8. Using automatic differentiation

• 8.1. Contributors
• 8.2. Introduction and features
• 8.3. Installation
  • 8.3.1. Open source AD Tool Tapenade
    • 8.3.1.1. Prerequisites for Linux or Mac OS X
    • 8.3.1.2. Steps for Mac OS
    • 8.3.1.3. Steps for Linux
    • 8.3.1.4. Prerequisites for Windows
    • 8.3.1.5. Steps for Windows
  • 8.3.2. SICOPOLIS-AD v2
    • 8.3.2.1. GNU GCC Compiler (gfortran+gcc) or Intel Compiler (ifort+icc)
    • 8.3.2.2. Lis (1.4.43 or newer)
    • 8.3.2.3. NetCDF (3.6.x or newer)
    • 8.3.2.4. Downloading SICOPOLIS-AD v2
    • 8.3.2.5. Initial configuration
• 8.4. Running SICOPOLIS-AD v2
• 8.5. Activating variables
• 8.6. Utilities
  • 8.6.1. Continuous Integration (CI)
  • 8.6.2. Code coverage
  • 8.6.3. Automation using Python scripts
    • 8.6.3.1. Automated Normal Mode
    • 8.6.3.2. Automated Finite Differences
    • 8.6.3.3. Automated Tangent Linear Mode
    • 8.6.3.4. Automated Adjoint Mode
    • 8.6.3.5. Input options
    • 8.6.3.6. Reading inputs from a file
    • 8.6.3.7. Validation
• 8.7. Some FAQs
  • 8.7.1. Does the Finite Differences use AD at all?
  • 8.7.2. Why use the Finite Differences when the adjoint is much faster?
• 8.8. Tutorials
  • 8.8.1. Tutorial 1: Writing your own tangent-linear and adjoint code
    • 8.8.1.1. Forward model
    • 8.8.1.2. Tangent linear mode
    • 8.8.1.3. Adjoint mode
  • 8.8.2. Tutorial 2: Using Tapenade for a mountain glacier model
    • 8.8.2.1. Mountain glacier model
    • 8.8.2.2. Forward Model code
    • 8.8.2.3. Finite Differences for validation
    • 8.8.2.4. Tangent Linear model
    • 8.8.2.5. Adjoint Model
    • 8.8.2.6. Driver Routine
    • 8.8.2.7. Combining all compilation commands into a Makefile
    • 8.8.2.8. Results
    • 8.8.2.9. A note on Binomial Checkpointing
  • 8.8.3. Tutorial 3: Validating results of SICOPOLIS adjoint and tangent linear mode with Finite differences
    • 8.8.3.1. Validation using automated Python script
    • 8.8.3.2. Results
• 8.9. Acknowledgements