Create experiment to profile the execution of similarity_calculator and analog_selector

radical-collaboration / hpc-workflows

NSF16514 EarthCube Project - Award Number:1639694

5 stars 0 forks source link

Create experiment to profile the execution of similarity_calculator and analog_selector #73

Closed vivek-bala closed 5 years ago

vivek-bala commented 5 years ago

Identified at least three parameters of the executables across which we want to profile the execution time and memory consumption.

Parameters: subgrid_size(SG), number of threads (NT), search space (SS) T_simcalc = f(SG, NT, SS) M_simcalc = g(SG, NT, SS) T_anasel = h(SG, NT, SS) M_anasel = j(SG, NT, SS)

Find f, g, h, and j when the executables are run on Cheyenne.

There may be more parameters than these three, please feel free to include them as well.

vivek-bala commented 5 years ago

Some plot ideas from F2F meeting:

img_20181209_153303

vivek-bala commented 5 years ago

Profiling to be done with the EnTK script developed in #77

mturilli commented 5 years ago

Profiled single-node. To be extended to multi-node.

mturilli commented 5 years ago

Blocked by uploading dataset to Cheyenne.

Weiming-Hu commented 5 years ago

Uploading data to Cheyenne.

mturilli commented 5 years ago

Block by #83

mturilli commented 5 years ago

Ready to go for the first two stages.

Weiming-Hu commented 5 years ago

The paper on PAnEn implementation and profiling has been submitted to the Journal of Earth Science Informatics. This paper covers profiling of the algorithm on a single node. Next step is to test AnEn with EnTK. But before that, I need make sure that the environment is properly set up (issue #88 and #87 ).