Performance-based workload scheduling in multi-core architectures

• US 7,788,670 B2
• Filed: 10/26/2004
• Issued: 08/31/2010
• Est. Priority Date: 10/26/2004
• Status: Expired due to Fees

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Nakajima, Jun, Bodas, Devadatta V.
• Primary Examiner(s): An; Meng-Ai
• Assistant Examiner(s): KAWSAR, ABDULLAH AL
• Application Number:
  US10/973,707
• Publication Number:
  US 20060090161A1
• Time in Patent Office: 2,135 Days
• Field of Search: 718/100, 718/102, 718/104, 718/108, 718/105, 718/1, 718/101, 718/103, 709223-226, 713/300, 713500-502, 713/320, 713/310, 713/322, 703/22, 702/182, 712/11, 712/12, 712/28, 712/32, 712/203
• US Class Current: 718/105
• CPC Class Codes: G06F 9/5044 considering hardware capabilities
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

Systems and methods of managing workloads provide for detecting a workload for a system having a first processor core with a first performance indicator and a second processor core with a second performance indicator. The workload is scheduled based on the first and second performance indicator settings. In one embodiment, a performance feasibility index is calculated for each core based on the core'"'"'s frequency setting and utilization and the workload is assigned to the core associated with the highest index.

Claims

1. A computer-implemented method comprising:

detecting a workload for a system having a first multi-core processor having a first processor core with a first performance indicator and a second multi-core processor having a second processor core with a second performance indicator, the first processor core sharing a first frequency setting with a first plurality of cores of the first multi-core processor and the second processor core sharing a second frequency setting with a second plurality of cores of the second multi-core processor; and

scheduling the workload to a processor core that corresponds to a highest performance indicator, wherein the first and second performance indicators are calculated by multiplying a determined availability based on a utilization of each processor core by a corresponding frequency setting of each processor core.

2. The method of claim 1, wherein the scheduling includes scheduling the workload based on the corresponding frequency settings, wherein the corresponding frequency settings comprise a first frequency setting and a second frequency setting.

3. The method of claim 2, wherein scheduling the workload includes:

determining the availability of the first processor core based on the utilization of the first processor core;

calculating the first performance indicator based on the availability and the first frequency setting;

repeating the determining and the calculating for the second processor core to obtain the second performance indicator; and

scheduling the workload based on the first and second performance indicators.

4. The method of claim 3, wherein scheduling the workload includes:

determining the highest performance indicator among the first and second performance indicators; and

assigning the workload to a core that corresponds to the highest performance indicator.

5. The method of claim 1, wherein the scheduling includes scheduling the workload based on a first cache size and a second cache size.

6. An apparatus comprising:

a first multi-core processor having a first processor core to have a first performance indicator, the first processor core to share a first frequency setting with a first plurality of cores of the first multi-core processor;

a second multi-core processor having a second processor core to have a second performance indicator, the second processor core to share a second frequency setting with a second plurality of cores of the second multi-core processor; and

scheduling logic to detect a workload for the apparatus and schedule the workload to a processor core that corresponds to a highest performance indicators, wherein the first and second performance indicators are calculated by multiplying a determined availability based on a utilization of each processor core by a corresponding frequency setting of each processor core.

7. The apparatus of claim 6, wherein the first performance indicator is to include a first frequency setting and the second performance indicator is to include a second frequency setting, and wherein the first and second frequency settings comprise the corresponding frequency settings of each processor core.

8. The apparatus of claim 7, wherein the scheduling logic is to determine the availability of the first processor core based on the utilization of the first processor core, calculate the first performance indicator based on the availability and the first frequency setting, repeat the determining and the calculating for the second processor core to obtain the second performance indicator and schedule the workload based on the first and second performance indicators.

9. The apparatus of claim 8, wherein the scheduling logic is to determine the highest performance indicator among the first and second performance indicators and assign the workload to a core that corresponds to the highest performance indicator.

10. The apparatus of claim 6, further including a plurality of counters, each counter being disposed within a processor core to multiply the availability by a corresponding frequency setting.

11. The apparatus of claim 6, wherein the scheduling logic is to schedule the workload based on a first cache size and a second cache size.

12. A system comprising:

a first socket coupled to a first processor, the first processor having a first plurality of cores and a first performance indicator that is shared by the first plurality of cores;

a second socket coupled to a second processor, the second processor having a second plurality of cores and a second performance indicator that is shared by the second plurality of cores; and

scheduling logic to detect a workload for the system and schedule the workload on a processor core that corresponds to a highest performance indicator, wherein the first and second performance indicators are calculated by multiplying a determined availability based on a utilization of each processor core by a corresponding frequency setting of each processor core.

13. The system of claim 12, wherein the first performance indicator is to include a first frequency setting and the second performance indicator is to include a second frequency setting, and wherein the first and second frequency settings comprise the corresponding frequency settings of each processor core.

14. The system of claim 13, wherein the scheduling logic is to determine an availability of each of the first and second plurality of cores based on utilizations of the first and second plurality of cores, calculate performance indicators based on the utilizations and the first and second frequency settings and schedule the workload based on the performance indicators.

15. The system of claim 14, wherein the scheduling logic is to determine the highest performance indicator among the first and second performance indicators and assign the workload to a core that corresponds to the highest performance indicator.

16. The system of claim 12, further including a plurality of counters, each counter being disposed within a core to multiply the availability by a corresponding frequency setting.

17. The system of claim 12, wherein the scheduling logic is to schedule the workload based on a first cache size and a second cache size.

18. An article comprising a memory containing instructions which if executed by one or more processors enable a system to:

detect a workload for a system having a first multi-core processor having a first processor core with a first performance indicator and a second multi-core processor having a second processor core with a second performance indicator, the first processor core sharing a first frequency setting with a first plurality of cores of the first multi-core processor and the second processor core sharing a second frequency setting with a second plurality of cores of the second multi-core processor; and

schedule the workload on a processor core that corresponds to a highest performance indicator, wherein the first and second performance indicators are calculated by multiplying a determined availability based on a utilization of each processor core by a corresponding frequency setting of each processor core.

19. The article of claim 18, comprising instructions that if executed enable the system to schedule the workload based on the corresponding frequency settings, wherein the corresponding frequency settings comprise a first frequency setting and a second frequency setting.

20. The article of claim 19, comprising instructions that if executed enable the system to:

determine the availability of the first processor core based on the utilization of the first processor core;

calculate the first performance indicator based on the availability and the first frequency setting;

repeat the determining and the calculating for the second processor core to obtain the second performance indicator; and

schedule the workload based on the first and second performance indicators.

21. The article of claim 18, comprising instructions that if executed enable the system to:

determine the highest performance indicator among the first and second performance indicators; and

assign the workload to a core that corresponds to the highest performance indicator.

22. The article of claim 18, comprising instructions that if executed enable the system to schedule the workload based on a first cache size and a second cache size.

23. A computer-implemented method comprising:

detecting a workload for a computing system having a first multi-core processor having a first processor core with a first performance indicator that includes a first frequency setting and a second multi-core processor having a second processor core with a second performance indicator that includes a second frequency setting, the first processor core sharing the first frequency setting with a first plurality of cores of the first multi-core processor and the second processor core sharing the second frequency setting with a second plurality of cores of the second multi-core processor;

determining the availability of the first processor core based on the utilization of the first processor core;

multiplying the availability by the first frequency setting to obtain the first performance indicator;

repeating the determining and the multiplying for the second processor core to obtain the second performance indicator;

determining a highest indicator among the first and second performance indicators; and

assigning the workload to a core that corresponds to the highest indicator.

24. The method of claim 23, further including using a plurality of counters to multiply the availabilities by the frequency settings.