Automated Analysis of Memory Access Behavior

Abstract— We developed an automated environment to measure the memory access behavior of applications on high performance clusters. Code optimization for processor caches is crucial for achieving good performance on such systems. Our environment is capable to automatically instrument OpenMP Fortran95 programs upon requests of programmer. The monitor can be configured to selectively collect hardware counter information. Limitations due to the number of physical hardware counters are automatically taken into account. The whole environment is controlled through a user interface based on Eclipse and is highly portable.

• Type: Conference
• Year: 2005
• Where: IPPS
• Authors: Michael Gerndt, Tianchao Li
• Keywords: Distributed And Parallel Computing, Hardware Counters, IPPS 2005, Memory Access Behavior, OpenMP Fortran95 Programs