Intel Corporation

Cabletron Systems, Inc.

Deepak J Aatresh

Gregory Scott Mathews

Sanjay Jain

Michael J Morrison

Nazar A Zaidi

Tosaku Nakanishi

A method and apparatus for forwarding packets from contending queues of a multiport switch to an output of a finite bandwidth involve first prioritizing the contending queues into different priorities that relate to priorities of the packets that are being forwarded in the network. Bandwidth of the output is then allocated among the prioritized contending queues and the bandwidth of the output is consumed by the queued packets according to the allocated proportions. Any unconsumed bandwidth is distributed to the queues on a priority basis such that the highest priority queue is offered the unconsumed bandwidth first and lower priority queues are offered the remaining unconsumed bandwidth in priority order. An advantage of the invention is that queues are not starved of bandwidth by higher priority queues and unconsumed bandwidth is not wasted when there are not enough packets to consume an allocated portion of the output bandwidth. The method is adjustable during normal network operations through a programming interface to provide a specified quality of service.

Method and apparatus for forwarding packets from a plurality of

A microprocessor includes an instruction fetch unit, a branch prediction unit, and a decode unit. The instruction fetch unit is adapted to retrieve a plurality of program instructions. The program instructions include serialization initiating instructions and branch instructions. The branch prediction unit is adapted to generate branch predictions for the branch instructions, direct the instruction fetch unit to retrieve the program instructions in an order corresponding to the branch predictions, and redirect the instruction fetch unit based on a branch misprediction. The branch prediction unit is further adapted to store a redirect address corresponding to the branch misprediction. The decode unit is adapted to decode the program instructions into microcode. The microcode for each of the serialization initiating instructions includes microcode for writing a serialization address of the program instruction following the serialization initiating instruction in the branch prediction unit as the redirect address and triggering the branch misprediction.

System for using a branch prediction unit to achieve serialization by

A clock throttling mechanism turns off certain processor components to minimize power consumption. The processor detects the issuance of certain bus cycles or the execution of certain instructions which typically cause the processor to be idle for a period of time. Control circuitry detects the existence of the instruction and/or bus cycle and shuts down the clock driving certain processor components during that idle period. The control circuitry then detects the occurrence or upcoming occurrence of an event to which the processor responds and becomes active. At detection of this event, the clock signal input to these components is then restarted such that the processor can continue normal execution.

Power management for low power processors through the use of auto

A microprocessor for use in a computer system which pipelines addresses for both burst and non-burst mode data transfers. By pipelining addresses, the microprocessor is able to increase the throughput of data transfers in the system. In the present invention, bits are used which may be programmed to disable and enable the address pipelining for the non-burst mode and burst mode transfers.

Central processing unit address pipelining

An innovative protocol and system for implementing the same enables quick release of the bus by the master device, such as a CPU, to permit slave devices access to the bus. In one embodiment, the arbiter can select between the original hold protocol and quick hold protocol according to predetermined criteria which indicates that a low latency response is requested. Upon assertion of a QHOLD signal, the CPU issues a burst last signal to prematurely terminate outstanding burst transactions on the bus in a manner transparent to the slave devices. Once the outstanding bus cycles are complete, the CPU performs an internal backoff to immediately release the bus for access by the slave device requesting access. Any pending burst cycles which were terminated prematurely by the QHOLD signal, are subsequently restarted for the data not transacted by the CPU after the slave device completes access to the bus. The internal backoff mechanism is similarly transparent to the slave devices and does not cause a backoff signal to be issued to the peripherals or devices coupled to the bus. Thus, the addition of a quick hold protocol is added without significant modification of the slave devices' bus interface.

Growing community of inventors

Sunnyvale, CA, United States of America

Deepak J Aatresh