文献类型：专利

英文题名：Memory access priority based general-purpose graphics processing unit performance level optimizing method, involves selecting access request corresponding to access priority, and continuing scheduling process when task execution is finished

作者：ZHANG J;HE Y;TAN H

机构：[1]UNIV EAST CHINA SCI & TECHNOLOGY

申请号：CN108279981-A

申请日：2018-01-30

公开日：2018-07-13

语种：英文

收录：DERWENT

摘要：NOVELTY - The method involves calculating item length a table item in a MSHR. Memory access dispersion of the table item in the MSHR is calculated. Highest access priority of the table in the MSHR is calculated. An access request is selected corresponding to the highest access priority of the table. Next access scheduling process is continued when task execution is finished. Judgment is made to check whether memory access dispersion of the table in the MSHR is equal to thread memory access address of the table. USE - Memory access priority based general-purpose graphics processing unit (GPGPU) performance level optimizing method. ADVANTAGE - The method enables reducing cache access contention problems on the GPGPU. DESCRIPTION OF DRAWING(S) - The drawing shows a block diagram of a memory access priority based GPGPU performance level optimizing system. '(Drawing includes non-English language text)'