Re: extended double precision: looking for a small C library

On Sep 4, 4:21 pm, Axel Vogt <&nore...@xxxxxxxxxxx> wrote:
It is not actually numerical Analysis ... I am looking for a tiny lib in C,
which provides basic arithmetics (addition, multiplication and their inverse)
for types 'double double' and converting from double precision and back for
complex floats (so C++ with std for simplifying noations would be fine), no
elementary functions.

I just want to have it handy if I have the need to incorporate it in C(++)
code (using VC), a kind of 'plugg in'.

Does anybody have a source or a link?

What I tried is 'doubledouble' by Briggs (for Reals only), Moshier's qlib
(I gave up to strip it down and bring to run), dq by Bailey & Hida (this
is far too much) and SCSlib by Defour (makes me crazy) ... besides the
full libraries like MPFR (or older and smaller).

Is there no 'slim' tool, do I have to do [and test :-( ] that by myself,
just basic arithmetics over the Complex in 'extended' precison as C code?

Obviously you've looked at several libraries
already without complete satisfaction. I'd
suggest looking at this:

http://www.softlist.net/program/cephes_mathematical_library__c_sources_-software.html

Your notion of extended precision may be just
"double double", possibly the equivalent of
the internal FPU format on Intel chips which
Borland used to support. But for more general
applications you might want additional precision.

So I think it makes sense to commit to a C++
template class for complex arithmetic (into
which various real extended precision types
can be assigned). It appears the above does
that, though if you really wanted a smallest
footprint that supplies only arithmetic and
not (say) elementary transcendental functions
of complex arguments, a roll-your-own approach
should be easy enough in C++.

regards, chip
.

Quantcast