Format Comparison
=================

This page presents benchmark results comparing TACO with other popular molecular dynamics trajectory formats.

The benchmark compares:

- **TACO**: Our format with different full-frame interval (FFI) settings
- **ASE trajectory**: Standard format from Atomic Simulation Environment
- **XYZ**: Simple text-based format widely used in computational chemistry
- **XTC**: GROMACS compressed trajectory format (requires MDAnalysis)

Benchmark Results
-----------------

The following results were generated using a synthetic trajectory with 500 atoms and 200 frames.

File Size Comparison
~~~~~~~~~~~~~~~~~~~~

.. figure:: _static/benchmark_results/benchmark_500atoms_200frames_filesize.svg
   :width: 80%
   :align: center
   
   File sizes of different trajectory formats.

Write Speed Comparison
~~~~~~~~~~~~~~~~~~~~~~

.. figure:: _static/benchmark_results/benchmark_500atoms_200frames_writespeed.svg
   :width: 80%
   :align: center
   
   Write speeds of different trajectory formats.

Read Speed Comparison
~~~~~~~~~~~~~~~~~~~~~

.. figure:: _static/benchmark_results/benchmark_500atoms_200frames_readspeed.svg
   :width: 80%
   :align: center
   
   Read speeds of different trajectory formats.

Accuracy Metrics
~~~~~~~~~~~~~~~~

Root Mean Square Error (RMSE)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

.. figure:: _static/benchmark_results/benchmark_500atoms_200frames_rmse.svg
   :width: 80%
   :align: center
   
   RMSE for different trajectory formats.

Mean Absolute Error (MAE)
^^^^^^^^^^^^^^^^^^^^^^^^^

.. figure:: _static/benchmark_results/benchmark_500atoms_200frames_mae.svg
   :width: 80%
   :align: center
   
   MAE for different trajectory formats.

Maximum Error
^^^^^^^^^^^^^

.. figure:: _static/benchmark_results/benchmark_500atoms_200frames_maxerror.svg
   :width: 80%
   :align: center
   
   Maximum error for different trajectory formats.

Numerical Results
-----------------

.. raw:: html
   :file: _static/benchmark_results/benchmark_table.html

Run Your Own Benchmarks
-----------------------

You can run your own benchmark comparisons using the ``format_comparison.py`` script in the docs directory:

.. code-block:: bash

   python format_comparison.py --atoms 1000 --frames 500 --output results_dir

This will generate plots and a CSV file with the benchmark results.