VLSI circuit structure for determining the edit distance between strings

Claim 2

A VLSI circuit structure for use in determining an edit distance between two strings, comprising (a) a systolic array of processing elements; (b) each processing element receiving as inputs a pair of string symbols; (c) each processing element including a comparator which compares the pair of string symbols input to the processing element and outputs a signal which has a first value if the pair of string symbols match and a second value if the pair of the string symbols do not match; (d) each processing element further including a difference computation module which receives difference values associated with the pair of string symbols input to the processing element and also receives the output of the comparator, said difference computation module having stored therein editing costs for predetermined editing operations associated with the string symbols, said difference computation module producing a pair of outputs for the pair of string symbols, each of the outputs comprising a new difference value related to the editing cost for the pair of string symbols that was input and compared by the comparator, and (e) each processing element producing a pair of outputs, each of which comprises one string symbol of the pair of string symbols and the new difference value related to the editing cost for the pair of string symbols.

Claim 3

A VLSI circuit structure as defined in claim 1, wherein the stystolic array of processing elements is designed to compute a distance matrix, each matrix element of which is represented by a pair of difference values and the left and top boundaries of which comprise accumulated edit values, each difference value representing the difference between a matrix element and its left and top neighbors, and wherein each difference value has a range from ±max (I,D), where I and D are insertion and deletion cost values stored in the computation modules of the processing elements.

Claim 4

A VLSI circuit structure as defined in claim 2, wherein the systolic array of processing elements is designed such that a processing element receives inputs from and direct outputs to processing elements on its opposite sides, the difference computation module of a processing element comprising a three-way minimum comparator which receives the 1 or 0 output from the comparator of the processing element and a pair of difference value inputs (left in , right in ) from the processing elements on the left and right sides of the processing element, and directs difference value outputs (left out , right out ) to the processing elements on the left and right sides of the processing element; the pair of difference value outputs (left out , right out ) being determined by the three-way minimum comparator as follows: right.sub.out =min(left.sub.in +I, right.sub.in +D, dia)-right.sub.in left.sub.out =min(left.sub.in +I, right.sub.in +D, dia)-left.sub.in Where dia=0 if the output of the comparator is 1 dia=S if the output of the comparator is 0 S=substitution cost value stored in the difference computation module D=deletion cost value stored in the difference computation module I=insertion cost value stored in the difference computation module.

Claim 5

A VLSI circuit structure according to claim 3 wherein said outputs of said processing elements comprise strings of string symbols with new difference values interleaved between the string symbols.

Claim 6

A VLSI circuit structure according to claim 4 further comprising an accumulator in circuit communication with one of said processing elements, said accumulator accumulating a value having a larger bit-width than the difference values.

Claim 7

A VLSI circuit structure according to claim 3 further comprising an accumulator in circuit communication with one of said processing elements, said accumulator accumulating a value having a larger bit-width than the difference values.

Claim 8

A VLSI circuit structure according to claim 2 wherein said outputs of said processing elements comprise strings of string symbols with new difference values interleaved between the string symbols.

Claim 9

A VLSI circuit structure according to claim 7 further comprising an accumulator in circuit communication with one of said processing elements, said accumulator accumulating a value having a larger bit-width than the difference values.

Claim 10

A VLSI circuit structure according to claim 2 further comprising an accumulator in circuit communication with one of said processing elements, said accumulator accumulating a value having a larger bit-width than the difference values.

Claim 11

A VLSI circuit structure according to claim 1 wherein said outputs of said processing elements comprise strings of string symbols with new difference values interleaved between the string symbols.

Claim 12

A VLSI circuit structure according to claim 10 further comprising an accumulator in circuit communication with one of said processing elements, said accumulator accumulating a value having a larger bit-width than the difference values.

Claim 13

A VLSI circuit structure according to claim 1 further comprising an accumulator in circuit communication with one of said processing elements, said accumulator accumulating a value having a larger bit-width than the difference values.

Claim 14

A circuit structure for use in determining an edit distance between a first string of string symbols and a second string of string symbols, comprising: (a) a linear systolic array of processing elements having a first end processing element and a second end processing element, the first string of string symbols being input into said linear systolic array at said first end processing element and the second string of string symbols being input into said linear systolic array at said second end processing element; (b) each processing element receiving as inputs string symbols with at least one value interleaved between the string symbols; (c) each processing element including a comparator which compares at least a pair of the string symbols input to the processing element and generates an output signal which has a first value if the pair of the string symbols match and a second value if the pair of the string symbols do not match; (d) each processing element further including a difference computation module which receives the at least one value interleaved between the string symbols input to the processing element and also receives a signal associated with the output signal of the comparator; said difference computation module having stored therein editing costs for predetermined editing operations associated with the string symbols; said difference computation module producing a pair of outputs for each pair of the string symbols, each of the outputs comprising a new scaled value related to the editing cost for the pair of the string symbols that was input and compared by the comparator; and (e) each processing element generating a pair of outputs, each of which comprises one string symbol of the pair of the string symbols and the new scaled value related to the editing cost for the pair of the string symbols.

Claim 15

A circuit structure for use in determining an edit distance between a first string of string symbols and a second string of string symbols, comprising: (a) a linear systolic array of processing elements having a first end processing element and a second end processing element, the first string of string symbols being input into said linear systolic array at said first end processing element and the second string of string symbols being input into said linear systolic array at said second end processing element; (b) each processing element receiving as inputs string symbols with at least one value interleaved between the string symbols; (c) each processing element including a comparator which compares at least a pair of the string symbols input to the processing element and generates an output signal which has a first value if the pair of the string symbols match and a second value if the pair of the string symbols do not match; (d) each processing element further including a difference computation module which receives the at least one value interleaved between the string symbols input to the processing element and also receives a signal associated with the output signal of the comparator; said difference computation module having stored therein editing costs for predetermined editing operations associated with the string symbols; said difference computation module producing a pair of outputs for each pair of the string symbols, each of the outputs comprising a new scaled value related to the editing cost for the pair of the string symbols that was input and compared by the comparator; and (e) each processing element generating a pair of outputs, each of which comprises one string symbol of the pair of the string symbols and the new scaled value related to the editing cost for the pair of the string symbols.

Claim 16

A circuit structure according to claim 13, wherein said linear systolic array of processing elements computes a distance matrix, each matrix element of which is represented by a pair of difference values and the left and top boundaries of which comprise accumulated edit values, each difference value representing the difference between a matrix element and its left and top neighbors, and wherein each difference value has a range from±max (I,D), where I and D are insertion and deletion cost values stored in the computation modules of the processing elements.

Claim 17

A circuit structure according to claim 14, said difference computation modules comprising a three-way minimum comparator which receives a signal associated with the output signal of the comparator and a pair of difference value inputs (left in , right in ) from the processing elements on the left and right sides of the processing element, and directs difference value outputs (left out , right out ) to the processing elements on the left and right sides of the processing element; the pair of difference value outputs (left out , right out ) being determined by the three-way minimum comparator as follows: right.sub.out =min(left.sub.in +I, right.sub.in +D, dia)-right.sub.in left.sub.out =min(left.sub.in +I, right.sub.in +D, dia)-left.sub.in Where dia=0 if the output of the comparator is 1 dia=S if the output of the comparator is 0 S=substitution cost value stored in the difference computation module D=deletion cost value stored in the difference computation module I=insertion cost value stored in the difference computation module.

Claim 18

A circuit structure according to claim 15 further comprising an accumulator in circuit communication with one of said end processing elements, said accumulator accumulating a value having a larger bit-width than the difference values.

Claim 19

A circuit structure according to claim 13 wherein said outputs comprise strings of string symbols with the new scaled values interleaved between the string symbols.

Claim 20

A circuit structure according to claim 13 further comprising an accumulator in circuit communication with one of said end processing elements, said accumulator accumulating a value having a larger bit-width than the new scaled values.

Claim 21

A circuit structure according to claim 13 wherein said values interleaved between the string symbols comprise difference values.

Claim 22

A circuit structure according to claim 19 wherein said outputs comprise strings of string symbols with the new scaled values interleaved between the string symbols.

Claim 23

A circuit structure according to claim 20 further comprising an accumulator in circuit communication with one of said end processing elements, said accumulator accumulating a value having a larger bit-width than the difference values.

Claim 24

A circuit structure according to claim 13 wherein said values interleaved between the string symbols consist of difference values.