Re: floating-point arithmetics

You can load the GMP library dynamically from an interactive ROOT session. For example with gmp-4.1.4 installed in my home directory, I can do

  root [0] gSystem.Load("/home/brun/gmp-4.1.4/.libs/libgmp.so")

Now you can modify your prime.C file to include all the relevant   extern "C" void gmp_xxx
prototype declarations and the equivalent includes. Then do

  root[1] .L prime.C+ (using ACLIC)
  root[2] prime(4567)

I agree that it would be nice to call the GMP library directly from CINT. This requires the dictionary for all GMP functions. If somebody succeeds in building it, let me know.

Rene Brun

Giannis Doxaras wrote:
>
> Hi all,
> I am experiencing a problem with high precision integer arithmetic in ROOT.
> After e^{20}, fmod() function seems to lose precision due to
> floating point error I presume.
> I wonder if there exists a class manipulating large integer, or
> floating point arithmetic in general , something similar to GMP (GNU
> Multiple Precision Arithmetic Library).
> If not, it should be rather helpful, if such a feature was included
> in forthcoming versions.
> Friendly,
>
> Doxaras John
> University Of Athens
> Division of Nuclear Physics & Elementary Particles
> d0x_at_r@$
>
> --------------------------------------------------------------------------------
> Name: prime.C
> prime.C Type: Plain Text (text/plain)
> Encoding: quoted-printable
Received on Wed Jan 05 2005 - 15:00:56 MET