摘要 |
PURPOSE:To reduce the overhead of exponent part conversion processing and to speed up the processing by detecting an exponent overflow, an exponent underflow, and whether a bias adjustment is allowed or not by using the output of an exponent part converting circuit. CONSTITUTION:The exponent part converting circuit 3 inputs the exponent part E of input floating point data 2 and performs specified accuracy conversion to generate an exponent overflow detection signal EXO or exponent underflow detection signal EXU when detecting the overflow or underflow at this time. Further, a state detection signal CC for detecting whether the bias adjustment is allowed or not is outputted with a value based upon k1=E or k2=2, k1<E<k2, or E<k1 or k2<E, where k1 is the lower-limit value of a predetermined bias adjustment permissible limit and k2 is the upper-limit value. A main conversion processing part 4 inputs a mantissa part F to perform the bias adjustment when allowed according to the respective signals EXO, EXU, and CC, and outputs the conversion output 5. Consequently, a waste of the procedure for the accuracy conversion processing is reduced and the processing is speeded up. |