Welcome to the cuDF documentation!
==================================
.. figure:: _static/RAPIDS-logo-purple.png
:width: 300px
:align: center
**cuDF** is a Python GPU DataFrame library (built on the `Apache Arrow
`_ columnar memory format) for loading, joining,
aggregating, filtering, and otherwise manipulating data. cuDF also provides a
pandas-like API that will be familiar to data engineers & data scientists, so
they can use it to easily accelerate their workflows without going into
the details of CUDA programming.
``cudf.pandas`` is built on cuDF and accelerates pandas code on the
GPU. It supports 100% of the pandas API, using the GPU for
supported operations, and automatically falling back to pandas for
other operations.
.. figure:: _static/duckdb-benchmark-groupby-join.png
:width: 750px
:align: center
Results of the `Database-like ops benchmark `_ including `cudf.pandas`. See details `here `_.
.. toctree::
:maxdepth: 1
:caption: Contents:
user_guide/index
cudf_pandas/index
developer_guide/index