Scholarlink

<p>General matrix multiplication is a vital operation in high-performance computing and has wide applications in areas such as computational fluid dynamics and deep learning (DL). While there are many optimization techniques available for large matrix multiplications on CPUs and GPUs, handling batches of small matrix operations requires innovative solutions. Digital Signal Processors (DSPs) offer a promising alternative for processing DL workloads; however, the architectural differences between DSPs and conventional processors like CPUs and GPUs necessitate the development of specialized optimization strategies. This paper introduces <span>mtSMM</span>, an optimization approach tailored for small matrix multiplications on multi-core DSPs. Our approach focuses on the batch-as-vector paradigm, efficient on-chip memory management, and a well-designed micro-kernel. By maximizing computational resources, optimizing instruction-level and thread-level parallelism, and enhancing memory access patterns, our approach significantly improves performance. Experimental results on the FT-M7032 DSP demonstrate that our method can achieve up to 83% of the theoretical peak performance of the hardware, significantly outperforming current state-of-the-art methods for batches of small matrices.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing small matrix multiplications via batch grouping on multi-core DSPs

  • 期刊论文

  • 发表日期：

    2025年10月28日

查看全文
  • Xiaotian Chen,
  • Pengyu Wang,
  • Jianbin Fang,
  • Peng Zhang,
  • Chun Huang

摘要

General matrix multiplication is a vital operation in high-performance computing and has wide applications in areas such as computational fluid dynamics and deep learning (DL). While there are many optimization techniques available for large matrix multiplications on CPUs and GPUs, handling batches of small matrix operations requires innovative solutions. Digital Signal Processors (DSPs) offer a promising alternative for processing DL workloads; however, the architectural differences between DSPs and conventional processors like CPUs and GPUs necessitate the development of specialized optimization strategies. This paper introduces mtSMM, an optimization approach tailored for small matrix multiplications on multi-core DSPs. Our approach focuses on the batch-as-vector paradigm, efficient on-chip memory management, and a well-designed micro-kernel. By maximizing computational resources, optimizing instruction-level and thread-level parallelism, and enhancing memory access patterns, our approach significantly improves performance. Experimental results on the FT-M7032 DSP demonstrate that our method can achieve up to 83% of the theoretical peak performance of the hardware, significantly outperforming current state-of-the-art methods for batches of small matrices.