US4622650A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Circuitry for generating scalar products and sums of floating point numbers with maximum accuracy
Assignee: Individual
•
Filing: 1985-08-09
•
Granted: 1986-11-11
•
Expires: 2005-08-09
Abstract
Abstract Circuitry for generating scalar products and sums of floating point numbers with maximum accuracy and circuitry and a method for electronic computers by which scalar products of floating point numbers of the type pi, qi, ES(b,l,e1,e2) are summed with full precision in a fixed point representation by means of a summing unit (ALU) and one or more accumulator registers (ARC1, ARC2) with cells (Ai, j) for storing of codes of a base b having a length (2l+ 2 e1+ 2e2) for fixed point representation and certain overflow positions. By control means (SHR, E, Contr) the mantissas of products are delivered depending on the value of the respective exponents into the summing unit (ALU). By control means (RD, Contro), rounding operations ( ○ , , ∇, .increment.) demanded by the higher level computer are performed, and a rounded floating point number (□ c ε S(b,l,e1,e2)) and overflow (OF) and underflow (UF) criteria are delivered. Parallel, serial and word organized summing units (ALU) and accumulator registers (ACR) are usable and in another embodiment, the multiplication of the factors (pi, qi) is performed using a table of multiples store.
Inventors
- 👤 Ulrich Kulisch
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-29 17:24
CPC Classifications
G; G06; G06F; G06F7/00; G06F7/38; G06F7/48; G06F7/544; G06F7/5443; G06F7/483; G06F7/499; G06F7/49905; G06F7/4991; G06F7/49942; G06F7/49947
Claims Analysis
Total Claims: 13
Claim 1
Independent Claim
244 words
In an electronic computer operating in a floating point system and wherein data is defined as a function of (b, l, e 1 , e 2 ) where b is the base, 1 is the number of digits of the mantissa, e 1 is the least exponent and e 2 is the highest exponent, the improvement comprising means receiving two floating point vectors each having a finite number of vector components for performing a scalar product operation thereon with maximum accuracy wherein the scalar product is produced from a sum of a plurality of partial scalar products of the vector components, including first means for performing fixed point arithmetic operations on the mantissas and exponents of each of the vector components to produce the plurality of partial scalar products, second means receiving the partial scalar products from the first means for storing each partial scalar product in fixed point format in a store having a length L which is 21+2e 2 +2 times greater than the absolute value of e 1 ; third means receiving the partial scalar products from the second means for summing the stored fixed point partial scalar products; fourth means receiving the summed partial scalar products from the third means for rounding the resulting sum of partial scalar products to the floating point number closest thereto to obtain maximum accuracy; and fifth means receiving the rounded sum from the fourth means for converting the rounded sum into floating point format.
Claim 2
Dependent Claim
244 words
In an electronic computer operating in a floating point system and wherein data is defined as a function of (b, l, e 1 , e 2 ) where b is the base, 1 is the number of digits of the mantissa, e 1 is the least exponent and e 2 is the highest exponent, the improvement comprising means receiving two floating point vectors each having a finite number of vector components for performing a scalar product operation thereon with maximum accuracy wherein the scalar product is produced from a sum of a plurality of partial scalar products of the vector components, including first means for performing fixed point arithmetic operations on the mantissas and exponents of each of the vector components to produce the plurality of partial scalar products, second means receiving the partial scalar products from the first means for storing each partial scalar product in fixed point format in a store having a length L which is 21+2e 2 +2 times greater than the absolute value of e 1 ; third means receiving the partial scalar products from the second means for summing the stored fixed point partial scalar products; fourth means receiving the summed partial scalar products from the third means for rounding the resulting sum of partial scalar products to the floating point number closest thereto to obtain maximum accuracy; and fifth means receiving the rounded sum from the fourth means for converting the rounded sum into floating point format.
Claim 3
Dependent Claim
33 words
The computer according to claim 1, wherein the first means for performing fixed point arithmetic operations includes an arithmetic unit and the second means for storing comprises an accumulator having the length L.
Claim 4
Dependent Claim
31 words
The computer according to claim 2, wherein the accumulator comprises two registers of length L and means for storing negative numbers in one register and positive numbers in the other register.
Claim 5
Dependent Claim
16 words
The computer according to claim 3, wherein the arithmetic unit comprises a register of length L.
Claim 6
Dependent Claim
76 words
The computer according to claim 2, wherein the first means comprises a mantissa register into which the mantissa of each number is loaded an exponent register into which the exponent of each number is loaded and the control and logic circuits are connected to the outputs of mantissa and exponent registers to feed the mantissa to the arithmetic unit to be summed and into the accumulator wherein the exponent defines the digit positions in the accumulator.
Claim 7
Dependent Claim
126 words
The computer according to claim 5, wherein the control and logic circuits produce a first control signal indicating the position of the cells in the accumulator containing the most significant digit, a second control signal indicating the position of the cells containing digits of significance, a third control signal indicating that the accumulator contains all zeros, a fourth control signal indicating if there are no zero digits contained in cells (Ai, j) outside a range of at least once the length of the mantissa of the factor (pi), a fifth signal indicating that digits of higher significance compared to a limit are present in the cells, and a sixth signal indicating that only digits of lower significance compared to a limit are present in the cells.
Claim 8
Dependent Claim
45 words
The computer according to claim 2, further comprising a gate arrangement connected between the outputs of a summing network of the arithmetic unit and the inputs of the accumulator for delivering the absolute value of the sum generated in the summing network to the accumulator.
Claim 9
Dependent Claim
53 words
The computer according to claim 5, wherein: the arithmetic unit comprises a parallel working unit of a length equal to the length of the accumulator and the control and logic circuits connect the outputs of the mantissa register to the inputs of the arithmetic unit defined by the contents of the exponent register.
Claim 10
Dependent Claim
90 words
The computer according to claim 5, wherein the accumulator comprises a serial working register and the arithmetic unit operates serially and the control and logic circuits and the exponent register comprise are counters and shift registers and the mantissa register is connected to the arithmetic unit during times given by the contents of the exponent register and that the times of rounding and delivering of the floating point number are defined relative to the timely appearance of the most significant digit of the serial information string of the accumulator contents.
Claim 11
Dependent Claim
46 words
The computer according to claim 2, wherein the accumulator register consists of sections with parallel addressable access a length equal to the length of the mantissa of the factor (pi) and the arithmetic is organized in parallel having at least the same length as the sections.
Claim 12
Dependent Claim
28 words
The computer according to claim 2, wherein the accumulator comprises two registers and multiplexer circuits for alternatively connecting same to the inputs and outputs of the arithmetic unit.
Claim 13
Dependent Claim
39 words
The computer according to claim 1, wherein the second means comprises accumulator register of the length of at least n-times the length of the mantissa plus n-times the sum of the absolute value of the smallest and largest exponent.
References Cited (Backward Citations)
US3388385A
Cited
US3699326A
Cited
US3871578A
Cited
US4054787A
Cited
US4149261A
Cited
US4075704A
Cited
US4135249A
Cited
US4130879A
Cited
US4229800A
Cited
US4295203A
Cited
US4338675A
Cited
US4409668A
Cited
US4490807A
Cited
US4534010A
Cited
US4429370A
Cited
US4476537A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1989-11-10 | 1990-05-12 | 1990-11-11 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1993-11-10 | 1994-05-12 | 1994-11-11 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 1997-11-10 | 1998-05-12 | 1998-11-11 | $7700.00 | Window Closed / Lapsed |