Application-level dispatcher control of application-level pseudo threads and operating system threads

• US 10,360,070 B2
• Filed: 12/22/2017
• Issued: 07/23/2019
• Est. Priority Date: 10/08/2014
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Cadarette, Paul M., Love, Robert D., Willoughby, Austin J.
• Primary Examiner(s): Khatri, Anil
• Application Number:
  US15/853,591
• Publication Number:
  US 20180121246A1
• Time in Patent Office: 578 Days
• Field of Search: 718101-105
• CPC Class Codes: G06F 2209/483 Multiproc
• CPC Class Codes: G06F 2209/5011 Pool
• CPC Class Codes: G06F 2209/5018 Thread allocation
• CPC Class Codes: G06F 9/4881 Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
• CPC Class Codes: G06F 9/505 considering the load
• CPC Class Codes: G06F 9/5055 considering software capabilities, i.e. software resources associated or available to the machine

Abstract

An application-level thread dispatcher that operates in a main full-weight operating system-level thread allocated to an application initializes at least one application-level pseudo threads that operates as an application-controlled thread within the main full-weight operating system-level thread allocated to the application. The application-level thread dispatcher migrates work associated with the application between the at least one application-level pseudo thread and a separate operating system-level thread in accordance with evaluated changes in run-time performance of the application.

Claims

1. A method, comprising:

initializing, by an application-level thread dispatcher that executes on a processor in a main full-weight operating system-level thread allocated to an application,

at least one application-level pseudo thread that operates as an application-controlled thread within the main full-weight operating system-level thread allocated to the application; and

migrating, by the application-level thread dispatcher in accordance with evaluated changes in run-time performance of the application, work associated with the application between the at least one application-level pseudo thread and a separate operating system-level thread.

2. The method of claim 1, further comprising:

establishing, by the application-level thread dispatcher, an application-level threading model under which the application-level thread dispatcher creates and schedules the at least one application-level pseudo thread, where scheduling the at least one application-level pseudo thread comprises providing a portion of overall time an operating system allocates to the main full-weight operating system-level thread to each created one of the at least one application-level pseudo thread.

3. The method of claim 1, where migrating, by the application-level thread dispatcher in accordance with the evaluated changes in the run-time performance of the application, the work associated with the application between the at least one application-level pseudo thread and the separate operating system-level thread comprises:

dispatching, from the main full-weight operating system-level thread by the application-level thread dispatcher in response to determining that the at least one application-level pseudo thread meets configuration requirements to operate within the separate operating system-level thread in parallel with the main full-weight operating system-level thread, the at least one application-level pseudo thread to the separate operating system-level thread.

4. The method of claim 1, where the separate operating system-level thread comprises one of another full-weight operating system-level thread and a separate lighter-weight operating system-level thread.

5. The method of claim 4, where, in response to the separate operating system-level thread comprising the separate lighter-weight operating system-level thread, the separate lighter-weight operating system-level thread is executed on a separate thread execution offload processor.

6. The method of claim 1, where migrating, by the application-level thread dispatcher in accordance with the evaluated changes in the run-time performance of the application, the work associated with the application between the at least one application-level pseudo thread and the separate operating system-level thread comprises:

placing a function pointer to a memory-mapped application function usable to perform the work from the at least one application-level pseudo thread within a work dispatching queue; and

invoking the separate operating system-level thread to process the work dispatching queue, where the function pointer to the memory-mapped application function allows the separate operating system-level thread to directly invoke the memory-mapped application function to perform the work.

7. The method of claim 1, further comprising:

dynamically adjusting, by the application-level thread dispatcher in response to the evaluated changes in the run-time performance of the application, at least one of;

a quantity of full-weight operating system-level threads allocated to the application;

a quantity of application-level pseudo threads allocated to the application; and

a quantity of lighter-weight operating system-level threads allocated to the application.

8. A system, comprising:

a memory; and

a processor programmed to;

initialize, by an application-level thread dispatcher that executes on the processor in a main full-weight operating system-level thread allocated to an application,

at least one application-level pseudo thread that operates as an application-controlled thread within the main full-weight operating system-level thread allocated to the application; and

migrate, by the application-level thread dispatcher in accordance with evaluated changes in run-time performance of the application, work associated with the application between the at least one application-level pseudo thread and a separate operating system-level thread.

9. The system of claim 8, where the processor is further programmed to:

establish, by the application-level thread dispatcher, an application-level threading model under which the application-level thread dispatcher creates and schedules the at least one application-level pseudo thread, where scheduling the at least one application-level pseudo thread comprises providing a portion of overall time an operating system allocates to the main full-weight operating system-level thread to each created one of the at least one application-level pseudo thread.

10. The system of claim 8, where, in being programmed to migrate, by the application-level thread dispatcher in accordance with the evaluated changes in the run-time performance of the application, the work associated with the application between the at least one application-level pseudo thread and the separate operating system-level thread, the processor is programmed to:

dispatch, from the main full-weight operating system-level thread by the application-level thread dispatcher in response to determining that the at least one application-level pseudo thread meets configuration requirements to operate within the separate operating system-level thread in parallel with the main full-weight operating system-level thread, the at least one application-level pseudo thread to the separate operating system-level thread.

11. The system of claim 8, where the separate operating system-level thread comprises one of another full-weight operating system-level thread and a separate lighter-weight operating system-level thread.

12. The system of claim 11, where, in response to the separate operating system-level thread comprising the separate lighter-weight operating system-level thread, the separate lighter-weight operating system-level thread is executed on a separate thread execution offload processor.

13. The system of claim 8, where the processor is further programmed to:

dynamically adjust, by the application-level thread dispatcher in response to the evaluated changes in the run-time performance of the application, at least one of;

a quantity of full-weight operating system-level threads allocated to the application;

a quantity of application-level pseudo threads allocated to the application; and

a quantity of lighter-weight operating system-level threads allocated to the application.

14. A computer program product, comprising:

a computer readable storage medium having computer readable program code embodied therewith, where the computer readable program code when executed on a computer causes the computer to;

initialize, by an application-level thread dispatcher that executes on the computer in a main full-weight operating system-level thread allocated to an application,

at least one application-level pseudo thread that operates as an application-controlled thread within the main full-weight operating system-level thread allocated to the application; and

migrate, by the application-level thread dispatcher in accordance with evaluated changes in run-time performance of the application, work associated with the application between the at least one application-level pseudo thread and a separate operating system-level thread.

15. The computer program product of claim 14, where the computer readable program code when executed on the computer further causes the computer to:

establish, by the application-level thread dispatcher, an application-level threading model under which the application-level thread dispatcher creates and schedules the at least one application-level pseudo thread, where scheduling the at least one application-level pseudo thread comprises providing a portion of overall time an operating system allocates to the main full-weight operating system-level thread to each created one of the at least one application-level pseudo thread.

16. The computer program product of claim 14, where, in causing the computer to migrate, by the application-level thread dispatcher in accordance with the evaluated changes in the run-time performance of the application, the work associated with the application between the at least one application-level pseudo thread and the separate operating system-level thread, the computer readable program code when executed on the computer causes the computer to:

dispatch, from the main full-weight operating system-level thread by the application-level thread dispatcher in response to determining that the at least one application-level pseudo thread meets the configuration requirements to operate within the separate operating system-level thread in parallel with the main full-weight operating system-level thread, the at least one application-level pseudo thread to the separate operating system-level thread.

17. The computer program product of claim 14, where the separate operating system-level thread comprises one of another full-weight operating system-level thread and a separate lighter-weight operating system-level thread.

18. The computer program product of claim 17, where, in response to the separate operating system-level thread comprising the separate lighter-weight operating system-level thread, the separate lighter-weight operating system-level thread is executed on a separate thread execution offload processor.

19. The computer program product of claim 14, where, in causing the computer to migrate, by the application-level thread dispatcher in accordance with the evaluated changes in the run-time performance of the application, the work associated with the application between the at least one application-level pseudo thread and the separate operating system-level thread, the computer readable program code when executed on the computer causes the computer to:

place a function pointer to a memory-mapped application function usable to perform the work from the at least one application-level pseudo thread within a work dispatching queue; and

invoke the separate operating system-level thread to process the work dispatching queue, where the function pointer to the memory-mapped application function allows the separate operating system-level thread to directly invoke the memory-mapped application function to perform the work.

20. The computer program product of claim 14, where the computer readable program code when executed on the computer further causes the computer to:

dynamically adjust, by the application-level thread dispatcher in response to the evaluated changes in the run-time performance of the application, at least one of;

a quantity of full-weight operating system-level threads allocated to the application;

a quantity of application-level pseudo threads allocated to the application; and

a quantity of lighter-weight operating system-level threads allocated to the application.