BLAS and LAPACK provide high performance dense matrix and vector routines. Atlas provides a portable, optimized implementation of the BLAS. BLAS and LAPACK are required by nearly all Trilinos packages.
The Message Passing Interface (MPI) is a standard that supports distributed memory parallel computing. Several vendors provide MPI implementations. If an MPI implementation is available, MPI can be turned on in Trilinos using --enable-mpi.
Expat is an XML parser. Teuchos can optionally use Expat. Expat support can be enabled using --enable-teuchos-expat.
PyTrilinos is a set of Python modules that make most of Trilinos' functionalities available in a Python script. PyTrilinos can be used in serial, or with MPI. In this latter case, a standard, out-of-the-box python interpreted is used. PyTrilinos requires SWIG, the Simplified Wrapper and Interface Generator.
METIS and ParMETIS are a family of graph partitioning algorithsm. ML can use both METIS and ParMETIS to create aggregates containing the number of nodes specified by the user. ML can optionally use METIS and ParMETIS. METIS support can be enabled using --with-ml_metis, while ParMETIS support can be enabled using --with-ml_parmetis3x.
Amesos is a Trilinos package that provides interfaces to a large number of popular direct solvers. The Amesos Homepage links to the direct solvers that Amesos supports.
EpetraExt contains, among others, a class to read and write distributed linear algebra objects to and from the parallel binary HDF5 library. This is done with the class EpetraExt::HDF5, whose compilation is enabled by adding --enable-epetraext-hdf5 to the configure line.
Y12m is an older library for performing certain sparse matrix operations. AztecOO links to y12m when the --enable-aztecoo-azlu option is provided at configure time.
GNU Make is required when using the --with-gnumake option, which removes redundancies in the link line that is generated automatically by the makefile.export system.
