IP Library Granted Patent US 12,326,813
Granted Patent B2
US 12,326,813 · App. 18/452,197 · Granted Jun 10, 2025

Heterogeneous architecture, delivered by cxl based cached switch SOC and extensible via cxloverethernet (COE) protocols

Inventors: Shreyas Shah (San Jose, CA); George Apostol, Jr. (Los Gatos, CA); Nagarajan Subramaniyan (San Jose, CA); Jack Regula (Durham, NC); Jeffrey S. Earl (San Jose, CA)
Assignee: Avago Technologies International Sales Pte. Limited
G06F12/0868G06F12/0646G06F12/0815G06F12/0837G06F12/0862G06F12/1466G06F13/1642G06F13/1668G06F13/1673G06F13/4022G06F13/4221G06F2213/0026G06N20/00
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 12,326,813
App. No.
18/452,197
Granted
Jun 10, 2025
Kind
B2
Abstract

Described herein are systems, methods, and products utilizing a cache coherent switch on chip. The cache coherent switch on chip may utilize Compute Express Link (CXL) interconnect open standard and allow for multi-host access and the sharing of resources. The cache coherent switch on chip provides for resource sharing between components while independent of a system processor, removing the system processor as a bottleneck. Cache coherent switch on chip may further allow for cache coherency between various different components. Thus, for example, memories, accelerators, and/or other components within the disclose systems may each maintain caches, and the systems and techniques described herein allow for cache coherency between the different components of the system with minimal latency.

Claims (43)

1. A system comprising:

a first Compute Express Link (CXL) device comprising:

a CXL interface; and

a networking component, wherein the CXL interface is configured to communicate with the networking component over a first software stack via a CXL protocol, and wherein the CXL protocol comprises a L2 layer comprising a configurable size interframe gap (IFG).

2. The system of claim 1 , wherein the networking component comprises a SerDes.

3. The system of claim 2 , wherein the networking component is an off-site SerDes, wherein the networking component and the CXL interface are communicatively coupled via Ethernet, and wherein the CXL protocol comprises a protocol configured to communicatively couple the CXL interface to an off-site SerDes via the L2 layer.

4. The system of claim 2 , wherein the software stack comprises a CXL 3.0 base communications.

5. The system of claim 1 , wherein the networking component comprises a memory prefetcher, and wherein the CXL interface comprises a hierarchy and is configured to:

receive a data indicating that the networking component is communicatively coupled to the CXL interface; and

assign the networking component to a first position within a first hierarchy based on the networking component being communicatively coupled to the CXL interface.

6. The system of claim 1 , wherein the networking component is a first networking component, and wherein the system further comprises:

a second networking component, wherein the second networking component is configured to communicate with the CXL interface via a non-CXL protocol, and wherein the CXL interface is configured to convert the non-CXL protocol to the CXL protocol for communication to the first networking component.

7. The system of claim 6 , wherein the first networking component is a first port, and wherein the second networking component is a second port.

8. The system of claim 6 , wherein the first networking component and the second networking component are communicatively coupled via Ethernet.

9. The system of claim 1 , wherein the CXL protocol further comprises a configurable size preamble.

10. The system of claim 1 , wherein the CXL protocol further comprises one or more of:

a DSP cache read request to a SRAM destination;

a DSP cache read request to a DSP destination;

a DSP cache read response to the SRAM destination;

a DSP cache read response to the DSP destination;

a DSP cache write request; and

a write acknowledgement.

11. A Compute Express Link (CXL) device comprising:

a CXL interface; and

a networking component, wherein the CXL interface is configured to communicate with the networking component over a first software stack via a CXL protocol, and wherein the CXL protocol comprises a L2 layer comprising a configurable size interframe gap (IFG).

12. The CXL device of claim 11 , wherein the networking component comprises a SerDes.

13. The CXL device of claim 12 , wherein the networking component is an off-site SerDes, wherein the networking component and the CXL interface are communicatively coupled via Ethernet, and wherein the CXL protocol comprises a protocol configured to communicatively couple the CXL interface to an off-site SerDes via the L2 layer.

14. The CXL device of claim 12 , wherein the software stack comprises a CXL 3.0 base communications.

15. The CXL device of claim 11 , wherein the networking component comprises a memory prefetcher, and wherein the CXL interface comprises a hierarchy and is configured to:

receive a data indicating that the networking component is communicatively coupled to the CXL interface; and

assign the networking component to a first position within a first hierarchy based on the networking component being communicatively coupled to the CXL interface.

16. The CXL device of claim 11 , wherein the networking component is a first networking component, and wherein the CXL device further comprises:

a second networking component, wherein the second networking component is configured to communicate with the CXL interface via a non-CXL protocol, and wherein the CXL interface is configured to convert the non-CXL protocol to the CXL protocol for communication to the first networking component.

17. The CXL device of claim 16 , wherein the first networking component is a first port, and wherein the second networking component is a second port.

18. The CXL device of claim 16 , wherein the first networking component and the second networking component are communicatively coupled via Ethernet.

19. The CXL device of claim 11 , wherein the CXL protocol further comprises a configurable size preamble.

20. The CXL device of claim 11 , wherein the CXL protocol further comprises one or more of:

a DSP cache read request to a SRAM destination;

a DSP cache read request to a DSP destination;

a DSP cache read response to the SRAM destination;

a DSP cache read response to the DSP destination;

a DSP cache write request; and

a write acknowledgement.

Assignments (2)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Dec 20, 2023
From: SHAH, SHREYAS; APOSTOL, GEORGE, JR.; SUBRAMANIYAN, NAGARAJAN; REGULA, JACK; EARL, JEFFREY S.
To: ELASTICS.CLOUD, INC.
Reel/Frame 066092/0042 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 3, 2023
From: ELASTICS.CLOUD, INC.
To: AVAGO TECHNOLOGIES INTERNATIONAL SALES PTE. LIMITED
Reel/Frame 065104/0547 →
Continuity (3)
Continuation 17809484 · Jun 28, 2022
Provisional Application 63223045 · Jul 18, 2021
Related Publication 20230393997A1 · Dec 7, 2023
References Cited (41)
US 11388268B1 · Siva et al. · 2022 [cited by applicant]
US 11573898B2 · Passint et al. · 2023 [cited by applicant]
US 20150169452A1 · Persson et al. · 2015 [cited by applicant]
US 20160299860A1 · Harriman · 2016 [cited by applicant]
US 20160381176A1 · Cherubini et al. · 2016 [cited by applicant]
US 20190042518A1 · Marolia et al. · 2019 [cited by applicant]
US 20200192798A1 · Natu · 2020 [cited by applicant]
US 20200322287A1 · Connor et al. · 2020 [cited by applicant]
US 20200341930A1 · Cannata et al. · 2020 [cited by applicant]
US 20210011755A1 · Shah · 2021 [cited by applicant]
US 20210075633A1 · Sen et al. · 2021 [cited by applicant]
US 20210117244A1 · Herdrich et al. · 2021 [cited by applicant]
US 20210132999A1 · Haywood et al. · 2021 [cited by applicant]
US 20210240655A1 · Das Sharma · 2021 [cited by applicant]
US 20210311643A1 · Shanbhouge et al. · 2021 [cited by applicant]
US 20210311646A1 · Malladi et al. · 2021 [cited by applicant]
US 20210311739A1 · Malladi et al. · 2021 [cited by applicant]
US 20210318976A1 · Zhang et al. · 2021 [cited by applicant]
US 20210320866A1 · Le et al. · 2021 [cited by applicant]
US 20210374056A1 · Malladi et al. · 2021 [cited by applicant]
US 20210382838A1 · Mittal et al. · 2021 [cited by applicant]
US 20220124038A1 · Leguay et al. · 2022 [cited by applicant]
US 20220147476A1 · Nam et al. · 2022 [cited by applicant]
US 20220164288A1 · Ramagiri et al. · 2022 [cited by applicant]
US 20220292026A1 · Hornung et al. · 2022 [cited by applicant]
US 20220350767A1 · Mcgraw et al. · 2022 [cited by applicant]
US 20220398207A1 · Norrie et al. · 2022 [cited by applicant]
US 20220405212A1 · Kakaiya et al. · 2022 [cited by applicant]
US 20230012822A1 · Shah et al. · 2023 [cited by applicant]
US 20230017583A1 · Shah et al. · 2023 [cited by applicant]
US 20230017643A1 · Shah et al. · 2023 [cited by applicant]
US 20230409302A1 · Kodama et al. · 2023 [cited by applicant]
“FlexPod Datacenter with Citrix VDI and VMware vSphere 7 for up to 2500 Seats”, Cisco, Published Apr. 2022, http:// www.cisco.com/go/designzone, 497 pages. [cited by applicant]
Amir Roozbeh, “Realizing Next-Generation Data Centers via Software-Defined ”Hardware“ Infrastructures and Resource Disaggregation”, Doctoral Thesis KTH Royal Institute of Technology, 227 pages. [cited by applicant]
Davide Giri et al., “NoC-Based Support of Heterogeneous Cache-Coherence Models for Accelerators”, 2018 Twelfth EEE/ACM International Symposium on Networks-on-Chip (NOCS), IEEE Oct. 4, 18, pp. 1-8, Section III and figure… [cited by applicant]
International Search Report on Serial No. PCT/US22/73233, ISR/WO mailed Oct. 14, 2022. [cited by applicant]
Kshitij Bhardwaj et al., “Determining Optimal Coherence Interface for Many-Accelerator SoC's Using Bayesian Optimization”, IEEE Computer Architecture Letters, IEEE Sep. 16, 2019, pp. 119-123 Section 3.1; and figure 2. [cited by applicant]
Kshitij Bhardwaj, et al., “A Comprehensive Methodology to Determine Optimal Coherence Interfaces for Many- Accelerator SoC's”, ISLPED '20 Proceedings of the ACM/IEEE International Symposium on Low Power Electronics %uDB… [cited by applicant]
Prateek Shantharama, et al., “Hardware Accelerated Platforms and Infrastructures for Network Functions: A Survey of Enabling Technologies and Research Studies”. IEEE Jul. 9, 2020, Digital Object Identifier 10.1109/ACCES… [cited by applicant]
V'Akun Sophia Shao, et al. “Co-Designing Accelerators and Soc Interfaces using gem5-Aladdin”, 2016 49th Annual EEE/ACM International Symposium on Microarchitecture (MICRO). IFEE, Oct. 15, 2016, pp. 1-12, pp. 3-5 and fig… [cited by applicant]
Zuckerman, et al. “Cohmeleon: Learning-Based Orchestration of Accelerator Coherence in Heterogeneous SoCs”, Columbia University, New York, New York, arXiv:2109.06382v1 [cs.AR] Sep. 14, 2021, 14 pages. [cited by applicant]