These pages show the results of performance tests for different dynamo codes on the TACC XSEDE Stampede system.
To clarify terminology, here are some definitions of terms used in the measurements:
All timings are measured in terms of wall time unless otherwise specified. They are measured using
MPI_Wtime() (If program does not use MPI,
system_clock(itime, irate, imax) is used). The time for one step is determined by taking the mean of the elapsed time for 100 time steps.
* Calypso Ver.1.1
* Calypso latest Ver.
Status: Two cases are missing
Status: Having a problem generating initial field. Request is pending to authors.
SPmodel(GFD-dennou code by Takehiro et al.)
Status: Strong scaling is done. Discussing with authors
Status: Routine has been written to generate initial field of multiple resolutions. Tests are underway. Request pending for how to deal with $dt$ adjustment.
Test again with different resolutions from better scaling.
Parody(IPGP JHU code)
LSD (Leeds code)
MagIC (by J. Wicht)
Status: Now testing.
SFEMANS (not finished)
Status: Strong scaling is done. Discussing with authors for warnings and changing resolutions
Status: Received new version. Now testing.
* UCSC-Glatzmaier code
Status: Strong scaling is done.
* UCSC-Glatzmaier code R3-p4
Status: Report was sent to authors.