The CADNA library
The following tar files include the complete CADNA
library. There are packaged with autotools. They contain a
README file, an INSTALL file, a user guide in the doc directory,
and examples with/without CADNA in the examples directory.
The CADNA library should work on any Linux or Unix system.
if you have any trouble with the installation, please send an e-mail
with the error message to the CADNA team.
License: CADNA is free software; you can redistribute it and/or modify it under
the terms of the GNU Lesser
General Public License as published by the Free Software
Foundation; either version 3 of the License, or (at your option) any
CADNA for C/C++ or Fortran codes:
cadna_c-3.1.11.tar.gz (21 Jul. 2021 version).
This package generates libraries that enable one to use CADNA in numerical codes
(sequential or parallel C/C++ codes with MPI and/or OpenMP,
sequential or parallel Fortran codes with MPI thanks to binding between Fortran and C++).
The configuration script supports clang, GNU compilers (g++, gfortran),
INTEL compilers (icpc, ifort), IBM compilers (xlc, xlf), and PGI compilers.
- cadna_c_half-3.1.11.tar.gz (21 Jul. 2021 version).
Same features. Can be used also to control
half precision computation in C/C++ codes:
- native half precision (tested on
ARM v8.2 processor)
- emulated half precision (thanks to the half
precision floating-point library developed by C. Rau available at http://half.sourceforge.net)
CADNA for C/C++ codes:
This version performs explicit changes of the rounding mode.
This version performs implicit changes of the rounding mode and benefits from inlining.
It can be used with sequential or OpenMP codes.
The configuration script supports g++ or icpc.
To use CADNA with OpenMP codes, a compiler that supports OpenMP reductions with stochastic variables is required.
This is possible since OpenMP 4.0.
CADNA for FORTRAN codes:
CADNA can be used to control accuracy in GPU codes written in CUDA. Feel free to contact the
CADNA team to get more information.
Tools related to CADNA