Subject: Precision problems with CCM3
From: Carlos Fernandez (carlosf@clima.cesga.es)
Date: Fri Mar 21 1997 - 02:50:10 MST
Hello CCM users:
I have found differences in the results of CCM3 (CCM3.2) runs on several
machines, and even on the same machine, using different compilation options.
Tested machines are:
- Indigo2 (R10000) with and without -O2, -mips4, -r10000, -32, -64 options
- Indigo2 (R4400) with and without -O2, -mips2, -mips3, -32, -64 options
- Fujitsu VP2400/10 with -Oe or -Ob
The differences are in several parameters, i.e., surface temperature after only
5 days, or even between monthly averages.
My questions are:
- Are there any dataset files of a CCM3.2 run (T21 and T42) tested at NCAR
public available? If yes, how can I get them?
- How can I know if the differences are significant? Is it an unsolved
numerical problem? Has anybody experienced similar problems?
Thanks in advance...
-- % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % Carlos Fernandez Sanchez CESGA % % Centro de Supercomputacion de Galicia % % E-mail: elcarlosf@uscmail.usc.es % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % %
This archive was generated by hypermail 2b27 : Thu Jun 01 2000 - 09:10:51 MDT