IP Library › Granted Patent US 12,248,783
Granted Patent B2
US 12,248,783 · App. 18/369,082 · Granted Mar 11, 2025

Frequency scaling for per-core accelerator assignments

Inventors: Stephen T. Palermo (Chandler, AZ); Srihari Makineni (Portland, OR); Shubha Bommalingaiahnapallya (East Brunswick, NJ); Neelam Chandwani (Portland, OR); Rany T. Elsayed (Folsom, CA); Udayan Mukherjee (Portland, OR); Lokpraveen Mosur (Gilbert, AZ); Adwait Purandare (Beaverton, OR)
Assignee: Intel Corporation
G06F9/30036G06F9/3887G06F9/3888
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,248,783
App. No.
18/369,082
Granted
Mar 11, 2025
Kind
B2
Abstract

Methods for frequency scaling for per-core accelerator assignments and associated apparatus. A processor includes a CPU (central processing unit) having multiple cores that can be selectively configured to support frequency scaling and instruction extensions. Under this approach, some cores can be configured to support a selective set of AVX instructions (such as AVX3/5G-ISA instructions) and/or AMX instructions, while other cores are configured to not support these AVX/AMX instructions. In one aspect, the selective AVX/AMX instructions are implemented in one or more ISA extension units that are separate from the main processor core (or otherwise comprises a separate block of circuitry in a processor core) that can be selectively enabled or disabled. This enables cores having the separate unit(s) disabled to consume less power and/or operate at higher frequencies, while supporting the selective AVX/AMX instructions using other cores. These capabilities enhance performance and provides flexibility to handle a variety of applications requiring use of advanced AVX/AMX instructions to support accelerated workloads.

Claims (41)

1. A processor, comprising:

a plurality of cores to be selectively configured on a per-core basis to operate in a first mode associated with execution of a particular type of instructions of an instruction set and a second mode not associated with the particular type of instructions; and

power control circuitry to:

receive a request to pre-grant licenses to a number of cores of the plurality of cores to operate in the first mode,

determine a first plurality of frequency values at which the number of cores are to operate,

pre-grant the licenses to the number of cores, and

determine a second plurality of frequency values at which other cores of the plurality of cores are to operate,

wherein the first plurality of frequency values and the second plurality of frequency values are based, at least in part, on a maximum power budget supported by the processor.

2. The processor of claim 1 , wherein the first mode is associated with a relatively higher power consumption when executed on a corresponding core and the second mode is associated with a relatively lower power consumption when executed on the corresponding core.

3. The processor of claim 1 , wherein the first plurality of frequency values includes a first base frequency value and a first maximum frequency value and the second plurality of frequency values includes a second base frequency value and a second maximum frequency value.

4. The processor of claim 1 , wherein the number is based on a plurality of workloads to be executed on the plurality of cores.

5. The processor of claim 4 , wherein the plurality of workloads comprise a one or more workloads which include the particular type of instructions and one or more workloads which do not include the particular type of instructions.

6. The processor of claim 5 , wherein the one or more workloads which include the particular type of instructions include at least one of: radio access network (RAN) workloads and media workloads.

7. The processor of claim 6 , wherein the RAN workloads comprise long-term evolution (LTE) and 5G physical layer signal processing workloads.

8. The processor of claim 1 , wherein the particular type of instructions comprises at least one of: AVX-512 instructions and 5G instructions.

9. A method, comprising:

receiving a request to pre-grant licenses to a number of cores of a plurality of cores to operate in a first mode, the first mode associated with execution of a particular type of instructions;

determining a first plurality of frequency values at which the number of cores are to operate;

pre-granting the licenses to the number of cores; and

determining a second plurality of frequency values at which other cores of the plurality of cores are to operate,

wherein the first plurality of frequency values and the second plurality of frequency values are based, at least in part, on a maximum power budget supported by the processor.

10. The method of claim 9 , wherein the first mode is associated with a relatively higher power consumption when executed on a corresponding core and the second mode is associated with a relatively lower power consumption when executed on the corresponding core.

11. The method of claim 9 , wherein the first plurality of frequency values includes a first base frequency value and a first maximum frequency value and the second plurality of frequency values includes a second base frequency value and a second maximum frequency value.

12. The method of claim 9 , wherein the number is based on a plurality of workloads to be executed on the plurality of cores.

13. The method of claim 12 , wherein the plurality of workloads comprise a one or more workloads which include the particular type of instructions and one or more workloads which do not include the particular type of instructions.

14. The method of claim 13 , wherein the one or more workloads which include the particular type of instructions include at least one of: radio access network (RAN) workloads and media workloads.

15. The method of claim 14 , wherein the RAN workloads comprise long-term evolution (LTE) and 5G physical layer signal processing workloads.

16. The method of claim 9 , wherein the particular type of instructions comprises at least one of: AVX-512 instructions and 5G instructions.

17. A non-transitory machine-readable medium having program code stored thereon which, when executed by a machine, causes the machine to perform operations, comprising:

receiving a request to pre-grant licenses to a number of cores of a plurality of cores to operate in a first mode, the first mode associated with execution of a particular type of instructions;

determining a first plurality of frequency values at which the number of cores are to operate;

pre-granting the licenses to the number of cores; and

determining a second plurality of frequency values at which other cores of the plurality of cores are to operate,

wherein the first plurality of frequency values and the second plurality of frequency values are based, at least in part, on a maximum power budget supported by the processor.

18. The non-transitory machine-readable medium of claim 17 , wherein the first mode is associated with a relatively higher power consumption when executed on a corresponding core and the second mode is associated with a relatively lower power consumption when executed on the corresponding core.

19. The non-transitory machine-readable medium of claim 17 , wherein the first plurality of frequency values includes a first base frequency value and a first maximum frequency value and the second plurality of frequency values includes a second base frequency value and a second maximum frequency value.

20. The non-transitory machine-readable medium of claim 17 , wherein the number is based on a plurality of workloads to be executed on the plurality of cores.

21. The non-transitory machine-readable medium of claim 20 , wherein the plurality of workloads comprise a one or more workloads which include the particular type of instructions and one or more workloads which do not include the particular type of instructions.

22. The non-transitory machine-readable medium of claim 21 , wherein the one or more workloads which include the particular type of instructions include at least one of: radio access network (RAN) workloads and media workloads.

23. The non-transitory machine-readable medium of claim 22 , wherein the RAN workloads comprise long-term evolution (LTE) and 5G physical layer signal processing workloads.

24. The non-transitory machine-readable medium of claim 17 , wherein the particular type of instructions comprises at least one of: AVX-512 instructions and 5G instructions.

Continuity (3)
Continuation 16933369 · Jul 20, 2020
Provisional Application 63015083 · Apr 24, 2020
Related Publication 20240111531A1 · Apr 4, 2024
References Cited (51)
US 5461266A · Koreeda · 1995 [cited by examiner]
US 6549999B2 · Kishida · 2003 [cited by examiner]
US 11405946B2 · Babaei · 2022 [cited by examiner]
US 20050193382A1 · Goring · 2005 [cited by examiner]
US 20060174151A1 · Lin · 2006 [cited by examiner]
US 20080072216A1 · Zhao · 2008 [cited by examiner]
US 20080275828A1 · Payton · 2008 [cited by examiner]
US 20080291995A1 · Graham · 2008 [cited by examiner]
US 20110225436A1 · Beard · 2011 [cited by examiner]
US 20120117403A1 · Bieswanger · 2012 [cited by examiner]
US 20130268742A1 · Yamada · 2013 [cited by examiner]
US 20130311758A1 · Caprioli · 2013 [cited by examiner]
US 20150220345A1 · Corbal · 2015 [cited by examiner]
US 20170168546A1 · Meswani · 2017 [cited by examiner]
US 20170242696A1 · Kidd · 2017 [cited by examiner]
US 20180095751A1 · Aminot · 2018 [cited by examiner]
US 20180324979A1 · North · 2018 [cited by examiner]
US 20190004585A1 · Halverson · 2019 [cited by examiner]
US 20190042244A1 · Henry · 2019 [cited by examiner]
US 20190068731A1 · Spicer · 2019 [cited by examiner]
US 20190196827A1 · Madduri · 2019 [cited by examiner]
US 20190347125A1 · Sankaran · 2019 [cited by examiner]
US 20200104126A1 · Pearce · 2020 [cited by examiner]
US 20200220759A1 · Katzav · 2020 [cited by examiner]
US 20200310809A1 · Hughes · 2020 [cited by examiner]
US 20200310872A1 · Jambur Sathyanarayana · 2020 [cited by examiner]
US 20210026708A1 · Weissmann · 2021 [cited by examiner]
US 20210303357A1 · Varma · 2021 [cited by examiner]
US 20210334101A1 · Palermo · 2021 [cited by examiner]
US 20220231701A1 · Furtek · 2022 [cited by examiner]
US 20230068290A1 · Li · 2023 [cited by examiner]
US 20230195199A1 · Chidambaram · 2023 [cited by examiner]
EP 1197847A2 · 2002 [cited by applicant]
Anonymous: “SPARC Assembly/SPARC Instructions—Wikibooks, open books for an open world”, Nov. 5, 2013 (Nov. 5, 2013), pp. 1-3, XP055917843, Retrieved from the Internet: URL:https://en.wikibooks.org/w/index.php?title=SPAR… [cited by applicant]
European First Office Action, (EP Exam Report Article 94(3) EPC), for Patent Application No. 2024598.3, dated May 11, 2022, 8 pages. [cited by applicant]
European Search Report and Search Opinion, EP App. No. 20214598.3, Jun. 16, 2021, 11 pages. [cited by applicant]
European Second Office Action, (EP Exam Report Article 94(3) EPC), for Patent Application No. 2024598.3, dated Sep. 29, 2022, 4 pages. [cited by applicant]
Final Office Action, U.S. Appl. No. 16/933,369, Jan. 14, 2022, 17 pages. [cited by applicant]
Final Office Action, U.S. Appl. No. 16/933,369, Mar. 9, 2023, 22 pages. [cited by applicant]
Ian Matthews, “SOLVED: What is Intel AVX-512?”, Nov. 2, 2017, Urtech.ca (Year: 2017). [cited by applicant]
Non-Final Office Action, U.S. Appl. No. 16/933,369, Aug. 22, 2022, 22 pages. [cited by applicant]
Non-Final Office Action, U.S. Appl. No. 16/933,369, Jul. 21, 2021, 10 pages. [cited by applicant]
Notice of Allowance, U.S. Appl. No. 16/933,369, Jun. 14, 2023, 3 pages. [cited by applicant]
Notice of Allowance, U.S. Appl. No. 16/933,369, May 25, 2023, 10 pages. [cited by applicant]
Office Action, CN App. No. 202011516266.3, Apr. 24, 2023, 2 pages (1 page of English Translation and 1 page of original document). [cited by applicant]
Office Action, EP App. No. 20214598.3, Apr. 5, 2024, 04 pages. [cited by applicant]
Office Action, EP App. No. 20214598.3, May 11, 2022, 8 pages. [cited by applicant]
Office Action, EP App. No. 20214598.3, Nov. 29, 2023, 7 pages. [cited by applicant]
Office Action, EP App. No. 20214598.3, Sep. 29, 2022, 4 pages. [cited by applicant]
Serres et al.: “An Architecture for Reconfigurable Multi-core Explorations”, Reconfigurable Computing and FPGAS (Reconfig), 2011 International Conference On, IEEE, Nov. 30, 2011, 6 pages. [cited by applicant]
Thermal Design Power (TDP) in Intel Processors, Nov. 21, 2019, Intel (Year: 2019). [cited by applicant]