Low-Cost Metaheuristics Optimization on Integrated GPUs: A Case Study on Jaya algorithm

This study investigates the computational and energy performance opportunities that arise from running metaheuristic algorithms on integrated GPUs, a class of low power GPUs found in consumer grade CPU chips. Leveraging OpenCL for parallelization, the paper presents an iGPU implementation of Jaya algorithm and compares its execution time, energy consumption, and optimization accuracy against a multicore CPU baseline. Experiments show that the iGPU implementation consistently outperforms the CPU counterpart in speed and energy efficiency for both single and double-precision arithmetic. Furthermore, the analysis takes into consideration specific architectural characteristics of iGPU devices to find execution sweet spots where performance and energy efficiency are maximized. Results confirm that iGPUs are a viable and efficient platform for lightweight metaheuristics in resource-constrained scenarios, offering practical advantages without compromising solution quality.