Optimizing power usage by factoring processor architectural events to PMU

• US 8,966,299 B2
• Filed: 02/18/2014
• Issued: 02/24/2015
• Est. Priority Date: 12/29/2006
• Status: Active Grant

• Current Assignee: Tahoe Research Limited (f/k/a Learndale Limited) (Vector Capital Corporation)
• Original Assignee: Intel Corporation
• Inventors: Liu, Yen-Cheng, Or, P Keong, Sistla, Krishnakanth, Srinivasa, Ganapati
• Primary Examiner(s): Abbaszadeh, Jaweed A
• Application Number:
  US14/182,385
• Publication Number:
  US 20140164799A1
• Time in Patent Office: 371 Days
• Field of Search: 713/300, 713/320, 713/322, 713/323, 713/324, 713/600
• US Class Current: 713/320
• CPC Class Codes: G06F 1/206 comprising thermal management
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode
• 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 1/3243 Power saving in microcontroller unit
• CPC Class Codes: G06F 1/3275 Power saving in memory, e.g. RAM, cache
• CPC Class Codes: G06F 1/3287 by switching off individual functional units in the computer system
• CPC Class Codes: G06F 12/0811 with multilevel cache hierarchies
• CPC Class Codes: G06F 15/80 comprising an array of processing units with common control, e.g. single instruction multiple data processors G06F15/82 takes precedence ; for correlation function computation G06F17/15
• CPC Class Codes: G06F 2212/283 Plural cache memories
• CPC Class Codes: G11C 7/1072 for memories with random access ports synchronised on clock signal pulse trains, e.g. synchronous memories, self timed memories
• CPC Class Codes: G11C 7/1075 for multiport memories each having random access ports and serial ports, e.g. video RAM
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A method and apparatus to monitor architecture events is disclosed. The architecture events are linked together via a push bus mechanism with each architectural event having a designated time slot. There is at least one branch of the push bus in each core. Each branch of the push bus may monitor one core with all the architectural events. All the data collected from the events by the push bus is then sent to a power control unit.

Claims

1. A processor comprising:

a plurality of cores, each core including;

a plurality of thermal sensors to provide thermal data for the core and a plurality of counters each to count a number of occurrences of an architectural event;

a bus to couple the plurality of counters and the plurality of thermal sensors; and

a power control unit coupled to the bus to modify a power state of one of the plurality of cores in response to an occurrence of an architectural event in the one of the plurality of cores.

2. The processor of claim 1, wherein the plurality of counters includes public counters and private counters.

3. The processor of claim 2, wherein the power control unit is to determine whether to modify the power state of the one of the plurality of cores based on the number of occurrences stored in at least one of the plurality of counters.

4. The processor of claim 2, wherein the power control unit is to determine whether to modify the power state of the one of the plurality of cores based on the number of occurrences stored in at least some of the plurality of counters.

5. The processor of claim 4, wherein the at least some counters include an instruction retired counter, a cache miss counter, and a branch misprediction counter.

6. The processor of claim 5, wherein the power control unit is to reduce a frequency of a first core when the number of occurrences stored in the cache miss counter exceeds a threshold.

7. The processor of claim 5, wherein the power control unit is to increase a frequency of the first core when the number of occurrences stored in the instruction retired counter exceeds a threshold.

8. The processor of claim 1, wherein the bus includes a plurality of branches, and a first branch is to monitor the plurality of counters of a first core in a designated time slot.

9. The processor of claim 8, wherein a second branch is to monitor the plurality of thermal sensors of the first core in another designated time slot.

10. The processor of claim 1, wherein the bus is to send data collected from each of the plurality of counters to the power control unit.

11. The processor of claim 1, wherein the power control unit is to manage thermal and power events for a first core, the power control unit further to receive the thermal data from the plurality of thermal sensors.

12. The processor of claim 1, further comprising logic circuitry to initiate a power up sequence responsive to a count of at least one of the plurality of counters and to determine that a turbo mode is available.

13. The processor of claim 12, wherein the logic circuitry is to initiate a power down sequence responsive to a count of at least one of the plurality of counters.

14. A method comprising:

generating a signal to indicate an occurrence of an architectural event at a core of a multicore processor, wherein the architectural event is to be monitored by a channel programmed with information corresponding to the architectural event;

communicating the signal to a power control unit of the multicore processor via a bus shared with thermal sensor circuitry of the multicore processor; and

causing a power sequence based on the occurrence of the architectural event, wherein the power control unit is to determine whether to modify a power state of the core based on a value of one or more of a plurality of counters each corresponding to an architectural event.

15. The method of claim 14, further comprising causing a power up sequence responsive to a first counter value, including powering up the core if reserve power is available, and causing a power down sequence responsive to a second counter value, including powering down the core.

16. The method of claim 14, wherein the plurality of counters includes a counter to indicate a number of instructions retired per cycle, a counter to indicate a number of branch mispredictions during a time period in the core, a counter to indicate a number of translation lookaside buffer (TLB) misses during a time period in the core, and a counter to indicate a number of cache misses during a time period in the core.

17. A system comprising:

a processor including logic circuitry to detect whether an architectural event has occurred within a first core using a channel programmed with information corresponding to the architectural event, a plurality of counters each to count a number of occurrences of one of a plurality of architectural events within the first core, a bus to couple the plurality of counters, and a power control unit to cause a second core to enter into a turbo mode in response to the number of occurrences of one or more of the plurality of counters; and

a dynamic random access memory (DRAM) coupled to the processor.

18. The system of claim 17, wherein the plurality of counters includes a counter to indicate a number of cache misses during a time period in the first core, the number of cache misses to indicate whether a corresponding program is to have access to a higher bandwidth on a system interconnect.

19. The system of claim 17, wherein the power control unit is to further decrease a frequency of the first core if one of the plurality of counters corresponding to cache misses is greater than a corresponding count threshold, to enable the second core to enter into the turbo mode.

20. The system of claim 17, further comprising a plurality of thermal sensors to provide thermal data for the first core, wherein the bus is to couple the plurality of counters and the plurality of thermal sensors, and the power control unit is to make a power control decision for the first core coherently based on values of the at least some of the plurality of counters and the thermal data of at least some of the plurality of thermal sensors.