| Sumario: | The article discusses using high-performance computing (HPC) applications with field-programmable gate arrays (FPGAs) to maximize performance. The authors contend that while significant speedups are possible, to achieve it without expending a lot of development effort (while keeping flexibility, portability, and maintainability) is difficult. They tested supercomputers that use FPGAs with applications using 12 specific design methods. The authors believe that each method may be beneficial depending on what the programmer is trying to achieve, but that a broader approach does not yet work because the HPC/FPGA hardware is still in its commercial infancy. They think that eventually new programming models and idioms will be capable of encompassing a broader spectrum.
|
|---|