Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-21717

Decouple the generated codes of consuming rows in operators under whole-stage codegen

Attach filesAttach ScreenshotVotersWatch issueWatchersCreate sub-taskLinkCloneUpdate Comment AuthorReplace String in CommentUpdate Comment VisibilityDelete Comments
    XMLWordPrintableJSON

Details

    • Improvement
    • Status: Resolved
    • Critical
    • Resolution: Fixed
    • 2.2.0
    • 2.3.0
    • SQL
    • None

    Description

      It has been observed in SPARK-21603 that whole-stage codegen suffers performance degradtion, if generated functions are too long to be optimized by JIT.

      We basically produce a single function to incorporate generated codes from all physical operators in whole-stage. Thus, it is possibly to grow the size of generated function over a threshold that we can't have JIT optimization for it anymore.

      This ticket is trying to decouple the logic of consuming rows in physical operators to avoid a giant function processing rows.

      Attachments

        Activity

          This comment will be Viewable by All Users Viewable by All Users
          Cancel

          People

            viirya L. C. Hsieh
            viirya L. C. Hsieh
            Votes:
            0 Vote for this issue
            Watchers:
            6 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved:

              Slack

                Issue deployment