>>>>> "Diego" == Diego Giacomolli <diegogiacomolli at gmail.com> writes:
Diego> This is about Maxima 5.24.0 for Windows XP.
Diego> The "bf_fmin_cobyla" package is bugged.
Diego> It only works correctly if I set "iprint=0".
Diego> With "iprint=1" it works well but gives an error message at the end of
Diego> the calculation.
Diego> With "iprint=2" or "iprint=3" it gives an error message in the first
Diego> step of the calculation and then stops.
Diego> It apparently does not want to print big float values for the figure
Diego> of merit being minimized.
Could be a bug. It would certainly help a lot if you actually
provided a (simple) example illustrating this. Sorry I can't read
your mind.
Diego> Another problem: the command "loadprint:false" does not work. That is,
Diego> loading messages are displayed when I load a package.
Again, an example is always helpful. I do not see any message about
loading a file. I do see that load returns a value which happens to
be the full name of file that is loaded, but that is different from
when loadprint is true, which prints out name of the file being loaded
and still returns the name of the file loaded as the value.
Diego> Is there a big float version of "quad_qagi"? I really need it.
No. quad_qagi uses a Fortran (translated to Lisp) routine, which only
exists in a double-float version.
Perhaps romberg will work (with a suitable change of variable of
integration to change the infinite limit to a finite limit, which is
what quad_qagi does too.)
Diego> Is there a big float version of "lbfgs"? I really need it.
lbfgs is also a Fortran routine that only exists in a double-float
version. However, I think it might be possible to make a bigfloat
version of that without too much trouble since I don't think it has
any hardwired numerical contansts. (This is unlike quad_qagi which
has some hardwired constants for the integration rules. These would
need to be converted on the fly to the appropriate bigfloat
precision.)
Ray