DIRACbenchmark¶
DIRAC Benchmark 2012 by Ricardo Graciani, and wrapper functions to run multiple copies in parallel by Andrew McNab.
This file (DIRACbenchmark.py) is intended to be the ultimate upstream shared by different users of the DIRAC Benchmark 2012 (DB12). The canonical version can be found at https://github.com/DIRACGrid/DB12
This script can either be imported or run from the command line:
./DIRACbenchmark.py NUMBER
where NUMBER gives the number of benchmark processes to run in parallel.
Run ./DIRACbenchmark.py help to see more options.
- DIRAC.WorkloadManagementSystem.Client.DIRACbenchmark.jobslotDiracBenchmark(copies=None, iterations=1, extraIteration=False)¶
Run as many copies as needed to occupy the job slot
- DIRAC.WorkloadManagementSystem.Client.DIRACbenchmark.multipleDiracBenchmark(copies=1, iterations=1, extraIteration=False)¶
Run multiple copies of the DIRAC Benchmark in parallel
- DIRAC.WorkloadManagementSystem.Client.DIRACbenchmark.singleDiracBenchmark(iterations=1, measuredCopies=None)¶
Get Normalized Power of one CPU in DIRAC Benchmark 2012 units (DB12)
- DIRAC.WorkloadManagementSystem.Client.DIRACbenchmark.singleDiracBenchmarkProcess(resultObject, iterations=1, measuredCopies=None)¶
Run singleDiracBenchmark() in a multiprocessing friendly way
- DIRAC.WorkloadManagementSystem.Client.DIRACbenchmark.wholenodeDiracBenchmark(copies=None, iterations=1, extraIteration=False)¶
Run as many copies as needed to occupy the whole machine