ATS-6 Benchmarks. ATTENTION: This page is a work in progress and nothing is considered to be final

Getting Started

• 1. Introduction
  • 1.1. Approvals

ModSim

• 1. Benchmark Overview
  • 1.1. Run Rules Synopsis
• 2. Tier 1
  • 2.1. Kripke
    • 2.1.1. Purpose
    • 2.1.2. Problems
    • 2.1.3. Figure of Merit
    • 2.1.4. Source code modifications
    • 2.1.5. Building
    • 2.1.6. Running
    • 2.1.7. Validation
    • 2.1.8. Memory Usage
    • 2.1.9. Baseline on El Capitan
    • 2.1.10. Performance Projection
  • 2.2. Laghos
    • 2.2.1. Purpose
    • 2.2.2. Characteristics
    • 2.2.3. Source code modifications
    • 2.2.4. Building
    • 2.2.5. Running
    • 2.2.6. Validation
    • 2.2.7. Memory Usage
    • 2.2.8. Baseline on El Capitan
    • 2.2.9. Performance Projection
    • 2.2.10. References
  • 2.3. RAJA Performance Suite
    • 2.3.1. Purpose
    • 2.3.2. Problems
    • 2.3.3. Figure of Merit
    • 2.3.4. Source code modifications
    • 2.3.5. Building
    • 2.3.6. Running
    • 2.3.7. Validation
    • 2.3.8. Memory Usage
    • 2.3.9. Baseline on El Capitan
    • 2.3.10. Baseline on NVIDIA H100
    • 2.3.11. Performance Projection
    • 2.3.12. References
  • 2.4. Branson
    • 2.4.1. Purpose
    • 2.4.2. Characteristics
    • 2.4.3. Source code modifications
    • 2.4.4. Building
    • 2.4.5. Running
    • 2.4.6. Validation
    • 2.4.7. Memory Usage
    • 2.4.8. Throughput
    • 2.4.9. References
  • 2.5. SPARTA
    • 2.5.1. Purpose
    • 2.5.2. Problems
    • 2.5.3. Source code modifications
    • 2.5.4. System Information
    • 2.5.5. Building
    • 2.5.6. Running
    • 2.5.7. Validation
    • 2.5.8. Memory Usage
    • 2.5.9. Baseline on El Capitan
    • 2.5.10. Performance Projection
    • 2.5.11. References
  • 2.6. MPI: Phloem
    • 2.6.1. Purpose
    • 2.6.2. Problems
    • 2.6.3. Figures of Merit
    • 2.6.4. Source code modifications
    • 2.6.5. Building
    • 2.6.6. Running
    • 2.6.7. Validation
    • 2.6.8. Memory Usage
    • 2.6.9. Baseline on El Capitan
    • 2.6.10. Performance Projections
  • 2.7. MPI: OMB
    • 2.7.1. Purpose
    • 2.7.2. Characteristics
    • 2.7.3. Source code modifications
    • 2.7.4. Building
    • 2.7.5. Running
    • 2.7.6. Scaling on El Capitan
    • 2.7.7. References
  • 2.8. MPI: SMB Message Rate
    • 2.8.1. Purpose
    • 2.8.2. Characteristics
    • 2.8.3. Source Code Modifications
    • 2.8.4. Building
    • 2.8.5. Verification
    • 2.8.6. Running
    • 2.8.7. Results
    • 2.8.8. Validation
    • 2.8.9. Example Scalability Results
    • 2.8.10. Strong Scaling on El Capitan
    • 2.8.11. Weak Scaling on El Capitan
    • 2.8.12. References
• 3. Tier 2
  • 3.1. AMG2023
    • 3.1.1. Purpose
    • 3.1.2. Problems
    • 3.1.3. Figure of Merit
    • 3.1.4. Source code modifications
    • 3.1.5. Building
    • 3.1.6. Running
    • 3.1.7. Validation
    • 3.1.8. Memory Usage
    • 3.1.9. Baseline on El Capitan
    • 3.1.10. Performance Projection
    • 3.1.11. References
  • 3.2. LAMMPS ACE
    • 3.2.1. Purpose
    • 3.2.2. Characteristics
    • 3.2.3. Source Code Modifications
    • 3.2.4. System Information
    • 3.2.5. Building
    • 3.2.6. Running
    • 3.2.7. Verification of Results
    • 3.2.8. References
  • 3.3. RAJA Performance Suite Tier 2 Kernels
    • 3.3.1. Problems
    • 3.3.2. Figure of Merit
    • 3.3.3. Baseline on El Capitan
    • 3.3.4. Baseline on NVIDIA H100
  • 3.4. Branson
    • 3.4.1. Problem
  • 3.5. Remhos
    • 3.5.1. Purpose
    • 3.5.2. Characteristics
    • 3.5.3. Source code modifications
    • 3.5.4. Building
    • 3.5.5. Running
    • 3.5.6. Validation
    • 3.5.7. Example Scalability Results
    • 3.5.8. Memory Usage
    • 3.5.9. Strong Scaling on El Capitan
    • 3.5.10. Weak Scaling on El Capitan
    • 3.5.11. References
  • 3.6. MiniEM
    • 3.6.1. Purpose
    • 3.6.2. Characteristics
    • 3.6.3. Source code modifications
    • 3.6.4. Building
    • 3.6.5. Running
    • 3.6.6. Validation
    • 3.6.7. Example Scalability Results
    • 3.6.8. Memory Usage
    • 3.6.9. Strong Scaling on El Capitan
    • 3.6.10. Weak Scaling on El Capitan
    • 3.6.11. References
• 4. Appendix: Building Libraries
  • 4.1. Adiak
  • 4.2. Caliper
  • 4.3. fmt
  • 4.4. camp
  • 4.5. Umpire
  • 4.6. RAJA
  • 4.7. CHAI
  • 4.8. Hypre
  • 4.9. MFEM

SI4Sim

• 1. Benchmark Overview
  • 1.1. Run Rules Synopsis
• 2. Tier 1
  • 2.1. ScaFFold
    • 2.1.1. Purpose
    • 2.1.2. Problem
    • 2.1.3. Figure of Merit
    • 2.1.4. Source code modifications
    • 2.1.5. Building
    • 2.1.6. Running
    • 2.1.7. Validation
    • 2.1.8. Memory usage
    • 2.1.9. Baseline on El Capitan
    • 2.1.10. Performance projections
    • 2.1.11. Appendix
  • 2.2. MLPerf Llama 3.1 405B training
    • 2.2.1. Purpose
    • 2.2.2. Characteristics
    • 2.2.3. Source code modifications
    • 2.2.4. Building
    • 2.2.5. Running
    • 2.2.6. Validation
    • 2.2.7. Example Scalability Results
    • 2.2.8. Memory Usage
    • 2.2.9. Strong Scaling on El Capitan
    • 2.2.10. Weak Scaling on El Capitan
    • 2.2.11. References
  • 2.3. Collectives
    • 2.3.1. MPI: OMB
    • 2.3.2. CCL Tests

Storage

• 1. Benchmark Overview
  • 1.1. IO500
    • 1.1.1. Purpose
    • 1.1.2. Characteristics
    • 1.1.3. Source code modifications
    • 1.1.4. Building
    • 1.1.5. Running
    • 1.1.6. Validation & Reporting
    • 1.1.7. Example Scalability Results
  • 1.2. MLPerf Storage
    • 1.2.1. Purpose
    • 1.2.2. Characteristics
    • 1.2.3. Source code modifications
    • 1.2.4. Installation
    • 1.2.5. Running
    • 1.2.6. Validation & Reporting
    • 1.2.7. Example Results

Appendices

• 1. Project Documentation
  • 1.1. Contributing
  • 1.2. Dependencies
  • 1.3. Building
  • 1.4. Plot Generation and Table Inclusion
    • 1.4.1. Creating a Table from CSV
    • 1.4.2. Creating a Figure with Gnuplot
    • 1.4.3. Creating a Figure with Matplotlib
    • 1.4.4. Ignore Relevant Files
  • 1.5. Build Script
• 2. Reference Systems
  • 2.1. El Capitan
    • 2.1.1. Node Architecture
  • 2.2. CTS-2
    • 2.2.1. Node Architecture
  • 2.3. H100 System
    • 2.3.1. Node Architecture