Hi,
I am running some test simulations using OpenMM and plan to run GaMD later to accelerate the simulations. I wanted to know if the platform functionality is currently limited to 'CUDA', 'Reference', and 'CPU'. As we run our MD simulations on Frontier HPC in Oak Ridge National Lab, the HIP platform is what we need for the GPU support in GaMD-OpenMM. Has it already been implemented? Using the help command to gamdRunner, I didn't see the mention of HIP anywhere.
(base) [skhatua@login11.frontier alanine-dipeptide]$ gamdRunner --help
usage: gamdRunner [-h] [-r] [-p PLATFORM] [-d DEVICE_INDEX] [-D] [-o OUTPUT_DIRECTORY]
INPUT_FILE_TYPE INPUT_FILE
This runner program ties together the different components and provides a command line interface to
run basic GaMD simulations under OpenMM.
positional arguments:
INPUT_FILE_TYPE The type of file being provided. Available options are: 'xml', ... More to
come later
INPUT_FILE name of input file for GaMD calculation. only XML format is currently
preferred.
options:
-h, --help show this help message and exit
-r, --restart Restart simulation from backup checkpoint in input file
-p PLATFORM, --platform PLATFORM
Define the platform that will run the simulations. Default is 'CUDA', but
other options include: 'Reference', 'CPU', and 'OpenCL'.
-d DEVICE_INDEX, --device_index DEVICE_INDEX
modify which device_index to run the simulation on. For example, the number 0
or 1 would suffice. To run on multiple GPU indices, simply enter comma
separated indices. Example: '0,1'. If a value is not supplied, the value '0'
will be used by default.
-D, --debug Whether to start the run in debug mode.
-o OUTPUT_DIRECTORY, --output OUTPUT_DIRECTORY
Provides the capability to override the configuration value in your
configuration for the output directory.
Best regards,
Satyajit Khatua
Hi,
I am running some test simulations using OpenMM and plan to run GaMD later to accelerate the simulations. I wanted to know if the platform functionality is currently limited to 'CUDA', 'Reference', and 'CPU'. As we run our MD simulations on Frontier HPC in Oak Ridge National Lab, the HIP platform is what we need for the GPU support in GaMD-OpenMM. Has it already been implemented? Using the help command to gamdRunner, I didn't see the mention of HIP anywhere.
(base) [skhatua@login11.frontier alanine-dipeptide]$ gamdRunner --help
usage: gamdRunner [-h] [-r] [-p PLATFORM] [-d DEVICE_INDEX] [-D] [-o OUTPUT_DIRECTORY]
INPUT_FILE_TYPE INPUT_FILE
This runner program ties together the different components and provides a command line interface to
run basic GaMD simulations under OpenMM.
positional arguments:
INPUT_FILE_TYPE The type of file being provided. Available options are: 'xml', ... More to
come later
INPUT_FILE name of input file for GaMD calculation. only XML format is currently
preferred.
options:
-h, --help show this help message and exit
-r, --restart Restart simulation from backup checkpoint in input file
-p PLATFORM, --platform PLATFORM
Define the platform that will run the simulations. Default is 'CUDA', but
other options include: 'Reference', 'CPU', and 'OpenCL'.
-d DEVICE_INDEX, --device_index DEVICE_INDEX
modify which device_index to run the simulation on. For example, the number 0
or 1 would suffice. To run on multiple GPU indices, simply enter comma
separated indices. Example: '0,1'. If a value is not supplied, the value '0'
will be used by default.
-D, --debug Whether to start the run in debug mode.
-o OUTPUT_DIRECTORY, --output OUTPUT_DIRECTORY
Provides the capability to override the configuration value in your
configuration for the output directory.
Best regards,
Satyajit Khatua