Marvell International Limited

Virage Logic Corporation

Synopsys, Inc.

Firquest LLC

Corigine (hong Kong) Limited

Gevorg Torjyan

Hillel Gazit

Sohail Syed

Yervant Zorian

Karen Darbinyan

Mher Mkhoyan

Albert Harutyunyan

Valery Vardanian

Apparatuses and methods for performing lookup operations are provided. A content addressable memory (CAM) module disposed on a first chip includes one or more content addressable memories. A random access memory (RAM) module disposed on the first chip is configured to receive a memory address from the CAM modules, the memory address being based on a search key received by the CAM module. The RAM module includes one or more random access memories configured to store data entries and counter values or timestamps for respective ones of the data entries. The RAM module also includes logic that is configured to compute the counter values or timestamps. The RAM module is configured to output a data entry corresponding to the search key, the data entry being stored in the one or more random access memories at the memory address, and a counter value or timestamp for the data entry.

Hybrid memory device for lookup operations

Apparatuses and methods for performing lookup operations are provided. A content addressable memory (CAM) module disposed on a first chip includes one or more content addressable memories. A random access memory (RAM) module disposed on the first chip is configured to receive a memory address from the CAM module, the memory address being based on a search key received by the CAM module. The RAM module includes one or more random access memories configured to store data entries and counter values or timestamps for respective ones of the data entries. The RAM module also includes logic that is configured to compute the counter values or timestamps. The RAM module is configured to output a data entry corresponding to the search key, the data entry being stored in the one or more random access memories at the memory address, and a counter value or timestamp for the data entry.

Computer circuitry is provided for fast caching, which includes a memory, a processor, and a cache. The memory stores a data block. The processor retrieves the data block from the memory and determines whether to store the data block in the cache. The cache performs a first hash function on the data block in response to the processor determining to store the data block in the cache. The cache performs a second hash function on the data block if the first hash function results in a collision.

Methods and apparatus for performing fast caching

Systems and techniques relating to processing of network communications include, according to an aspect, a network device that includes circuitry configured to receive value bits selected from a group consisting of a zero bit, a one bit, and a don't care bit; and circuitry configured to store encoded representations of the value bits for use in network packet routing, wherein the encoded representations are position bits selected from a group consisting of a zero bit and a one bit; wherein the circuitry configured to store includes a first memory location and a second memory location that each eliminate a different combination of the value bits from being available for storage respectively in the first memory location and the second memory location.

Apparatus and methods to compress data in a network device and perform ternary content addressable memory (TCAM) processing

Multi-port memory circuitry includes single-port memory circuitry, and arbitration logic circuitry that accepts multiple memory queries for the single-port memory circuitry and prevents the multiple memory queries from addressing conflicting portions of the single-port memory circuitry within a single clock cycle. The arbitration logic circuitry may include conflict-resolution logic circuitry that determines whether multiple memory queries address conflicting portions of the single-port memory circuitry. The single-port memory circuitry may be divided into a plurality of sub-arrays, and the conflict-resolution logic circuitry determines whether the multiple memory queries address overlapping groups of sub-arrays. The single-port memory circuitry may be a content-addressable memory or a random-access memory. The multi-port memory circuitry may be part of a shared-memory, multi-processor apparatus. Each memory query may include a profile identifying at least one portion of the single-port memory circuitry targeted by the query, and the conflict-resolution logic circuitry would examine each profile.

Multi-port memory

The present disclosure describes systems and techniques relating to accessing data stored in Ternary Content Addressable Memory (TCAM). According to an aspect of the described systems and techniques, a device includes: several blocks of TCAM (Ternary Content Addressable Memory); a hash RAM (Random Access Memory); and processor electronics configured to pre-process control records to (i) identify a subset of bits of the control records, giving priority to bits with no X value, (ii) load the hash RAM based on the identified subset of the bits to be used for hashing of search records to find locations in the several blocks of TCAM, and (iii) order the control records in the several blocks of TCAM in accordance with the identified subset of the bits.

Reducing power consumption in ternary content addressable memory (TCAM)

The present disclosure describes systems and techniques relating to processing of network communications. According to an aspect of the described systems and techniques, a network device includes a content addressable memory (CAM) device including random access memory (RAM) devices; and a register configured to store a value for the RAM devices of the CAM device; wherein the CAM device is configured to retrieve data stored in the RAM devices of the CAM device, at a received address offset by the value stored in the register, for comparison to at least a portion of a search string received from a network processor to handle network packet processing.

Network device architecture to support algorithmic content addressable memory (CAM) processing

The present disclosure describes systems and techniques relating to calculation of array statistics. According to an aspect of the described systems and techniques, a device includes: a memory configured to store a data array and a counter array, wherein the data array includes multiple values, and each of the multiple values is encoded in a respective row of the data array, and wherein the counter array includes multiple counters, respective columns of the counter array correspond to respective ones of the counters, and rows of the counter array correspond with bit significance positions spanning the multiple counters; and processor electronics configured to add up a number bits found in respective columns of the data array using respective ones of the multiple counters.

Accelerated calculation of array statistics

A memory hard macro designed to support multiple design for test (DFT) techniques having signal paths associated with the DFT techniques and the functional operation of the memory instance that share logic devices or components. The memory hard macro includes a functional input port and a functional output port, forming a functional memory data path, which includes input latches from the memory instance. The memory hard macro also includes a scan input port and a scan output port, forming a scan data path, which includes input latches from the array of data buffer circuits and output latches from the array of sense amplifiers. The memory hard macro further includes a BIST input port and a BIST output port, forming a BIST data path, which includes at least one input latch from the array of data buffer circuits and at least one output latch from the array of sense amplifiers.

Memory hard macro partition optimization for testing embedded memories

The present disclosure describes systems and techniques relating to processing of network communications. According to an aspect of the described systems and techniques, a network device includes a random access memory (RAM); a buffer coupled with the RAM; circuitry configured to copy data from a location in the RAM to the buffer responsive to a received identifier corresponding to a search key corresponding to a received packet; and circuitry configured to compare the data copied to the buffer with the search key to provide a result for use in forwarding of the packet, wherein don't care bits for the comparison are determined from a count of don't care bits encoded in a portion of the location in the RAM indicated by the identifier.

Apparatus and methods to store data in a network device and perform longest prefix match (LPM) processing

The present disclosure describes systems and techniques relating to processing of network communications. According to an aspect of the described systems and techniques, a network device includes a content addressable memory (CAM); and processing circuitry configured to receive records to be stored in the CAM, compare the records to identify similar bit values at respective bit positions of at least a portion of the records, store in the CAM the similar bit values in a single sample record corresponding to the portion of the records, store in the CAM remaining non-similar bit values of the portion of the records, thereby compressing the portion of the records stored in the CAM, store in the CAM one or more remaining records of the received records not included in the portion of the records, and search the CAM including the compressed portion of the records and the one or more remaining records.

Apparatus and methods to compress data in a network device and perform content addressable memory (CAM) processing

Architecture, system and method for providing compression of repair data in an IC design having a plurality of memory instances. In one embodiment, the repair data storage method includes determining repair data for each of the memory instances and compressing the repair data into a compressed format that is stored in a shared nonvolatile storage common to the memory instances.

Architecture, system and method for compressing repair data in an integrated circuit (IC) design

Various methods and apparatuses are described in which a repair data container may store a concatenated repair signature for multiple memories having one or more redundant components associated with each memory. A processor contains redundancy allocation logic to execute one or more repair algorithms to generate a repair signature for each memory. The repair data container may store actual repair signatures for each memory having one or more defective memory cells detected during fault testing and dummy repair signatures for each memory with no defective memory cells. The processor may contain logic configured to compress an amount of bits making up the concatenated repair signature, to decompress the amount of bits making up the concatenated repair signature, and to compose the concatenated repair signature for all of the memories sharing the repair data container. The repair data container may have an amount of fuses to store the actual repair signatures for an adjustable subset of the multiple memories.

Methods and apparatuses that reduce the size of a repair data container for repairable memories

Methods and apparatuses in which two or more memories share a processor for Built In Self Test algorithms and features are described. The processor initiates a Built In Self Test for the memories. Each memory has an intelligence wrapper bounding that memory. Each intelligence wrapper contains control logic to decode a command from the processor. Each intelligence wrapper contains logic to execute a set of test vectors on a bounded memory. The processor sends a command based self-test to each intelligence wrapper at a first clock speed and the control logic executes the operations associated with that command at a second clock speed asynchronous with the first speed. The processor loads the command containing representations of a march element and data to one or more of the intelligence wrappers via a serial bus.

Growing community of inventors

Fremont, CA, United States of America

Gevorg Torjyan