EP0461240A4 - Methods and apparatus for efficient resource allocation for error and exception handling in convergent division - Google Patents
Methods and apparatus for efficient resource allocation for error and exception handling in convergent divisionInfo
- Publication number
- EP0461240A4 EP0461240A4 EP19910902487 EP91902487A EP0461240A4 EP 0461240 A4 EP0461240 A4 EP 0461240A4 EP 19910902487 EP19910902487 EP 19910902487 EP 91902487 A EP91902487 A EP 91902487A EP 0461240 A4 EP0461240 A4 EP 0461240A4
- Authority
- EP
- European Patent Office
- Prior art keywords
- value
- input value
- seed
- input
- convergence function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F7/38—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation
- G06F7/48—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using non-contact-making devices, e.g. tube, solid state device; using unspecified devices
- G06F7/52—Multiplying; Dividing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F7/38—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation
- G06F7/48—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using non-contact-making devices, e.g. tube, solid state device; using unspecified devices
- G06F7/52—Multiplying; Dividing
- G06F7/535—Dividing only
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2207/00—Indexing scheme relating to methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F2207/535—Indexing scheme relating to groups G06F7/535 - G06F7/5375
- G06F2207/5355—Using iterative approximation not using digit recurrence, e.g. Newton Raphson or Goldschmidt
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F7/38—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation
- G06F7/48—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using non-contact-making devices, e.g. tube, solid state device; using unspecified devices
- G06F7/483—Computations with numbers represented by a non-linear combination of denominational numbers, e.g. rational numbers, logarithmic number system or floating-point numbers
- G06F7/487—Multiplying; Dividing
- G06F7/4873—Dividing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F7/38—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation
- G06F7/48—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using non-contact-making devices, e.g. tube, solid state device; using unspecified devices
- G06F7/499—Denomination or exception handling, e.g. rounding or overflow
- G06F7/49905—Exception handling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F7/38—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation
- G06F7/48—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using non-contact-making devices, e.g. tube, solid state device; using unspecified devices
- G06F7/499—Denomination or exception handling, e.g. rounding or overflow
- G06F7/49905—Exception handling
- G06F7/49926—Division by zero
Definitions
- efficient resource allocation together with an appropriate mapping of data input, allow processing of all data inputs in a continuous, accurate, and efficient manner in a digital computer such as a digital signal processor utilizing a computer program storage medium having a computer program stored thereon that modifies and propagates data inputs through a convergent division algorithm.
- a divisor input is assigned a seed value according to a predetermined categorization of divisor inputs, and is modified to indicate a divisor value suitable for efficient propagation through a convergent division algorithm.
- This selection of two feasible values for implementation of the convergent division algorithm is a useful preliminary determination for efficient computation.
- the solution of a convergent division algorithm for all inputs by careful choice of seed values and modified divisor values constrains the computation to a substantially shorter duration than in the past, improving cost- effectiveness of signal processing. Utilization of fewer resources to allocate the seed values and to modify the divisors also is cost-effective.
- Figure 1 is a general flow chart depicting implementation of a floating-point division algorithm under the prior art.
- Figure 2 is a flow chart of an implementation of the invention wherein a seed and a modified input value are assigned in parallel.
- Figure 3 is a flow chart of an implementation of the invention wherein a seed and a modified input value are assigned in series.
- Figure 4 is a block diagram of a computer hardware implementation of the invention.
- Figure 5 is a table setting forth the error matrix achieved without exception correction, according to the prior art.
- Figure 6 is a table setting forth an application of the present invention for exception correction.
- Figure 7 is a table illustrating the output seed and corrected divisor for a given input divisor according to the present invention.
- FIG. 1 is a graphical representation, depicted generally by the numeral 100, of the constraints on an implementation of floating-point division, according to the prior art.
- a programming model may provide a division mode utilizing two numbers, N and D (102), typically represented in the floating-point number form.
- the program determines whether or not a divisor D is zero (104). If the divisor D is zero (104), the program may substitute a very small number for the divisor so that division may continue (106), or may skip division altogether (105).
- the program may determine whether or not the divisor is infinity (108). If the divisor is infinity (108), the program may substitute a very large number so that division may continue (106), or may skip division altogether (109). Following the check for division by zero and for division by infinity, the program executes the division algorithm (110) in accordance with prior art technique.
- the program checks to see whether or not the desired degree of accuracy has been reached (112). If insufficient accuracy has been attained, the program recycles to repeat the division algorithm computation (110) until the desired degree of accuracy is obtained (112), then stops. If the desired degree of accuracy has already been attained (112), the program stops.
- FIG. 2 illustrates the steps executed by a programming model incorporating the present invention, wherein the implementation of a mapping of an input value D to a seed S and a modified input value D occurs in parallel, depicted generally by the numeral 200.
- the program obtains an input value D and a second value, N, related to the first value, for a division algorithm computation (202).
- the program checks for a plus or minus zero input value D, implemented as a divisor (104). If the divisor D is plus or minus zero, a seed S is assigned a corresponding infinity value with the same sign as D and a modified input value D, implemented as a divisor, is assigned a non-zero value with the opposite sign as the seed S (208), -S being a workable value.
- the program checks to see if the input value D, implemented as a divisor, is plus or minus infinity (108). If the divisor D is plus or minus infinity, a seed S is set to a corresponding plus or minus zero and a modified input value D, implemented as a divisor, is assigned a non-infinite number (210), S being a workable value. If the divisor D is not plus or minus infinity, the divisor D may be a plus or minus normalized or denormalized number according to the IEEE 754-1985 standard.
- a seed S is set to an approximation corresponding to plus or minus 1/D and the divisor D remains the same (212).
- the checking for division by zero (104) and for division by infinity (108) may be accomplished in alternate order if desired. In this embodiment all arithmetic operations, addition, subtraction, and multiplication, conform to the IEEE 754-1985 standard for binary floating-point arithmetic.
- the second value N and the divisor D, modified or unchanged are each premultiplied by the seed value S (214), yielding a numerator and a denominator, respectively.
- a convergence function value f1 is obtained by subtracting the denominator from the value 2.0.
- the program multiplies both the numerator and the denominator by the convergence function value f1 , obtaining a quotient having a next numerator and a next denominator, respectively (110).
- the next denominator is subtracted from 2.0, yielding a next convergence function value, f2.
- ⁇ D * S * fl * f2 is utilized in this invention, and a convergence factor f is obtained by subtracting the modified divisor D from the value 2.0.
- the approximated seed S must have a value such that D * S is less than 2.0.
- FIG. 3 illustrates the steps executed by a programming model incorporating the present invention wherein the implementation of a mapping of an input value D to a seed S and a modified input value D occurs in series, depicted generally by the numeral 300.
- the program obtains an input value D and a second value, N, related to the first value, for a division algorithm computation (202).
- the program checks for a plus or minus zero input value D, implemented as a divisor (104). If the divisor D is plus or minus zero, a seed S is assigned a corresponding plus or minus infinity value (308). If the divisor D is not plus or minus zero (104), the program checks to see if the divisor D is plus or minus infinity (108).
- a seed S is set to a corresponding plus or minus zero (310). If the divisor D is not plus or minus infinity (108), a corresponding seed S is set to an approximation of plus or minus 1/D (312).
- the checking for division by zero (104) and for division by infinity (108) may be accomplished in alternate order if desired.
- the program now checks to see if the seed value S is plus or minus zero (314). If the seed value S is plus or minus zero, or alternatively, since S is directly related to D, if a divisor D is equal to plus or minus infinity, a modified input value D, implemented as a divisor, is assigned a non-infinite number (318), S being a workable value . If the seed value S is not plus or minus zero (314), the program checks to see if the seed value S is plus or minus infinity (316).
- a divisor D is equal to plus or minus zero (316)
- a modified input value D is assigned a non-zero value with the opposite sign of S (320), -S being a workable value. If the seed value S is not plus or minus infinity (316), the divisor value remains the same (322).
- FIG. 4 illustrates a hardware implementation of the present invention, generally depicted by the numeral 400.
- a computer program for implementation of the present invention may be stored in the program memory (404), other memory (412), or may be embodied in hardware in the ALU (406).
- the program control unit (402) utilizes the bus (410) to select the program to implement the present invention, and the status register (408) determines whether division by zero or by infinity is taking place.
- the ALU (406) performs the previously described manipulations of values in application of the convergent division algorithm set forth above.
- the ALU (406) generates an initial seed S with a mantissa that is stored in 8 binary bits of memory.
- Figure 5 illustrates the prior art, setting forth the error matrix achieved without exception correction.
- This prior art implementation allows all operands to propagate through the quadratic convergent algorithm without any operand checking. Division by zero or division by infinity results in a NaN, a result that is inconsistent with the IEEE 754-1985 floating point standard.
- Figure 6 depicts application of the present invention for exception correction.
- the operands are modified according to the present invention, and the modified operands propagate through the quadratic convergent division algorithm using IEEE 754-1985 arithmetic, outputting quotient and exception results, including signs for zeros and infinities for exception cases, in accordance with the IEEE 754-1985 floating point standard.
- Figure 7 illustrates the output seed and corrected divisor for a given input divisor according to the present invention.
Abstract
Procédé utilisant une seule étape de prétraitement (214, 216) afin de modifier un diviseur (D) en division à point flottant, et afin de valider une opération de correction d'exception et de contrôle d'erreurs (104, 108) ayant lieu simultanément avec une opération de division. L'invention permet d'obtenir le résultat prédéterminé correct s'il existe une exception, ou le quotient si aucune exception ne se produit.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US45909089A | 1989-12-29 | 1989-12-29 | |
US459090 | 1989-12-29 |
Publications (2)
Publication Number | Publication Date |
---|---|
EP0461240A1 EP0461240A1 (fr) | 1991-12-18 |
EP0461240A4 true EP0461240A4 (en) | 1993-08-11 |
Family
ID=23823363
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP19910902487 Withdrawn EP0461240A4 (en) | 1989-12-29 | 1990-12-17 | Methods and apparatus for efficient resource allocation for error and exception handling in convergent division |
Country Status (5)
Country | Link |
---|---|
EP (1) | EP0461240A4 (fr) |
JP (1) | JPH04504774A (fr) |
KR (1) | KR940008617B1 (fr) |
CA (1) | CA2047180A1 (fr) |
WO (1) | WO1991010190A1 (fr) |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0377992A2 (fr) * | 1989-01-13 | 1990-07-18 | International Business Machines Corporation | Méthode et dispositif de division à virgule flottante |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4724529A (en) * | 1985-02-14 | 1988-02-09 | Prime Computer, Inc. | Method and apparatus for numerical division |
-
1990
- 1990-12-17 CA CA002047180A patent/CA2047180A1/fr not_active Abandoned
- 1990-12-17 KR KR1019910701004A patent/KR940008617B1/ko not_active IP Right Cessation
- 1990-12-17 WO PCT/US1990/007431 patent/WO1991010190A1/fr not_active Application Discontinuation
- 1990-12-17 EP EP19910902487 patent/EP0461240A4/en not_active Withdrawn
- 1990-12-17 JP JP3502826A patent/JPH04504774A/ja active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0377992A2 (fr) * | 1989-01-13 | 1990-07-18 | International Business Machines Corporation | Méthode et dispositif de division à virgule flottante |
Non-Patent Citations (2)
Title |
---|
See also references of WO9110190A1 * |
WEITEK APPLICATION NOTE 'WTL 1032/1033 Floating Point Division/Square Root/ IEEE Arithmetic' 1983 , WEITEK CORPORATION , SANTA CLARA, CA, USA * |
Also Published As
Publication number | Publication date |
---|---|
KR940008617B1 (ko) | 1994-09-24 |
JPH04504774A (ja) | 1992-08-20 |
WO1991010190A1 (fr) | 1991-07-11 |
CA2047180A1 (fr) | 1991-06-30 |
EP0461240A1 (fr) | 1991-12-18 |
KR920701904A (ko) | 1992-08-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109661647B (zh) | 数据处理装置和方法 | |
US6138135A (en) | Propagating NaNs during high precision calculations using lesser precision hardware | |
US10579338B2 (en) | Apparatus and method for processing input operand values | |
JPH0210427A (ja) | 精密浮動小数点例外用の方法及び装置 | |
CN111027690B (zh) | 执行确定性推理的组合处理装置、芯片和方法 | |
JPH06202850A (ja) | データ処理装置 | |
US5195052A (en) | Circuit and method for performing integer power operations | |
JPH0145649B2 (fr) | ||
US6912559B1 (en) | System and method for improving the accuracy of reciprocal square root operations performed by a floating-point unit | |
US9582469B1 (en) | System and methods for determining attributes for arithmetic operations with fixed-point numbers | |
US10459688B1 (en) | Encoding special value in anchored-data element | |
EP0461240A4 (en) | Methods and apparatus for efficient resource allocation for error and exception handling in convergent division | |
US10963245B2 (en) | Anchored data element conversion | |
US10936285B2 (en) | Overflow or underflow handling for anchored-data value | |
JP3493085B2 (ja) | 演算器 | |
US5305247A (en) | Method and processor for high-speed convergence factor determination | |
JP3522387B2 (ja) | パイプライン演算装置 | |
JP3541086B2 (ja) | 除算結果、及び開平算結果に正確な丸め処理を行う方法、及び装置 | |
JPH0317738A (ja) | 演算処理装置 | |
WO1991010188A1 (fr) | Procede et processeur pour determiner tres rapidement le facteur de convergence | |
JPS63285603A (ja) | プログラマブルコントロ−ラ | |
JPH04117520A (ja) | 浮動小数点演算装置 | |
JPS63279321A (ja) | マイクロプログラム制御装置 | |
JPS63254524A (ja) | デイジタル信号処理装置 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
17P | Request for examination filed |
Effective date: 19910829 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): DE FR GB |
|
A4 | Supplementary search report drawn up and despatched |
Effective date: 19930624 |
|
AK | Designated contracting states |
Kind code of ref document: A4 Designated state(s): DE FR GB |
|
17Q | First examination report despatched |
Effective date: 19960112 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN |
|
18D | Application deemed to be withdrawn |
Effective date: 19960723 |