blazingsql

Why we included this project

BlazingSQL is a SQL engine that pushes query execution onto GPUs, using RAPIDS' cuDF for its dataframes. You register cuDF or Dask-cuDF dataframes, local files, or Amazon S3 buckets, then issue ordinary SQL through a Python context object; the engine parses and runs the query as CUDA kernels, including heavy filters, joins, and aggregations, and hands results back as GPU dataframes. That avoids moving data through CPU memory first, which is the usual bottleneck in big data science jobs. One honest caveat: the project hasn't seen active development since 2022, so treat it as a useful reference for GPU-accelerated query design, not as a dependency for new production stacks.