Bug ID: JDK-4969969 JDWP: Performance regression on using debbuger (step speed improvements)

Versions (Unresolved/Resolved/Fixed)

The Version table provides details related to the release that this issue/RFE will be addressed.

Unresolved : Release in which this issue/RFE will be addressed.
Resolved: Release in which this issue/RFE has been resolved.
Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.

To download the current JDK release, click here.

Other
5.0 b40Fixed

###@###.### 2003-12-17

J2SE Version (please include all output from java -version flag):
    java version "1.5.0-beta"
    Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0-beta-b30)
    Java HotSpot(TM) Client VM (build 1.5.0-beta-b30, mixed mode) 

Does this problem occur on J2SE 1.3 or 1.4.x?  Yes / No (pick one) 
    No, 1.4.2 only takes 1 sec. b30 takes 18 sec.

Operating System Configuration Information (be specific): 
    Windows XP (version 5.1 Service Pack 1 (Build 2600)) 

Hardware Configuration Information (be specific): 

    Platform configuration:

    CPU:     Intel 32bit Single Processor
    Processor Speed: 1893 MHz

    Operating System:
    Windows XP (version 5.1 Service Pack 1 (Build 2600))

    DirectX version: DirectX 8.1 or greater

    DirectX display devices:
    Device 0:
      Resolution (width x height x depth): 1024 x 768 x 32
      Device driver nv4_disp.dll version 6.13.10.3638

    GDI display devices:
    Primary: Name = \\.\DISPLAY1, Description = NVIDIA GeForce4 460 Go (Toshiba)

Bug Description: 

   Testing the VisualMust debugger with B30, there is a problem :
   when running the application step by step and entering into a "super()"
   to create a frame for example, the step takes 18 seconds and 100 % CPU 
   (with a Pentium 4, 1900 MHz). With the same machine, with Java 1.4.2. 
   the step takes less than 1 second.

Steps to Reproduce (be specific):
  - unzip VisualMust.zip
  - run java -Xmx128m -classpath ./VisualMust.jar;./VisualMustDocs.jar;./JbcBeansEvaluation.jar;c:/jdk/lib/tools.jar VisualMust

CONVERTED DATA BugTraq+ Release Management Values COMMIT TO FIX: tiger-beta2 FIXED IN: tiger-beta2 INTEGRATED IN: tiger-b40 tiger-beta2

14-06-2004

PUBLIC COMMENTS Initial step into a JFrame is much slower with 1.5. There were some basic performance improvements made for this bug in the debugger backend, but the JFrame changes for 1.5 has caused a significant increase in the number of methods called, which will impact the step performance significantly.

10-06-2004

EVALUATION ###@###.### 2004-01-09 Must fix for beta2. -- Looks like we're victim to changes in the javax.swing.JFrame constructor. With 1.4.2 stepping into the JFrame constructor when there's a javax.* filter in place takes <1 second. The resulting step processes 15,300 MethodEnter events. With 1.5.0 the equivalent step takes 8 second on the machine that we have the VisualMust test case setup. The resulting step processes 377,400 MethodEnter events. However the test is useful as it has identified a number of cases where the step performance can be improved. ###@###.### 2004-02-05 ==================================================================== We have made some good speed improvements in the debugger backend and will submit these changes under this bug, however, to be clear about this, the significant increase in methods entered was the trigger for the performance issue and the Swing team may want to investigate this. So "fixed" in this case is applying to making stepping faster in 1.5 in general, and although this benefits this testcase, it can't completely fix all the symptoms. A separate bug will be filed on the JFrame issue. Bug id 4990399. ###@###.### 2004-02-06

06-02-2004