JavaMemoryModel: Synchronization overhead in Spec JVM 98

• Next message: Bill Pugh: "Re: JavaMemoryModel: Re: Double-check and Symantec JIT"
• Previous message: Gregory Thain: "JavaMemoryModel: Re: Double-check and Symantec JIT"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
• Mail actions: [ respond to this message ] [ mail a new topic ]

Hans Boehm asked me for some more details about exactly which JVM I
used in my synchronization experiments.

I've attached an HTML file with some more details. This is also available at

    http://www.cs.umd.edu/~pugh/java/memoryModel/synchronizationOverhead.html

None of the synchronization in Spec JVM is useful, so it can all be
removed. For these experiments, I removed all synchronization from

        BufferedInputStream
        BufferedOutputStream
        ByteArrayInputStream
        ByteArrayOutputStream
        Vector
        Hashtable
        Stack

A separate web page

   http://www.cs.umd.edu/~pugh/java/memoryModel/syncCost.html

gives some information about the number of synchronization operations
in the benchmarks.

        Bill

• text/html attachment: synchronizationOverhead.html

• Next message: Bill Pugh: "Re: JavaMemoryModel: Re: Double-check and Symantec JIT"
• Previous message: Gregory Thain: "JavaMemoryModel: Re: Double-check and Symantec JIT"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
• Mail actions: [ respond to this message ] [ mail a new topic ]

This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:26 EDT