Title :
Performance results of several High Performance Fortran benchmarks
Author :
Meadows, Larry F. ; Miles, Douglas ; Young, Mark
Author_Institution :
Portland Group Inc., Wilsonville, OR, USA
Abstract :
The pghpf High Performance Fortran (HPF) compiler from The Portland Group, Inc. (PGI) has been ported to a variety of shared- and distributed-memory platforms. To track compiler improvements and assist in tuning of the pghpf runtime system, 3 types of benchmarks are used. Low level runtime benchmarks measure the efficiency of the runtime system relative to hardware specifications. Kernel benchmarks, usually only a few hundred lines long, measure effectiveness on sequences of code that dominate the execution time of many common applications. Full application benchmarks measure effectiveness on production-class codes from a variety of disciplines. Application benchmarks are typically long-running, include I/O of varying amounts, and are significantly more complicated than the kernel benchmarks. We describe these benchmarks below in further detail
Keywords :
FORTRAN; distributed memory systems; program compilers; shared memory systems; software performance evaluation; HPF; compiler; distributed-memory systems; kernel benchmarks; pghpf High Performance Fortran; production-class codes; shared-memory systems; Benchmark testing; Delay; Kernel; Motion measurement; Performance evaluation; Power measurement; Runtime; System testing; Time measurement; Velocity measurement;
Conference_Titel :
Parallel Processing Symposium, 1995. Proceedings., 9th International
Conference_Location :
Santa Barbara, CA
Print_ISBN :
0-8186-7074-6
DOI :
10.1109/IPPS.1995.395979