[CIG-SHORT] improved Okada matlab script(s) for benchmarks
Brad Aagaard
baagaard at usgs.gov
Mon Jul 24 17:28:57 PDT 2006
Hi all:
I just uploaded improved matlab scripts (the .m files) for the reverse
and strike-slip benchmarks. As I was trying to compute the analytic
solution for use with the benchmarking infrastructure that Luis is
working on, I noticed the memory use rising as the script progressed.
I altered the script to preallocate the memory for the arrays and
removed some temporary variables. The memory use should now be steady
as the script progresses and much less for large numbers of
datapoints.
I also factored out the common parts of the two scripts and put it in
a separate file. The ReverseSlip_ng.m and StrikeSlip_ng.m files are
now just lightweight wrappers around a new script called
okadasoln_ng.m.
This should make it easier to generate the boundary conditions and
solution for the higher resolution benchmarks. On my machine, the
script computes about 12,000 points per minute. For the points.in in
the tarball (nearly a million points) the script uses about 212 MB of
memory.
http://www.geodynamics.org:8080/cig/workinggroups/eq/workarea/benchmarks/utilities/elastic-okada/
Brad
More information about the CIG-SHORT
mailing list