CN104378120B - A kind of Hsiao coding checkout matrix generating methods detected for continuous N BU - Google Patents
A kind of Hsiao coding checkout matrix generating methods detected for continuous N BU Download PDFInfo
- Publication number
- CN104378120B CN104378120B CN201410649686.7A CN201410649686A CN104378120B CN 104378120 B CN104378120 B CN 104378120B CN 201410649686 A CN201410649686 A CN 201410649686A CN 104378120 B CN104378120 B CN 104378120B
- Authority
- CN
- China
- Prior art keywords
- column vector
- check matrix
- vector
- search space
- continuous
- 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.)
- Active
Links
Landscapes
- Detection And Correction Of Errors (AREA)
Abstract
The invention discloses a kind of Hsiao coding checkout matrix generating methods detected for continuous N BU, check code bit wide L of this method first according to needed for being determined Hamming inequality, then list based on L bit checks code, meet the column vector formation search space of Hsiao coding criterions, and according in column vector 1 number to search space priority arrangement；Secondly initiation verification matrix, again according to the continuous overturning digit i for being actually needed detection, is utilized respectively search and discriminant function is tried one's best from acquirement legal column vector in high priority search space adds to check matrix, ultimately produce satisfactory check matrix.Error correction and detection circuit based on this check matrix, on the premise of the delay of traditional SEC DED checking circuits is not increased, can detect continuous long numeric data upset, reply aerospace processor Cache MBU phenomenons, the application of suitable Cache reliabilities Scheme of Strengthening.
Description
Technical field
The invention belongs to aerospace processor reliability field, it is related to a kind of Hsiao codings school detected for continuous N BU
Test matrix generating method.
Background technology
In-line memory area on 30% and 60%, therefore piece respectively more than depositing in contemporary processor and SoC chip
The reliability of reservoir is most important.Due to the influence of space single particle effect, the data in aerospace processor Cache easily occur
Upset, and with integrated circuit technology progress, error pattern is developed from traditional unit upset (SBU) to Multiple-bit upsets (MBU).
Space and the experiment of ground single-particle find that heavy particle can cause the upset of 10~15 Coutinuous store positions in 90nm SRAM (continuous
Multiple-bit upsets are MBU key characters).Based on this, memory bank relies primarily on high-order error correcting code (such as RS, BCH code in current slice
Deng) realize it is system-level fault-tolerant, high-order error correcting code possess correct and detection Multiple-bit upsets ability, but its checking procedure is complicated, school
Encoding and decoding during testing generally require several clock cycle, therefore the read-write that can greatly prolong memory using high-order error correcting code is visited
Ask the cycle.The Cache read and write access cycle can have a strong impact on processor performance, and the high latency characteristic of high-order error correcting code makes it can not
Apply the reinforcement protection scheme in Cache.
In order to reduce Cache read-write cycle to the full extent, generally using delays such as SEC-DED codings or even-odd checks
Less Scheme of Strengthening is protected to Cache data, when processor finds that Cache data have correctable error, is then entered
Row is corrected, and when finding not correctable error, then forces Cache to be not hit by, Cache corresponding data pair is read from main memory
This.But SEC-DED or even-odd check can not successfully manage the MBU phenomenons in Cache, i.e., it can not recover or detect that space is single
The continuous Multiple-bit upsets of data in Cache memory cell caused by particle, once Cache data can not 100% recovery or inspection
Data flipping is surveyed, processor is likely to result in " run and fly " etc. unpredictable by using the instruction or data for occurring continuous Multiple-bit upsets
Situation.
In order to tackle aerospace processor Cache MBU problems, it is necessary to construct a kind of low latency, can correct or detect
The error correcting code of continuous Multiple-bit upsets, the EDAC circuits based on this error correcting code are not on the premise of Cache access delays are increased, it is ensured that
The reliability of Cache data, maintains the reliability service of aerospace processor.
The content of the invention
It is an object of the invention to overcome the shortcoming of above-mentioned prior art to be detected there is provided a kind of for continuous N BU
Any column vector in Hsiao coding checkout matrix generating methods, the Hsiao coding checkout matrixes of this method generation is met
Hsiao coding criterions are to facilitate delay of optimizing hardware.
To reach above-mentioned purpose, the Hsiao coding checkout matrix generating methods of the present invention detected for continuous N BU
Comprise the following steps：
1) according to Hamming inequality 2r>=k+r+1 determines the check code bit wide r needed for k information bits；
2) all L dimensional vectors for meeting Hsiao coding criterions are listed, the L dimensional vectors sum is equal to
2n+1 is the number of binary one in L dimensional vectors, and search space A is belonged to according in L dimensional vectors 1 number2n+1,
The search space A2n+1IncludingIndividual L dimensional vectors, n be integer and
3) the binary system check matrix C of initialization L × M dimensions, wherein, M=k+L, binary system check matrix C preceding k row
Vector is null vector, and rear L column vector is made up of L rank unit matrixs, then binary system check matrix C column vector form is C
=[d1 d2 … dk c1 … cL]；
4) the continuous overturning digit i of detection needed for judging, as continuous overturning digit i≤L+1 of required detection, it is determined that
Search function Func1 and search function Func2, wherein, search function Func1 ensures from search space A2n+1Middle taking-up it is new to
Amount b is not repeated with already present any column vector in binary system check matrix C, and search function Func2 ensures from search space
A2n+1L column vector linear independence adjacent in binary system check matrix C with it the new vectorial b of middle taking-up, wherein, l ∈ [3,
I], from search space A2n+1Middle taking-up meets search function Func1 and search function Func2 column vector is filled into binary system school
The correspondence position of Matrix C is tested to substitute the null vector in initialization, Hsiao coding checkout matrixes are obtained；
5) as the continuous overturning digit i ＞ L+1 of required detection, then generation can detect the verification of L+1 continuous overturnings
Matrix C1, decision function Func3 is determined, by decision function Func3 to check matrix C1Detected, judge check matrix C1
X column vector of middle arbitrary continuation whether linear correlation, wherein, x ∈ [L+2, i], when the linear phase of x column vector of arbitrary continuation
Guan Shi, then to search space A2n+1In column vector random alignment, the position of random inversion column vector；
6) repeat step 5) until check matrix C1Untill x column vector linear independence for meeting arbitrary continuation, then will
Check matrix C1Check matrix C is assigned to, Hsiao coding checkout matrixes are obtained.
Described search function Func1 meets all column vectors, any one column vector in traversal binary system check matrix C
With search space A2n+1The new vectorial b of middle taking-up is differed.
Described search function Func2 is met from A2n+1The new vectorial b of middle taking-up can be filled into binary system check matrix C
In any one row in interior 1~kth row；Meanwhile, search function Func2 column vector index p successively decreases i from M- (L-i+1), right
In M- (L-i+1) to any one column vector index p in i, the i-1 column vector including column vector p is taken, passes through the i-
1 column vector is from search space A2n+1Legal new vectorial b is searched, and ensures any in the i-1 column vector and new vector b
L adjacent column vector linear independence.
The searchable index r that the decision function Func3 meets decision function Func3 is incremented to M-i+1 from 1, from any one
I column vector of continuous drawing certainly is taken in the function Func3 corresponding column vectors of searchable index r, decision function Func3 disclosure satisfy that
X column vector linear independence of arbitrary neighborhood in this i vector.
Step 4) described in from search space A2n+1Middle taking-up meets search function Func1 and search function Func2 row
Vector be filled into binary system check matrix C correspondence position using substitute initialization in null vector concrete operations as：Search for letter
Number Func2 is first from the A of high priority3Scan for, work as A3It can not meet, then search space be extended into A2n+1For A3+A5, work as A3
+A5It can not still meet, then once gradually expand search space A2n+1, until binary system check matrix C is changed into without null vector
L × M dimension matrix untill.
The invention has the advantages that：
The Hsiao coding checkouts matrix generating method of the present invention detected for continuous N BU is first according to Hamming inequality
It is determined that required check code bit wide, then lists based on L bit checks code and meets the column vector formation search of Hsiao coding criterions
Space A2n+1, and according in column vector 1 number to search space A2n+1Priority arrangement；Next initiation verification matrix, according to
The continuous overturning digit i for being actually needed detection is utilized respectively search function and discriminant function and taken as far as possible from high priority search space
Obtain legal column vector to add in check matrix, obtain Hsiao coding checkout matrixes, the Hsiao coding checkouts that the present invention is generated
Any column vector is satisfied by Hsiao coding criterions to facilitate hardware optimization in matrix, based on the Hsiao coding checkouts matrix
The EDAC circuits towards aerospace processor Cache can be designed, data are overturn into Cache when EDAC electric circuit inspections
When, then force Cache to be not hit by, force Cache controllers to be transmitted to request data is hosted to processor, and returned using hosting
Data update Cache and realize the correction of Cache wrong data, the EDAC circuits do not increase traditional SEC-DED checking circuits
Delay, the application of suitable Cache reliabilities Scheme of Strengthening.
Brief description of the drawings
Fig. 1 is check matrix generating method flow chart of the invention.
Embodiment
The present invention is described in further detail with reference to specific embodiment and accompanying drawing,
It is of the present invention to comprise the following steps for the continuous N BU Hsiao coding checkout matrix generating methods detected：
1) according to Hamming inequality 2r>=k+r+1 determines the check code bit wide r needed for k information bits；
2) all L dimensional vectors for meeting Hsiao coding criterions are listed, the L dimensional vectors sum is equal to
2n+1 is the number of binary one in L dimensional vectors, and search space A is belonged to according in L dimensional vectors 1 number2n+1,
The search space A2n+1IncludingIndividual L dimensional vectors, n be integer and
3) the binary system check matrix C of initialization L × M dimensions, wherein, M=k+L, binary system check matrix C preceding k row
Vector is null vector, and rear L column vector is made up of L rank unit matrixs, then binary system check matrix C column vector form is C
=[d1 d2 … dk c1 … cL]；
4) the continuous overturning digit i of detection needed for judging, as continuous overturning digit i≤L+1 of required detection, it is determined that
Search function Func1 and search function Func2, wherein, search function Func1 ensures from search space A2n+1Middle taking-up it is new to
Amount b is not repeated with already present any column vector in binary system check matrix C, and search function Func2 ensures from search space
A2n+1L column vector linear independence adjacent in binary system check matrix C with it the new vectorial b of middle taking-up, wherein, l ∈ [3,
I], from search space A2n+1Middle taking-up meets search function Func1 and search function Func2 column vector is filled into binary system school
The correspondence position of Matrix C is tested to substitute the null vector in initialization, Hsiao coding checkout matrixes are obtained；
5) as the continuous overturning digit i ＞ L+1 of required detection, then generation can detect the verification of L+1 continuous overturnings
Matrix C1, decision function Func3 is determined, by decision function Func3 to check matrix C1Detected, judge check matrix C1
X column vector of middle arbitrary continuation whether linear correlation, wherein, x ∈ [L+2, i], when the linear phase of x column vector of arbitrary continuation
Guan Shi, then to search space A2n+1In column vector random alignment, the position of random inversion column vector；
6) repeat step 5) until check matrix C1Untill x column vector linear independence for meeting arbitrary continuation, then will
Check matrix C1Check matrix C is assigned to, Hsiao coding checkout matrixes are obtained.
Described search function Func1 meets all column vectors, any one column vector in traversal binary system check matrix C
With search space A2n+1The new vectorial b of middle taking-up is differed.
Described search function Func2 is met from A2n+1The new vectorial b of middle taking-up can be filled into binary system check matrix C
In any one row in interior 1~kth row；Meanwhile, search function Func2 column vector index p successively decreases i from M- (L-i+1), right
In M- (L-i+1) to any one column vector index p in i, the i-1 column vector including column vector p is taken, passes through the i-
1 column vector is from search space A2n+1Legal new vectorial b is searched, and ensures any in the i-1 column vector and new vector b
L adjacent column vector linear independence.
The searchable index r that the decision function Func3 meets decision function Func3 is incremented to M-i+1 from 1, from any one
I column vector of continuous drawing certainly is taken in the function Func3 corresponding column vectors of searchable index r, decision function Func3 disclosure satisfy that
X column vector linear independence of arbitrary neighborhood in this i vector.
Step 4) described in from search space A2n+1Middle taking-up meets search function Func1 and search function Func2 row
Vector be filled into binary system check matrix C correspondence position using substitute initialization in null vector concrete operations as：Search for letter
Number Func2 is first from the A of high priority3Scan for, work as A3It can not meet, then search space be extended into A2n+1For A3+A5, work as A3
+A5It can not still meet, then once gradually expand search space A2n+1, until binary system check matrix C is changed into without null vector
L × M dimension matrix untill.
Embodiment one
32 information bit generations can be detected continuous by the check matrix generating method flow chart provided using the present invention respectively
8 and two check matrixes of continuous 25 bit flipping.
Comprise the following steps：
1) according to Hamming inequality 2r>=k+r+1 determines the check code bit wide 6 needed for 32 information bits, disclosed by the invention
Check code bit wide needed for check matrix is equal to 7.
Step2：According to Hsiao coding criterions meet its requirement 7 dimensional vectors (binary vector) have 56 (), wherein, high priority search space A3In the element that includes as shown in table 1, sub-priority search space A5In include
Element as shown in table 2, for convenience of expressing, the element in table 1 and table 2 is binary row vector, can to each element transposition
It is changed into column vector.
Table 1
| Sequence number | Vector | Sequence number | Vector |
| 1 | (1110000) | 19 | (0110001) |
| 2 | (1101000) | 20 | (0101100) |
| 3 | (1100100) | 21 | (0101010) |
| 4 | (1100010) | 22 | (0101001) |
| 5 | (1100001) | 23 | (0100110) |
| 6 | (1011000) | 24 | (0100101) |
| 7 | (1010100) | 25 | (0100011) |
| 8 | (1010010) | 26 | (0011100) |
| 9 | (1010001) | 27 | (0011010) |
| 10 | (1001100) | 28 | (0011001) |
| 11 | (1001010) | 29 | (0010110) |
| 12 | (1001001) | 30 | (0010101) |
| 13 | (1000110) | 31 | (0010011) |
| 14 | (1000101) | 32 | (0001110) |
| 15 | (1000011) | 33 | (0001101) |
| 16 | (0111000) | 34 | (0001011) |
| 17 | (0110100) | 35 | (0000111) |
| 18 | (0110010) |
Table 2
| Sequence number | Vector | Sequence number | Vector |
| 1 | (1111100) | 12 | (1011101) |
| 2 | (1111010) | 13 | (1011011) |
| 3 | (1111001) | 14 | (1010111) |
| 4 | (1110110) | 15 | (1001111) |
| 5 | (1110101) | 16 | (0111110) |
| 6 | (1110011) | 17 | (0111101) |
| 7 | (1101110) | 18 | (0111011) |
| 8 | (1101101) | 19 | (0110111) |
| 9 | (1101011) | 20 | (0101111) |
| 10 | (1100111) | 21 | (0011111) |
| 11 | (1011110) |
3) before binary system the check matrix C, wherein M=k+L, binary system check matrix C of initialization L × M dimensions k arrange to
Amount is null vector, and last L column vector is made up of L rank unit matrixs, and this example generates 7x39 binary system verification square
Battle array C, wherein rear the 7 of binary system check matrix C are classified as 7 rank unit matrixs, before 32 be classified as null vector, while being also letter to be searched
The region of number filling, initialization matrix is as follows.
4) continuous 8 can be detected to 32 information bit generations in the continuous overturning digit of detection, the present embodiment needed for judging
The check matrix of upset, because 8≤8 (7+1)；
5) search function of check matrix is determined.All column vectors in Func1 search functions traversal binary system check matrix C
(being 39 in this example), if any of which column vector and search space A2n+1The new vectorial b of middle taking-up is identical, then the function
It is unsatisfactory for, return value is 0.Func2 search functions are from search space A2n+1The new vectorial b of middle taking-up may be filled into binary system school
Any one row tested in the row of 1~kth in Matrix C, the column vector index p of search function Func2 traversals successively decreases 8 from 39, for
From 39 to 8 any one column vector index p, take including column vector p i-1 column vector (including p, p-1, p-2 ...,
P-i+1), this i-1 column vector removal search space A is utilized2n+1A legal new vectorial b is searched, and ensures i-1 column vectors and new
The l column vector linear independence (l ∈ [3, i]) of arbitrary neighborhood in i vector of vectorial b combinations, it is ensured that any phase in i vector
L adjacent column vector linear independence needs i-l+1 equation.
6) from search space A2n+1Middle taking-up meets Func1 search functions and the column vector of Func2 search functions is filled into i
In individual vectorCCorrespondence position, substitute initialization in null vector, the A from high priority first3Scan for, A3Search refer to
Number p is incremented by 1 to 35 from 1, and the check matrix for obtaining can detect continuous 8 Data flippings is：
Embodiment two
Step 4 in embodiment one) in when 32 information bits are generated with the check matrix that can detect continuous 25 bit flipping, then
1)~4) identical with embodiment one；
5) the check matrix C of 8 continuous overturnings of generation energy1, it is assumed that check matrix C1It is identical with the matrix in embodiment 1.
6) decision function Func3 is determined.Decision function Func3 searchable index r (the present embodiment correspondences from 1 to M-i+1
39-25+1=14), for from the corresponding column vectors of 1 to 14 any one searchable index r, taking r including column vector r to arrange
Vector, decision function Func3 ensures l column vector linear independence of arbitrary neighborhood in this i vector, it is ensured that appoint in i vector
L adjacent column vector linear independence of meaning needs i-l+1 equation.
7) assume using decision function Func3 to check matrix C1Judged, it is found that it can not detect continuous 16
Overturn (check matrix C1In deposit), then to search space A2n+1Middle A3Column vector random alignment, the position of random inversion column vector
Put；
8) and then re-execute 7), until check matrix C1X column vector linear independence of arbitrary continuation is met, will be verified
Matrix C1Check matrix C is assigned to, the check matrix of continuously 25 bit flippings is obtained.
The check matrix of continuous 25 bit flipping is as follows：
Patent of the present invention discloses a kind of Hsiao coding checkout matrix generating methods detected for continuous N BU, this method
The Hsiao coding checkouts matrix of generation does not increase the hardware consumption of traditional SECDED EDAC circuits, can detect in memory
Continuous N BU (Multiple-bit upsets) phenomenon.On the premise of Cache access delays are not increased, the MBU phenomenons for reply Cache are carried
Supplied a kind of effective low overhead encoding scheme, can directly substitute the EDAC circuits that are used in current Cache protection schemes or
Parity checker.
Claims (1)
1. a kind of Hsiao coding checkout matrix generating methods detected for continuous N BU, it is characterised in that comprise the following steps：
1) according to Hamming inequality 2r>=k+r+1 determines the check code bit wide r needed for k information bits；
2) all L dimensional vectors for meeting Hsiao coding criterions are listed, the L dimensional vectors sum is equal to2n+1
For the number of binary one in L dimensional vectors, search space A is belonged to according in L dimensional vectors 1 number2n+1, it is described
Search space A2n+1IncludingIndividual L dimensional vectors, n be integer and
3) the binary system check matrix C of initialization L × M dimensions, wherein, M=k+L, binary system check matrix C preceding k column vector
It is null vector, rear L column vector is made up of L rank unit matrixs, then binary system check matrix C column vector form is C=[d1
d2 … dk c1 … cL]；
4) the continuous overturning digit i of detection needed for judging, as continuous overturning digit i≤L+1 of required detection, it is determined that search
Function Func1 and search function Func2, wherein, search function Func1 ensures from search space A2n+1The new vectorial b of middle taking-up
Do not repeated with already present any column vector in binary system check matrix C, search function Func2 ensures from search space A2n+1In
The new vectorial b l column vector linear independences adjacent in binary system check matrix C with it taken out, wherein, l ∈ [3, i], from
Search space A2n+1The column vector that middle taking-up meets search function Func1 and search function Func2 is filled into binary system verification square
Battle array C correspondence position obtains Hsiao coding checkout matrixes to substitute the null vector in initialization；
5) as the continuous overturning digit i ＞ L+1 of required detection, then generation can detect the check matrix of L+1 continuous overturnings
C1, decision function Func3 is determined, by decision function Func3 to check matrix C1Detected, judge check matrix C1In appoint
Continuous x column vector of anticipating whether linear correlation, wherein, x ∈ [L+2, i], when x column vector linear correlation of arbitrary continuation
When, then to search space A2n+1In column vector random alignment, the position of random inversion column vector；
6) repeat step 5) until check matrix C1Untill x column vector linear independence for meeting arbitrary continuation, square then will be verified
Battle array C1Check matrix C is assigned to, Hsiao coding checkout matrixes are obtained；
Described search function Func2 is met from A2n+1The new vectorial b of middle taking-up can be filled into the 1st in binary system check matrix C
In any one row in~kth row；Meanwhile, search function Func2 column vector index p successively decreases i from M- (L-i+1), for M-
(L-i+1) to any one column vector index p in i, the i-1 column vector including column vector p is taken, passes through the i-1
Column vector is from search space A2n+1Legal new vectorial b is searched, and ensures any phase in the i-1 column vector and new vector b
L adjacent column vector linear independence；
Step 4) described in from search space A2n+1Middle taking-up meets search function Func1 and search function Func2 column vector is filled out
Be charged to binary system check matrix C correspondence position using substitute initialization in null vector concrete operations as：Search function Func2
First from the A of high priority3Scan for, work as A3It can not meet, then search space be extended into A2n+1For A3+A5, work as A3+A5Still
It can not meet, then gradually expand search space A successively2n+1, until binary system check matrix C is changed into L × M without null vector
Untill tieing up matrix；
The searchable index r that the decision function Func3 meets decision function Func3 is incremented to M-i+1 from 1, from any one judgement
I column vector of continuous drawing is taken in the function Func3 corresponding column vectors of searchable index r, decision function Func3 disclosure satisfy that this
X column vector linear independence of arbitrary neighborhood in i vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410649686.7A CN104378120B (en) | 2014-11-14 | 2014-11-14 | A kind of Hsiao coding checkout matrix generating methods detected for continuous N BU |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410649686.7A CN104378120B (en) | 2014-11-14 | 2014-11-14 | A kind of Hsiao coding checkout matrix generating methods detected for continuous N BU |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104378120A CN104378120A (en) | 2015-02-25 |
| CN104378120B true CN104378120B (en) | 2017-07-25 |
Family
ID=52556817
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410649686.7A Active CN104378120B (en) | 2014-11-14 | 2014-11-14 | A kind of Hsiao coding checkout matrix generating methods detected for continuous N BU |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104378120B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10476525B2 (en) * | 2017-01-09 | 2019-11-12 | Qualcomm Incorporated | Low latency bit-reversed polar codes |
| CN112398485B (en) * | 2020-11-20 | 2023-03-21 | 中国人民解放军国防科技大学 | Check matrix generation method for correcting one-error-free code and two-error-correction adjacent code |
| CN114244374B (en) * | 2021-11-22 | 2022-08-05 | 成都博尔微晶科技有限公司 | Check matrix generation method and device, electronic equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8516339B1 (en) * | 2011-04-01 | 2013-08-20 | Xilinx, Inc. | Method of and circuit for correcting adjacent bit errors in a memory |
| CN103886915A (en) * | 2012-12-19 | 2014-06-25 | 英飞凌科技股份有限公司 | Circuitry and method for correcting 3-bit errors containing adjacent 2-bit error |
| US8875002B1 (en) * | 2012-01-20 | 2014-10-28 | Cypress Semiconductor Corporation | Low cost adjacent double error correcting code |
-
2014
- 2014-11-14 CN CN201410649686.7A patent/CN104378120B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8516339B1 (en) * | 2011-04-01 | 2013-08-20 | Xilinx, Inc. | Method of and circuit for correcting adjacent bit errors in a memory |
| US8875002B1 (en) * | 2012-01-20 | 2014-10-28 | Cypress Semiconductor Corporation | Low cost adjacent double error correcting code |
| CN103886915A (en) * | 2012-12-19 | 2014-06-25 | 英飞凌科技股份有限公司 | Circuitry and method for correcting 3-bit errors containing adjacent 2-bit error |
Non-Patent Citations (1)
| Title |
|---|
| Multiple Bit Upset Tolerant Memory Using a Selective Cycle Avoidance Based SEC-DED-DAEC Code;Avijit Dutta 等;《25th IEEE VLSI Test Symposium 2007》;20070510;摘要，1-5页第2-5节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104378120A (en) | 2015-02-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9213595B2 (en) | Handling errors in ternary content addressable memories | |
| Alameldeen et al. | Energy-efficient cache design using variable-strength error-correcting codes | |
| Du et al. | Soft error resilient QR factorization for hybrid system with GPGPU | |
| CN104378120B (en) | A kind of Hsiao coding checkout matrix generating methods detected for continuous N BU | |
| CN102122959B (en) | Data compression device for improving main memory reliability of computer, and method thereof | |
| CN104409103A (en) | Novel two-dimensional coding reinforcing method and circuit arrangement for aerospace memory | |
| EP2975522A2 (en) | Digital error correction | |
| Gracia-Moran et al. | Correction of adjacent errors with low redundant matrix error correction codes | |
| Reviriego et al. | A method to extend orthogonal Latin square codes | |
| US8661320B2 (en) | Independent orthogonal error correction and detection | |
| Neelima et al. | Efficient adjacent 3D parity error detection and correction codes for embedded memories | |
| Moldaschl et al. | Fault tolerant communication-optimal 2.5 D matrix multiplication | |
| CN108021467B (en) | Memory fault tolerance protection method, device, equipment and storage medium | |
| CN108897637A (en) | The error correction method, device of reading data, equipment and medium in memory | |
| Anker et al. | Flag gadgets based on classical codes | |
| CN110751977B (en) | Memory chip fault-tolerant device and fault-tolerant error-correcting method based on LDPC code | |
| CN109032833B (en) | Correction method, device, equipment and storage medium for multi-bit error data | |
| Liu et al. | A method to recover critical bits under a double error in SEC-DED protected memories | |
| CN108984340B (en) | Fault-tolerant protection method, device, equipment and storage medium for storage data | |
| US9053050B2 (en) | Determining a desirable number of segments for a multi-segment single error correcting coding scheme | |
| CN108733628A (en) | A kind of reinforcement means of parallel matrix multiplication algorithm | |
| Evans et al. | Optimal correction of concatenated fault-tolerant quantum codes | |
| Liu et al. | Fault secure encoder and decoder designs for matrix codes | |
| Austin et al. | Resilient matrix multiplication of hierarchical semi-separable matrices | |
| Kustov et al. | Efficiency Estimation of Single Error Correction, Double Error Detection and Double-Adjacent-Error Correction Codes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |