Issues on Building an MPI Cluster on Microblaze

MPI is the traditional paradigm to parallelize applications for High Performance Computing environments. AzequiaMPI is an implementation of the MPI-1.3 standard. Its thread-based architecture enables it to run on high-end HPC machines as well as on embedded environments as soft-core processor in FPGAs. This article describes the experience of building a maintainable cluster of fourteen popular Xilinx XUP2-Pro FPGA development boards around the PetaLinux/ Micro blaze base. Figures on the AzequiaMPI performance are given.