Abstract—Heterogeneous computing Systems-on-Chip (SoCs)
have emerged as powerful platforms integrating diverse process-
ing units such as CPUs, GPUs, and FPGAs to enhance com-
putational capabilities and efficiency. However, the complexity
of managing these resources necessitates advanced optimiza-
tion techniques, particularly with runtime parameter adaptation
to handle dynamic workloads and resource availability. This
survey comprehensively reviews the state-of-the-art methods in
optimizing heterogeneous computational resources through real-
time parameter tuning. This survey is structured to cover key
themes including power management, energy efficiency, system-
level optimization, and performance improvement. Each section
discusses various methodologies, highlighting their strengths and
weaknesses. Our critical analysis identifies current gaps in the
literature and suggests potential research directions. We empha-
size the integration of evolutionary algorithms with reinforcement
learning (ERL) as a promising approach to dynamically adapt
system parameters, thereby achieving optimal resource utilization
and performance. Key contributions of this survey include: A
comprehensive review of dynamic power management techniques
and AI-based solutions for heterogeneous SoCs. In-depth analysis
of energy efficiency strategies and their impact on system
performance. Detailed examination of system-level optimization
methods, including collaborative scheduling and edge computing
frameworks. Evaluation of performance improvement techniques
and their implications on computational efficiency. Identification
of open challenges and future research directions in the field of
computational resource optimization. This survey aims to provide
researchers and practitioners with a holistic understanding of
the current landscape and future prospects in optimizing hetero-
geneous computational resources through real-time parameter
tuning.
Index Terms—Heterogeneous Computing, Systems-on-Chips,
Optimization, CPU, GPU, FPGA
