Even if they are no noticeable differences for this or that orchestra, I´m for
implementing double precision for a rather large number of specific needs. Anyway
a 12% of increased compilation time seems to me a very reasonable price to pay!
I´ll prepare this weekend an orchestra showing (hopefully) how much better doulbe
precision should sound compared to the current float implementation.
I´ll send it to Ed to see if that´s true ;-)


Ed Hall wrote:

> It took
> all of fifteen minutes to produce two binaries for Csound 3.54, one
> with floats, the other with doubles.

> Here are the results:
>
>     float:    95.72 seconds
>     double:  107.47 seconds
>
> So the double version is about 12% slower.

--
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/

> . . .
> I´ll prepare this weekend an orchestra showing (hopefully) how much better doulbe
> precision should sound compared to the current float implementation.
> I´ll send it to Ed to see if that´s true ;-)

Even better, folks can try it for themselves.  Linux binaries (for Intel)
are at:

    http://www.weirdnoise.com/csound/csound-d   # double-prec version
    http://www.weirdnoise.com/csound/csound-f   # single-prec version

while the shell script I used to wrap GCC with is at:

    http://www.weirdnoise.com/csound/doubleup.sh

Copies of the files I used for benchmarks are at:

    http://www.weirdnoise.com/csound/xanadu44.orc
    http://www.weirdnoise.com/csound/xanadus.sco

Oh, and if you aren't already using Version 3.54 of Csound, you'll need:

    http://www.weirdnoise.com/csound/csound.txt

or it will cry mysteriously about a missing "DB" file.

		-Ed

PS.  Please don't use www.weirdnoise.com as a general site for Csound--
bandwidth is limited!