Ruifeng Zheng created SPARK-59532:
-------------------------------------
Summary: Optimize VectorDotProduct for foldable inputs
Key: SPARK-59532
URL: https://issues.apache.org/jira/browse/SPARK-59532
Project: Spark
Issue Type: Improvement
Components: ML
Affects Versions: 5.0.0
Reporter: Ruifeng Zheng
VectorDotProduct decodes both vector structs and accesses both operands through
ArrayData for every input row. When one operand is foldable, evaluate it once
during code generation and cache its size, indices, and values using primitive
data. Preserve the existing fallback for two dynamic operands and cover dense
and sparse vectors in both operand positions.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
