Not at all. We are talking about 64 bit samples (double) instead of 32 bit samples (float) internally, not 64 bit code for 64 bit processors. 

I imagine Csound would sound exactly, or almost exactly, the same on a 64 bit processor as on a 32 bit processor.

I believe the scons option is:

useDouble=1

In other words, in both tests, Csound should be compiled for 32 bit code. In one test useDouble=0, in the other test useDouble=1.

Best,
Mike

-----Original Message-----
From: Iain Duncan 
Sent: Nov 7, 2005 4:33 PM
To: csound-devel@lists.sourceforge.net
Subject: Re: [Cs-dev] CVS Reorganization Request

Oops, which options do I need to change in SConstruct, just the "build 
for 64 bit machine"?

Thanks
iain

Michael Gogins wrote:
> It's not a question whether it's a lot of work, or a little work. The question is simply important, just as the question of sound quality is important. 
> 
> I think tests would be very interesting. I would be particularly interested if you had live pieces that stress Csound, and that could be "tuned" to be more or less demanding.
> 
> Best,
> Mike
> 
> -----Original Message-----
> From: Iain Duncan 
> Sent: Nov 7, 2005 3:37 PM
> To: csound-devel@lists.sourceforge.net
> Subject: Re: [Cs-dev] CVS Reorganization Request
> 
> If we're talking about a *lot* of extra work, I could run heavy load 
> tests of both if you'd like.
> 
> Iain
> 
> Michael Gogins wrote:
> 
>>How much speed are you talking about here? 
>>
>>Have you tried using 64 bits and found it too slow? Or are you guessing?
>>
>>Best,
>>Mike
>>
>>-----Original Message-----
>>From: Iain Duncan 
>>Sent: Nov 7, 2005 1:55 PM
>>To: csound-devel@lists.sourceforge.net
>>Subject: Re: [Cs-dev] CVS Reorganization Request
>>
>>
>>
>>
>>>Well, as you know, I think having 32 bits and 64 bits is a bad idea. It should just be 64 bits. So I am utterly indifferent to this question. I never build or use 32 bits.
>>>
>>>It continues to escape me why everyone else doesn't do the same. If it sometimes sounds better and is inarguably more precise and only runs a little slower, and oh yes it's much simpler to only build for 64 bits, what's the big deal? Are we musicians here or software engineering speed demons? 
>>>
>>>At the end of the day, Csound is a musical instrument and its only purpose is to make music.
>>
>>
>>Ah, but some of us are live performers, and the speed *does* matter, 
>>very very much. I am and will be for many years yet picking which 
>>corners to cut in order to get close to what I want running in real 
>>time. So I think it is good to still be able to run 32 bits.
>>
>>Iain
>>
>>
>>-------------------------------------------------------
>>SF.Net email is sponsored by:
>>Tame your development challenges with Apache's Geronimo App Server. Download
>>it for free - -and be entered to win a 42" plasma tv or your very own
>>Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
>>_______________________________________________
>>Csound-devel mailing list
>>Csound-devel@lists.sourceforge.net
>>https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
>>
>>
>>
>>
>>-------------------------------------------------------
>>SF.Net email is sponsored by:
>>Tame your development challenges with Apache's Geronimo App Server. Download
>>it for free - -and be entered to win a 42" plasma tv or your very own
>>Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
>>_______________________________________________
>>Csound-devel mailing list
>>Csound-devel@lists.sourceforge.net
>>https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
> 
> 
> 
> -------------------------------------------------------
> SF.Net email is sponsored by:
> Tame your development challenges with Apache's Geronimo App Server. Download
> it for free - -and be entered to win a 42" plasma tv or your very own
> Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
> 
> 
> 
> 
> 
> -------------------------------------------------------
> SF.Net email is sponsored by:
> Tame your development challenges with Apache's Geronimo App Server. Download
> it for free - -and be entered to win a 42" plasma tv or your very own
> Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
> 


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/csound-devel





-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

> Not at all. We are talking about 64 bit samples (double) instead of 32 bit samples (float) internally, not 64 bit code for 64 bit processors.

> I imagine Csound would sound exactly, or almost exactly, the same on a 64 bit processor as on a 32 bit processor.

Right, but would not a 64 bit processor likely have less overhead in 
dealing with 64 bit samples then a 32 bit would? I imagined that it 
would have to execute fewer instructions to do some of the 64 bit math.

Iain


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Ok, at ksmp=64, sr=44100, on my AMD64 3000 running in 32 bit mode, my 
gig ensemble stays steady at 69% cpu with 32 bit, and 82% cpu at 64bit. 
For me that is significant, especially as, at that krate, I could not 
hear an audible difference on my studio monitors. I will definitely 
stick to 32 bit for shows so I would ask that it not be discontinued.

I am also using no FFT code in there.

Can test rendering time later if you'd like, or provide more details on 
the code.

Iain

Michael Gogins wrote:
> Not at all. We are talking about 64 bit samples (double) instead of 32 bit samples (float) internally, not 64 bit code for 64 bit processors. 
> 
> I imagine Csound would sound exactly, or almost exactly, the same on a 64 bit processor as on a 32 bit processor.
> 
> I believe the scons option is:
> 
> useDouble=1
> 
> In other words, in both tests, Csound should be compiled for 32 bit code. In one test useDouble=0, in the other test useDouble=1.
> 
> Best,
> Mike
> 
> -----Original Message-----
> From: Iain Duncan 
> Sent: Nov 7, 2005 4:33 PM
> To: csound-devel@lists.sourceforge.net
> Subject: Re: [Cs-dev] CVS Reorganization Request
> 
> Oops, which options do I need to change in SConstruct, just the "build 
> for 64 bit machine"?
> 
> Thanks
> iain
> 
> Michael Gogins wrote:
> 
>>It's not a question whether it's a lot of work, or a little work. The question is simply important, just as the question of sound quality is important. 
>>
>>I think tests would be very interesting. I would be particularly interested if you had live pieces that stress Csound, and that could be "tuned" to be more or less demanding.
>>
>>Best,
>>Mike
>>
>>-----Original Message-----
>>From: Iain Duncan 
>>Sent: Nov 7, 2005 3:37 PM
>>To: csound-devel@lists.sourceforge.net
>>Subject: Re: [Cs-dev] CVS Reorganization Request
>>
>>If we're talking about a *lot* of extra work, I could run heavy load 
>>tests of both if you'd like.
>>
>>Iain
>>
>>Michael Gogins wrote:
>>
>>
>>>How much speed are you talking about here? 
>>>
>>>Have you tried using 64 bits and found it too slow? Or are you guessing?
>>>
>>>Best,
>>>Mike
>>>
>>>-----Original Message-----
>>>From: Iain Duncan 
>>>Sent: Nov 7, 2005 1:55 PM
>>>To: csound-devel@lists.sourceforge.net
>>>Subject: Re: [Cs-dev] CVS Reorganization Request
>>>
>>>
>>>
>>>
>>>
>>>>Well, as you know, I think having 32 bits and 64 bits is a bad idea. It should just be 64 bits. So I am utterly indifferent to this question. I never build or use 32 bits.
>>>>
>>>>It continues to escape me why everyone else doesn't do the same. If it sometimes sounds better and is inarguably more precise and only runs a little slower, and oh yes it's much simpler to only build for 64 bits, what's the big deal? Are we musicians here or software engineering speed demons? 
>>>>
>>>>At the end of the day, Csound is a musical instrument and its only purpose is to make music.
>>>
>>>
>>>Ah, but some of us are live performers, and the speed *does* matter, 
>>>very very much. I am and will be for many years yet picking which 
>>>corners to cut in order to get close to what I want running in real 
>>>time. So I think it is good to still be able to run 32 bits.
>>>
>>>Iain
>>>
>>>
>>>-------------------------------------------------------
>>>SF.Net email is sponsored by:
>>>Tame your development challenges with Apache's Geronimo App Server. Download
>>>it for free - -and be entered to win a 42" plasma tv or your very own
>>>Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
>>>_______________________________________________
>>>Csound-devel mailing list
>>>Csound-devel@lists.sourceforge.net
>>>https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>
>>>
>>>
>>>
>>>
>>>-------------------------------------------------------
>>>SF.Net email is sponsored by:
>>>Tame your development challenges with Apache's Geronimo App Server. Download
>>>it for free - -and be entered to win a 42" plasma tv or your very own
>>>Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
>>>_______________________________________________
>>>Csound-devel mailing list
>>>Csound-devel@lists.sourceforge.net
>>>https://lists.sourceforge.net/lists/listinfo/csound-devel
>>>
>>
>>
>>
>>-------------------------------------------------------
>>SF.Net email is sponsored by:
>>Tame your development challenges with Apache's Geronimo App Server. Download
>>it for free - -and be entered to win a 42" plasma tv or your very own
>>Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
>>_______________________________________________
>>Csound-devel mailing list
>>Csound-devel@lists.sourceforge.net
>>https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
>>
>>
>>
>>
>>-------------------------------------------------------
>>SF.Net email is sponsored by:
>>Tame your development challenges with Apache's Geronimo App Server. Download
>>it for free - -and be entered to win a 42" plasma tv or your very own
>>Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
>>_______________________________________________
>>Csound-devel mailing list
>>Csound-devel@lists.sourceforge.net
>>https://lists.sourceforge.net/lists/listinfo/csound-devel
>>
> 
> 
> 
> -------------------------------------------------------
> SF.Net email is sponsored by:
> Tame your development challenges with Apache's Geronimo App Server. Download
> it for free - -and be entered to win a 42" plasma tv or your very own
> Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
> 
> 
> 
> 
> 
> -------------------------------------------------------
> SF.Net email is sponsored by:
> Tame your development challenges with Apache's Geronimo App Server. Download
> it for free - -and be entered to win a 42" plasma tv or your very own
> Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
> _______________________________________________
> Csound-devel mailing list
> Csound-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/csound-devel
> 


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

32bit floats:
72.339u 0.457s 1:16.77 94.8%    0+0k 0+0io 0pf+0w

64bit doubles:
74.435u 0.453s 1:21.24 92.1%    0+0k 0+0io 15pf+0w

This is rendering to a file, with normal printing.  Suggests that the
speed loss is not much, less than 3%
==John ffitch


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

> 32bit floats:
> 72.339u 0.457s 1:16.77 94.8%    0+0k 0+0io 0pf+0w
> 
> 64bit doubles:
> 74.435u 0.453s 1:21.24 92.1%    0+0k 0+0io 15pf+0w
> 
> This is rendering to a file, with normal printing.  Suggests that the
> speed loss is not much, less than 3%
> ==John ffitch

Hmm, is there a reason that the real time loss would be so drastic ( 82% 
cpu vs 69% ) while real time would be so close? Could the actual limiter 
in the John's scenario be the writing to disk instead of the processing? 
'Cause 13% of an AMD 3000+ seems like a lot to me in comparsion.

I will conduct further real time tests later and report.

Iain


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

My quick test was a piece I had on disk rendered to file.  The machine
is a laptop so disk speed could be quite slow.
  I will try the same on the 64bit machine when I have a moment.

If someone writes a heap of oscilator test for realtime I will happily
test that as well.

==John ffitch


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

> 32bit floats:
> 72.339u 0.457s 1:16.77 94.8%    0+0k 0+0io 0pf+0w
> 
> 64bit doubles:
> 74.435u 0.453s 1:21.24 92.1%    0+0k 0+0io 15pf+0w
> 

Rendering to -n -m0 -d same machine, same orc/sco

32bit:
  71.086u 0.069s 1:11.54 99.4%    0+0k 0+0io 7pf+0w
64 bit:
  73.730u 0.090s 1:15.05 98.3%    0+0k 0+0io 9pf+0w

So taking elapsed time that is 3.51s in 71.56 -> 4.9%
Not very different.

==John ffitch


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Here are the render times (with -d -m128 -H0 -f -h -n --expression-opt)
of some example files from the manual and CVS sources:

                        floats       doubles

babo_expert.{orc,sco}:  7.880       11.430
ftconv.csd:             5.310        7.510
oscbnk.{orc,sco}:      16.190       22.960
trapped.csd:           10.880       12.680
xanadu.csd:            60.910       66.750


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net

Iain Duncan wrote:

> Hmm, is there a reason that the real time loss would be so drastic ( 82% 
> cpu vs 69% ) while real time would be so close? Could the actual limiter 
> in the John's scenario be the writing to disk instead of the processing? 

It is more likely that the two tests use a different set of opcodes
(probably you also have a much larger and more complex orchestra),
and John's test requires less memory access and more arithmetics.


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net