Uploaded image for project: 'Hive'
  1. Hive
  2. HIVE-10179 Optimization for SIMD instructions in Hive
  3. HIVE-10189

Create a micro benchmark tool for vectorization to evaluate the performance gain after SIMD optimization

    XMLWordPrintableJSON

    Details

    • Type: Sub-task
    • Status: Resolved
    • Priority: Major
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: None
    • Component/s: None
    • Labels:
      None

      Description

      We should show the performance gain from SIMD optimization.
      Current score is as follows:
      Benchmark Mode Samples Score Error Units
      o.a.h.b.v.VectorizationBench.DoubleAddDoubleExpr.bench avgt 2 20719.882 ? NaN ns/op
      o.a.h.b.v.VectorizationBench.DoubleAddLongExpr.bench avgt 2 22216.747 ? NaN ns/op
      o.a.h.b.v.VectorizationBench.DoubleDivideDoubleExpr.bench avgt 2 54319.682 ? NaN ns/op
      o.a.h.b.v.VectorizationBench.DoubleDivideLongExpr.bench avgt 2 34774.870 ? NaN ns/op
      o.a.h.b.v.VectorizationBench.LongAddDoubleExpr.bench avgt 2 47144.954 ? NaN ns/op
      o.a.h.b.v.VectorizationBench.LongAddLongExpr.bench avgt 2 21483.787 ? NaN ns/op
      o.a.h.b.v.VectorizationBench.LongDivideDoubleExpr.bench avgt 2 49765.990 ? NaN ns/op
      o.a.h.b.v.VectorizationBench.LongDivideLongExpr.bench avgt 2 34117.538 ? NaN ns/op

        Attachments

        1. HIVE-10189.2.patch
          8 kB
          Ferdinand Xu
        2. HIVE-10189.1.patch
          8 kB
          Ferdinand Xu
        3. HIVE-10189.patch
          5 kB
          Ferdinand Xu
        4. avx-64.docx
          29 kB
          Ferdinand Xu

          Issue Links

            Activity

              People

              • Assignee:
                Ferd Ferdinand Xu
                Reporter:
                Ferd Ferdinand Xu
              • Votes:
                0 Vote for this issue
                Watchers:
                6 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved: