Automatic detection of optimal compute unit partitioning

Patent metadata

Field	Value
Publication number	US-10437646-B2
Application number	US-201715430657-A
Country	US
Kind code	B2
Filing date	Feb 13, 2017
Priority date	Feb 13, 2017
Publication date	Oct 8, 2019
Grant date	Oct 8, 2019

Abstract

A system, method, and computer-readable medium are disclosed for performing an automatic partition optimization operation on a processor, comprising: monitoring compute unit usage of the processor over time; determining whether an application executing on the processor comprises a lightly threaded application or a heavily threaded application; and, automatically adjusting compute unit partitioning of the processor based upon whether the application executing on the processor comprises the lightly threaded application or the heavily threaded application.

First claim

What is claimed is:

1. A computer-implementable method for performing an automatic partition optimization operation on a processor, comprising:
monitoring compute unit usage of the processor over time;
determining whether an application executing on the processor comprises a lightly threaded application or a heavily threaded application, a lightly threaded application corresponding to a percentage of processor usage below 30 percent, a heavily threaded application corresponding to the percentage of processor usage above 50 percent;
determining whether the application executing on the processor results in a mixed use compute usage;
providing programmable thresholds for use when adjusting compute unit partitioning, the programmable thresholds providing a tenability function and a hysteresis function when adjusting compute partitioning of processor; and,
automatically adjusting compute unit partitioning of the processor based upon whether the application executing on the processor comprises the lightly threaded application or the heavily threaded application.

2. The method of claim 1 , wherein:
for lighter threaded compute unit usage, the compute units are automatically partitioned to maximize performance by eliminating conflicts due to resource sharing.

3. The method of claim 1 , wherein:
for heavily threaded compute unit usage, the compute units are partitioned to maximize symmetric multi-processing capabilities.

4. The method of claim 1 , wherein:
the determining whether an application executing on the processor comprises a lightly threaded application or a heavily threaded application comprises identifying a percentage of processor usage.

5. The method of claim 4 , wherein:
the percentage of processor usage is analyzed across a plurality of regular time intervals.

6. The method of claim 1 , further comprising:
querying a basic input output system (BIOS) to determine a currently configured compute unit partitioning scheme.

7. A system comprising:
a processor;
a data bus coupled to the processor; and
a non-transitory, computer-readable storage medium embodying computer program code, the non-transitory, computer-readable storage medium being coupled to the data bus, the computer program code interacting with a plurality of computer operations and comprising instructions executable by the processor and configured for:
monitoring compute unit usage of the processor over time;
determining whether an application executing on the processor comprises a lightly threaded application or a heavily threaded application, a lightly threaded application corresponding to a percentage of processor usage below 30 percent, a heavily threaded application corresponding to the percentage of processor usage above 50 percent;
determining whether the application executing on the processor results in a mixed use compute usage;
providing programmable thresholds for use when adjusting compute unit partitioning, the programmable thresholds providing a tenability function and a hysteresis function when adjusting compute partitioning of processor; and,
automatically adjusting compute unit partitioning of the processor based upon whether the application executing on the processor comprises the lightly threaded application or the heavily threaded application.

8. The system of claim 7 , wherein:
for lighter threaded compute unit usage, the compute units are automatically partitioned to maximize performance by eliminating conflicts due to resource sharing.

9. The system of claim 7 , wherein:
for heavily threaded compute unit usage, the compute units are partitioned to maximize symmetric multi-processing capabilities.

10. The system of claim 7 , wherein:
the determining whether an application executing on the processor comprises a lightly threaded application or a heavily threaded application comprises identifying a percentage of processor usage.

11. The system of claim 10 , wherein:
the percentage of processor usage is analyzed across a plurality of regular time intervals.

12. The system of claim 7 , wherein the instructions are further configured for:
querying a basic input output system (BIOS) to determine a currently configured compute unit partitioning scheme.

13. A non-transitory, computer-readable storage medium embodying computer program code, the computer program code comprising computer executable instructions configured for:
monitoring compute unit usage of a processor over time;
determining whether an application executing on the processor comprises a lightly threaded application or a heavily threaded application, a lightly threaded application corresponding to a percentage of processor usage below 30 percent, a heavily threaded application corresponding to the percentage of processor usage above 50 percent;
determining whether the application executing on the processor results in a mixed use compute usage;
providing programmable thresholds for use when adjusting compute unit partitioning, the programmable thresholds providing a tenability function and a hysteresis function when adjusting compute partitioning of processor; and,
automatically adjusting compute unit partitioning of the processor based upon whether the application executing on the processor comprises the lightly threaded application or the heavily threaded application.

14. The non-transitory, computer-readable storage medium of claim 13 , wherein:
for lighter threaded compute unit usage, the compute units are automatically partitioned to maximize performance by eliminating conflicts due to resource sharing.

15. The non-transitory, computer-readable storage medium of claim 13 , wherein:
for heavily threaded compute unit usage, the compute units are partitioned to maximize symmetric multi-processing capabilities.

16. The non-transitory, computer-readable storage medium of claim 13 , wherein:
the determining whether an application executing on the processor comprises a lightly threaded application or a heavily threaded application comprises identifying a percentage of processor usage.

17. The non-transitory, computer-readable storage medium of claim 16 , wherein:
the percentage of processor usage is analyzed across a plurality of regular time intervals.

18. The non-transitory, computer-readable storage medium of claim 13 , wherein the instructions are further configured for:
querying a basic input output system (BIOS) to determine a currently configured compute unit partitioning scheme.

Assignees

• Dell Products Lp

Inventors

• Shows Thomas Alexander
• Morton Charles G

Classifications

• G06F9/5061 Primary

  Partitioning or combining of resources · CPC title

• G06F2209/501

  Performance criteria · CPC title

• G06F2209/508

  Monitor · CPC title