File: batch_script_mpi_runit_summit_ibm.sh

package info (click to toggle)
superlu-dist 8.2.1%2Bdfsg1-5.1
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid
  • size: 9,576 kB
  • sloc: ansic: 132,713; sh: 3,073; f90: 827; cpp: 505; makefile: 431; fortran: 51; csh: 9
file content (113 lines) | stat: -rw-r--r-- 4,262 bytes parent folder | download | duplicates (3)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
#!/bin/bash
# Bash script to submit many files to Cori/Edison/Queue

#BSUB -P CSC289
#BSUB -W 2:00
#BSUB -nnodes 45
#BSUB -alloc_flags gpumps
#BSUB -J superlu_gpu




EXIT_SUCCESS=0
EXIT_HOST=1
EXIT_PARAM=2

module load essl
module load cmake/3.11.3
module load cuda/10.1.168

CUR_DIR=`pwd`
FILE_DIR=$CUR_DIR/EXAMPLE
INPUT_DIR=$MEMBERWORK/csc289/matrix/HTS
# INPUT_DIR=$CUR_DIR/../EXAMPLE
FILE_NAME=pddrive
FILE=$FILE_DIR/$FILE_NAME


nprows=(  1  )
npcols=(  1  )  
 
for ((i = 0; i < ${#npcols[@]}; i++)); do
NROW=${nprows[i]}
NCOL=${npcols[i]}

# NROW=36
CORE_VAL=`expr $NCOL \* $NROW`
RANK_PER_RS=1
GPU_PER_RANK=0


#PARTITION=debug
PARTITION=regular
LICENSE=SCRATCH
TIME=00:20:00

if [[ $NERSC_HOST == edison ]]
then
  CONSTRAINT=0
fi
if [[ $NERSC_HOST == cori ]]
then
  CONSTRAINT=haswell
fi

for GPU_PER_RANK in  1
do
for NTH in 7
do

RS_VAL=`expr $CORE_VAL / $RANK_PER_RS`
MOD_VAL=`expr $CORE_VAL % $RANK_PER_RS`
if [[ $MOD_VAL -ne 0 ]]
then
  RS_VAL=`expr $RS_VAL + 1`
fi
OMP_NUM_THREADS=$NTH
TH_PER_RS=`expr $NTH \* $RANK_PER_RS`
GPU_PER_RS=`expr $RANK_PER_RS \* $GPU_PER_RANK`

# for MAT in copter2.mtx
 # for MAT in rajat16.mtx
# for MAT in ExaSGD/118_1536/globalmat.datnh
# for MAT in copter2.mtx gas_sensor.mtx matrix-new_3.mtx xenon2.mtx shipsec1.mtx xenon1.mtx g7jac160.mtx g7jac140sc.mtx mark3jac100sc.mtx ct20stif.mtx vanbody.mtx ncvxbqp1.mtx dawson5.mtx 2D_54019_highK.mtx gridgena.mtx epb3.mtx torso2.mtx torsion1.mtx boyd1.bin hvdc2.mtx rajat16.mtx hcircuit.mtx 
# for MAT in copter2.mtx gas_sensor.mtx matrix-new_3.mtx av41092.mtx xenon2.mtx c-71.mtx shipsec1.mtx xenon1.mtx g7jac160.mtx g7jac140sc.mtx mark3jac100sc.mtx ct20stif.mtx vanbody.mtx ncvxbqp1.mtx dawson5.mtx c-59.mtx 2D_54019_highK.mtx gridgena.mtx epb3.mtx torso2.mtx finan512.mtx twotone.mtx torsion1.mtx jan99jac120.mtx boyd1.mtx c-73b.mtx hvdc2.mtx rajat16.mtx hcircuit.mtx 
  # for MAT in matrix05_ntime=2/s1_mat_0_126936.bin
# for MAT in matrix05_ntime=2/s1_mat_0_126936.bin A30_P0/A30_015_0_25356.bin
# for MAT in matrix05_ntime=2/s1_mat_0_126936.bin
# for MAT in ../mathias/Li4244.bin ../mathias/DG_GrapheneDisorder_32768.bin
#for MAT in  copter2.mtx epb3.mtx gridgena.mtx vanbody.mtx shipsec1.mtx dawson5.mtx gas_sensor.mtx rajat16.mtx  
#for MAT in rajat16.mtx
#for MAT in epb3.mtx
#for MAT in ../globalmat118_1536.bin
#for MAT in ../mat_it_001_sml.mtx ../mat_it_001_med.mtx
#for MAT in ../mat_it_001_med.mtx 
#for MAT in ../s1_mat_0_126936.bin
for MAT in copter2.mtx
#for MAT in ../full_1000.rua

# for MAT in A30_P0/A30_015_0_25356.bin
 # for MAT in A64/A64_001_0_1204992.bin
#for MAT in ../big.rua
#for MAT in ../g4.rua
# for MAT in  StocF-1465.bin 
# for MAT in /mathias/DG_GrapheneDisorder_8192.bin /mathias/DNA_715_64cell.bin /mathias/LU_C_BN_C_4by2.bin /mathias/Li4244.bin 
# for MAT in HTS/copter2.mtx 
		
# for MAT in LU_C_BN_C_4by2.bin Li4244.bin atmosmodj.bin Ga19As19H42.bin Geo_1438.bin StocF-1465.bin
  do
    export OMP_NUM_THREADS=$OMP_NUM_THREADS
    mkdir -p ${MAT}_summit
	echo "jsrun -n $RS_VAL -a $RANK_PER_RS -c $TH_PER_RS -g $GPU_PER_RS -b packed:$NTH $FILE -c $NCOL -r $NROW $INPUT_DIR/$MAT | tee ./${MAT}_summit_new_LU/SLU.o_mpi_${NROW}x${NCOL}_OMP_${OMP_NUM_THREADS}_GPU_${GPU_PER_RANK}"
    jsrun -n $RS_VAL -a $RANK_PER_RS -c $TH_PER_RS -g $GPU_PER_RS -b packed:$NTH '--smpiargs=-x PAMI_DISABLE_CUDA_HOOK=1 -disable_gpu_hooks' nvprof --profile-from-start off $FILE -c $NCOL -r $NROW $INPUT_DIR/$MAT | tee ./${MAT}_summit/SLU.o_mpi_${NROW}x${NCOL}_OMP_${OMP_NUM_THREADS}_GPU_${GPU_PER_RANK}
    # jsrun -n $RS_VAL -a $RANK_PER_RS -c $TH_PER_RS -g $GPU_PER_RS -b packed:$NTH '--smpiargs=-x PAMI_DISABLE_CUDA_HOOK=1 -disable_gpu_hooks' nvprof --profile-from-start off --print-gpu-trace --print-api-trace --kernels "dlsum_fmod_inv_cuda" --metrics "achieved_occupancy" $FILE -c $NCOL -r $NROW $INPUT_DIR/$MAT | tee ./${MAT}_summit/SLU.o_mpi_${NROW}x${NCOL}_OMP_${OMP_NUM_THREADS}_GPU_${GPU_PER_RANK}
    # jsrun -n $RS_VAL -a $RANK_PER_RS -c $TH_PER_RS -g $GPU_PER_RS -b packed:$NTH '--smpiargs=-x PAMI_DISABLE_CUDA_HOOK=1 -disable_gpu_hooks' nvprof $FILE -c $NCOL -r $NROW $INPUT_DIR/$MAT | tee ./${MAT}_summit/SLU.o_mpi_${NROW}x${NCOL}_OMP_${OMP_NUM_THREADS}_GPU_${GPU_PER_RANK}													  
  done
#one

done
done
done
exit $EXIT_SUCCESS