\\
**[[cluster:0|Back]]**


===== Amber GPU Testing (EC) =====


We are interested in benchmarking the serial, MPI, cuda and cuda.MPI versions of pmemd.

==== Results ====

  * Verified the MPI threads and GPU invocations
  * Verified the output data
  * pmemd.cuda.MPI errors
  * Script used is listed at end of this page

^ PMEMD implementation of SANDER, Release 12  ^
|Minimzing the system with 25 kcal/mol restraints on protein, 500 steps of steepest descent and 500 of conjugated gradient - Surjit Dixit problem set|

^CPU Jobs (1,000 steps)^ Serial ^ -np 2 ^ -np 4 ^ -np 8 ^ -np 16 ^ -np 24 ^ -np 32 ^
|Wall Time (secs)|  211  |  120  |  64  |  35  |  29  |  26  |  33  |

  * MPI speedup near -np 24 is 8x serial

^GPU Jobs^ Serial ^ -np 2 ^ -np 4 ^ -np 8 ^ -np 16 ^ -np 24 ^ -np 32 ^
|Wall Time (secs)|  12  |    |    |    |    |    |    |

  * GPU serial speedup is 17.5x CPU serial performance and outperforms MPI by at least 2x
  * GPU parallel unable to measure

^AMBER BENCHMARK EXAMPLES^^^^^^
|JAC_PRODUCTION_NVE - 23,558 atoms PME||||||
|  16 cpu cores  | 1xK20  |  2xK20   |  3xK20  | 4xK20  |  measure  |
|  12.87  |  80.50  |  88.76  |  103.09  |  122.45  |  ns/day  |
|  6713.99  |  1073.23  |  973.45  |  838.09  |  705.61  |  seconds/ns  |
|FACTOR_IX_PRODUCTION_NVE - 90,906 atoms PME||||||
|  16 cpu cores  | 1xK20  |  2xK20   |  3xK20  | 4xK20  |  measure  |
|  3.95  |  22.25  |  27.47  |  32.56   |  39.52   |  ns/day  |
|  21865.59  |  3883.38  |  3145.32  |  2653.65  |  2186.28  |  seconds/ns  |
|CELLULOSE_PRODUCTION_NVE - 408,609 atoms PME||||||
|  16 cpu cores  | 1xK20  |  2xK20   |  3xK20  | 4xK20  |  measure  |
|  0.91   |  5.40  |  6.44  |  7.51   |   8.85  |  ns/day  |
|  95235.87  |  15986.42  |  13406.15  |  11509.28  |  9768.23   |  seconds/ns  |
|NUCLEOSOME_PRODUCTION - 25,095 atoms GB||||||
|  16 cpu cores  | 1xK20  |  2xK20   |  3xK20  | 4xK20  |  measure  |
|  0.06   |  2.79  |  3.65  |  3.98   |  ???   |  ns/day  |
|  1478614.67  |  31007.58  |  23694.29  |  21724.33  |  ???   |  seconds/ns  |


  * 5-6x performance speed ups using one GPU versus 16 CPU cores
  * 9-10x perrformance speedups using four GPUs versus 16 CPU cores


==== Setup ====

First we get some CPU based data.

<code>

# serial run of pmemd
nohup $AMBERHOME/bin/pmemd -O -i mdin -o mdout -p prmtop \
-c inpcrd -r restrt -x mdcrd </dev/null &

# parallel run, note that you will need create the machinefile
# if -np=4 it would would contain 4 lines with the string 'localhost'...does not work, use hostname
mpirun --machinefile=nodefile -np 4 $AMBERHOME/bin/pmemd.MPI \
-O -i mdin -o mdout -p prmtop \
-c inpcrd -r restrt -x mdcrd </dev/null &

</code>


The following script should be in your path ... located in ~/bin

You need to allocate one or more GPUs for your cuda runs.

<code>
node2$ gpu-info
====================================================
Device  Model           Temperature     Utilization
====================================================
0       Tesla K20       27 C             0 %
1       Tesla K20       28 C             0 %
2       Tesla K20       27 C             0 %
3       Tesla K20       30 C             0 %
====================================================
</code>


Next we need to expose these GPUs to pmemd ... 

<code>

# expose one
export CUDA_VISIBLE_DEVICES="0"

# serial run of pmemd.cuda
nohup $AMBERHOME/bin/pmemd.cuda -O -i mdin -o mdout -p prmtop \
-c inpcrd -r restrt -x mdcrd </dev/null &

# parallel run, note that you will need create the machinefile
# if -np=4 it would could contain 4 lines with the string 'localhost'
mpirun --machinefile=nodefile -np 4 $AMBERHOME/bin/pmemd.cuda.MPI \
-O -i mdin -o mdout -p prmtop \
-c inpcrd -r restrt -x mdcrd </dev/null &

</code>


You may want to try to run your pmemd problem across multiple GPUs if problem set is large enough.

<code>

# expose multiple (for serial or parallel runs)
export CUDA_VISIBLE_DEVICES="0,2"

</code>

==== Script ====

<code>

[TestDriveUser0@K20-WS]$ cat run
#!/bin/bash
rm -rf err out logfile mdout restrt mdinfo

echo CPU serial
pmemd -O -i inp/mini.in -p 1g6r.cd.parm \
 -c 1g6r.cd.randions.crd.1 -ref 1g6r.cd.randions.crd.1 2>&1
cp mdout 1core.serial.log

echo CPU parallel 2,4,8,16 /usr/local/mpich2-1.4.1p1/bin/mpirun
for i in 2 4 8 16 24 32
do
echo $i
mpirun --machinefile=nodefile$i -np $i pmemd.MPI -O -i inp/mini.in -p 1g6r.cd.parm \
 -c 1g6r.cd.randions.crd.1 -ref 1g6r.cd.randions.crd.1 2>&1
cp mdout ${i}core.parallel.log 
done

echo GPU serial
export CUDA_VISIBLE_DEVICES="2"
pmemd.cuda -O -i inp/mini.in -p 1g6r.cd.parm \
 -c 1g6r.cd.randions.crd.1 -ref 1g6r.cd.randions.crd.1 2>&1
cp mdout 1gpu.serial.log

echo GPU parallel 2,4,8,16 /usr/local/mpich2-1.4.1p1/bin/mpirun
export CUDA_VISIBLE_DEVICES="2"
for i in 2
do
echo $i
mpirun --machinefile=nodefile$i -np $i pmemd.cuda.MPI -O -i inp/mini.in -p 1g6r.cd.parm \
 -c 1g6r.cd.randions.crd.1 -ref 1g6r.cd.randions.crd.1 2>&1
cp mdout ${i}gpu.parallel.log 
done

</code>

\\
**[[cluster:0|Back]]**