Weighted Least Squares (WLS) Library in C++

This project implements the kernels for performing WLS computations by taking advantage of LAPACK optimiations. This library is template-baed header-only; just include wls.hpp and you are ready to go!

Fortran name mangling and LAPACK integer customization

In wls.hpp, controling Fortran name mangling is done via a preprocessor macro WLS_FC, which takes values from one to six:

1. all lower cases without appending underscore,
2. all lower cases followed by a single underscore (default),
3. all lower cases followed by two underscores,
4. all upper cases without appending underscore,
5. all upper cases followed by a single underscore, and
6. all upper cases followed by two single underscores.

Example compiling command for appending a single understore and converting to all lower cases is

g++ -c -DWLS_FC=2 proj.cpp

Also, depending on LAPACK configurations, the size of integer type may vary. For instance, MATLAB comes with a customized installation with integer type of mwSignedIndex, which is ptrdiff_t (i.e., 64-bit signed integer on 64-bit machines). wls.hpp takes an optional macro of WLS_LAPACK_INT that allows the user to customize the integer type, e.g.,

g++ -c -DWLS_LAPACK_INT=mwSignedIndex proj.cpp

Notice that the default value of WLS_LAPACK_INT is int.

Copyright

Copyright (C) 2020 NumGeom Group at Stony Brook University

wls.hpp is distributed under GNU Lesser General Public License v3 (LGPLv3). Refer to LICENSE for more details.