After some hardware upgrades, finally the model started running without erroring out all the time. (Not on the laptop, I simply gave up there, fortran errors were finally the last straw. Fortan? You\'ve got to be kidding me.) However, on the desktop, I\'ve gotten to nearly 4% complete without errors, and I think that\'s great. However, somewhere around 1.5 percent, the clouds stopped displaying correctly. Now, my oceans show up like smooth grey expanses. If I play with the control to turn off clouds, then suddenly the ocean blue shows up with the clouds are toggled off. So, I\'m thinking that somewhere in the computations the clouds lost elevation, so far that they think they\'re below sea-level. Is this happening to anyone else, and is there a fix to it?

---

Hi Jindai

The model code is indeed written in Fortran, I believe over a million lines of code from the UK Met Office......

Have you got one or two HADSM slab models showing in the Tasks tab of your desktop? It would be better if we know for sure because the model you describe running on your desktop seems to have gone wrong and we need to know exactly which model it is to make a definite diagnosis.

I think the slab model you describe has turned into what, for want of a better phrase, we call an ice world. Model slowed right down, graphics no longer displaying properly. If this is the case it won\'t be processing properly either and will need to be aborted.

Read about these iceworld slab models in Geophi\'s thread near the top of Number Crunching:

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_forum.php?id=2

There\'s a discussion thread about the same problem a bit further down.

If your desktop is overclocked, the problem could have been caused by your computer. If it isn\'t overclocked, we will assume that the model was flawed and doomed. (We need to know about possible O/Cing please.)

You\'ll see from Geophi\'s thread that after a definite diagnosis, ice world models have to be aborted - they don\'t correct themselves. Bad luck with this model, but at least you\'ve spotted the problem early.

---

Cpdn news

Hi Jindai

The model code is indeed written in Fortran, I believe over a million lines of code from the UK Met Office......

Wow, fortran, thought that was seriously outdated, but okay. It does keep giving me errors on the laptop, which only had XP Home, so I detached from it there. I miss the lovely graphics, but whatchagonnado?

Have you got one or two HADSM slab models showing in the Tasks tab of your desktop? It would be better if we know for sure because the model you describe running on your desktop seems to have gone wrong and we need to know exactly which model it is to make a definite diagnosis.

Only one HADSM slam model showing. hadsm3fub_0563_005942533_4, if that\'s any help.

I think the slab model you describe has turned into what, for want of a better phrase, we call an ice world. Model slowed right down, graphics no longer displaying properly. If this is the case it won\'t be processing properly either and will need to be aborted.

Ugg, again. Guess I\'ll abort.

If your desktop is overclocked, the problem could have been caused by your computer. If it isn\'t overclocked, we will assume that the model was flawed and doomed. (We need to know about possible O/Cing please.)

Not overclocking.. Flawed and doomed. Just my luck. I finally get it so it doesn\'t just fail out, but it still fails.
I\'ll abort, and see what happens next time. *sigh*

Thanks

---

This is the model:

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=7394932
Before aborting it you need to check that it really has slowed right down (typically 10X slower) compared with the sec/timestep speeds shown for its first two trickles.

Open its graphics window, press Z then 8 to show the extra info including timesteps. These models keep counting down from 144 to 0, then back to 144 again. Note the model\'s timestep time on a watch with a seconds hand, close the graphics window, let the model run for 5 minutes, then look again and calculate how many seconds the model takes for each timestep. I\'d repeat the exercise to be sure.

It would help if you tell us what you find.

A couple of computers running models in the same workunit have got past the point your computer\'s reached, but they\'re Intels whereas yours is an AMD. Sometimes an iceworld develops on one sort of CPU only.

---

Cpdn news

Dang, I\'m sorry. I pulled the trigger too quickly. I read the post on ice-worlds, and based on that, it sounded like it was doomed, so I aborted it. So I can\'t do the diagnosis you are asking for. Sorry about that. Well, if it happens again, I\'ll wait to abort until I have done all you ask. Sorry, again.

---

Well, I expect you\'d read about the ice world graphics problem in Geophi\'s post and assuming that you were seeing the same, the diagnosis would be safe. I just wanted to double-check.

Better luck with your next model!

---

Cpdn news