5 papers
FlashAccel: Leveraging High-Bandwidth Flash (HBF) for High-Throughput LLM Inference
Xinyu Wang, Yalong Xue, Xiaotian Sun +5
Large language model (LLM) inference is increasingly limited by the capacity of High-Bandwidth Memory (HBM) in GPUs, as model weights and KV cache grow rapidly. High-Bandwidth Flas…
PIMCOMP: An End-to-End DNN Compiler for Processing-In-Memory Accelerators
Xiaotian Sun, Xinyu Wang, Wanqian Li +2
Various processing-in-memory (PIM) accelerators based on various devices, micro-architectures, and interfaces have been proposed to accelerate deep neural networks (DNNs). How to d…
PIMSYN: Synthesizing Processing-in-memory CNN Accelerators
Wanqian Li, Xiaotian Sun, Xinyu Wang +3
Processing-in-memory architectures have been regarded as a promising solution for CNN acceleration. Existing PIM accelerator designs rely heavily on the experience of experts and r…
PIMSIM-NN: An ISA-based Simulation Framework for Processing-in-Memory Accelerators
Xinyu Wang, Xiaotian Sun, Yinhe Han +1
Processing-in-memory (PIM) has shown extraordinary potential in accelerating neural networks. To evaluate the performance of PIM accelerators, we present an ISA-based simulation fr…
PIMCOMP: A Universal Compilation Framework for Crossbar-based PIM DNN Accelerators
Xiaotian Sun, Xinyu Wang, Wanqian Li +3
Crossbar-based PIM DNN accelerators can provide massively parallel in-situ operations. A specifically designed compiler is important to achieve high performance for a wide variety…