Author: Carsten Burstedde, Lucas C. Wilcox, Tobin Isaac et.al.

Copyright: GNU Lesser General Public License version 2.1 (or, at your option, any later version)

The sc library provides functions and data types that can be useful in scientific computing. It is part of the p4est project, which uses it extensively. Consequently, it is also a dependency of all projects that use p4est. Some of the more important features are the following:

The primary build system is based on the GNU autotools. Several macro files reside in the config/ subdirectory. These are used in sc but also available to other projects for convenience. It is possible to nest packages this way, conveniently via git sumbodules.
CMake support has been added and is still in development.
The library provides MPI wrappers named sc_mpi_*. With MPI enabled, they just call the corresponding MPI function. In case sc is configured without MPI (the conservative default), they implement initialize/finalize and collective calls as noops, which means that the sc_mpi_* calls do not have to be protected with the #ifdef SC_MPI construct in the code.
The library provides a logging framework that can be adapted by other packages. Multiple log levels are available, as well as options to output on just one or all MPI processes and to customize the style.
The library provides a set of data containers, such as dynamically resizable arrays and memory pools, which are thin and lightweight.

To build from a fresh checkout, call ./bootstrap first. Then proceed as described next (building from a tar distribution). The bootstrap script usually does not need to be re-executed!

To build the sc library from a tar distribution, use the standard procedure of the GNU autotools. The configure script is best called from a new empty directory with a relative path to the source. It takes the following options:

--enable-debug (recommended for development) lowers the log level for increased verbosity and activates the SC_ASSERT macro for consistency checks.
--enable-mpi (recommended in general) expects a working MPI build environment, pulls in the mpi.h include file and activates the MPI compiler wrappers. If this option is not given, trivial wrappers for MPI routines are activated, the simpler ones of which run fine, and the code is expected to be executed in serial only.

A typical development configure line looks as follows:

relative/path/to/configure CFLAGS="-Wall -O0 -g" --enable-mpi --enable-debug

A typical production configure line looks as follows:

relative/path/to/configure CFLAGS="-Wall -O2" --enable-mpi

To specify a target path for make install add a --prefix=<installdir> option.

The CMake build works as usual with that system. There are analogous options to enable the MPI compile and debug mode.

See also: https://www.p4est.org/; https://www.gnu.org/licenses/licenses.html