Why I think lapack matrices should not be converted to list-of-lists.

Setting fpprec to some number affects the arithmetic, and some elementary
functions. It does not appropriately set the error tolerances, iteration
limits, etc. for functions. Thus a method that is supposed to compute (say)
pi to double precision using a series, will ordinariliy not compute pi to
100 decimals by doing fpprec:100.

One can construct programs that work this way, but they are generally quite
different from the usual efficient methods.  A double-precision cosine
routine might use a fixed-degree polynomial approximation.  A bigfloat
cosine cannot.

RJF


> -----Original Message-----
> From: Raymond Toy [mailto:raymond.toy at ericsson.com]
> Sent: Wednesday, January 03, 2007 9:38 AM
> To: sen1 at math.msu.edu
> Cc: Richard Fateman; maxima at math.utexas.edu; 'Robert Dodier'
> Subject: Re: [Maxima] Why I think lapack matrices should not be converted
> to list-of-lists.
> 
> >>>>> "sen1" == sen1  <sen1 at math.msu.edu> writes:
> 
>     sen1> This brings up another question.
>     sen1> Is it reasonable to set some kind of switch, maybe at compile
> time, so
>     sen1> that some routines loaded in can be run in a given preset
> precision?
> 
>     sen1> Perhaps there could be an option to create two "maxima"
> packages. One
>     sen1> as it now is, and one where all computations in certain routines
> can
>     sen1> be run with a preset precision.  Initially, the user would be
> responsible for
>     sen1> conversion problems.
> 
> You can already do that with bfloats.  Just set fpprec to the desired
> precision.
> 
> Unfortunately, not all functions have bfloat implementations.
> 
> Ray