CN105790808B - A kind of reconfigurable arrays framework and its detection method towards MIMO detections - Google Patents
A kind of reconfigurable arrays framework and its detection method towards MIMO detections Download PDFInfo
- Publication number
- CN105790808B CN105790808B CN201610099729.8A CN201610099729A CN105790808B CN 105790808 B CN105790808 B CN 105790808B CN 201610099729 A CN201610099729 A CN 201610099729A CN 105790808 B CN105790808 B CN 105790808B
- Authority
- CN
- China
- Prior art keywords
- reconfigurable arrays
- ped
- general
- node
- matrix
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/0413—MIMO systems
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/0202—Channel estimation
- H04L25/024—Channel estimation channel estimation algorithms
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Power Engineering (AREA)
- Radio Transmission System (AREA)
Abstract
The present invention provides a kind of reconfigurable arrays framework detected towards MIMO, is applied in reconfigurable processor system.Reconfigurable arrays framework includes general reconfigurable arrays, special reconfigurable arrays and shared memory bank.Compared to traditional reconfigurable arrays computing architecture, the complete K best algorithms under a variety of matrix operations and a variety of K values may be implemented in the framework of this paper, by combining the reconfigurable arrays of isomery, by flexibly sharing storage mode, complete efficient MIMO detections and calculate.
Description
Technical field
The present invention relates to the reconstruction structure design fields of communications baseband.
Background technology
MIMO (Multiple-Input Multiple-Output Multiple Input Multiple Output) technology refers in transmitting terminal and reception
End uses multiple transmitting antennas and reception antenna respectively, and signal is made to pass through the mutiple antennas transmission of transmitting terminal and receiving terminal and connect
It receives, so as to improve communication quality.It can make full use of space resources, and multiple-input multiple-output is realized by mutiple antennas, not increase frequency
In the case of spectrum resource and antenna transmission power, system channel capacity can be increased exponentially, show apparent advantage, by regarding
For the core technology of next generation mobile communication.
In current reconstruction structure, there are many support a variety of inspections such as MMSE detection algorithms, SIC-MMSE detection algorithms
The reconstruction structure of method of determining and calculating, but few reconstruction structures for supporting K-best detection algorithms.K-best detection algorithms have
There is excellent performance, preferable performance, and moderate complexity have been remained able in the case where channel condition is poor, has used model
It encloses extensively, there is very high researching value.K-best detection algorithms presence itself largely computes repeatedly and parallel computation, restructural
Framework can fully excavate the degree of parallelism of K-best detection algorithms, effectively improve algorithm performance.
Invention content
The present invention provides a kind of reconfigurable arrays architecture designs towards MIMO detections, for Parallel Implementation K-best inspections
Node operation present in method of determining and calculating and corresponding PED are calculated, and can effectively accelerate PED sorting operations, substantially reduce inspection
The method of determining and calculating period improves detection algorithm throughput.The detection method that the reconfigurable arrays framework is realized, passes through the calculating battle array of isomery
Row, complete a variety of matrixes and calculate, and fully meet the signal-data processing demand in MIMO detections.
The technical scheme is that：A kind of reconfigurable arrays framework towards MIMO detections, including first to fourth is logical
With reconfigurable arrays, 1 special reconfigurable arrays, shared storage #1, shared storage #2 and shared storage #3；It is described it is general can
Restructuring array scale is 4 × 4 and includes one and be locally stored, for realizing the section in Matrix QR Decomposition, K-best detection algorithms
Point calculates and the corresponding PED of node is calculated；The special reconfigurable arrays scale is 8 × 4, for realizing K-best algorithms
In PED sequences, the sequence of k PED data only needs [k/2]+1 period；Accelerate data sorting, when reduction sequence is occupied
The clock period.The shared storage #1 is shared by the first to the second general reconfigurable arrays and special reconfigurable arrays, for real
Existing data interaction between the first to the second general reconfigurable arrays and special reconfigurable arrays；The shared storage #2 is by third
It is shared to the 4th general reconfigurable arrays and special reconfigurable arrays, for realizing third to the 4th general reconfigurable arrays and spy
Data interaction between different reconfigurable arrays；The shared storage #3 is shared by 4 general reconfigurable arrays, for realizing general
Data interaction between reconfigurable arrays.
Further, the special reconfigurable arrays are made of the PE and routing infrastructure of the row customization of 4 rows 8, can be realized
A variety of quicksort methods；The basic unit of the hardware configuration of each quicksort method is register, selector and compares
Device；The PE of the customization includes a register, No. two selectors and a comparator, and each PE can be configured to one individually
Register, selector or comparator；And by routing infrastructure, each PE can carry out data friendship with adjacent eight PE around
Mutually.The PE of customization and flexible and efficient routing infrastructure enable special reconfigurable arrays to realize a variety of quick sorting algorithms, subtract
The clock cycle that few sequence occupies, improve K-best detection algorithm performances.
Further, a variety of quicksort methods are common sequence or odd even ordering.
Further, described first to fourth general reconfigurable arrays can realize a variety of computing functions；It is each restructural
Array scale is 4 × 4, supports that the Matrix QR Decomposition of a variety of scales, node calculate and the corresponding PED of node is calculated；Each PE
Input there are two types of type：One is row interconnection line transmission data, it is another then come from above three PE and control two PE
Output；There are five directions for the output of each PE, i.e.,：Three PE of left and right two PE and lower section can be transferred to；It is each general
Reconfigurable arrays are locally stored comprising one, for storing local data；Each PE include a register, a multiplier,
The functional shift of PE can be multiplier, divider, ALU logical operations or register by one divider and ALU units.
The present invention also provides a kind of inspections realized using the coarse-grained reconfigurable array framework towards MIMO detections
Survey method, is as follows：
(1) the general reconfigurable arrays are decomposed for the QR of matrix, and the R matrix elements after decomposition are stored in shared deposit
In storing up #1, shared storage #2 or being locally stored；
(2) general reconfigurable arrays read the element of upper triangular matrix R, carry out node operation and PED is calculated, PED is calculated
As a result it is stored in shared storage #1 or shared storages #2；
(3) special reconfigurable arrays read PED result of calculations and are ranked up, ranking results be stored in shared deposit storage #1 or
In shared storage #2, general reconfigurable arrays reading ranking results carry out new one layer of node calculating and corresponding PED is calculated；
(4) circulation step (2) and step (3) terminate until calculating.
Further, there are a large amount of parallel node operations in K-best detection algorithms to calculate with PED, and node calculates packet
Containing multiplication, subtraction and divide operations, PED is calculated to be operated comprising multiplication and subtraction；General reconfigurable arrays can be realized multiple
Node operation and corresponding PED are calculated, and effectively improve the performance of detection algorithm.It is specific as follows：Same day line gauge mould no more than 4 ×
When 4, i.e., access matrix is small in 4 × 4, only needs the QR that a general reconfigurable arrays can complete matrix to decompose, institute
The element of upper triangular matrix R be stored in and be locally stored, calculate for subsequent node operation and PED；Same day line gauge mould
When not less than 8 × 8, i.e., access matrix scale is not less than 8 × 8, and all general reconfigurable arrays both participate in QR operation splittings, QR
The intermediate data of decomposable process is stored in shared storage #3, so that four general reconfigurable arrays are read, after the completion of QR is decomposed,
Obtained upper triangular matrix element is stored in respectively in being locally stored in 4 general reconfigurable arrays, is used for subsequent node
Operation and corresponding PED are calculated；When same day line gauge mould is between 4 × 4 and 8 × 8, general weighed using 2 or 3 as needed
The QR that structure array completes matrix is decomposed, and the intermediate data of QR decomposable processes is stored in shared storage #3, so as to general restructural
Array is read, and after the completion of QR is decomposed, obtained upper triangular matrix element is stored in the local in the reconfigurable arrays for participating in calculating
In storage, calculated for subsequent node operation and corresponding PED.
Further, storage #1, shared storage #2, shared storage #3 and being locally stored is shared mutually to be highly coupled.Tightly
Close cooperation is to improve data interaction efficiency.After the completion of the PED that general reconfigurable arrays are realized is calculated, result of calculation is stored to close
The shared storage #1 of array or shared storage #2, special reconfigurable arrays are read out and carry out PED sequences.It is locally stored and is used for
The data needed for a general reconfigurable arrays are stored, when same day line gauge mould is not more than 4 × 4, only need a general restructural battle array
Row, during the result that QR is decomposed is placed directly in and is locally stored, so that subsequent node calculates and PED is calculated, without sending other to
Array.Same day line gauge mould be more than 8 × 8 when, need 4 general reconfigurable arrays to work at the same time and could complete, at this time 4 it is general can
Data interaction between restructuring array is completed by shared storage #3, is effectively improved data transmission efficiency.
Further, the coarse-grained reconfigurable array framework towards MIMO detections can realize a variety of antenna scales
The K-best detection algorithms of a variety of K values；It is specific as follows：K-best algorithm flows are that access matrix QR is decomposed, node calculates, PED
It calculates and PED sorts；Wherein, node calculates, PED is calculated and sequencing cycle executes K times；Access matrix QR decomposition, node
It calculates and PED calculating is completed by general reconfigurable arrays, PED sequences are completed by special reconfigurable arrays, the data between each array
Storage and interaction completed by 3 shared storages；
Different antenna scale meanings need the scale for carrying out the access matrix of QR decomposition different, utilize claim 6 institute
The detection method stated can complete the QR operation splittings of the access matrix of different scales；
The K values represent the parallel computation of K node and the corresponding PED of K node is calculated and K PED data
Sequence；General reconfigurable arrays, which read pre-designed configuration information, which can complete a variety of K, is worth node to calculate and PED meters
It calculates, special reconfigurable arrays, which read pre-designed configuration information, can complete the PED sorting operations of a variety of K values；Each is special
The K-best algorithms of set pattern mould can all have a set of corresponding configuration information.
Advantageous effect：The present invention provides a kind of reconfigurable arrays architecture designs towards MIMO detections, for parallel real
Node operation present in existing K-best detection algorithms and corresponding PED are calculated, and can effectively accelerate PED sorting operations,
The detection algorithm period is substantially reduced, detection algorithm throughput is improved.The detection method that the reconfigurable arrays framework is realized, by different
The computing array of structure, completes a variety of matrixes and calculates, and fully meets the signal-data processing demand in MIMO detections.
Description of the drawings
Fig. 1 is a kind of reconfigurable arrays architecture system result block diagram towards MIMO detections of the present invention.
Fig. 2 is special reconfigurable array structure figure in a kind of reconfigurable arrays framework towards MIMO detections of the present invention.
Fig. 3 is general reconfigurable array structure figure in a kind of reconfigurable arrays framework towards MIMO detections of the present invention.
Specific implementation mode
The present invention is further described below in conjunction with the accompanying drawings.
Fig. 1 is a kind of reconfigurable arrays architecture system result block diagram towards MIMO detections of the present invention.The restructural battle array
Column structure includes 4 general reconfigurable arrays, 1 special reconfigurable arrays and 3 shared storages.Described is general restructural
Array scale is 4 × 4 and includes one and be locally stored, for realizing the node meter in Matrix QR Decomposition, K-best detection algorithms
Calculation and the corresponding PED calculating of node etc..The special reconfigurable arrays scale is 8 × 4, for realizing in K-best algorithms
The operations such as PED sequences, accelerate data sorting, reduce the occupied clock cycle of sorting.The shared storage #1 and shared
Storage #2 shared by two general reconfigurable arrays and special reconfigurable arrays, for realizing general reconfigurable arrays with it is special can
Data interaction between restructuring array.The shared storage #3 is shared by 4 general reconfigurable arrays, is weighed for realizing general
Data interaction between structure array.
General reconfigurable arrays scale is 4 × 4, that is, includes 16 PE, see Fig. 3.Each PE include a register, one
Multiplier, a divider and ALU units.By configuring corresponding information, PE can be configured to multiplier, divider, ALU and patrol
Collect operation or register.In addition, the input of each PE is either the data that row interconnection line transmits, can also be three PE in top
With the output of two PE in left and right, i.e., the output of each PE can be transferred to three PE of left and right two PE and lower section.Same day line gauge
When mould is not more than 4 × 4, i.e., access matrix is small in 4 × 4, only needs a general reconfigurable arrays that matrix can be completed
QR is decomposed, and the element of the upper triangular matrix R of gained, which is stored in, to be locally stored, and is calculated for subsequent node operation and PED；When
When antenna scale is not less than 8 × 8, i.e., access matrix scale is not less than 8 × 8, and all general reconfigurable arrays are required to participate in
QR operation splittings, the intermediate data in this QR decomposable process is stored in shared storage #3, so as to four general restructural battle arrays
Row are read.When same day line gauge mould is between 4 × 4 and 8 × 8, square is completed using 2 or 3 general reconfigurable arrays as needed
The QR of battle array is decomposed, and the intermediate data of QR decomposable processes is stored in shared storage #3, so that general reconfigurable arrays are read, QR points
After the completion of solution, obtained upper triangular matrix element is stored in being locally stored in the reconfigurable arrays for participating in calculating, after being used for
Continuous node operation and corresponding PED are calculated.After the completion of QR is decomposed, obtained upper triangular matrix element be stored in respectively 4 it is logical
In being locally stored in reconfigurable arrays, the corresponding PED of node operation and node after being used for is calculated.General reconfigurable arrays
The upper triangular matrix element for being locally stored or sharing in storage is read, node calculating and corresponding PED calculating are carried out.It is general can
The corresponding PED of K parallel node operations and K is had in restructuring array every time to calculate, K PED result of calculations are stored in
In shared storage #1 or shared storages #2.Special reconfigurable arrays are made of the PE and routing infrastructure of the row customization of 4 rows 8, customization
PE includes a register, No. two selectors and a comparator.By configuration information, each PE can be configured to one individually
Register, selector or comparator.In addition, by flexible and efficient routing infrastructure, each PE can be with around adjacent eight
A PE carries out data interaction, sees Fig. 2, and special reconfigurable arrays are read simultaneously is stored in K in shared storage #1 or shared storages #2
The value of a PED, is ranked up, and the result of sequence is stored in shared storage #1 or shared storages #2, is transported as next node layer
The input of calculation.General reconfigurable arrays read the node that the ranking results in shared storage #1 or shared storages #2 start the second layer
Operation and corresponding PED are calculated, repeatedly until obtaining K layers of PED ranking results.
The above is only a preferred embodiment of the present invention, it should be pointed out that：For the ordinary skill people of the art
For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also answered
It is considered as protection scope of the present invention.
Claims (7)
1. a kind of reconfigurable arrays framework towards MIMO detections, it is characterised in that：Including first to fourth general restructural battle array
Row, 1 special reconfigurable arrays, shared storage #1, shared storage #2 and shared storage #3；The general reconfigurable arrays rule
Mould is 4 × 4 and includes one and be locally stored, calculated for realizing the node in Matrix QR Decomposition, K-best detection algorithms and
The corresponding PED of node is calculated；The special reconfigurable arrays scale is 8 × 4, for realizing the PED rows in K-best algorithms
The sequence of sequence, k PED data only needs [k/2]+1 period；The shared storage #1 is by the first to the second general restructural battle array
Row and special reconfigurable arrays are shared, for realizing the number between the first to the second general reconfigurable arrays and special reconfigurable arrays
According to interaction；The shared storage #2 is shared by third to the 4th general reconfigurable arrays and special reconfigurable arrays, for real
Show third to the data interaction between the 4th general reconfigurable arrays and special reconfigurable arrays；The shared storage #3 is by 4
General reconfigurable arrays are shared, for realizing the data interaction between general reconfigurable arrays；
The special reconfigurable arrays are made of the PE and routing infrastructure of the row customization of 4 rows 8, can realize a variety of quicksort sides
Method；The basic unit of the hardware configuration of each quicksort method is register, selector and comparator；The PE of the customization
Including a register, No. two selectors and a comparator, each PE can be configured to individual register, a selector
Or comparator；And by routing infrastructure, each PE can carry out data interaction with adjacent eight PE around.
2. the reconfigurable arrays framework according to claim 1 towards MIMO detections, it is characterised in that：It is described a variety of quick
Sort method is common sequence or odd even ordering.
3. the reconfigurable arrays framework according to claim 1 towards MIMO detections, it is characterised in that：Described first to
Four general reconfigurable arrays can realize a variety of computing functions；Each reconfigurable arrays scale is 4 × 4, supports a variety of scales
Matrix QR Decomposition, node calculate and the corresponding PED of node is calculated；There are two types of types for the input of each PE：One is rows to interconnect
The data of line transmission, output that is another then coming from above two PE of three PE and left and right；Each direction there are five the outputs of PE,
I.e.：Three PE of left and right two PE and lower section can be transferred to；Each general reconfigurable arrays are locally stored comprising one, are used for
Store local data；Each PE includes a register, a multiplier, a divider and ALU units, can be by the work(of PE
Multiplier, divider, ALU logical operations or register can be changed into.
4. a kind of detection side realized using any reconfigurable arrays frameworks towards MIMO detections of claim 1-3
Method, it is characterised in that：It is as follows：
(1) the general reconfigurable arrays are decomposed for the QR of matrix, the R matrix elements after decomposition be stored in shared storage #1,
In sharing storage #2 or being locally stored；
(2) general reconfigurable arrays read the element of upper triangular matrix R, carry out node operation and PED is calculated, PED result of calculations
It is stored in shared storage #1 or shared storages #2；
(3) special reconfigurable arrays read PED result of calculations and are ranked up, and ranking results are stored in shared deposit storage #1 or shared
It stores in #2, general reconfigurable arrays read ranking results and carry out new one layer of node calculating and corresponding PED calculating；
(4) circulation step (2) and step (3) terminate until calculating.
5. detection method according to claim 4, it is characterised in that：There are a large amount of parallel sections in K-best detection algorithms
Point processing is calculated with PED, and it includes multiplication, subtraction and divide operations that node, which calculates, and PED is calculated to be grasped comprising multiplication and subtraction
Make；General reconfigurable arrays can realize that multiple node operations and corresponding PED calculate, specific as follows：Same day line gauge mould is not
When more than 4 × 4, i.e., access matrix is small in 4 × 4, only needs a general reconfigurable arrays that can complete the QR of matrix
It decomposes, the element of the upper triangular matrix R of gained, which is stored in, to be locally stored, and is calculated for subsequent node operation and PED；The same day
When line gauge mould is not less than 8 × 8, i.e., access matrix scale is not less than 8 × 8, and all general reconfigurable arrays both participate in QR decomposition
The intermediate data of operation, QR decomposable processes is stored in shared storage #3, so that four general reconfigurable arrays are read, QR is decomposed
After the completion, the upper triangular matrix element obtained is stored in respectively in being locally stored in 4 general reconfigurable arrays, for follow-up
Node operation and corresponding PED calculate；It is logical using 2 or 3 as needed when same day line gauge mould is between 4 × 4 and 8 × 8
The QR that matrix is completed with reconfigurable arrays is decomposed, and the intermediate data of QR decomposable processes is stored in shared storage #3, so as to general
Reconfigurable arrays are read, and after the completion of QR is decomposed, obtained upper triangular matrix element is stored in the reconfigurable arrays for participating in calculating
Be locally stored, calculated for subsequent node operation and corresponding PED.
6. detection method according to claim 4, it is characterised in that：Shared storage #1, shared storage #2, shared storage #3
And it is locally stored and intercouples.
7. detection method according to claim 5, it is characterised in that：The restructural battle array of coarseness towards MIMO detections
Column structure can realize the K-best detection algorithms of a variety of K values of a variety of antenna scales；It is specific as follows：K-best algorithm flows are
Access matrix QR is decomposed, node calculates, PED is calculated and PED sequences；Wherein, node calculates, PED is calculated and sequencing cycle
It executes K times；Access matrix QR is decomposed, node calculates and PED calculating is completed by general reconfigurable arrays, PED sequences by it is special can
Restructuring array is completed, and the storage and interaction of the data between each array are completed by 3 shared storages；
Different antenna scale meanings need the scale for carrying out the access matrix of QR decomposition different, using described in claim 6
Detection method can complete the QR operation splittings of the access matrix of different scales；
The K values represent the parallel computation of K node and the corresponding PED of K node is calculated and the row of K PED data
Sequence；General reconfigurable arrays, which read pre-designed configuration information, which can complete a variety of K, is worth node calculating and PED to calculate,
Special reconfigurable arrays, which read pre-designed configuration information, can complete the PED sorting operations of a variety of K values；Each specific rule
The K-best algorithms of mould can all have a set of corresponding configuration information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610099729.8A CN105790808B (en) | 2016-02-23 | 2016-02-23 | A kind of reconfigurable arrays framework and its detection method towards MIMO detections |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610099729.8A CN105790808B (en) | 2016-02-23 | 2016-02-23 | A kind of reconfigurable arrays framework and its detection method towards MIMO detections |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105790808A CN105790808A (en) | 2016-07-20 |
| CN105790808B true CN105790808B (en) | 2018-08-28 |
Family
ID=56402897
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610099729.8A Active CN105790808B (en) | 2016-02-23 | 2016-02-23 | A kind of reconfigurable arrays framework and its detection method towards MIMO detections |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105790808B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112486905A (en) * | 2020-12-18 | 2021-03-12 | 清华大学 | Reconfigurable isomerization PEA interconnection method |
| CN113055060B (en) * | 2021-03-08 | 2022-04-05 | 上海交通大学 | Coarse-grained reconfigurable architecture system for large-scale MIMO signal detection |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103984560A (en) * | 2014-05-30 | 2014-08-13 | 东南大学 | Embedded reconfigurable system based on large-scale coarseness and processing method thereof |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8553796B2 (en) * | 2009-12-23 | 2013-10-08 | Intel Corporation | Distortion-aware multiple input multiple output precoding |
-
2016
- 2016-02-23 CN CN201610099729.8A patent/CN105790808B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103984560A (en) * | 2014-05-30 | 2014-08-13 | 东南大学 | Embedded reconfigurable system based on large-scale coarseness and processing method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105790808A (en) | 2016-07-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109034373A (en) | The parallel processor and processing method of convolutional neural networks | |
| CN109086076A (en) | Processing with Neural Network device and its method for executing dot product instruction | |
| Berntsen | Communication efficient matrix multiplication on hypercubes | |
| CN110245752A (en) | A kind of connection operation method and device entirely | |
| Mohanty et al. | Memory footprint reduction for power-efficient realization of 2-D finite impulse response filters | |
| CN107704916A (en) | A kind of hardware accelerator and method that RNN neutral nets are realized based on FPGA | |
| CN106203621A (en) | The processor calculated for convolutional neural networks | |
| CN105930902A (en) | Neural network processing method and system | |
| CN105790808B (en) | A kind of reconfigurable arrays framework and its detection method towards MIMO detections | |
| CN103677739A (en) | Configurable multiply accumulation cell and multiply accumulation array consisting of same | |
| US8406334B1 (en) | Overflow resistant, fixed precision, bit optimized systolic array for QR decomposition and MIMO decoding | |
| CN109271138A (en) | A kind of chain type multiplication structure multiplied suitable for big dimensional matrix | |
| CN107341133A (en) | The dispatching method of Reconfigurable Computation structure based on Arbitrary Dimensions LU Decomposition | |
| CA1289261C (en) | Polymorphic mesh network image processing system | |
| CN105790809B (en) | Towards coarse-grained reconfigurable array and routing infrastructure in mimo channel detecting system | |
| CN100438211C (en) | Method and unit for beam control of antenna array | |
| CN111160542A (en) | Integrated circuit chip device and related product | |
| CN107368459B (en) | Scheduling method of reconfigurable computing structure based on arbitrary dimension matrix multiplication | |
| Johnsson et al. | Multiple tridiagonal systems, the alternating direction method, and Boolean cube configured multiprocessors | |
| CN103914429B (en) | Multimode data for coarseness dynamic reconfigurable array transmits connectors | |
| CN109615061A (en) | A kind of convolution algorithm method and device | |
| CN113055060B (en) | Coarse-grained reconfigurable architecture system for large-scale MIMO signal detection | |
| CN103873203B (en) | Signal detection method and device | |
| CN105844009A (en) | Efficient sparse matrix storage and numerical reservoir simulation method and apparatus | |
| JP2020177641A (en) | Chip device and related products |
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 |