I had some time to kill on the plane today so I decided to see what  
was the matter with the parcs build/code. As I expected, it was not  
only me being tired last night, but there was something wrong.

There was a chunk of code missing from main.c to activate the threads,  
so no wonder that I only got a single thread out of csound. Anyway,  
this fixed (and I will commit the fix as soon as I can), I was able to  
do some tests. Btw, doing this I had the opportunity of reading the  
parcs code. It's very impressive.

This is what I get:

trapped -- kr = 441 sr = 44100
-j 1  => 2.419s
-j 2 => 9.250s

trapped --kr= 100
-j 1 => 2.220s
-j 2 => 4.704s

trapped -- kr =10
-j 1 => 2.221s
-j 2 => 2.088s  !!!!!!

===

xanadu -- kr=441
-j 1 = 2.226
-j 2 = 8.750

xanadu -- kr=100
-j 1 = 2.118
-j 2 = 2.989  (see the difference!)

xanadu -- kr=10
-j 1 = 2.282
-j 2 = 1.151 !!!!!! (almost 50% speedup)

=============

So increasing the granularity pays off big time. I am wondering  
whether there is not a way of supporting this a bit more in Csound.  
Working with large ksmps is not really practical in most cases. Could  
we envisage a buffer-size granularity?

Dr Victor Lazzarini
Senior Lecturer
Dept. of Music
NUI Maynooth Ireland
tel.: +353 1 708 3545
Victor dot Lazzarini AT nuim dot ie




------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2d-oct
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Thanks for fixing this! I had inferred that something like this must
be the case based on my own timings, but I have been too busy to do
anything about it. I will of course rebuild and repackage.

Regards,
Mike

On Thu, Oct 13, 2011 at 11:25 AM, Victor Lazzarini
 wrote:
> I had some time to kill on the plane today so I decided to see what
> was the matter with the parcs build/code. As I expected, it was not
> only me being tired last night, but there was something wrong.
>
> There was a chunk of code missing from main.c to activate the threads,
> so no wonder that I only got a single thread out of csound. Anyway,
> this fixed (and I will commit the fix as soon as I can), I was able to
> do some tests. Btw, doing this I had the opportunity of reading the
> parcs code. It's very impressive.
>
> This is what I get:
>
> trapped -- kr = 441 sr = 44100
> -j 1  => 2.419s
> -j 2 => 9.250s
>
> trapped --kr= 100
> -j 1 => 2.220s
> -j 2 => 4.704s
>
> trapped -- kr =10
> -j 1 => 2.221s
> -j 2 => 2.088s  !!!!!!
>
> ===
>
> xanadu -- kr=441
> -j 1 = 2.226
> -j 2 = 8.750
>
> xanadu -- kr=100
> -j 1 = 2.118
> -j 2 = 2.989  (see the difference!)
>
> xanadu -- kr=10
> -j 1 = 2.282
> -j 2 = 1.151 !!!!!! (almost 50% speedup)
>
> =============
>
> So increasing the granularity pays off big time. I am wondering
> whether there is not a way of supporting this a bit more in Csound.
> Working with large ksmps is not really practical in most cases. Could
> we envisage a buffer-size granularity?
>
> Dr Victor Lazzarini
> Senior Lecturer
> Dept. of Music
> NUI Maynooth Ireland
> tel.: +353 1 708 3545
> Victor dot Lazzarini AT nuim dot ie
>
>
>
>
> ------------------------------------------------------------------------------
> All the data continuously generated in your IT infrastructure contains a
> definitive record of customers, application performance, security
> threats, fraudulent activity and more. Splunk takes this data and makes
> sense of it. Business sense. IT sense. Common sense.
> http://p.sf.net/sfu/splunk-d2d-oct
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
>



-- 
Michael Gogins
Irreducible Productions
http://www.michael-gogins.com
Michael dot Gogins at gmail dot com

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2d-oct
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

I pushed the updates now, so it should be all working.

Victor
On 13 Oct 2011, at 17:02, Michael Gogins wrote:

> Thanks for fixing this! I had inferred that something like this must
> be the case based on my own timings, but I have been too busy to do
> anything about it. I will of course rebuild and repackage.
>
> Regards,
> Mike
>
> On Thu, Oct 13, 2011 at 11:25 AM, Victor Lazzarini
>  wrote:
>> I had some time to kill on the plane today so I decided to see what
>> was the matter with the parcs build/code. As I expected, it was not
>> only me being tired last night, but there was something wrong.
>>
>> There was a chunk of code missing from main.c to activate the  
>> threads,
>> so no wonder that I only got a single thread out of csound. Anyway,
>> this fixed (and I will commit the fix as soon as I can), I was able  
>> to
>> do some tests. Btw, doing this I had the opportunity of reading the
>> parcs code. It's very impressive.
>>
>> This is what I get:
>>
>> trapped -- kr = 441 sr = 44100
>> -j 1  => 2.419s
>> -j 2 => 9.250s
>>
>> trapped --kr= 100
>> -j 1 => 2.220s
>> -j 2 => 4.704s
>>
>> trapped -- kr =10
>> -j 1 => 2.221s
>> -j 2 => 2.088s  !!!!!!
>>
>> ===
>>
>> xanadu -- kr=441
>> -j 1 = 2.226
>> -j 2 = 8.750
>>
>> xanadu -- kr=100
>> -j 1 = 2.118
>> -j 2 = 2.989  (see the difference!)
>>
>> xanadu -- kr=10
>> -j 1 = 2.282
>> -j 2 = 1.151 !!!!!! (almost 50% speedup)
>>
>> =============
>>
>> So increasing the granularity pays off big time. I am wondering
>> whether there is not a way of supporting this a bit more in Csound.
>> Working with large ksmps is not really practical in most cases. Could
>> we envisage a buffer-size granularity?
>>
>> Dr Victor Lazzarini
>> Senior Lecturer
>> Dept. of Music
>> NUI Maynooth Ireland
>> tel.: +353 1 708 3545
>> Victor dot Lazzarini AT nuim dot ie
>>
>>
>>
>>
>> ------------------------------------------------------------------------------
>> All the data continuously generated in your IT infrastructure  
>> contains a
>> definitive record of customers, application performance, security
>> threats, fraudulent activity and more. Splunk takes this data and  
>> makes
>> sense of it. Business sense. IT sense. Common sense.
>> http://p.sf.net/sfu/splunk-d2d-oct
>> _______________________________________________
>> Csound-devel mailing list
>> Csound-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
>
>
>
> -- 
> Michael Gogins
> Irreducible Productions
> http://www.michael-gogins.com
> Michael dot Gogins at gmail dot com
>
> ------------------------------------------------------------------------------
> All the data continuously generated in your IT infrastructure  
> contains a
> definitive record of customers, application performance, security
> threats, fraudulent activity and more. Splunk takes this data and  
> makes
> sense of it. Business sense. IT sense. Common sense.
> http://p.sf.net/sfu/splunk-d2d-oct
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel

Dr Victor Lazzarini
Senior Lecturer
Dept. of Music
NUI Maynooth Ireland
tel.: +353 1 708 3545
Victor dot Lazzarini AT nuim dot ie




------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a
definitive record of customers, application performance, security
threats, fraudulent activity and more. Splunk takes this data and makes
sense of it. Business sense. IT sense. Common sense.
http://p.sf.net/sfu/splunk-d2d-oct
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net