1 |
For interactive session, Ivy Bridge nodes: |
2 |
qsub -I -q devel -l select=300:ncpus=20:model=ivy,walltime=02:00:00 -m abe |
3 |
qsub -I -q normal -l select=103:ncpus=20:model=ivy,walltime=8:00:00 -m abe |
4 |
qsub -I -q long -l select=300:ncpus=20:model=ivy,walltime=120:00:00 -m abe |
5 |
qsub -I -q long -l select=103:ncpus=20:model=ivy,min_walltime=30:00,max_walltime=120:00:00 -m abe |
6 |
|
7 |
########################## |
8 |
# 72x72_7666 configuration |
9 |
qsub -I -q long -l select=400:ncpus=20:model=ivy,walltime=120:00:00 -m abe |
10 |
module purge |
11 |
module load comp-intel/2016.2.181 mpi-hpe/mpt.2.17r13 hdf4/4.2.12 hdf5/1.8.18_mpt netcdf/4.4.1.1_mpt |
12 |
cd ~/llc_2160 |
13 |
git clone git@github.com:MITgcm/MITgcm.git |
14 |
cvs co MITgcm_contrib/llc_hires/llc_2160 |
15 |
cd ~/llc_2160/MITgcm |
16 |
mkdir build run |
17 |
cd ~/llc_2160/MITgcm/build |
18 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code/SIZE.h_72x72_7666 SIZE.h |
19 |
../tools/genmake2 -of \ |
20 |
../../MITgcm_contrib/llc_hires/llc_2160/code-async/linux_amd64_ifort+mpi_ice_nas -mpi -mods \ |
21 |
'../../MITgcm_contrib/llc_hires/llc_2160/code ../../MITgcm_contrib/llc_hires/llc_2160/code-async' |
22 |
make depend |
23 |
make -j 16 |
24 |
|
25 |
cd ~/llc_2160/MITgcm/run |
26 |
cp ../build/mitgcmuv mitgcmuv_72x72_7666 |
27 |
ln -sf /nobackup/dmenemen/tarballs/llc_2160/run_template/* . |
28 |
ln -sf /nobackup/dmenemen/forcing/ECMWF_operational/* . |
29 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/* . |
30 |
cp data.exch2_72x72x7666 data.exch2 |
31 |
emacs data |
32 |
|
33 |
mpiexec -n 8000 ./mitgcmuv_72x72_7666 |
34 |
|
35 |
tail -f STDOUT.00000 | grep advcfl_W |
36 |
|
37 |
################################################################### |
38 |
# 90x90_5004 configuration with newer code on electra special queue |
39 |
qsub -I -q electra -l select=200:ncpus=28:model=bro_ele:aoe=sles12 -l walltime=10:00:00 |
40 |
module purge |
41 |
module load comp-intel/2016.2.181 mpi-hpe/mpt.2.17r13 hdf4/4.2.12 hdf5/1.8.18_mpt netcdf/4.4.1.1_mpt |
42 |
cd ~/llc_2160 |
43 |
git clone git@github.com:MITgcm/MITgcm.git |
44 |
cvs co MITgcm_contrib/llc_hires/llc_2160 |
45 |
cd ~/llc_2160/MITgcm |
46 |
mkdir build run |
47 |
cd ~/llc_2160/MITgcm/build |
48 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code/SIZE.h_90x90_5004 SIZE.h |
49 |
../tools/genmake2 -of \ |
50 |
../../MITgcm_contrib/llc_hires/llc_2160/code-async/linux_amd64_ifort+mpi_ice_nas -mpi -mods \ |
51 |
'../../MITgcm_contrib/llc_hires/llc_2160/code ../../MITgcm_contrib/llc_hires/llc_2160/code-async' |
52 |
make depend |
53 |
make -j 16 |
54 |
|
55 |
cd ~/llc_2160/MITgcm/run |
56 |
cp ../build/mitgcmuv mitgcmuv_90x90_5004 |
57 |
ln -sf /nobackup/dmenemen/tarballs/llc_2160/run_template/* . |
58 |
ln -sf /nobackup/dmenemen/forcing/ECMWF_operational/* . |
59 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/* . |
60 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/data.exch2_90x90x5004 data.exch2 |
61 |
emacs data |
62 |
|
63 |
mpiexec -n 5500 ./mitgcmuv_90x90_5004 |
64 |
|
65 |
tail -f STDOUT.00000 | grep advcfl_W |
66 |
|
67 |
############################# |
68 |
# 60x60x10882 configuration |
69 |
|
70 |
qsub -I -q long -l select=600:ncpus=20:model=ivy,min_walltime=30:00,max_walltime=120:00:00 -m abe |
71 |
module purge |
72 |
module load comp-intel/2012.0.032 mpi-sgi/mpt.2.10r6 netcdf/4.0 |
73 |
cd ~/llc_2160 |
74 |
cvs co -r checkpoint64t MITgcm_code |
75 |
cvs co MITgcm_contrib/llc_hires/llc_2160 |
76 |
cd MITgcm |
77 |
mkdir build run |
78 |
cd build |
79 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code/SIZE.h_60x60_10882 SIZE.h |
80 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code-async/readtile_mpiio.c . |
81 |
emacs readtile_mpiio.c |
82 |
tileSizeX = 60; |
83 |
tileSizeY = 60; |
84 |
../tools/genmake2 -of \ |
85 |
../../MITgcm_contrib/llc_hires/llc_2160/code-async/linux_amd64_ifort+mpi_ice_nas -mpi -mods \ |
86 |
'../../MITgcm_contrib/llc_hires/llc_2160/code ../../MITgcm_contrib/llc_hires/llc_2160/code-async' |
87 |
make depend |
88 |
make -j 16 |
89 |
|
90 |
cd ~/llc_2160/MITgcm/run |
91 |
cp ../build/mitgcmuv mitgcmuv_60x60x10882 |
92 |
ln -sf /nobackup/dmenemen/tarballs/llc_2160/run_template/* . |
93 |
ln -sf /nobackup/dmenemen/forcing/ECMWF_operational/* . |
94 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/* . |
95 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/data.exch2_60x60x10882 data.exch2 |
96 |
emacs data |
97 |
|
98 |
export MPI_BUFS_PER_PROC=1024 |
99 |
export MPI_REQUEST_MAX=65536 |
100 |
export MPI_GROUP_MAX=1024 |
101 |
export MPI_NUM_MEMORY_REGIONS=8 |
102 |
export MPI_UNBUFFERED_STDIO=1 |
103 |
export MPI_MEMMAP_OFF=1 |
104 |
export MPI_UD_TIMEOUT=100 |
105 |
mpiexec -n 12000 ./mitgcmuv_60x60x10882 |
106 |
|
107 |
tail -f STDOUT.00000 | grep advcfl_W |
108 |
|
109 |
################################################ |
110 |
# 144x144x2047 configuration for grid generation |
111 |
|
112 |
qsub -I -q devel -l select=103:ncpus=20:model=ivy,walltime=02:00:00 -m abe -M YOUR_EMAIL_HERE |
113 |
module purge |
114 |
module load comp-intel/2012.0.032 mpi-sgi/mpt.2.10r6 netcdf/4.0 |
115 |
cd ~/llc_2160 |
116 |
cvs co MITgcm_code |
117 |
cvs co MITgcm_contrib/llc_hires/llc_2160 |
118 |
cd MITgcm |
119 |
mkdir build run_grid |
120 |
lfs setstripe -c -1 run_grid |
121 |
cd build |
122 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code/SIZE.h_144x144x2047 SIZE.h |
123 |
../tools/genmake2 -of \ |
124 |
../../MITgcm_contrib/llc_hires/llc_2160/code-async/linux_amd64_ifort+mpi_ice_nas \ |
125 |
-mpi -mods ../../MITgcm_contrib/llc_hires/llc_2160/code |
126 |
make depend |
127 |
make -j 16 |
128 |
|
129 |
cd ~/llc_2160/MITgcm/run_grid |
130 |
cp ../build/mitgcmuv mitgcmuv_144x144x2047 |
131 |
ln -sf /nobackup/dmenemen/tarballs/llc_2160/run_template/* . |
132 |
ln -sf /nobackup/dmenemen/forcing/ECMWF_operational/* . |
133 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/* . |
134 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/data.exch2_144x144x2047 data.exch2 |
135 |
|
136 |
emacs data |
137 |
debuglevel=3, |
138 |
useSingleCPUio=.TRUE., |
139 |
endtime=0., |
140 |
deltaT = 1., |
141 |
|
142 |
mpiexec -n 2047 ./mitgcmuv_144x144x2047 |
143 |
|
144 |
============== |
145 |
|
146 |
cd ~/llc_2160 |
147 |
cvs co MITgcm_code |
148 |
cvs co MITgcm_contrib/llc_hires/llc_2160 |
149 |
cd MITgcm |
150 |
module purge |
151 |
module load comp-intel/2012.0.032 mpi-sgi/mpt.2.08r7 netcdf/4.0 |
152 |
mkdir build run |
153 |
lfs setstripe -c -1 run |
154 |
cd build |
155 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code/SIZE.h_90x90_5004 SIZE.h |
156 |
../tools/genmake2 -of \ |
157 |
../../MITgcm_contrib/llc_hires/llc_2160/code-async/linux_amd64_ifort+mpi_ice_nas -mpi -mods \ |
158 |
'../../MITgcm_contrib/llc_hires/llc_2160/code ../../MITgcm_contrib/llc_hires/llc_2160/code-async' |
159 |
make depend |
160 |
make -j 16 |
161 |
cd ~/llc_2160/MITgcm/run |
162 |
ln -sf ../build/mitgcmuv . |
163 |
ln -sf /nobackup/dmenemen/tarballs/llc_2160/run_template/* . |
164 |
ln -sf /nobackup/dmenemen/forcing/ECMWF_operational/* . |
165 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/* . |
166 |
mv data.exch2_90x90x5004 data.exch2 |
167 |
export MPI_NUM_MEMORY_REGIONS=256 |
168 |
mpiexec -n 6000 ./mitgcmuv |
169 |
|
170 |
============== |
171 |
|
172 |
look at output |
173 |
|
174 |
for ts=[0 120 600:10:980 1080:120:2280] |
175 |
fld=quikread_llc(['Eta.' myint2str(ts,10) '.data'],2160); |
176 |
clf,quikplot_llc(fld),caxis([-2.5 2]),thincolorbar |
177 |
title(ts) |
178 |
pause(.1) |
179 |
end |
180 |
|
181 |
============== |
182 |
|
183 |
to determine empty tiles: |
184 |
grep Empty STDOUT.* > empty.txt |
185 |
|
186 |
============== |
187 |
|
188 |
# generate 30x30 blank tiles |
189 |
qsub -I -q normal -l select=339:ncpus=28:model=bro,walltime=2:00:00 -m abe |
190 |
module purge |
191 |
module load comp-intel/2016.2.181 mpi-sgi/mpt.2.14r19 hdf4/4.2.12 hdf5/1.8.18_mpt netcdf/4.4.1.1_mpt |
192 |
cd ~/llc_2160/MITgcm |
193 |
mkdir run_30x30 |
194 |
lfs setstripe -c -1 run_30x30 |
195 |
cd build |
196 |
rm * |
197 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code-async/readtile_mpiio.c . |
198 |
emacs readtile_mpiio.c |
199 |
tileSizeX = 30; |
200 |
tileSizeY = 30; |
201 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code/SIZE.h_60x60_10882 SIZE.h |
202 |
emacs SIZE.h |
203 |
& sNx = 30, |
204 |
& sNy = 30, |
205 |
& nSx = 8, |
206 |
& nPx = 8424, |
207 |
& Nr = 2 ) |
208 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code-async/eeboot_minimal.F . |
209 |
emacs eeboot_minimal.F |
210 |
C standardMessageUnit=errorMessageUnit |
211 |
WRITE(fNam,'(A,A)') 'STDOUT.', myProcessStr(1:5) |
212 |
OPEN(standardMessageUnit,FILE=fNam,STATUS='unknown') |
213 |
../tools/genmake2 -of \ |
214 |
../../MITgcm_contrib/llc_hires/llc_2160/code-async/linux_amd64_ifort+mpi_ice_nas -mpi -mods \ |
215 |
'../../MITgcm_contrib/llc_hires/llc_2160/code ../../MITgcm_contrib/llc_hires/llc_2160/code-async' |
216 |
make depend |
217 |
make -j 16 |
218 |
cd ~/llc_2160/MITgcm/run_30x30 |
219 |
cp ../build/mitgcmuv . |
220 |
ln -sf /nobackup/dmenemen/tarballs/llc_2160/run_template/* . |
221 |
ln -sf /nobackup/dmenemen/forcing/ECMWF_operational/* . |
222 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/* . |
223 |
mv data.exch2_144x144x2047 data.exch2 |
224 |
emacs data.exch2 |
225 |
# remove the blank tile list |
226 |
emacs data |
227 |
tRef = 18.89, 18.89, |
228 |
sRef = 34.84, 34.84, |
229 |
endtime=45., |
230 |
delR = 1.00, 1.14, |
231 |
# hydrogThetaFile='THETA_llc1080_14jan2011_2160x28080x90_r4', |
232 |
# hydrogSaltFile ='SALT_llc1080_14jan2011_2160x28080x90_r4', |
233 |
# uVelInitFile ='UVEL_llc1080_14jan2011_2160x28080x90_r4', |
234 |
# vVelInitFile ='VVEL_llc1080_14jan2011_2160x28080x90_r4', |
235 |
# pSurfInitFile ='ETAN_llc1080_14jan2011_2160x28080_r4', |
236 |
mpiexec -n 9492 ./mitgcmuv |
237 |
|
238 |
============== |
239 |
|
240 |
memory requirements: |
241 |
nPx sNx sNy nSx cpu node0 total rank0 rankm |
242 |
936 180 180 2 san node ran out of memory and crashed with singlecpuio |
243 |
1053 240 240 1 san node ran out of memory and crashed with singlecpuio |
244 |
1300 216 216 1 san node ran out of memory and crashed with singlecpuio |
245 |
1872 180 180 1 wes node ran out of memory and crashed with singlecpuio |
246 |
1872 180 180 1 wes 21,377,644kb 3,294,676,080kb node ran out of memory with singlecpuio and bigmem=true:mem=90GB for node 0 |
247 |
1872 180 180 1 san node ran out of memory and crashed with singlecpuio |
248 |
1872 180 180 1 san 11,558,588kb 1,356,676,140kb singlecpuio=.FALSE. |
249 |
2925 144 144 1 san 8,374,668kb 1,538,454,112kb 886MB 892MB singlecpuio=.FALSE. |
250 |
2925 144 144 1 san 27,284,996kb 4,942,949,704kb node ran out of memory and crashed with singlecpuio |
251 |
3328 135 135 1 san rank 0 run out of memory |
252 |
3328 135 135 1 san some random node run out of memory (full node for rank 0) |
253 |
4212 120 120 1 san node ran out of memory |
254 |
5200 108 108 1 san node ran out of memory |
255 |
|
256 |
============= |
257 |
|
258 |
2 = 2 |
259 |
3 = 3 |
260 |
2*2 = 4 |
261 |
5 = 5 |
262 |
2*3 = 6 |
263 |
2*2*2 = 8 |
264 |
3*3 = 9 |
265 |
2*5 = 10 |
266 |
2*2*3 = 12 |
267 |
3*5 = 15 |
268 |
2*2*2*2 = 16 |
269 |
2*3*3 = 18 |
270 |
2*2*5 = 20 |
271 |
2*2*2*3 = 24 |
272 |
3*3*3 = 27 |
273 |
2*3*5 = 30 |
274 |
2*2*3*3 = 36 |
275 |
2*2*2*5 = 40 |
276 |
3*3*5 = 45 |
277 |
2*2*2*2*3 = 48 * 45 |
278 |
2*3*3*3 = 54 * 40 |
279 |
2*2*3*5 = 60 * 36 |
280 |
2*2*2*3*3 = 72 * 30 |
281 |
2*2*2*2*5 = 80 * 27 |
282 |
2*3*3*5 = 90 * 24 |
283 |
2*2*3*3*3 = 108 * 20 |
284 |
2*2*2*3*5 = 120 * 18 |
285 |
3*3*3*5 = 135 * 16 |
286 |
2*2*2*2*3*3 = 144 * 15 |
287 |
2*2*3*3*5 = 180 * 12 |
288 |
2*2*2*3*3*3 = 216 * 10 |
289 |
2*2*2*2*3*5 = 240 * 9 |
290 |
2*3*3*3*5 = 270 * 8 |
291 |
2*2*2*3*3*5 = 360 * 6 |
292 |
2*2*2*2*3*3*3 = 432 * 5 |
293 |
2*2*3*3*3*5 = 540 * 4 |
294 |
2*2*2*2*3*3*5 = 720 * 3 |
295 |
2*2*2*3*3*3*5 = 1080 * 2 |