Advanced bandwidth allocation in PCI bus architecture

• US 7,231,475 B1
• Filed: 01/30/2004
• Issued: 06/12/2007
• Est. Priority Date: 01/30/2004
• Status: Active Grant

• Current Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Original Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Inventors: Nakil, Harshad, Singla, Ankur, Reddy, Rajashekar
• Primary Examiner(s): Myers; Paul R.
• Application Number:
  US10/768,924
• Time in Patent Office: 1,229 Days
• Field of Search: 710/117, 710/124, 370/395.21
• US Class Current: 710/117
• CPC Class Codes: G06F 13/3625 using a time dependent access

Abstract

A bus arbitration system and method allocates total bus bandwidth between latency sensitive and latency insensitive interfaces by utilizing windows to divide the total bandwidth into latency sensitive and latency insensitive portions. Each interface is initially allocated top-up numbers of latency sensitive and latency insensitive tokens to proportionally allocate bus accesses between the interfaces according to their requirements. For an interface having access to the bus, the number of tokens is decremented for each successful bus transfer.

Claims

1. A method comprising:

assigning a top-up of number of latency sensitive (LS) tokens associated with each requestor, with the number of LS tokens indicating the LS bandwidth of the bus allocated to an associated requestor;

assigning a top-up of number of latency insensitive (LI) tokens associated with each requestor, with the number of LI tokens indicating the LI bandwidth of the bus allocated to an associated requestor;

setting the widths of alternately occurring LS and LI resource access windows with the width of the LI resource access window not greater than a required LS time interval where the width of the LS resource access window is an allocation of total bus bandwidth to LS accesses and the remaining bus bandwidth is allocated to LI accesses;

applying either an LS resource access window or an LI resource access window to a current arbitration cycle;

granting access to a requestor having the most LS tokens when a number of requestors request access to the bus during a current arbitration cycle in the LS resource access window and decrementing the number of LS tokens assigned to a winning requestor granted access to the bus in proportion to the number of successful bus data transactions completed during the current arbitration cycle; and

granting access to a requestor having the most LI tokens when a number of requestors request access to the bus during a current arbitration cycle in the LI resource access window and decrementing the number of LI tokens assigned to a winning requestor granted access to the bus in proportion to the number of successful bus data transactions completed during the current arbitration cycle.

2. The method of claim 1 further comprising the steps of:

reloading all LS and LI token values of devices to the top-up values if the token numbers assigned to all the devices equal zero.

3. The method of claim 1 further comprising the steps of:

reloading all LS and LI token values of devices to the top-up numbers if the token numbers assigned to all the requestors requesting access to the bus during a current arbitration cycle equal zero.

4. The method of claim 1 further comprising the step of:

reloading all LS and LI token values of devices to the top-up numbers if a set time interval expires.

5. The method of claim 1 where the step of setting the time interval widths comprises the step of:

setting a counter with an LI interval value to indicate a frequency of latency cycles determined by the LI interval value.

6. The method of claim 1 further comprising the step of:

not granting the bus to a particular requestor during a current arbitration cycle if the particular requestor won an immediately preceding arbitration cycle.

7. The method of claim 1 further comprising the step of:

monitoring bus protocol timing signals to determine the number of successful bus data transactions completed during the current arbitration cycle.

8. A system comprising:

a programmable window signal generator that outputs a window control signal indicating whether a current access window is a latency sensitive (LS) or latency insensitive (LI) access window with the duration of an LS interval between occurrences of the successive LS access windows being programmable;

a selector block that receives a number of LS tokens and LI tokens, associated with each requesting device requesting access to the bus during a current arbitration cycle, at first and second input ports respectively, a control input coupled to the programmable window signal generator to receive the window control signal, and an output coupled to the first input port during an LS window and to the second input port during an LI window; and

an arbiter block, coupled to the output port of the selector block, that grants access to the bus to a requesting device having the most tokens and that decrements the number of tokens assigned to a winning device granted access to the bus in proportion to the number of successful bus data transactions completed during the current arbitration cycle.

9. The system of claim 8 where the programmable window signal generator comprises:

a programmable counter that counts arbitration cycles and indicates an LS cycle at a fixed programmable interval.

10. The system of claim 9 where the arbiter block is coupled to receive bus timing signals and monitors the signals to determine the number of successful bus data transactions completed during the current arbitration cycle.

11. The system of claim 8 where the arbiter block receives the grant signal from an immediately preceding arbitration cycle and does not grant the bus to a winner of the previous arbitration cycle.

12. A system comprising:

means for assigning a top-up of number of latency sensitive (LS) tokens associated with each requestor, with the number of LS tokens indicating the LS bandwidth of the bus allocated to an associated requestor;

means for assigning a top-up of number of latency insensitive (LI) tokens associated with each requestor, with the number of LI tokens indicating the LI bandwidth of the bus allocated to an associated requestor;

means for setting the widths of alternately occurring LS and LI resource access windows with the width of the LI resource access window not greater than a required LS time interval where the width of the LS resource access window is an allocation of total bus bandwidth to LS accesses and the remaining bus bandwidth is allocated to LI accesses;

means for applying either an LS resource access window or an LI resource access window to a current arbitration cycle;

means for granting access to a requestor having the most LS tokens when a number of requestors request access to the bus during a current arbitration cycle in the LS resource access window and decrementing the number of LS tokens assigned to a winning requestor granted access to the bus in proportion to the number of successful bus data transactions completed during the current arbitration cycle; and

means for granting access to a requestor having the most LI tokens when a number of requestors request access to the bus during a current arbitration cycle in the LI resource access window and decrementing the number of LI tokens assigned to a winning requestor granted access to the bus in proportion to the number of successful bus data transactions completed during the current arbitration cycle.

13. The system of claim 12 further comprising:

means for reloading all LS and LI token values of devices to the top-up values if the token numbers assigned to all the devices equal zero.

14. The system of claim 12 further comprising:

means for reloading all LS and LI token values of devices to the top-up numbers if the token numbers assigned to all the requestors requesting access to the bus during a current arbitration cycle equal zero.

15. The system of claim 12 further comprising:

means for reloading all LS and LI token values of devices to the top-up numbers if a set time interval expires.

16. The system of claim 12 where the means for setting the time interval widths comprises:

means for setting a counter with an LI interval value to indicate a frequency of latency cycles determined by the LI interval value.

17. The system of claim 12 further comprising:

means for not granting the bus to a particular requestor during a current arbitration cycle if the particular requestor won an immediately preceding arbitration cycle.

18. The system of claim 12 further comprising:

means for monitoring bus protocol timing signals to determine the number of successful bus data transactions completed during the current arbitration cycle.

19. Software embodied on one or more computer readable media when executed operable to:

assign a top-up of number of latency sensitive (LS) tokens associated with each requestor, with the number of LS tokens indicating the LS bandwidth of the bus allocated to an associated requestor;

assign a top-up of number of latency insensitive (LI) tokens associated with each requestor, with the number of LI tokens indicating the LI bandwidth of the bus allocated to an associated requestor;

set the widths of alternately occurring LS and LI access resource windows with the width of the LI resource access window not greater than a required LS time interval, where the width of the LS resource access window is an allocation of total bus bandwidth to LS accesses and the remaining bus bandwidth is allocated to LI accesses;

apply either an LS resource access window or an LI resource access window to a current arbitration cycle;

grant access to a requestor having the most LS tokens when a number of requestors request access to the bus during a current arbitration cycle in the LS resource access window and decrement the number of LS tokens assigned to a winning requestor granted access to the bus in proportion to the number of successful bus data transactions completed during the current arbitration cycle; and

grant access to a requestor having the most LI tokens when a number of requestors request access to the bus during a current arbitration cycle in the LI resource access window and decrement the number of LI tokens assigned to a winning requestor granted access to the bus in proportion to the number of successful bus data transactions completed during the current arbitration cycle.

20. The software of claim 19 when executed further operable to:

reload all LS and LI token values of devices to the top-up values if the token numbers assigned to all the devices equal zero.

21. The software of claim 19 when executed further operable to:

reload all LS and LI token values of devices to the top-up numbers if the token numbers assigned to all the requestors requesting access to the bus during a current arbitration cycle equal zero.

22. The software of claim 19 when executed further operable to:

reload all LS and LI token values of devices to the top-up numbers if a set time interval expires.

23. The software of claim 19 operable to set the time interval widths when executed further operable to:

set a counter with an LI interval value to indicate a frequency of latency cycles determined by the LI interval value.

24. The software of claim 19 further operable to:

not grant the bus to a particular requestor during a current arbitration cycle if the particular requestor won an immediately preceding arbitration cycle.

25. The software of claim 19 further operable to:

monitor bus protocol timing signals to determine the number of successful bus data transactions completed during the current arbitration cycle.