New "feature" request

Kevin Ryde user42 at
Fri Feb 27 08:27:49 CET 2004

"J. Foug" <jim_fougeron at> writes:
> I would love some interface (function(s) or data objects), which
> reported certain "build envirionment" information.

That idea has come up once or twice before, but I don't think it's
excited much interest.  We probably prefer programs are written to
what's documented, and left at that.

> I have wanted
> something like this for a while but never asked. With the recent
> anouncement of the FFT problem, it has again risen up as an issue.

It may or may not be a problem in 4.1.2.  It looks doubtful, but we
don't have a failing case.  It was definitely a bug in the current
unreleased development sources, but 4.1.2 is subtly different.  If
only the author of that code would deign to review it :-(.

If it turns out to be a bug the suggestion would probably be to run a
case provoking it, rather than enquiring about the configuration
options.  Or if we make a point release then just demand a new enough
gmp for all users.

> 2.  If alloca (or malloc) are used.

This certainly can be a problem.  Lobby Torbjorn to consider what
Niels M suggested for malloc on "big enough" blocks, thereby keeping
stack usage under control.  :)

> 4.  Tuning values (basecase, karatusba, toom-n, fft, ...)

Hopefully such things are merely optimizations and won't affect

> Yes, some of the above IS available in the header files.  However, I
> currently provide many prebuild Win32 DLL's of GMP (I build about 10
> different libgmp-3.dll's for different CPU's)

The next release will have a scheme of run-time detection and
selection of low level routines, which may help.

> Before, I wanted these data to simply
> be able to "nag" the user, due to non-optimal GMP (of course
> providing a way to allow the user to turn off the nagging).

Oh, well, we've always hoped anyone who cares enough would take the
trouble to do what's necessary.

More information about the gmp-discuss mailing list