L. C. Hsieh created SPARK-57112:
-----------------------------------
Summary: Add putNotNulls case to
WritableColumnVectorBulkFillBenchmark
Key: SPARK-57112
URL: https://issues.apache.org/jira/browse/SPARK-57112
Project: Spark
Issue Type: Improvement
Components: Tests
Affects Versions: 4.3.0
Reporter: L. C. Hsieh
Assignee: L. C. Hsieh
SPARK-57042 added WritableColumnVectorBulkFillBenchmark covering the
constant-value bulk-fill APIs on WritableColumnVector. The benchmark
does not yet cover putNotNulls(rowId, count), which is the inverse of
putNulls and runs once per batch from WritableColumnVector.reset()
when numNulls > 0.
This change adds a putNotNulls case mirroring the existing putNulls
case. It seeds one null into the column vector at setup so that
putNotNulls' `if (!hasNull()) return;` early-out does not skip the
fill during measurement.
The benchmark addition lands separately so that an upstream-master
baseline exists before SPARK-57111 (which switches putNotNulls to
Arrays.fill / Platform.setMemory) is measured. With this case in
master, the GHA `Run benchmarks` workflow will produce a like-for-like
before/after comparison when SPARK-57111 is benchmarked.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
