Abstract: A memory unit is controlled by a first word line and a second word line. The memory unit includes a memory cell and a multi-bit input local computing cell. The memory cell stores a weight. The memory cell is controlled by the first word line and includes a local bit line transmitting the weight. The multi-bit input local computing cell is connected to the memory cell and receives the weight via the local bit line. The multi-bit input local computing cell includes a plurality of input lines and a plurality of output lines. Each of the input lines transmits a multi-bit input value, and the multi-bit input local computing cell is controlled by the second word line to generate a multi-bit output value on each of the output lines according to the multi-bit input value multiplied by the weight.

Abstract: Embodiments provide a write operation circuit, a semiconductor memory, and write operation method. The write operation circuit includes: a data determination module that determines whether to flip an input data of the semiconductor memory depending on the number of high data bits in the input data so as to generate a flip flag data and a first intermediate data; a data buffer module that determines whether to flip a global bus according to a second intermediate data, where the second intermediate data is an inverted data of the first intermediate data; a data receiving module that decodes the global bus data according to the flip flag data and writes the decoded data into a memory bank of the semiconductor, where the decoding including determining whether to flip the global bus data; and a precharge module that sets the initial state of the global bus to low.

Abstract: According to one embodiment, a memory system includes a semiconductor memory and a controller. The semiconductor memory includes blocks each containing memory cells. The controller is configured to instruct the semiconductor memory to execute a first operation and a second operation. In the first operation and the second operation, the semiconductor memory selects at least one of the blocks, and applies at least one voltage to all memory cells contained in said selected blocks. A number of blocks to which said voltage is applied per unit time in the second operation is larger than that in the first operation.

Abstract: A method for finding an extreme value among a plurality of numbers in an associative memory includes creating a spread-out representation (SOR) for each number of the plurality of numbers, storing each SOR in a column of the associative memory array and performing a horizontal bit-wise Boolean operation on rows of the associative memory array to produce an extreme SOR (ESOR) having the extreme value. A system for finding an extreme value includes an associative memory array to store the plurality of numbers, each number storable in a column; a spread-out representation (SOR) creator to create a SOR for each number of the plurality of numbers and to store each SOR in a column of the associative memory array, and an extreme SOR (ESOR) finder to find an extreme value using a horizontal bit-wise Boolean operation on rows of the associative memory array storing bits of the SORs.

Abstract: Examples disclosed herein relate, in one aspect, to method of searching a content addressable memory (CAM) that stores a plurality of entries. The method may include obtaining a search word corresponding to a matching data word stored in a matching entry of the CAM, where the matching entry may include a plurality of data words. The method may also include determining, based at least on a value of a predetermined bit of the search word, a search mask to mask off any data words within the matching entry other than the matching data word. The method may also determine, based on the search mask and a search key that includes the search word, the address of the matching entry within the CAM.

Abstract: A method and apparatus of a network element that processes data by a network element with a data processing pipeline is described. In an exemplary embodiment, the network element receives network data and performs a policy-based routing lookup using one or more characteristics of the network data to retrieve a next hop identifier. In addition, the network element generates a key for the next hop identifier and performs a longest prefix match lookup to retrieve a forwarding result. The network element further determines a next hop interface based on the forwarding result.

Abstract: A method for finding an extreme value among a plurality of numbers in an associative memory includes creating a spread-out representation (SOR) for each number of the plurality of numbers, storing each SOR in a column of the associative memory array and performing a horizontal bit-wise Boolean operation on rows of the associative memory array to produce an extreme SOR (ESOR) having the extreme value. A system for finding an extreme value includes an associative memory array to store the plurality of numbers, each number storable in a column; a spread-out representation (SOR) creator to create a SOR for each number of the plurality of numbers and to store each SOR in a column of the associative memory array, and an extreme SOR (ESOR) finder to find an extreme value using a horizontal bit-wise Boolean operation on rows of the associative memory array storing bits of the SORs.

Abstract: Examples disclose a crossbar memory with a first crossbar to write data values corresponding to a word. The crossbar memory further comprises a second crossbar, substantially parallel to the first crossbar, to receive voltage for activation of data values across the second crossbar. Additionally, the examples of the crossbar memory provide an output line that interconnects with the crossbars at junctions, to read the data values at the junctions. Further, the examples of the crossbar memory provide a logic module to determine whether the second crossbar data values correspond to the word written in the first crossbar.

Abstract: The hybrid dynamic-static encoder described herein may combine dynamic and static structural and logical design features that strategically partition dynamic nets and logic to substantially eliminate redundancy and thereby provide area, power, and leakage savings relative to a fully dynamic encoder with an equivalent logic delay. For example, the hybrid dynamic-static encoder may include identical top and bottom halves, which may be combined to produce final encoded index, hit, and multi-hit outputs. Each encoder half may use a dynamic net for each index bit with rows that match a search key dotted. If a row has been dotted to indicate that the row matches the search key, the dynamic nets associated therewith may be evaluated to reflect the index associated with the row. Accordingly, the hybrid dynamic-static encoder may have a reduced set of smaller dynamic nets that leverage redundant pull-down structures across the index, hit, and multi-hit dynamic nets.

Abstract: A method for managing data stored in a content addressable memory (CAM) device includes at least the following steps: performing a partial write operation to overwrite only a portion of original write data stored in an entry of the CAM device, and storing updated write data in the entry; and updating a parity flag by a first value to indicate that parity data corresponding to the entry of the CAM device is invalid. Besides, a CAM system employing the method has a CAM device, a storage device and a parity flag controller.

Abstract: A method for content addressable memory (CAM) error recovery that includes detecting an error in an entry of a CAM, identifying an address of the entry in the CAM, copying data from the address in the backup random access memory (RAM) into the entry of the CAM to obtain a corrected CAM, clearing a results (first in first out) FIFO structure based on detecting the error, performing, using the corrected CAM, a match request stored in a replay FIFO structure to obtain a revised result, and storing the revised result in the results FIFO structure.

Abstract: A digital design and technique may be used to implement a Manhattan Nearest Neighbor content addressable memory function by augmenting a serial content addressable memory design with additional memory and counters for bit serially accumulating in parallel and subsequently comparing in parallel all the Manhattan distances between a serially inputted vector and all corresponding vectors resident in the CAM. Other distance measures, besides a Manhattan distance, may optionally be used in conjunction with similar techniques and designs.

Abstract: This invention discloses a 2T-string NOR-based CAM logic cell comprising two physical NAND cells connected in series with two horizontal WLs and one vertical BL and one vertical SL. Additionally, a sector of NOR-based CAM logic cell array is configured with N vertical cell strings each including M 2T-string NOR-based CAM logic cells connected in parallel sharing a local vertical SL and one dedicated vertical ML as an Operand word vertical page. Each 2T-string NOR-based CAM logic cell can be either a binary or ternary CAM cell associated with two or three physical states assigned to NAND cells' Vt levels for defining CAM logic states. Logic match of M-logic-bit inputs is found for at least one vertical page if the corresponding M 2T-string NOR-based CAM logic cells are in non-conduction state, providing M times faster Compare performance over the NAND-based CAM and 2 time faster than SRAM-based CAM.

Abstract: An associative memory that can reduce search errors is provided. An associative memory includes R distance/time conversion circuits DT1 to DTR. The R distance/time conversion circuits DT1 to DTR each include a NAND circuit 40 and N bit stages 41 to 4k. The N bit stages 41 to 4k delay a signal from the NAND circuit 40 by longer delay time as the distance between reference data and search data is greater and oscillate the signal. Among R oscillation signals output from the distance/time conversion circuits DT1 to DTR, the earliest changing oscillation signal is detected as an oscillation signal for the Winner row.

Abstract: A method within a hybrid ternary content addressable memory (TCAM) includes comparing a first portion of a search word to a first portion of a stored word in a first TCAM stage. The method further includes interfacing an output of the first TCAM stage to an input of the second TCAM stage. The method also includes comparing a second portion of the search word to a second portion of the stored word in a second TCAM stage when the first portion of the search word matches the first portion of the stored word. The first TCAM stage is different from the second TCAM stage.

Abstract: The present invention provides a content addressable memory capable of higher frequency operation than conventional. When a search enable signal supplied from a search control circuit is asserted, each of search line drivers transfers search data to each CAM cell of a CAM memory array via a search line pair. The search line enable signal is transmitted to the search line drivers via a single control signal line coupled to the search control circuit. The control signal line is coupled to the search line drivers in such a manner that the search line enable signal passes through coupling nodes between the search line drivers and the control signal line in an arrangement order of the search line drivers from the side far away as viewed from match amplifiers.

Abstract: In one embodiment, a first search operation is performed based on a base lookup word on a first plurality of content-addressable memory entries of an overall plurality of priority-ordered content-addressable memory entries to identify a first matching entry and a corresponding first overall search position of the first matching entry within the overall plurality of priority-ordered content-addressable memory entries. A second search operation is performed based on the base lookup word on a second plurality of content-addressable memory entries of the overall plurality of priority-ordered content-addressable memory entries to identify a second matching entry and a corresponding second overall search position of the second matching entry within the overall plurality of priority-ordered content-addressable memory entries. The corresponding first overall search position is compared to the corresponding second overall search position to determine the overall search result.

Abstract: A multi-priority encoder includes a plurality of interconnected, single-priority encoders arranged in descending priority order. The multi-priority encoder includes circuitry for blocking a match output by a lower level single-priority encoder if a higher level single-priority encoder outputs a match output. Match data is received from a content addressable memory, and the priority encoder includes address encoding circuitry for outputting the address locations of each highest priority match line flagged by the highest priority indicator. Each single-priority encoder includes a highest priority indicator which has a plurality of indicator segments, each indicator segment being associated with a match line input.

Abstract: Embodiments of content addressable memories for internet protocol devices and operations are described herein. Other examples and related methods are also disclosed herein.

Abstract: A memory includes a plurality of content-addressable memory (CAM) cells and a summary circuit associated with the plurality of CAM cells. The summary circuit includes a first level of logic gates and a second level of logic gates. The first level of logic gates have inputs each configured to receive an output of a corresponding one of the plurality of CAM cell. The second level of logic gates have inputs each configured to receive an output of a corresponding one of the first level of logic gates.

Abstract: Embodiments of content addressable memories for internet protocol devices and operations are described herein. Other examples and related methods are also disclosed herein.

Abstract: An associative list processing unit and method comprising employing a plurality of prioritized cell blocks and permitting inserts to occur in a single clock cycle if all of the cell blocks are not full.

Abstract: Methods and circuits for CAM cells using PMCs are disclosed herein. In one embodiment, a BCAM cell can include: (i) a first PMC coupled to a first access transistor and a bit node, where the first access transistor is coupled to a true bit line; (ii) a second PMC cell coupled to a second access transistor and the bit node, where the second access transistor is coupled to a complement bit line, and the first and second access transistors are controllable by a word line; (iii) a program enable transistor coupled to the bit node, and configured to couple a program control voltage to the bit node when enabled; and (iv) a match indication transistor configured to discharge a match line in response to states of the true and complement bit lines relative to the bit node.

Abstract: Quaternary CAM cells are provided that include a compare circuit having a discharge path between a match line and ground potential, the single discharge path consisting essentially of a single transistor. In an embodiment, the single transistor has a gate coupled to a pull-down node and the compare circuit includes first and second gating transistors connected in series between the pull-down node and a ground potential, the first gating transistor having a gate to receive a comparand bit, and the second gating transistor having a gate to receive a complemented comparand bit.

Abstract: Array area and power consumption are reduced in a content addressable memory. A comparator for performing a match determination and a size determination is provided commonly for plural entries each storing data to be retrieved. Each entry includes data storage cells for storing data and mask cells for storing mask bits. The number of mask cells is smaller than that of the data storage cells. Search data is transmitted to the comparator via a search data bus. One of the entries is selected according to a predetermined rule. The comparator decodes the mask bits, generates a mask instruction signal, and performs match comparison and size comparison between the search data and data to be retrieved which is stored in the selected entry.

Abstract: A multi-priority encoder includes a plurality of interconnected, single-priority encoders arranged in descending priority order. The multi-priority encoder includes circuitry for blocking a match output by a lower level single-priority encoder if a higher level single-priority encoder outputs a match output. Match data is received from a content addressable memory, and the priority encoder includes address encoding circuitry for outputting the address locations of each highest priority match line flagged by the highest priority indicator. Each single-priority encoder includes a highest priority indicator which has a plurality of indicator segments, each indicator segment being associated with a match line input.

Abstract: Aspects of the disclosure provide a method for encoding ranges in a ternary content addressable memory (TCAM). The method includes determining first positive ranges and first negative ranges corresponding to a first encoding range to be encoded in the TCAM. The first encoding range is in association with a first action. The first positive ranges include the first encoding range. The first negative ranges exclude the first encoding range. At least a first positive range and a first negative range are overlapping. Further, the method includes encoding the first positive ranges in first TCAM entries, and encoding the first negative ranges in second TCAM entries. At least one of the second TCAM entries has a higher priority than one of the first TCAM entries. Then, the method includes associating the first TCAM entries to the first action, and associating the second TCAM entries to a reject action.

Abstract: The present invention provides a content addressable memory capable of higher frequency operation than conventional. When a search enable signal supplied from a search control circuit is asserted, each of search line drivers transfers search data to each CAM cell of a CAM memory array via a search line pair. The search line enable signal is transmitted to the search line drivers via a single control signal line coupled to the search control circuit. The control signal line is coupled to the search line drivers in such a manner that the search line enable signal passes through coupling nodes between the search line drivers and the control signal line in an arrangement order of the search line drivers from the side far away as viewed from match amplifiers.

Abstract: A content addressable memory device capable of making simultaneous pursuit of low power consumption and speeding up is provided. A match amplifier A determines coincidence or non-coincidence of search data and data stored in a content addressable memory in an entry of a memory array A, according to a voltage of a match line MLA. A match amplifier B determines coincidence or non-coincidence of search data and data stored in a content addressable memory in an entry of a memory array B, according to a voltage of a match line MLB. A block-B control circuit directs to start searching in the memory array B after two cycles after searching has been started in the memory array A. A block-B activation control circuit directs to stop searching in the memory array B according to a voltage of the match line MLA after searching in the memory array A.

Abstract: A static CAM includes a plurality of entries E each including a number of CAM cells B and a summary S. Each CAM cell B is associated with a memory cell M and a comparator C. Generally, the CAM receives as input i number of lookup data lines. When data is received, memory cells M provide compared data for corresponding comparators C in CAM cells B to compare the compared data to the received data. If all compared data match all received data lines for an entry, then there is a hit for that entry. But if any compared data does not match the corresponding data line, then there is a miss for that line and therefore a miss for that entry. Depending on applications, the CAM returns an address if there is a hit for one or a plurality of entries.

Abstract: The power consumption of a search engine such as a CAM device is dynamically adjusted to prevent performance degradation and/or damage resulting from overheating. For some embodiments, the CAM device is continuously sampled to generate sampling signals indicating the number of active states and number of compare operations performed during each sampling period. The sampling signals are accumulated to generate an estimated device power profile, which is compared with reference values corresponding to predetermined power levels to generate a dynamic power control signal indicating predicted increases in the device's operating temperature resulting from its power consumption. The dynamic power control signal is then used to selectively reduce the input data rate of the CAM device, thereby reducing power consumption and allowing the device to cool.

Abstract: Methods and circuits for CAM cells using PMCs are disclosed herein. In one embodiment, a BCAM cell can include: (i) a first PMC coupled to a first access transistor and a bit node, where the first access transistor is coupled to a true bit line; (ii) a second PMC cell coupled to a second access transistor and the bit node, where the second access transistor is coupled to a complement bit line, and the first and second access transistors are controllable by a word line; (iii) a program enable transistor coupled to the bit node, and configured to couple a program control voltage to the bit node when enabled; and (iv) a match indication transistor configured to discharge a match line in response to states of the true and complement bit lines relative to the bit node.

Abstract: The present invention is directed to reduce array area and power consumption in a content addressable memory. A comparator for performing a match determination and a size determination is provided commonly for plural entries each storing data to be retrieved. Each entry includes data storage cells for storing data and mask cells for storing mask bits. The number of mask cells is smaller than that of the data storage cells. Search data is transmitted to the comparator via a search data bus. One of the entries is selected according to a predetermined rule. The comparator decodes the mask bits, generates a mask instruction signal, and performs match comparison and size comparison between the search data and data to be retrieved which is stored in the selected entry.

Abstract: A content addressable memory and method of operation uses a memory array having a plurality of rows of stored content addressable memory data and compare circuitry for comparing received comparand data with the stored content addressable memory data. A hit signal and one or more parity bits is provided for each row. Erroneous hit detection circuitry coupled to the memory array for each row generates a row error indicator in response to a comparison between parity of the comparand data and parity of a row that is correlated to the hit signal as qualified by assertion of a hit signal of that row. The erroneous hit detection circuitry uses the row error indicator for each row to provide an output which indicates whether at least one asserted hit signal corresponds to an erroneous hit.

Abstract: A network apparatus comprises a plurality of ports, and a forwarding engine coupled to the plurality of ports. The forwarding engine is configured to transfer data units received via at least some of the plurality of ports to one or more appropriate ports in the plurality of ports. The forwarding engine comprises a content addressable memory (CAM) device to store a plurality of data patterns organized in a plurality of groups, wherein the CAM device is configured to, responsive to input data, output in a single cycle a plurality of match indications corresponding to the plurality of groups. The forwarding engine also comprises a logic device coupled to the CAM device and configured to generate an action value based on the plurality of match indications, wherein the action value indicates an action to be taken by the forwarding engine.

Abstract: Embodiments of content addressable memories for internet protocol devices and operations are described herein. Other examples and related methods are also disclosed herein.

Abstract: The power consumption of a search engine such as a CAM device is dynamically adjusted to prevent performance degradation and/or damage resulting from overheating. For some embodiments, the CAM device is continuously sampled to generate sampling signals indicating the number of active states and number of compare operations performed during each sampling period. The sampling signals are accumulated to generate an estimated device power profile, which is compared with reference values corresponding to predetermined power levels to generate a dynamic power control signal indicating predicted increases in the device's operating temperature resulting from its power consumption. The dynamic power control signal is then used to selectively reduce the input data rate of the CAM device, thereby reducing power consumption and allowing the device to cool.

Abstract: Aspects of the disclosure provide a CAM module that can be used independent of a defective entry line. The CAM module can include at least a CAM array having at least X CAM entry lines, and an additional CAM entry line. Each CAM entry line may include a selection line for enabling the CAM entry line for writing and/or reading and an entry output for indicating matching to a search key. Further, the CAM module can include a decoder unit that can decode an address to enable one out of X word-lines, and an encoder unit that can encode X matching outputs to a matching address according to a predetermined priority sequence. Additionally, the CAM module can include a switching unit coupling the CAM array with the decoder unit and the encoder unit.

Abstract: A CAM (Content Addressable Memory) cell includes first and second data storage portions storing data, horizontal port write gates for storing data applied through a match line pair in the data storage portions in a data write through a horizontal port, and search/read gates for driving the match lines of the match line pair in accordance with the data stored in the data storage portions in a search operation and in a data read through the horizontal port. The match lines are used as horizontal bit line pair, or signal lines for accessing the horizontal port. As the first and second data storage portions are used, it becomes possible to store ternary data, and accordingly, a write mask function of inhibiting a data write at a destination of data transfer is realized. Further, as the CAM cell is used, an arithmetic/logic operation following a search process can be executed selectively, and high speed data writing/reading becomes possible.

Abstract: An embedded processor system including a flash process semiconductor die and a digital process semiconductor die. The flash process semiconductor die includes i) first cache memory configured to cache information associated with an embedded processor, and ii) a first cache controller configured to control the first cache memory. The digital process semiconductor die includes i) a translator configured to translate the information between the flash process semiconductor die and the digital process semiconductor die, and ii) the embedded processor. The embedded processor is configured to process the information.

Abstract: An encoding scheme is disclosed that allows a CAM device to selectively store, within each cell of a row of the CAM device, either a single bit of a binary value or two bits of an encoded data word encoded from the binary value. By storing two bits of the encoded data word in each CAM cell, data may be stored more efficiently and CAM systems may consume less power. The encoded data words can be balanced data words that have equal number of logic high and logic low values.

Abstract: Present embodiments describe a CAM device having a segmented CAM array. Each segment of the CAM array, or cell blocks, includes one or more rows of CAM cells. One or more of the cell blocks in the CAM array are selectively enabled during a search operation based on a detected matching condition of another cell block. By selectively enabling cell blocks during search operations only when needed, energy consumption is reduced. Selectively enabling a cell block includes selectively pre-charging match lines to the cell block, selectively enabling word lines to the cell block, and selectively enabling comparand line to the cell block. In accordance with certain embodiments, the CAM device is configurable to perform search operations in a pipelined manner. Accordingly, the CAM device is capable of performing multiple search operations simultaneously.

Abstract: The power consumption of a search engine such as a CAM device is dynamically adjusted to prevent performance degradation and/or damage resulting from overheating. For some embodiments, the CAM device is continuously sampled to generate sampling signals indicating the number of active states and number of compare operations performed during each sampling period. The sampling signals are accumulated to generate an estimated device power profile, which is compared with reference values corresponding to predetermined power levels to generate a dynamic power control signal indicating predicted increases in the device's operating temperature resulting from its power consumption. The dynamic power control signal is then used to selectively reduce the input data rate of the CAM device, thereby reducing power consumption and allowing the device to cool.

Abstract: Present embodiments allow a search engine to quickly save and restore state information to and from an external state memory when switching between multiple data flows by transferring the state information between the search engine and the external state memory in a parallel manner. More specifically, for CAM-based search engines configured according to present embodiments, the CAM array includes state information gating circuits that selectively allow state information stored in the CAM array's match latches to be transposed onto the array's bit lines and then read from the array using the array's sense amplifiers.

Abstract: A CAM device includes an array of CAM cells each having a spin torque transfer (STT) storage cell to store a data bit. Each STT storage cell includes a first magnetic tunnel junction (MTJ) element coupled between a first input node and an output node of the CAM cell, a second MTJ element coupled between a second input node and the output node of the CAM cell, and a first match transistor coupled between the match line and ground potential and having a gate coupled to the output node. The logic state of the data bit is represented by the relative resistances of the first and second MTJ elements.

Abstract: Present embodiments allow a search engine to quickly save and restore state information to and from an external state memory when switching between multiple data flows by transferring the state information between the search engine and the external state memory in a parallel manner. More specifically, for CAM-based search engines configured according to present embodiments, the CAM array includes state information gating circuits that selectively allow state information stored in the CAM array's match latches to be transposed onto the array's bit lines and then read from the array using the array's sense amplifiers.

Abstract: Approaches for an improved encoding scheme that allows a CAM device to selectively store, within each cell of a row of the CAM device, either a single bit of a binary value or two bits of an encoded data word encoded from the binary value. By storing two bits of the encoded data word in each CAM cell, data may be stored more efficiently and CAM systems may consume less power as compared to traditional binary CAMS when performing certain types of operations, such as exact matching and longest prefix matching. Encoded data words may be, but need not be, balanced data words which have equal number of logic high and logic low values.

Abstract: Disclosed herein is a method and apparatus for multiple string searching using a ternary content addressable memory. The method includes receiving a text string having a plurality of characters and performing an unanchored search of a database of stored patterns matching one or more characters of the text string using a state machine, wherein the state machine comprises a ternary content addressable memory (CAM) and wherein the performing comprises comparing a state and one of the plurality of characters with contents of a state field and a character field, respectively, stored in the ternary CAM. In the method and apparatus described herein, one or more of the following search features may be supported: exact string matching, inexact string matching, single character wildcard matching, multiple character wildcard matching, case insensitive matching, parallel matching and rollback.

Abstract: Associative memories capable of outputting multiple reference data close to search data are provided. A memory array compares each of the multiple reference data with the search data in parallel and generates multiple comparison current signals representing the result of the comparison. A WLA converts the multiple comparison current signals into voltages. During the first cycle, the WLA detects the lowest voltage among the voltages as Winner and detects the remaining voltages as Loser. After the second cycle, based on feedback signals, the WLA detects all the voltages other than a voltage detected as Winner during the last preceding cycle, and detects the lowest voltage among the detected voltages as Winner and detects the remaining detected voltages as Loser. The WLA repeats these operations k times.

Abstract: For receiving an input data, a pattern data and a data clock signal and outputting a hit signal and an address signal, a content addressable memory includes a plurality of content addressable memory units connected in series, each content addressable memory unit being adapted to receive the input data and the data clock signal and to output a comparison result signal, and an encoder coupled to the comparison result signal of each content addressable memory unit and adapted for outputting a hit signal and a memory address signal subject to the comparison result signal received.