/[MITgcm]/MITgcm/verification/global_ocean.90x40x15/README
ViewVC logotype

Diff of /MITgcm/verification/global_ocean.90x40x15/README

Parent Directory Parent Directory | Revision Log Revision Log | View Revision Graph Revision Graph | View Patch Patch

revision 1.3 by jmc, Mon Mar 10 01:35:16 2014 UTC revision 1.4 by heimbach, Tue Mar 11 16:06:44 2014 UTC
# Line 1  Line 1 
1    ============================================
2  Example: "4x4 Global Simulation with Seasonal Forcing"  Example: "4x4 Global Simulation with Seasonal Forcing"
3  ============================================  ============================================
4  (see also similar set-up in: verification/tutorial_global_oce_latlon/)  (see also similar set-up in: verification/tutorial_global_oce_latlon/)
# Line 49  to run this 2nd example: Line 50  to run this 2nd example:
50    ../build/mitgcmuv > output.dwnslp.txt    ../build/mitgcmuv > output.dwnslp.txt
51    cd ..    cd ..
52    
53    
54    ============================================
55    Use of "blank tiles" in conjunction with exch2 package:
56    ============================================
57    
58    This verification experiment also demonstrate the omission of tiles
59    (or processors) for tiles that are fully land cover and don't need computation.
60    The relevant config. files to be manipulated are:
61    * at compile time, in dir. code/: packages.conf, SIZE.h
62    * at run time, in dir. input/: data.exch2
63    
64    To enable this feature requires package "exch2"
65    (see Section 6.2.4 "exch2: Extended Cubed Sphere Topology" of online manual),
66    i.e. in code/packages.conf add "exch2"
67    
68    The basic layout of the experiment is Nx*Ny = 90x40.
69    In a single-processor config. with very small tile sizes (sNx*sNy=10*10)
70    this can be represented, e.g. via
71         &           sNx =  10,
72         &           sNy =  10,
73         &           OLx =   3,
74         &           OLy =   3,
75         &           nSx =   9,
76         &           nSy =   4,
77         &           nPx =   1,
78         &           nPy =   1,
79    i.e. we use nSx*nSy=9*4=36 virtual processors.
80    
81    An equivalent parallel setup using 9 virtual and 4 real processors
82    would look like:
83         &           sNx =  10,
84         &           sNy =  10,
85         &           OLx =   3,
86         &           OLy =   3,
87         &           nSx =   9,
88         &           nSy =   1,
89         &           nPx =   1,
90         &           nPy =   4,
91    
92    In this layout it turns out that tile number 30 is "empty", i.e. fully land covered.
93    We wish to remove this tile from our calculation. How to proceed?
94    
95    1. Find out which tiles to eliminate via a config. that uses all tiles
96    ---
97    1.1. At compile time:
98    * add line "exch2" to packages.conf
99    * configure SIZE.h using your desired individual tile size, e.g. sNx*sNy=10*10, as follows:
100         &           sNx =  10,
101         &           sNy =  10,
102         &           OLx =   3,
103         &           OLy =   3,
104         &           nSx =   9,
105         &           nSy =   1,
106         &           nPx =   1,
107         &           nPy =   4,
108      As described above you are using 4 real processors with 9 virtual tiles per processor.
109    * compile (don't forget to compile via -mpi or similar)
110    
111    1.2 At runtime:
112    * need to reflect your basic layout in data.exch2
113      This is simple: since you are not using any non-trivial topology with
114      multiple facets (such as cubed-sphere),
115      you only need to specify one basic facet (Nx,Ny)=(90,40) layout via following line:
116      dimsFacets = 90, 40,
117      Then run the moodel, e.g. via:
118      mpirun -np 4 ./mitgcmuv
119    
120    1.3 Diagnose which tile numbers are empty:
121    * from STDOUT.000[0-3] you can infer which tiles are empty using following grep:
122      grep "Empty tile" STDOUT.* | awk '{print "  " $6 ","}' > empty_tiles.txt
123    * In this example there's only one empty tile, and it is #30.
124    
125    2. Configuring with empty tiles removed
126    ---
127    2.1 At compile time:
128       We've determined that 1 out of nSx*nPx*nSy*nPy=36 tiles is empty and can be removed,
129       leaving 36-1=35 non-empty tiles. We are free to re-order nSx,nPx,nSy,nPy in any way,
130       as long as nSx*nPx*nSy*nPy=35. Here's how it's chosen in the verif. experiment
131       (see file SIZE.h_mpi)
132         &           sNx =  10,
133         &           sNy =  10,
134         &           OLx =   3,
135         &           OLy =   3,
136         &           nSx =   7,
137         &           nSy =   1,
138         &           nPx =   1,
139         &           nPy =   5,
140       and for which, as required nSx*nPx*nSy*nPy=35
141    
142    2.2 At runtime:
143    We now need to specify in data.exch2 the number of the empty tile to be removed.
144    This is done as follows (see file data.exch2.mpi):
145      blankList  = 30,
146    If there were more empty tiles, this would be a more extensive index array, e.g.
147      blankList = tile1, tile2, tile3, ...
148    Now run the model (note that we've selected nPy=5, so in this example we actually
149    *increase* the number of "real" processors used, despite reducing the number of
150    total, i.e. "real" plus virtual tiles from 36 to 35):
151    mpirun -np 5 ./mitgcmuv
152    
153    
154  ============================================  ============================================
155  Adjoint set-up example:  Adjoint set-up example:
156    ============================================
157    
158  Configure and compile the code:  Configure and compile the code:
159    cd build    cd build
Legend:
Removed from v.1.3  
changed lines
  Added in v.1.4
  ViewVC Help
Powered by ViewVC 1.1.22