You can use Quantify, in conjunction with a Java virtual machine (JVM), to profile the performance of Java applets, class files, JAR files, and Java code launched by container programs.
You can profile Java code:
§ from Quantify standalone (start by specifying your JVM in the JVM tab of the Preferences dialog box, then run your code using the Run Program) dialog box
§ from the command line.
With Quantify's PowerTune options (click Settings in the Run Program dialog box), you can choose to collect performance data at line or function level, or by timing. You can also limit instrumentation and profiling to selected classes.
By default, Quantify collects line-level data for Java methods that are compiled to bytecode and for which debug data, which is stored in class files, is available. If debug data in not available for a module, Quantify collects function level data.
You can use Quantify's data analysis windows and tools to analyze performance. For example, after you exit an applet, file, or code, you can filter out methods such as *<clinit> that do not interest you as a developer, and focus on bottlenecks.
Notes:
§ For information about configuring your Java Runtime Environment for PurifyPlus, click .
§ Quantify uses Byte Code Insertion (BCI) technology to instrument Java code dynamically in memory. Because of this, Quantify does not display the Instrumentation progress dialog box or save copies of instrumented files.
§ For Java2-compliant Java code, Quantify collects both line-level and function-level data by timing.
§ Some methods are always timed. For example, native methods, some well-known Java methods, such as java.lang.object.wait(), and various read and write methods on Stream classes, are always timed. Quantify also times certain VM internal methods, such as Garbage Collector, Class Loader, and JIT Compiler. You can identify these methods by the prefix JVM, such as JVM Garbage Collector. By default, Quantify filters out the JIT Compiler methods from the current dataset but retains their time.
(C) Copyright IBM Corporation 1993, 2009.