WO2005078599A1 - Rekonfigurierbares schaltwerk zur parallelen berechung beliebiger algorithmen - Google Patents
Rekonfigurierbares schaltwerk zur parallelen berechung beliebiger algorithmen Download PDFInfo
- Publication number
- WO2005078599A1 WO2005078599A1 PCT/EP2005/050500 EP2005050500W WO2005078599A1 WO 2005078599 A1 WO2005078599 A1 WO 2005078599A1 EP 2005050500 W EP2005050500 W EP 2005050500W WO 2005078599 A1 WO2005078599 A1 WO 2005078599A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- stage
- architecture according
- architecture
- implicants
- switching
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F15/00—Digital computers in general; Data processing equipment in general
- G06F15/76—Architectures of general purpose stored program computers
- G06F15/78—Architectures of general purpose stored program computers comprising a single central processing unit
- G06F15/7867—Architectures of general purpose stored program computers comprising a single central processing unit with reconfigurable architecture
Definitions
- the present invention relates to an architecture of a computer device for parallel calculation of algorithms with at least one switching mechanism.
- Known computer models 5 In general, the so-called Von Neumann architecture, as can be seen from the basic illustration in FIG. 1, is regarded as a universal computer device. This means that with the help of a computer, on this architecture with the components Central Processing Unit [CPU, it contains the Control Unit (CU) and Arithmetical-Logical Unit (ALU)], memory (memory), input / output (Input / output) and bus system based, all algorithmic problems can be calculated in principle.
- CPU Central Processing Unit
- ALU Arithmetical-Logical Unit
- memory memory
- input / output Input / output
- bus system bus system based
- All algorithmic problems can be calculated in principle.
- Such a computer is set to the respective problem by a program, ie a textual description of the algorithm, for example in the programming languages C, C ++ or Java. This program is translated into a machine-readable program by a translator (compiler), which is itself a program.
- the program processing takes place in the computer, which is based on the Von Neumann architecture according to FIG. 1 and similar architectures (eg a Harvard or modified Harvard architecture), in principle in a sequential manner. 5 This is to be understood as the algorithm consisting of a set of machine instructions, processed by knowing the first instruction. The one loading The subsequent instruction is either the one in the memory at the next address (normal program flow), or the last instruction executed was a jump instruction that moves the current program status to another location.
- the internal register, which stores the current program status is referred to as a 'Program Counter' (PC).
- This principle of sequential processing means that exactly one instruction is currently being processed. It is called the Von Neumann principle. Modern architectures called RISC (Reduced Instruction Set Computing), superscalar or VLIW (Very Long Instruction Word) do more than one instruction at a time; however, the principle of sequentiality remains. Overall, this execution principle is referred to as 'time sequential' (computing in time), which indicates that the algorithm requires time.
- RISC Reduced Instruction Set Computing
- VLIW Very Long Instruction Word
- PLDs Programmable Logic Devices
- FPGAs Field-Programmable Gate Arrays
- This architecture is also universal, ie can be used for any algorithmable problem.
- the programming takes place in such a way that elementary computing units, usually defined at the bit level and therefore referred to as logic elements, are interconnected in a network. This form of programming is usually referred to as "configuration" * .
- Cycle time often associated with an external clock, then depends on the complexity of the composition. pending. As a result, a clock rate that is lower than that of processors is used, but this is more than compensated for by the parallelism of the execution. This principle of execution is referred to as 'computing in space' * .
- DEAs deterministic finite automata
- DFAs deterministic finite automata
- FSMs Finite State Machines
- '1' for yj means that i k is taken unchanged, '0' means that i must be selected inverted (noted as / i k ), and '-' stands for don't care; that is, i k is not used. If only the values ⁇ 0, 1 ⁇ are used as attributes for yj, one speaks of the 'canonically disjunctive normal form' 1 .
- RAM Random Addressable Read / Write Memory
- Memory elements which are called CAM and directly support this behavior, are considered special building blocks and are by no means found frequently. For practical applications, however, the functionality of the CAM can be emulated using conventional RAM modules. For this purpose, for all data that would be stored directly in the CAM, the query does not give the value but the memory address, the corresponding addresses must be calculated beforehand and saved at the RAM address that corresponds to the date.
- Cellular automata [CAs (cellular automata)] are a set of finite automata that are in a fixed field Topology are arranged and have further properties (see literature references [1] and [4]).
- Each FSM clearly has neighbors with whom it can communicate.
- CAs are considered universal computers like the previously discussed architectures; they also work completely in parallel. If a network of CAs is to be mapped in a hardware circuit, for example an ASIC or PLD, the number of connections increases linearly with the number of machines. Depending on the chosen topology, the connections themselves are only relatively short and fixed. The effort to communicate the CAs with each other is therefore relatively low. If the memory requirement required to map this switching function into a RAM is considered to be the complexity of a switching function, the maximum complexity of the switching function corresponding to the behavior of a cell increases exponentially with the number of input variables and linearly with the number of output variables of the function. The number of input variables here is the sum of all bits that are necessary to encode the states of all CAs connected to the cell, including the state of the cell itself. The maximum complexity of the cell is essentially determined by the number of connections fertilize any machine.
- GCAs Global cellular automata
- each individual machine As with conventional CAs, the complexity of the individual machines themselves or the underlying switching function essentially depends on the number of connections in each cell. If a GCA is to be mapped in a reconfigurable circuit (PLD), each individual cell, which can implement any switching functions, must allow the maximum possible complexity.
- PLD reconfigurable circuit
- each cell can also implement complex switching functions.
- the effort for any communication all cells increases quadratically with the number of cells.
- the granularity of the circuit is then determined by the number of cells or the bit width of the connections between the cells.
- Such a circuit can do very well Realize GCAs that correspond to the specifications of the circuit in terms of number of FSMs and bit width.
- Complex switching functions can also be implemented in each individual cell.
- GCAs which do not match the specified granularity in the number and required bit width of the connections can only be mapped with difficulty on the circuit.
- each cell must be able to process data from all other cells, including its own status. For this reason, each switching network must be able to implement switching functions that can contain all binary-coded states of all cells as input variables.
- the number of output variables of the switching function only has to enable binary coding of all states of a single cell.
- the disadvantage here is that the complexity of the switching function increases exponentially with the number of input variables.
- Another disadvantage is the increasing polynomial effort for the communication of the cells with one another.
- a (re) configurable architecture (PLD) that is suitable for the inclusion of a GCA must therefore be able to accommodate any complexity per FSM. This means that - if the FSMs are to be mapped directly into a cell of the PLD - the cells must take on any function. It is known that this results in an exponential growth in cell size.
- the configurable network in the PLD must also be fully implemented, i.e. every cell output must be connectable to every other cell. The network grows square, so does the configuration of the network.
- the object of the present invention is to design the architecture specified in introduction 1 in such a way that the aforementioned problems are at least alleviated.
- the possibility is to be shown of being able to map the CGAs to be regarded as an excellent model, which can actually use the parallelism contained in the algorithm, onto programmable hardware such as a PLD.
- a PLD architecture should be specified that enables the inclusion of any CGA.
- the stated object is achieved with the measures specified in claim 1. Accordingly, to take environmentally with the features mentioned at least one individually configurable and / or reconfigurable switching mechanism having a stage input and a stage output architecture, wherein the output variable of at least some of the outputs at a time t n _ ⁇ the input variables to the associated Form inputs of the switching mechanism at time t n and means are provided for storing the output variables of the switching mechanism between times t n - ⁇ and t n . Not all inputs of (stored) outputs need to be occupied; but there can also be free inputs. In addition, there may also be outputs that are not coupled through, that is to say, for example, only represent an end result that is not reused.
- t n _ ⁇ and t n successive designated evaluations of the switching function.
- k is constant, for different applications it can be chosen differently.
- the (re) configurable architecture is suitable for accommodating GCAs.
- it consists of at least one individual configurable switching mechanism, the output variables of which form the input variables of the switching mechanism at the time t n - ⁇ at the time t n .
- the outputs of the switching mechanism are stored in a memory, in particular in registers, so that a complex switching mechanism or an FSM is created.
- the individual levels in the structure also represent derailleurs; only the configuration lies in individual switching networks of each level, whereby each level consists of a configurable switching network and storage elements.
- switching networks are accordingly present as Boolean switching function units, the respective output of which depends on current values at the input of the architecture, while the at least one switching mechanism is to be understood as a Boolean switching function unit, the output of which from current values at the input to this and finally depends on many previous times.
- Switchgear assemblies are therefore composed of switching networks and mostly clock-edge-controlled storage elements.
- Advantageous refinements of the architecture according to the invention emerge from the claims dependent on claim 1.
- the embodiment according to claim 1 can be combined with the features of one of the subclaims or preferably also those from a plurality of subclaims. Accordingly, the architecture can also be designed as follows: Register storage elements can be provided as storage means, these generally being clock-edge-controlled flip-flop elements.
- the switching mechanism is preferably designed in three stages, the three stages being connected in series.
- the second stage with memory elements arranged between the stage with the inputs and the stage with the outputs serves for processing data generated in the first stage and for forwarding processed data to the third stage.
- a first stage can advantageously contain a plurality of memory elements connected in parallel, which can be addressed via input lines, with each memory element being supplied with a subset of the input variables bound in an associated, determined implicant, the second stage being followed by a second stage with memory elements which are identified by identifiers of the individual Implicants are to be addressed, and the second stage is followed by a third stage with means for disjunctively linking the output values of the individual implicants from the memory elements of the second stage.
- the individual implicants can preferably be determined using minimization methods.
- the first stage can be linked to one another via at least one crossbar switch.
- CAMs and / or RAMs can advantageously be provided as memory elements. It is particularly advantageous to integrate at least one CGA.
- Magnetoresistive components in particular of the TMR type, can be provided as memory elements. Corresponding tunnel magnetoresistive elements are known per se.
- a memory element of the architecture is understood to mean any internal element that generally consists of a few transistors or comparable components as well as the actual elements that enable the memory function, e.g. the mentioned magnetoresistive elements (per storage unit) is composed.
- the complexity of a switching function in the selected switching mechanism increases linearly with the number of output variables, but all the effort for communication between the cells is eliminated.
- many individual machines with a few output variables can be mapped, or a few GCAs with many output variables or a mixture of different cells. This means that no granularity is specified, and communication between the FSMs is in principle completely possible.
- a limit is set by the maximum complexity of the switching function that the switching network can accommodate.
- the input variables of the switching function form the address bits and address a memory cell for every possible combination of input variables.
- the content of this memory cell then corresponds to the value of the switching function, the data lines of the memory element form the output variables of the function.
- the advantage of this concept lies in the simple structure, the circuit, the simple reconfigurability, the high integration density of memory elements and the fixed time period which the evaluation of the switching function requires.
- the number of memory cells required that is to say the size of the memory element required, increases exponentially with the number of input variables. For this reason, only small switching functions can be mapped in this way.
- a second option for mapping switching functions in a switching network is to arrange gates in a configurable manner in a 2- or multi-stage network. So that's it possible to map switching function with a minimal consumption of gates in switching networks.
- the circuit theory provides cheap forms of representation of switching functions such as the Reed-Muller form, or powerful algorithms for logic minimization.
- the advantage of this approach is the minimal consumption of gates and the possibility to use powerful methods and algorithms for minimization (cf. references [2] and [3]).
- This method can be used to represent a fixed switching function, for example the implementation of a ROM.
- the individual gates can be reconfigurably connected via hardware modules such as crossbar switches, however, the effort for reconfigurability increases exponentially with the number of input variables of the switching function.
- the basic idea of the adopted embodiment is the development of a reconfigurable switching mechanism which represents as much logical functionality as possible in RAM modules in order to take advantage of the high integration density.
- These RAM modules are to be reconfigurably connected to one another in a multi-stage circuit in order not to have to save complete switching functions.
- the input variables of the switching function are represented by the input lines of the switching network.
- the first stage the circuit consists of several memory elements connected in parallel, which are addressed by the input lines. A subset of the input lines and thus the input variables of the switching function are assigned to each storage element.
- the implicants of a switching function (minimized zi from Eq. (2) or their string term representation) determined by minimization methods are stored in memory elements of the input stage. For this purpose, the assignment of the input variables in the implicant and a unique identifier for this implicant are stored in each storage element to which the input variables are assigned. Since only a part of the input variables are assigned to each block, only a partial implicant is saved.
- no part of the implicant is stored in storage elements to which no variables bound in the implicant are assigned. If there is a bit combination on the input lines of the switching network, all memory elements which contain partial implicants matching the bit combination pass on their identifiers via the data lines to the second stage of the circuit.
- the identifiers of the (partial) implicants address a memory in a second stage.
- the bit patterns of the identifiers belonging to the respective implicant and the output values of the switching function are stored in this memory. If the bit pattern of identifiers supplied by the first stage corresponds to that of a stored implicant, this implicant is present on the input lines of the circuit.
- the second stage of the circuit then forwards the output values of all implicants that are present on the input lines via the data lines to the third stage.
- 3rd stage In the 3rd stage, the output values of the individual implicants are linked disjunctively (OR) and thus form the result of the switching function.
- OR disjunctively
- FIG. 1 shows the basic structure of a Von Neumann architecture according to the prior art
- FIG. 2 shows the general structure of a PLD according to the prior art
- FIG. 3 shows the basic structure of an FSM in the form of a Mealy machine according to the prior art
- FIG. 4 shows an exemplary embodiment of a memory structure of an architecture according to the invention
- FIG. 5 shows the mapping of spray terms to RAM, partial figure a) partial spray terms, partial figure b) the mapping to a tag RAM and partial figure c) the mapping to a conventional RAM
- FIG. 6 shows the mapping of the result of stage 1 of an architecture according to the invention onto a RAM combination in stage 2
- FIG. 7 shows a final architecture for the example
- FIG. 8 shows an architecture according to the invention for a switching mechanism with a large switching network for accommodating a GCA and
- FIG. 9 shows a reconfigurable architecture according to the invention for recording CGAs.
- ne architecture according to the invention with three stages, as mentioned above, for a switching function with 12 input variables, 10 implicants and 8 output variables: Table 1 shows all the implicants (also referred to as 'minterme * ') for an example function. The representation of the string terms is chosen so that three groups of four are created.
- a memory element which stores a partial implicant with the bit combination present returns the identifier of the associated implicant.
- This storage element of the first stage is designed as a 3-value CAM in FIG. 4, i.e. the input vectors which are actually present as two-value information on the address bus are compared with stored three-value information. An identifier, also three-valued, stored for the hit is output as output.
- identifiers together form the identifier of the implicit combination present at the input of the switching network.
- This combination of implicants can include several implicants. For example, in the example shown, the implicants 3, 4 and 8 or the combination of the implicants 4, 5 and 8 can be present. In the first case, the bit combination 100001 is at the 2nd level, in the second case the combination 110001.
- the identifier of the implicit combinations is recognized in the 2nd stage of the circuit and provides for everyone involved
- This second stage now consists of a three-valued RAM, i.e. the address bus information of this stage is trivalent, but the stored data is two-valued.
- a very special building block or a very special architecture can be used as the storage element for accommodating the partial implicants, here denoted by trivalent CAM according to FIG. 4.
- trivalent CAM partial implicants
- a fully associative cache is also possible as an insert for level 1.
- the partial implicants can be saved as a so-called day, the cached date serves as an identifier of the recognized implicant.
- a partially implicit contains unbound variables, which are expressed as Don 't-Care (DC) digits when compared with adjacent bit combinations, a tag must be created in the Tag RAM for all assignments of this implicant that fulfill the comparison with DC who- the.
- DC Don 't-Care
- the comparison with DC overlaps of partial implicants For example, the bit combination 0011 in the first partial implicant memory of the above example belongs to both the implicant 1 and the implicant 3. Not only combinations of implicants are possible, but also combinations of partial implicants.
- each of these RAMs is addressed by part of the input lines of the switching network.
- the same identifier is stored in the memory at the addresses whose bit combination corresponds to the same partial implicant with DC digits (each DC digit in a string term means that the number of applicable digits is increased by a factor of 2 for binary coding).
- each DC digit in a string term means that the number of applicable digits is increased by a factor of 2 for binary coding).
- the output width of the RAM used is larger than the bit width required to display the identifiers, the other bits can be used as a context value. This context can e.g. indicate an invalid assignment of the input variables of the switching function.
- the memory of the second level must be divided: the bit combinations of the first level address a RAM of the second level. An index is stored there for each valid implicant combination, which in turn addresses a RAM, which in turn contains the output variables of the implicants involved. So the different addresses, which are characterized by the combination of implicants
- Don 1 t-Care positions result in being mapped to the same index of the output values of the switching function.
- FIG. 6 shows a corresponding mapping of the result of stage 1 to a RAM combination in stage 2.
- the disjunctively linked output values of all implicants involved in the identified combination must be recorded in the 1st stage of a switching mechanism in the memory of the output values.
- the output RAM must be able to store all possible function values of the switching function at this stage. Since the number of possible function values increases exponentially with the number of output variables of a switching function, several combination and output memories of the 2nd stage of the switching mechanism are used in parallel and their output combinations are linked disjunctively. This means that all the function values of a switching function can be generated.
- all data lines of the implicit memories are connected to all address lines of the combination memories via a crossbar switch. This means that any data lines can address the combination RAMs. Unused data lines can be accessed via the Crossbar switch can be forwarded as context information.
- the generated output value is bit-wise via the exclusive
- the 3rd stage of the circuit thus consists of the disjunctive combination of the output combinations and the subsequent possible inversion of individual output bits.
- the intention of the invention is to offer both a RAM-based architecture for implementing large switching networks and, to a certain extent as an application for accommodating a universal machine, this architecture for accommodating GCAs.
- the memory space in the memory must grow exponentially with the number of inputs (and linearly with the number of outputs). In the case of the example above, this means that any function with 12 inputs and 12 outputs would have a memory requirement of 4096 * 12 bits corresponding to 6144 bytes. With 32 inputs and 8 outputs, this would already be 4 GB of storage capacity.
- the proposed architecture of a multi-level network contains only 211.5 bytes of RAM, namely:
- the first condition is that the number of different partial string terms that are present in a column can be mapped to the RAMs of the first level. This is automatically achieved by the fact that these RAMs record all combinations (because they emulate CAMs), only the input width of the circuit must be sufficient for the application.
- the second condition follows here: The number of different partial string term combinations that are contained in the application after minimization must be codable. This means that a number of storage locations must be available. For efficiency assessment, let m be the input width of the switching network. This would mean that 2m memory cells would be required to display the complete function.
- Condition 3 means that the starting width must be selected appropriately.
- GCAs to architecture
- storage elements must be introduced which store the progress in the calculation in a clock-controlled manner. This is because GCAs are defined as an array of FSMs, and these are usually synchronized. Here it is assumed that a global clock is used for synchronization. All implementations of non-global, in particular not interrelated clocks would lead to significant problems, but are rarely found in practice.
- FIG. 8 shows a further example of a configurable architecture of the proposed type, now equipped with registers for storing states.
- another crossbar switch is inserted, which, among other things, at this point serves to provide input and output interfaces for the computer.
- FIG. 9 shows the basic structure of a reconfigurable architecture according to the invention, as is suitable for the reception of CGAs.
- This structure represents a generalization of the structure according to FIG. 8.
- RAM levels 1 and 2 are illustrated by dashed lines.
Abstract
Description
Claims
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US10/588,860 US7352205B2 (en) | 2004-02-13 | 2005-02-07 | Reconfigurable switching device parallel calculation of any particular algorithms |
JP2006552599A JP5027515B2 (ja) | 2004-02-13 | 2005-02-07 | 任意のアルゴリズムを並列計算するための再構成可能な論理回路装置 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE102004007232.9 | 2004-02-13 | ||
DE102004007232 | 2004-02-13 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2005078599A1 true WO2005078599A1 (de) | 2005-08-25 |
Family
ID=34832667
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/EP2005/050500 WO2005078599A1 (de) | 2004-02-13 | 2005-02-07 | Rekonfigurierbares schaltwerk zur parallelen berechung beliebiger algorithmen |
Country Status (5)
Country | Link |
---|---|
US (1) | US7352205B2 (de) |
JP (2) | JP5027515B2 (de) |
CN (1) | CN100388264C (de) |
DE (1) | DE102005005073B4 (de) |
WO (1) | WO2005078599A1 (de) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7509479B2 (en) * | 2004-09-16 | 2009-03-24 | Siemens Aktiengesellschaft | Reconfigurable global cellular automaton with RAM blocks coupled to input and output feedback crossbar switches receiving clock counter value from sequence control unit |
US7801835B2 (en) * | 2005-06-15 | 2010-09-21 | International Business Machines Corporation | Method for constructing autonomic advisors and learning procedural knowledge from scored examples |
CN101325559B (zh) * | 2008-07-28 | 2010-08-18 | 腾讯科技(深圳)有限公司 | 一种推荐游戏房间的方法、系统和游戏服务器 |
KR101274115B1 (ko) * | 2009-10-06 | 2013-06-13 | 한국전자통신연구원 | 스크램블 장치 및 이의 동작 방법 |
US11113176B2 (en) | 2019-01-14 | 2021-09-07 | Microsoft Technology Licensing, Llc | Generating a debugging network for a synchronous digital circuit during compilation of program source code |
US11144286B2 (en) | 2019-01-14 | 2021-10-12 | Microsoft Technology Licensing, Llc | Generating synchronous digital circuits from source code constructs that map to circuit implementations |
US11275568B2 (en) | 2019-01-14 | 2022-03-15 | Microsoft Technology Licensing, Llc | Generating a synchronous digital circuit from a source code construct defining a function call |
US11106437B2 (en) * | 2019-01-14 | 2021-08-31 | Microsoft Technology Licensing, Llc | Lookup table optimization for programming languages that target synchronous digital circuits |
US11093682B2 (en) | 2019-01-14 | 2021-08-17 | Microsoft Technology Licensing, Llc | Language and compiler that generate synchronous digital circuits that maintain thread execution order |
Family Cites Families (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3948494B2 (ja) * | 1995-04-28 | 2007-07-25 | ザイリンクス,インコーポレイテッド | プログラム可能論理装置によってアクセス可能な分散レジスタを有するマイクロプロセサ |
US6247036B1 (en) * | 1996-01-22 | 2001-06-12 | Infinite Technology Corp. | Processor with reconfigurable arithmetic data path |
DE19843640A1 (de) | 1998-09-23 | 2000-03-30 | Siemens Ag | Verfahren zum Konfigurieren eines konfigurierbaren Hardware-Blocks |
JP4285877B2 (ja) * | 1999-02-23 | 2009-06-24 | 株式会社リコー | 動的再構成計算のためのメタアドレス指定アーキテクチャ及び動的再構成計算のためのメタアドレス指定方法 |
JP4558879B2 (ja) * | 2000-02-15 | 2010-10-06 | 富士通株式会社 | テーブルを用いたデータ処理装置および処理システム |
JP2002026721A (ja) * | 2000-07-10 | 2002-01-25 | Fuji Xerox Co Ltd | 情報処理装置 |
WO2003017095A2 (de) * | 2001-08-16 | 2003-02-27 | Pact Xpp Technologies Ag | Verfahren zum übersetzen von programmen für rekonfigurierbare architekturen |
DE10347975B4 (de) | 2002-10-24 | 2008-10-09 | Siemens Ag | Einrichtung der programmierbaren Logik |
JP4011007B2 (ja) * | 2003-01-15 | 2007-11-21 | 三洋電機株式会社 | リコンフィギュラブル回路を備えた集積回路装置および処理装置 |
DE102004044976A1 (de) * | 2004-09-16 | 2006-03-30 | Siemens Ag | Rechnereinrichtung mit rekonfigurierbarer Architektur |
-
2005
- 2005-02-03 DE DE102005005073A patent/DE102005005073B4/de not_active Expired - Fee Related
- 2005-02-07 US US10/588,860 patent/US7352205B2/en not_active Expired - Fee Related
- 2005-02-07 CN CNB2005800048743A patent/CN100388264C/zh not_active Expired - Fee Related
- 2005-02-07 WO PCT/EP2005/050500 patent/WO2005078599A1/de active Application Filing
- 2005-02-07 JP JP2006552599A patent/JP5027515B2/ja not_active Expired - Fee Related
-
2011
- 2011-11-02 JP JP2011241298A patent/JP2012074051A/ja active Pending
Non-Patent Citations (6)
Title |
---|
HOFFMANN R ET AL: "GCA: a massively parallel model", PARALLEL AND DISTRIBUTED PROCESSING SYMPOSIUM, 2003. PROCEEDINGS. INTERNATIONAL APRIL 22-26, 2003, PISCATAWAY, NJ, USA,IEEE, 22 April 2003 (2003-04-22), pages 270 - 276, XP010645542, ISBN: 0-7695-1926-1 * |
HOFFMANN R, VÖLKMANN K P, HEENES W: "Globaler Zellularautomat (CGA): Ein neues massivparalleles Berechnungsmodell", PARS WORKSHOP, October 2001 (2001-10-01), MUNICH, XP002326432 * |
JOZWIAK L ET AL: "Effective and efficient FPGA synthesis through general functional decomposition", JOURNAL OF SYSTEMS ARCHITECTURE, ELSEVIER SCIENCE PUBLISHERS BV., AMSTERDAM, NL, vol. 49, no. 4-6, September 2003 (2003-09-01), pages 247 - 265, XP004464564, ISSN: 1383-7621 * |
MIRSALEHI M M, GAYLORD T K: "Logical minimization of multilevel coded functions", APPLIED OPTICS, vol. 25, no. 18, 15 September 1986 (1986-09-15), pages 3078 - 3088, XP002326433 * |
RECHENBERG, POMBERGER: "Informatik-Handbuch, 3. Auflage", 2002, HANSER, MÜNCHEN, XP002327006 * |
WIEGAND C, SIEMERS C, RICHTER H: "Definition of a Configurable Architecture for Implementation of Global Cellular Automaton", LECTURE NOTES IN COMPUTER SCIENCE, vol. 2981, February 2004 (2004-02-01), pages 140 - 155, XP002326434, ISSN: 0302-9743, ISBN: 3-540-21238-8, Retrieved from the Internet <URL:http://www.springerlink.com/media/16GPUPQTM0U7JAMNGADT/Contributions/6/L/U/6/6LU6M1C0YCR9MEV2.pdf> [retrieved on 20050428] * |
Also Published As
Publication number | Publication date |
---|---|
JP2012074051A (ja) | 2012-04-12 |
CN1918560A (zh) | 2007-02-21 |
US20070171101A1 (en) | 2007-07-26 |
DE102005005073B4 (de) | 2009-05-07 |
DE102005005073A1 (de) | 2005-09-08 |
JP5027515B2 (ja) | 2012-09-19 |
US7352205B2 (en) | 2008-04-01 |
JP2007522571A (ja) | 2007-08-09 |
CN100388264C (zh) | 2008-05-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2005078599A1 (de) | Rekonfigurierbares schaltwerk zur parallelen berechung beliebiger algorithmen | |
DE69827589T2 (de) | Konfigurierbare Verarbeitungsanordnung und Verfahren zur Benutzung dieser Anordnung, um eine Zentraleinheit aufzubauen | |
DE4416881C2 (de) | Verfahren zum Betrieb einer Datenverarbeitungseinrichtung | |
DE60100476T2 (de) | Unterstützung mehrerer konfigurationszustände eines fpga mittels eines auf dem integrierten schaltkreis dafür vorgesehenen prozessors | |
DE2716369C2 (de) | ||
DE69534812T2 (de) | Programmierbare logische Vorrichtung, die mehr als eine Konfiguration speichert, und Mittel zum Umschalten der Konfiguration | |
EP1222739B1 (de) | Rekonfigurierbares gate-array | |
EP1329816B1 (de) | Verfahren zum selbständigen dynamischen Umladen von Datenflussprozessoren (DFPs) sowie Bausteinen mit zwei- oder mehrdimensionalen programmierbaren Zellstrukturen (FPGAs, DPGAs, o.dgl.) | |
DE60118622T2 (de) | Benutzer-konfigurierbares on-chip speichersystem | |
DE2524046C2 (de) | Elektronische Datenverarbeitungsanlage | |
DE19704044A1 (de) | Verfahren zur automatischen Adressgenerierung von Bausteinen innerhalb Clustern aus einer Vielzahl dieser Bausteine | |
DE4417575A1 (de) | Verbesserte Array-Architektur für programmierbare logische Zellen | |
DE2621882A1 (de) | Dynamischer digitalspeicher mit ununterbrochen umlaufendem datenfluss | |
EP1599794A2 (de) | Prozessor mit verschiedenartigen steuerwerken für gemeinsam genutzte ressourcen | |
DE3114921A1 (de) | Datenverarbeitungssystem | |
DE2926322A1 (de) | Speicher-subsystem | |
DE102013019278A1 (de) | Generator für gefalteten FIFO-Speicher | |
DE2944757A1 (de) | Prozessrechner | |
EP1789889B1 (de) | Rechnereinrichtung mit rekonfigurierbarer architektur zur aufnahme eines globalen zellularen automaten | |
DE10347975B4 (de) | Einrichtung der programmierbaren Logik | |
DE102017217908A1 (de) | Recheneinheit und Betriebsverfahren hierfür | |
DE602004005806T2 (de) | Serieller Speicher mit Mitteln zur Integration eines erweiterten Speicherfeldes | |
EP1116127B1 (de) | Programmgesteuerte einheit | |
DE4447707B4 (de) | Videointerface mit zur Laufzeit rekonfigurierbarer Schaltlogik | |
DE4447706B4 (de) | Partiell zur Laufzeit rekonfigurierbarer Datenflussprozessor |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): BW GH GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
DPEN | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed from 20040101) | ||
WWE | Wipo information: entry into national phase |
Ref document number: 2007171101 Country of ref document: US Ref document number: 10588860 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2006552599 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 200580004874.3 Country of ref document: CN |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWW | Wipo information: withdrawn in national office |
Country of ref document: DE |
|
122 | Ep: pct application non-entry in european phase | ||
WWP | Wipo information: published in national office |
Ref document number: 10588860 Country of ref document: US |