I again move for the canonical Csound build to use double-precision floating
point numbers for all real numbers. I would add also to use 32-bit signed
ints for all integers (unless there is a need for 64 bits).

-----Original Message-----
From: Josep M Comajuncosas 
To: csound@maths.ex.ac.uk 
Date: Friday, June 04, 1999 6:12 PM
Subject: Higher numerical precission in Csound


Hi all the programmers,
I´m currently developing some physical models and in order to implement
them in Csound it should work using 32bit floating point but better
64bit floats. Would it be possible to have such numerical resolution as
an option maybe? This is crucial for my project.
Thanks in advance!

--
Josep M Comajuncosas
C/ Circumval.lacio 75  08790 Gelida - Penedes
Catalunya - SPAIN
home phone : 93 7792243 / 00 34 3 7792243

Csound page at http://members.tripod.com/csound/

"Michael Gogins"  wrote:
> I again move for the canonical Csound build to use double-precision floating
> point numbers for all real numbers. I would add also to use 32-bit signed
> ints for all integers (unless there is a need for 64 bits).

Ouch!  Except in limited circumstances, is there any perceivable gain
from doubling the size of virtually all data in Csound?  Not only does
it greatly increase memory requirements, but on many architectures 64-bit
FP computations are significantly slower.  Main memories have certainly
grown enough to support larger data in the past few years, but all-important
cache memories have not--halving the number of cachable data elements is
going to hit performance pretty hard even if the CPU is as fast for 64-bit
FP as 32-bit.

I'd like to have an option for 64-bit floating point (especially if it
can be requested on a per-variable and per-table basis), but burdening
Csound with a 64-bit default seems entirely unnecessary.

		-Ed

Dear Canonical,

8 byte doubles would be real peachy for Macintosh as well.


Philip

Philip Aker
philip@vcn.bc.ca


On Fri, 4 Jun 1999, Michael Gogins wrote:

}I again move for the canonical Csound build to use double-precision floating
}point numbers for all real numbers. I would add also to use 32-bit signed
}ints for all integers (unless there is a need for 64 bits).
}
}-----Original Message-----
}From: Josep M Comajuncosas 

}I´m currently developing some physical models and in order to implement
}them in Csound it should work using 32bit floating point but better
}64bit floats. Would it be possible to have such numerical resolution as
}an option maybe? This is crucial for my project.