Cache input data in GPU memory between consecutive queries

[andygongyb]

The FAQ page https://docs.nvidia.com/spark-rapids/user-guide/latest/faq.html states that:

Are cache and persist supported?
Yes cache and persist are supported, the cache is GPU accelerated but still stored on the host memory. Refer to RAPIDS Cache Serializer for more details.

Can I cache data into GPU memory?
No, that isn’t currently supported. It would require much larger changes to Apache Spark to be able to support this.

May I ask if anyone could confirm that, if I understand it correctly, the input data always need to be transferred from the host memory to the device memory again between consecutive queries?

[mattahrens]

Yes, your understanding is correct. Caching data onto GPU memory is not currently supported and input data will need to be transferred from host memory to device memory between consecutive queries even if the RAPIDS Cache Serializer is used.

[andygongyb]

Thank you so much for the clarification!