The list below outlines the release history including dates and brief descriptions describing the changes and fixes since the previous release.

• Version 2.0.6 for KVM, CLDC/MIDP and WABA Sept. 8th 2002
  A small bug fix release. The function toLong (and toInt in Waba and KVM) had problems rounding to the correct nearest natural number for negative values (Ben Hui). This release also includes an updated demo for the MIDP on PalmOS. MathFP.toLong(long fp) should be if rounding is desired otherwise a simple bit shift can be used using the bit precision of the number (e.g. fp >> 24).

• Version 2.0.5 for KVM, CLDC and WABA Feb 9th 2001
  Fixed arcsine sign issue and added atan2 function. Included Test programs for PalmOS (J2ME or Waba) and Windows CE (waba).

• Version 2.0.4 Beta2 for J2ME CLDC and WABA Dec 13th 2000
  New package structure defined to support multiple platforms (See documentation) and provide backwards compatibilty. Both 64bit and 32bit fixed-point integer support.

• Version 2.0 Beta1 for J2ME CLDC Nov 29th 2000
  This version contains all intended functions for release 2. This includes the convert function to change numbers with different precision to the current runtime precision. Log and pow functions enhanced for greater precision; Small numbers for log are now more precise and as a concequence pow. Documentation completed.

• Version 2.0 Alpha 2 for J2ME CLDC Oct 29th 2000
  MathFP Version 2 Aplha 2 is fully functional. All functions from Version 1 and more (i.e. atan) are implemented. Again this is just a pre-release. Some additional size and speed enhancements will be made and the documentation needs to be completed.

• Version 2.0 Alpha 1 for J2ME CLDC Oct 26th 2000
  MathFP Version 2 is based on the long data type and implements variable precision. Precision can be adjusted during runtime. This release is intended as a preview only.

• Version 1.2.2 for J2ME CLDC May 25th 2000
  Improved performance and precision for log, exp, pow. The exp function is now extremly fast and as a result pow function as well. The exp function also has overflow detection for exponents that are to large. Round function added. Updated 1.2.2 documentation.

• Version 1.2.1 for J2ME CLDC May 23rd 2000
  Improved performance and precision for log, exp, pow, mul and div. I also added two new performance tips to the fixed-point basic theory.

• Version 1.2 for J2ME CLDC May 2th 2000
  Improved performance and precision for sqrt, mul and div. Reduced the size of the class (less than 4K) and added a new sqrt method with precision control. The new sqrt method is 2 to 10 times faster than the previous implementation. Fixed a rounding bug. Rounding for negative numbers is now away from 0. In addition MathFP.toInt(n) rounds before converting to an int.

• Version 1.1.3 for J2ME CLDC April 8th 2000
  Discovered a sign problem in the arcsine function. For all numbers between -1 and 0 the function returned a positive number. This is obviously incorrect. In this new release then sign has been taken into account.

• Version 1.1.2 for J2ME CLDC march 30th 2000
  Generated the new MathFPClassDB.pdb file with the preverification of the MathFP class for J2ME CLDC beta 1.0. Added a small demonstration program called GoFP.prc to validate the correct installation of MathFP

• Version 1.1.2 march 12th 2000
  Improved performance for div's for fixed point integers with no fraction value. Reduced the size of the class by inlining of constants.This saved about 200 bytes. The class size is now 4.04kb

• Version 1.1.1 march 6th 2000
  Improved the performance of mul and div. Because these function are part of many functions it will speedup most other functions too. Range for the div function increased to MAX_VALUE for numbers that have no fractional part and for fractional numbers the maximum is now 131071. More overflow detection added (i.e. mul) Although size optimizations have been implemented they where nulled by the other improvements. The library did only increase by a few bytes.

• Version 1.1.0 february 23rd 2000
  Added the natural logarithm function log() and the pow() function to raise fixed-point integers to fixed-point integers. Renamed the EXP constant (representing the natural number) to E.

• Version 1.0.3 february 3rd 2000
  MathFP finally belongs to a package. The package is net.jscience.util. Some development tools did not support the JAVA default package and hence could not work with the MathFP library.

• Version 1.0.2 january 31st 2000 This release contains a bug fix for handeling fractions like 1.9999. The toFP translator ignored the overflow and the internal representaion resulted in something not to close to 2 (fraction wise). The fix will handle the overflow and convert the number to 2.0000.
  Additionally this release will automatically chop-off excessive fractional digits. The maximum nuber of fractional digits is 4. So if you specify more then 4 digits then the extra digits will be ignored.

• Version 1.0.1 january 28th 2000
  This version includes an additional toString member function which allows you to specify the number of fractional digits to display. The result will be correctly rounded.

• Version 1.0.0 january 18th 2000

  First public release of the MathFP library.

Onno Hommes, Rochester NY
jScience Technologies
www.jscience.net