Multiplier with index transforms modulo a prime or modulo a fermat prime and the fermat prime less one
OwnerAisuke Katayamainventor
Life of this patent
Estimating the term…
Abstract
An asynchronous data-processing system for multiplying two binary numbers, by a use of read-only memories storing tables of data for transforming the numbers into exponents of a prime number. The exponents are added and then transformed back into the terms of the original numbers system. The transforms produce the product of the two numbers merely by addressing read-only memories and, therefore, accomplish the multiplication at a very high speed. Sophistications of the system compensate for variations in the bit patterns of input signals, inabilities to handle some numbers in the numbering systems, and the like. A use of Fermat prime numbers eliminates rounding errors which occur in systems using logarithmic transforms.
Claims20
A multiplication system for calculating a product of two integers by a use of a prime number m modulus and a predetermined natural number raised to the power of consecutive whole numbers which are congruent with residues modulo of said prime number m, said predetermined natural number being a primitive root inherent in said prime number m, and when arranged in an ascending order, said residues form a set of consecutive integers, whereby a one-to-one correspondence is established between said whole numbers and said consecutive integers, said two integers being given as a first and a second of said consecutive integers, said multiplication system comprising:
- first transforming means for transforming said first and said second integers into a first and a second index, respectively, with said first and said second indices being given by those two of said whole numbers which are in a one-to-one correspondence with said first and said second integers, respectively;
- summing means for deriving a modulus sum responsive to said first and said second indices, said modulus sum being a sum of said first and said second indices which is congruent with said modulus sum modulo of said prime number m less one;
- and
- second transforming means for transforming said modulus sum to said product with said product given by that one of said consecutive integers which is in a one-to-one correspondence with one of said whole numbers, said one whole number being given by said modulus sum.
A multiplication system for calculating a product of two odd integers by a use of a first and a second modulus equal to a prime number m which is given by a Fermat number which is not less than five and equal to said prime number m less one, respectively, and by a use of a first and a second congruence relationship, said first congruence relationship being such that a first predetermined natural number raised to the power of consecutive whole numbers of a first group, (m-1)/2 in number, are congruent with those residues modulo of said prime number m which form a set of consecutive odd integers, when said residues are arranged in the ascending order, said first predetermined natural number being a primitive root inherent to said prime number m, said second congruence relationship being such that a negative unity raised to the powers 0 and 1 multiplied by a second predetermined natural number raised to the powers of consecutive whole numbers of a second group, (m-1)/4 in number, are congruent with those residues of said second modulus which form said set of consecutive odd integers when said second modulus is arranged in the ascending order, said second predetermined natural number being a quasi-primitive root inherent to said second modulus, whereby a one-to-one correspondence is established between said first-group of whole numbers and said consecutive odd integers according to said first congruence relationship and is established between combinations of 0 and 1 with said second group of whole numbers and said consecutive odd integers according to said second congruence relationship, said two odd integers being given as a first and a second of said consecutive odd integers, said second odd integer being enabled to be identical with said first odd integer, said multiplication system comprising:
- first transforming means for transforming said first and said second odd integers into a first and a second index of a first group, respectively, with said first and said second first-group indices being given by those two of said first group of whole numbers which are in a one-to-one correspondence with said first and said second odd integers, respectively, according to said first congruence relationship;
- first summing means responsive to said first and said second first-group indices for deriving a modulus sum of a first kind with which a sum of said first and said second first-group indices is congruent with respect to said second modulus;
- second transforming means for transforming said first kind of modulus sum into a first residue with said first residue being given by that one of said consecutive odd integers which is in a one-to-one correspondence with one of said first group of whole numbers according to said first congruence relationship, said one first group of whole numbers being given by said first kind of modulus sum;
- third transforming means for transforming said first odd integer into a first index of a first set and a first index of a second set, both of said sets being in a second group of indices, said third transforming means transforming said second odd integer into a second index of said first set and a second index of said second set, a first combination of said first first-set and said first second-set indices and a second combination of said second first-set and said second second-set indices being given by those two of said combinations of 0 and 1 with said second group of whole numbers which are in one-to-one correspondence with said first and said second odd integers, respectively, according to said second congruence relationship;
- second summing means responsive to said first and said second first-set indices for deriving a first modulus sum of a second kind with which a sum of said first and said second first-set indices is a congruent modulo two and responsive to said first and said second second-set indices for deriving a second modulus sum of said second kind with which a sum of said first and said second second-set indices is a congruent modulo (m-1)/4;
- fourth transforming means for transforming a combination of said first and said second second-kind of modulus sums into a second residue with said second residue being given by that one of said consecutive odd integers which is in a one-to-one correspondence with one of said combinations of 0 and 1 with said second group of whole numbers, said one combination being given by said combination of first and second second-kind modulus sums;
- and
- combining means for combining said first and said second residues into said product.
A multiplication system for calculating a product of two integers by a use of a first and a second modulus which is equal to a prime number m given by a Fermat number which is not less than five and is equal to said prime number m less one, respectively, and by a use of first and second congruence relationships, said first congruence relationship being such that a first predetermined natural number raised to the power of those consecutive whole numbers (which are (m-1)/2 in number) of a first group which are congruent with those residues modulo of said prime number m which form a set of consecutive odd integers, when the residues are arranged in the ascending order, said set of odd integers also being (m-1)/2 in number and being arranged in the ascending order, said first predetermined natural number being a primitive root which is inherent to said prime number m, said second congruence relationship being such that minus unity raised to the power of 0 and 1 multiplied by a second predetermined natural number raised to the power of (m- 1)/4 consecutive whole numbers of a second group, are congruent with those residues with respect to said second modulus which, when arranged in the ascending order, form said set of consecutive odd integers, said second predetermined natural number being a quasi-primitive root inherent to said second modulus, whereby a one-to-one correspondence is established between said first-group of whole numbers and said consecutive odd integers according to said first congruence relationship and between combinations of 0 and 1 with said second group of whole numbers and said consecutive odd integers according to said second congruence relationship, said two integers being given by a first and a second h-bit binary signal where h represents that exponent of the number "2" which represents said Fermat number as (2.sup.h +1), said multiplication system comprising:
- means responsive to said first and said second h-bit binary signals for producing a first and a second (h-1)-bit binary signal with the least significant bit being removed from each of said first and said second h-bit binary signals, said first and said second (h-1)-bit binary signals representing first and second odd integers when a binary "1" bit is added as a new least significant bit;
- first transforming means for transforming said first and said second (h-1)-bit binary signals into a first and a second index signal of a first group, respectively, each of said first and said second first-group index signals being of h binary bits, said first and said second first-group index signals being representative of a first and a second index of a first group, respectively, said first and said second first-group indices being those two of said first-group of whole numbers which are in a one-to-one correspondence with said first and said second odd integers, respectively, according to said first congruence relationship;
- first summing means responsive to said first and said second first-group of index signal for producing a modulus sum signal of a first kind of h binary bits, said first-kind of modulus sum signals being representative of a modulus sum of a first kind with which a sum of said first and said second first-group indices is congruent with respect to said second modulus;
- second transforming means for transforming said first kind of modulus sum signal into a first residue signal of h binary bits, said first residue signal being representative of a first residue given by that one of said consecutive odd integers which is in a one-to-one correspondence with one of said first group of whole numbers according to said first congruence relationship, said one first group of whole number being given by said first kind of modulus sum;
- third transforming means for transforming said first (h-1)-bit binary signal into a first combination of a first index signal of a first set and a first index signal of a second set, both sets being in a second group of index signals, and for transforming said second (h-1)-bit binary signal into a second combination of a second index signal of said first set and a second index signal of said second set, each of said first set of index signals being of single binary bits, each of said second-set index signals being of (h-2) binary bits, said first and said second first-set index signals being representative of a first and a second index of a first set, respectively, said first and said second second-set index signals being representative of a first and a second index of a second set, both sets of indices being in a second group of indices, said first combination of a first first-set and a first second-set indices and said second combination of a second first-set and a second second-set indices being given by those two of said combinations of 0 and 1 with said second group of whole numbers being in a one-to-one correspondence with said first and said second odd integers, respectively, according to said second congruence relationship;
- second summing means responsive to said first and said second first-set index signals for producing a first modulus sum signal of a second kind and responsive to said first and said second second-set index signals for producing a second modulus sum signal of said second kind, said first second-kind of modulus sum signal being a single binary bit and being representative of a first modulus sum of a second kind with which a sum of said first and said second first-set indices is a congruent modulo "2", said second second-kind of modulus sum signal being (h-2) binary bits and being representative of a second modulus sum of said second kind with which a sum of said first and said second second-set indices is congruent modulo (m-1)/4;
- fourth transforming means for transforming a combination of said first and said second second-kind of modulus sum signals to a second residue signal of h binary bits, said second residue signal being representative of a second residue given by that one of said consecutive odd integers which is in a one-to-one correspondence with one of said combination of 0 and 1 with said second group of whole numbers, said one combination being given by said combination of first and second second-kind of modulus sum;
- combining means for combining said first and said second residue signals into a resulting signal of 2h binary bits and being representative of a result of a multiplication of said first and said second odd integers;
- correction signal-producing means responsive to said first and said second h-bit binary signal for producing a correction signal representative of a correction integer which is equal to zero when each of two integers is odd, equal to the integer represented by said second h-bit binary signal when the integer represented by said first h-bit binary signal is even, equal to, the integer represented by said first h-bit binary when the integer represented by said second h-bit binary signal is even, and equal to a sum of said two integers, and unity, when the integer represented by said second h-bit binary signal is even, and each of said two integers is even;
- and
- adjusting means responsive to said resulting signal and said correction signal for producing a product signal of 2h binary bits and which is representative of said product with said product being given by subtracting said correction integer from said results of said multiplication.
An asynchronous binary multiplication system driven by input signals identifying first and second numbers which are a multiplicand and a multiplier, said system comprising first and second read-only memory means, each of said memory means storing a table of data for transforming individually associated input signals into non-logarithmic exponent signals corresponding to said input signals, modulus adding means for adding said exponent signals, and inverse transform means responsive to said adding means for transforming said added exponent signals into a signal indicating the product of said first number multiplied by said second number, said product being free of rounding errors, said inverse transform means comprising a read-only memory storing a table of data for converting said adding exponent signals into said product.
An asynchronous binary multiplication system comprising a pair of read-only memory means, a pair of demultiplexing means for applying multiplicand and multiplier signals with a predetermined bit pattern to individually associated ones of said read-only memory means, said read-only memory means containing stored tables of data for transforming the output of the associated demultiplexing means into corresponding non-logarithmic exponent signals, means for adding the transformed output of said memory means, means responsive to said adding means for transforming the added outputs of said memory means into signals representing a product of said multiplication, said product being free of rounding errors, demultiplexing means for converting said product signals into predetermined bit patterns, and control means on each of said demultiplexing means for selecting said bit pattern responsive to at least one signal in said multiplicand and multiplier signals.
A method of asynchronously multiplying two of a set of binary numbers representing a multiplicand and a multiplier, said method comprising the steps of:
- a. separately transforming each of the two numbers into non-logarithmic exponents which individually correspond to said two numbers, said transforms being accomplished by selectively addressing read-only memories storing tables of data representing sets of numbers including at least said two numbers;
- b. adding the non-logarithmic transform exponents;
- and
- c. transforming the sum of the added exponents back into said set of numbers by addressing another read-only memory containing tables of data representing a product of the multiplication in terms of said added exponents, said product being free of rounding errors.
Dependencies are inferred from the claim text.