Abstract: Methods, apparatus, and program product are disclosed for use in a computer system in which one or more multiprocessor nodes comprise the computer system. The methods and apparatus provide for configurable allocation of a memory in a node memory controller. In a single node implementation of the computer system, substantially all of the memory is allocated to a snoop directory used to store directory entries for cache lines used by processors in the node. In computer system implementations having more than one node, the amount of the memory allocated to the snoop directory and the amount of the memory allocated to a remote memory directory is controlled respondent to predetermined sizes respondent to the number of nodes in the computer system.

Abstract: Disclosed is a method for and an apparatus using various factors including system performance feedback data to optimize the time suggested to attempt retry of a request. Among the factors used there is included present system performance, type of request, status of pending action, current number of retries pending, a predefined fixed interval, a pseudo random interval, a random interval, past history of retry requests, heuristically determined interval, and an interval based upon hang detection.

Abstract: A data processing system, circuit arrangement, integrated circuit device, program product, and method selectively prefetch a non-cached target memory address for a branch instruction when the target memory address is in a predetermined portion of a memory address space, e.g., within a predetermined number of cache lines from a branch instruction being processed. By prefetching the non-cached target memory addresses for this subclass of branch instructions, the delays associated with retrieving the target memory addresses from higher order memory are minimized. Moreover, by limiting such prefetching to only this subclass of branch instructions, the frequency of retrieval of unneeded data into the cache is often reduced.

Abstract: The method and apparatus for decreasing thread switch latency in a multithread processor stores instructions for an active thread in a primary instruction queue, and stores instructions for a dormant thread in a thread switch instruction queue. The active thread is the thread currently being processed by the multithread processor, while the dormant thread is a thread not currently being executed by the multithread processor. During execution of the active thread, instructions are dispatched from the primary instruction queue for processing. When a thread switch occurs, instructions are dispatched from the thread switch instruction queue for execution. Simultaneously, instructions stored in the thread switch instruction queue are transferred to the primary instruction queue. In this manner, the thread switch latency resulting from the amount of time to refill the primary instruction queue with instructions of the dormant thread is eliminated.