A hardware/software co-synthesis method is presented for SoC designs consisting of both hardware IP cores and software components on a graph-theoretic formulation. Given a SoC integrated with a set of functions and a ...A hardware/software co-synthesis method is presented for SoC designs consisting of both hardware IP cores and software components on a graph-theoretic formulation. Given a SoC integrated with a set of functions and a set of performance factors, a core for each function is selected from a set of alternative IP cores and software components, and optimal partitions is found in a way to evenly balance the performance factors and to ultimately reduce the overall cost, size, power consumption and runtime of the core-based SoC. The algorithm formulates IP cores and components into the corresponding mathematical models, presents a graph-theoretic model for finding the optimal partitions of SoC design and transforms SoC hardware/software co-synthesis problem into finding optimal paths in a weighted, directed graph. Overcoming the three main deficiencies of the traditional methods, this method can work automatically, evaluate more performance factors at the same time and meet the particularity of SoC designs. At last, the approach is illustrated that is practical and effective through partitioning a practical system.展开更多
基于软件实现的多核系统模拟器执行计算密集/数据密集任务的时效性极差,且存在模拟精度和性能评估准确性差的不足,限制其在多核系统结构优化探索中的应用。文章提出一种周期精确的软硬件协同多核系统模拟器(cycle accurate hardware-sof...基于软件实现的多核系统模拟器执行计算密集/数据密集任务的时效性极差,且存在模拟精度和性能评估准确性差的不足,限制其在多核系统结构优化探索中的应用。文章提出一种周期精确的软硬件协同多核系统模拟器(cycle accurate hardware-software co-simulator,CAHSCS),通过在传统模拟器架构中引入硬件计算和存储模块,CAHSCS能有效改善全系统的模拟速度、精度,提高性能评估的准确性。复杂真实任务加载实验结果表明,CAHSCS将大规模复杂数据的运算效率提高了10倍,显著加快了系统设计收敛速度。展开更多
人工智能(AI)和高性能计算(HPC)是计算机领域的两大重要技术。随着计算机技术的飞速发展,二者的联系逐渐紧密,并呈现出互相依赖、互相促进的关系。一方面,高性能计算系统面临的各种新问题与新挑战,需要人工智能方法技术辅助解决(AI for ...人工智能(AI)和高性能计算(HPC)是计算机领域的两大重要技术。随着计算机技术的飞速发展,二者的联系逐渐紧密,并呈现出互相依赖、互相促进的关系。一方面,高性能计算系统面临的各种新问题与新挑战,需要人工智能方法技术辅助解决(AI for HPC);另一方面,人工智能领域理论的突破,依赖于HPC提供的强大的计算能力(HPC for AI)。在这样的背景下,AI和HPC两领域交叉融合,深度发展。文中系统回顾了近年来AI和HPC两个领域各自技术的发展脉络,着重从以下几方面展开分析:1)AI技术在解决HPC硬件体系结构、操作系统资源管理、编译优化和软件开发等几个方面问题的贡献;2)HPC为AI在硬件基础设施及软件应用上的支持;3)AI和HPC领域融合的未来发展前景与挑战。展开更多
基金This project was supported by the Defense Pre-Research Project of the ‘Tenth Five-Year-Plan’ of China(41315040106) and the National"863"High Technology Research and Development Programof China (2003AAIZ2210)
文摘A hardware/software co-synthesis method is presented for SoC designs consisting of both hardware IP cores and software components on a graph-theoretic formulation. Given a SoC integrated with a set of functions and a set of performance factors, a core for each function is selected from a set of alternative IP cores and software components, and optimal partitions is found in a way to evenly balance the performance factors and to ultimately reduce the overall cost, size, power consumption and runtime of the core-based SoC. The algorithm formulates IP cores and components into the corresponding mathematical models, presents a graph-theoretic model for finding the optimal partitions of SoC design and transforms SoC hardware/software co-synthesis problem into finding optimal paths in a weighted, directed graph. Overcoming the three main deficiencies of the traditional methods, this method can work automatically, evaluate more performance factors at the same time and meet the particularity of SoC designs. At last, the approach is illustrated that is practical and effective through partitioning a practical system.
文摘基于软件实现的多核系统模拟器执行计算密集/数据密集任务的时效性极差,且存在模拟精度和性能评估准确性差的不足,限制其在多核系统结构优化探索中的应用。文章提出一种周期精确的软硬件协同多核系统模拟器(cycle accurate hardware-software co-simulator,CAHSCS),通过在传统模拟器架构中引入硬件计算和存储模块,CAHSCS能有效改善全系统的模拟速度、精度,提高性能评估的准确性。复杂真实任务加载实验结果表明,CAHSCS将大规模复杂数据的运算效率提高了10倍,显著加快了系统设计收敛速度。
文摘人工智能(AI)和高性能计算(HPC)是计算机领域的两大重要技术。随着计算机技术的飞速发展,二者的联系逐渐紧密,并呈现出互相依赖、互相促进的关系。一方面,高性能计算系统面临的各种新问题与新挑战,需要人工智能方法技术辅助解决(AI for HPC);另一方面,人工智能领域理论的突破,依赖于HPC提供的强大的计算能力(HPC for AI)。在这样的背景下,AI和HPC两领域交叉融合,深度发展。文中系统回顾了近年来AI和HPC两个领域各自技术的发展脉络,着重从以下几方面展开分析:1)AI技术在解决HPC硬件体系结构、操作系统资源管理、编译优化和软件开发等几个方面问题的贡献;2)HPC为AI在硬件基础设施及软件应用上的支持;3)AI和HPC领域融合的未来发展前景与挑战。
