arXiv cs.AI· Leandro Fiorin, Luigi Altamura, Cristina Silvano·· 12 小时前AI 评分18
qGAMMA：面向多精度加速器的最优 DNN 映射框架
qGAMMA: A Framework for Optimal DNN Mapping on Multi-Precision Accelerators
AI 导读
qGAMMA 是一个用于探索和优化混合精度 DNN 模型在多精度空间加速器上映射的框架，通过领域特定遗传算法与基于硬件综合结果的解析性能模型搜索设计空间，找出使目标指标最优的映射与加速器配置。该框架在多种 DNN 负载上评估，可为混合精度应用设计提供性能、能效与资源利用率之间的权衡参考。论文已被 SAMOS XXV 2025 接收，将发表于 Springer LNCS。
正文 · 原文
Abstract:This work describes qGAMMA, a flexible framework for exploring and optimizing the mapping of mixed-precision Deep Neural Network (DNN) models onto general-purpose spatial accelerators that support a range of fixed- and floating-point precisions. qGAMMA navigates the design space using a domain-specific genetic algorithm and an analytical performance model based on hardware synthesis results. The framework identifies the optimal mapping and accelerator configuration that minimizes target metrics. Extensive evaluations across DNN workloads show that qGAMMA offers valuable insights for designing mixed-precision applications, enabling better trade-offs between performance, energy efficiency, and resource utilization.
| Comments: | Accepted at the 25th International Conference on Embedded Computer Systems: Architectures, Modeling, and Simulation (SAMOS XXV), 2025. To appear in Springer LNCS |
| Subjects: | Hardware Architecture (cs.AR); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.23222 [cs.AR] |
| (or arXiv:2610.23222v1 [cs.AR] for this version) | |
| https://doi.org/10.48550/arXiv.2610.23222 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Luigi Altamura [view email]
[v1]
Fri, 16 Oct 2026 09:54:13 UTC (1,072 KB)
来源：arXiv cs.AI · arxiv.org