Standard language compilers, software libraries, debugging and performance tools are essential components of any high performance computing resource. The languages that must be available are FORTRAN-90, C, C++ and Assembler. These should have fully integrated debug and performance analysis tools. The computational science community as a whole will build applications that use one or more of these languages, therefore, language interoperability is also important.
For MPP supercomputers it is important to support third party applications and libraries for data management and performance measurement. For example, the UPSHOT performance tools (from Argonne National Laboratory) should be available since they are in use at many DOE sites. Parallel software development and environment tools such as HPF, MPI, PVM, and the Global Array tools are used by a variety of ER computational scientists and access to these tools on the central NERSC MPP resources are essential. Furthermore, software libraries such as ScaLAPack and vendor specific parallel linear algebra tools are required for the development of many scientific applications.