- Top: XFloat Documentation
2. XFloat Basics
Here describes the basic concepts you need to know to before using XFloat.
XFloat split a large integer into 24-bit size internally. This 24-bit unit is called 'word'. Thus one word matches a 24-bit integer. Numeric expressed in XFloat can be stored in one word or more, depending on the precision of that number.
Floating point stores numbers into the mantissa and exponent parts. Thus, representing the number of 12,340,000 in floating point is 1, 2, 3, 4 and 7 because 12,340,000 = 1.234 * 107, instead of storing all digits like 1,2,3,4,0,0,0,0.
In IEEE 754 floating-point format, double-precision can not exceed 53-bit precision. Therefore, the error will occur in the lower-digit if more than 53-bit of precision is required to store the value. In XFloat, however, the precision of mantissa is adjusted as needed, so you can lower the error of the value below a desired.
XFloat, as distinct from IEEE 754 format, handles a sign of value into three. One is positive (greater than 0), another is 0, and the other is negative (smaller than 0). Thus, XFloat has unique 0 value while IEEE 754 divides zero into +0 and -0.
In IEEE 754 format, the result is set to Infinity or NaN if an exception, such as dividing by zero and square root of a negative number, is occured. In contrast, XFloatError object is thrown as an exception when unrecoverable error is occurred in XFloat. So you need to write a routine to handle an exception occurred in operations.