J.A.D.E.
Java^TM Addition to Default Environment.

      Just as the constant increase of entropy is the basic law of the universe,
      so it is the basic law of life to be ever more highly structured and to
      struggle against entropy!                    Václav Havel

• Real-time facility to accelerate calculations and avoid garbage collection interruptions.
• units package (e.g. SI units) developped in collaboration with the JSR-108 experts group.
• Strongly typed quantity classes to avoid interface errors such as the one which caused the loss of the Mars climate orbiter.
• Automatic mapping between a quantity's unit and its class (e.g. the division of a Length by a Duration returns a Velocity instance).
• Support for Standard, Relativistic, High-Energy, Quantum and Natural physical models.
• Error calculations including numeric errors to avoid disasters such as the patriot missile failure and the explosion of Ariane 5.
• A math package which includes a generic Matrix class, Real and Complex numbers (and soon LargeInteger and LargeFraction numbers).
• Support for the encoding of objects into XML; and the complementary reconstruction of the object graph from XML.
• XML parsing with no dynamic memory allocation! It allows our RealtimeParser to run much faster than conventional SAX2 parsers.
• Monetary package fully integrated with units (JSR-108) and physical quantities.
• And miscellaneous public domain classes.

This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License (LGPL).

JADE Version 4.6.7

           

Runtime Environment: J2SE 1.4+

Compilers: Java^TM, Jikes^TM or GNU Compiler for Java (GCJ)

Build tool: Apache Ant

Collaborative Tool: java.net (code development, comments, issues, bug reporting).

Benchmark

The JADE library is highly optimized for speed. It also provides real-time classes to accelerate code execution by avoiding or postponing memory allocation and garbage collection.

As shown below, gain can be of several orders of magnitude.

"Performs 100 successive concatenations of this string."	HotSpot 1.4.1 (java -jar jade.jar perf)	GCJ 3.2 (jade.exe perf)
String (immutable)	4.10 ms	13.89 ms
StringBuffer (mutable)	0.31 ms	0.96 ms
StringRT in heap context (immutable)	0.31 ms	1.02 ms
StringRT in pool context (mutable)	0.12 ms	0.13 ms

Below the benchmark results on Windows 2000, Compaq at 2.26GHz:

D:\jade-4.6>java -jar jade.jar perf

Java(TM) Addition to Default Environment (http://jade.dautelle.com/)
Version 4.6.6 June 25 2003

Benchmark...

Initialization: 0.17 s

JVM Heap allocation versus Pool-Context
java.lang.Double (8 bytes) Heap Creation: 42 ns
char[128] Heap Creation: 650 ns
char[256] Heap Creation: 1263 ns
com.dautelle.math.Real (8 bytes) Pool-Context Creation: 33 ns
char[128] Pool-Context Creation: 56 ns
char[256] Pool-Context Creation: 61 ns

Mathematical operations:
com.dautelle.math.Real add: 37 ns
com.dautelle.math.Complex add: 56 ns
com.dautelle.physics.Quantity add: 97 ns
com.dautelle.math.Real multiply: 39 ns
com.dautelle.math.Complex multiply: 64 ns
com.dautelle.physics.Quantity multiply: 122 ns

StringRT versus String/StringBuffer:
"Performs 100 successive concatenations of this string."
java.lang.String Concatenation: 1.497 ms
java.lang.StringBuffer Concatenation: 0.106 ms
com.dautelle.realtime.StringRT Concatenation(Heap-Context): 0.113 ms
com.dautelle.realtime.StringRT Concatenation(Pool-Context): 0.031 ms

XML Parsing
org.apache.crimson.parser.XMLReaderImpl (Sax2 Parser): 2922
com.dautelle.xml.sax.RealtimeParser: 594
com.dautelle.xml.sax.XmlReaderImpl (Sax2 Wrapper of RealtimeParser): 984

More performance analysis in future versions...

History

• Summer 2003...

  OpenGL^TM/DirectX^TM bridge API (any help welcome!)

• Coming soon:
  • Additional math classes: LargeInteger and LargeFraction sensitive to the current modulus context (not set by default). They implements the Operable interface, making it easy to resolve "modulo" equations (Ref. Number Theory) and to find the exact solutions to linear equations.
  • Examples on how to quickly solve differential equations to the machine precision (numerical errors only) using DiscreteFunction and Quantity classes.
  • Polynomial class (extends Function).

• July 6, 2003: JADE 4.6.7
  
  • Much faster algorithms for the generic Matrix class.
  • Member of java.net.
  • UTF-8 encoding for JADE source code (default encoding on Linux).
  • Corrected minor bugs in Utf8StreamReader (buffer overflow) and TypeFormat (precision for negative number).
  • Real-time XML parser, extremely fast (faster than any conventional SAX2 parser) and small (15Kbytes compressed). It has the unique characteristic of not allocating dynamic memory and still maintains a very small memory footprint (e.g. less than 16Kbytes while parsing 32Mbytes files). Typical applications include SOAP messaging, embedded/realtime systems, web servers (possibly thousands instances running concurrently), etc.
  • Our XML serialization/deserialization facility is now using the new RealtimeParser (300% improvement in deserialization speed).

• April 27, 2003: JADE 4.4.5
  
  • Utility class (public domain) TypeFormat to parse text into primitive types (fast and unlike the standard library, does not create temporary objects).
  • Preallocation for PredictiveContext can be performed concurrently (background task).
  • StringRT is fully integrated with the new (1.4) java.lang.CharSequence interface (allows for search, concatenation and comparison with any CharSequence such as itself, java.lang.String or java.lang.StringBuffer).

• March 25, 2003: JADE 4.3
  
  • Updated real-time FAQ.
  • New class StringRT. This class provides an immutable String several order of magnitude faster than java.lang.String. Concatenation for example, executing in the default JVM heap context (where immutability is guaranteed) is basically as fast as if StringBuffer were being used! For even greater performance, you may execute within a real-time context, then StringRT is 2x-3x time faster than java.lang.StringBuffer (see Benchmark).
  • math package has been extended with generic functions. The combination of discrete functions and physical quantities allows for the numerical resolution of any real-life system (differential). Furthermore, because quantity calculations take into account numeric errors, the solution obtained can be guaranteed (interval solution always contains the exact "theorical" solution)!

• February 24, 2003: JADE 4.2
  
  • Monetary quantities have been added to the money package.
  • The physics package has been re-written to support multiple physical models (Standard, Relativistic, HighEnergy, Quantum and Natural).
  • The xml package documentation has been improved.

• November 24, 2002: JADE 3.1
  

  The real-time package has been finalized. This framework was initially intended for real-time systems, but its "accelerating" effect might appeal to any system dynamically allocating memory.

  The cost of memory allocations/initialization can be huge (90% of CPU is not rare) and using mutable objects is unsafe... The framework provides "immutable objects" safety and convenience (e.g. no need to specify a "returnValue" parameter) with the performance of "mutable objects".

• September 4, 2002: JADE 3.0
• July 28, 2002: JADE 2.0
• November 12, 2001: JADE 1.4
• February 24, 2001: JADE 1.3
  Article focusing upon JADE XML support, published in the February issue of Dr Dobbs Journal.
• December 11, 2000: JADE 1.2
• December 9, 2000: JADE 1.1
• September 12, 2000: Official release of JADE 1.0

Comments To: Jean-Marie Dautelle
Revision: July 6, 2003

Java^TM is a trademark of Sun Microsystems, Inc

Copyright© 2003 The Dautelle Family