hi, i would like to know some options to do when a single cpu thread 
goes to 100% with csound and producing audio glitches while the other 
cpu threads  idle. i have tried the option e.g -j6 but this results in 
600% cpu  (i.e. 6 cpu threads goto 100%) and same glitches. i wonder for 
what this parameter -j is used actually.

are there some options to use the other cpu threads as well in a manner 
of balancing the computing?

i have minimal global variables - almost all is done in instruments. use 
the channel system for control parameters, in/outleta/connect for audio 
routing and i am already on ksmps=64 (with B=512 and b=128 and a NI 
Komplete Audio 6 interface on a 6 core 2.6-4GHZ linux laptop )

are there concepts/ideas to handle this besides starting several csound 
processes which may have some uncomfortable issues?

thx.

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

The -j option works best with global variables as they are distinct while 
all channels are treated as one global.  There are papers in LAC and ICMC 
that describe the parallelism on multi-core.  If you run with -v option 
you should see the available parallelism laid out as semantic analysis. 
May not be easy to red but I can explain it.

Use of larger ksmps and --sample-accurate also reduces load

Or you can roll you own parallel execution with the API.

We would need to see what you were doing to give further advice

On Wed, 26 Feb 2020, Christian Bacherr wrote:

> hi, i would like to know some options to do when a single cpu thread goes to 
> 100% with csound and producing audio glitches while the other cpu threads  
> idle. i have tried the option e.g -j6 but this results in 600% cpu  (i.e. 6 
> cpu threads goto 100%) and same glitches. i wonder for what this parameter -j 
> is used actually.
>
> are there some options to use the other cpu threads as well in a manner of 
> balancing the computing?
>
> i have minimal global variables - almost all is done in instruments. use the 
> channel system for control parameters, in/outleta/connect for audio routing 
> and i am already on ksmps=64 (with B=512 and b=128 and a NI Komplete Audio 6 
> interface on a 6 core 2.6-4GHZ linux laptop )
>
> are there concepts/ideas to handle this besides starting several csound 
> processes which may have some uncomfortable issues?
>
> thx.
>
> 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

(unfortunately the -v option produces an init log file of >50000 lines so it is kind of hard to read)

My goal would be still to try to enabling the use of more cpu threads.

Of course i may speed up some stuff with some measures but some time later i could face the same problem again.

Currently i can generate the csd with a script which takes the instruments files as input (i.e. including them in the csd and generate a mixer instrument inside)

The next question would be: Does the use of outleta/inleta/connect influence the possibility of concurrent processing?

A side question: Does the chebychevpoly opcode with 11 harmonics consume a lot of cpu?

Thx and best regards,

Christian

After re-reading your email I advise you to use a single instrument definition with a single instance to manage all control channels. This should be an always on instrument that reads channel values and assigns them to global variables that are distinguished by instrument number, name, or ID. This works as well for sending values, the other instruments can assign values to global variables that the one always on control channel instrument sends to the appropriate named channels. Doing this has significantly speeded up my orchestras. For an example of how this works you may read this, https://github.com/gogins/csound-extended/blob/develop/silencio/patches/modular_csound.pdf, page 5.

The key to this system is to build the instrument names or numbers, and parameter names, into the global variable names which are exactly the same as the channel names.

Regards,

Mike

-----------------------------------------------------

Michael Gogins
Irreducible Productions
http://michaelgogins.tumblr.com
Michael dot Gogins at gmail dot com

On Wed, Feb 26, 2020 at 8:00 PM Michael Gogins <michael.gogins@gmail.com> wrote:

The -j option is more efficient than a single thread only in certain cases. These are when there are a few instrument definitions and several copies of each instrument definition running for a fairly long time, or some combination of these factors. In such cases I have observed a significant speedup, e.g. -j4 running almost 2 x as fast. Without these favorable factors that essentially enable instances of the same instrument definition to run each in a separate thread without incurring too much overhead, most likely using more than one thread will actually degrade performance.

Regards,

Mike

-----------------------------------------------------

Michael Gogins
Irreducible Productions
http://michaelgogins.tumblr.com
Michael dot Gogins at gmail dot com

On Wed, Feb 26, 2020 at 7:05 PM Christian Bacherr <000005d2745f1ec0-dmarc-request@listserv.heanet.ie> wrote:

hi, i would like to know some options to do when a single cpu thread
goes to 100% with csound and producing audio glitches while the other
cpu threads  idle. i have tried the option e.g -j6 but this results in
600% cpu  (i.e. 6 cpu threads goto 100%) and same glitches. i wonder for
what this parameter -j is used actually.

are there some options to use the other cpu threads as well in a manner
of balancing the computing?

i have minimal global variables - almost all is done in instruments. use
the channel system for control parameters, in/outleta/connect for audio
routing and i am already on ksmps=64 (with B=512 and b=128 and a NI
Komplete Audio 6 interface on a 6 core 2.6-4GHZ linux laptop )

are there concepts/ideas to handle this besides starting several csound
processes which may have some uncomfortable issues?

thx.

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