For some reason, ParCS is not producing any audio with the sheppard  
tone CSD, whatever -j setting. It  plays OK with trapped up until  
section 2, when it hangs (with -j 2).

Victor

------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Does the ParCS code run the DAG analysis for the next kperiod in its own thread?

Regards,
Mike

On Mon, Aug 2, 2010 at 5:54 AM, Victor Lazzarini
 wrote:
> For some reason, ParCS is not producing any audio with the sheppard
> tone CSD, whatever -j setting. It  plays OK with trapped up until
> section 2, when it hangs (with -j 2).
>
> Victor
>
> ------------------------------------------------------------------------------
> The Palm PDK Hot Apps Program offers developers who use the
> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
> of $1 Million in cash or HP Products. Visit us here for more details:
> http://p.sf.net/sfu/dev2dev-palm
> _______________________________________________
> 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

------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

I think there is also something wrong with the new parser, because the  
sheppard example produces the wrong output with ordinary csound and  
new parser.

Victor
On 2 Aug 2010, at 14:01, Michael Gogins wrote:

> Does the ParCS code run the DAG analysis for the next kperiod in its  
> own thread?
>
> Regards,
> Mike
>
> On Mon, Aug 2, 2010 at 5:54 AM, Victor Lazzarini
>  wrote:
>> For some reason, ParCS is not producing any audio with the sheppard
>> tone CSD, whatever -j setting. It  plays OK with trapped up until
>> section 2, when it hangs (with -j 2).
>>
>> Victor
>>
>> ------------------------------------------------------------------------------
>> The Palm PDK Hot Apps Program offers developers who use the
>> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
>> of $1 Million in cash or HP Products. Visit us here for more details:
>> http://p.sf.net/sfu/dev2dev-palm
>> _______________________________________________
>> 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
>
> ------------------------------------------------------------------------------
> The Palm PDK Hot Apps Program offers developers who use the
> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
> of $1 Million in cash or HP Products. Visit us here for more details:
> http://p.sf.net/sfu/dev2dev-palm
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel


------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

>>>>> "Victor" == Victor Lazzarini  writes:

 Victor> For some reason, ParCS is not producing any audio with the sheppard  
 Victor> tone CSD, whatever -j setting. It  plays OK with trapped up until  
 Victor> section 2, when it hangs (with -j 2).

Is this still the case after the fix to division?

==John ffitch

------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

I'll test later today and let you know.

Victor
On 4 Aug 2010, at 10:25, john ffitch wrote:

>>>>>> "Victor" == Victor Lazzarini  writes:
>
> Victor> For some reason, ParCS is not producing any audio with the  
> sheppard
> Victor> tone CSD, whatever -j setting. It  plays OK with trapped up  
> until
> Victor> section 2, when it hangs (with -j 2).
>
> Is this still the case after the fix to division?
>
> ==John ffitch
>
> ------------------------------------------------------------------------------
> The Palm PDK Hot Apps Program offers developers who use the
> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
> of $1 Million in cash or HP Products. Visit us here for more details:
> http://p.sf.net/sfu/dev2dev-palm
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel


------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Sheppard tones now play OK. Timing has also improved (for -j 1), 0.908  
s  (compared to 0.76 sec with the old parser).

-j 2 still slow on this machine, but the HW is only dual core.

Victor

On 4 Aug 2010, at 10:25, john ffitch wrote:

>>>>>> "Victor" == Victor Lazzarini  writes:
>
> Victor> For some reason, ParCS is not producing any audio with the  
> sheppard
> Victor> tone CSD, whatever -j setting. It  plays OK with trapped up  
> until
> Victor> section 2, when it hangs (with -j 2).
>
> Is this still the case after the fix to division?
>
> ==John ffitch
>
> ------------------------------------------------------------------------------
> The Palm PDK Hot Apps Program offers developers who use the
> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
> of $1 Million in cash or HP Products. Visit us here for more details:
> http://p.sf.net/sfu/dev2dev-palm
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel


------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

> Sheppard tones now play OK. Timing has also improved (for -j 1), 0.908
> s  (compared to 0.76 sec with the old parser).
>
> -j 2 still slow on this machine, but the HW is only dual core.
>
> Victor
>

Thanks;

There is a performance issue that is concerning me.  If the instruments
are very expensive then all seems oK, but without th eload-balancing code
fully implemented it seems that the overhead is way too much.  That
suggests (a) getting the load-balance operationsal and (b) a critial look
at the dispatch code

Both on my agenda but as ever assistance most welcome

And for Michael, I think that the DAG is constructed by the main thread
after tall processes have done the k-cycle.  So not a dedicated thread.  I
am certainly concerned about this area.  Highest priority now

==John ff


------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

About ParCS performance in general, the fact that most instruments
will end up using an output opcode is a setup for "false sharing"
whereby the cache lines become invalidated and cache accesses must be
serialized. It may be worth thinking about changing opcodes like the
output opcodes for parallel Csound such that data is reduced instead
of shared. Reducing the data within the opcode probably will call for
some thought. Perhaps the data can be reduced when it finally goes to
the driver output buffer. This would mean that outs etc. enqueue
buffers that are then summed when the outputs are written to the
driver.

At a minimum, we should try to determine experimentally whether the
output opcodes do in fact cause false sharing or not.

Regards,
Mike

On Wed, Aug 4, 2010 at 6:15 AM,   wrote:
>> Sheppard tones now play OK. Timing has also improved (for -j 1), 0.908
>> s  (compared to 0.76 sec with the old parser).
>>
>> -j 2 still slow on this machine, but the HW is only dual core.
>>
>> Victor
>>
>
> Thanks;
>
> There is a performance issue that is concerning me.  If the instruments
> are very expensive then all seems oK, but without th eload-balancing code
> fully implemented it seems that the overhead is way too much.  That
> suggests (a) getting the load-balance operationsal and (b) a critial look
> at the dispatch code
>
> Both on my agenda but as ever assistance most welcome
>
> And for Michael, I think that the DAG is constructed by the main thread
> after tall processes have done the k-cycle.  So not a dedicated thread.  I
> am certainly concerned about this area.  Highest priority now
>
> ==John ff
>
>
> ------------------------------------------------------------------------------
> The Palm PDK Hot Apps Program offers developers who use the
> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
> of $1 Million in cash or HP Products. Visit us here for more details:
> http://p.sf.net/sfu/dev2dev-palm
> _______________________________________________
> 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

------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

As far as the sheppard tone example I am using to test is concerned,  
there is no real difference between running it with or without the  
'out' opcode.
BTW, I am using this code because it should be where the gains would  
definitely show up as it is based on 10 parallel instances of the same  
instrument.

Victor
On 4 Aug 2010, at 21:16, Michael Gogins wrote:

> About ParCS performance in general, the fact that most instruments
> will end up using an output opcode is a setup for "false sharing"
> whereby the cache lines become invalidated and cache accesses must be
> serialized. It may be worth thinking about changing opcodes like the
> output opcodes for parallel Csound such that data is reduced instead
> of shared. Reducing the data within the opcode probably will call for
> some thought. Perhaps the data can be reduced when it finally goes to
> the driver output buffer. This would mean that outs etc. enqueue
> buffers that are then summed when the outputs are written to the
> driver.
>
> At a minimum, we should try to determine experimentally whether the
> output opcodes do in fact cause false sharing or not.
>
> Regards,
> Mike
>
> On Wed, Aug 4, 2010 at 6:15 AM,   wrote:
>>> Sheppard tones now play OK. Timing has also improved (for -j 1),  
>>> 0.908
>>> s  (compared to 0.76 sec with the old parser).
>>>
>>> -j 2 still slow on this machine, but the HW is only dual core.
>>>
>>> Victor
>>>
>>
>> Thanks;
>>
>> There is a performance issue that is concerning me.  If the  
>> instruments
>> are very expensive then all seems oK, but without th eload- 
>> balancing code
>> fully implemented it seems that the overhead is way too much.  That
>> suggests (a) getting the load-balance operationsal and (b) a  
>> critial look
>> at the dispatch code
>>
>> Both on my agenda but as ever assistance most welcome
>>
>> And for Michael, I think that the DAG is constructed by the main  
>> thread
>> after tall processes have done the k-cycle.  So not a dedicated  
>> thread.  I
>> am certainly concerned about this area.  Highest priority now
>>
>> ==John ff
>>
>>
>> ------------------------------------------------------------------------------
>> The Palm PDK Hot Apps Program offers developers who use the
>> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
>> of $1 Million in cash or HP Products. Visit us here for more details:
>> http://p.sf.net/sfu/dev2dev-palm
>> _______________________________________________
>> 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
>
> ------------------------------------------------------------------------------
> The Palm PDK Hot Apps Program offers developers who use the
> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
> of $1 Million in cash or HP Products. Visit us here for more details:
> http://p.sf.net/sfu/dev2dev-palm
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel


------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

OK, that's good to know.

Mike

On Wed, Aug 4, 2010 at 4:56 PM, Victor Lazzarini
 wrote:
> As far as the sheppard tone example I am using to test is concerned,
> there is no real difference between running it with or without the
> 'out' opcode.
> BTW, I am using this code because it should be where the gains would
> definitely show up as it is based on 10 parallel instances of the same
> instrument.
>
> Victor
> On 4 Aug 2010, at 21:16, Michael Gogins wrote:
>
>> About ParCS performance in general, the fact that most instruments
>> will end up using an output opcode is a setup for "false sharing"
>> whereby the cache lines become invalidated and cache accesses must be
>> serialized. It may be worth thinking about changing opcodes like the
>> output opcodes for parallel Csound such that data is reduced instead
>> of shared. Reducing the data within the opcode probably will call for
>> some thought. Perhaps the data can be reduced when it finally goes to
>> the driver output buffer. This would mean that outs etc. enqueue
>> buffers that are then summed when the outputs are written to the
>> driver.
>>
>> At a minimum, we should try to determine experimentally whether the
>> output opcodes do in fact cause false sharing or not.
>>
>> Regards,
>> Mike
>>
>> On Wed, Aug 4, 2010 at 6:15 AM,   wrote:
>>>> Sheppard tones now play OK. Timing has also improved (for -j 1),
>>>> 0.908
>>>> s  (compared to 0.76 sec with the old parser).
>>>>
>>>> -j 2 still slow on this machine, but the HW is only dual core.
>>>>
>>>> Victor
>>>>
>>>
>>> Thanks;
>>>
>>> There is a performance issue that is concerning me.  If the
>>> instruments
>>> are very expensive then all seems oK, but without th eload-
>>> balancing code
>>> fully implemented it seems that the overhead is way too much.  That
>>> suggests (a) getting the load-balance operationsal and (b) a
>>> critial look
>>> at the dispatch code
>>>
>>> Both on my agenda but as ever assistance most welcome
>>>
>>> And for Michael, I think that the DAG is constructed by the main
>>> thread
>>> after tall processes have done the k-cycle.  So not a dedicated
>>> thread.  I
>>> am certainly concerned about this area.  Highest priority now
>>>
>>> ==John ff
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> The Palm PDK Hot Apps Program offers developers who use the
>>> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
>>> of $1 Million in cash or HP Products. Visit us here for more details:
>>> http://p.sf.net/sfu/dev2dev-palm
>>> _______________________________________________
>>> 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
>>
>> ------------------------------------------------------------------------------
>> The Palm PDK Hot Apps Program offers developers who use the
>> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
>> of $1 Million in cash or HP Products. Visit us here for more details:
>> http://p.sf.net/sfu/dev2dev-palm
>> _______________________________________________
>> Csound-devel mailing list
>> Csound-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>
>
> ------------------------------------------------------------------------------
> The Palm PDK Hot Apps Program offers developers who use the
> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
> of $1 Million in cash or HP Products. Visit us here for more details:
> http://p.sf.net/sfu/dev2dev-palm
> _______________________________________________
> 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

------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

But you raised a good point there. It's not the problem right now, but  
we need to keep an eye out for it.

Victor

On 4 Aug 2010, at 21:59, Michael Gogins wrote:

> OK, that's good to know.
>
> Mike
>
> On Wed, Aug 4, 2010 at 4:56 PM, Victor Lazzarini
>  wrote:
>> As far as the sheppard tone example I am using to test is concerned,
>> there is no real difference between running it with or without the
>> 'out' opcode.
>> BTW, I am using this code because it should be where the gains would
>> definitely show up as it is based on 10 parallel instances of the  
>> same
>> instrument.
>>
>> Victor
>> On 4 Aug 2010, at 21:16, Michael Gogins wrote:
>>
>>> About ParCS performance in general, the fact that most instruments
>>> will end up using an output opcode is a setup for "false sharing"
>>> whereby the cache lines become invalidated and cache accesses must  
>>> be
>>> serialized. It may be worth thinking about changing opcodes like the
>>> output opcodes for parallel Csound such that data is reduced instead
>>> of shared. Reducing the data within the opcode probably will call  
>>> for
>>> some thought. Perhaps the data can be reduced when it finally goes  
>>> to
>>> the driver output buffer. This would mean that outs etc. enqueue
>>> buffers that are then summed when the outputs are written to the
>>> driver.
>>>
>>> At a minimum, we should try to determine experimentally whether the
>>> output opcodes do in fact cause false sharing or not.
>>>
>>> Regards,
>>> Mike
>>>
>>> On Wed, Aug 4, 2010 at 6:15 AM,   wrote:
>>>>> Sheppard tones now play OK. Timing has also improved (for -j 1),
>>>>> 0.908
>>>>> s  (compared to 0.76 sec with the old parser).
>>>>>
>>>>> -j 2 still slow on this machine, but the HW is only dual core.
>>>>>
>>>>> Victor
>>>>>
>>>>
>>>> Thanks;
>>>>
>>>> There is a performance issue that is concerning me.  If the
>>>> instruments
>>>> are very expensive then all seems oK, but without th eload-
>>>> balancing code
>>>> fully implemented it seems that the overhead is way too much.  That
>>>> suggests (a) getting the load-balance operationsal and (b) a
>>>> critial look
>>>> at the dispatch code
>>>>
>>>> Both on my agenda but as ever assistance most welcome
>>>>
>>>> And for Michael, I think that the DAG is constructed by the main
>>>> thread
>>>> after tall processes have done the k-cycle.  So not a dedicated
>>>> thread.  I
>>>> am certainly concerned about this area.  Highest priority now
>>>>
>>>> ==John ff
>>>>
>>>>
>>>> ------------------------------------------------------------------------------
>>>> The Palm PDK Hot Apps Program offers developers who use the
>>>> Plug-In Development Kit to bring their C/C++ apps to Palm for a  
>>>> share
>>>> of $1 Million in cash or HP Products. Visit us here for more  
>>>> details:
>>>> http://p.sf.net/sfu/dev2dev-palm
>>>> _______________________________________________
>>>> 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
>>>
>>> ------------------------------------------------------------------------------
>>> The Palm PDK Hot Apps Program offers developers who use the
>>> Plug-In Development Kit to bring their C/C++ apps to Palm for a  
>>> share
>>> of $1 Million in cash or HP Products. Visit us here for more  
>>> details:
>>> http://p.sf.net/sfu/dev2dev-palm
>>> _______________________________________________
>>> Csound-devel mailing list
>>> Csound-devel@lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
>>
>> ------------------------------------------------------------------------------
>> The Palm PDK Hot Apps Program offers developers who use the
>> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
>> of $1 Million in cash or HP Products. Visit us here for more details:
>> http://p.sf.net/sfu/dev2dev-palm
>> _______________________________________________
>> 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
>
> ------------------------------------------------------------------------------
> The Palm PDK Hot Apps Program offers developers who use the
> Plug-In Development Kit to bring their C/C++ apps to Palm for a share
> of $1 Million in cash or HP Products. Visit us here for more details:
> http://p.sf.net/sfu/dev2dev-palm
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel


------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

I would help if I understood the code. Where should I be looking? Is  
the load-balancing code there already but not yet enabled?

Victor
On 4 Aug 2010, at 11:15, jpff@cs.bath.ac.uk wrote:

> There is a performance issue that is concerning me.  If the  
> instruments
> are very expensive then all seems oK, but without th eload-balancing  
> code
> fully implemented it seems that the overhead is way too much.  That
> suggests (a) getting the load-balance operationsal and (b) a critial  
> look
> at the dispatch code
>
> Both on my agenda but as ever assistance most welcome


------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

The out opcodes are not treated as a shared resource, but the inner
code is protected by a spinlock.  We thought about that one!
Similarly adding into a global variable only needs a spinlock -- not
serialisation.
==John ffitch

------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

The changes are mainly in three files

Support code (Engine/cs_par_base.c)
Code for DAG creation etc (Engine/cs_par_dispatch.c)
The threaded perf (Top/csound.c)

There are minor changes elsewhere, like spinlocks round outX opcodes.

Other major area is the parser/semantics where spinlocks are inserted
and the DAG shape is determined/

==John ffitch

------------------------------------------------------------------------------
The Palm PDK Hot Apps Program offers developers who use the
Plug-In Development Kit to bring their C/C++ apps to Palm for a share
of $1 Million in cash or HP Products. Visit us here for more details:
http://p.sf.net/sfu/dev2dev-palm
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net