Hi,
I am running v6p3r2 on my university cluster and have experienced a few strange issues which some of you may be able to shed some light on.
Initially I was running my model using 8 cores. I began to refine my mesh incrementally to track any problems and after introducing a couple of piers (my domain is coastal) hit a problem whereby the model would not converge (see example section from log file below). After trying the suggestions in post #14821 I still go the same problem.
ITERATION 71736 TIME: 3 D 6 H 3 MN 45.3001 S ( 281025.3001 S)
--------------------------------------------------------------------------------
ADVECTION STEP
--------------------------------------------------------------------------------
DIFFUSION-PROPAGATION STEP
GRACJG (BIEF) : 2 ITERATIONS, RELATIVE PRECISION: 0.2467060E-06
--------------------------------------------------------------------------------
K-EPSILON MODEL
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: NaN
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: 0.7667694E+14
--------------------------------------------------------------------------------
BALANCE OF WATER VOLUME
VOLUME IN THE DOMAIN : 0.5383677E+34 M3
FLUX BOUNDARY 1: -0.4003055E+42 M3/S ( >0 : ENTERING <0 : EXITING )
RELATIVE ERROR IN VOLUME AT T = 0.2810E+06 S : -0.6102714E-08
MAXIMUM COURANT NUMBER: 0.2548289E-04
TIME-STEP : 0.2602325E-11
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: 0.1189718E+38
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: 30573.14
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: NaN
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: 0.1713843E+18
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: 0.8415477E+39
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: 120754.4
However when I run the same model using 2 cores, 4 cores or 10 cores I do not get this problem and the model runs to completion without an issue. I think this may be due to a boundary problem and how the domain is split up. I haven't yet looked into the source code to find the cause of the problem as for now I'm happy to just accept that the model doesn't work with 8 cores. Does anyone have a suggestion as to what the cause may be?
More interestingly however, the job submissions using 2 cores, 4 cores and 10 cores were identical other than the number of processors used but give slightly different results. When loading in to Blue Kenue:
10 cores:
This appears to be a byte swapped file.
2D Grid Geometry:
142010 nodes, 280638 elements
Scanning Variable: VELOCITY UV
Min: 0, Max: 3.1713
Scanning Variable: WATER DEPTH
Min: -0.11113, Max: 65
Scanning Variable: FREE SURFACE
Min: -4.9323, Max: 6.1287
Scanning Variable: BOTTOM
Min: -62.037, Max: 5.4512
4 cores:
This appears to be a byte swapped file.
2D Grid Geometry:
142010 nodes, 280638 elements
Scanning Variable: VELOCITY UV
Min: 0, Max: 3.1615
Scanning Variable: WATER DEPTH
Min: -0.3174, Max: 65
Scanning Variable: FREE SURFACE
Min: -4.9336, Max: 6.1279
Scanning Variable: BOTTOM
Min: -62.037, Max: 5.4512
2 cores:
This appears to be a byte swapped file.
2D Grid Geometry:
142010 nodes, 280638 elements
Scanning Variable: VELOCITY UV
Min: 0, Max: 3.1834
Scanning Variable: WATER DEPTH
Min: -0.063878, Max: 64.999
Scanning Variable: FREE SURFACE
Min: -4.9327, Max: 6.1274
Scanning Variable: BOTTOM
Min: -62.037, Max: 5.4512
On the whole there isn't an issue with the results as the differences are so small but inspecting my results there is consistently a variation of 1-2cm in water levels throughout my domain. Does anyone know what may be causing this and how to resolve?
Many thanks
Jonathan
