How to use GMM++ with QD type (double-double and quad-double)

The QD library (see http://www.cs.berkeley.edu/~yozo or http://www.nersc.gov/~dhb/mpdist/mpdist.html) is an efficient library for double-double (32 decimal digits) and quad-double (approx. 64 decimal digits). Once you installed this library on your system you have to link your program with QD library (with -lqd). In your program, include the header files of QD with

#include <qd/dd.h>
#include <qd/qd.h>
#include <qd/fpu.h>

Then the two type dd_real and qd_real will be usable with GMM++ . You will also be able to use std::complex<dd_real> and std::complex<qdreal>
IMPORTANT : do not forget to initialize QD before using it with the following call

unsigned int old_cw;
fpu_fix_start(&old_cw);

This disables the 80 bits precision of x86 processors which conflicts with QD. Once you finished to use QD you can reactivate it with

fpu_fix_end(&old_cw);

(see the QD documentation for more details).

Gmm++ User Documentation

How to use GMM++ with QD type (double-double and quad-double)