Timemory SC2021 HPCTools Tutorial

Overview

This repository is intended to provide walk-through tutorials for using timemory (Github) (Documentation) for software monitoring (performance analysis, debugging, logging) with Python. The pre-built tools and libraries primarily focus on performance analysis.

Example Applications

Visit the apps/README for information on the examples.

Containers

CPU Only

Configuration:

• Python 3.8
• Caliper
• GOTCHA
• gperftools
• LIKWID
• PAPI
• Perfetto
• Dyninst (runtime instrumentation and binary rewrite)
• Hatchet
• line-profiler

Docker

docker pull jrmadsen/timemory:hpctools-sc2021-cpu
docker run -it --rm --cap-add=SYS_ADMIN -v $PWD:/home/timemory-tutorials jrmadsen/timemory:hpctools-sc2021-cpu /bin/bash -l

Shifter (Cori-Haswell, Cori-KNL)

salloc -q interactive -C haswell -A ntrain -N 1 -c 8 -t 01:00:00
srun --pty shifter --image=jrmadsen/timemory:hpctools-sc2021-cpu -- /bin/bash

CUDA Support

Configuration:

• Python 3.8
• CUDA 11.4
• Caliper
• GOTCHA
• gperftools
• LIKWID
• PAPI
• Perfetto
• Dyninst (runtime instrumentation and binary rewrite)
• Hatchet
• line-profiler

Docker

docker pull jrmadsen/timemory:hpctools-sc2021-gpu
docker run -it --rm --cap-add=SYS_ADMIN -v $PWD:/home/timemory-tutorials jrmadsen/timemory:hpctools-sc2021-gpu /bin/bash -l

Shifter (Cori-GPU)

salloc -A ntrain -N 1 -C gpu -G 1 -n 1 -c 20 -t 01:00:00
srun --pty shifter --image=jrmadsen/timemory:hpctools-sc2021-gpu -- /bin/bash