Deployment of GPUs in high-performance computing depends on the availability of efficient numerical algorithms running on this new generation of super computers. Over the past years a number of GPU-based libraries of numerical algorithms as well as software products for numerical analysis with dedicated GPU support have emerged.
Three factors specific to GPU computing necessitate continued research and development of numerical algorithms in this otherwise well-established field:
• The GPU hardware architecture and the GPU programming model are substantially different from previous generations of parallel computer architectures.
• The GPUs’ capabilities continue to be expanded from hardware ...