基于TMS320C80的BP算法并行实现的研究

Parallel Implementation Research of BP Algorithms Based on TMS 320C80

结合 TMS32 0 C80的编程结构 ,利用神经网络的权并行性和神经元并行性 ,提出了BP算法的并行实现方法 ,并解决了存储器访问冲突、数据双缓冲传输、PP细粒度并行和死锁等问题。结果表明 ,该并行 BP算法十分有效。

Combining TMS320C80's programmable structure,the paper realized parallel BP algorithms by using Weight parallelism and neuron pa ralleism and solved many questions such as Double-Buffer model,memory accessing interfere,PP's tiny granularity parallelism and deadlock.The result shows that t he parallel BP algorithms is effective.