Floating point adding circuit

Abstract

Abstract PURPOSE:To perform operations at a high speed and to detect exceptions from result at a high speed by extracting a part of an exponential part and operating it in preference to a mantissa part in a normalizing circuit after operation of a floating point adding circuit. CONSTITUTION:A counting circuit 13 counts the number of continuous ''0''s doem the most significant bit of the intermediate result of the mantissa part to output an m-bit counted result. Before the output of the counting circuit 13 is settled, high-order n bits of the exponential part having an (n+m)-bit data width are extracted by an extracting circuit 15 and are defined as EXP(n), and one is subtracted from EXP(n) by the first operator 11. If the intermediate result of the mantissa part is not overflowed, m bits of the output of the counting circuit 13 are subtracted from low-order m bits of the exponential part by the second operator 12 to output low-order m bits. On a basis of the borrow output of the second operator 12, EXP(n)-1 is selected by a selecting circuit 14 in case of borrow but EXP(n) is selected in case of borrow to determine high-order n bits of the operation result of the exponential part.

Inventors

• 👤 Kenzo Tanimoto
• 👤 谷本　謙造

Bibliographic Details