Multiprocessing Performance Analysis Using Numerical Integration Techniques

Weddle’s rule is one of the best for high-order numerical integration solving techniques. For better accuracy, more step sizes are used to compute the end result, but this leads to an increased number of computations. Parallel and multicore computations help us better enhance this computation. To maximize the performance of Weddle's rule on contemporary processors, this research will investigate the role of multiprocessing. Utilizing multiprocessing and parallel computing techniques can significantly enhance the efficiency, accuracy, and scalability of these computations. In theory, this should reduce the computation time of the integration process as the workload is divided among multiple processors, resulting in better performance. Despite the theoretical promise of reduced computation times through multiprocessing, practical applications often reveal a discrepancy between expected and actual performance gains. This study analyzes the time consumed in computation over the number of processors (1 to 12) and the number of subintervals used in computation during the execution of the integration process. Also, the machine learning technique Gradient Descent can be used to predict time consumption using features like time taken, input size, and number of processors. The model has a 95% accuracy.