[SYSTEMDS-3902] Replace py4j with faster Unix pipes for Python-Java DataFrame transfer #2363
+2,318
−451
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch extends the previously added data transfer with unix pipes to SystemDS frame transfer capabilities. Additionally, the matrix transfer was further improved by fusing the non-zero value computation into data reading and reducing unnecessary array allocations. The impact is nicely visible in the flame graph of the profiler:
Numpy transfer w/ fused DenseBlock creation and nnz count
Experiment setup: Python --> Java --> Python
py4j profiling
runtime: 16s

unix-pipe profiling
runtime: 0.7s

Updated Python --> Java numpy transfer time:
DataFrame Transfer stats: