摘要 |
An information processing network includes multiple processing devices, a main storage memory, and an interface coupling the processing devices to the main storage memory. All processing devices contend for control of the interface on an equal basis, subject to a dynamically shifting sequence of priority rankings, invoked to resolve contentions for the interface or for one of a plurality of hardware class locks. The class locks are uniquely associated with different capabilities or classes of data operations, which reduces the number of contentions and allows multiple operations to proceed simultaneously. Arbitration logic encompassing all of the processing devices is duplicated in each of the processing devices, and kept coherent through an interconnection of multiple data buses. One bus is associated with each processing device, receives the output of the associated processing device and provides the output to each of the other processing devices. A lock sequencer includes a store-tracking feature that minimizes the time that any processor, after losing a contention for a lock, is in a hold condition. |