Loop parallelization in source code for internet of things computing using hybrid heuristic algorithm

Area of Science:

• Computer Science
• Parallel Computing
• Internet of Things (IoT)

Background:

• Task scheduling in High-Performance Computing (HPC) and IoT systems faces challenges due to sequential loop execution limiting parallelism.
• Efficiently parallelizing nested loops is crucial for optimizing performance in heterogeneous IoT environments.

Purpose of the Study:

• To propose a hybrid approach, PSOALS, for dynamic parallelization of nested loops in heterogeneous IoT environments.
• To enhance makespan, resource utilization, and scalability by minimizing communication overhead in task scheduling.

Main Methods:

• PSOALS combines Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and wave-angle scheduling.
• Nested loops are modeled as two-dimensional iteration spaces, with loop iterations encoded as particles.
• A dependency-aware fitness function and mutation/diversity techniques are employed for solution optimization.

Main Results:

• PSOALS demonstrates superior performance compared to block-based, cyclic, and GA-based scheduling methods.
• The proposed method shows improvements in convergence speed, stability, and execution time across various IoT configurations.
• Wave-angle scheduling balances load and communication delays, enhancing task execution order.

Conclusions:

• PSOALS offers a scalable and adaptive solution for efficient loop parallelization and dependency management in IoT.
• The approach addresses future IoT challenges including real-time processing, energy efficiency, and large-scale deployment.
• Dynamic scheduling and optimized task execution order are key to improved IoT system performance.