The Research Of Heterogeneous Platform Based LHS Algorithm For Simulation Of Electrocardiogram

ECG simulation is an important topic of cardiac bioelectric phenomena studied, with a high scientific and practical value. With the development of computer technology, ECG simulation with computer has become an important means of ECG research field. However, due to the computationally intensive of ECG simulation, complex of simulation models, supercomputers often be used, but these supercomputers are expensive, bulky, difficult to use and maintain. If ECG simulation algorithm is accelerated greatly, we may even use a general personal computer to calculating ECG simulation, it will undoubtedly be a huge role in promoting diagnosis and treatment of heart disease.Aimed at the problem of computing power is currently experiencing ECG simulation insufficiency, I focus on the ECG simulation of parallel computing. Through analyzing and studying the ECG simulation model and serial ECG simulation algorithm, this paper attempts to construct a multi-core CPU and GPU based on the heterogeneous computing platform to accelerate the ECG simulation calculation. In order to fully use of the calculation ability of heterogeneous platform, this paper also proposes a scheduling algorithm to allocate tasks reasonable. The main work of this paper is as follows:The ECG simulation algorithm based on GPU. In this paper, through the experiment analysis of the ECG simulation serial algorithm, the ECG simulation calculation can be divided into three parts, simulation of excitation propagation, simulation calculation of potential, the treatment of calculation results of the calculation. In the three parts, ECG simulation of potential is the key part of parallel computing. Using the Amdahl law of accelerated ratio analysis, this paper choose to use NVIDIA CUDA as software and GPU as hardware respectively for the realization of an ECG simulation algorithm based on GPU.The Load-prediction based Hybrid Scheduling Algorithm, called LHS algorithm. In order to make full use of the multi-core CPU, Open MP is used to implement the parallelism of CPU.LHS algorithm was proposed in order to allocate tasks of the ECG simulation reasonably to components with different properties in heterogeneous. With LHS algorithm, computing tasks can be effectively allocated according to the characteristics of the CPU and GPU, make use of idle CPU when GPU compute the potential of heart, reduce the data transmission of GPU which waste the power of GPU, improve the comprehensive utilization rate of CPU and GPU, thus further improving the overall performance of the system.The experimental results show that, the proposed ECG simulation algorithm based on GPU achieves a speedup of 16.83 compared with serial algorithm of ECG simulation, which achieved the desire effect of acceleration. The computing ability of the system improves further after using the LHS algorithm, and the result is better compared with general dynamic scheduling algorithm.