检索结果-南通市图书馆

Developing a Clang Libtooling-Based Refactoring Tool for CUDA GPU Programming

Journal of Software Engineering and Applications 2024年第2期17卷 89-108页

作者： Kian Nejadfard Janche Sang Champ Titles Inc. Cleveland USA Department of Computer Science Cleveland State University Cleveland USA

Refactoring tools, whether fully automated or semi-automated, are essential components of the software development life cycle. As software libraries and frameworks evolve over time, it’s crucial for programs utilizing them to also evolve to remain compatible with modern advancements. Take, for example, NVIDIA CUDA’s platform for general-purpose GPU programming. Embracing the more contemporary unified memory architecture offers several benefits, such as simplifying program source code, reducing bugs stemming from manual memory management between host and device memory, and optimizing memory transfer through automated memory handling. This paper describes our development of a refactoring tool based on Clang’s Libtooling to facilitate this transition automatically, thereby relieving developers from the burden and risks associated with manually refactoring large code bases.

关键词： Refactoring CUDA unified memory Clang Libtooling

来源：

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

A perceptual and predictive batch-processing memory scheduling strategy for a CPU-GPU heterogeneous system

引用

Frontiers of Information Technology & Electronic Engineering 2023年第7期24卷 994-1006页

作者： Juan FANG Sheng LIN Huijing YANG Yixiang XU Xing SU SU Faculty of Information Technology Beijing University of TechnologyBeijing 100124China

When multiple central processing unit(CPU)cores and integrated graphics processing units(GPUs)share off-chip main memory,CPU and GPU applications compete for the critical memory *** causes serious resource competition and has a negative impact on the overall performance of the *** describe the competition for shared-memory resources in a CPU-GPU heterogeneous multi-core architecture,and a sharedmemory request scheduling strategy based on perceptual and predictive batch-processing is *** sensing the CPU and GPU memory request conditions in the request buffer,the proposed scheduling strategy estimates the GPU latency tolerance and reduces mutual interference between CPU and GPU by processing CPU or GPU memory requests in *** to the simulation results,the scheduling strategy improves CPU performance by8.53%and reduces mutual interference by 10.38%with low hardware complexity.

关键词： CPU-GPU heterogeneous Multi-core unified memory Access scheduling

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

Transparent partial page migration between CPU and GPU

引用

Frontiers of Computer Science 2020年第3期14卷 1-13页

作者： Shiqing ZHANG Zheng QIN Yaohua YANG Li SHEN Zhiying WANG Department of Computer Science National University of Defense TechnologyChangsha 410073China

Despite the increasing investment in integrated GPU and next-generation interconnect research,discrete GPU connected by PCIe still account for the dominant position of the market,the management of data communication between CPU and GPU continues to ***,the programmer explicitly controls the data transfer between CPU and *** simplify programming and enable systemwide atomic memory operations,GPU vendors have developed a programming model that provides a single,virtual address space for accessing all CPU and GPU memories in the *** page migration engine in this model automatically migrates pages between CPU and GPU on *** meet the needs of high-performance workloads,the page size tends to be *** by low bandwidth and high latency interconnects compared to GDDR,larger page migration has longer delay,which may reduce the overlap of computation and transmission,waste time to migrate unrequested data,block subsequent requests,and cause serious performance *** this paper,we propose partial page migration that only migrates the requested part of a page to reduce the migration unit,shorten the migration latency,and avoid the performance degradation of the full page migration when the page becomes *** show that partial page migration is possible to largely hide the performance overheads of full page *** with programmer controlled data transmission,when the page size is 2MB and the PCIe bandwidth is 16GB/sec,full page migration is 72.72×slower,while our partial page migration achieves 1.29×*** the PCIe bandwidth is changed to 96GB/sec,full page migration is 18.85×slower,while our partial page migration provides 1.37×***,we examine the performance impact that PCIe bandwidth and migration unit size have on execution time,enabling designers to make informed decisions.

关键词： unified memory data communication page migration partial page migration

来源：

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

D-Cubicle:boosting data transfer dynamically for large-scale analytical queries in single-GPU systems

引用

Frontiers of Computer Science 2023年第4期17卷 141-153页

作者： Jialun WANG Wenhao PANG Chuliang WENG Aoying ZHOU School of Data Science and Engineering East China Normal UniversityShanghai 200062China

In analytical queries,a number of important operators like JOIN and GROUP BY are suitable for parallelization,and GPU is an ideal accelerator considering its power of parallel ***,when data size increases to hundreds of gigabytes,one GPU card becomes insufficient due to the small capacity of global memory and the slow data transfer between host and device.A straightforward solution is to equip more GPUs linked with high-bandwidth connectors,but the cost will be highly *** utilize unified memory(UM)produced by NVIDIA CUDA(Compute unified Device Architecture)to make it possible to accelerate large-scale queries on just one GPU,but we notice that the transfer performance between host and UM,which happens before kernel execution,is often significantly slower than the theoretical *** important reason is that,in singleGPU environment,data processing systems usually invoke only one or a static number of threads for data copy,leading to an inefficient transfer which slows down the overall performance *** this paper,we present D-Cubicle,a runtime module to accelerate data transfer between host-managed memory and unified memory.D-Cubicle boosts the actual transfer speed dynamically through a self-adaptive *** our experiments,taking data transfer into account,D-Cubicle processes 200 GB of data on a single GPU with 32 GB of global memory,achieving 1.43x averagely and 2.09x maximally the performance of the baseline system.

关键词： data analytics GPU unified memory

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

欢迎您,

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

在线全文

在线全文

在线全文

在线全文

请选择保存的检索档案：

请选择收藏分类：

通借通还

欢迎您,

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

在线全文

在线全文

在线全文

在线全文

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：