| 摘要 |
PURPOSE: To provide a floating point computing system capable of performing fast arithmetic processing consequently by simplifying the working process of floating point computation by extended precision. CONSTITUTION: This system is provided with a conversion part 2 which converts a bit format 1 of floating point expression format with (n) bit precision to a bit format 3 of floating point expression format with extended precision, and an arithmetic part 4 which executes the floating point computation with extended precision by using the bit format 3 with extended precision. The bit format 3 with extended precision is constituted, for example, setting the high- order numeric value format EH of floating point expression format of high-order 20 bits (the number of effective bits = 21) and the low-order numeric value format EL of floating point expression format of low-order 52 bits with double precision of 52 bits (the number of effective bits = 53) for a mantissa part as a pair. |
