Hybrid floating point/logarithmic number system arithmetic processor

A processor for performing an arithmetic operation on a pair of input operands in floating point format to provide an output in floating point format having an integer exponent part e.sub.s and a mantissa part m.sub.s, said processor comprising:

• an input means for converting the floating point input operands to intermediate operands in logarithmic number system format;
• a logarithmic number system arithmetic means for performing an arithmetic operation on the logarithmic number system intermediate operands and providing an intermediate output in logarithmic number system format represented by an exponent e'.sub.s ;
• and
• an output means for converting the logarithmic number system intermediate output in floating point format, said output means including
• a ceiling function truncator for determining, from the exponent e'.sub.s representing the intermediate output in logarithmic number system format, an integer value e'.sub.s , the integer value e'.sub.s being the exponent partr e.sub.s of the floating point output, and
• an output mapping device for determining the mantissa part m.sub.s of the floating point output, said output mapping device satisfying the function
• m.sub.s =.phi..sup.-1 (e'.sub.s - e'.sub.s )=r.sup.(e'.sub.s - e'.sub.s )
• where r is the radix.

An accumulator for summing a series of input operands in floating point format each having an integer exponent e.sub.x and a mantissa part m.sub.x to provide an output in floating point format having an integer exponent part e.sub.s and a mantissa part m.sub.s, said accumulator comprising:

• an input means for converting the floating point input operands to a series of intermediate operands in logarithmic number system format, said input means including:
• an input mapping device for converting the mantissa m.sub.x of a floating point input operand to an exponent .DELTA., said input mapping device satisfying the function
• .phi.(m)=.DELTA.=log.sub.r (m),
• where m is the mantissa m.sub.x of the floating point input operand, and r is the radix, and
• an adder for summing the exponent .DELTA. with the exponent e.sub.x of the floating point input operand to provide an exponent e' representing the operand in logarithmic number system format;
• a logarithmic number system accumulator for successively adding the intermediate operands to accumulate a sum in logarithmic number system format represented by an exponent e'.sub.s ;
• and
• an output means for converting the logarithmic number system sum to the output in floating point format.

An accumulator for summing a series of input operands in floating point format to provide an output in floating point format having an integer exponent part e.sub.s and a mantissa part m.sub.s, said accumulator comprising:

• an input means for converting the floating point input operands to a series of intermediate operands in logarithmic number system format;
• a logarithmic number system accumulator for successively adding the intermediate operands to accumulate a sum in logarithmic number system format represented by an exponent e'.sub.s ;
• and
• an output means for converting the logarithmic number system sum to the output in floating point format, said output means including:
• a ceiling function truncator for determining, from the exponent e'.sub.s representing the sum in logarithmic number system format, an integer value e'.sub.s , the integer value e'.sub.s being the exponent part e.sub.s of the floating point output, and
• an output mapping device for determining the mantissa part ms of the floating point output, said output mapping device satisfying the function
• m.sub.s =.phi..sup.-1 (e'.sub.s - e'.sub.s )=r.sup.(e's- e's )
• wherein r is the radix.

A converter for converting a number expressed in floating point format to logarithmic number system format, said converter comprising:

• an input mapping device for converting a mantissa m of a floating point input operand to an exponent .DELTA., said input mapping device satisfying the function
• .phi.(m)=.DELTA.=log.sub.r (m),
• where m is the mantissa of the floating point input operand, and r is the radix;
• and
• an adder for summing the exponent .DELTA. with an exponent e of the floating point input operand to provide an exponent e' representing an operand in the logarithmic number system format.

A converter for converting a number expressed in logarithmic number system format to normalized floating point format, said converter comprising:

• a ceiling function truncator for determining, from an exponent e'.sub.s representing an intermediate output in the logarithmic number system format, an integer value e'.sub.s , the integer value e'.sub.s the exponent e.sub.s of the floating point output;
• and
• an output mapping device for determining the mantissa m.sub.s of the floating point output, said output mapping device satisfying the function
• m.sub.s =.sup..phi.-1 (e's- e'.sub.s =r.sup.(e'.sub.s - e'.sub.s
• where r is the radix.