NPB-BT - Scorep/1.2.3

[jg piccinali]

MPI/GNU

Get the src

• git submodule add https://github.com/eth-cscs/parallel-debuggers.git src
• cd src/npbmz.git/src

Compile

• module swap PrgEnv-cray PrgEnv-gnu # gcc/4.8.2
• module load craype-accel-nvidia35 # cudatoolkit/5.5.20-1.0501.7945.8.2
• module load scorep # 1.2.3
• make clean
• make CLASS=A NPROCS=16 FFLAGS=-O3 F77="scorep --mpi ftn"

Run

• cd ../bin/ # $SCRATCH
• export SCOREP_TOTAL_MEMORY=314572800
• salloc -N1
• aprun -j0 -n16 ./GNU.SANTIS.A.16

Reports

Sampling

• cube scorep-20140520_1735_25575634161358/profile.cubex

Tracing

• export SCOREP_ENABLE_TRACING=true
• vampir821 scorep-20140520_1738_25752056122527/traces.otf2

Troubleshooting

• If aprun -j0 -n16 ./INTEL.SANTIS.A.16 crashes, just rerun

[Score-P] Error: Can't write measurement configuration 
into `scorep-measurement-tmp/scorep.cfg'
[Score-P] Fatal: Assertion 'err == OTF2_SUCCESS' failed
[Score-P] Please report this to [e-mail]. Thank you.
[Score-P] Try also to preserve any generated core dumps.
forrtl: error (76): Abort trap signal

[jg piccinali]

MPI / INTEL

Compile

• module swap PrgEnv-cray PrgEnv-intel # intel/14.0.1
• module load craype-accel-nvidia35 # cudatoolkit/5.5.20-1.0501.7945.8.2
• module load scorep # 1.2.3
• make clean
• make CLASS=A NPROCS=16 FFLAGS=-O3 F77="scorep --mpi ftn"

Run

• cd ../bin/ # $SCRATCH
• export SCOREP_TOTAL_MEMORY=314572800
• salloc -N1
• aprun -j0 -n16 ./INTEL.SANTIS.A.16

Reports

Sampling

• cube scorep-20140520_1800_27078303843413/profile.cubex

Tracing

• export SCOREP_ENABLE_TRACING=true
• vampir821 scorep-20140520_1800_27078303843413/traces.otf2

[jg piccinali]

MPI / PGI

Compile

• module swap PrgEnv-cray PrgEnv-pgi
• module swap pgi pgi/14.4 # pgi/14.4
• module load cudatoolkit # cudatoolkit/5.5.20-1.0501.7945.8.2
• module load scorep # 1.2.3
• make clean
• make CLASS=A NPROCS=16 FFLAGS=-O3 F77="scorep --mpi ftn"

Run

• cd ../bin/ # $SCRATCH
• export SCOREP_TOTAL_MEMORY=314572800
• salloc -N1
• aprun -j0 -n16 ./PGI.SANTIS.A.16

Reports

Sampling

• cube scorep-20140520_1816_27995980213431/profile.cubex

Tracing

• export SCOREP_ENABLE_TRACING=true
• aprun -j0 -n16 ./PGI.SANTIS.A.16
• vampir821 scorep-20140520_1816_27995980213431/traces.otf2

[jg piccinali]

MPI / CCE

Compile

• module load PrgEnv-cray # cce/8.2.6
• module load craype-accel-nvidia35 # cudatoolkit/5.5.20-1.0501.7945.8.2
• module load scorep # 1.2.3
• make clean
• make CLASS=A NPROCS=16 FFLAGS=-O3 F77="scorep --mpi ftn"

Run

• cd ../bin/ # $SCRATCH
• export SCOREP_TOTAL_MEMORY=314572800
• salloc -N1
• aprun -j0 -n16 ./CRAY.SANTIS.A.16

Reports

Sampling

• cube scorep-20140520_1834_29081903313549/profile.cubex

Tracing

• export SCOREP_ENABLE_TRACING=true
• aprun -j0 -n16 ./CRAY.SANTIS.A.16
• vampir821 scorep-20140520_1834_29081903313549/traces.otf2