high performance computing on graphics processing units: hgpu.org

Computer clock speeds which had been increasing tremendously over years is now slowing down and has reached its limit of saturation. In order to overcome this saturation of the clock speed, aggressively pursuing optimizations techniques are being developed to get more work done in each clock cycle in favor of parallel computing and concurrent programming. The GPUs massive parallel computing is now evolving as significantly faster processor than any other multi-core processors.The measurement analysis algorithm also has to be modified in order to make use of this parallelism. This paper presents one such measurement analysis, transition localization, which basically computes the high to low and vice versa transitions of a stream of signals in parallel. Measuring signals have dependencies on their previous signals when computed serially. However, now there is a parallel solution developed on the GPU which not only makes this algorithm efficient but also faster than any multi-core processors.