high performance computing on graphics processing units: hgpu.org

To monitor bad information spreading in microblog system, large-scale data from microblog must be processed in real time. This needs high cost-effective parallel schemes. A parallel processing method on GPUs was put forward to monitor massive microblog. The proposed scheme can fully exploit the GPU feature to schedule massive threads for data-intensive tasks. The detailed parallel algorithms were provided. The key technologies including memory optimization and dynamic thread configuration were solved. The experiment results show that this parallel scheme can obtain the average speed-up ratio of about 16 and can process more than 500MB online data of microblog per second in real time.