Job rescheduling using a dynamic number of processors

Inventors: Abhijit Prakash Paithankar (San Jose, CA); Suryanarayan Ramamurthy (Sunnyvale, CA); Arpit Singh (Cupertino, CA)

Assignee: NVIDIA Corporation

G06F11/2028 G06F11/1407

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 12675379
App. No.: 18/640,668
Granted: Jul 7, 2026
Kind: B1

External Links

USPTO Patent Center↗ Full Patent PDF↗

Abstract

Apparatuses, systems, and techniques to adjust a number of processors to be used by software. In at least one embodiment, as a result of a failure of one or more processors of a first group of processors, software performed by this first group of processors to be restarted, using a second smaller group of processors.

Claims (34)

1 . One or more processors, comprising:

circuitry to:

detect a failure of one or more processors of a first group of processors comprising a first number of processors performing a software program;

determine a second smaller number of processors to perform the software program, wherein a size of the second smaller number of processors is within a range defined by the software program;

determine whether a second smaller group of processors comprising the second smaller number of processors is available; and

cause the software program to be performed by the second smaller group of processors in response to determining that the second smaller group of processors is available.

2 . The one or more processors of claim 1 , wherein the circuitry is to determine the second smaller group of processors by excluding one or more failed processors within the first group of processors.

3 . The one or more processors of claim 1 , wherein the range defines a maximum number of processors to be used to perform the software program.

4 . The one or more processors of claim 1 , wherein the range defines a minimum number of processors to be used to perform the software program.

5 . The one or more processors of claim 1 , wherein the size of the second smaller group of processors is determined by decrementing a predetermined number from the first number of processors.

6 . The one or more processors of claim 1 , wherein the software program is restarted to be performed by the second smaller group of processors.

7 . The one or more processors of claim 1 , wherein the software program is reverted to a checkpoint to be performed by the second smaller group of processors.

8 . A method, comprising:

detecting a failure of one or more processors of a first group of processors comprising a first number of processors performing a software program;

determining a second smaller number of processors to perform the software program, wherein a size of the second smaller number of processors is within a range defined by the software program;

determining whether a second smaller group of processors comprising the second smaller number of processors is available; and

causing the software program to be performed by the second smaller group of processors in response to determining that the second smaller group of processors is available.

9 . The method of claim 8 , wherein determining the second smaller group of processors is to be performed to exclude one or more failed processors within the first group of processors.

10 . The method of claim 8 , wherein the range defines a maximum number of processors to be used to perform the software program.

11 . The method of claim 8 , wherein the range defines a minimum number of processors to be used to perform the software program.

12 . The method of claim 8 , wherein the size of the second smaller group of processors is determined by subtracting a predetermined number from the first number of processors.

13 . The method of claim 8 , wherein the software program is restarted to be performed by the second smaller group of processors.

14 . The method of claim 8 , wherein the software program is reverted to a checkpoint to be performed by the second smaller group of processors.

15 . A system, comprising:

one or more processors to:

detect a failure of one or more processors of a first group of processors comprising a first number of processors performing a software program;

determine a second smaller number of processors to perform the software program, wherein a size of the second smaller number of processors is within a range defined by the software program;

determine whether a second smaller group of processors comprising the second smaller number of processors is available; and

cause the software program to be performed by the second smaller group of processors in response to determining that the second smaller group of processors is available.

16 . The system of claim 15 , wherein the one or more processors are to determine the second smaller group of processors by excluding one or more failed processors within the first group of processors.

17 . The system of claim 15 , wherein the range defines a maximum number of processors to be used to perform the software program.

18 . The system of claim 15 , wherein the range defines a minimum number of processors to be used to perform the software program.

19 . The system of claim 15 , wherein the size of the second smaller group of processors is determined by subtracting a predetermined number from the first number of processors.

20 . The system of claim 15 , wherein the software program is restarted to be performed by the second smaller group of processors.