Peter Rudenko

Fixes #31 For some reason on the user's system `ibv_context->num_comp_vectors` returns negative number (probably not initialized). To prevent `DivisionByZeroException` we can use the default value of 1, to assign CQ...

#37 Before `postRecv.execute()` need to call `endpoint.getWcEvents().take();` to not get stall buffer.

[Proposal] Include libdisni.so into assembly jar

9

This will include libdisni into assembly jar, thus not require to mess with `LD_LIBRARY_PATH`, etc, allowing to have different disni versions for different components. Making PR for discussion if everything's...

Useful function for cache purpose

Configurable null values

8

There's datasets where each column has it's own marker for missing values. spark-csv assumes only empty string for missing values. To avoid additional data transformation and saving on user's side...

Text parameter type

1

We need from user to enter/edit long config value. Currently we use `"type" : "string"` but in UI it's very inconvenient, need to move cursor. Would be great to have...

For reference. This branch seems like worked for GPU unified API.

[WIP] Spark Shuffle manager implementation.

2

Working implementation of Spark Shuffle manager with UcxShuffleTransport API. TODO: 1. Run tests 2. Publish to maven 3. Port one-sided API protocol

Performance benchmark

1

To run a benchmark: ``` mvn package (there would be 2 jars: ucx-spark-2.0-for-spark-3.0.0-jar-with-dependencies.jar and ucx-spark-2.0-for-spark-3.0.0-tests.jar ``` Put them to some jars folder with cudf. Server: ``` java -cp /PATH_TO_UCX/lib/:spark/jars/*:jars/* org.apache.spark.shuffle.ucx.perf.UcxShuffleTransportPerfTool...