Dynamic Profiling Methodology for Resource Optimization in Heterogeneous Computing System
Mahendra Vucha, Rajendra Patel, Arvind Rajawat · 2013
Embedded Systems combine one or more processing cores to support dedicated logic running on an ASIC or FPGA and meets design goals with optimized resource utilization. Optimization of resource utilization can be achieved by estimating the requirement of an application with a variety of aspects like performance, memory usage, resource usage, cache hit versus cache misses, energy consumption, etc. Out of these, estimation of resources is more important to enhance the execution speed of an application with minimum resource utili- zation. Since ever increasing system and application complexities, it becomes quite necessary to estimate the resources of Computing Systems for an application. This paper addresses a profiling methodology for resource optimization in Heterogeneous Systems. The addressed profiling methodology estimate the resources required for an application execution and prepares a resource utilization chart for the Heterogeneous Systems. The profiling methodology also estimates the performance of an application i.e. speed enhancement based on the attributes of an application and resources of Heterogeneous Systems. Here, the Heterogeneous System (HS) is a computing platform which contains an array of Programmable Logic Devices such as FPGAs in combination with a General Purpose Processor as Processing Elements (PEs). The behavior of profiling methodology for resource optimization is described by using SystemC language to support Heterogeneous Systems i.e. hardware software co-design.