Floating-point arithmetic apparatus

Claim 2

A floating-point arithmetic apparatus for first and second floating point data each having an exponent section, a mantissa section expressed in an absolute value, and a sign section representing a sign of the mantissa section, the apparatus comprising: first means for calculating the difference between the exponent sections of the first and second floating point data, determining which of the exponent sections of the first and second floating point data is greater, and selecting the greater exponent section of the first and second floating point data; second means responsive to an output from the first means for selecting the mantissa section of one of the first and second floating point data which has the smaller exponent section, and shifting rightward the selected mantissa section by a bit number corresponding to the difference between the exponent sections; third means responsive to the output from the first means for selecting the mantissa section of one of the first and second floating point data which has the greater exponent section; an arithmetic means for performing addition and subtraction between outputs from the second and third means; fourth means for normalizing a mantissa section which is an output from said arithmetic means, and correcting the greater exponent section of the first and second floating point data in correspondence with the normalizing; fifth means for rounding off a mantissa section of an output from the fourth means; sixth means for calculating an absolute value of the mantissa section of the output from the fourth means; seventh means for generating a signal representative of a final arithmetic mode on the basis of the sign sections of the first and second floating point data, a result of the determination by the first means about which of the exponent sections of the first and second floating data is greater, and an arithmetic mode designation signal, for outputting the final arithmetic mode signal to said arithmetic means, and for, in cases where the final arithmetic mode corresponds to subtraction and the output from the first means indicates that there is a difference in exponent between the first and second floating point data, instructing said arithmetic means to subtract the output of the second means from the output of the third means; eighth means for, in cases where the output from the first means represents that there is a difference in exponent between the first and second floating point data, enabling the fifth means to execute its function, for in cases where the output from the first means represents that there is no difference in exponent between the first and second floating point data and the final arithmetic mode corresponds to addition, enabling the fifth means to execute its function, and for, in cases where the output from the first means represents that there is no difference in exponent between the first and second floating point data and the final arithmetic mode corresponds to subtraction, enabling the sixth means to execute its function; and ninth means for determining a sign of a final arithmetic result on the basis of the sign sections of the first and second floating point data, the result of the determination by the first means about which of the exponent sections of the first and second floating point data is greater, the arithmetic mode designation signal, and a sign bit of the output from said arithmetic means.

Claim 3

The apparatus of claim 1 further comprising means for, when an output from the fifth means overflows, shifting rightward the output from the fifth means by one bit and correcting an exponent section of an output from the fourth means.

Claim 4

The apparatus of claim 1, wherein the said arithmetic means and the sixth means each include an adder.

Claim 5

A floating-point arithmetic apparatus for first and second floating point data each having an exponent section, a mantissa section expressed in an absolute value, and a sign section representing a sign of the mantissa section, the apparatus comprising: first means for calculating the difference between the exponent sections of the first and second floating point data, determining which of the exponent sections of the first and second floating point data is greater, and selecting the greater exponent section of the first and second floating point data; second means responsive to an output from the first means for selecting the mantissa section of one of the first and second floating point data which has the smaller exponent section, and shifting rightward the selected mantissa section by a bit number corresponding to the difference between the exponent sections; third means responsive to the output from the first means for selecting the mantissa section of one of the first and second floating point data which has the greater exponent section; an arithmetic means for performing addition and subtraction between outputs from the second and third means; fourth means for normalizing a mantissa section which is an output from said arithmetic means, and correcting the greater exponent section of the first and second floating point data in correspondence with the normalizing; fifth means for rounding off a mantissa section of an output from the fourth means; sixth means for calculating an absolute value of the mantissa section of the output from the fourth means; seventh means for generating a signal representative of a final arithmetic mode on the basis of the sign sections of the first and second floating point data, a result of the determination by the firs means about which of the exponent sections of the first and second floating point data is greater, and an arithmetic mode designation signal, for outputting the final arithmetic mode signal to said arithmetic means, and for, in cases where the final arithmetic mode corresponds to subtraction and the output from the first means indicates that there is a difference in exponent between the first and second floating point data, instructing said arithmetic means to subtract the output of the second means from the output of the third means; eighth means for, in cases where the output from the first means represents that there is a difference in exponent between the first and second floating point data, enabling the fifth means to execute its function, for, in cases where the output from the first means represents that there is no difference in exponent between the first and second floating point data and the final arithmetic mode corresponds to addition, enabling the fifth means to execute its function, and for, in cases where the output from the first means represents that there is no difference in exponent between the first and second floating point data and the final arithmetic mode corresponds to subtraction, enabling the sixth means to execute its function; and ninth means for determining a sign of a final arithmetic result on the basis of the sign sections of the first and second floating point data, the result of the determination by the first means about which of the exponent sections of the first and second floating point data is greater, the arithmetic mode designation signal, and a sign bit of the output from said arithmetic means; wherein the fifth means and the sixth means each utilize an adding means, the adding means being common to the fifth means and the sixth means.

Claim 6

A floating-point arithmetic apparatus for first and second floating point data each having an exponent section, a mantissa section expressed in an absolute value, and a sign section representing a sign of the mantissa section, the apparatus comprising: first means for calculating the difference between the exponent sections of the first and second floating point data, determining which of the exponent sections of the first and second floating point data is greater, and selecting the greater exponent section of the first and second floating point data; second means responsive to an output from the first means for selecting the mantissa section of one of the first and second floating point data which has the smaller exponent section, and shifting rightward the selected mantissa section by a bit number corresponding to the difference between the exponent sections; third means responsive to the output from the first means for selecting the mantissa section of one of the first and second floating point data which has the greater exponent section; an arithmetic means for performing addition and subtraction between outputs from the second and third means; fourth means for normalizing a mantissa section which is an output from said arithmetic means, and correcting the greater exponent section of the first and second floating point data in correspondence with the normalizing; fifth means for rounding off a mantissa section of an output from the fourth means; sixth means for calculating an absolute value of the mantissa section of the output from the fourth means; seventh means for generating a signal representative of a final arithmetic mode on the basis of the sign sections of the first and second floating point data, a result of the determination by the firs means about which of the exponent sections of the first and second floating point data is greater, and an arithmetic mode designation signal, for outputting the final arithmetic mode signal to said arithmetic means, and for, in cases where the final arithmetic mode corresponds to subtraction and the output from the first means indicates that there is a difference in exponent between the first and second floating point data, instructing said arithmetic means to subtract the output of the second means from the output of the third means; eighth means for, in cases where the output from the first means represents that there is a difference in exponent between the first and second floating point data, enabling the fifth means to execute its function, for, in cases where the output from the first means represents that there is no difference in exponent between the first and second floating point data and the final arithmetic mode corresponds to addition, enabling the fifth means to execute its function, and for, in cases where the output from the first means represents that there is no difference in exponent between the first and second floating point data and the final arithmetic mode corresponds to subtraction, enabling the sixth means to execute its function; and ninth means for determining a sign of a final arithmetic result on the basis of the sign sections of the first and second floating point data, the result of the determination by the first means about which of the exponent sections of the first and second floating point data is greater, the arithmetic mode designation signal, and a sign bit of the output from said arithmetic means; wherein the fifth means and the sixth means each utilize an adding means, the adding means being common to the fifth means and the sixth means.

Claim 7

The apparatus of claim 4, further comprising means for, when output from the fifth means overflows, shifting rightward the output from the fifth means by one bit and correcting an exponent section of an output from the fourth means.

Claim 8

The apparatus of claim 4, wherein the said arithmetic means and the sixth means each include an adder.