Optimizations for ARM11 MPCore on Computational Capabilities and Interrupt Processing

In this contribution, several optimization strategies for fully exploiting the performance of ARM11 MPcore have been discussed. Experimental studies are conducted for the basic computation performance and the multimedia processing performance of ARM11 MPcore. The basic computation operation includes integer operation and floating-point operation. Results show that an efficiency gain of a factor 3.8 for integer operation could be achieved when being optimized by using the OpenMP parallel computing model, and for floating-point operation, the value is 3 to 5 when adopting the VFP (vector float point) optimization method. With regard to the multimedia processing, two optimization methods are put forward, namely, the DCT optimization with VFP and the parallelization of audio decoding, video decoding and playing. The experimental results show that both methods could efficiently improve the performance of multimedia processing for ARM11 MPCore. In addition, considering that the interrupt load imbalance on multicore processor may bottleneck the entire system's performance improvement, workload balancing methods basing on interrupt affinity are proposed for ARM11 MPcore, and results indicate these methods are effective.