.png){kind=icon}
.png){kind=icon}
Return
Optimizing Standard Convolution for Diverse Precision on DCU
.png){kind=icon}
PRE
AI
DOI：10.1007/s42514-025-00253-y.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Standard convolution remains a major performance bottleneck in modern deep neural networks. Although existing optimization libraries demonstrate effectiveness, they often underutilize key architectural features of emerging accelerators like DCUs, leading to suboptimal performance. To address this limitation, we propose a holistic, architecture-aware framework that systematically co-optimizes memory hierarchy and computational pipelines. The framework dynamically adapts to convolution parameters for maximal hardware utilization, with core contributions including: an innovative memory management strategy mitigating access conflicts, an adaptive computation pipeline balancing parallelism and data reuse, and a method bypassing API limitations to leverage underlying hardware instructions. On DCU hardware, our framework achieves significant speedups over MIOpen - delivering \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$3.09\times $$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.64\times $$\end{document} average acceleration for FP16 and FP32 precision respectively, while reducing end-to-end training time for ResNet and EfficientNet by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$6.7\%$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$12.1\%$$\end{document}.
Keywords:
Convolution
Implicit GEMM
DCU
Performance optimization
Memory optimization
Journal
C
IF:
1.9
Papers:
38
Citations:
253