The "f" version operates on floats rather than doubles. It is not built by default. If building fftw yourself, you need to configure it with the --enable-float option.
but this will build only the float variant. And the fftw3 build procedure is not able to build both variants ...
does anyone have any clues about how to solve this in a "packageful" way? thanks
Because a prog (I think brutefir) uses both variants, I have build my fftw3 rpm (with checkinstall) the way, Suse build their package:
- first build the normal version
- then in the install step install the normal version and build and install the float-version.
You can get the src.rpm from the Suse server. Look into the SuSE-9.0 tree.
Uwe
-- voiceINTERconnect www.voiceinterconnect.de ... smart speech applications from germany
