Electrical computers and digital processing systems: multicomput – Computer-to-computer protocol implementing – Computer-to-computer data transfer regulating
Reexamination Certificate
1999-11-09
2003-07-22
An, Meng-Al T. (Department: 2154)
Electrical computers and digital processing systems: multicomput
Computer-to-computer protocol implementing
Computer-to-computer data transfer regulating
C709S232000, C709S227000, C709S229000, C370S230000, C370S259000
Reexamination Certificate
active
06598086
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Technical Field
The present invention relates in general to high frequency digital systems and, in particular, to controlling the flow of information between a producer and a buffering consumer in a high frequency digital system. Still more particularly, the present invention relates to a method and system for controlling the flow of information in a high frequency digital system from a memoriless producer to a buffering consumer via an intermediate buffer.
2. Description of the Related Art
As processor micro-architectures are optimized toward implementations that support higher clock frequencies, the complexity of the work that can be performed within each clock cycle decreases. This phenomenon has a tendency to introduce increased cycle-time latency into mechanisms which control the flow of information between components in the micro-architecture, effectively delaying critical feedback within such mechanisms and eroding the aggregate bandwidth of the information flow.
This bandwidth erosion occurs when the buffering capacity of downstream components approaches full occupancy, such that the fill time of the remaining To available capacity approaches the latency of feedback in the information flow control mechanism. In such cases, the information flow control mechanism often must pessimistically assume worst case information flow to avoid exceeding the buffering capacity of the downstream components. Thus, for example, the information flow control mechanism may assume that information is being sent (and stall information flow appropriately based on downstream capacity considerations) whether or not information is actually sent in order to avoid overflowing the downstream buffers.
In many cases, such bandwidth erosion can be averted by increasing the buffering capacity of downstream components to minimize the likelihood that high utilizations will dictate precautionary stalls. However, an increase in buffering capacity also raises component cost without providing additional benefit beyond a reduction in bandwidth erosion. In other cases, more sophisticated flow protocols can be introduced, which tolerate information loss due to aggressive capacity speculation by incorporating retry mechanisms. However, such mechanisms can greatly increase complexity, thus inhibiting design verification and testability. Moreover, retry protocols also increase the utilization of the upstream component, possibly requiring an increase in its size and cost.
The present invention offers an attractive alternative for addressing the obstacles to efficient information flow that arise in high frequency digital systems that control, monitor, or perform transformations on streaming information.
SUMMARY OF THE INVENTION
In accordance with the present invention, an information handling system, such as a processor, includes a plurality of sequentially connected units including a first unit, a second unit and a third unit. Packets of information flow from the first unit directly to the second unit and then to the third unit, and each of the plurality of units provides a respective dynamic output indication indicating if that unit output a packet. The information handling system further includes a control unit that determines, utilizing all of the plurality of dynamic output indications, packet buffering capacities of the plurality of units, and guaranteed packet flows between adjacent ones of the plurality of units, if the first unit can output a packet directly to the second unit without packet loss. In response to this determination, the control unit outputs a control signal to the first unit.
All objects, features, and advantages of the present invention will become apparent in the following detailed written description.
REFERENCES:
patent: 5291481 (1994-03-01), Doshi et al.
patent: 5392280 (1995-02-01), Zheng
patent: 5646943 (1997-07-01), Elwalid
patent: 5650994 (1997-07-01), Daley
patent: 5898670 (1999-04-01), Hoebeke et al.
patent: 5938749 (1999-08-01), Rusu et al.
patent: 5996013 (1999-11-01), Delp et al.
patent: 6052738 (2000-04-01), Muller et al.
patent: 6259698 (2001-07-01), Shin et al.
patent: 6438104 (2002-08-01), Fodor et al.
Sakurai et al., “Large-Scale ATM Multistage Switching Network with Shared Buffer Memory Switches,” IEEE Communications, vol. 29, No. 1 pp. 90-96 (Jan. 1991).*
Fong et al., “Performance analysis of shared buffer ATM switch with different cell departure models”, Performance, Computing and Communications, 1998. IPCCC '98., IEEE International , 1998 pp.: 386-392.
Bell, Jr. Robert Henry
Cargnoni Robert Alan
Clark Leo James
Starke William John
An Meng-Al T.
Bracewell & Patterson L.L.P.
Emile Volel
International Business Machines - Corporation
Lin Wen-Tai
LandOfFree
Method and system for controlling information flow in a high... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for controlling information flow in a high..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for controlling information flow in a high... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3045047