GPU Acceleration Of Key Algorithms In Blockchain

With the increasing complexity of general-purpose computing and increasing data size,the way to improve the processing efficiency by using traditional multi-core processors and high-performance CPUs has increasingly failed to meet the needs of scientific computing.Therefore,it is urgent to use numerical computations to efficiently process data,Thereby improving the efficiency of the application.Blockchain is a brand new decentralized infrastructure and distributed computing paradigm that has gradually emerged with the increasing popularity of digital cryptocurrencies,such as bitcoin,which has drawn great attention from government agencies,financial institutions,technology companies and capital markets With wide attention.The SHA-256 algorithm,as a reliable encryption algorithm nowadays,has been fully used in the blockchain technology and is the most crucial algorithm.At present,the development of GPU parallel computing technology is quite mature,with strong floating-point computing power and good programmability.The adoption of CPU-GPU heterogeneous architecture has become one of the trends of high performance computer systems.Based on the heterogeneous computing environment of CPU-GPU,this paper uses the CUDA platform provided by NVIDIA to develop the parallel processing of SHA-256 and Keccak algorithms,and studies the key technologies of data stream optimization in heterogeneous computing.Dataflow processing model framework,elaborating the application of the implementation process and data flow specific processing flow.Optimize data stream processing with the Hyper feature in the Kepler architecture and reduce access latency with on-device texture memory and light memory to further enhance application processing efficiency.The data dependencies appearing in the process of data flow are described and corresponding solutions are provided.In this paper,SHA-256 and Keccak algorithms are applied to the architecture of CPU-GPU heterogeneous architecture,GPU-based data parallel computing algorithm to optimize the data flow in the process and improve the calculation efficiency of the program,and finally get some acceleration,Has achieved the anticipated goal,has certain promotion significance to the GPU research in other fields.