Improve Hmc Based Graph Processing By Adding Compressdecompress Unit

Graphs play an important role in various practical application areas from social science tornmachine learning. However, due to the irregular data access pattern of graph computation,rnthere is a major challenge in graph processing.rnThe emergence of the technology called Hybrid memory cube(HMC) has helped graph processingrnaccelerators to overcome this issue. This hardware provides e cient bandwidth tornthe graph computation, however, the communication tra c between memory cubes limitsrnthe performance. To overcome this issue we proposed a new approach for HMCs basedrnaccelerators by adding a packet compression/ decompression unit. We used Message Fussionrnand Tesseract as our baseline system. In our approach, the data sent between thernmemory cubes will be compressed before being sent into the network. From the experimentalrnresult, the proposed approach showed 1.7x performance improvement on averagernover the baseline systems. In addition, the energy consumption by the transmission of thernnetwork is reduced by 47.28% over the baseline system and the compressor/decompressorrnunit takes 25% of the total area.

Get Full Work

Report copyright infringement or plagiarism