Research On Approximate Communication For Energy-efficient Network-on-chip

With the advancement of semiconductor technology and computer architecture design,more and more cores are integrated into one chip.Now the number of cores of a multi-core processor has become an important indicator of the computing performance of the processor.However,in the face of the ever-increasing number of cores,the interconnection scheme between cores has become one of the main factors affecting the overall energy efficiency of the chip.Compared with the bus structure,the network on chip(NoC)has better bandwidth,lower latency and higher scalability,thus becoming the mainstream on-chip interconnection solution for chip multiprocessors(CMP).The trend of energy-efficient design in massively parallel multicore systems and the fault tolerance of applications,with the advent of the big data era,have brought upon alternative solutions based on approximate computing.Approximate computing is being touted as a viable solution for energy-efficient architecture design by relaxing the accuracy constraints of applications,and it has achieved significant energy efficiency gains in the computing unit and storage unit.With the increasing communication demand as well as the optimization bottleneck of NoC performance and energy consumption,the approximate communication design has become an important research issue for energy-efficient NoC.This paper,with a focus on the approximate communication design for energyefficient NoC,mainly conducts exploratory research in the following four aspects:First,a dynamic traffic regulation scheme is proposed for approximate communication of NoC.Network congestion is one of the main factors that affect transmission delay,and different traffic flows have different impacts on network congestion.This method designs a approximation based traffic regulation structure in the network interface,which reduces the amount of injected data through data approximation,and can regulate the injection rate of each node.In addition,it also designs a dynamic traffic regulation algorithm to dynamically adjust the injection rate of each node according to the impact of traffic flow on network congestion.Thus,it improves the NoC performance.Based on the PARSEC benchmark experiments,the results show that this method can reduce the average transmission delay up to 44.7% within an application error of 8%,increase the network bandwidth up to 40%,and achieve an average application acceleration of 1.12 times.Second,an NoC performance optimization method based on dynamic lossy compression is proposed.Network workload is one of the factors that affect network performance,and data compression is an effective way to reduce network workload.However,the existing lossy compression and lossless compression designs use static compression encoding scheme,ignoring the difference in the impact of data on network performance.In order to improve the NoC performance optimization benefits,this method proposes a dynamic lossy compression design which encodes the non-approximate data losslessly and encodes the approximable data lossy.Furthermore,it also dynamically adjusts the error threshold of lossy compression to optimizate network performance under the transmission quality limitation.Based on the PARSEC benchmark experiments,the results show that compared with the lossy compression scheme,this method has a lower compression rate and lower transmission delay under the same transmission quality,and it improves the NoC throughput up to 37.5% compared with to lossless compression scheme,while keeping low application error.Third,a performance optimization method for bufferless NoC based on approximate communication is proposed.By removing the buffers,The bufferless NoC reduces power consumption and area overhead,but also leads to an increase in transmission delay and a decrease in network throughput.Through the performance analysis of the bufferles NoC,in the retransmission based bufferless NoC,packet retransmission is a key factor affecting the NoC performance.In order to improve the performance of the bufferless NoC,this method designs a new bufferless NoC architecture,which reduces packet retransmission through lossy transmission and improves the NoC performance.Moreover,it also proposes a packet approximate codec design to approximate the missing data.Thus,This method improves the performance of bufferless NoC with extremely low quality loss.Based on the PARSEC benchmark experiments,the results show that compared with the existing bufferless NoC,this design reduces the retransmission by 83.6%,reduces the transmission delay by 46.7%,increases the network throughput by 92%,and achieves application acceleration 1.2 times,while maintaining low application error.Fourth,an NoC energy optimization method based on multiplane network design is proposed.The NoC performance optimization usually leads to an increase in area overhead and affects the energy consumption of NoC.In order to reduce the energy consumption of NoC,this method designs a two-plane network structure which includes a lossy subnetwork and a lossless subnetwork.Based on lossy transmission,the lossy subnetwork realizes a lightweight,low-delay,bufferless architecture design.In addition,based on the multiplane transmission design,this method speeds up part of the data transfer and achieves transmission quality control.Thus,this method improves NoC performance while reducing NoC area overhead and power consumption.Based on the PARSEC benchmark experiments,the results show that compared with the single-plane NoC,this method reduces the transmission delay by 42.2%,and saves 37.8% of the NoC area overhead and 28.9% of the NoC power consumption under the same throughput.