| 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 -M email |
| 3 |
qsub -I -q normal -l select=103:ncpus=20:model=ivy,walltime=8:00:00 -m abe -M email |
| 4 |
qsub -I -q long -l select=300:ncpus=20:model=ivy,walltime=120:00:00 -m abe -M email |
| 5 |
qsub -I -q long -l select=103:ncpus=20:model=ivy,min_walltime=30:00,max_walltime=120:00:00 -m abe -M email |
| 6 |
|
| 7 |
############################# |
| 8 |
# 60x60x10882 configuration |
| 9 |
|
| 10 |
qsub -I -q long -l select=600:ncpus=20:model=ivy,min_walltime=30:00,max_walltime=120:00:00 -m abe -M menemenlis@me.com |
| 11 |
module purge |
| 12 |
module load comp-intel/2012.0.032 netcdf/4.0 |
| 13 |
module use -a ~kjtaylor/modulefiles |
| 14 |
module load sles11sp3/mpt-2.10-nasa201311271217 |
| 15 |
cd ~/llc_2160 |
| 16 |
cvs co MITgcm_code |
| 17 |
cvs co MITgcm_contrib/llc_hires/llc_4320 |
| 18 |
cd MITgcm |
| 19 |
mkdir build run |
| 20 |
lfs setstripe -c -1 run |
| 21 |
cd build |
| 22 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code/SIZE.h_60x60_10882 SIZE.h |
| 23 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code-async/readtile_mpiio.c . |
| 24 |
emacs readtile_mpiio.c |
| 25 |
tileSizeX = 60; |
| 26 |
tileSizeY = 60; |
| 27 |
../tools/genmake2 -of \ |
| 28 |
../../MITgcm_contrib/llc_hires/llc_2160/code-async/linux_amd64_ifort+mpi_ice_nas -mpi -mods \ |
| 29 |
'../../MITgcm_contrib/llc_hires/llc_2160/code ../../MITgcm_contrib/llc_hires/llc_2160/code-async' |
| 30 |
make depend |
| 31 |
make -j 16 |
| 32 |
|
| 33 |
cd ~/llc_2160/MITgcm/run |
| 34 |
cp ../build/mitgcmuv mitgcmuv_60x60x10882 |
| 35 |
ln -sf /nobackup/dmenemen/tarballs/llc_2160/run_template/* . |
| 36 |
ln -sf /nobackup/dmenemen/forcing/ECMWF_operational/* . |
| 37 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/* . |
| 38 |
mv ../../MITgcm_contrib/llc_hires/llc_2160/input/data.exch2_60x60x10882 data.exch2 |
| 39 |
emacs data |
| 40 |
|
| 41 |
export MPI_BUFS_PER_PROC=1024 |
| 42 |
export MPI_REQUEST_MAX=65536 |
| 43 |
export MPI_GROUP_MAX=1024 |
| 44 |
export MPI_NUM_MEMORY_REGIONS=8 |
| 45 |
export MPI_UNBUFFERED_STDIO=1 |
| 46 |
export MPI_MEMMAP_OFF=1 |
| 47 |
mpiexec -n 12000 ./mitgcmuv_60x60x10882 |
| 48 |
|
| 49 |
tail -f STDOUT.00000 | grep advcfl_W |
| 50 |
|
| 51 |
============== |
| 52 |
|
| 53 |
cd ~/llc_2160 |
| 54 |
cvs co MITgcm_code |
| 55 |
cvs co MITgcm_contrib/llc_hires/llc_2160 |
| 56 |
cd MITgcm |
| 57 |
module purge |
| 58 |
module load comp-intel/2012.0.032 mpi-sgi/mpt.2.08r7 netcdf/4.0 |
| 59 |
mkdir build run |
| 60 |
lfs setstripe -c -1 run |
| 61 |
cd build |
| 62 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/code/SIZE.h_90x90_5004 SIZE.h |
| 63 |
../tools/genmake2 -of \ |
| 64 |
../../MITgcm_contrib/llc_hires/llc_2160/code-async/linux_amd64_ifort+mpi_ice_nas -mpi -mods \ |
| 65 |
'../../MITgcm_contrib/llc_hires/llc_2160/code ../../MITgcm_contrib/llc_hires/llc_2160/code-async' |
| 66 |
make depend |
| 67 |
make -j 16 |
| 68 |
cd ~/llc_2160/MITgcm/run |
| 69 |
ln -sf ../build/mitgcmuv . |
| 70 |
ln -sf /nobackup/dmenemen/tarballs/llc_2160/run_template/* . |
| 71 |
ln -sf /nobackup/dmenemen/forcing/ECMWF_operational/* . |
| 72 |
cp ../../MITgcm_contrib/llc_hires/llc_2160/input/* . |
| 73 |
mv data.exch2_90x90x5004 data.exch2 |
| 74 |
export MPI_NUM_MEMORY_REGIONS=256 |
| 75 |
mpiexec -n 6000 ./mitgcmuv |
| 76 |
|
| 77 |
============== |
| 78 |
|
| 79 |
look at output |
| 80 |
|
| 81 |
for ts=[0 120 600:10:980 1080:120:2280] |
| 82 |
fld=quikread_llc(['Eta.' myint2str(ts,10) '.data'],2160); |
| 83 |
clf,quikplot_llc(fld),caxis([-2.5 2]),thincolorbar |
| 84 |
title(ts) |
| 85 |
pause(.1) |
| 86 |
end |
| 87 |
|
| 88 |
============== |
| 89 |
|
| 90 |
to determine empty tiles: |
| 91 |
grep Empty STDOUT.* |
| 92 |
|
| 93 |
============== |
| 94 |
|
| 95 |
memory requirements: |
| 96 |
nPx sNx sNy nSx cpu node0 total rank0 rankm |
| 97 |
936 180 180 2 san node ran out of memory and crashed with singlecpuio |
| 98 |
1053 240 240 1 san node ran out of memory and crashed with singlecpuio |
| 99 |
1300 216 216 1 san node ran out of memory and crashed with singlecpuio |
| 100 |
1872 180 180 1 wes node ran out of memory and crashed with singlecpuio |
| 101 |
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 |
| 102 |
1872 180 180 1 san node ran out of memory and crashed with singlecpuio |
| 103 |
1872 180 180 1 san 11,558,588kb 1,356,676,140kb singlecpuio=.FALSE. |
| 104 |
2925 144 144 1 san 8,374,668kb 1,538,454,112kb 886MB 892MB singlecpuio=.FALSE. |
| 105 |
2925 144 144 1 san 27,284,996kb 4,942,949,704kb node ran out of memory and crashed with singlecpuio |
| 106 |
3328 135 135 1 san rank 0 run out of memory |
| 107 |
3328 135 135 1 san some random node run out of memory (full node for rank 0) |
| 108 |
4212 120 120 1 san node ran out of memory |
| 109 |
5200 108 108 1 san node ran out of memory |
| 110 |
|
| 111 |
============= |
| 112 |
|
| 113 |
2 = 2 |
| 114 |
3 = 3 |
| 115 |
2*2 = 4 |
| 116 |
5 = 5 |
| 117 |
2*3 = 6 |
| 118 |
2*2*2 = 8 |
| 119 |
3*3 = 9 |
| 120 |
2*5 = 10 |
| 121 |
2*2*3 = 12 |
| 122 |
3*5 = 15 |
| 123 |
2*2*2*2 = 16 |
| 124 |
2*3*3 = 18 |
| 125 |
2*2*5 = 20 |
| 126 |
2*2*2*3 = 24 |
| 127 |
3*3*3 = 27 |
| 128 |
2*3*5 = 30 |
| 129 |
2*2*3*3 = 36 |
| 130 |
2*2*2*5 = 40 |
| 131 |
3*3*5 = 45 |
| 132 |
2*2*2*2*3 = 48 * 45 |
| 133 |
2*3*3*3 = 54 * 40 |
| 134 |
2*2*3*5 = 60 * 36 |
| 135 |
2*2*2*3*3 = 72 * 30 |
| 136 |
2*2*2*2*5 = 80 * 27 |
| 137 |
2*3*3*5 = 90 * 24 |
| 138 |
2*2*3*3*3 = 108 * 20 |
| 139 |
2*2*2*3*5 = 120 * 18 |
| 140 |
3*3*3*5 = 135 * 16 |
| 141 |
2*2*2*2*3*3 = 144 * 15 |
| 142 |
2*2*3*3*5 = 180 * 12 |
| 143 |
2*2*2*3*3*3 = 216 * 10 |
| 144 |
2*2*2*2*3*5 = 240 * 9 |
| 145 |
2*3*3*3*5 = 270 * 8 |
| 146 |
2*2*2*3*3*5 = 360 * 6 |
| 147 |
2*2*2*2*3*3*3 = 432 * 5 |
| 148 |
2*2*3*3*3*5 = 540 * 4 |
| 149 |
2*2*2*2*3*3*5 = 720 * 3 |
| 150 |
2*2*2*3*3*3*5 = 1080 * 2 |
Parent Directory
| 
Revision Log
| 
Revision Graph