[Maxima] Long-float variant giving more precision to numerical computations.

[Raymond Toy (RT/EUS)]

Douglas Crosher wrote:
> Hello Ray,
> 
> The changes have been committed, except for the changes to the Fortran 
> code.  

I didn't expect you to check the changes in so quickly! :-)
Patches
> to the Fortran code are temporarily available here:
> 
> Patches to use long-float (SCL, CLISP):
> http://www.scieneer.com/files/maxima-lf-patch1.txt.gz
> 
> Patches to use CMU CL double-double-float:
> http://www.scieneer.com/files/maxima-ddf-patch1.txt.gz
> 
> The CLISP version is working relatively well.  The CMUCL double-double 
> float version has
> some problems that have not been explored yet but can be reproduced with 
> the above patch.

I'm not sure I would want double-double's.  They're fairly slow.  But 
more importantly, we don't have double-double accuracy for the special 
functions.  It will take quite a bit of work to get that in place.  I 
think that effort would be better spent on getting bfloat versions of them.

The same goes for x86 long-float (double extended) support, except they 
should be fast.

Dunno.

Ray