Single-proc 64-bit
|
Single-proc 32-bit
|
MPP 64-bit
|
MPP 32-bit
|
Ocean
|
Coupled
|
Opteron
(native 64 bit)
(External)
Alan Iwi's UM page
|
(External)
UWERN
(External)
newsletter
|
Timings
|
(External)
AH's timings
(needs password)
|
(External)
CGAM
and (External)
UMIS
|
hints-and-tips
|
Bit comparison
|
Misc
|
Cluster activity
|
Utilities
(inc
ancil creation
)
|
Gcom
Science
Modsets
|
Userprestash files
|
Verify
|
Technical Problems Database
|
(External)
vn4.5
and
(External)
vn5.3
code browser
Basis Jobs
Platforms
Beowulf and the PUM vn4.5 at BAS
BAS has a 2-node dual Opteron cluster and a 13-node cluster (8 dual-board Athlon MP 1600's, with 5 dual-board 2200's recently [2003/02] added).
The older 4-node cluster (dual PIII's 1GHz from Compusys) is now turned off.
Generally the machines are trouble free. On 2002/12 we had our first problem:
the power supply has failed in one of the older nodes
- though other cluster folk indicate a higher rate of PSU failure.
We have the UKMO/Hadley Centre Ported Unified Model vn4.5. You
want one? Go ask them. But if you want help making it work, then hopefully
there are usful things here.
Don't expect it to work "out of the box". Lifes not like that.
Status of the model at BAS
On the whole, this is for the Fujitsu compiler (old version).
ATMOS
works OK.
COUPLED
OK at 64 bit.
-
64-bit almost certainly OK, see
yacjc
. Earlier versions included
climate change mods inappropriate to a control run: see:
yabwb
and
Verification of 32/64 bit model
.
-
32-bit definitely not OK for extended runs (past 10 years).
NB: QTPOS problems are less with the GHG bug fixed...
Past last modified: 17/6/2005
/
wmc@bas.ac.uk
©
Copyright Natural Environment Research Council - British Antarctic Survey 2002
Change Text Only Settings
Graphic version of this page