quadruple precision
For high-accuracy solar system body or interplanetary spacecraft orbit propagation or determination, it is recommended that state vectors and position vectors are represented in quadruple precision. Ideally, all floating point arithmetics in the orbit propagation/determination program should be in quadruple precision. When the software is written in FORTRAN, it is possible with some compilers to promote all double precision variables to quadruple precision using a compiler option that emulates quadruple precision arithmetics. Then you can get two version of the software almost for free: a fast version in double precision, and a quadruple precision version for high accuracy (extremely slow because the hardware does not support quadruple precision arithmetics so this is in software). In C**, one would use template to achieve something similar, although this requires that the software was originally designed with the possibility to change the scalar type. Is it possible to achieve something similar with Apache Commons Math and Orekit? I have been browsing for a java virtual machine that would allow promotion of all double to quad without success.
(from redmine: issue id 99, created on 2012-05-17, closed on 2015-05-29)