Please keep in mind us "realtimers."

Unless you are saying that for recent machines there is no longer a 
significant difference between realtime load capability between float and 
double?

(Then, too, there is OLPC. What would you propose in that case? And 
currently, is it float or double?)

Art Hunkins

----- Original Message ----- 
From: "Andres Cabrera" 
To: "Developer discussions" 
Sent: Wednesday, April 29, 2009 8:30 AM
Subject: Re: [Cs-dev] check for float size


The floats build will still be of use for older machines. Will the
floats option be completely removed, or the idea is to distribute only
the doubles version, but have the floats version as an alternative
compile option?

Cheers,
Andrés

On Wed, Apr 29, 2009 at 7:19 AM, Michael Gogins
 wrote:
> Thank you. On Linux, then, the decision would seem not to require any 
> brains.
>
> I will perform similar tests myself on both Linux and Windows, when I
> get the time, and obviously this kind of thing does not actually take
> a lot of time to do.
>
> Regards,
> Mike
>
> On 4/28/09, Felipe Sateler  wrote:
>> El 29/04/09 06:02 Michael Gogins escribió:
>>> I beg your indulgence as I repeat my call to omit the float version of
>>> csound, and to produce only the double version.
>>>
>>> Alternatively, to omit the double version, and produce only the float
>>> version.
>>>
>>> The criteria would be, of course, to measure how fast each build runs,
>>> and do double-blind listening and mathematical comparison tests to see
>>> how good each build sounds.
>>>
>>> The current situation is a big waste of time unless, and only unless,
>>> one of the builds is hopeless at efficiency, and the other is hopeless
>>> at precision.
>>>
>>
>> In 64bit linux, doubles seem to be even faster.
>>
>> http://www.nabble.com/Re:-Re:-Re:-64-bit---double-question-p17329282.html
>>
>> Saludos,
>> Felipe Sateler
>>
>
>
> --
> Michael Gogins
> Irreducible Productions
> http://www.michael-gogins.com
> Michael dot Gogins at gmail dot com
>
> ------------------------------------------------------------------------------
> Register Now & Save for Velocity, the Web Performance & Operations
> Conference from O'Reilly Media. Velocity features a full day of
> expert-led, hands-on workshops and two days of sessions from industry
> leaders in dedicated Performance & Operations tracks. Use code vel09scf
> and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
>



-- 


Andrés

------------------------------------------------------------------------------
Register Now & Save for Velocity, the Web Performance & Operations
Conference from O'Reilly Media. Velocity features a full day of
expert-led, hands-on workshops and two days of sessions from industry
leaders in dedicated Performance & Operations tracks. Use code vel09scf
and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/csound-devel 


------------------------------------------------------------------------------
Register Now & Save for Velocity, the Web Performance & Operations 
Conference from O'Reilly Media. Velocity features a full day of 
expert-led, hands-on workshops and two days of sessions from industry 
leaders in dedicated Performance & Operations tracks. Use code vel09scf 
and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

I do not know about OLPC.

For regular Csound, realtime performance, float versus double, the
current (sketchy) evidence is that double is 30% faster on Linux, and
float is 15% faster on Windows.

Since I now regularly use both Linux and Windows, I plan to do a less
sketchy evaluation soon. I will probably write a program that
automatically performs these kinds of tests.

Regards,
Mike

On 4/29/09, Art Hunkins  wrote:
> Please keep in mind us "realtimers."
>
> Unless you are saying that for recent machines there is no longer a
> significant difference between realtime load capability between float and
> double?
>
> (Then, too, there is OLPC. What would you propose in that case? And
> currently, is it float or double?)
>
> Art Hunkins
>
> ----- Original Message -----
> From: "Andres Cabrera" 
> To: "Developer discussions" 
> Sent: Wednesday, April 29, 2009 8:30 AM
> Subject: Re: [Cs-dev] check for float size
>
>
> The floats build will still be of use for older machines. Will the
> floats option be completely removed, or the idea is to distribute only
> the doubles version, but have the floats version as an alternative
> compile option?
>
> Cheers,
> Andrés
>
> On Wed, Apr 29, 2009 at 7:19 AM, Michael Gogins
>  wrote:
>> Thank you. On Linux, then, the decision would seem not to require any
>> brains.
>>
>> I will perform similar tests myself on both Linux and Windows, when I
>> get the time, and obviously this kind of thing does not actually take
>> a lot of time to do.
>>
>> Regards,
>> Mike
>>
>> On 4/28/09, Felipe Sateler  wrote:
>>> El 29/04/09 06:02 Michael Gogins escribió:
>>>> I beg your indulgence as I repeat my call to omit the float version of
>>>> csound, and to produce only the double version.
>>>>
>>>> Alternatively, to omit the double version, and produce only the float
>>>> version.
>>>>
>>>> The criteria would be, of course, to measure how fast each build runs,
>>>> and do double-blind listening and mathematical comparison tests to see
>>>> how good each build sounds.
>>>>
>>>> The current situation is a big waste of time unless, and only unless,
>>>> one of the builds is hopeless at efficiency, and the other is hopeless
>>>> at precision.
>>>>
>>>
>>> In 64bit linux, doubles seem to be even faster.
>>>
>>> http://www.nabble.com/Re:-Re:-Re:-64-bit---double-question-p17329282.html
>>>
>>> Saludos,
>>> Felipe Sateler
>>>
>>
>>
>> --
>> Michael Gogins
>> Irreducible Productions
>> http://www.michael-gogins.com
>> Michael dot Gogins at gmail dot com
>>
>> ------------------------------------------------------------------------------
>> Register Now & Save for Velocity, the Web Performance & Operations
>> Conference from O'Reilly Media. Velocity features a full day of
>> expert-led, hands-on workshops and two days of sessions from industry
>> leaders in dedicated Performance & Operations tracks. Use code vel09scf
>> and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
>> _______________________________________________
>> Csound-devel mailing list
>> Csound-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
>
>
>
> --
>
>
> Andrés
>
> ------------------------------------------------------------------------------
> Register Now & Save for Velocity, the Web Performance & Operations
> Conference from O'Reilly Media. Velocity features a full day of
> expert-led, hands-on workshops and two days of sessions from industry
> leaders in dedicated Performance & Operations tracks. Use code vel09scf
> and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
>
>
> ------------------------------------------------------------------------------
> Register Now & Save for Velocity, the Web Performance & Operations
> Conference from O'Reilly Media. Velocity features a full day of
> expert-led, hands-on workshops and two days of sessions from industry
> leaders in dedicated Performance & Operations tracks. Use code vel09scf
> and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
> _______________________________________________
> 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

------------------------------------------------------------------------------
Register Now & Save for Velocity, the Web Performance & Operations 
Conference from O'Reilly Media. Velocity features a full day of 
expert-led, hands-on workshops and two days of sessions from industry 
leaders in dedicated Performance & Operations tracks. Use code vel09scf 
and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Regarding the Linux benchmarks, Felipe was running a 64-bit version of
Linux.  I know that generally refers to integer size, but could data
transfer operations also be better optimized on a 64-bit CPU when using a
64-bit OS vs. a 32-bit OS ?  I.E., can we safely assume that we will get
similar performance gains with a 32-bit OS on modern CPUs?  (32-bit OSes are
probably still significantly more prevalent).

Anthony

Michael Gogins wrote on 4/29/09 3:59 PM:

> For regular Csound, realtime performance, float versus double, the
> current (sketchy) evidence is that double is 30% faster on Linux, and
> float is 15% faster on Windows.


------------------------------------------------------------------------------
Register Now & Save for Velocity, the Web Performance & Operations 
Conference from O'Reilly Media. Velocity features a full day of 
expert-led, hands-on workshops and two days of sessions from industry 
leaders in dedicated Performance & Operations tracks. Use code vel09scf 
and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net