Recently me and @Dzandaa prepared an experimental software implementation of 32-bit floating point numbers. The module float32.pas include string to float (without scientific notation) and float to string conversion, addition, subtraction, division and other basic routines. Moreover, @Dzandaa prepared routines for square root, sinus and cosinus. At the time we are testing main software float routines comparing them with results of their RTL counterparts. After tests, float32.pas will be added to UnoLib library. Source code and test program attached.