ATS Benchmarks Project

Main:

• 1. Introduction
  • 1.1. Benchmark Changes from Crossroads
  • 1.2. Benchmark Overview
  • 1.3. Microbenchmark Overview
  • 1.4. Run Rules Synopsis
  • 1.5. Scaled Single Node Improvement
  • 1.6. SSNI Weights and SSNI problem sizes
  • 1.7. SSNI Baseline
  • 1.8. System Information
    • 1.8.1. ATS-3/Crossroads
    • 1.8.2. ATS-2/Sierra
  • 1.9. Approvals
• 2. Branson
  • 2.1. Purpose
  • 2.2. Source code optimizations
  • 2.3. Characteristics
    • 2.3.1. Problem
    • 2.3.2. Figure of Merit
    • 2.3.3. Problem Sizes
  • 2.4. Building
  • 2.5. Running
    • 2.5.1. AMD Epyc + Nvidia A100
    • 2.5.2. Correctness
    • 2.5.3. Crossroads
    • 2.5.4. AMD Epyc + Nvidia A100
  • 2.6. Multi-node scaling on Crossroads
  • 2.7. References
• 3. AMG2023
  • 3.1. Purpose
  • 3.2. Characteristics
    • 3.2.1. Problems
    • 3.2.2. Figure of Merit
  • 3.3. Source code modifications
  • 3.4. Building
  • 3.5. Running
  • 3.6. Validation
  • 3.7. Example Scalability Results
  • 3.8. Memory Usage
  • 3.9. Strong Scaling on Crossroads
  • 3.10. V-100
  • 3.11. Multi-node scaling on Crossroads
  • 3.12. References
• 4. Parthenon-VIBE
  • 4.1. Purpose
    • 4.1.1. Problem
    • 4.1.2. Figure of Merit
  • 4.2. Building
  • 4.3. Running
    • 4.3.1. Crossroads
    • 4.3.2. Nvidia Grace Hopper
  • 4.4. Multi-node scaling on Crossroads
  • 4.5. Validation
  • 4.6. References
• 5. MLMD
  • 5.1. Purpose
  • 5.2. Characteristics
    • 5.2.1. Problem
    • 5.2.2. Figure of Merit
  • 5.3. Building
    • 5.3.1. Building on Chicoma
    • 5.3.2. Building on Crossroads
  • 5.4. Running
    • 5.4.1. External Files
    • 5.4.2. Model Training
    • 5.4.3. Running the Benchmark
  • 5.5. Results
    • 5.5.1. Training HIPNN Model
    • 5.5.2. Simulation+Inference
    • 5.5.3. Single GPU Throughput Scaling on Chicoma
  • 5.6. References
• 6. UMT
  • 6.1. Purpose
  • 6.2. Characteristics
    • 6.2.1. Problem
    • 6.2.2. Figure of Merit
    • 6.2.3. Source code modifications
  • 6.3. Building
    • 6.3.1. Accessing the Source
    • 6.3.2. Build Requirements
  • 6.4. Running
  • 6.5. Example FOM Results
  • 6.6. Verification of Results
  • 6.7. References
• 7. MiniEM
  • 7.1. Purpose
  • 7.2. Characteristics
    • 7.2.1. Application Version
    • 7.2.2. Problem
    • 7.2.3. Figure of Merit
    • 7.2.4. Correctness Check
    • 7.2.5. Permissable Modifications
  • 7.3. System Information
    • 7.3.1. Sandia National Laboratories’ “ascicgpu030”
  • 7.4. Building
    • 7.4.1. Crossroads
    • 7.4.2. Sandia National Laboratories’ “ascicgpu030”
  • 7.5. Running
    • 7.5.1. Crossroads
    • 7.5.2. Sandia National Laboratories’ “ascicgpu030”
  • 7.6. Verification of Results
    • 7.6.1. Crossroads
    • 7.6.2. Sandia National Laboratories’ “ascicgpu030”
  • 7.7. References
• 8. SPARTA
  • 8.1. Purpose
  • 8.2. Characteristics
    • 8.2.1. Application Version
    • 8.2.2. Problem
    • 8.2.3. Figure of Merit
    • 8.2.4. Correctness
    • 8.2.5. SSNI & SSI
  • 8.3. System Information
  • 8.4. Building
    • 8.4.1. Generic
    • 8.4.2. Crossroads
    • 8.4.3. Sierra
  • 8.5. Running
    • 8.5.1. Crossroads
    • 8.5.2. Sierra
  • 8.6. Verification of Results
    • 8.6.1. Crossroads - Single Node
    • 8.6.2. Sierra - Single Node
    • 8.6.3. Crossroads - Many Nodes
  • 8.7. References

Microbenchmarks:

• 1. Microbenchmark Introduction
  • 1.1. Microbenchmarks
• 2. STREAM
  • 2.1. Purpose
  • 2.2. Characteristics
    • 2.2.1. Problem
    • 2.2.2. Figure of Merit
    • 2.2.3. Run Rules
  • 2.3. Building
  • 2.4. Running
  • 2.5. Example Results
    • 2.5.1. Crossroads
• 3. Spatter
  • 3.1. Purpose
  • 3.2. Characteristics
    • 3.2.1. Problem
    • 3.2.2. Figure of Merit
  • 3.3. Building
    • 3.3.1. Optional Manual Build
  • 3.4. Running
    • 3.4.1. Crossroads
    • 3.4.2. H100
  • 3.5. References
• 4. OSU Microbenchmarks
  • 4.1. Purpose
  • 4.2. Characteristics
    • 4.2.1. Problem
    • 4.2.2. Run Rules
  • 4.3. Building
  • 4.4. Running
  • 4.5. Example Results
    • 4.5.1. Crossroads
• 5. Sandia Microbenchmarks - Message rate
  • 5.1. Purpose
  • 5.2. Characteristics
    • 5.2.1. Problem
    • 5.2.2. Figure of Merit
  • 5.3. Building
  • 5.4. Running
  • 5.5. Results
  • 5.6. References
• 6. DGEMM
  • 6.1. Purpose
  • 6.2. Characteristics
    • 6.2.1. Problem
    • 6.2.2. Figure of Merit
    • 6.2.3. Run Rules
  • 6.3. Building
  • 6.4. Running
  • 6.5. Example Results
    • 6.5.1. Crossroads
• 7. IOR
  • 7.1. Purpose
  • 7.2. Characteristics
    • 7.2.1. Problem
    • 7.2.2. Run Rules
  • 7.3. Building
  • 7.4. Running
    • 7.4.1. Input
  • 7.5. Example Results
    • 7.5.1. Crossroads
• 8. MDTEST
  • 8.1. Purpose
  • 8.2. Characteristics
    • 8.2.1. Problem
    • 8.2.2. Run Rules
  • 8.3. Building
  • 8.4. Running
  • 8.5. Example Results
    • 8.5.1. Crossroads

Appendices:

• 1. Project Documentation
  • 1.1. Dependencies
  • 1.2. Building
  • 1.3. Contributing
  • 1.4. Build Script

Reference Systems:

• 1. Crossroads
  • 1.1. Single-Node Strong Scaling
  • 1.2. References