Lsi Corporation

Avago Technologies General IP (Singapore) Pte. Ltd.

Intel Corporation

Agere Systems Inc.

Leonid Dubrovin

Alexander Rabinovitch

Amichay Amitay

Nimrod Alexandron

Eliahou Arviv

Dmitry Podvalny

Hagit Margolin

Noam Abda

An apparatus having a first memory and a circuit is disclosed. The first memory may be configured to store a list having a plurality of read requests. The read requests generally (i) correspond to a plurality of blocks of a reference picture and (ii) are used to decode a current picture in a bitstream carrying video. The circuit may be configured to (i) rearrange the read requests in the list based on at least one of (a) a size of a buffer in a second memory and (b) a width of a data bus of the second memory and (ii) copy a portion of the reference picture from the second memory to a third memory using one or more direct memory access transfers in response to the list.

Adjusting direct memory access transfers used in video decoding

An apparatus having a memory and a circuit is disclosed. The memory may be configured to store a picture being encoded. The circuit may be configured to calculate a plurality of first arrays directly from a plurality of neighboring samples around a current block of the picture. Each first array generally represents a respective one of a plurality of intra-prediction modes. Each first array may be spatially smaller than the current block. The circuit may also be configured to calculate a second array from a plurality of current samples in the current block. The second array may spatially match the first arrays. The circuit may be further configured to generate a plurality of scores of the intra-prediction modes by comparing the first arrays with the second array and select a given one of the intra-prediction modes corresponding to a lowest of the scores to encode the current block.

Intra-prediction mode selection while encoding a picture

An apparatus having a cache and a processor. The cache may be configured to (i) buffer a first subset of reference samples of a reference picture to facilitate a motion estimation of a current block and (ii) prefetch a second subset of the reference samples while a first search pattern is being tested. The first search pattern used in the motion estimation generally defines multiple motion vectors to test. The reference samples of the second subset may be utilized by a second search pattern in the motion estimation of the current block. The prefetch of the second subset may be based on a geometry of the first search pattern and scores of the motion vectors already tested. The processor may be configured to calculate the scores of the motion vectors by a block comparison of the reference samples to the current block according to the first search pattern.

Cache prefetch during motion estimation

An apparatus having a buffer and a circuit is disclosed. The buffer may be configured to store a plurality of fetch sets. Each fetch set generally includes a prefix word and a plurality of instruction words. Each prefix word may include a plurality of symbols. Each symbol generally corresponds to a respective one of the instruction words. The circuit may be configured to (i) identify each of the symbols in each of the fetch sets having a predetermined value and (ii) parse the fetch sets into a plurality of execution sets in response to the symbols having the predetermined value.

Efficient extraction of execution sets from fetch sets

A method of subtracting floating-point numbers includes determining whether a first sign associated with a first floating-point number is unequal to a second sign associated with a second floating-point number, determining whether a first exponent associated with the first floating-point number is less than a second exponent associated with the second floating-point number, negating a first mantissa associated with the first floating-point number when the first sign is unequal to the second sign and determining that the first exponent is less than the second exponent, and adding the first mantissa to a second mantissa associated with the second floating-point number when the first sign is unequal to the second sign and determining that the first exponent is less than the second exponent. Embodiments of a corresponding computer-readable medium and device are also provided.

Method and apparatus to perform floating point operations

A multiplier circuit for generating a product of at least first and second multiplicands includes encoding circuitry comprising a plurality of encoders. Each of the encoders is operative to receive at least a subset of bits of the first multiplicand and to generate a partial product corresponding to the subset of bits of the first multiplicand. The encoding circuitry is further operative to incorporate a negation of the product as a function of at least a first control signal supplied to the multiplier circuit. The multiplier circuit further includes summation circuitry coupled with the encoding circuitry. The summation circuitry is operative to sum each of the partial products generated by the encoding circuitry to thereby generate the product without performing post-incrementation.

Implementation of negation in a multiplication operation without post-incrementation

An apparatus generally having a first circuit and a second circuit is disclosed. The first circuit may be configured to generate (i) a plurality of first code bits in response to an index value and (ii) a plurality of first intermediate bits in response to the index value. The first code bits may be generated in parallel with the first intermediate bits. The second circuit may be configured to generate a plurality of second code bits in response to all of (i) the index value, (ii) the first code bits and (iii) the first intermediate bits. A combination of the first code bits and the second code bits generally forms one of a plurality of orthogonal codes.

Orthogonal variable spreading factor code sequence generation

An apparatus having a memory and a controller is disclosed. The controller may be configured to (i) receive a read request from a processor, the read request comprising a first value and a second value, (ii) where the read request is an indirect memory access, (a) generate a first address in response to the first value, (b) read data stored in the memory at the first address and (c) generate a second address in response to the second value and the data, (iii) where the read request is a direct memory access, generate the second address in response to the second value and (iv) read a requested data stored in the memory at the second address.

Low access time indirect memory accesses

An apparatus generally having a processor, a cache and a circuit is disclosed. The processor may be configured to generate (i) a plurality of access addresses and (ii) a plurality of program counter values corresponding to the access addresses. The cache may be configured to present in response to the access addresses (i) a plurality of data words and (ii) a plurality of address information corresponding to the data words. The circuit may be configured to record a plurality of events in a file in response to a plurality of cache misses. A first of the events in the file due to a first of the cache misses generally includes (i) a first of the program counter values, (ii) a first of the address information and (iii) a first time to prefetch a first of the data word from a memory to the cache.

Cache prefetch learning

An apparatus for performing data caching comprises at least one cache memory including multiple cache lines arranged into multiple segments, each segment having a subset of the cache lines associated therewith. The apparatus further includes a first plurality of counters, each of the counters being operative to track a number of active cache lines associated with a corresponding one of the segments. At least one controller included in the apparatus is operative to receive information relating to the number of active cache lines associated with a corresponding segment from the first plurality of counters and to implement a cache segment replacement policy for determining which of the segments to replace as a function of at least the information relating to the number of active cache lines associated with a corresponding segment.

Cache replacement using active cache line counters

A BIIR system includes a first delay line for receiving at least one input data sample and generating delayed input samples as a function of the input data sample. The BIIR system further includes a second delay line including multiple delay elements connected in series for generating delayed output samples. An input of one of the delay elements receives at least one output data sample of the BIIR system. A summation element in the BIIR system generates the output data sample of the BIIR system as a function of an addition of at least first and second signals and a subtraction of at least a third signal. The third signal includes a first delayed output sample generated by the second delay line multiplied by a first prescribed value. The first delayed output sample and the output data sample are temporally nonadjacent to one another.

Biquad infinite impulse response system transformation

An apparatus having a first cache and a controller is disclosed. The first cache may be configured to assert a first signal after receiving given information in response to being ready to receive additional information. The controller may be configured to (i) fetch the given information from a memory to the first cache and (ii) prefetch first information in a direct memory access transfer from the memory to the first cache in response to the assertion of the first signal.

Copying data to a cache using direct memory access

A data processing system comprises data processing circuitry, a cache memory, and memory access circuitry. The memory access circuitry is operative to assign a memory address region to be allocated in the cache memory with a predefined initialization value. Subsequently, a portion of the cache memory is allocated to the assigned memory address region only after the data processing circuitry first attempts to perform a memory access on a memory address within the assigned memory address region. The allocated portion of the cache memory is then initialized with the predefined initialization value.

On-demand allocation of cache memory for use as a preset buffer

An apparatus having a first circuit and a second circuit is disclosed. The first circuit may be configured to generate a plurality of packed items by extracting-and-packing a plurality of input data words based on a bit mask. The second circuit may be configured to (i) receive the packed items from the first circuit, (ii) sequentially buffer the packed items in a plurality of registers, at least one of the packed items crossing a boundary between a current one of the registers and a next one of the registers, and (iii) write the packed items in the current register to a memory in response to the current register becoming full.

Sequentially packing mask selected bits from plural words in circularly coupled register pair for transferring filled register bits to memory

An apparatus having first and second circuits is disclosed. The first circuit may be disposed on a first side of a bus and configured to store thresholds in a first memory. Each threshold generally represents a respective one of a plurality of regular bit patterns in first data. The first circuit may also be configured to generate second data by representing each respective first data as (i) an index to one of the thresholds and (ii) a difference between the one threshold and the respective first data. A width of the bus may be narrower than the respective first data. The second circuit may be disposed on a second side of the bus and configured to (i) store the thresholds and a plurality of items in a second memory and (ii) reconstruct the first data by adding the respective thresholds to the second data in response to the items.

Growing community of inventors

Karney Shomron, Israel

Leonid Dubrovin