Granted Patent B2

US 12,499,071 · App. 18/680,970 · Granted Dec 16, 2025

Inventors: Francesc Guim Bernat (Barcelona, ES); Da-Ming Chiang (San Jose, CA); Kshitij A. Doshi (Tempe, AZ); Suraj Prabhakaran (Aachen, DE); Mark A. Schmisseur (Phoenix, AZ)

Assignee: Intel Corporation

G06F13/4068G06F9/45533G06F9/5027G06F9/54G06F13/362G06F13/4265G06F13/4282G06N3/02G06N3/04G06N3/045G06N3/08G06F2213/0026

Abstract

Claims (99)

1 . Accelerator hardware for use in a computing platform, the computing platform comprising multiple central processing units (CPUs) and forwarding hardware, the accelerator hardware comprising:

at least one graphics processing unit (GPU) and at least one other GPU, the at least one GPU and the at least one other GPU being configurable to execute respective multiple virtualized instances, the respective multiple virtualized instances being configurable to generate respective data;

wherein:

when the computing platform is in operation:

the at least one GPU is configurable to transfer, via the forwarding hardware, the respective data of a certain one of the respective multiple virtualized instances of the at least one GPU to another certain one of the respective multiple virtualized instances of the at least one other GPU;

data transfer between the at least one GPU and the at least one other GPU is to be in accordance with at least one communication protocol;

the at least one GPU and at least one of the multiple CPUs are configurable for communication with each other in accordance with at least one other communication protocol; and

the at least one communication protocol and the at least one other communication protocol are different from each other, at least in part.

2 . The accelerator hardware of claim 1 , wherein:

the respective data are associated with cloud computing.

3 . The accelerator hardware of claim 1 , wherein:

the computing platform comprises a physical chassis; and

the physical chassis comprises the multiple CPUs and the forwarding hardware.

4 . The accelerator hardware of claim 1 , wherein:

the at least one other communication protocol comprises Peripheral Component Interconnect Express protocol.

5 . At least one non-transitory machine-readable storage medium storing instructions for being executed, at least in part, by accelerator hardware, the accelerator hardware to be used in a computing platform, the computing platform comprising multiple central processing units (CPUs) and forwarding hardware, the accelerator hardware comprising at least one graphics processing unit (GPU) and at least one other GPU, the instructions, when executed, at least in part, by the accelerator hardware, resulting in performance of operations comprising:

configuring the at least one GPU and the at least one other GPU to execute respective multiple virtualized instances, the respective multiple virtualized instances being configurable to generate respective data;

wherein:

when the computing platform is in operation:

the at least one GPU is configurable to transfer, via the forwarding hardware, the respective data of a certain one of the respective multiple virtualized instances of the at least one GPU to another certain one of the respective multiple virtualized instances of the at least one other GPU;

data transfer between the at least one GPU and the at least one other GPU is to be in accordance with at least one communication protocol;

the at least one GPU and at least one of the multiple CPUs are configurable for communication with each other in accordance with at least one other communication protocol; and

the at least one communication protocol and the at least one other communication protocol are different from each other, at least in part.

6 . The at least one non-transitory machine-readable storage medium of claim 5 , wherein:

the respective data are associated with cloud computing.

7 . The at least one non-transitory machine-readable storage medium of claim 5 , wherein:

the computing platform comprises a physical chassis; and

the physical chassis comprises the multiple CPUs and the forwarding hardware.

8 . The at least one non-transitory machine-readable storage medium of claim 5 , wherein:

the at least one other communication protocol comprises Peripheral Component Interconnect Express protocol.

9 . Computing platform comprising:

multiple central processing units (CPUs);

forwarding hardware; and

accelerator hardware comprising at least one graphics processing unit (GPU) and at least one other GPU, the at least one GPU and the at least one other GPU being configurable to execute respective multiple virtualized instances, the respective multiple virtualized instances being configurable to generate respective data;

wherein:

when the computing platform is in operation:

the at least one GPU is configurable to transfer, via the forwarding hardware, the respective data of a certain one of the respective multiple virtualized instances of the at least one GPU to another certain one of the respective multiple virtualized instances of the at least one other GPU;

data transfer between the at least one GPU and the at least one other GPU is to be in accordance with at least one communication protocol;

the at least one GPU and at least one of the multiple CPUs are configurable for communication with each other in accordance with at least one other communication protocol; and

the at least one communication protocol and the at least one other communication protocol are different from each other, at least in part.

10 . The computing platform of claim 9 , wherein:

the respective data are associated with cloud computing.

11 . The computing platform of claim 9 , wherein:

the computing platform comprises a physical chassis; and

the physical chassis comprises the multiple CPUs and the forwarding hardware.

12 . The computing platform of claim 9 , wherein:

the at least one other communication protocol comprises Peripheral Component Interconnect Express protocol.

13 . At least one non-transitory machine-readable storage medium storing instructions for being executed, at least in part, by a computing platform, the computing platform comprising multiple central processing units (CPUs), forwarding hardware, and accelerator hardware, the accelerator hardware comprising at least one graphics processing unit (GPU) and at least one other GPU, the instructions when executed, at least in part, by the computing platform resulting in performance of operations comprising:

configuring the at least one GPU and the at least one other GPU to execute respective multiple virtualized instances, the respective multiple virtualized instances being configurable to generate respective data;

wherein:

when the computing platform is in operation:

the at least one GPU is configurable to transfer, via the forwarding hardware, the respective data of a certain one of the respective multiple virtualized instances of the at least one GPU to another certain one of the respective multiple virtualized instances of the at least one other GPU;

data transfer between the at least one GPU and the at least one other GPU is to be in accordance with at least one communication protocol;

the at least one GPU and at least one of the multiple CPUs are configurable for communication with each other in accordance with at least one other communication protocol; and

the at least one communication protocol and the at least one other communication protocol are different from each other, at least in part.

14 . The at least one non-transitory machine-readable storage medium of claim 13 , wherein:

the respective data are associated with cloud computing.

15 . The at least one non-transitory machine-readable storage medium of claim 13 , wherein:

the computing platform comprises a physical chassis; and

the physical chassis comprises the multiple CPUs and the forwarding hardware.

16 . The at least one non-transitory machine-readable storage medium of claim 13 , wherein:

the at least one other communication protocol comprises Peripheral Component Interconnect Express protocol.

17 . Computing platform for use with accelerator hardware and at least one network, the computing platform being for use in providing, when the computing platform is in operation, at least one service, the computing platform comprising:

at least one central processing unit (CPU) core;

physical accelerator logic;

multiple physical network interface controllers (NICs); and

forwarding hardware;

wherein:

the at least one CPU core, the accelerator hardware, the physical accelerator logic, and the multiple physical NICs are comprised in multiple circuit boards to be communicatively coupled together in the computing platform;

the multiple physical NICs are for use in network traffic communication via the at least one network;

the accelerator hardware comprises accelerator circuitry comprising field programmable gate array circuitry, the accelerator circuitry being configurable to execute multiple virtualized instances for use in association with implementation of neural network-related operations for use in association with the providing of the at least one service; and

when the computing platform is in the operation:

the accelerator circuitry is configurable for communication, using the forwarding hardware and at least one communication protocol, with the physical accelerator logic;

the physical accelerator logic is also configurable for communication, using at least one other communication protocol, with the at least one CPU core; and

the least one communication protocol and the at least one other communication protocol are different from each other, at least in part.

18 . The computing platform of claim 17 , wherein:

the at least one service is associated, at least in part, with cloud computing.

19 . At least one non-transitory machine-readable storage medium storing instructions for being executed, at least in part, by a computing platform, the computing platform to be used with accelerator hardware and at least one network, the computing platform being configurable for use in providing, when the computing platform is in operation, at least one service, the computing platform comprising at least one central processing unit (CPU) core, physical accelerator logic, multiple physical network interface controllers (NICs), and forwarding hardware, the accelerator hardware comprising accelerator circuitry that comprises field programmable gate array circuitry, the instructions when executed, at least in part, by the computing platform resulting in performance of operations comprising:

configuring the accelerator circuitry to execute multiple virtualized instances for use in association with implementation of neural network-related operations for use in association with the providing of the at least one service;

wherein:

the at least one CPU core, the accelerator hardware, the physical accelerator logic, and the multiple physical NICs are comprised in multiple circuit boards to be communicatively coupled together in the computing platform;

the multiple physical NICs are for use in network traffic communication via the at least one network; and

when the computing platform is in the operation:

the accelerator circuitry is configurable for communication, using the forwarding hardware and at least one communication protocol, with the physical accelerator logic;

the physical accelerator logic is also configurable for communication, using at least one other communication protocol, with the at least one CPU core; and

the least one communication protocol and the at least one other communication protocol are different from each other, at least in part.

20 . The at least one non-transitory machine-readable storage medium of claim 19 , wherein:

the at least one service is associated, at least in part, with cloud computing.

21 . A method implemented using a computing platform, the computing platform being for use with accelerator hardware and at least one network, the computing platform being for use in providing, when the computing platform is in operation, at least one service, the computing platform comprising at least one central processing unit (CPU) core, physical accelerator logic, multiple physical network interface controllers (NICs), and forwarding hardware, the accelerator hardware comprising accelerator circuitry that comprises field programmable gate array circuitry, the method comprising:

configuring the accelerator circuitry to execute multiple virtualized instances for use in association with implementation of neural network-related operations for use in association with the providing of the at least one service;

wherein:

the at least one CPU core, the accelerator hardware, the physical accelerator logic, and the multiple physical NICs are comprised in multiple circuit boards to be communicatively coupled together in the computing platform;

the multiple physical NICs are for use in network traffic communication via the at least one network; and

when the computing platform is in the operation:

the accelerator circuitry is configurable for communication, using the forwarding hardware and at least one communication protocol, with the physical accelerator logic;

the physical accelerator logic is also configurable for communication, using at least one other communication protocol, with the at least one CPU core; and

the least one communication protocol and the at least one other communication protocol are different from each other, at least in part.

22 . The method of claim 21 , wherein:

the at least one service is associated, at least in part, with cloud computing.

Continuity (5)

References Cited (42)