Data layout in a file system is the organization of data stored in external storages. The data layout has a huge impact on performance of storage systems. We survey three main kinds of data layout in traditional file ...Data layout in a file system is the organization of data stored in external storages. The data layout has a huge impact on performance of storage systems. We survey three main kinds of data layout in traditional file systems： in-place update file system, log-structured file system, and copy-on-write file sys- tem. Each file system has its own strengths and weaknesses under different circumstances. We also include a recent us- age of persistent layout in a file system that combines both flash memory and byte- addressable non- volatile memory. With this survey, we conclude that persistent data layout in file systems may evolve dramatically in the era of emerging non-volatile memory.展开更多
Over the past decade, Graphics Processing Units (GPUs) have revolutionized high-performance computing, playing pivotal roles in advancing fields like IoT, autonomous vehicles, and exascale computing. Despite these adv...Over the past decade, Graphics Processing Units (GPUs) have revolutionized high-performance computing, playing pivotal roles in advancing fields like IoT, autonomous vehicles, and exascale computing. Despite these advancements, efficiently programming GPUs remains a daunting challenge, often relying on trial-and-error optimization methods. This paper introduces an optimization technique for CUDA programs through a novel Data Layout strategy, aimed at restructuring memory data arrangement to significantly enhance data access locality. Focusing on the dynamic programming algorithm for chained matrix multiplication—a critical operation across various domains including artificial intelligence (AI), high-performance computing (HPC), and the Internet of Things (IoT)—this technique facilitates more localized access. We specifically illustrate the importance of efficient matrix multiplication in these areas, underscoring the technique’s broader applicability and its potential to address some of the most pressing computational challenges in GPU-accelerated applications. Our findings reveal a remarkable reduction in memory consumption and a substantial 50% decrease in execution time for CUDA programs utilizing this technique, thereby setting a new benchmark for optimization in GPU computing.展开更多
基金supported by ZTE Industry-Academia-Research Cooperation Funds
文摘Data layout in a file system is the organization of data stored in external storages. The data layout has a huge impact on performance of storage systems. We survey three main kinds of data layout in traditional file systems： in-place update file system, log-structured file system, and copy-on-write file sys- tem. Each file system has its own strengths and weaknesses under different circumstances. We also include a recent us- age of persistent layout in a file system that combines both flash memory and byte- addressable non- volatile memory. With this survey, we conclude that persistent data layout in file systems may evolve dramatically in the era of emerging non-volatile memory.
文摘Over the past decade, Graphics Processing Units (GPUs) have revolutionized high-performance computing, playing pivotal roles in advancing fields like IoT, autonomous vehicles, and exascale computing. Despite these advancements, efficiently programming GPUs remains a daunting challenge, often relying on trial-and-error optimization methods. This paper introduces an optimization technique for CUDA programs through a novel Data Layout strategy, aimed at restructuring memory data arrangement to significantly enhance data access locality. Focusing on the dynamic programming algorithm for chained matrix multiplication—a critical operation across various domains including artificial intelligence (AI), high-performance computing (HPC), and the Internet of Things (IoT)—this technique facilitates more localized access. We specifically illustrate the importance of efficient matrix multiplication in these areas, underscoring the technique’s broader applicability and its potential to address some of the most pressing computational challenges in GPU-accelerated applications. Our findings reveal a remarkable reduction in memory consumption and a substantial 50% decrease in execution time for CUDA programs utilizing this technique, thereby setting a new benchmark for optimization in GPU computing.
