The field of High-Performance Computing (HPC) and Artificial Intelligence (AI) is witnessing significant developments in energy efficiency and computing-in-memory (CIM) technologies. Researchers are exploring innovative approaches to reduce energy consumption and improve performance in HPC systems, including the use of energy-efficient processors, novel system architectures, and advanced scheduling policies. Additionally, CIM technologies are being developed to accelerate AI workloads by computing directly within memory arrays, reducing data movement and energy consumption. Notable advancements include the development of scalable neural network accelerators, end-to-end CIM accelerators, and compression techniques to enable larger models to be executed within on-chip memory constraints. Noteworthy papers in this area include:
- Register Dispersion, which proposes a compact Vector Register File design to reduce area and power consumption in low-cost processors.
- CIMPool, which introduces a CIM-aware compression and acceleration framework to enable significantly larger neural networks to be accommodated within on-chip memory constraints.
- CIMR-V, which presents an end-to-end CIM accelerator with RISC-V that incorporates CIM layer fusion, convolution/max pooling pipeline, and weight fusion, resulting in reduced latency and improved energy efficiency.