Uploaded image for project: 'JDK'
  1. JDK
  2. JDK-6343888

REGRESSION: JTableLoad test performance regression

    XMLWordPrintable

    Details

      Description

      Attached is the benchmark test program of JTable initialization:

      Source code:
      ------------------------------------
      import javax.swing.*;

      public class JTableLoad {

           public static void main(String args[]) throws Exception {
      long t1 = System.currentTimeMillis();
               new JTable();
      long t2 = System.currentTimeMillis();
      System.out.println("new JTable() time: "+(t2-t1));
      System.out.println("Press Ctrl+C to end.");
      System.in.read();

           }

      }
      -------------------------------------


      Test comment:
      Each test was run 11 times. First run was ignored to ensure the time is
      not affected by swapping or disk I/O operations.

      Test results:

      Windows:

      JVM 1.6b54
      844,594,516,860,687,593,828,859,875,875ms
      average: 753ms

      JVM 1.5.0
      437,422,438,422,422,438,422,437,422,437ms
      average: 430ms

        Linux:

      JVM 1.6b58
      244,244,241,242,243,242,242,240,240,242ms
      average: 242ms
      JVM 1.5.0
      206,205,207,195,207,207,206,205,208,205ms
      average: 205ms


      Test Observation:
      1. On Windows there is high performance slowdown (about 75%) in loading
         Swing JTable in Mustang JVM.
      2. Results on Windows are much worse then on Linux

      Test Configuration:

      Hardware: Mobile Pentium 4 2.66GHz, 512 MB RAM
      OS1: Windows XP Professional
      OS2: Fedora Linux 2
      Customer exact system configuration:
      Notebook Acer TravelMate 541LCi
      Mobile Intel Pentium 4 - 2.66GHz, no Hyperthreading
      512 MB RAM
      40GB Ultra ATA/100 HDD
      ATI Mobility Radeon 9200
      Windows XP Proffesional, Service Pack 1, Czech Locale
      Antivirus disabled
      200 MB RAM occupied by system and applications

      performed tests again with settings as suggested in
      bug#6343888evaluation. Results are still the same:


      1.6.0 b54

      java -Xshare:on
      687,734,672,672,688,687,735,781,672,672
      average: 700


      java -Xshare:off
      656,750,781,875,859,859,718,766,844,766
      average: 787


      1.5.0_05

      java -Xshare:off
      453,438,438,438,438,437,437,438,453,438
      average:440

      java -Xshare:on
      266,250,235,234,234,281,313,312,312,312
      average:274


      they are uniprocessor, without
       hyperthreading and multicore features.


      more testing result with build57
      ===============================

      performing the same test on another machine with Mustang build 57:

      JDK 1.6.0: -Xshare:off
      390,391,390, ... still the same numbers

      JDK 1.6.0: -Xshare:on
      328,344, ... still the same numbers

      JDK 1.5.0: -Xshare:on
      141,156,156,141,... still the same numbers


      Machine configuration:
      Dell Precision 360
      Pentium 4 3GHz, Hyperthreading
      Windows XP professional, Czech Locale
      1 GB RAM

        Attachments

          Issue Links

            Activity

              People

              Assignee:
              bchristi Brent Christian
              Reporter:
              tyao Ting-Yun Ingrid Yao (Inactive)
              Votes:
              0 Vote for this issue
              Watchers:
              0 Start watching this issue

                Dates

                Created:
                Updated:
                Resolved:
                Imported:
                Indexed: