Skip to content

updated timemory metric tracking around fit#15

Open
jbalma wants to merge 1 commit intosparticlesteve:timemoryfrom
jbalma:timemory
Open

updated timemory metric tracking around fit#15
jbalma wants to merge 1 commit intosparticlesteve:timemoryfrom
jbalma:timemory

Conversation

@jbalma
Copy link

@jbalma jbalma commented Jul 31, 2020

Updated train.py with timemory tracking for communication, I/O, CPU utilization and walltime

#python setup.py install <ARGS> -- <CMAKE_ARGS>
#python setup.py install --help
python -m pip install scikit-build
export CMAKE_CXX_COMPILER=$MPI_CXX

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just a side note, this env variable isn't used by CMake. You can use either -DCMAKE_CXX_COMPILER=$MPI_CXX or export CXX=$MPI_CXX. The former will force the compiler to change if there is an existing build and the CXX environment variable is used if there is no existing build and the latter is not specified on the command line.

Also, out of curiosity, is there a particular reason you are using the MPI compiler wrappers here? The find_package for MPI in CMake should take care of that but if you've had build errors, please lmk.

timemory.settings.flat_profile = False
timemory.settings.timeline_profile = False

components = ['wall_clock','peak_rss','read_bytes','written_bytes']

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You should add in the 'user_global_bundle' here. It will essentially be zero overhead and it will allow you to add other components on-the-fly without and modifications to the script, e.g. export TIMEMORY_GLOBAL_COMPONENTS="papi_vector, virtual_memory, gperftools_cpu_profiler"

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants