================================================================== === === === GENESIS Distributed Memory Benchmarks === === === === TICK2 === === === === Measuring the clock multiplier === === === === Versions: Std F77, Subset HPF(F90) === === === === Author: Roger Hockney === === Department of Electronics and Computer Science === === University of Southampton === === Southampton SO9 5NH, U.K. === === fax.:+44-703-593045 e-mail:rwh@uk.ac.soton.pac === === === === Last update: May 1993; Release: 2.2 === === === ================================================================== 1. Description -------------- Most computers measure time by counting the number of clock ticks, then use a multiplier to convert this number to seconds. This benchmark tests the validity of this multiplier, by measuring the same time interval with both the computer TIMER subroutine, and an external wall-clock, which can be your own wrist-watch. The time interval is started and ended by a keyboard input. We suggest you time in interval of 5 minutes, which is timeable to about 0.1% with a wrist-watch. If this multiplier is very variable or badly in error, it probably means that you are using a CPU-for-this-job-only timer not a wall-clock timer. You must use a wall-clock timer for parallel benchmarking. *** BEWARE *** : The returned value of the popular Sun Unix timer ETIME is not a wall-clock timer, even though it is called the elapsed time. 2. Operating Instructions ------------------------- To compile and link the benchmark type: `make slave'. On some systems it may be necessary to allocate the appropriate resources before running the benchmark, eg. on the iPSC/860 to reserve a single processor, type: getcube -t1. To run the benchmark type: tick2 Output from the benchmark is written to the file "result" $Id: ReadMe,v 1.2 1994/04/20 13:17:50 igl Rel igl $
Submitted by Mark Papiani,
last updated on 10 Jan 1995.