The implementation of parallel computation on CPU and GPU

Abstract

High performance computations have never been failed to attract people since the first computer was created. Before using parallel computations, people tried to increase the performance of single core processor. But the large power consumption and heat releasing became to the biggest performance limit for single core processor. Therefore, the technology of parallel programming was introduced couple years ago and replaced the role of single core processor. Parallel computation is the algorithm that control multiple operations simultaneously. It can be implement not only on multiple processor device, but also implement on single processor. This study is about researching implementation of parallel computing with C. After introducing the basic idea and structures of parallel computation, the implementation of parallel computation will be created with tree algorithm. On the single processor like CPU, parallel computation is applied with pthread function. And for the multiple processor like GPU, CUDA is the best choice for us. Meanwhile, the serial function is also needed for testing the correctness of parallel computing functions and performance analysis.