Abstract: A memory controller, method and program product for memory controller optimization by active re-prioritization of requests based on memory bus activity. The invention continually monitors memory bus activity and, where necessary, re-prioritizes a queue of pending memory requests. In addition, mechanisms are provided to abort an existing request on a memory bus, and interrupt an existing request to complete a pending request.

Abstract: A memory controller, method and program product for memory controller optimization by active re-prioritization of requests based on memory bus activity. The invention continually monitors memory bus activity and, where necessary, re-prioritizes a queue of pending memory requests. In addition, mechanisms are provided to abort an existing request on a memory bus, and interrupt an existing request to complete a pending request.

Abstract: A data transfer engine, method and article of manufacture are provided for use in a functional unit of an integrated system for enhanced access and transfer of data over a shared data bus. The data transfer technique comprises pipelining data transfer requests onto the shared bus. The technique involves receiving and storing a first transfer parameter set in a primary parameter register within the data transfer structure, and receiving and storing a second transfer parameter set in a secondary parameter register within the transfer structure. Thereafter, data is transferred using the shared memory bus in response to the first transfer parameter set and the second transfer parameter set without relinquishing access to the shared memory bus between transferring of data in response to the different parameter sets.

Abstract: Method and apparatus for encoding digital motion video where a motion vector is formed to describe the translation of a set of picture elements from one picture to another. This is accomplished by calculating a difference or prediction error between a current luminance macroblock and a best match reference luminance macroblock, and between a current chrominance macroblock and a best match reference chrominance macroblock. This is done by selecting a chrominance prediction mode from among a set of chrominance prediction modes. Using the selected chrominance prediction mode, a full or half pel interpolation is calculated and from this a difference or prediction error is calculated between a current chrominance macroblock and a best match reference chrominance macroblock. The required reference chrominance data is fetched, and a chrominance difference or prediction error is calculated. The output is the chrominance difference or prediction error data.