cannot run the Lammps OpenCl version on the laptop

Dear Lammps Users,

recently I discovered that the new version of Lammps can be built on the machine with the Amd video card.
I also discovered the a short manual of Andrew Jewett on how to do that. The manual is really great and useful.
Andrew, thank you for that.

that helped me a lot and I managed to build the gpu/mpi version of the Lammps. To do that i do

Makefile.mylinux_opencl_amd (560 Bytes)

Makefile.ubuntu_gpu_opencl_my (1.43 KB)

ocl_timer.h (4.34 KB)

Dear Lammps Users,

recently I discovered that the new version of Lammps can be built on the machine with the Amd video card.
I also discovered the a short manual of Andrew Jewett on how to do that. The manual is really great and useful.
Andrew, thank you for that.

http://lammps.sandia.gov/threads/msg36437.html

that helped me a lot and I managed to build the gpu/mpi version of the Lammps.

To do that i do

1. Build the libgpu library using the script
    Makefile.ubuntu_gpu_opencl_amd
    <http://lammps.sandia.gov/threads/binQ4bRDBFQGO.bin>
2. Build the Lammps using the Makefile Makefile.ubuntu_parallel
    <http://lammps.sandia.gov/threads/binnzZOqmnDPe.bin>

The only thing that I changed was the folder , my folder was ./x86 instead of ./x86_64

after building the lammps I got the binary file and I could run a parallel simulation on my machine.
but when I try to run an example from the lammps/examples/gpu

i get an error
mpirun -np 4 ../../src/lmp_ubuntu_gpu_opencl_my -sf gpu < in.gpu.melt.2.5

/
//LAMMPS (30 Sep 2013)//
//
//Please see the log.cite file for references relevant to this simulation//
//
//Please see the log.cite file for references relevant to this simulation//
//
//Please see the log.cite file for references relevant to this simulation//
//
//Lattice spacing in x,y,z = 1.6796 1.6796 1.6796//
//Created orthogonal box = (0 0 0) to (67.1838 67.1838 67.1838)//
// 1 by 1 by 1 MPI processor grid//
//Created 256000 atoms//
//
//--------------------------------------------------------------------------//
//- Using GPGPU acceleration for lj/cut://
//- with 1 proc(s) per device.//
//- with OpenCL Parameters for: GENERIC//
//--------------------------------------------------------------------------//
//GPU 0: Intel(R) Core(TM) i3 CPU M 380 @ 2.53GHz, 4 cores, 3.5 GB, 2.5 GHZ (Single Precision)//
//--------------------------------------------------------------------------//
//
//Initializing GPU and compiling on process 0...Done.//
//Initializing GPUs 0-1 on core 0...Done.//
//
//OpenCL error in file 'geryon/ocl_timer.h' in line 106 : -7.//
//application called MPI_Abort(MPI_COMM_WORLD, -1) - process 0//
//OpenCL error in file 'geryon/ocl_timer.h' in line 106 : -7.//
//application called MPI_Abort(MPI_COMM_WORLD, -1) - process 0//
//
//+++++++++++++++++/

the same thing happens if I use the ../../src/lmp_ubuntu_gpu_opencl_my -sf gpu < in.gpu.melt.2.5 command.

What is wrong with ocl_timer.h??? Please tell me whether you may have any ideas about this?

I am attaching the ocl_timer.h, and makefiles
Thank you,
Vasiliy.

Makefile.mylinux_opencl_amd (561 Bytes)

Makefile.ubuntu_gpu_opencl_my (1.43 KB)

ocl_timer.h (4.34 KB)

Dear Lammps Users,

recently I discovered that the new version of Lammps can be built on the
machine with the Amd video card.
I also discovered the a short manual of Andrew Jewett on how to do that. The
manual is really great and useful.
Andrew, thank you for that.

LAMMPS Molecular Dynamics Simulator

that helped me a lot and I managed to build the gpu/mpi version of the
Lammps.

To do that i do

Build the libgpu library using the script Makefile.ubuntu_gpu_opencl_amd
Build the Lammps using the Makefile Makefile.ubuntu_parallel

The only thing that I changed was the folder , my folder was ./x86 instead
of ./x86_64

after building the lammps I got the binary file and I could run a parallel
simulation on my machine.
but when I try to run an example from the lammps/examples/gpu

i get an error
mpirun -np 4 ../../src/lmp_ubuntu_gpu_opencl_my -sf gpu < in.gpu.melt.2.5

there are two major issues that seem to be wrong here, and both of
which are not a LAMMPS problem.

either you didn't properly compile a parallel version of LAMMPS, or
the mpirun command you use is not compatible with the MPI library that
you use. in any case you *MUST NOT* get all output replicated 4 times
when you run with 4 MPI tasks.

you need to correct this *first* and best do so without GPU support.

second, you appear to have the AMD OpenCL support installed, but not
configured to use the GPU devices, or are using the wrong assignment
in the package gpu command. please keep in mind, that the AMD OpenCL
support library not only exposes a the GPU as OpenCL device but also
the CPU. please run ocl_get_devices first and verify how the OpenCL
devices are enumerated. this second issue is likely the reason for the
timer errors.

axel.