Here is a plugin that lets you use FFTW instead of the default FFT code
in Csound (this version only supports real FFT with power of two sizes,
but that is used by almost all opcodes anyway).

Build it with one of the following command lines (replace the include
path as needed, and make sure that the file csGblMtx.h is also found):

gcc -Wall -O2 -shared -I/usr/local/include/csound fftw_wrapper.c        \
    -o libfftw_wrapper.so -lfftw3 -lm -lpthread

  (for single precision Csound and double precision FFTW)

gcc -Wall -O2 -shared -DFFTW_USE_FLOAT                                  \
    -I/usr/local/include/csound fftw_wrapper.c -o libfftw_wrapper.so    \
    -lfftw3f -lm -lpthread

  (for single precision Csound and single precision FFTW - this is
   faster than the above, and is recommended if you have the FFTW library
   with 32 bit floats)

gcc -Wall -O2 -shared -DUSE_DOUBLE -I/usr/local/include/csound          \
    fftw_wrapper.c -o libfftw_wrapper.so -lfftw3 -lm -lpthread

  (for double precision Csound and double precision FFTW)

So far, I have only tested the first type of build (the others may or may
not work), and the code is not quite perfect yet, but if it does work for
you, then this plugin may be useful.

On Sunday 09 April 2006 12:55, Iain Duncan wrote:

> I had it working running 32bit linux on this chip, which was fine. I
> have only recently switched to pure 64 bit gentoo ( and you can see why
> I waited! ). If I can't get csound performing better than with 32 bit, I
> will likely switch back as that's really the only reason for the fast
> machine.