Back close

Gpu Hierarchy and Memory Optimization Techniques: Focusing on Shared Memory Usage

Publication Type : Conference Paper

Publisher : IEEE

Source : 2026 International Conference on Computing Theory and Wireless Communications (ICCTWC)

Url : https://doi.org/10.1109/icctwc68241.2026.11583329

Campus : Amritapuri

School : School of Engineering

Center : Humanitarian Technology (HuT) Labs

Department : Electronics and Communication

Year : 2026

Abstract : GPUs are powerful accelerators, which have been widely used in deep learning, scientific computing, and large-scale simulations due to due to their ability to handle massive parallel workloads. However, achieving high performance depends on how well the The GPU memory hierarchy is managed. This paper discusses detailed overview of the various levels of GPU memory: global, constant, register, and especially shared memory-and explains why shared memory plays an essential role in reducing latency and improving throughput. We reviewed recent techniques that make Shared-memory usage, more efficient, which includes optimized data. placement, automatic memory allocation strategies, compiler- Assisted tuning and performanceprediction models. We also Compared shared memory to other memory optimization techniques to provide details on when and why it offers advantages in terms of speed and data. reused. Additionally, we consider practical means of reducing data access delays, avoiding memory bank conflicts, and maxi- Maximizing data processed per kernel by tiling and cooperative thread loading. These optimization strategies are essential for performancecritical tasks such as training deep neural networks or running high-resolution simulations. Following insights from Based on the results of the existing research, we propose ideas for future improvements, such as as dynamic shared-memory allocation at runtime and tailoring memory use to specific workload patterns. Overall, The present work wishes to bring clear orientation on proper shared- memory management, helping developers better exploit the GPU.

Cite this Research Publication : Akash Moorkkanattumana Madhu, Arjun Krishna Muraleedharan, Rajesh Kannan Megalingam, Gpu Hierarchy and Memory Optimization Techniques: Focusing on Shared Memory Usage, 2026 International Conference on Computing Theory and Wireless Communications (ICCTWC), IEEE, 2026, https://doi.org/10.1109/icctwc68241.2026.11583329

Admissions Apply Now