CN105790809B - Towards coarse-grained reconfigurable array and routing infrastructure in mimo channel detecting system - Google Patents
Towards coarse-grained reconfigurable array and routing infrastructure in mimo channel detecting system Download PDFInfo
- Publication number
- CN105790809B CN105790809B CN201610101435.4A CN201610101435A CN105790809B CN 105790809 B CN105790809 B CN 105790809B CN 201610101435 A CN201610101435 A CN 201610101435A CN 105790809 B CN105790809 B CN 105790809B
- Authority
- CN
- China
- Prior art keywords
- mode
- array
- reconfigurable
- routing infrastructure
- computing unit
- 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 invention discloses one kind towards coarse-grained reconfigurable array and routing infrastructure in mimo channel detecting system, and system is made of 14 coarseness Reconfigurable Computation domains, and each coarseness Reconfigurable Computation domain includes:Multi-level high usage route structure, multi-mode reconfigureable computing array structure and multi-functional accelerating module;Multi-mode reconfigureable computing array structure is made of 8*2 configurable computing unit micro-structures, and multi-functional accelerating module accelerates block and sequence that block is accelerated to form by division;Division accelerates block for the solution for first coefficient that disappears in LU decomposition, and sorts and accelerate block for the sequence to each part Euclidean distance in K best algorithms;Configurable computing unit micro-structure is to form the minimum unit of multi-mode reconfigureable computing array structure.The present invention can not only meet the flexibility switched between many algorithms in base band signal process, but also can meet the needs of higher data throughput, more antennas number in Large-scale Mobile communication.
Description
Technical field
The present invention relates to the implementation towards base band signal process in high energy efficiency mobile communication field, it is specifically a kind of can
Applied to coarse-grained reconfigurable array and the implementation of routing infrastructure in extensive, multiple antennas MIMO detecting systems.
Background technology
In forth generation mobile communication system, MIMO technology can be in the case where not increasing bandwidth or total sending power consumption
Significantly increase system data throughput and transmission range thus be widely applied.In mimo systems, channel is estimated
Meter, channel pretreatment and Channel Detection etc. are widely studied as most important key technology.In existing, traditional MIMO systems
In the realization of system, it is primarily present two problems:1) existing realization method is mostly based on VLSI, and flexibility is short of, to not
Lack flexibility in the adaptation of same algorithm, different scales etc. etc.;2) in traditional implementation, many designs institute
The antenna scale that can meet is limited, and does not have expansibility, cannot meet the needs of next generation mobile communication antenna number is up to a hundred.
Coarseness Reconfiguration Technologies having both flexibility and high performance implementation as one kind and increase sharply in, antenna number various towards algorithm
Next generation mobile communication system in have very high feasibility.
Invention content
Goal of the invention:Aiming at the existing problems and shortcomings of the prior art, the present invention provides a kind of logical towards high energy efficiency movement
Realization system in letter system about the coarse-grained reconfigurable array and routing infrastructure of MIMO detections.Using this kind of array architecture and
The MIMO detecting systems of routing infrastructure design can both meet the flexibility switched between many algorithms in base band signal process, together
When again can meet Large-scale Mobile communicate in higher data throughput, more antennas number the needs of.
Technical solution:One kind is towards coarse-grained reconfigurable array and road in high energy efficiency mobile communication mimo channel detecting system
By structure, entire mimo channel detecting system is made of 14 coarseness Reconfigurable Computation domains, each coarseness Reconfigurable Computation
Domain includes three parts:Multi-level high usage route structure, multi-mode reconfigureable computing array structure and multi-functional accelerating module.Its
In, multi-mode reconfigureable computing array structure is made of 8*2 configurable computing unit micro-structures, and multi-functional accelerating module is by removing
Method accelerates block and sequence to accelerate block composition.Division accelerates block for the solution for first coefficient that disappears in LU decomposition, and sorts and block is accelerated to use
To the sequence of each part Euclidean distance in K-best algorithms.Configurable computing unit micro-structure is that composition multi-mode can
The minimum unit of reconstruction calculations array structure, it is interior comprising multiple plus arithmetic element, shift operation unit and a bit register, it can lead to
It crosses configuration and realizes following four arithmetic operation:QR based on Givens rotation decompose in circulating CORDIC operations and linear
CORDIC is operated, addition or accumulation operations, multiplication operation;Using hybrid routing infrastructure, system includes multi-level high usage route
The routing infrastructure of many levels:Routing infrastructure in multi-mode reconfigureable computing array, multi-mode reconfigurable arrays and multi-functional
Routing infrastructure between accelerating module, the routing infrastructure between Reconfigurable Computation domain.Wherein, include in multi-mode reconfigureable computing array
Following two routing infrastructures:Neighbouring rice shape network routing infrastructure and the parallel multi-emitting routing infrastructure of array.Both routing infrastructures
Multiply-add for meeting matrix, QR decomposes, matrix inversion, in K-best the interconnection of the calculating operation of part Euclidean distance want
It asks.Between multi-mode reconfigurable arrays and multi-functional accelerating module use two-way mutual contact mode, and multi-mode Reconfigurable Computation domain it
Between interconnection realized using shared register file;14 coarseness Reconfigurable Computation domains of whole system are divided into two groups, every group
Between include 7 Reconfigurable Computation domains and a shared register file.Shared register file is used to store every group of multiple restructural meter
Calculate the data in domain.The interaction of data is realized between two groups of Reconfigurable Computation domain by sharing register file.Whole system handles number
According to flow it is as follows:Under the control of processor core, each coarseness Reconfigurable Computation domain obtains matching from configuration memory
Confidence ceases.The configuration information of different operation determines the routing infrastructure of the operation, data flow and array scale.Reconfigurable Computation
Multi-mode reconfigurable arrays, multi-functional accelerating module and multi-level routing infrastructure in domain are according to different configurations, coarseness
Reconfigurable Computation domain receives the operation behaviour that the data from memory or other coarseness Reconfigurable Computation domains realize algorithms of different
Make, the demand of data interaction.Configurable computing unit micro-structure is to form the minimum list of multi-mode reconfigureable computing array structure
Member can meet the needs of many algorithms in MIMO detections.Its micro-structure includes 16 shift operation units, 16 add operations
Unit and 8 bit registers.8 add operation units and 8 shift operation units are one group, each can configure computing unit
Micro-structure is divided into two groups.In each grouping, a shift operation unit and an add operation unit are a subgroup.It can match
8 bit registers in computing unit micro-structure are set, not only can be used to implement the storage in Givens rotation direction, but also can use
Some input multiplies the factor storage of each when multiplying.This 8 bit registers and the coupling of each subgroup, are each
Subgroup provides input or judges signal.Each configurable computing unit micro-structure (can be moved by configuring each subgroup
Bit arithmetic unit, add operation unit and a bit register) realize circulating CORDIC in the QR decomposition based on Givens rotation
Operation, linear CORDIC operations, the addition in matrix operation or accumulation operations, multiplication operation meet the needs of algorithms of different.It is single
Computing unit in a configurable computing unit micro-structure can be adjusted according to application demand, and ordinary circumstance can increase subgroup
Number.For example, it is desired to improve the computational accuracy of QR decomposition, the number that can increase subgroup completes more Robin Givens rotation
Turn.Part in configurable computing unit micro-structure designed by the present invention for CORDIC calculating is revolved for based on Robin Givens
Optimization algorithm that the QR turned is decomposed and design, it is only necessary to circulating CORDIC operation and linear CORDIC operations can be realized QR points
Solution operation.
Single multi-mode reconfigurable arrays are made of 8*2 configurable computing unit micro-structures.According in Channel Detection
The feature of each algorithm, between each coarseness Reconfigurable Computation domain, inside each multi-mode reconfigureable computing array and multi-mode can
Multi-level high usage route structure design is used between reconstruction calculations array and multi-functional accelerating module.Wherein, the restructural battle array of multi-mode
Internal row include two kinds of routing infrastructures:Neighbouring rice shape network routing and the parallel multi-emitting routing of array, for meeting Matrix Multiplication, square
Battle array plus, the route need that QR is decomposed, the calculating of part Euclidean distance, LU such as decompose at the operations in K-best algorithms.Such as LU points
Disappear the rapid Cover of first coefficient in an array in solution, the quick row of R matrix elements during required QR is decomposed in K-best calculating process
Covering, single multi-mode reconfigureable computing array structure carry out row covering of configurable computing unit micro-structure when accumulation operations etc.
Operation then needs the parallel multi-emitting routing infrastructure of array;Colleague in being decomposed applied to QR involved by systolic arrays is adjacent configurable
Diagonally can configure the data transmission of computing unit micro-structure between data transmission and adjacent rows between computing unit micro-structure needs
It will be adjacent to the routing infrastructure of rice shape network;And LU is decomposed or other matrix operations involved by routing infrastructure (such as single-row match
Set data interaction between computing unit micro-structure, the data that can configure computing unit micro-structure between adjacent row non-adjacent in the ranks are handed over
It mutually) is also contained in the routing infrastructure of neighbouring rice shape network.In addition, multi-mode reconfigureable computing array and multi-functional accelerating module
Between use two-way mutual contact mode, meet the needs of algorithms of different.For the data interaction between coarseness Reconfigurable Computation domain, this hair
The bright mode using shared register file.
Description of the drawings
Fig. 1 mimo channel detecting system schematic diagrams;
Fig. 2 is configurable computing unit micro-structure structure chart under 4 kinds of operations;
Fig. 3 is that can configure computing unit micro-structure figure in multi-mode reconfigureable computing array.
Specific implementation mode
With reference to specific embodiment, the present invention is furture elucidated, it should be understood that these embodiments are merely to illustrate the present invention
Rather than limit the scope of the invention, after having read the present invention, various equivalences of the those skilled in the art to the present invention
The modification of form falls within the application range as defined in the appended claims.
Towards coarse-grained reconfigurable array and routing infrastructure, such as Fig. 1 in high energy efficiency mobile communication mimo channel detecting system
Shown, entire mimo channel detecting system is made of 14 coarseness Reconfigurable Computation domains, wherein coarseness Reconfigurable Computation domain
In being detected applied to mimo channel, the computation-intensive operations such as K-best algorithms, QR are decomposed, matrix is multiply-add, LU is decomposed, at many levels
High usage route structure is applied to the data interaction between coarseness Reconfigurable Computation domain in coarseness Reconfigurable Computation domain.Coarseness
Reconfigurable Computation domain and multi-level high usage route structure realize the Channel Detection of extensive mimo system jointly.Coarseness can weigh
Structure computational domain can be weighed coarseness based on traditional QR decomposition, the operating characteristic that K-best algorithms, LU are decomposed and matrix is multiply-add
Structure computational domain is divided into multi-mode reconfigureable computing array and multi-functional accelerating module.Each coarseness Reconfigurable Computation domain includes three
A part:Multi-level high usage route structure, multi-mode reconfigureable computing array structure and multi-functional accelerating module.Wherein, multimode
Formula reconfigureable computing array structure is made of 8*2 configurable computing unit micro-structures, and multi-functional accelerating module is accelerated by division
Block and sequence accelerate block composition.Division accelerates block for the solution for first coefficient that disappears in LU decomposition, and sorts and block is accelerated to be used for K-
To the sequence of each part Euclidean distance in best algorithms.Configurable computing unit micro-structure is that composition multi-mode is restructural
The minimum unit of computing array structure, it is interior comprising multiple plus arithmetic element, shift operation unit and a bit register, it can be by matching
It sets and realizes following four arithmetic operation (as shown in Figure 2):Circulating CORDIC during QR based on Givens rotation is decomposed is operated
It is operated with linear CORDIC, addition or accumulation operations, multiplication operation;
Multi-level high usage route uses hybrid routing infrastructure, system to include the routing infrastructure of many levels:Multi-mode can
Routing infrastructure in reconstruction calculations array, the routing infrastructure between multi-mode reconfigurable arrays and multi-functional accelerating module are restructural
Routing infrastructure between computational domain;Requirement for meeting MIMO detection algorithm data flows.Wherein, multi-mode reconfigureable computing array
Interior includes following two routing infrastructures:Neighbouring rice shape network routing infrastructure and the parallel multi-emitting routing infrastructure of array.Both roads
Multiply-add for meeting matrix by structure, QR decomposes, matrix inversion, and the calculating operation of part Euclidean distance is mutual in K-best
Connection requires.Two-way mutual contact mode, and multi-mode Reconfigurable Computation are used between multi-mode reconfigurable arrays and multi-functional accelerating module
Interconnection between domain is realized using shared register file;14 coarseness Reconfigurable Computation domains of whole system are divided into two groups,
Include 7 Reconfigurable Computation domains and a shared register file between every group.Shared register file is used to store every group and multiple weighs
The data of structure computational domain.The interaction of data is realized between two groups of Reconfigurable Computation domain by sharing register file.At whole system
The flow for managing data is as follows:Under the control of processor core, each coarseness Reconfigurable Computation domain, which obtains, comes from configuration memory
Configuration information.The configuration information of different operation determines the routing infrastructure of the operation, data flow and array scale.It is restructural
Multi-mode reconfigurable arrays, multi-functional accelerating module and multi-level routing infrastructure in computational domain are according to different configurations, slightly
Granularity Reconfigurable Computation domain receives the fortune that the data from memory or other coarseness Reconfigurable Computation domains realize algorithms of different
Calculate the demand of operation, data interaction.
As shown in figure 3, configurable computing unit micro-structure is to form the minimum list of multi-mode reconfigureable computing array structure
Member can meet the needs of many algorithms in MIMO detections.Its micro-structure includes 16 shift operation units, 16 add operations
Unit and 8 bit registers.8 add operation units and 8 shift operation units are one group, each can configure computing unit
Micro-structure is divided into two groups.In each grouping, a shift operation unit and an add operation unit are a subgroup.It can match
8 bit registers in computing unit micro-structure are set, not only can be used to implement the storage in Givens rotation direction, but also can use
Some input multiplies the factor storage of each when multiplying.This 8 bit registers and the coupling of each subgroup, are each
Subgroup provides input or judges signal.Each configurable computing unit micro-structure (can be moved by configuring each subgroup
Bit arithmetic unit, add operation unit and a bit register) realize circulating CORDIC in the QR decomposition based on Givens rotation
Operation, linear CORDIC operations, the addition in matrix operation or accumulation operations, multiplication operation meet the needs of algorithms of different.It is single
Computing unit in a configurable computing unit micro-structure can be adjusted according to application demand, and ordinary circumstance can increase subgroup
Number.For example, it is desired to improve the computational accuracy of QR decomposition, the number that can increase subgroup completes more Robin Givens rotation
Turn.Part in configurable computing unit micro-structure designed by the present invention for CORDIC calculating is revolved for based on Robin Givens
Optimization algorithm that the QR turned is decomposed and design, it is only necessary to circulating CORDIC operation and linear CORDIC operations can be realized QR points
Solution operation.
Single multi-mode reconfigurable arrays are made of 8*2 configurable computing unit micro-structures.According in Channel Detection
The feature of each algorithm, between each coarseness Reconfigurable Computation domain, inside each multi-mode reconfigureable computing array and multi-mode can
Multi-level high usage route structure design is used between reconstruction calculations array and multi-functional accelerating module.Wherein, the restructural battle array of multi-mode
Internal row include two kinds of routing infrastructures:Neighbouring rice shape network routing and the parallel multi-emitting routing of array, for meeting Matrix Multiplication, square
Battle array plus, the route need that QR is decomposed, the calculating of part Euclidean distance, LU such as decompose at the operations in K-best algorithms.Such as LU points
Disappear the rapid Cover of first coefficient in an array in solution, the quick row of R matrix elements during required QR is decomposed in K-best calculating process
Covering, single multi-mode reconfigureable computing array structure carry out row covering of configurable computing unit micro-structure when accumulation operations etc.
Operation then needs the parallel multi-emitting routing infrastructure of array;Colleague in being decomposed applied to QR involved by systolic arrays is adjacent configurable
Diagonally can configure the data transmission of computing unit micro-structure between data transmission and adjacent rows between computing unit micro-structure needs
It will be adjacent to the routing infrastructure of rice shape network;And LU is decomposed or other matrix operations involved by routing infrastructure (such as single-row match
Set data interaction between computing unit micro-structure, the data that can configure computing unit micro-structure between adjacent row non-adjacent in the ranks are handed over
It mutually) is also contained in the routing infrastructure of neighbouring rice shape network.In addition, multi-mode reconfigureable computing array and multi-functional accelerating module
Between use two-way mutual contact mode, meet the needs of algorithms of different.For the data interaction between coarseness Reconfigurable Computation domain, this hair
The bright mode using shared register file.
, can be for parallel, ping-pong operation by 96 rank matrix operations in the present invention, whole system calculating section includes more than 14
Pattern reconfigureable computing array, 14 multi-functional accelerating modules are for completing algorithm requirements needed for MIMO detections.More than 14
Pattern reconfigureable computing array and 14 multi-functional accelerating modules are divided into two groups, every group of 7 multi-mode reconfigureable computing arrays and 7
A multi-functional accelerating module.Every group has data of the respective shared register file for storing the group.Every group of 7 multi-modes can
Reconstruction calculations array and 7 multi-functional accelerating modules can disposably handle the QR operation splittings of 96 ranks:96 rank QR are decomposed,
Totally 96 configurable computing unit micro-structures (i.e. 8*2*6=96) are used for matrix to every group of 6 multi-mode reconfigureable computing arrays
96 elements disposable parallel input, the last one multi-mode reconfigureable computing array be used for systolic arrays in last
The calculating of row.Due to each configurable computing unit micro-structure isomorphism, each multi-mode reconfigureable computing array isomorphism, Mei Geduo
Function accelerating module isomorphism, i.e. each coarseness Reconfigurable Computation domain are identical structures.It can be according to Massive-MIMO pairs
The requirement of antenna number, which adjusts in multi-mode reconfigureable computing array, multi-functional accelerating module and configurable computing unit micro-structure, to be added
The number of method, shift operation unit and a bit register meets more massive demand.Wherein, the increase of antenna number can lead to
The number realization for increasing coarseness Reconfigurable Computation domain is crossed, and for the requirement of higher precision, such as increase Givens rotation number
Or multiply factor data bit wide in multiplication operation, can be can configure by increase shift operation unit in computing unit micro-structure,
The number of add operation unit and a bit register meets.
Also include 14 multi-functional accelerating modules in whole system.Multi-functional accelerating module is a customized module, can be with
It is customized acceleration for the part needed for different detection algorithms.The present invention relates to LU decomposition and K-best algorithms, need every
A accelerator module block is made of two parts:LU disappears the solution operation of first coefficient in decomposing, i.e. division arithmetic block;K-best algorithms
Part Euclidean distance sequence.Multi-functional accelerating module and multi-mode reconfigureable computing array two-way communication, can meet LU
Disappear data transmission of first coefficient to multi-mode reconfigureable computing array, can also meet multi-mode reconfigureable computing array operation
Data transmission of the part Euclidean distance result to multi-functional accelerating module.
Coarse-grained reconfigurable array has expansibility.Each configurable computing unit micro-structure isomorphism, each multi-mode
Reconfigureable computing array isomorphism, each multi-functional accelerating module isomorphism, i.e. each coarseness Reconfigurable Computation domain is identical knot
Structure.Can be adjusted according to requirements of the extensive MIMO to antenna number multi-mode reconfigureable computing array, multi-functional accelerating module and
The number of addition, shift operation unit and a bit register in computing unit micro-structure be can configure to meet more massive need
It asks.Wherein, increasing for antenna number can be realized by increasing the number in coarseness Reconfigurable Computation domain, and for higher precision
It is required that as multiplied factor data bit wide in increase Givens rotation number or multiplication operation, list can be calculated by increasing can configure
The number of shift operation unit, add operation unit and a bit register in first micro-structure meets.
Claims (4)
1. it is a kind of towards coarse-grained reconfigurable array and routing infrastructure in high energy efficiency mobile communication mimo channel detecting system, it is special
Sign is:Entire mimo channel detecting system is made of 14 coarseness Reconfigurable Computation domains, each coarseness Reconfigurable Computation
Domain includes three parts:Multi-level high usage route structure, multi-mode reconfigureable computing array structure and multi-functional accelerating module;Its
In, multi-mode reconfigureable computing array structure is made of 8*2 configurable computing unit micro-structures, and multi-functional accelerating module is by removing
Method accelerates block and sequence to accelerate block composition;Division accelerates block for the solution for first coefficient that disappears in LU decomposition, and sorts and block is accelerated to use
To the sequence of each part Euclidean distance in K-best algorithms;Configurable computing unit micro-structure is that composition multi-mode can
The minimum unit of reconstruction calculations array structure, it is interior comprising multiple plus arithmetic element, shift operation unit and a bit register, it can lead to
It crosses configuration and realizes following four arithmetic operation:QR based on Givens rotation decompose in circulating CORDIC operations and linear
CORDIC is operated, addition or accumulation operations, multiplication operation;Using hybrid routing infrastructure, system includes multi-level high usage route
The routing infrastructure of many levels:Routing infrastructure in multi-mode reconfigureable computing array, multi-mode reconfigurable arrays and multi-functional
Routing infrastructure between accelerating module, the routing infrastructure between Reconfigurable Computation domain.
2. as described in claim 1 towards coarse-grained reconfigurable array in high energy efficiency mobile communication mimo channel detecting system and
Routing infrastructure, it is characterised in that:Include following two routing infrastructures in multi-mode reconfigureable computing array:Neighbouring rice shape network road
By structure and the parallel multi-emitting routing infrastructure of array;Two-way interconnection is used between multi-mode reconfigurable arrays and multi-functional accelerating module
Mode, and the interconnection between multi-mode Reconfigurable Computation domain is realized using shared register file;14 coarse grain of whole system
Degree Reconfigurable Computation domain includes 7 Reconfigurable Computation domains and a shared register file between being divided into two groups, every group;Shared register
Heap is used to store the data in every group of multiple Reconfigurable Computation domains;It is realized by shared register file between two groups of Reconfigurable Computation domain
The interaction of data.
3. as described in claim 1 towards coarse-grained reconfigurable array in high energy efficiency mobile communication mimo channel detecting system and
Routing infrastructure, it is characterised in that:The micro-structure of configurable computing unit micro-structure includes 16 shift operation units, 16 additions
Arithmetic element and 8 bit registers;8 add operation units and 8 shift operation units are one group, each configurable calculating
Elementary microstructure is divided into two groups;In each grouping, a shift operation unit and an add operation unit are a subgroup;
8 bit registers in configurable computing unit micro-structure, not only can be used to implement the storage in Givens rotation direction, but also can
Multiply the factor storage of each for some input when multiplying;This 8 bit registers and the coupling of each subgroup are
Each subgroup provides input or judges signal;Each configurable computing unit micro-structure can be real by configuring each subgroup
Now QR based on Givens rotation circulating CORDIC operations, linear CORDIC operations in decomposing, addition in matrix operation or
Accumulation operations, multiplication operation, meet the needs of algorithms of different.
4. as described in claim 1 towards coarse-grained reconfigurable array in high energy efficiency mobile communication mimo channel detecting system and
Routing infrastructure, it is characterised in that:Coarse-grained reconfigurable array has expansibility;Each configurable computing unit micro-structure is same
Structure, each multi-mode reconfigureable computing array isomorphism, each multi-functional accelerating module isomorphism, i.e., each coarseness Reconfigurable Computation
Domain is identical structure;Multi-mode reconfigureable computing array, more work(can be adjusted according to requirements of the extensive MIMO to antenna number
The number of addition, shift operation unit and a bit register it can meet more in accelerating module and configurable computing unit micro-structure
Large-scale demand;Wherein, increasing for antenna number can be realized by increasing the number in coarseness Reconfigurable Computation domain, and for
The requirement of higher precision increases in Givens rotation number or multiplication operation and multiplies factor data bit wide, can be matched by increasing
The number of the shift operation unit in computing unit micro-structure, add operation unit and a bit register is set to meet.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610101435.4A CN105790809B (en) | 2016-02-24 | 2016-02-24 | Towards coarse-grained reconfigurable array and routing infrastructure in mimo channel detecting system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610101435.4A CN105790809B (en) | 2016-02-24 | 2016-02-24 | Towards coarse-grained reconfigurable array and routing infrastructure in mimo channel detecting system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105790809A CN105790809A (en) | 2016-07-20 |
CN105790809B true CN105790809B (en) | 2018-08-21 |
Family
ID=56403617
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610101435.4A Active CN105790809B (en) | 2016-02-24 | 2016-02-24 | Towards coarse-grained reconfigurable array and routing infrastructure in mimo channel detecting system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105790809B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108717402B (en) * | 2018-05-16 | 2021-03-30 | 清华大学 | Memory for reconfigurable processing system and reconfigurable processing system |
CN114065122A (en) * | 2020-07-31 | 2022-02-18 | 深圳市中兴微电子技术有限公司 | Data processing method, device and storage medium |
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 |
CN113225161B (en) * | 2021-04-14 | 2022-09-13 | 上海微波技术研究所(中国电子科技集团公司第五十研究所) | 5G terminal PDCCH MIMO detection accelerator and design method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102508816A (en) * | 2011-11-15 | 2012-06-20 | 东南大学 | Configuration method applied to coarse-grained reconfigurable array |
CN103970720A (en) * | 2014-05-30 | 2014-08-06 | 东南大学 | Embedded reconfigurable system based on large-scale coarse granularity and processing method of system |
CN104933008A (en) * | 2015-06-24 | 2015-09-23 | 东南大学 | Reconfigurable system and reconfigurable array structure and application of reconfigurable array structure |
-
2016
- 2016-02-24 CN CN201610101435.4A patent/CN105790809B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102508816A (en) * | 2011-11-15 | 2012-06-20 | 东南大学 | Configuration method applied to coarse-grained reconfigurable array |
CN103970720A (en) * | 2014-05-30 | 2014-08-06 | 东南大学 | Embedded reconfigurable system based on large-scale coarse granularity and processing method of system |
CN104933008A (en) * | 2015-06-24 | 2015-09-23 | 东南大学 | Reconfigurable system and reconfigurable array structure and application of reconfigurable array structure |
Non-Patent Citations (4)
Title |
---|
A 684Mbps 57mW Joint QR Decomposition and MIMO Processor for 4×4 MIMO-OFDM Systems;Chiu PoLin等;《Solid State Circuits Conference(A-SSCC)》;IEEE;20120109;309-312 * |
A Heterogeneous Reconfigurable Cell Array for MIMO Signal Processing;Zhang Chenxin等;《IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—I: REGULAR PAPERS》;IEEE;20141203;第62卷(第3期);733-742 * |
An Energy-Efficient Coarse-Grained Dynamically Reconfigurable Fabric for Multiple-Standard Video Decoding Applications;Leibo Liu等;《Custom Integrated Circuits Conference(2013)》;IEEE;20131111;1-4 * |
BilRC: An Execution Triggered Coarse Grained Reconfigurable Architecture;Oguzhan Atak等;《IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS》;IEEE;20120731;第21卷(第7期);1285-1298 * |
Also Published As
Publication number | Publication date |
---|---|
CN105790809A (en) | 2016-07-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105790809B (en) | Towards coarse-grained reconfigurable array and routing infrastructure in mimo channel detecting system | |
US7249242B2 (en) | Input pipeline registers for a node in an adaptive computing engine | |
CN101782893B (en) | Reconfigurable data processing platform | |
Ma et al. | End-to-end scalable FPGA accelerator for deep residual networks | |
US8543795B2 (en) | Adaptive integrated circuitry with heterogeneous and reconfigurable matrices of diverse and adaptive computational units having fixed, application specific computational elements | |
US8533431B2 (en) | Adaptive integrated circuitry with heterogeneous and reconfigurable matrices of diverse and adaptive computational units having fixed, application specific computational elements | |
CN108780492A (en) | Simulate coprocessor | |
CN105912501B (en) | A kind of SM4-128 Encryption Algorithm realization method and systems based on extensive coarseness reconfigurable processor | |
CN105335331A (en) | SHA256 realizing method and system based on large-scale coarse-grain reconfigurable processor | |
CN107817708B (en) | High-compatibility programmable neural network acceleration array | |
TW200828044A (en) | Pipeline structure reconfigurable mixed-radix Fast Fourier Transform | |
Joardar et al. | REGENT: A heterogeneous ReRAM/GPU-based architecture enabled by NoC for training CNNs | |
CN106294278A (en) | The pre-configured controller of adaptive hardware of system is calculated for dynamic reconfigurable array | |
Peker | A fully customizable hardware implementation for general purpose genetic algorithms | |
CN101617306A (en) | Fast fourier transform architecture | |
Noh et al. | FlexBlock: A flexible DNN training accelerator with multi-mode block floating point support | |
Xu et al. | HeSA: Heterogeneous systolic array architecture for compact CNNs hardware accelerators | |
CN106021171A (en) | An SM4-128 secret key extension realization method and system based on a large-scale coarseness reconfigurable processor | |
CN104063357A (en) | Processor And Processing Method | |
JP2010535370A (en) | An electronic system that emulates the “DNA” structure of a chromosome | |
Li et al. | A unified reconfigurable floating-point arithmetic architecture based on CORDIC algorithm | |
EP0239634A1 (en) | Multizone processor | |
CN105790808B (en) | A kind of reconfigurable arrays framework and its detection method towards MIMO detections | |
CN111160543A (en) | Integrated circuit chip device and related product | |
Gao et al. | Fca-bnn: Flexible and configurable accelerator for binarized neural networks on fpga |
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 |