Operations related to a retransmission buffer

• US 9,766,681 B2
• Filed: 06/17/2014
• Issued: 09/19/2017
• Est. Priority Date: 06/17/2014
• Status: Expired due to Fees

• Current Assignee: Empire Technology Development LLC (Allied Inventors Management, LLC)
• Original Assignee: Empire Technology Development LLC (Allied Inventors Management, LLC)
• Inventors: Solihin, Yan, Samih, Ahmad
• Primary Examiner(s): Ghowrwal, Omar
• Assistant Examiner(s): BROCKMAN, ANGEL T
• Application Number:
  US14/306,406
• Publication Number:
  US 20150362975A1
• Time in Patent Office: 1,190 Days
• CPC Class Codes: G06F 1/3206 Monitoring of events, devices or parameters that trigger a change in power modality
• CPC Class Codes: G06F 1/324 by lowering clock frequency
• CPC Class Codes: G06F 1/3278 Power saving in modem or I/O interface
• CPC Class Codes: G06F 1/3296 by lowering the supply or operating voltage
• CPC Class Codes: H04L 45/123 Evaluation of link metrics techniques for monitoring network metrics H04L43/08
• CPC Class Codes: H04W 40/10 based on available power or energy
• CPC Class Codes: H04W 40/125 using a measured number of retransmissions as a link metric
• CPC Class Codes: H04W 52/48 during retransmission after error or non-acknowledgment
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management
• CPC Class Codes: Y02D 30/70 in wireless communication networks

Abstract

A method is provided for each router to individually manage retransmissions at run time in a single chip computer die or a single computer that includes cores or compute nodes and routers that interconnect the cores or the compute nodes. Each router compares static energy saving and dynamic energy increase from turning off a retransmission buffer of the router in a monitoring phase. When the static energy saving is greater than the dynamic energy increase, the router turns off the retransmission buffer in a subsequent monitoring phase. When the static energy saving is less than the dynamic energy increase, the router turns on the retransmission buffer in the subsequent monitoring phase.

Claims

1. A method for each router to individually manage retransmissions at run time in a single chip computer die or a single computer that includes cores or compute nodes and routers that interconnect the cores or the compute nodes, the method comprising:

comparing static energy saving and dynamic energy increase from turning off a retransmission buffer of the router in a monitoring phase, wherein the dynamic energy increase comprises a product of an average number of transmissions requested by a downstream router, an average number of hops per retransmission, and an average dynamic energy per hop;

when the static energy saving is greater than the dynamic energy increase in the monitoring phase, turning off the retransmission buffer in a subsequent monitoring phase so that being transmitted to the downstream router in the subsequent monitoring phase is not temporarily stored in the retransmission buffer; and

when the static energy saving is less than the dynamic energy increase in the monitoring phase;

turning on the retransmission buffer in the subsequent monitoring phase, and

temporarily storing, in the retransmission buffer, the data that is transmitted to the downstream router in the subsequent monitoring phase.

2. The method of claim 1, wherein:

the average number of transmissions requested by the downstream router is averaged over multiple monitoring phases including the monitoring phase;

the average number of hops per retransmission is averaged over the multiple monitoring phases including the monitoring phase; and

the average dynamic energy per hop is averaged over the multiple monitoring phases including the monitoring phase.

3. The method of claim 1, wherein each retransmission comprises an end-to-end retransmission from a source router to a destination router and a hop count for each retransmission comprises a number of hops from the source router to the destination router.

4. The method of claim 1, wherein the average dynamic energy per hop in the monitoring phase depends on paths taken, delays experienced, temperatures at the routers, voltages of the routers, currents of the routers, or ages of the routers.

5. The method of claim 1, further comprising, when the retransmission buffer is turned off:

determining whether incoming data has an uncorrectable error; and

when the incoming data has the uncorrectable error, flagging the incoming data to indicate that a hop-by-hop retransmission is not available from an upstream router to the downstream router.

6. The method of claim 1, further comprising:

determining whether incoming data has an uncorrectable error;

when the incoming data has the uncorrectable error, recording a hop count for an end-to-end retransmission of the incoming data;

when the downstream router signals the uncorrectable error in the incoming data, adding the hop count to a running total; and

at an end of the monitoring phase, determining a number of hops per retransmission in the monitoring phase from the running total and a number of the uncorrectable error signals by the downstream router.

7. The method of claim 1, further comprising:

determining whether incoming data has an uncorrectable error;

when the incoming data has the uncorrectable error, determining whether the incoming data is flagged to indicate that a hop-by-hop retransmission is not available from an upstream router to the downstream router;

when the incoming data is not flagged, requesting the hop-by-hop retransmission from the upstream router to the downstream router; and

when the incoming data is flagged, discarding the incoming data and requesting an end-to-end retransmission from a source router.

8. The method of claim 7, further comprising, when the incoming data is flagged, signaling the upstream router that the incoming data has the uncorrectable error.

9. A non-transitory computer-readable storage medium, including a set of computer-readable instructions stored thereon which, in response to execution by a retransmission control logic, cause the retransmission control logic to perform the method of any of claims 1-8.

10. A single chip computer die or a single computer, comprising:

cores or compute nodes; and

routers that interconnects the cores or the compute nodes, each router comprising;

a transmission buffer to a downstream router;

an error detection unit;

a retransmission buffer to the downstream router;

a power gate circuit to turn off the retransmission buffer; and

a retransmission control logic for;

comparing static energy saving and dynamic energy increase from turning off the retransmission buffer of the router in a monitoring phase, wherein the dynamic energy increase comprises a product of an average number of transmissions requested by the downstream router, an average number of hops per retransmission, and an average dynamic energy per hop;

when the static energy saving is greater than the dynamic energy increase in the monitoring phase, turning off the retransmission buffer in a subsequent monitoring phase so the retransmission buffer does not temporarily store data that is transmitted to the downstream router in the subsequent monitoring phase; and

when the static energy saving is less than the dynamic energy increase in the monitoring phase, turning on the retransmission buffer in the subsequent monitoring phase so the retransmission buffer temporarily stores the data that is transmitted to the downstream router in the subsequent monitoring phase.

11. The single chip computer die or the single computer of claim 10, wherein:

the average number of transmissions requested by the downstream router is averaged over multiple monitoring phases including the monitoring phase;

the average number of hops per retransmission is averaged over the multiple monitoring phases including the monitoring phase; and

the average dynamic energy per hop is averaged over the multiple monitoring phases including the monitoring phase.

12. The single chip computer die or the single computer of claim 10, wherein each retransmission comprises an end-to-end retransmission from a source router to a destination router and a hop count for each retransmission comprises a number of hops from the source router to the destination router.

13. The single chip computer die or the single computer of claim 10, wherein the average dynamic energy per hop in the monitoring phase depends on paths taken, delays experienced, temperatures at the routers, voltages of the routers, currents of the routers, or ages of the routers.

14. The single chip computer die or the single computer of claim 10, wherein the retransmission control logic, when the retransmission buffer is turned off:

determines whether incoming data has an uncorrectable error; and

when the incoming data has the uncorrectable error, flags the incoming data to indicate that a hop-by-hop retransmission is not available from an upstream router to the downstream router.

15. The single chip computer die or the single computer of claim 10, wherein the retransmission control logic:

determines whether incoming data has an uncorrectable error;

when the incoming data has the uncorrectable error, records a hop count for an end-to-end retransmission of the data;

when the downstream router signals the uncorrectable error in the incoming data, adds the hop count to a running total; and

at an end of the monitoring phase, determines a number of hops per retransmission in the monitoring phase from the running total and a number of the uncorrectable error signals by the downstream router.

16. The single chip computer die or the single computer of claim 10, wherein the retransmission control logic:

determines whether incoming data has an uncorrectable error;

when the incoming data has the uncorrectable error, determines whether the incoming data is flagged to indicate that a hop-by-hop retransmission is not available from an upstream router to the downstream router;

when the incoming data is not flagged, requests the hop-by-hop retransmission from the upstream router to the downstream router; and

when the incoming data is flagged, discards the incoming data and request an end-to-end retransmission from a source router.

17. The single chip computer die or the single computer of claim 16, wherein the retransmission control logic, when the incoming data is flagged, signals the upstream router that the incoming data has the uncorrectable error.

18. A method of managing power consumption in a router of a network system, the router having a retransmission buffer, the method comprising:

determining a static energy saving from power gating the retransmission buffer;

determining a dynamic energy increase from using an end-to-end retransmission scheme at the router when the retransmission buffer is power gated;

comparing the static energy saving and the dynamic energy increase, wherein the determination of the static energy saving comprises multiplication of a static energy saving per cycle by a number of cycles when the retransmission buffer is power gated;

power gating the retransmission buffer and using the end-to-end retransmission scheme when the static energy saving exceeds the dynamic energy increase; and

powering the retransmission buffer when the dynamic energy increase exceeds the static energy saving.

19. The method of claim 18, wherein powering the retransmission buffer comprises disabling the end-to-end retransmission scheme and using a hop-by-hop retransmission scheme.

20. The method of claim 18, wherein determining the dynamic energy increase includes determining a number of end-to-end retransmissions, and determining the dynamic energy increase using the number of end-to-end retransmissions and an average energy per end-to-end retransmission.

21. The method of claim 18, further comprising:

operating the router in a monitoring mode for a monitoring time, the retransmission buffer being power gated and the router using the end-to-end retransmission for error correction during a monitoring phase;

determining a number of retransmissions during the monitoring phase;

determining an average end-to-end latency as a number of hops during the monitoring phase; and

determining the dynamic energy increase from the number of retransmissions, the average end-to-end latency as the number of hops, and an average energy per hop.

22. The method of claim 21, further comprising operating the router in the monitoring mode at intervals.

23. The method of claim 18, further comprising:

operating the router in a first mode, wherein the retransmission buffer is power gated when the router is in the first mode;

determining a first error rate when the router is in the first mode; and

ending the power gating of the retransmission buffer when the first error rate exceeds a first threshold value.

24. The method of claim 23, further comprising:

operating the router in a second mode, wherein the retransmission buffer is operational when the router is in the second mode;

determining a second error rate when the router is in the second mode; and

power gating the retransmission buffer when the second error rate is less than a second threshold value.

25. A computer system comprising a plurality of routers, at least one router of the plurality of routers comprising:

a transmission buffer;

an error detection unit;

a retransmission buffer;

a power gate circuit to turn off the retransmission buffer; and

a retransmission controller configured to;

compare a static energy saving from turning off the retransmission buffer of the at least one router in a monitoring phase with a dynamic energy increase from using an end-to-end retransmission scheme, wherein the dynamic energy increase comprises a product of an average number of transmissions requested by a downstream router, an average number of hops per retransmission, and an average dynamic energy per hop;

when the static energy saving is greater than the dynamic energy increase in the monitoring phase, turn off the retransmission buffer; and

when the static energy saving is less than the dynamic energy increase in the monitoring phase, turn on the retransmission buffer.

26. The computer system of claim 25, wherein the retransmission controller is configured to determine the dynamic energy increase during the monitoring phase, and wherein the monitoring phase being performed at intervals.

27. The computer system of claim 25, further comprising a single chip computer.

28. The computer system of claim 25, further comprising a multicore processor.

29. The computer system of claim 25, further comprising a plurality of networked compute nodes.

30. The computer system of claim 25, wherein the error detection unit is configured to receives an error signal from the downstream router when there is an uncorrectable error in data transmission to the downstream router.

31. A router comprising:

a transmission buffer;

an error detection unit;

a retransmission buffer;

a power gate circuit operable to turn off the retransmission buffer; and

a retransmission controller configured to;

compare a static energy saving from turning off the retransmission buffer of the router in a monitoring phase with a dynamic energy increase from using an end-to-end retransmission scheme, wherein the dynamic energy increase comprises a product of an average number of transmissions requested by a downstream router, an average number of hops per retransmission, and an average dynamic energy per hop;

when the static energy saving is greater than the dynamic energy increase in the monitoring phase, turn off the retransmission buffer; and

when the static energy saving is less than the dynamic energy increase in the monitoring phase, turn on the retransmission buffer.

32. The router of claim 31, wherein the router is associated with a core of a multicore processor.

33. The router of claim 31, wherein the router is associated with a compute node of a computer system, wherein the computer system includes a plurality of compute nodes including the compute node.