HPC-benchmarks

This repository gathers HPC benchmarks:

• Source code or a script to get it
• SbatchMan Configurations and Jobs YAML files
• Scripts for parsing results and plotting

List of Shared-Memory Benchmarks

• STREAM
• Pointer Chasing
• ThreadsSynchronization

List of Distributed-Memory Benchmarks (estimated runtime)

• Graph500 (HAICGU)
  • 2 nodes, scale 20, edgefactor 64 , ~4min
  • 4 nodes, scale 20, edgefactor 64, ~2min
• HPL (2 hours)
• HPCG (HAICGU)
  • 2 nodes, 128x128x128 grid, min runtime 30s , ~5min
  • 4 nodes, 128x128x128 grid, min runtime 30s , ~5min
• DNNProxies (15 min)

Estimated runtimes refer to experiments up to 8 nodes on a single partition

Please refer to README of the individual folders for more details (WIP)