Abstract: A novel scalable-port non-blocking shared-memory output-buffered variable length queued data switching method and apparatus wherein successive data in each of a plurality of queues of data traffic is distributed to corresponding cells of each of successive memory channels in striped fashion across a shared memory space.

Abstract: A sequential data packet addressing technique and system, particularly adapted for shared memory output-buffered switch fabrics and related memories, using a ring of successive subaddress generators each assigning addresses for predetermined size data byte packets received in successive time slots, and creating therefrom super packets ordered based on arrival time; and sequentially allocating memory therefor in the shared memory without overlap among the packets and with no holes between adjacent packets, and assigning addresses for the individual packets in the super packets upon the assigning of an address in the memory for each super packet.

Abstract: A sequential data packet addressing technique and system, particularly adapted for shared memory output-buffered switch fabrics and related memories, using a ring of successive subaddress generators each assigning addresses for predetermined size data byte packets received in successive time slots, and creating therefrom super packets ordered based on arrival time; and sequentially allocating memory therefor in the shared memory without overlap among the packets and with no holes between adjacent packets, and assigning addresses for the individual packets in the super packets upon the assigning of an address in the memory for each super packet.

Abstract: A novel scalable-port non-blocking shared-memory output-buffered variable length queued data switching method and apparatus.

Abstract: A video decoder is provided with robust error handling and concealment. In one embodiment, the video decoder detects syntactic, semantic, and coding errors in encoded slices of macroblocks. An error handler determines the number of remaining un-decoded macroblocks in the corrupted slice and replaces these corrupted macroblocks using substitute DCT coefficient matrices and motion vectors. The zero-frequency DCT coefficient of each substitute matrix is set equal to the zero-frequency DCT coefficient of the last uncorrupted macroblock, while the higher frequency DCT coefficients are set equal to zero. The substitute motion vectors are provided from a concealment vector memory which buffers the motion vectors of the previous macroblock row. In this way, intelligent approximations are made for the missing macroblocks, effectively masking the video bitstream error.

Abstract: A system and method for motion vector extraction and computation is embodied in an architecture adapted to overlap a data extraction process with a computation process and to provide 2-frame store decode with letterbox scaling capability, to extract a plurality of parameters usable for calculating a motion vector, and to compute motion vectors. The architecture is adapted to compute vertical and horizontal components of motion vectors in back-to-back cycles. The architecture includes a motion vector compute pipeline which, in a preferred embodiment includes a delta compute engine, a raw vector compute engine, a motion vector with respect to top left corner of picture block, or a combination of these logic circuits. The delta compute engine is adapted to generate a delta from a motion code and a motion residual and to compute a predicted motion vector in consideration of a motion vector of a previous macroblock.

Abstract: A system and method for performing motion compensation in an MPEG video decoder. The system comprises a horizontal half pixel compensation arrangement including multiple adders and multiplexers which perform horizontal half pixel compensation using an addition function, a division function, and a modulo function on pixel data. The system also includes a register bank which provides the ability to store an array of reference data when vertical half pixel compensation is required. The system also includes a verical half pixel compensation arrangement, which also includes multiple adders and multiplexers which perform vertical half pixel compensation using an addition function, a division function, and a modulo function on pixel data. Reference data and odd pixel data is transferred into and within the system in a predetermined arrangement. Reference and odd pel data may comprise either luma or chroma data.

Abstract: A system and method for low delay mode operation video decoding embodied in a prefetch buffer and an mbcore including an mbcore pipeline. The mbcore is adapted to check a status of the prefetch buffer at predetermined times and to implement a low delay mode to delay the mbcore pipeline when a data level of the prefetch buffer goes below a threshold at the predetermined times. The mbcore is adapted to ensure that there is a sufficient quantity of data in the prefetch buffer for a particular operation and, in a preferred embodiment, is adapted to check the status of the prefetch buffer at a start of a slice, at a beginning of dct decoding of each coded block. The prefetch buffer and the mbcore operate asynchronously, with the mbcore being adapted to prevent a symbol from splitting between the prefetch buffer and the mbcore.

Abstract: A system and method for decoding an MPEG video bitstream comprising several macroblocks of data is disclosed. The system comprises a macroblock core (MBCORE) which processes video bitstream data and computes discrete cosine transform data corresponding to the processed video bitstream, and a parser which parses the video bitstream macroblocks into multiple data blocks used in subsequent stages of decoding. The system further includes a transformation/motion compensation core (TMCCORE) which is divided into multiple stages. The TMCCORE includes an IDCT first stage, an intermediate memory (transpose RAM), and an IDCT second stage. The IDCT first stage passes data to memory and the IDCT second stage receives data from memory. The IDCT first stage has the ability to operate on a first data block while the second stage simultaneously operates on a second data block. The TMCCORE receives the discrete cosine transform data from the MBCORE and calculates and reconstructs a frame therefrom using motion compensation.

Abstract: A novel chip layout for a network wherein pluralities of I/O data ports are each connected to transmit/receive SRAM buffer banks operable under arbitration units to access pluralities of internally cached DRAM banks via internal busses to enable switching data connections amongst all data ports through the appropriate buffers, the chip layout having, data ports substantially symmetrically placed with each data port connected to each arbitration unit and each transmit/receive buffer bank, and with each data port enabled to write into any DRAM bank, with the connections being effected such that each data port is substantially symmetric with respect to DRAM bank, arbitration unit and transmit/receive buffer banks and busses; and with timing clocks centrally placed on the chip to minimize clock skew by symmetric clock distribution.

Abstract: An audio decoder is provided with a programmable and re-configurable downmixing process. In one embodiment, the audio decoder includes a control module and a data path. The data path is configured to read, scale, add, and write audio samples to and from various audio channel frame buffers. The control module implements state diagrams which specify various control signals for directing the operations of the data path. The control module implements state diagrams for directing windowing and downmixing operations. The order in which these operations are performed may be reconfigurable, i.e. downmixing may be performed before or after windowing. This reconfigurability advantageously permits the system designer to trade a slight audio quality enhancement for a decreased memory requirement for some speaker configurations. The downmixing operation requires scaling coefficients which are provided by the control module.

Abstract: An audio decoder is provided with a programmable and re-configurable downmixing process. In one embodiment, the audio decoder includes a control module and a data path. The data path is configured to read, scale, add, and write audio samples to and from various audio channel frame buffers. The control module implements state diagrams which specify various control signals for directing the operations of the data path. The control module implements state diagrams for directing windowing and downmixing operations. The order in which these operations are performed may be reconfigurable, i.e. downmixing may be performed before or after windowing. This reconfigurability advantageously permits the system designer to trade a slight audio quality enhancement for a decreased memory requirement for some speaker configurations. The downmixing operation requires scaling coefficients which are provided by the control module.

Abstract: A system and method for decoding an MPEG video bitstream comprises, comprising a macroblock core (MBCORE) which processes video bitstream data and computes discrete cosine transform data and a parser which parses the video bitstream macroblocks into multiple data blocks used in subsequent stages of decoding. fixed length data words comprising variable length objects using a novel rotating register arrangement. A multistage transformation/motion compensation core (TMCCORE) uses intermediate memory. The IDCT first stage has the ability to operate on a first data block while the second stage simultaneously operates on a second data block. The TMCCORE receives the discrete cosine transform data from the MBCORE and calculates and reconstructs a frame therefrom using motion compensation. The MBCORE can operate on data from a first macroblock while the TMCCORE simultaneously operates on data from a second macroblock.

Abstract: A system and method for decoding fixed length data words comprising variable length objects is disclosed having the ability to decode a variable length DCT in every clock cycle. The system includes multiple floating point registers, preferably two, for holding the fixed length data words, and a tracking arrangement, including a summation block and a total used bits register, where the summation block sums bits used for each variable length object with the contents of the total bits used register to form the total number of used bits. The total used bits are fed back and summed within the total used bits register.The system also has a rotating shift register, which is a circular buffer, and a multiplexer arrangement which transfers variable length objects from the floating point registers to the rotating shift register. The tracking arrangement counts the bits used in transferring variable length objects to the rotating shift register.

Abstract: A computer interconnect system uses packet data transmission over serial links connecting nodes of a network. The serial links provide simultaneous dual paths for transmit/receive. An adapter couples a CPU or the like at a node to the serial link. The adapter includes a packet memory for temporarily storing transmit packets and receive packets, along with a port processor for executing the protocol. Packets of data are transferred between the system bus of the CPU and the packet memory of a pair of data movers, one for read and one for write. The packet memory is accessed upon demand by the serial link, the port processor and the data movers, using interleaved cycles. To accommodate this access upon demand without request/grant cycles, parking registers are provided to store read or write data until a later cycle, and the data rate on the packet memory port is high enough to allow ample time for simultaneous use of both channels as well as packet processing and moving to and from the CPU.