What is the comparison between __sync_lock_test_and_set and
pthread_spin_lock/pthread_spin_unlock ?

It seems that Csound used the former until the weekend; just wondered
why?

==John ffitch

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Read my prior emails. I have changed it back to
__sync_lock_test_and_set because pthread spinlocks didn't seem to
actually work. Don't know why, it should be the way to go.

Regards,
Mike

On Tue, Aug 24, 2010 at 8:16 AM, John ff  wrote:
> What is the comparison between __sync_lock_test_and_set and
> pthread_spin_lock/pthread_spin_unlock ?
>
> It seems that Csound used the former until the weekend; just wondered
> why?
>
> ==John ffitch
>
> ------------------------------------------------------------------------------
> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
> Be part of this innovative community and reach millions of netbook users
> worldwide. Take advantage of special opportunities to increase revenue and
> speed time-to-market. Join now, and jumpstart your future.
> http://p.sf.net/sfu/intel-atom-d2d
> _______________________________________________
> 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

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

JFTR, the __sync_lock_test_and_set stuff never actually worked for me
(there appears to be a deadlock somewhere).

On 24/08/10 08:26, Michael Gogins wrote:
> Read my prior emails. I have changed it back to
> __sync_lock_test_and_set because pthread spinlocks didn't seem to
> actually work. Don't know why, it should be the way to go.
> 
> Regards,
> Mike
> 
> On Tue, Aug 24, 2010 at 8:16 AM, John ff  wrote:
>> What is the comparison between __sync_lock_test_and_set and
>> pthread_spin_lock/pthread_spin_unlock ?
>>
>> It seems that Csound used the former until the weekend; just wondered
>> why?
>>
>> ==John ffitch
>>
>> ------------------------------------------------------------------------------
>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>> Be part of this innovative community and reach millions of netbook users
>> worldwide. Take advantage of special opportunities to increase revenue and
>> speed time-to-market. Join now, and jumpstart your future.
>> http://p.sf.net/sfu/intel-atom-d2d
>> _______________________________________________
>> Csound-devel mailing list
>> Csound-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
> 
> 
> 


-- 
Saludos,
Felipe Sateler

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Can you give more details? It is working for me in both ParCS and head
for Windows and for ParCS for Linux.

Regards,
Mike

On Tue, Aug 24, 2010 at 3:25 PM, Felipe Sateler  wrote:
> JFTR, the __sync_lock_test_and_set stuff never actually worked for me
> (there appears to be a deadlock somewhere).
>
> On 24/08/10 08:26, Michael Gogins wrote:
>> Read my prior emails. I have changed it back to
>> __sync_lock_test_and_set because pthread spinlocks didn't seem to
>> actually work. Don't know why, it should be the way to go.
>>
>> Regards,
>> Mike
>>
>> On Tue, Aug 24, 2010 at 8:16 AM, John ff  wrote:
>>> What is the comparison between __sync_lock_test_and_set and
>>> pthread_spin_lock/pthread_spin_unlock ?
>>>
>>> It seems that Csound used the former until the weekend; just wondered
>>> why?
>>>
>>> ==John ffitch
>>>
>>> ------------------------------------------------------------------------------
>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>> Be part of this innovative community and reach millions of netbook users
>>> worldwide. Take advantage of special opportunities to increase revenue and
>>> speed time-to-market. Join now, and jumpstart your future.
>>> http://p.sf.net/sfu/intel-atom-d2d
>>> _______________________________________________
>>> Csound-devel mailing list
>>> Csound-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>
>>
>>
>>
>
>
> --
> Saludos,
> Felipe Sateler
>
> ------------------------------------------------------------------------------
> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
> Be part of this innovative community and reach millions of netbook users
> worldwide. Take advantage of special opportunities to increase revenue and
> speed time-to-market. Join now, and jumpstart your future.
> http://p.sf.net/sfu/intel-atom-d2d
> _______________________________________________
> 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

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Well, it's been some time since I actually tried it. I just checked.

Building head with no arguments, and specifying --num-threads=2 I can
render xanadu and trapped, but no speedup (only one cpu core is used).

With the ParCS branch, building with buildNewParser=1 and specifying -j
2 also gives me no speedup, and only one cpu core used.

Or are special parameters needed on scons?

On 24/08/10 15:29, Michael Gogins wrote:
> Can you give more details? It is working for me in both ParCS and head
> for Windows and for ParCS for Linux.
> 
> Regards,
> Mike
> 
> On Tue, Aug 24, 2010 at 3:25 PM, Felipe Sateler  wrote:
>> JFTR, the __sync_lock_test_and_set stuff never actually worked for me
>> (there appears to be a deadlock somewhere).
>>
>> On 24/08/10 08:26, Michael Gogins wrote:
>>> Read my prior emails. I have changed it back to
>>> __sync_lock_test_and_set because pthread spinlocks didn't seem to
>>> actually work. Don't know why, it should be the way to go.
>>>
>>> Regards,
>>> Mike
>>>
>>> On Tue, Aug 24, 2010 at 8:16 AM, John ff  wrote:
>>>> What is the comparison between __sync_lock_test_and_set and
>>>> pthread_spin_lock/pthread_spin_unlock ?
>>>>
>>>> It seems that Csound used the former until the weekend; just wondered
>>>> why?
>>>>
>>>> ==John ffitch
>>>>
>>>> ------------------------------------------------------------------------------
>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>> Be part of this innovative community and reach millions of netbook users
>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>> speed time-to-market. Join now, and jumpstart your future.
>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>> _______________________________________________
>>>> Csound-devel mailing list
>>>> Csound-devel@lists.sourceforge.net
>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>
>>>
>>>
>>>
>>
>>
>> --
>> Saludos,
>> Felipe Sateler
>>
>> ------------------------------------------------------------------------------
>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>> Be part of this innovative community and reach millions of netbook users
>> worldwide. Take advantage of special opportunities to increase revenue and
>> speed time-to-market. Join now, and jumpstart your future.
>> http://p.sf.net/sfu/intel-atom-d2d
>> _______________________________________________
>> Csound-devel mailing list
>> Csound-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
> 
> 
> 


-- 
Saludos,
Felipe Sateler

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Yes. useOpenMP=1 in the current head branch, buildNewParser=1 in the
ParCS branch.

Regards,
Mike

On Tue, Aug 24, 2010 at 3:59 PM, Felipe Sateler  wrote:
> Well, it's been some time since I actually tried it. I just checked.
>
> Building head with no arguments, and specifying --num-threads=2 I can
> render xanadu and trapped, but no speedup (only one cpu core is used).
>
> With the ParCS branch, building with buildNewParser=1 and specifying -j
> 2 also gives me no speedup, and only one cpu core used.
>
> Or are special parameters needed on scons?
>
> On 24/08/10 15:29, Michael Gogins wrote:
>> Can you give more details? It is working for me in both ParCS and head
>> for Windows and for ParCS for Linux.
>>
>> Regards,
>> Mike
>>
>> On Tue, Aug 24, 2010 at 3:25 PM, Felipe Sateler  wrote:
>>> JFTR, the __sync_lock_test_and_set stuff never actually worked for me
>>> (there appears to be a deadlock somewhere).
>>>
>>> On 24/08/10 08:26, Michael Gogins wrote:
>>>> Read my prior emails. I have changed it back to
>>>> __sync_lock_test_and_set because pthread spinlocks didn't seem to
>>>> actually work. Don't know why, it should be the way to go.
>>>>
>>>> Regards,
>>>> Mike
>>>>
>>>> On Tue, Aug 24, 2010 at 8:16 AM, John ff  wrote:
>>>>> What is the comparison between __sync_lock_test_and_set and
>>>>> pthread_spin_lock/pthread_spin_unlock ?
>>>>>
>>>>> It seems that Csound used the former until the weekend; just wondered
>>>>> why?
>>>>>
>>>>> ==John ffitch
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>>> Be part of this innovative community and reach millions of netbook users
>>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>> _______________________________________________
>>>>> Csound-devel mailing list
>>>>> Csound-devel@lists.sourceforge.net
>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>> --
>>> Saludos,
>>> Felipe Sateler
>>>
>>> ------------------------------------------------------------------------------
>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>> Be part of this innovative community and reach millions of netbook users
>>> worldwide. Take advantage of special opportunities to increase revenue and
>>> speed time-to-market. Join now, and jumpstart your future.
>>> http://p.sf.net/sfu/intel-atom-d2d
>>> _______________________________________________
>>> Csound-devel mailing list
>>> Csound-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>
>>
>>
>>
>
>
> --
> Saludos,
> Felipe Sateler
>
> ------------------------------------------------------------------------------
> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
> Be part of this innovative community and reach millions of netbook users
> worldwide. Take advantage of special opportunities to increase revenue and
> speed time-to-market. Join now, and jumpstart your future.
> http://p.sf.net/sfu/intel-atom-d2d
> _______________________________________________
> 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

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

OK, so ParCS doesn't use more than one core for me, and the current head
does... but is not effective:

xanadu.csd
1 thread:  real: 51.050s, CPU: 50.660s
2 threads: real: 64.312s, CPU: 117.460s

trapped.csd
1 thread:  real: 13.686s, CPU: 13.350s
2 threads: real: 17.876s, CPU: 30.860s

ParCS does not seem to be able to use more than one core.


On 24/08/10 16:05, Michael Gogins wrote:
> Yes. useOpenMP=1 in the current head branch, buildNewParser=1 in the
> ParCS branch.
> 
> Regards,
> Mike
> 
> On Tue, Aug 24, 2010 at 3:59 PM, Felipe Sateler  wrote:
>> Well, it's been some time since I actually tried it. I just checked.
>>
>> Building head with no arguments, and specifying --num-threads=2 I can
>> render xanadu and trapped, but no speedup (only one cpu core is used).
>>
>> With the ParCS branch, building with buildNewParser=1 and specifying -j
>> 2 also gives me no speedup, and only one cpu core used.
>>
>> Or are special parameters needed on scons?
>>
>> On 24/08/10 15:29, Michael Gogins wrote:
>>> Can you give more details? It is working for me in both ParCS and head
>>> for Windows and for ParCS for Linux.
>>>
>>> Regards,
>>> Mike
>>>
>>> On Tue, Aug 24, 2010 at 3:25 PM, Felipe Sateler  wrote:
>>>> JFTR, the __sync_lock_test_and_set stuff never actually worked for me
>>>> (there appears to be a deadlock somewhere).
>>>>
>>>> On 24/08/10 08:26, Michael Gogins wrote:
>>>>> Read my prior emails. I have changed it back to
>>>>> __sync_lock_test_and_set because pthread spinlocks didn't seem to
>>>>> actually work. Don't know why, it should be the way to go.
>>>>>
>>>>> Regards,
>>>>> Mike
>>>>>
>>>>> On Tue, Aug 24, 2010 at 8:16 AM, John ff  wrote:
>>>>>> What is the comparison between __sync_lock_test_and_set and
>>>>>> pthread_spin_lock/pthread_spin_unlock ?
>>>>>>
>>>>>> It seems that Csound used the former until the weekend; just wondered
>>>>>> why?
>>>>>>
>>>>>> ==John ffitch
>>>>>>
>>>>>> ------------------------------------------------------------------------------
>>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>>>> Be part of this innovative community and reach millions of netbook users
>>>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>>> _______________________________________________
>>>>>> Csound-devel mailing list
>>>>>> Csound-devel@lists.sourceforge.net
>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Saludos,
>>>> Felipe Sateler
>>>>
>>>> ------------------------------------------------------------------------------
>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>> Be part of this innovative community and reach millions of netbook users
>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>> speed time-to-market. Join now, and jumpstart your future.
>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>> _______________________________________________
>>>> Csound-devel mailing list
>>>> Csound-devel@lists.sourceforge.net
>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>
>>>
>>>
>>>
>>
>>
>> --
>> Saludos,
>> Felipe Sateler
>>
>> ------------------------------------------------------------------------------
>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>> Be part of this innovative community and reach millions of netbook users
>> worldwide. Take advantage of special opportunities to increase revenue and
>> speed time-to-market. Join now, and jumpstart your future.
>> http://p.sf.net/sfu/intel-atom-d2d
>> _______________________________________________
>> Csound-devel mailing list
>> Csound-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
> 
> 
> 


-- 
Saludos,
Felipe Sateler

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

For ParCS the parameter for threads is -j, e.g. -j4.

There is too much threading overhead for you to get a speedup in head.
Try increasing ksmps to 500 or 1000 or so and see what happens. The
threading overhead (acquiring locks, putting threads to sleep and
waking them up, waiting on barriers, etc.). happens every kperiod, so
the more ksmps the less overhead per frame.

It seems to me that ParCS involves less threading overhead than head,
that is why it runs faster for me with smaller ksmps.

Regards,
Mike

On Tue, Aug 24, 2010 at 4:22 PM, Felipe Sateler  wrote:
> OK, so ParCS doesn't use more than one core for me, and the current head
> does... but is not effective:
>
> xanadu.csd
> 1 thread:  real: 51.050s, CPU: 50.660s
> 2 threads: real: 64.312s, CPU: 117.460s
>
> trapped.csd
> 1 thread:  real: 13.686s, CPU: 13.350s
> 2 threads: real: 17.876s, CPU: 30.860s
>
> ParCS does not seem to be able to use more than one core.
>
>
> On 24/08/10 16:05, Michael Gogins wrote:
>> Yes. useOpenMP=1 in the current head branch, buildNewParser=1 in the
>> ParCS branch.
>>
>> Regards,
>> Mike
>>
>> On Tue, Aug 24, 2010 at 3:59 PM, Felipe Sateler  wrote:
>>> Well, it's been some time since I actually tried it. I just checked.
>>>
>>> Building head with no arguments, and specifying --num-threads=2 I can
>>> render xanadu and trapped, but no speedup (only one cpu core is used).
>>>
>>> With the ParCS branch, building with buildNewParser=1 and specifying -j
>>> 2 also gives me no speedup, and only one cpu core used.
>>>
>>> Or are special parameters needed on scons?
>>>
>>> On 24/08/10 15:29, Michael Gogins wrote:
>>>> Can you give more details? It is working for me in both ParCS and head
>>>> for Windows and for ParCS for Linux.
>>>>
>>>> Regards,
>>>> Mike
>>>>
>>>> On Tue, Aug 24, 2010 at 3:25 PM, Felipe Sateler  wrote:
>>>>> JFTR, the __sync_lock_test_and_set stuff never actually worked for me
>>>>> (there appears to be a deadlock somewhere).
>>>>>
>>>>> On 24/08/10 08:26, Michael Gogins wrote:
>>>>>> Read my prior emails. I have changed it back to
>>>>>> __sync_lock_test_and_set because pthread spinlocks didn't seem to
>>>>>> actually work. Don't know why, it should be the way to go.
>>>>>>
>>>>>> Regards,
>>>>>> Mike
>>>>>>
>>>>>> On Tue, Aug 24, 2010 at 8:16 AM, John ff  wrote:
>>>>>>> What is the comparison between __sync_lock_test_and_set and
>>>>>>> pthread_spin_lock/pthread_spin_unlock ?
>>>>>>>
>>>>>>> It seems that Csound used the former until the weekend; just wondered
>>>>>>> why?
>>>>>>>
>>>>>>> ==John ffitch
>>>>>>>
>>>>>>> ------------------------------------------------------------------------------
>>>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>>>>> Be part of this innovative community and reach millions of netbook users
>>>>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>>>> _______________________________________________
>>>>>>> Csound-devel mailing list
>>>>>>> Csound-devel@lists.sourceforge.net
>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Saludos,
>>>>> Felipe Sateler
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>>> Be part of this innovative community and reach millions of netbook users
>>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>> _______________________________________________
>>>>> Csound-devel mailing list
>>>>> Csound-devel@lists.sourceforge.net
>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>> --
>>> Saludos,
>>> Felipe Sateler
>>>
>>> ------------------------------------------------------------------------------
>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>> Be part of this innovative community and reach millions of netbook users
>>> worldwide. Take advantage of special opportunities to increase revenue and
>>> speed time-to-market. Join now, and jumpstart your future.
>>> http://p.sf.net/sfu/intel-atom-d2d
>>> _______________________________________________
>>> Csound-devel mailing list
>>> Csound-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>
>>
>>
>>
>
>
> --
> Saludos,
> Felipe Sateler
>
> ------------------------------------------------------------------------------
> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
> Be part of this innovative community and reach millions of netbook users
> worldwide. Take advantage of special opportunities to increase revenue and
> speed time-to-market. Join now, and jumpstart your future.
> http://p.sf.net/sfu/intel-atom-d2d
> _______________________________________________
> 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

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Indeed, increasing ksmps makes it perform better:

500
1 thread: real: 6.130s, CPU: 6.060s
2 thread: real: 5.284s, CPU: 10.160s

1000
1 thread: real: 6.145s, CPU: 6.110s
2 thread: real: 5.869s, CPU: 10.220s

I just noticed that parcs cannot render trapped when called with -j2:

Expected a root to perform. Found none (1 remaining)
Early return from csoundPerform().


On 24/08/10 16:31, Michael Gogins wrote:
> For ParCS the parameter for threads is -j, e.g. -j4.
> 
> There is too much threading overhead for you to get a speedup in head.
> Try increasing ksmps to 500 or 1000 or so and see what happens. The
> threading overhead (acquiring locks, putting threads to sleep and
> waking them up, waiting on barriers, etc.). happens every kperiod, so
> the more ksmps the less overhead per frame.
> 
> It seems to me that ParCS involves less threading overhead than head,
> that is why it runs faster for me with smaller ksmps.
> 
> Regards,
> Mike
> 
> On Tue, Aug 24, 2010 at 4:22 PM, Felipe Sateler  wrote:
>> OK, so ParCS doesn't use more than one core for me, and the current head
>> does... but is not effective:
>>
>> xanadu.csd
>> 1 thread:  real: 51.050s, CPU: 50.660s
>> 2 threads: real: 64.312s, CPU: 117.460s
>>
>> trapped.csd
>> 1 thread:  real: 13.686s, CPU: 13.350s
>> 2 threads: real: 17.876s, CPU: 30.860s
>>
>> ParCS does not seem to be able to use more than one core.
>>
>>
>> On 24/08/10 16:05, Michael Gogins wrote:
>>> Yes. useOpenMP=1 in the current head branch, buildNewParser=1 in the
>>> ParCS branch.
>>>
>>> Regards,
>>> Mike
>>>
>>> On Tue, Aug 24, 2010 at 3:59 PM, Felipe Sateler  wrote:
>>>> Well, it's been some time since I actually tried it. I just checked.
>>>>
>>>> Building head with no arguments, and specifying --num-threads=2 I can
>>>> render xanadu and trapped, but no speedup (only one cpu core is used).
>>>>
>>>> With the ParCS branch, building with buildNewParser=1 and specifying -j
>>>> 2 also gives me no speedup, and only one cpu core used.
>>>>
>>>> Or are special parameters needed on scons?
>>>>
>>>> On 24/08/10 15:29, Michael Gogins wrote:
>>>>> Can you give more details? It is working for me in both ParCS and head
>>>>> for Windows and for ParCS for Linux.
>>>>>
>>>>> Regards,
>>>>> Mike
>>>>>
>>>>> On Tue, Aug 24, 2010 at 3:25 PM, Felipe Sateler  wrote:
>>>>>> JFTR, the __sync_lock_test_and_set stuff never actually worked for me
>>>>>> (there appears to be a deadlock somewhere).
>>>>>>
>>>>>> On 24/08/10 08:26, Michael Gogins wrote:
>>>>>>> Read my prior emails. I have changed it back to
>>>>>>> __sync_lock_test_and_set because pthread spinlocks didn't seem to
>>>>>>> actually work. Don't know why, it should be the way to go.
>>>>>>>
>>>>>>> Regards,
>>>>>>> Mike
>>>>>>>
>>>>>>> On Tue, Aug 24, 2010 at 8:16 AM, John ff  wrote:
>>>>>>>> What is the comparison between __sync_lock_test_and_set and
>>>>>>>> pthread_spin_lock/pthread_spin_unlock ?
>>>>>>>>
>>>>>>>> It seems that Csound used the former until the weekend; just wondered
>>>>>>>> why?
>>>>>>>>
>>>>>>>> ==John ffitch
>>>>>>>>
>>>>>>>> ------------------------------------------------------------------------------
>>>>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>>>>>> Be part of this innovative community and reach millions of netbook users
>>>>>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>>>>> _______________________________________________
>>>>>>>> Csound-devel mailing list
>>>>>>>> Csound-devel@lists.sourceforge.net
>>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Saludos,
>>>>>> Felipe Sateler
>>>>>>
>>>>>> ------------------------------------------------------------------------------
>>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>>>> Be part of this innovative community and reach millions of netbook users
>>>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>>> _______________________________________________
>>>>>> Csound-devel mailing list
>>>>>> Csound-devel@lists.sourceforge.net
>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Saludos,
>>>> Felipe Sateler
>>>>
>>>> ------------------------------------------------------------------------------
>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>> Be part of this innovative community and reach millions of netbook users
>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>> speed time-to-market. Join now, and jumpstart your future.
>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>> _______________________________________________
>>>> Csound-devel mailing list
>>>> Csound-devel@lists.sourceforge.net
>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>
>>>
>>>
>>>
>>
>>
>> --
>> Saludos,
>> Felipe Sateler
>>
>> ------------------------------------------------------------------------------
>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>> Be part of this innovative community and reach millions of netbook users
>> worldwide. Take advantage of special opportunities to increase revenue and
>> speed time-to-market. Join now, and jumpstart your future.
>> http://p.sf.net/sfu/intel-atom-d2d
>> _______________________________________________
>> Csound-devel mailing list
>> Csound-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
> 
> 
> 


-- 
Saludos,
Felipe Sateler

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

That's right, I forgot, there is a bug in ParCS that segfaults on
trapped. Other csds seem to work OK. I will be debugging this very
soon, I will at least take a look at it tonight.

Regards,
Mike

On Tue, Aug 24, 2010 at 4:38 PM, Felipe Sateler  wrote:
> Indeed, increasing ksmps makes it perform better:
>
> 500
> 1 thread: real: 6.130s, CPU: 6.060s
> 2 thread: real: 5.284s, CPU: 10.160s
>
> 1000
> 1 thread: real: 6.145s, CPU: 6.110s
> 2 thread: real: 5.869s, CPU: 10.220s
>
> I just noticed that parcs cannot render trapped when called with -j2:
>
> Expected a root to perform. Found none (1 remaining)
> Early return from csoundPerform().
>
>
> On 24/08/10 16:31, Michael Gogins wrote:
>> For ParCS the parameter for threads is -j, e.g. -j4.
>>
>> There is too much threading overhead for you to get a speedup in head.
>> Try increasing ksmps to 500 or 1000 or so and see what happens. The
>> threading overhead (acquiring locks, putting threads to sleep and
>> waking them up, waiting on barriers, etc.). happens every kperiod, so
>> the more ksmps the less overhead per frame.
>>
>> It seems to me that ParCS involves less threading overhead than head,
>> that is why it runs faster for me with smaller ksmps.
>>
>> Regards,
>> Mike
>>
>> On Tue, Aug 24, 2010 at 4:22 PM, Felipe Sateler  wrote:
>>> OK, so ParCS doesn't use more than one core for me, and the current head
>>> does... but is not effective:
>>>
>>> xanadu.csd
>>> 1 thread:  real: 51.050s, CPU: 50.660s
>>> 2 threads: real: 64.312s, CPU: 117.460s
>>>
>>> trapped.csd
>>> 1 thread:  real: 13.686s, CPU: 13.350s
>>> 2 threads: real: 17.876s, CPU: 30.860s
>>>
>>> ParCS does not seem to be able to use more than one core.
>>>
>>>
>>> On 24/08/10 16:05, Michael Gogins wrote:
>>>> Yes. useOpenMP=1 in the current head branch, buildNewParser=1 in the
>>>> ParCS branch.
>>>>
>>>> Regards,
>>>> Mike
>>>>
>>>> On Tue, Aug 24, 2010 at 3:59 PM, Felipe Sateler  wrote:
>>>>> Well, it's been some time since I actually tried it. I just checked.
>>>>>
>>>>> Building head with no arguments, and specifying --num-threads=2 I can
>>>>> render xanadu and trapped, but no speedup (only one cpu core is used).
>>>>>
>>>>> With the ParCS branch, building with buildNewParser=1 and specifying -j
>>>>> 2 also gives me no speedup, and only one cpu core used.
>>>>>
>>>>> Or are special parameters needed on scons?
>>>>>
>>>>> On 24/08/10 15:29, Michael Gogins wrote:
>>>>>> Can you give more details? It is working for me in both ParCS and head
>>>>>> for Windows and for ParCS for Linux.
>>>>>>
>>>>>> Regards,
>>>>>> Mike
>>>>>>
>>>>>> On Tue, Aug 24, 2010 at 3:25 PM, Felipe Sateler  wrote:
>>>>>>> JFTR, the __sync_lock_test_and_set stuff never actually worked for me
>>>>>>> (there appears to be a deadlock somewhere).
>>>>>>>
>>>>>>> On 24/08/10 08:26, Michael Gogins wrote:
>>>>>>>> Read my prior emails. I have changed it back to
>>>>>>>> __sync_lock_test_and_set because pthread spinlocks didn't seem to
>>>>>>>> actually work. Don't know why, it should be the way to go.
>>>>>>>>
>>>>>>>> Regards,
>>>>>>>> Mike
>>>>>>>>
>>>>>>>> On Tue, Aug 24, 2010 at 8:16 AM, John ff  wrote:
>>>>>>>>> What is the comparison between __sync_lock_test_and_set and
>>>>>>>>> pthread_spin_lock/pthread_spin_unlock ?
>>>>>>>>>
>>>>>>>>> It seems that Csound used the former until the weekend; just wondered
>>>>>>>>> why?
>>>>>>>>>
>>>>>>>>> ==John ffitch
>>>>>>>>>
>>>>>>>>> ------------------------------------------------------------------------------
>>>>>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>>>>>>> Be part of this innovative community and reach millions of netbook users
>>>>>>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>>>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>>>>>> _______________________________________________
>>>>>>>>> Csound-devel mailing list
>>>>>>>>> Csound-devel@lists.sourceforge.net
>>>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> Saludos,
>>>>>>> Felipe Sateler
>>>>>>>
>>>>>>> ------------------------------------------------------------------------------
>>>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>>>>> Be part of this innovative community and reach millions of netbook users
>>>>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>>>> _______________________________________________
>>>>>>> Csound-devel mailing list
>>>>>>> Csound-devel@lists.sourceforge.net
>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Saludos,
>>>>> Felipe Sateler
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>>> Be part of this innovative community and reach millions of netbook users
>>>>> worldwide. Take advantage of special opportunities to increase revenue and
>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>> _______________________________________________
>>>>> Csound-devel mailing list
>>>>> Csound-devel@lists.sourceforge.net
>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>> --
>>> Saludos,
>>> Felipe Sateler
>>>
>>> ------------------------------------------------------------------------------
>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>> Be part of this innovative community and reach millions of netbook users
>>> worldwide. Take advantage of special opportunities to increase revenue and
>>> speed time-to-market. Join now, and jumpstart your future.
>>> http://p.sf.net/sfu/intel-atom-d2d
>>> _______________________________________________
>>> Csound-devel mailing list
>>> Csound-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>
>>
>>
>>
>
>
> --
> Saludos,
> Felipe Sateler
>
> ------------------------------------------------------------------------------
> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
> Be part of this innovative community and reach millions of netbook users
> worldwide. Take advantage of special opportunities to increase revenue and
> speed time-to-market. Join now, and jumpstart your future.
> http://p.sf.net/sfu/intel-atom-d2d
> _______________________________________________
> 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

------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

This is a problem thta for me "magically" went away over the weekend.

There is a simple fix tpo treat that Die as nothing abd return NULL as in
the no-work case.  I think it happens when there is work remaining but it
is locked by an instrunment that is still running.

==John ff

> That's right, I forgot, there is a bug in ParCS that segfaults on
> trapped. Other csds seem to work OK. I will be debugging this very
> soon, I will at least take a look at it tonight.
>
> Regards,
> Mike
>
> On Tue, Aug 24, 2010 at 4:38 PM, Felipe Sateler 
> wrote:
>> Indeed, increasing ksmps makes it perform better:
>>
>> 500
>> 1 thread: real: 6.130s, CPU: 6.060s
>> 2 thread: real: 5.284s, CPU: 10.160s
>>
>> 1000
>> 1 thread: real: 6.145s, CPU: 6.110s
>> 2 thread: real: 5.869s, CPU: 10.220s
>>
>> I just noticed that parcs cannot render trapped when called with -j2:
>>
>> Expected a root to perform. Found none (1 remaining)
>> Early return from csoundPerform().
>>
>>
>> On 24/08/10 16:31, Michael Gogins wrote:
>>> For ParCS the parameter for threads is -j, e.g. -j4.
>>>
>>> There is too much threading overhead for you to get a speedup in head.
>>> Try increasing ksmps to 500 or 1000 or so and see what happens. The
>>> threading overhead (acquiring locks, putting threads to sleep and
>>> waking them up, waiting on barriers, etc.). happens every kperiod, so
>>> the more ksmps the less overhead per frame.
>>>
>>> It seems to me that ParCS involves less threading overhead than head,
>>> that is why it runs faster for me with smaller ksmps.
>>>
>>> Regards,
>>> Mike
>>>
>>> On Tue, Aug 24, 2010 at 4:22 PM, Felipe Sateler 
>>> wrote:
>>>> OK, so ParCS doesn't use more than one core for me, and the current
>>>> head
>>>> does... but is not effective:
>>>>
>>>> xanadu.csd
>>>> 1 thread:  real: 51.050s, CPU: 50.660s
>>>> 2 threads: real: 64.312s, CPU: 117.460s
>>>>
>>>> trapped.csd
>>>> 1 thread:  real: 13.686s, CPU: 13.350s
>>>> 2 threads: real: 17.876s, CPU: 30.860s
>>>>
>>>> ParCS does not seem to be able to use more than one core.
>>>>
>>>>
>>>> On 24/08/10 16:05, Michael Gogins wrote:
>>>>> Yes. useOpenMP=1 in the current head branch, buildNewParser=1 in the
>>>>> ParCS branch.
>>>>>
>>>>> Regards,
>>>>> Mike
>>>>>
>>>>> On Tue, Aug 24, 2010 at 3:59 PM, Felipe Sateler 
>>>>> wrote:
>>>>>> Well, it's been some time since I actually tried it. I just checked.
>>>>>>
>>>>>> Building head with no arguments, and specifying --num-threads=2 I
>>>>>> can
>>>>>> render xanadu and trapped, but no speedup (only one cpu core is
>>>>>> used).
>>>>>>
>>>>>> With the ParCS branch, building with buildNewParser=1 and specifying
>>>>>> -j
>>>>>> 2 also gives me no speedup, and only one cpu core used.
>>>>>>
>>>>>> Or are special parameters needed on scons?
>>>>>>
>>>>>> On 24/08/10 15:29, Michael Gogins wrote:
>>>>>>> Can you give more details? It is working for me in both ParCS and
>>>>>>> head
>>>>>>> for Windows and for ParCS for Linux.
>>>>>>>
>>>>>>> Regards,
>>>>>>> Mike
>>>>>>>
>>>>>>> On Tue, Aug 24, 2010 at 3:25 PM, Felipe Sateler
>>>>>>>  wrote:
>>>>>>>> JFTR, the __sync_lock_test_and_set stuff never actually worked for
>>>>>>>> me
>>>>>>>> (there appears to be a deadlock somewhere).
>>>>>>>>
>>>>>>>> On 24/08/10 08:26, Michael Gogins wrote:
>>>>>>>>> Read my prior emails. I have changed it back to
>>>>>>>>> __sync_lock_test_and_set because pthread spinlocks didn't seem to
>>>>>>>>> actually work. Don't know why, it should be the way to go.
>>>>>>>>>
>>>>>>>>> Regards,
>>>>>>>>> Mike
>>>>>>>>>
>>>>>>>>> On Tue, Aug 24, 2010 at 8:16 AM, John ff 
>>>>>>>>> wrote:
>>>>>>>>>> What is the comparison between __sync_lock_test_and_set and
>>>>>>>>>> pthread_spin_lock/pthread_spin_unlock ?
>>>>>>>>>>
>>>>>>>>>> It seems that Csound used the former until the weekend; just
>>>>>>>>>> wondered
>>>>>>>>>> why?
>>>>>>>>>>
>>>>>>>>>> ==John ffitch
>>>>>>>>>>
>>>>>>>>>> ------------------------------------------------------------------------------
>>>>>>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer
>>>>>>>>>> Program
>>>>>>>>>> Be part of this innovative community and reach millions of
>>>>>>>>>> netbook users
>>>>>>>>>> worldwide. Take advantage of special opportunities to increase
>>>>>>>>>> revenue and
>>>>>>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>>>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>>>>>>> _______________________________________________
>>>>>>>>>> Csound-devel mailing list
>>>>>>>>>> Csound-devel@lists.sourceforge.net
>>>>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> --
>>>>>>>> Saludos,
>>>>>>>> Felipe Sateler
>>>>>>>>
>>>>>>>> ------------------------------------------------------------------------------
>>>>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer
>>>>>>>> Program
>>>>>>>> Be part of this innovative community and reach millions of netbook
>>>>>>>> users
>>>>>>>> worldwide. Take advantage of special opportunities to increase
>>>>>>>> revenue and
>>>>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>>>>> _______________________________________________
>>>>>>>> Csound-devel mailing list
>>>>>>>> Csound-devel@lists.sourceforge.net
>>>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Saludos,
>>>>>> Felipe Sateler
>>>>>>
>>>>>> ------------------------------------------------------------------------------
>>>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer
>>>>>> Program
>>>>>> Be part of this innovative community and reach millions of netbook
>>>>>> users
>>>>>> worldwide. Take advantage of special opportunities to increase
>>>>>> revenue and
>>>>>> speed time-to-market. Join now, and jumpstart your future.
>>>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>>>> _______________________________________________
>>>>>> Csound-devel mailing list
>>>>>> Csound-devel@lists.sourceforge.net
>>>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Saludos,
>>>> Felipe Sateler
>>>>
>>>> ------------------------------------------------------------------------------
>>>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>>>> Be part of this innovative community and reach millions of netbook
>>>> users
>>>> worldwide. Take advantage of special opportunities to increase revenue
>>>> and
>>>> speed time-to-market. Join now, and jumpstart your future.
>>>> http://p.sf.net/sfu/intel-atom-d2d
>>>> _______________________________________________
>>>> Csound-devel mailing list
>>>> Csound-devel@lists.sourceforge.net
>>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>>
>>>
>>>
>>>
>>
>>
>> --
>> Saludos,
>> Felipe Sateler
>>
>> ------------------------------------------------------------------------------
>> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
>> Be part of this innovative community and reach millions of netbook users
>> worldwide. Take advantage of special opportunities to increase revenue
>> and
>> speed time-to-market. Join now, and jumpstart your future.
>> http://p.sf.net/sfu/intel-atom-d2d
>> _______________________________________________
>> 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
>
> ------------------------------------------------------------------------------
> Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
> Be part of this innovative community and reach millions of netbook users
> worldwide. Take advantage of special opportunities to increase revenue and
> speed time-to-market. Join now, and jumpstart your future.
> http://p.sf.net/sfu/intel-atom-d2d
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
>
>
>



------------------------------------------------------------------------------
Sell apps to millions through the Intel(R) Atom(Tm) Developer Program
Be part of this innovative community and reach millions of netbook users 
worldwide. Take advantage of special opportunities to increase revenue and 
speed time-to-market. Join now, and jumpstart your future.
http://p.sf.net/sfu/intel-atom-d2d
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net