Abstract: A multiprocessor data processing system includes a memory controller controlling access to a memory subsystem, multiple processor buses coupled to the memory controller, and at least one of multiple processors coupled to each processor bus. In response to receiving a first read request of a first processor via a first processor bus, the memory controller initiates a speculative access to the memory subsystem and a lookup of the target address in a central coherence directory. In response to the central coherence directory indicating that a copy of the target memory block is cached by a second processor, the memory controller transmits a second read request for the target address on a second processor bus. In response to receiving a clean snoop response to the second read request, the memory controller provides to the first processor the target memory block retrieved from the memory subsystem by the speculative access.

Abstract: In a first aspect, a first method is provided for servicing commands. The first method includes the steps of (1) receiving a first command for servicing in a memory controller including a plurality of memory ports, wherein the first command is of a first priority; (2) receiving a second command for servicing in the memory controller, wherein the second command is of a second priority that is higher than the first priority; (3) determining whether the first and second commands will be serviced through the same memory port; and (4) if the first and second commands will not be serviced through the same memory port, servicing the first and second commands during the same time period. Numerous other aspects are provided.

Abstract: A method, computer system, and PCI Express device/protocol for enabling high performance IO data transfers for multiple, different IO configurations, which include variable packet sizes and/or variable/different numbers of transactions on the IO link. PCI Express protocol is enhanced to support utilization of counters and dynamically variable queue sizes. In addition to the standard queue entries, several (or a selected number of) dynamically changeable queue entries are provided/reserved and a dynamic queue modification (DQM) utility is provided within the enhanced PCI Express protocol to monitor ongoing, current data transfer and manage when the size(s) of the queue entries are modified (increased or decreased) based on current data traffic transmitting on the PCI Express IO link. The enhanced PCI Express protocol provides an equilibrium point at which many large data packets are transferred efficiently, while imposing a limit on the number of each size of packets outstanding.

Abstract: In an aspect, a method is provided for using memory. The method includes the steps of (1) employing memory stacking, memory mirroring and memory interleaving in a total memory to reduce a number of memory entries that are written to an input/output (I/O) device while a portion of the total memory is replaced; and (2) storing data in the total memory. Numerous other aspects are provided.