Abstract
With rapidly increasing demand on high-performance and low-power computing applications, multicore platforms have become popular in information and communication devices. To improve the CPU utilization for multicore platform, we propose a scheme to combine system information and compiler directives for programmer to dynamically balance the CPU load for parallel programs. Our procedure is (1) obtain the CPU usage information, (2) incorporate the CPU usage information in the programs, and (3) dynamically balance the load. Our experimental results in Duo-Core platform show that performance improvement can be obtained using the proposed scheme.