Analyzing and Implementing Graph Coloring Algorithm on GPU

• Main Authors: Thiagarajan, Sindhu, 辛度
• Other Authors: Chou, Jerry Chi Yuan
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/92829994044280848164

碩士 === 國立清華大學 === 資訊工程學系 === 103 === GPU is a multithreaded multicore unit used for high speed parallel processing of huge amounts of data. Data has to be divided among the threads for independent working. If the threads have to coordinate, threads have to be synchronized to share the information with each other. Graph problems can be a challenge to be solved in parallel as firstly it has to be partitioned evenly for a proper load balancing. After which results may conflict between partitions which have to be solved. Graph coloring is one of the problems to reflect this problem clearly. This research aims to color the graph by reducing the number of iterations taken to resolve the conflicts by coloring the nodes which will cause the conflict first. A Sequential coloring algorithm and a Parallel coloring algorithm are used.