jubench-picongpu
JUPITER Benchmark Suite: PIConGPU Benchmark
Science Score: 62.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
✓CITATION.cff file
Found CITATION.cff file -
✓codemeta.json file
Found codemeta.json file -
✓.zenodo.json file
Found .zenodo.json file -
○DOI references
-
✓Academic publication links
Links to: zenodo.org -
○Academic email domains
-
✓Institutional organization owner
Organization fzj-jsc has institutional domain (www.fz-juelich.de) -
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (9.6%) to scientific vocabulary
Repository
JUPITER Benchmark Suite: PIConGPU Benchmark
Basic Info
- Host: GitHub
- Owner: FZJ-JSC
- License: mit
- Language: Jupyter Notebook
- Default Branch: main
- Size: 198 KB
Statistics
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
- Releases: 2
Metadata Files
README.md
JUPITER Benchmark Suite: PIConGPU
This benchmark is part of the JUPITER Benchmark Suite. See the repository of the suite for some general remarks.
This repository contains the PIConGPU benchmark. DESCRIPTION.md contains details for compilation, execution, and evaluation.
The source code of PIConGPU is included in the ./src/ subdirectory as a submodule from the upstream PIConGPU repository at github.com/ComputationalRadiationPhysics/picongpu.
Quickstart
```
Obtain the benchmark if not present
git clone --recursive $REPO
Run the benchmark using JUBE
jube run benchmark/jube/benchmark.xml -t {baseline,highlarge,highmedium,high_small}
``` This will obtain the required sources from GitHub, perform a full build and run the benchmark. After the benchmark runs are complete, one can analyse the benchmark runs and generate a results table by executing the following,
``` jube analyse benchmark/jube/run --id XYZ jube result benchmark/jube/run --id XYZ
```
Results
Example results are given in the following
Baseline
The baseline test uses 8 nodes on the JUWELS booster with 4 GPUs per node and uses a grid size -g 1024 1024 512. Following is the example output on the JUWELS booster using "jube run benchmark.xml -t baseline"
|version |system |nnodes |nconfig |gridconfig |simtime| |-------------|-------------|--------|----------------|----------------|-------| |26cbd45b2c9 |juwelsbooster|8 |-d 4 4 2 |-g 1024 1024 512|466.155|
Large Scale
This is the second mode of the benchmark, designed to run at a scale of around 1000 PFLOP/s. Below is the example out sourced from JUWELS Booster and its validation data.
|version |system |nnodes|nconfig |gridconfig |simtime| |-----------|-------------|------|-----------|-----------------|-------| |26cbd45b2c9|juwelsbooster|256 |-d 16 8 8 |-g 4096 2048 2048|466.567| |26cbd45b2c9|juwelsbooster|384 |-d 16 16 6 |-g 4096 3072 2048|470.815| |26cbd45b2c9|juwelsbooster|512 |-d 16 16 8 |-g 4096 4096 2048|466.618| |26cbd45b2c9|juwelsbooster|640 |-d 16 16 10|-g 4096 4096 2560|466.785|
Owner
- Name: Jülich Supercomputing Centre
- Login: FZJ-JSC
- Kind: organization
- Location: Germany
- Website: https://www.fz-juelich.de/en/ias/jsc
- Twitter: fzj_jsc
- Repositories: 29
- Profile: https://github.com/FZJ-JSC
Jülich Supercomputing Centre provides HPC resources and expertise. Part of Forschungszentrum Jülich.
Citation (CITATION.cff)
cff-version: 1.2.0
title: "JUPITER Benchmark Suite: PIConGPU"
message: >-
In addition to citing this benchmark repository, please also cite either the JUPITER Benchmark Suite or the accompanying SC24 paper
authors:
- given-names: Ujjwal
family-names: Sinha
affiliation: Forschungszentrum Jülich, Jülich Supercomputing Centre
orcid: 'https://orcid.org/0000-0002-4609-0940'
- given-names: Jayesh
family-names: Badwaik
affiliation: Forschungszentrum Jülich, Jülich Supercomputing Centre
orcid: 'https://orcid.org/0000-0002-5252-8179'
- given-names: Andreas
family-names: Herten
affiliation: Forschungszentrum Jülich, Jülich Supercomputing Centre
orcid: 'https://orcid.org/0000-0002-7150-2505'
- given-names: Sebastian
family-names: Achilles
affiliation: Forschungszentrum Jülich, Jülich Supercomputing Centre
orcid: 'https://orcid.org/0000-0002-1943-6803'
abstract: "The PIConGPU benchmark of the JUPITER Benchmark Suite"
identifiers:
- type: doi
value: 10.5281/zenodo.12788381
description: Version-agnostic Zenodo Identifier
repository-code: 'https://github.com/FZJ-JSC/jubench-picongpu/'
license: MIT
date-released: '2024-07-20'
references:
- title: "JUPITER Benchmark Suite"
type: software
doi: 10.5281/zenodo.12737073