GMAP developed by Wolfgang P. Poenitz is a code to obtain evaluations of cross sections and their uncertainties based on the combined data from various experiments. The code employs the Bayesian version of the Generalized Least Squares method and is named after the mathematicians Gauss, Markov, and Aitken, who all contributed to the statistical theory around the linear least squares method. Input to GMAP are experimental data preprocessed by the DATP code, also developed by Poenitz.
Notably, these codes have been employed for the evaluation of neutron cross section standards. Among other available documents, there is a short user guide for DATP and GMAP and a more detailed report on the technicalities of the codes.
The purpose of this repository is to track modifications to GMAP and tro provide a reference for future developments related to this code.
-
Converted single precision numeric literals to double precision, e.g.,
0.5to0.5D0, as otherwise the implicit conversion to double precision in assignments introduces small numerical differences beyond the representational capabilities of single precision floats, e.g.,x=0.99may get converted tox=0.990000090...ifxis double precision. -
Modified a few format descriptors for printing table headings. Those format descriptors which involved strings and spanned several lines were not interpreted in the same way using the Intel and GNU Fortran compiler, i.e., printed strings had a different number of whitespace characters.
Modifications can also be traced by inspecting the commit log.
During the translation of this Fortran version to Python, statements have
been introducedd to write additional information to file debug.out for
debugging and ensuring the equivalence of the codes. The Fortran code
with these additional statements is stored in the debug branch and
those extra code blocks are enclosed by preprocessor directives #ifdef VERIFY ... #endif to facilitate enabling and disabling them.