>>> value. Chorus/delay/pitch shift fx's could deliver convincing
>>>>> resault for lesser cpu consumption. But dont know for sure.
>>>>> 
>>>>> On Jan 3, 2016 12:30 PM, "Hlöðver Sigurðsson" <>>>> class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=0"
>>>>> target="_top" rel="nofollow" link="external" class="">[hidden
>>>>> email]> wrote:
>>>>> 
>>>>>    You can read a recent paper John fittch wrote about how csound
>>>>>    uses multithreading. But that is very "low level". Unlike
>>>>>    supercollider's supernova csound has no way for user to control
>>>>>    multithreading, besodes just understanding its concepts, and the
>>>>>    -j4 compiler flag. There are many theories and many different
>>>>>    opinions even within csound when and if multithreading is a good
>>>>>    idea and for convolution i think parallelism would come in good
>>>>>    use. Besides convolition, csound would handle formant synthesis
>>>>>    very easily on modern machine. I just live coded 2 days ago with
>>>>>    4-5 voices fof plus many samples and poscil based synths and the
>>>>>    computer didnt even sweat. So my point is for the optimization
>>>>>    is that its better to do it the last because the things that
>>>>>    slow your processor down are many times something much more
>>>>>    simple that using multiple cores.
>>>>> 
>>>>>    On Jan 3, 2016 11:13 AM, "MarieCurie" <>>>>    class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=1"
>>>>>    target="_top" rel="nofollow" link="external" class="">[hidden
>>>>>    email]> wrote:
>>>>> 
>>>>>        In my project the audio content (text to be spoken by
>>>>>        synthetic voices) is
>>>>>        expected to change dynamically which means the live audio
>>>>>        output capability
>>>>>        of Csound is planned to be used. Multichannel output
>>>>>        (currently surround
>>>>>        5.1) will be used. The different voices will be placed at
>>>>>        different
>>>>>        positions in a virtual room or are even moving in this room.
>>>>>        The pulse
>>>>>        responses of the virtual room are/will be provided for the
>>>>>        different
>>>>>        positions of the static voices or have to be calculated for
>>>>>        the moving
>>>>>        voices during the performance. The voices have to be folded
>>>>>        with the proper
>>>>>        pulse response during the performance. This does not impose
>>>>>        a problem as
>>>>>        long as only few voices are active simultaneously. However,
>>>>>        in case of a
>>>>>        large number voices being active at the same time either as
>>>>>        a chorus or as a
>>>>>        cacophony I expect limitation by the CPU speed using a
>>>>>        single core. Are
>>>>>        there any multicore capabilities implemented in Csound (e.g.
>>>>>        distributing
>>>>>        the processing of different instruments to different cores,
>>>>>        in this case the
>>>>>        convolution)?
>>>>> 
>>>>> 
>>>>> 
>>>>>        --
>>>>>        View this message in context:
>>>>>        http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800.html
>>>>>        Sent from the Csound - General mailing list archive at
>>>>>        Nabble.com .
>>>>> 
>>>>>        Csound mailing list
>>>>>        >>>>        class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=2"
>>>>>        target="_top" rel="nofollow" link="external"
>>>>>        class="">[hidden email]
>>>>>        https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>>>>>        Send bugs reports to
>>>>>        https://github.com/csound/csound/issues
>>>>>        Discussions of bugs and features can be posted here
>>>>> 
>>>>> Csound mailing list >>>> class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=3"
>>>>> target="_top" rel="nofollow" link="external" class="">[hidden email]
>>>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>>>>> https://github.com/csound/csound/issues Discussions of bugs and
>>>>> features can be posted here
>>>>> 
>>>>> ------------------------------------------------------------------------
>>>>> If you reply to this email, your message will be added to the
>>>>> discussion below:
>>>>> http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745808.html
>>>>> 
>>>>> To unsubscribe from Does Csound have multicore capabilities?, click
>>>>> here.
>>>>> NAML
>>>>> 
>>>>> 
>>>> 
>>>> 
>>>> ------------------------------------------------------------------------
>>>> View this message in context: Re: [Csnd] Does Csound have multicore
>>>> capabilities?
>>>> 
>>>> Sent from the Csound - General mailing list archive
>>>>  at
>>>> Nabble.com .
>>>> Csound mailing list >>> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745813&i=1"
>>>> target="_top" rel="nofollow" link="external" class="">[hidden email]
>>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>>>> https://github.com/csound/csound/issues Discussions of bugs and
>>>> features can be posted here
>>> Csound mailing list >> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745813&i=2" target="_top"
>>> rel="nofollow" link="external" class="">[hidden email]
>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>>> https://github.com/csound/csound/issues Discussions of bugs and
>>> features can be posted here
>>> 
>>> ------------------------------------------------------------------------
>>> If you reply to this email, your message will be added to the
>>> discussion below:
>>> http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745813.html
>>> 
>>> To unsubscribe from Does Csound have multicore capabilities?, click here.
>>> NAML
>>> 
>>> 
>> 
>> cheers,
>> 
>> Karin
>> 
>> ------------------------------------------------------------------------
>> View this message in context: Re: [Csnd] Does Csound have multicore
>> capabilities?
>> 
>> Sent from the Csound - General mailing list archive
>>  at
>> Nabble.com.
>> Csound mailing list Csound@listserv.heanet.ie
>> 
>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>> https://github.com/csound/csound/issues Discussions of bugs and features
>> can be posted here
> 
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>       https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

thanks — and in a way even worse for me because i wrote this csoundqt 
example ... (ages ago)

would you mind to tell me what is buggy?  i think i used it some months 
ago and it seemed to work.

	joachim


On 04/01/16 09:03, Karin Daum wrote:
> hi Joachim,
>
>
> sorry, the ftconv example in the FLOSS manual looks fine. I mixed it with the example which comes with CsoundQt (Examples>Miscellaneous>Universal Convolution) which is buggy.
>
> cheers,
>
> Karin
>
>> On 3 Jan 2016, at 19:49, joachim heintz  wrote:
>>
>> hi karin -
>>
>> thanks for your feedback on the FLOSS manual example!
>>
>> if you can write me your corrections to the example (including its number), i am happy to correct this.  or i could give you write access, in case you want to do it on your own (maybe adding a multi-channel example).
>>
>> you can write me also off-list, if you prefer.
>>
>> cheers -
>> 	joachim
>>
>>
>> On 03/01/16 13:27, MarieCurie wrote:
>>> thanks, I’m just getting started with Csound, I’m doing some simple
>>> examples using CsoundQt, like using ftconv with impulse response for
>>> surround 5.1 which I've generated the for the virtual room by another
>>> program. This really works nicely. The dry mono voice is placed by
>>> ftconv in surround at the right position and gets reverberation at the
>>> same time according the room parameters and the position at which the
>>> impulse response is  simulated.
>>>
>>> Btw. FTCONV works GEN01 (instead of GEN52 which is advised in the FLOSS
>>> manual for multichannel use) also well for surround when using
>>>
>>> giSurround ftgen 0, 0, iftlen, 1, Sfile2, 0, 0, 0
>>>
>>> not as in the example for Stereo in CsoundQt, where it is done
>>> independently for the different channels. This yields the wrong
>>>   reverberation because the correlation between the channels gets lost
>>> due to the arbitrary normalisation.
>>>
>>>
>>>> On 3 Jan 2016, at 13:06, Victor Lazzarini [via Csound] <[hidden email]
>>>> > wrote:
>>>>
>>>> That should not be a problem. You can
>>>> place your impulse responses in different
>>>> tables and use ftconv with the correct
>>>> function table by passing the table
>>>> number as a parameter to the instrument.
>>>>
>>>> Victor Lazzarini
>>>> Dean of Arts, Celtic Studies, and Philosophy
>>>> Maynooth University
>>>> Ireland
>>>>
>>>> On 3 Jan 2016, at 11:53, MarieCurie <>>> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745813&i=0" target="_top"
>>>> rel="nofollow" link="external" class="">[hidden email]> wrote:
>>>>
>>>>> Thanks for the answers. Well I could imagine something like 20 to 50
>>>>> voices. The problem I see may be the convolution of so many voices
>>>>> with different pulse response function in order to get automatically
>>>>> for voices placed closer to the sweet spot a larger ratio (and
>>>>> different time structure) in the volumes for the direct signal to the
>>>>> reflections that for voices far away.
>>>>>
>>>>> Karin
>>>>>
>>>>>> On 3 Jan 2016, at 12:37, hlolli [via Csound] <>>>>> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745810&i=0"
>>>>>> target="_top" rel="nofollow" link="external" class="">[hidden
>>>>>> email]> wrote:
>>>>>>
>>>>>> Just a question still. How many voices do you want for cacophony?
>>>>>> Because yes if you are talking about numbers in thousands then yes
>>>>>> its going to make things slow depending on buffer size and kr/sr
>>>>>> value. Chorus/delay/pitch shift fx's could deliver convincing
>>>>>> resault for lesser cpu consumption. But dont know for sure.
>>>>>>
>>>>>> On Jan 3, 2016 12:30 PM, "Hlöðver Sigurðsson" <>>>>> class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=0"
>>>>>> target="_top" rel="nofollow" link="external" class="">[hidden
>>>>>> email]> wrote:
>>>>>>
>>>>>>     You can read a recent paper John fittch wrote about how csound
>>>>>>     uses multithreading. But that is very "low level". Unlike
>>>>>>     supercollider's supernova csound has no way for user to control
>>>>>>     multithreading, besodes just understanding its concepts, and the
>>>>>>     -j4 compiler flag. There are many theories and many different
>>>>>>     opinions even within csound when and if multithreading is a good
>>>>>>     idea and for convolution i think parallelism would come in good
>>>>>>     use. Besides convolition, csound would handle formant synthesis
>>>>>>     very easily on modern machine. I just live coded 2 days ago with
>>>>>>     4-5 voices fof plus many samples and poscil based synths and the
>>>>>>     computer didnt even sweat. So my point is for the optimization
>>>>>>     is that its better to do it the last because the things that
>>>>>>     slow your processor down are many times something much more
>>>>>>     simple that using multiple cores.
>>>>>>
>>>>>>     On Jan 3, 2016 11:13 AM, "MarieCurie" <>>>>>     class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=1"
>>>>>>     target="_top" rel="nofollow" link="external" class="">[hidden
>>>>>>     email]> wrote:
>>>>>>
>>>>>>         In my project the audio content (text to be spoken by
>>>>>>         synthetic voices) is
>>>>>>         expected to change dynamically which means the live audio
>>>>>>         output capability
>>>>>>         of Csound is planned to be used. Multichannel output
>>>>>>         (currently surround
>>>>>>         5.1) will be used. The different voices will be placed at
>>>>>>         different
>>>>>>         positions in a virtual room or are even moving in this room.
>>>>>>         The pulse
>>>>>>         responses of the virtual room are/will be provided for the
>>>>>>         different
>>>>>>         positions of the static voices or have to be calculated for
>>>>>>         the moving
>>>>>>         voices during the performance. The voices have to be folded
>>>>>>         with the proper
>>>>>>         pulse response during the performance. This does not impose
>>>>>>         a problem as
>>>>>>         long as only few voices are active simultaneously. However,
>>>>>>         in case of a
>>>>>>         large number voices being active at the same time either as
>>>>>>         a chorus or as a
>>>>>>         cacophony I expect limitation by the CPU speed using a
>>>>>>         single core. Are
>>>>>>         there any multicore capabilities implemented in Csound (e.g.
>>>>>>         distributing
>>>>>>         the processing of different instruments to different cores,
>>>>>>         in this case the
>>>>>>         convolution)?
>>>>>>
>>>>>>
>>>>>>
>>>>>>         --
>>>>>>         View this message in context:
>>>>>>         http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800.html
>>>>>>         Sent from the Csound - General mailing list archive at
>>>>>>         Nabble.com .
>>>>>>
>>>>>>         Csound mailing list
>>>>>>         >>>>>         class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=2"
>>>>>>         target="_top" rel="nofollow" link="external"
>>>>>>         class="">[hidden email]
>>>>>>         https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>>>>>>         Send bugs reports to
>>>>>>         https://github.com/csound/csound/issues
>>>>>>         Discussions of bugs and features can be posted here
>>>>>>
>>>>>> Csound mailing list >>>>> class="">x-msg://39/user/SendEmail.jtp?type=node&node=5745808&i=3"
>>>>>> target="_top" rel="nofollow" link="external" class="">[hidden email]
>>>>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>>>>>> https://github.com/csound/csound/issues Discussions of bugs and
>>>>>> features can be posted here
>>>>>>
>>>>>> ------------------------------------------------------------------------
>>>>>> If you reply to this email, your message will be added to the
>>>>>> discussion below:
>>>>>> http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745808.html
>>>>>>
>>>>>> To unsubscribe from Does Csound have multicore capabilities?, click
>>>>>> here.
>>>>>> NAML
>>>>>> 
>>>>>>
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------------
>>>>> View this message in context: Re: [Csnd] Does Csound have multicore
>>>>> capabilities?
>>>>> 
>>>>> Sent from the Csound - General mailing list archive
>>>>>  at
>>>>> Nabble.com .
>>>>> Csound mailing list >>>> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745813&i=1"
>>>>> target="_top" rel="nofollow" link="external" class="">[hidden email]
>>>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>>>>> https://github.com/csound/csound/issues Discussions of bugs and
>>>>> features can be posted here
>>>> Csound mailing list >>> href="x-msg://43/user/SendEmail.jtp?type=node&node=5745813&i=2" target="_top"
>>>> rel="nofollow" link="external" class="">[hidden email]
>>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>>>> https://github.com/csound/csound/issues Discussions of bugs and
>>>> features can be posted here
>>>>
>>>> ------------------------------------------------------------------------
>>>> If you reply to this email, your message will be added to the
>>>> discussion below:
>>>> http://csound.1045644.n5.nabble.com/Does-Csound-have-multicore-capabilities-tp5745800p5745813.html
>>>>
>>>> To unsubscribe from Does Csound have multicore capabilities?, click here.
>>>> NAML
>>>> 
>>>>
>>>
>>> cheers,
>>>
>>> Karin
>>>
>>> ------------------------------------------------------------------------
>>> View this message in context: Re: [Csnd] Does Csound have multicore
>>> capabilities?
>>> 
>>> Sent from the Csound - General mailing list archive
>>>  at
>>> Nabble.com.
>>> Csound mailing list Csound@listserv.heanet.ie
>>> 
>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND Send bugs reports to
>>> https://github.com/csound/csound/issues Discussions of bugs and features
>>> can be posted here
>>
>> Csound mailing list
>> Csound@listserv.heanet.ie
>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>> Send bugs reports to
>>        https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>          https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
>

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

hi Joachim,

MultiChannelPulseResponseConvolution.csd (51K) Download Attachment
1pingHall000_2.5_05_3611mode0.wav (6M) Download Attachment
faltungshall_mono.wav (3M) Download Attachment