A bus bridge which intercepts synchronization events and selectively flushes data in buffers within the bridge is disclosed. The bridge insures data consistency by actively intercepting synchronization events, including, interrupts, processor accesses of a control status registers, and I/O master accesses of shared memory space. Interrupt signals may be routed through the bridge, which includes a bridge control unit comprised of state machine logic for managing data transfers through the bridge. In response to an interrupt signal from an agent on a bus, the bridge control unit flushes posted data before allowing a processor to process the interrupt signal. The bridge control unit further requires that the bridge complete all posted writes generated from a first bus before the bridge accepts a read generated from a second bus. The bridge control unit additionally insures strict ordering of accesses through the bridge. Data consistency is thereby realized without requiring the bridge to participate in the cache coherency protocol of the primary bus.
An interrupt management system that enables a user to handle interrupt events either in a real time mode of operation, or in a batch mode of operation. In the real time mode, an interrupt request signal is asserted in response to each interrupt event. In the batch mode, an interrupt request signal is delayed until a predetermined number of interrupt events is detected, or until a predetermined time interval has elapsed since the last interrupt event is captured. In response to an interrupt event, the corresponding bit in an interrupt register is set to an active state. A control interrupt bit is provided in an interrupt control register for each interrupt to enable the activation of an interrupt request pin in response to the interrupt event. A batch enable bit is provided in a batch register for each interrupt event to enable the batching of the interrupt event.
A method of handling read transactions to improve latency and promote streaming across a bridge. When an initiator returns to retrieve the read data associated with a previously enqueued transaction, the bridge will insert one or more wait states if the enqueued transaction is being mastered and no data has been received in a buffer of the bridge from the target, or if some but not a sufficient amount of data has been received from the target. The bridge continues to hold the initiator until the buffer contains sufficient read data whereupon the bridge will deliver the read data from the buffer to the initiator.
A bridge for a computer system comprising at least a first processing set and a second processing set each connected to the bridge via an I/O bus. A resource control mechanism in the bridge comprises: an interface for exchanging signals with one or more resource slots of a device bus that is capable of being connected to the bridge, each of the resource slots being capable of communicating with a system resource; and a register associated with each system resource, the register having switchable indicia that indicate an operating state of the associated system resource, the control mechanism being operable in use to direct signals to and/or from respective system resources of the computer system.
A technique to provide device status information includes obtaining device status information, determining when a bus retry operation is being executed, and routing the device status information to a bus if a bus retry operation is being executed.
Method and apparatus for tuning the performance of bridge devices, including PCI-to-PCI bridges as well as PCI local bus bridges (or host bridges). The embodiments of the invention permit a multiple-bus computer system to be tuned in view of the application and the bridge queue sizes. Such applications include those concerned with raw bandwidth (such as disk storage), and those that are sensitive to latency (such as networking and videoconferencing). The embodiments of the invention feature a control register that specifies storage conditions to be met by the read and write queues of the bridge. The programmed storage conditions are trigger points which cause the bridge to transfer data into or remove data from the queues during read and write transactions in order to promote the performance (throughput or latency) desired from the bridge.
