Why now?
Until now, FPGAs have been too small in terms of capacity and too expensive to make a supercomputer system viable.
However, recently, FPGAs have been improving in capability much faster than microprocessors - a trend which is
expected to continue. As a result, FPGAs are becoming widely adopted in mainstream embedded computing to build
complex mission-critical systems. Their attractiveness to the high-performance computing sector has at the same
time been growing steadily.
For the past decade, supercomputers have been built by combining many conventional microprocessors to work together on the same application. However, as organisations become dependent on increasingly sophisticated applications with ever-growing volumes of data, clusters of conventional microprocessors are struggling to provide the required processing power.
The FHPCA is studying how to provide this power by building highly scalable, easy-to-program computers using arrays of FPGAs.
The Parallel Toolkit
The following presentation provides a high-level introduction to the concepts behind the FHPCA's Parallel Toolkit
approach to programming Maxwell, the 64 node FPGA supercomputer:
