Uploaded image for project: 'JDK'
  1. JDK
  2. JDK-8056025

jdk.nashorn.internal.codegen.CompilationPhase.setStates() is hot in class installation phase

    XMLWordPrintable

    Details

    • Type: Enhancement
    • Status: Resolved
    • Priority: P4
    • Resolution: Fixed
    • Affects Version/s: 8u40
    • Fix Version/s: 9
    • Component/s: core-libs
    • Subcomponent:
    • Resolved In Build:
      b29
    • CPU:
      generic
    • OS:
      generic

      Backports

        Description

        Profiling a simple scenario:

        $ ~/Install/jdk9u20/bin/java -jar dist/nashorn.jar -Dnashorn.typeInfo.disabled=false --class-cache-size=0 --persistent-code-cache=false -scripting --log=time test/script/basic/compile-octane.js -- --iterations 5

        ...yields a few simple low-hanging fruits, here is one of them. Out of 12 seconds spent in class installation time, at least 2 seconds are spent here:

        12.690 jdk.nashorn.internal.codegen.CompilationPhase$12.transform(jdk.nashorn.internal.codegen.Compiler, jdk.nashorn.internal.codegen.Compiler$CompilationPhases, jdk.nashorn.internal.ir.FunctionNode)
        2.110 jdk.nashorn.internal.codegen.CompilationPhase.access$100(jdk.nashorn.internal.ir.FunctionNode, jdk.nashorn.internal.ir.FunctionNode$CompilationState)
        2.110 jdk.nashorn.internal.codegen.CompilationPhase.setStates(jdk.nashorn.internal.ir.FunctionNode, jdk.nashorn.internal.ir.FunctionNode$CompilationState)
        2.100 jdk.nashorn.internal.ir.FunctionNode.accept(jdk.nashorn.internal.ir.visitor.NodeVisitor)

        Attila says setStates should not be hot, since it is apparently the internal invariant-checking thing. The dumbest patch of all:

        diff -r 494092ee7a01 src/jdk.scripting.nashorn/share/classes/jdk/nashorn/internal/codegen/CompilationPhase.java
        --- a/src/jdk.scripting.nashorn/share/classes/jdk/nashorn/internal/codegen/CompilationPhase.java Mon Aug 25 22:36:05 2014 +0200
        +++ b/src/jdk.scripting.nashorn/share/classes/jdk/nashorn/internal/codegen/CompilationPhase.java Tue Aug 26 12:24:39 2014 +0400
        @@ -567,7 +567,7 @@
                         log.fine(sb.toString());
                     }
         
        - return setStates(fn.setRootClass(null, rootClass), BYTECODE_INSTALLED);
        + return fn.setRootClass(null, rootClass);
                 }
         
                 @Override

        ...on the current jdk9/dev and this simple benchmark:

        $ for I in `seq 1 5`; do ~/trunks/jdk9-dev/build/linux-x86_64-normal-server-release/images/j2sdk-image/bin/java -jar ~/trunks/jdk9-dev/build/linux-x86_64-normal-server-release/images/j2sdk-image/jre/lib/ext/nashorn.jar -Dnashorn.typeInfo.disabled=false --class-cache-size=0 --persistent-code-cache=false -scripting --log=time test/script/basic/compile-octane.js -- --iterations 1 2>&1 | grep "Class Installation"; done

        ...yields a big improvement.

        Before (setStates is called):
         [time] 'Class Installation' 2926 ms
         [time] 'Class Installation' 3426 ms
         [time] 'Class Installation' 3079 ms
         [time] 'Class Installation' 2844 ms
         [time] 'Class Installation' 2893 ms

        After (setStates is not called):
         [time] 'Class Installation' 1668 ms
         [time] 'Class Installation' 1680 ms
         [time] 'Class Installation' 1674 ms
         [time] 'Class Installation' 1681 ms
         [time] 'Class Installation' 1720 ms

        This is with JDK-8055954 applied, so may be the relative improvement without JDK-8055954 is worse.

          Attachments

            Issue Links

              Activity

                People

                Assignee:
                lagergren Marcus Lagergren
                Reporter:
                shade Aleksey Shipilev
                Votes:
                0 Vote for this issue
                Watchers:
                2 Start watching this issue

                  Dates

                  Created:
                  Updated:
                  Resolved: