[linux-audio-user] ANN: swh-plugins 0.3.7

[Jaakko Prättälä]

On Sunday 09 March 2003 03:57, Steve Harris wrote:
> On Sun, Mar 09, 2003 at 01:26:38 +0200, Jaakko Prättälä wrote:
> >
> > I think there'll be no problems with the precision since
> > (to my knowledge) double precision is fftw's default.
> > Maybe swh's configure-script should look for 'd'-prefixed files too?
>
> I think single used to be the default. Double, /may/ work but it will be
> slow if it does.
>
> The FFTW people should really sort this out.
I totally agree, but the latest fftw (2.1.3) release is
from 1999 :-)

I'm quite sure that
the default for fftw is _double_,
unless you specify --enable-float for the build.

Here are some findings
to support this presumption:

These babblings are not so relevant, but I included
them for completeness' sake...
<cryptic_cyberbabble>

  - from the default fftw.h
    (no configure-script options, fftw version 2.1.3):

  /* Define for using single precision */
  /*
   * If you can, use configure --enable-float instead of changing this
   * flag directly
   */
   /* #undef FFTW_ENABLE_FLOAT */

   /* our real numbers */
   #ifdef FFTW_ENABLE_FLOAT
   typedef float fftw_real;
   #else
   typedef double fftw_real;
   #endif

   - make |grep "FFTW_ENABLE_FLOAT" produces no output
     other than compiler warnings.

</cryptic_cyberbabble>

...but the next says a lot:
- from the faq (fftw-2.1.3/FAQ/fftw-faq.ascii):
  Question 2.9.  How do I compile FFTW to run in single precision?

  On a Unix system: configure --enable-float.  On a non-Unix system: edit
  fftw/fftw.h to #define the symbol FFTW_ENABLE_FLOAT.  In both cases, you
  must then recompile FFTW.


The logical solution would be to
build one's fftw with
./configure --enable-shared --enable-float --enable-type-prefix
and make non-prefixed symlinks to the double version.

In this case (I have symlinks to the shared lib),
Steve, your configure-script says
checking for fftw_one in -lsfftw... (cached) no
checking for fftw_one in -lfftw... (cached) yes
(strange)
so I guess it will link against the double-precision
version of the lib. I haven't gotten into testing
imp or pitch_scale yet, so I can't say if there are
any problems with this. Last night I tried out
delay-o-rama + flanger. Tapping a mic never
sounded so good :-)

-- 
Jaakko Prättälä
Jaakko.Prattala at Helsinki.FI