GitHub - NVIDIA/spark-rapids: Spark RAPIDS plugin - accelerate Apache Spark with GPUs (original) (raw)

RAPIDS Accelerator For Apache Spark

NOTE: For the latest stable README.md ensure you are on the main branch.

The RAPIDS Accelerator for Apache Spark provides a set of plugins forApache Spark that leverage GPUs to accelerate processing via the RAPIDS libraries.

Documentation on the current release can be found here.

To get started and try the plugin out use the getting started guide.

Compatibility

The SQL plugin tries to produce results that are bit for bit identical with Apache Spark. Operator compatibility is documented here

Tuning

To get started tuning your job and get the most performance out of it please start with thetuning guide.

Configuration

The plugin has a set of Spark configs that control its behavior and are documentedhere.

Issues & Questions

We use github to track bugs, feature requests, and answer questions. File anissue for a bug or feature request. Ask or answer a question on the discussion board.

Download

The jar files for the most recent release can be retrieved from the downloadpage.

Building From Source

See the build instructions in the contributing guide.

Testing

Tests are described here.

Integration

The RAPIDS Accelerator For Apache Spark does provide some APIs for doing zero copy data transfer into other GPU enabled applications. It is describedhere.

Currently, we are working with XGBoost to try to provide this integration out of the box.

You may need to disable RMM caching when exporting data to an ML library as that library will likely want to use all of the GPU's memory and if it is not aware of RMM it will not have access to any of the memory that RMM is holding.

Qualification and Profiling tools

The Qualification and Profiling tools have been moved tonvidia/spark-rapids-tools repo.

Please refer to Qualification tool documentationand Profiling tool documentationfor more details on how to use the tools.

Dependency for External Projects

If you need to develop some functionality on top of RAPIDS Accelerator For Apache Spark (we currently limit support to GPU-accelerated UDFs) we recommend you declare our distribution artifact as a provided dependency.