Tokenized streams for concurrent execution between asymmetric multiprocessors

• US 9,542,192 B1
• Filed: 08/15/2008
• Issued: 01/10/2017
• Est. Priority Date: 08/15/2008
• Status: Active Grant

• Current Assignee: NVIDIA Corporation
• Original Assignee: NVIDIA Corporation
• Inventors: Wilt, Nicholas Patrick, Buck, Ian, Cuadra, Philip
• Primary Examiner(s): WU, BENJAMIN C
• Application Number:
  US12/192,912
• Time in Patent Office: 3,070 Days
• Field of Search: 718/101, 717/143, 719/323
• US Class Current: 1/1
• CPC Class Codes: G06F 2209/509 Offload
• CPC Class Codes: G06F 2209/521 Atomic
• CPC Class Codes: G06F 9/30087 Synchronisation or serialisation instructions
• CPC Class Codes: G06F 9/3851 from multiple instruction streams, e.g. multistreaming
• CPC Class Codes: G06F 9/3885 using a plurality of independent parallel functional units
• CPC Class Codes: G06F 9/52 Program synchronisation; Mutual exclusion, e.g. by means of semaphores
• CPC Class Codes: G06F 9/526 Mutual exclusion algorithms

Abstract

A method for executing an application program using streams. A device driver receives a first command within an application program and parses the first command to identify a first stream token that is associated with a first stream. The device driver checks a memory location associated with the first stream for a first semaphore, and determines whether the first semaphore has been released. Once the first semaphore has been released, a second command within the application program is executed. Advantageously, embodiments of the invention provide a technique for developers to take advantage of the parallel execution capabilities of a GPU.

Claims

1. A method for executing application commands associated with a plurality of streams, the method comprising:

receiving from an application program a first command that is associated with a first stream token;

parsing, via a device driver, the first command to identify the first stream token, wherein the first stream token indicates to the device driver that commands associated with the first stream, including the first command, can be executed in parallel with commands associated with another stream across different execution units within a graphics processing unit;

querying a memory location associated with the first stream for a first semaphore;

determining that the first semaphore has not yet been released; and

until the first semaphore has been released, either blocking execution of each command in the application program subsequent to the first command until the graphics processing unit has finished executing all operations associated with the first stream, or waiting until all operations associated with the first stream have finished executing before executing any operations associated with a second stream.

2. The method of claim 1, further comprising:

receiving a second command from the application program, wherein the second command causes the device driver to determine whether the processing unit has finished executing each operation associated with the first stream; and

returning an indication to the application program whether the operations associated with the first stream have finished executing.

3. The method of claim 1, further comprising the steps of:

receiving a second command from the application program;

parsing, via the device driver, the second command to identify a second stream token indicating that the second command is associated with the second stream;

determining that the second stream has a dependency on the first stream; and

waiting until all operations associated with the first stream have finished executing before executing any operations associated with the second stream.

4. The method of claim 3, further comprising the steps of:

releasing the first semaphore when the processing unit has finished executing each operation associated with the first stream; and

acquiring a second semaphore associated with the second stream from a second memory location.

5. The method of claim 3, further comprising the steps of:

receiving a third command from the application program

parsing, via the device driver, the third command to identify a third stream token indicating that the third command is associated with a third stream;

determining that the third stream does not have a dependency on the first stream;

acquiring a third semaphore associated with the third stream; and

transmitting one or more operations associated with the third command to the graphics processing unit for execution.

6. The method of claim 5, wherein the first command and the third command are sub-commands that represent different portions of a fourth command.

7. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to execute application commands associated with a plurality of streams by performing the steps of:

receiving from an application program a first command that is associated with a first stream token;

parsing the first command to identify the first stream token, wherein the first stream token indicates to a device driver that commands associated with the first stream, including the first command, can be executed in parallel with commands associated with another stream across different execution units within a graphics processing unit

querying a memory location associated with the first stream for a first semaphore;

determining that the first semaphore has not yet been released; and

until the first semaphore has been released, either blocking execution of each command in the application program subsequent to the first command until the graphics processing unit has finished executing all operations associated with the first stream, or waiting until all operations associated with the first stream have finished executing before executing any operations associated with a second stream.

8. The non-transitory computer-readable storage medium of claim 7, the steps further comprising:

receiving a second command from the application program, wherein the second command causes the device driver to determine whether the processing unit has finished executing each operation associated with the first stream; and

returning an indication to the application program whether the operations associated with the first stream have finished executing.

9. The non-transitory computer-readable storage medium of claim 7, further comprising the steps of:

receiving a second command from the application program;

parsing the second command to identify a second stream token indicating that the second command is associated with the second stream;

determining that the second stream has a dependency on the first stream; and

waiting until all operations associated with the first stream have finished executing before executing any operations associated with a second.

10. The non-transitory computer-readable storage medium of claim 9, further comprising the steps of:

releasing the first semaphore when the processing unit has finished executing each operation associated with the first stream; and

acquiring a second semaphore associated with the second stream from a second memory location.

11. The non-transitory computer-readable storage medium of claim 9, further comprising the steps of:

receiving a third command from the application program

parsing the third command to identify a third stream token indicating that the third command is associated with a third stream;

determining that the third stream does not have a dependency on the first stream;

acquiring a third semaphore associated with the third stream; and

transmitting one or more operations associated with the third command to the graphics processing unit for execution.

12. The non-transitory computer-readable storage medium of claim 11, wherein the first command and the third command are sub-commands that represent different portions of a fourth command.

13. The non-transitory computer-readable storage medium of claim 7, wherein the graphics processing unit comprises a co-processor within a computing system that includes a central processing unit.

14. A computing system configured to execute application commands from an application program using a plurality of streams, comprising:

a graphics processing unit; and

a system memory coupled to the graphics processing unit and storing a device driver and the application program, wherein the device driver is configured to;

receive from an application program a first command that is associated with a first stream token;

parse the first command to identify the first stream token, wherein the first stream token indicates to the device driver that commands associated with the first stream, including the first command, can be executed in parallel with commands associated with another stream across different execution units within a graphics processing unit;

querying a memory location associated with the first stream for a first semaphore;

determine that the first semaphore has not yet been released; and

until the first semaphore has been released, either block execution of each command in the application program subsequent to the first command until the graphics processing unit has finished executing all operations associated with the first stream, or wait until all operations associated with the first stream have finished executing before executing any operations associated with a second stream.