[gmx-users] About gpu

[陈照云]

I have tested gromacs-4.6.1 with k20. But when I run the mdrun, I met some
problems.

1.Configure options are -DGMX_MPI=ON ,-DGMX_DOUBLE=ON -DGMX_GPU=OFF .

But if I run parallely with mpirun, it would get wrong.

"Note: file tpx version 58, software tpx version 83

Fatal error in PMPI_Bcast: Invalid buffer pointer, error stack:

PMPI_Bcast(2011): MPI_Bcast(buf=(nil), count=56, MPI_BYTE, root=0,
MPI_COMM_WORLD) failed

PMPI_Bcast(1919): Null buffer pointer

APPLICATION TERMINATED WITH THE EXIT STRING: Hangup (signal 1)

"

2. Configure options are -DGMX_MPI=ON ,-DGMX_GPU=ON -DGMX_DOUBLE=OFF . But
if I run with gpu, the program would get wrong.

run one process with gpu:

"Reading file topol.tpr, VERSION 4.5.1-dev-20100917-b1d66 (single precision)

Note: file tpx version 73, software tpx version 83

NOTE: GPU(s) found, but the current simulation can not use GPUs

To use a GPU, set the mdp option: cutoff-scheme = Verlet

(for quick performance testing you can use the -testverlet option)

Using 1 MPI process

1 GPU detected on host node11:

#0: NVIDIA Tesla K20c, compute cap.: 3.5, ECC: yes, stat: compatible

Back Off! I just backed up ener.edr to ./#ener.edr.4#

starting mdrun 'Protein'

-1 steps, infinite ps.

Segmentation Fault (core dumped)

run eight processes with gpu:

Reading file topol.tpr, VERSION 4.5.1-dev-20100917-b1d66 (single precision)

Note: file tpx version 73, software tpx version 83

NOTE: GPU(s) found, but the current simulation can not use GPUs

To use a GPU, set the mdp option: cutoff-scheme = Verlet

(for quick performance testing you can use the -testverlet option)

Non-default thread affinity set, disabling internal thread affinity

Using 8 MPI processes

1 GPU detected on host node11:

#0: NVIDIA Tesla K20c, compute cap.: 3.5, ECC: yes, stat: compatible

Back Off! I just backed up ener.edr to ./#ener.edr.6#

starting mdrun 'Protein'

-1 steps, infinite ps.

APPLICATION TERMINATED WITH THE EXIT STRING: Hangup (signal 1)"

Thanks for your help!