On Mar 2, 2007, at 3:37 PM, Josh Hursey wrote:
> Certainly the end user could specify another file to use in addition
> to the default MCA parameter files (mca_param_files), but then they
> must also specify the other locations that already exist in that path
> (e.g., $HOME/.openmpi/mca-params.conf:$SYSCONFDIR/openmpi-mca-
> params.conf). This is a short cut in a sense, so the end user doesn't
> have to know all of this uglyness every time they want to run a
> benchmark, or ...
This is the main point of AMCA's: user convenience. Yes, you could
supply 20 MCA params on the command line, or set them in the
environment, or add them to your $HOME/.openmpi .conf file, or add an
individual file to your MCA param file list, or ... The --amca param
is effectively syntactic sugar for adding a .conf file to the MCA
param file list. It seemed a whole lot easier to do:
mpirun --amca my_benchmark_settings.conf ...
foo=`ompi_info --param mca base --parsable | grep
mca_param_files:value | cut -d: -f7-`
mpirun --mca mca_base_path my_benchmark_settings.conf:$foo ...
The intent is that various AMCA files can be introduced and installed
around the code base (I have one for the openib BTL for benchmarking,
for example -- settings that you wouldn't use in production, but you
would use with netpipe).
Server Virtualization Business Unit