Hi

While porting a project from supercollider to csound I noticed that csound uses a considerable amount of cpu just for signal output. When run in real time, the csd below uses ~45% cpu for just copying to spout. If instead of using "out" in instr 1 the signals are accumulated in a global variable and output once in another instrument, cpu drops down to ~30%. Is this something known which should be avoided? The same processing in supercollider (outputting 1000 silent signals), results in ~15% CPU

<CsInstruments>
sr = 48000
nchnls = 2
ksmps = 128
0dbfs = 1

instr 1
  a1 init 0
  out a1
endin

instr 2  
  iN = 1000
  idx = 0
  while idx < iN do
    event_i "i", 1, 0, 100, 1/iN
    idx += 1
  od 
  turnoff
endin
</CsInstruments>

<CsScore>
i2 0 100
</CsScore>


(
1000.do {
  { Out.ar(0, Silent.ar(1)) }.play
}
)

Is your build optimised? (-O3)
========================
Prof. 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 Feb 2017, at 10:02, Eduardo Moguillansky  wrote:
> 
> Hi
> 
> While porting a project from supercollider to csound I noticed that csound uses a considerable amount of cpu just for signal output. When run in real time, the csd below uses ~45% cpu for just copying to spout. If instead of using "out" in instr 1 the signals are accumulated in a global variable and output once in another instrument, cpu drops down to ~30%. Is this something known which should be avoided? The same processing in supercollider (outputting 1000 silent signals), results in ~15% CPU
> 
> 
> 
> sr = 48000
> nchnls = 2
> ksmps = 128
> 0dbfs = 1
> 
> 
> instr 1
> 
>   a1 init 0
>   out a1
> 
> endin
> 
> instr 2  
> 
>   iN = 1000
>   idx = 0
>   while idx < iN do
>     event_i "i", 1, 0, 100, 1/iN
>     idx += 1
>   od
>   turnoff
> endin
> 
> 
> 
> i2 0 100
> 
> 
> 
> (
> 1000.do {
>   { Out.ar(0, Silent.ar(1)) }.play
> }
> )
> 
> 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

But I also noted this here. Your code runs for about 9s for 100s of output; with a single global out in another instrument,
it runs for 5s. But if I replace out for outch 1, a1, it runs for 4s. So it’s out that seems to be slow. Needs a check.

========================
Prof. 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 Feb 2017, at 11:38, Victor Lazzarini  wrote:
> 
> Is your build optimised? (-O3)
> ========================
> Prof. 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 Feb 2017, at 10:02, Eduardo Moguillansky  wrote:
>> 
>> Hi
>> 
>> While porting a project from supercollider to csound I noticed that csound uses a considerable amount of cpu just for signal output. When run in real time, the csd below uses ~45% cpu for just copying to spout. If instead of using "out" in instr 1 the signals are accumulated in a global variable and output once in another instrument, cpu drops down to ~30%. Is this something known which should be avoided? The same processing in supercollider (outputting 1000 silent signals), results in ~15% CPU
>> 
>> 
>> 
>> sr = 48000
>> nchnls = 2
>> ksmps = 128
>> 0dbfs = 1
>> 
>> 
>> instr 1
>> 
>>  a1 init 0
>>  out a1
>> 
>> endin
>> 
>> instr 2  
>> 
>>  iN = 1000
>>  idx = 0
>>  while idx < iN do
>>    event_i "i", 1, 0, 100, 1/iN
>>    idx += 1
>>  od
>>  turnoff
>> endin
>> 
>> 
>> 
>> i2 0 100
>> 
>> 
>> 
>> (
>> 1000.do {
>>  { Out.ar(0, Silent.ar(1)) }.play
>> }
>> )
>> 
>> 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

Yes, I think so. I do

mkdir build

cd build

cmake-gui ..  

make

sudo make install

After having done that I this in build/CMakeCache.txt

Is your build optimised? (-O3)
========================
Prof. 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 Feb 2017, at 10:02, Eduardo Moguillansky <eduardo.moguillansky@gmail.com> wrote:

Hi

While porting a project from supercollider to csound I noticed that csound uses a considerable amount of cpu just for signal output. When run in real time, the csd below uses ~45% cpu for just copying to spout. If instead of using "out" in instr 1 the signals are accumulated in a global variable and output once in another instrument, cpu drops down to ~30%. Is this something known which should be avoided? The same processing in supercollider (outputting 1000 silent signals), results in ~15% CPU

<CsInstruments>

sr = 48000
nchnls = 2
ksmps = 128
0dbfs = 1


instr 1

  a1 init 0
  out a1

endin

instr 2  

  iN = 1000
  idx = 0
  while idx < iN do
    event_i "i", 1, 0, 100, 1/iN
    idx += 1
  od
  turnoff
endin
</CsInstruments>

<CsScore>
i2 0 100
</CsScore>


(
1000.do {
  { Out.ar(0, Silent.ar(1)) }.play
}
)

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

I tried the different out opcodes and all have the same impact. Can it be the spinlocks?

But I also noted this here. Your code runs for about 9s for 100s of output; with a single global out in another instrument,
it runs for 5s. But if I replace out for outch 1, a1, it runs for 4s. So it’s out that seems to be slow. Needs a check.

========================
Prof. 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 Feb 2017, at 11:38, Victor Lazzarini <Victor.Lazzarini@nuim.ie> wrote:

Is your build optimised? (-O3)
========================
Prof. 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 Feb 2017, at 10:02, Eduardo Moguillansky <eduardo.moguillansky@gmail.com> wrote:

Hi

While porting a project from supercollider to csound I noticed that csound uses a considerable amount of cpu just for signal output. When run in real time, the csd below uses ~45% cpu for just copying to spout. If instead of using "out" in instr 1 the signals are accumulated in a global variable and output once in another instrument, cpu drops down to ~30%. Is this something known which should be avoided? The same processing in supercollider (outputting 1000 silent signals), results in ~15% CPU

<CsInstruments>

sr = 48000
nchnls = 2
ksmps = 128
0dbfs = 1


instr 1

 a1 init 0
 out a1

endin

instr 2  

 iN = 1000
 idx = 0
 while idx < iN do
   event_i "i", 1, 0, 100, 1/iN
   idx += 1
 od
 turnoff
endin
</CsInstruments>

<CsScore>
i2 0 100
</CsScore>


(
1000.do {
 { Out.ar(0, Silent.ar(1)) }.play
}
)

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

I tried the different out opcodes and all have the same impact. Can it be the spinlocks?

On 03.02.2017 13:01, Victor Lazzarini wrote:

But I also noted this here. Your code runs for about 9s for 100s of output; with a single global out in another instrument,
it runs for 5s. But if I replace out for outch 1, a1, it runs for 4s. So it’s out that seems to be slow. Needs a check.

========================
Prof. 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 Feb 2017, at 11:38, Victor Lazzarini <Victor.Lazzarini@nuim.ie> wrote:

Is your build optimised? (-O3)
========================
Prof. 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 Feb 2017, at 10:02, Eduardo Moguillansky <eduardo.moguillansky@gmail.com> wrote:

Hi

While porting a project from supercollider to csound I noticed that csound uses a considerable amount of cpu just for signal output. When run in real time, the csd below uses ~45% cpu for just copying to spout. If instead of using "out" in instr 1 the signals are accumulated in a global variable and output once in another instrument, cpu drops down to ~30%. Is this something known which should be avoided? The same processing in supercollider (outputting 1000 silent signals), results in ~15% CPU

<CsInstruments>

sr = 48000
nchnls = 2
ksmps = 128
0dbfs = 1


instr 1

 a1 init 0
 out a1

endin

instr 2  

 iN = 1000
 idx = 0
 while idx < iN do
   event_i "i", 1, 0, 100, 1/iN
   idx += 1
 od
 turnoff
endin
</CsInstruments>

<CsScore>
i2 0 100
</CsScore>


(
1000.do {
 { Out.ar(0, Silent.ar(1)) }.play
}
)

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

Could be.  All the code does is protect the adding in of the signal.  On a uniprocessor the protection should cost little.

Sent from TypeApp

On 3 Feb 2017, at 12:06, Eduardo Moguillansky <eduardo.moguillansky@GMAIL.COM> wrote:

I tried the different out opcodes and all have the same impact. Can it be the spinlocks?

On 03.02.2017 13:01, Victor Lazzarini wrote:

But I also noted this here. Your code runs for about 9s for 100s of output; with a single global out in another instrument,
it runs for 5s. But if I replace out for outch 1, a1, it runs for 4s. So it’s out that seems to be slow. Needs a check.

========================
Prof. 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 Feb 2017, at 11:38, Victor Lazzarini <Victor.Lazzarini@nuim.ie> wrote:

Is your build optimised? (-O3)
========================
Prof. 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 Feb 2017, at 10:02, Eduardo Moguillansky <eduardo.moguillansky@gmail.com> wrote:

Hi

While porting a project from supercollider to csound I noticed that csound uses a considerable amount of cpu just for signal output. When run in real time, the csd below uses ~45% cpu for just copying to spout. If instead of using "out" in instr 1 the signals are accumulated in a global variable and output once in another instrument, cpu drops down to ~30%. Is this something known which should be avoided? The same processing in supercollider (outputting 1000 silent signals), results in ~15% CPU

<CsInstruments>

sr = 48000
nchnls = 2
ksmps = 128
0dbfs = 1


instr 1

 a1 init 0
 out a1

endin

instr 2  

 iN = 1000
 idx = 0
 while idx < iN do
   event_i "i", 1, 0, 100, 1/iN
   idx += 1
 od
 turnoff
endin
</CsInstruments>

<CsScore>
i2 0 100
</CsScore>


(
1000.do {
 { Out.ar(0, Silent.ar(1)) }.play
}
)

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

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

I have checked this and you are right, spinlocks costs nothing. The cost is solely in the copying of data to an interleaved format using
a loop. Replacing the loop by a memcpy brings the cost down to < 1/3 of the original. Of course that only works for mono.
========================
Prof. 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 Feb 2017, at 13:14, Victor Lazzarini  wrote:
> 
> Here on OSX, cost of outch is 1/2 that of out.
> 
> Victor Lazzarini
> Dean of Arts, Celtic Studies, and Philosophy
> Maynooth University
> Ireland
> 
> On 3 Feb 2017, at 12:05, Eduardo Moguillansky  wrote:
> 
>> I tried the different out opcodes and all have the same impact. Can it be the spinlocks? 
>> 
>> On 03.02.2017 13:01, Victor Lazzarini wrote:
>>> But I also noted this here. Your code runs for about 9s for 100s of output; with a single global out in another instrument,
>>> it runs for 5s. But if I replace out for outch 1, a1, it runs for 4s. So it’s out that seems to be slow. Needs a check.
>>> 
>>> ========================
>>> Prof. 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 Feb 2017, at 11:38, Victor Lazzarini 
>>>>  wrote:
>>>> 
>>>> Is your build optimised? (-O3)
>>>> ========================
>>>> Prof. 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 Feb 2017, at 10:02, Eduardo Moguillansky 
>>>>>  wrote:
>>>>> 
>>>>> Hi
>>>>> 
>>>>> While porting a project from supercollider to csound I noticed that csound uses a considerable amount of cpu just for signal output. When run in real time, the csd below uses ~45% cpu for just copying to spout. If instead of using "out" in instr 1 the signals are accumulated in a global variable and output once in another instrument, cpu drops down to ~30%. Is this something known which should be avoided? The same processing in supercollider (outputting 1000 silent signals), results in ~15% CPU
>>>>> 
>>>>> 
>>>>> 
>>>>> sr = 48000
>>>>> nchnls = 2
>>>>> ksmps = 128
>>>>> 0dbfs = 1
>>>>> 
>>>>> 
>>>>> instr 1
>>>>> 
>>>>>  a1 init 0
>>>>>  out a1
>>>>> 
>>>>> endin
>>>>> 
>>>>> instr 2  
>>>>> 
>>>>>  iN = 1000
>>>>>  idx = 0
>>>>>  while idx < iN do
>>>>>    event_i "i", 1, 0, 100, 1/iN
>>>>>    idx += 1
>>>>>  od
>>>>>  turnoff
>>>>> endin
>>>>> 
>>>>> 
>>>>> 
>>>>> i2 0 100
>>>>> 
>>>>> 
>>>>> 
>>>>> (
>>>>> 1000.do {
>>>>>  { 
>>>>> Out.ar(0, Silent.ar
>>>>> (1)) }.play
>>>>> }
>>>>> )
>>>>> 
>>>>> 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
> 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

sorry do t undestand.   i would epet about 100% cpu as it is doing nothing 
else.  What are you measurg?  I am seeing it un in 9s at 99.%


On Fri, 3 Feb 2017, Eduardo Moguillansky wrote:

> 
> Hi
> 
> While porting a project from supercollider to csound I noticed that csound
> uses a considerable amount of cpu just for signal output. When run in real
> time, the csd below uses ~45% cpu for just copying to spout. If instead of
> using "out" in instr 1 the signals are accumulated in a global variable and
> output once in another instrument, cpu drops down to ~30%. Is this something
> known which should be avoided? The same processing in supercollider
> (outputting 1000 silent signals), results in ~15% CPU
> 
> 
> sr = 48000
> nchnls = 2
> ksmps = 128
> 0dbfs = 1
> 
> instr 1
>   a1 init 0
>   out a1
> endin
> 
> instr 2
>   iN = 1000
>   idx = 0
>   while idx < iN do
>     event_i "i", 1, 0, 100, 1/iN
>     idx += 1
>   od
>   turnoff
> endin
> 
> 
> 
> i2 0 100
> 
> 
> 
> (
> 1000.do {
>   { Out.ar(0, Silent.ar(1)) }.play
> }
> )
> 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

I suspect he is running in realtime. My measurement is running it with -n and checking the processing time.
========================
Prof. 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 Feb 2017, at 14:39, jpff  wrote:
> 
> sorry do t undestand.   i would epet about 100% cpu as it is doing nothing else.  What are you measurg?  I am seeing it un in 9s at 99.%
> 
> 
> On Fri, 3 Feb 2017, Eduardo Moguillansky wrote:
> 
>> Hi
>> While porting a project from supercollider to csound I noticed that csound
>> uses a considerable amount of cpu just for signal output. When run in real
>> time, the csd below uses ~45% cpu for just copying to spout. If instead of
>> using "out" in instr 1 the signals are accumulated in a global variable and
>> output once in another instrument, cpu drops down to ~30%. Is this something
>> known which should be avoided? The same processing in supercollider
>> (outputting 1000 silent signals), results in ~15% CPU
>> 
>> sr = 48000
>> nchnls = 2
>> ksmps = 128
>> 0dbfs = 1
>> instr 1
>>  a1 init 0
>>  out a1
>> endin
>> instr 2
>>  iN = 1000
>>  idx = 0
>>  while idx < iN do
>>    event_i "i", 1, 0, 100, 1/iN
>>    idx += 1
>>  od
>>  turnoff
>> endin
>> 
>> 
>> i2 0 100
>> 
>> (
>> 1000.do {
>>  { Out.ar(0, Silent.ar(1)) }.play
>> }
>> )
>> 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

I was running in realtime in order to compare the performance with supercollider. The percents given are not significant in themselves, but I would have expected to see a similar performance of "out" vs accumulating in an a-variable and doing one "out" at the end of each k-cycle.

sorry do t undestand.   i would epet about 100% cpu as it is doing nothing else.  What are you measurg?  I am seeing it un in 9s at 99.%


On Fri, 3 Feb 2017, Eduardo Moguillansky wrote:

Hi

While porting a project from supercollider to csound I noticed that csound
uses a considerable amount of cpu just for signal output. When run in real
time, the csd below uses ~45% cpu for just copying to spout. If instead of
using "out" in instr 1 the signals are accumulated in a global variable and
output once in another instrument, cpu drops down to ~30%. Is this something
known which should be avoided? The same processing in supercollider
(outputting 1000 silent signals), results in ~15% CPU

<CsInstruments>
sr = 48000
nchnls = 2
ksmps = 128
0dbfs = 1

instr 1
  a1 init 0
  out a1
endin

instr 2
  iN = 1000
  idx = 0
  while idx < iN do
    event_i "i", 1, 0, 100, 1/iN
    idx += 1
  od
  turnoff
endin
</CsInstruments>

<CsScore>
i2 0 100
</CsScore>


(
1000.do {
  { Out.ar(0, Silent.ar(1)) }.play
}
)
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

No, because the difference is that  ga1 += a1  is cheap and then we only have
to interleave the data to spout once. When the instrument is writing to out, we have
1000 times the loop that interleaves the data. So it’s more expensive.

========================
Prof. 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 Feb 2017, at 14:59, Eduardo Moguillansky  wrote:
> 
> I was running in realtime in order to compare the performance with supercollider. The percents given are not significant in themselves, but I would have expected to see a similar performance of "out" vs accumulating in an a-variable and doing one "out" at the end of each k-cycle. 
> 
> On 03.02.2017 15:39, jpff wrote:
>> sorry do t undestand.   i would epet about 100% cpu as it is doing nothing else.  What are you measurg?  I am seeing it un in 9s at 99.% 
>> 
>> 
>> On Fri, 3 Feb 2017, Eduardo Moguillansky wrote: 
>> 
>>> 
>>> Hi 
>>> 
>>> While porting a project from supercollider to csound I noticed that csound 
>>> uses a considerable amount of cpu just for signal output. When run in real 
>>> time, the csd below uses ~45% cpu for just copying to spout. If instead of 
>>> using "out" in instr 1 the signals are accumulated in a global variable and 
>>> output once in another instrument, cpu drops down to ~30%. Is this something 
>>> known which should be avoided? The same processing in supercollider 
>>> (outputting 1000 silent signals), results in ~15% CPU 
>>> 
>>>  
>>> sr = 48000 
>>> nchnls = 2 
>>> ksmps = 128 
>>> 0dbfs = 1 
>>> 
>>> instr 1 
>>>   a1 init 0 
>>>   out a1 
>>> endin 
>>> 
>>> instr 2 
>>>   iN = 1000 
>>>   idx = 0 
>>>   while idx < iN do 
>>>     event_i "i", 1, 0, 100, 1/iN 
>>>     idx += 1 
>>>   od 
>>>   turnoff 
>>> endin 
>>>  
>>> 
>>>  
>>> i2 0 100 
>>>  
>>> 
>>> 
>>> ( 
>>> 1000.do { 
>>>   { Out.ar(0, Silent.ar(1)) }.play 
>>> } 
>>> ) 
>>> 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

And couldn't this be done, so that the out opcodes accumulate all writes to a contiguous array and interleave to spout at the end of the cycle?

No, because the difference is that  ga1 += a1  is cheap and then we only have
to interleave the data to spout once. When the instrument is writing to out, we have
1000 times the loop that interleaves the data. So it’s more expensive.

========================
Prof. 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 Feb 2017, at 14:59, Eduardo Moguillansky <eduardo.moguillansky@GMAIL.COM> wrote:

I was running in realtime in order to compare the performance with supercollider. The percents given are not significant in themselves, but I would have expected to see a similar performance of "out" vs accumulating in an a-variable and doing one "out" at the end of each k-cycle. 

On 03.02.2017 15:39, jpff wrote:

sorry do t undestand.   i would epet about 100% cpu as it is doing nothing else.  What are you measurg?  I am seeing it un in 9s at 99.% 


On Fri, 3 Feb 2017, Eduardo Moguillansky wrote: 

Hi 

While porting a project from supercollider to csound I noticed that csound 
uses a considerable amount of cpu just for signal output. When run in real 
time, the csd below uses ~45% cpu for just copying to spout. If instead of 
using "out" in instr 1 the signals are accumulated in a global variable and 
output once in another instrument, cpu drops down to ~30%. Is this something 
known which should be avoided? The same processing in supercollider 
(outputting 1000 silent signals), results in ~15% CPU 

<CsInstruments> 
sr = 48000 
nchnls = 2 
ksmps = 128 
0dbfs = 1 

instr 1 
  a1 init 0 
  out a1 
endin 

instr 2 
  iN = 1000 
  idx = 0 
  while idx < iN do 
    event_i "i", 1, 0, 100, 1/iN 
    idx += 1 
  od 
  turnoff 
endin 
</CsInstruments> 

<CsScore> 
i2 0 100 
</CsScore> 


( 
1000.do { 
  { Out.ar(0, Silent.ar(1)) }.play 
} 
) 
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

The only fix i can think of is te per instance spout whic we have 
discussed before; ot make spout not intertwinned and fix nly once

On Fri, 3 Feb 2017, Victor Lazzarini wrote:

> No, because the difference is that  ga1 += a1  is cheap and then we only have
> to interleave the data to spout once. When the instrument is writing to out, we have
> 1000 times the loop that interleaves the data. So it’s more expensive.
>
> ========================
> Prof. 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 Feb 2017, at 14:59, Eduardo Moguillansky  wrote:
>>
>> I was running in realtime in order to compare the performance with supercollider. The percents given are not significant in themselves, but I would have expected to see a similar performance of "out" vs accumulating in an a-variable and doing one "out" at the end of each k-cycle.
>>
>> On 03.02.2017 15:39, jpff wrote:
>>> sorry do t undestand.   i would epet about 100% cpu as it is doing nothing else.  What are you measurg?  I am seeing it un in 9s at 99.%
>>>
>>>
>>> On Fri, 3 Feb 2017, Eduardo Moguillansky wrote:
>>>
>>>>
>>>> Hi
>>>>
>>>> While porting a project from supercollider to csound I noticed that csound
>>>> uses a considerable amount of cpu just for signal output. When run in real
>>>> time, the csd below uses ~45% cpu for just copying to spout. If instead of
>>>> using "out" in instr 1 the signals are accumulated in a global variable and
>>>> output once in another instrument, cpu drops down to ~30%. Is this something
>>>> known which should be avoided? The same processing in supercollider
>>>> (outputting 1000 silent signals), results in ~15% CPU
>>>>
>>>> 
>>>> sr = 48000
>>>> nchnls = 2
>>>> ksmps = 128
>>>> 0dbfs = 1
>>>>
>>>> instr 1
>>>>   a1 init 0
>>>>   out a1
>>>> endin
>>>>
>>>> instr 2
>>>>   iN = 1000
>>>>   idx = 0
>>>>   while idx < iN do
>>>>     event_i "i", 1, 0, 100, 1/iN
>>>>     idx += 1
>>>>   od
>>>>   turnoff
>>>> endin
>>>> 
>>>>
>>>> 
>>>> i2 0 100
>>>> 
>>>>
>>>>
>>>> (
>>>> 1000.do {
>>>>   { Out.ar(0, Silent.ar(1)) }.play
>>>> }
>>>> )
>>>> 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
>

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

We could try and see what exactly is the slowdown, but yes, that could be done alright.
========================
Prof. 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 Feb 2017, at 15:12, Eduardo Moguillansky  wrote:
> 
> And couldn't this be done, so that the out opcodes accumulate all writes to a contiguous array and interleave to spout at the end of the cycle? 
> 
> On 03.02.2017 16:03, Victor Lazzarini wrote:
>> No, because the difference is that  ga1 += a1  is cheap and then we only have
>> to interleave the data to spout once. When the instrument is writing to out, we have
>> 1000 times the loop that interleaves the data. So it’s more expensive.
>> 
>> ========================
>> Prof. 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 Feb 2017, at 14:59, Eduardo Moguillansky 
>>>  wrote:
>>> 
>>> I was running in realtime in order to compare the performance with supercollider. The percents given are not significant in themselves, but I would have expected to see a similar performance of "out" vs accumulating in an a-variable and doing one "out" at the end of each k-cycle. 
>>> 
>>> On 03.02.2017 15:39, jpff wrote:
>>> 
>>>> sorry do t undestand.   i would epet about 100% cpu as it is doing nothing else.  What are you measurg?  I am seeing it un in 9s at 99.% 
>>>> 
>>>> 
>>>> On Fri, 3 Feb 2017, Eduardo Moguillansky wrote: 
>>>> 
>>>> 
>>>>> Hi 
>>>>> 
>>>>> While porting a project from supercollider to csound I noticed that csound 
>>>>> uses a considerable amount of cpu just for signal output. When run in real 
>>>>> time, the csd below uses ~45% cpu for just copying to spout. If instead of 
>>>>> using "out" in instr 1 the signals are accumulated in a global variable and 
>>>>> output once in another instrument, cpu drops down to ~30%. Is this something 
>>>>> known which should be avoided? The same processing in supercollider 
>>>>> (outputting 1000 silent signals), results in ~15% CPU 
>>>>> 
>>>>>  
>>>>> sr = 48000 
>>>>> nchnls = 2 
>>>>> ksmps = 128 
>>>>> 0dbfs = 1 
>>>>> 
>>>>> instr 1 
>>>>>   a1 init 0 
>>>>>   out a1 
>>>>> endin 
>>>>> 
>>>>> instr 2 
>>>>>   iN = 1000 
>>>>>   idx = 0 
>>>>>   while idx < iN do 
>>>>>     event_i "i", 1, 0, 100, 1/iN 
>>>>>     idx += 1 
>>>>>   od 
>>>>>   turnoff 
>>>>> endin 
>>>>>  
>>>>> 
>>>>>  
>>>>> i2 0 100 
>>>>>  
>>>>> 
>>>>> 
>>>>> ( 
>>>>> 1000.do { 
>>>>>   { Out.ar(0, Silent.ar(1)) }.play 
>>>>> } 
>>>>> ) 
>>>>> 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
>> 
> 
> 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

I think  we could try the latter, which would be simpler, then see if that makes a difference. 
========================
Prof. 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 Feb 2017, at 15:14, jpff  wrote:
> 
> The only fix i can think of is te per instance spout whic we have discussed before; ot make spout not intertwinned and fix nly once
> 
> On Fri, 3 Feb 2017, Victor Lazzarini wrote:
> 
>> No, because the difference is that  ga1 += a1  is cheap and then we only have
>> to interleave the data to spout once. When the instrument is writing to out, we have
>> 1000 times the loop that interleaves the data. So it’s more expensive.
>> 
>> ========================
>> Prof. 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 Feb 2017, at 14:59, Eduardo Moguillansky  wrote:
>>> 
>>> I was running in realtime in order to compare the performance with supercollider. The percents given are not significant in themselves, but I would have expected to see a similar performance of "out" vs accumulating in an a-variable and doing one "out" at the end of each k-cycle.
>>> 
>>> On 03.02.2017 15:39, jpff wrote:
>>>> sorry do t undestand.   i would epet about 100% cpu as it is doing nothing else.  What are you measurg?  I am seeing it un in 9s at 99.%
>>>> 
>>>> 
>>>> On Fri, 3 Feb 2017, Eduardo Moguillansky wrote:
>>>> 
>>>>> 
>>>>> Hi
>>>>> 
>>>>> While porting a project from supercollider to csound I noticed that csound
>>>>> uses a considerable amount of cpu just for signal output. When run in real
>>>>> time, the csd below uses ~45% cpu for just copying to spout. If instead of
>>>>> using "out" in instr 1 the signals are accumulated in a global variable and
>>>>> output once in another instrument, cpu drops down to ~30%. Is this something
>>>>> known which should be avoided? The same processing in supercollider
>>>>> (outputting 1000 silent signals), results in ~15% CPU
>>>>> 
>>>>> 
>>>>> sr = 48000
>>>>> nchnls = 2
>>>>> ksmps = 128
>>>>> 0dbfs = 1
>>>>> 
>>>>> instr 1
>>>>>  a1 init 0
>>>>>  out a1
>>>>> endin
>>>>> 
>>>>> instr 2
>>>>>  iN = 1000
>>>>>  idx = 0
>>>>>  while idx < iN do
>>>>>    event_i "i", 1, 0, 100, 1/iN
>>>>>    idx += 1
>>>>>  od
>>>>>  turnoff
>>>>> endin
>>>>> 
>>>>> 
>>>>> 
>>>>> i2 0 100
>>>>> 
>>>>> 
>>>>> 
>>>>> (
>>>>> 1000.do {
>>>>>  { Out.ar(0, Silent.ar(1)) }.play
>>>>> }
>>>>> )
>>>>> 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
>> 
> 
> 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

Firstly, thanks Eduardo for identifying this!  Secondly, and perhaps
obviously, if we modify output to use a non-interleaved layout, we
will, at least for Csound 6, require spin/spout to be interleaved so
that API users using spin/spout will still function.  (We could use an
internal spin/spout that get used so that the external spin/spout can
remain compatible.)

On Fri, Feb 3, 2017 at 10:16 AM, Victor Lazzarini
 wrote:
> I think  we could try the latter, which would be simpler, then see if that makes a difference.
> ========================
> Prof. 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 Feb 2017, at 15:14, jpff  wrote:
>>
>> The only fix i can think of is te per instance spout whic we have discussed before; ot make spout not intertwinned and fix nly once
>>
>> On Fri, 3 Feb 2017, Victor Lazzarini wrote:
>>
>>> No, because the difference is that  ga1 += a1  is cheap and then we only have
>>> to interleave the data to spout once. When the instrument is writing to out, we have
>>> 1000 times the loop that interleaves the data. So it’s more expensive.
>>>
>>> ========================
>>> Prof. 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 Feb 2017, at 14:59, Eduardo Moguillansky  wrote:
>>>>
>>>> I was running in realtime in order to compare the performance with supercollider. The percents given are not significant in themselves, but I would have expected to see a similar performance of "out" vs accumulating in an a-variable and doing one "out" at the end of each k-cycle.
>>>>
>>>> On 03.02.2017 15:39, jpff wrote:
>>>>> sorry do t undestand.   i would epet about 100% cpu as it is doing nothing else.  What are you measurg?  I am seeing it un in 9s at 99.%
>>>>>
>>>>>
>>>>> On Fri, 3 Feb 2017, Eduardo Moguillansky wrote:
>>>>>
>>>>>>
>>>>>> Hi
>>>>>>
>>>>>> While porting a project from supercollider to csound I noticed that csound
>>>>>> uses a considerable amount of cpu just for signal output. When run in real
>>>>>> time, the csd below uses ~45% cpu for just copying to spout. If instead of
>>>>>> using "out" in instr 1 the signals are accumulated in a global variable and
>>>>>> output once in another instrument, cpu drops down to ~30%. Is this something
>>>>>> known which should be avoided? The same processing in supercollider
>>>>>> (outputting 1000 silent signals), results in ~15% CPU
>>>>>>
>>>>>> 
>>>>>> sr = 48000
>>>>>> nchnls = 2
>>>>>> ksmps = 128
>>>>>> 0dbfs = 1
>>>>>>
>>>>>> instr 1
>>>>>>  a1 init 0
>>>>>>  out a1
>>>>>> endin
>>>>>>
>>>>>> instr 2
>>>>>>  iN = 1000
>>>>>>  idx = 0
>>>>>>  while idx < iN do
>>>>>>    event_i "i", 1, 0, 100, 1/iN
>>>>>>    idx += 1
>>>>>>  od
>>>>>>  turnoff
>>>>>> endin
>>>>>> 
>>>>>>
>>>>>> 
>>>>>> i2 0 100
>>>>>> 
>>>>>>
>>>>>>
>>>>>> (
>>>>>> 1000.do {
>>>>>>  { Out.ar(0, Silent.ar(1)) }.play
>>>>>> }
>>>>>> )
>>>>>> 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
>>>
>>
>> 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