In addition to the libraries required by the various programming languages, we provide several groups of independent libraries for parallel programming, data management, mathematical operations, plotting software and profiling.
The main parallel programming paradigm that is used on TIGRESS is MPI. We strongly recommend OpenMPI. While we do have the MPICH2 version as well, this is deprecated. All of our clusters have Infiniband connections and MPICH2 does not support them. There is a variant of MPICH2, called MVAPICH2 which does support Infiniband, but it is substantially less stable than OpenMPI.
- Intel's MKL
Cluster head nodes have the oprofile package. Intel's Vtune package, which is installed on an on-demand basis, can also be used for profiling as well as debugging. Additionally, the tau profiler is available on an on-demand basis.
|BLAS||Basic Linear Algebra||http://www.netlib.org/blas/|
|LAPACK||Linear Algebra Package||http://www.netlib.org/lapack/|
|FFTW2||Fast Fourier transform||http://www.fftw.org/|
|FFTW3||Fast Fourier transform||http://www.fftw.org/|
|Intel's MKL||Math Kernel library||http://software.intel.com/en-us/articles/intel-math-kernel-library-documentation/|
|Intel's Vtune||Execution profiling||http://software.intel.com/en-us/articles/intel-vtune-performance-analyzer-for-linux-documentation/|