REORDERING OF COMMAND STREAMS FOR GRAPHICAL PROCESSING UNITS (GPUs)

• US 20140184623A1
• Filed: 02/25/2013
• Published: 07/03/2014
• Est. Priority Date: 12/28/2012
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Frascati, Christopher Paul, Seetharamaiah, Avinash
• Granted Patent: US 9,384,522 B2
• Time in Patent Office: Days
• US Class Current: 345/522
• CPC Class Codes: G06T 1/20 Processor architectures; Processor configuration, e.g. pipelining
• CPC Class Codes: G06T 15/005 General purpose rendering architectures
• CPC Class Codes: G09G 5/363 Graphics controllers

Abstract

In general, techniques are described for analyzing a command stream that configures a graphics processing unit (GPU) to render one or more render targets. A device comprising a processor may perform the techniques. The processor may be configured to analyze the command stream to determine a representation of the one or more render targets defined by the command stream. The processor may also be configured to, based on the representation of the render targets, and identify one or more rendering inefficiencies that will occur upon execution of the command stream by the GPU. The processor may also be configured to re-order one or more commands in the command stream so as to reduce the identified rendering inefficiencies that will occur upon execution of the command stream by the GPU.

Claims

1. A method for analyzing a command stream that configures a graphics processing unit (GPU) to render one or more render targets, the method comprising:

analyzing the command stream to determine a representation of the one or more render targets defined by the command stream;

based on the representation of the render targets, identifying one or more rendering inefficiencies that will occur upon execution of the command stream by the GPU; and

re-ordering one or more commands in the command stream so as to reduce the identified rendering inefficiencies that will occur upon execution of the command stream by the GPU.

2. The method of claim 1,

wherein analyzing the command stream comprises:

buffering the command stream; and

generating a graph data structure that represents the one or more render targets as nodes of the graph data structure and any dependencies between the render targets as edges of the graph data structure that interconnect two or more of the nodes of the graph data structure, and

wherein identifying the one or more rendering inefficiencies comprises traversing the graph data structure to identify the one or more rendering inefficiencies that will occur upon execution of the command stream by the GPU.

3. The method of claim 1, further comprising:

based on the representation of the render targets, determining a rendering mode by which to render the render targets with the GPU;

configuring the GPU to operate in the determined rendering mode; and

executing the command stream in the determined rendering mode with the GPU.

4. The method of claim 3, wherein determining a rendering mode comprises determining whether to execute the command stream in a direct rendering mode or a tile-based rendering mode.

5. The method of claim 1, further comprising, when no inefficiencies that will occur upon execution of the command stream by the GPU are identified based on the representation of the one or more render targets, disabling the analysis of the command stream, disabling the identification of the one or more rendering inefficiencies and disabling the re-ordering of the one or more commands in the command stream.

6. The method of claim 5, further comprising:

upon determining that no inefficiencies will occur upon execution of the command stream by the GPU, storing one of a plurality of application profiles that includes data indicating that the analysis of the command stream has been disabled for a corresponding application that generated the command stream, wherein the application profile is associated with the application that generated the command stream, and wherein the command stream comprises a first command stream generated by the application;

receiving a second command stream that is also generated by the application;

accessing the one of the plurality of application profiles associated with the application to determine whether to perform the analysis of the second command stream;

determining, based on the one of the plurality of application profiles, that the analysis of the command stream has been disabled; and

forwarding the second command stream to the GPU without performing the analysis of the second command stream, the identification of the one or more rendering inefficiencies and the re-ordering of the one or more commands in the second command stream.

7. The method of claim 1, further comprising:

forwarding the re-ordered command stream to the GPU;

generating, with the GPU, rendered image data rendered in accordance with the re-ordered command stream; and

presenting the rendered image data.

8. The method of claim 1,

wherein analyzing the command stream comprises analyzing the command stream with a graphics driver to determine the representation of the one or more render targets defined by the command stream during execution of an application,

wherein identifying one or more rendering inefficiencies comprises identifying, based on the representation of the render targets, the one or more rendering inefficiencies with the graphics driver that will occur upon execution of the command stream by the GPU, and

wherein re-ordering the one or more commands comprises re-ordering the one or more commands in the command stream with the graphics driver so as to reduce the identified inefficiencies that will occur upon execution of the command stream by the GPU.

9. The method of claim 1,

wherein analyzing the command stream comprises analyzing the command stream with a graphics library operating in a developer environment to determine the representation of the one or more render targets defined by the command stream during development of an application,

wherein identifying one or more rendering inefficiencies comprises identifying, based on the representation of the render targets, the one or more rendering inefficiencies with the graphics library that will occur upon execution of the command stream by the GPU, and

wherein re-ordering the one or more commands comprises re-ordering the one or more commands in the command stream with the graphics library so as to reduce the identified inefficiencies that will occur upon execution of the command stream by the GPU.

10. The method of claim 1, further comprising generating one or more notifications alerting a developer of the identified one or more rendering inefficiencies.

11. A device configured to analyze a command stream that configures a graphics processing unit (GPU) to render one or more render targets, the device comprising:

one or more processors configured to analyze the command stream to determine a representation of the one or more render targets defined by the command stream, based on the representation of the render targets, identify one or more rendering inefficiencies that will occur upon execution of the command stream by the GPU, and re-order one or more commands in the command stream so as to reduce the identified rendering inefficiencies that will occur upon execution of the command stream by the GPU.

12. The device of claim 11,

wherein the processor is further configured to, when analyzing the command stream, buffer the command stream and generate a graph data structure that represents the one or more render targets as nodes of the graph data structure and any dependencies between the render targets as edges of the graph data structure that interconnect two or more of the nodes of the graph data structure, and

wherein the processor is further configured to, when identifying the one or more rendering inefficiencies, traverse the graph data structure to identify the one or more rendering inefficiencies that will occur upon execution of the command stream by the GPU.

13. The device of claim 11, wherein the processor is further configured to, based on the representation of the render targets, determine a rendering mode by which to render the render targets with the GPU, configure the GPU to operate in the determined rendering mode, and execute the command stream in the determined rendering mode with the GPU.

14. The device of claim 13, wherein the processor is further configured to, when determining a rendering mode, determine whether to execute the command stream in a direct rendering mode or a tile-based rendering mode.

15. The device of claim 11, wherein the processor is further configured to, when no inefficiencies that will occur upon execution of the command stream by the GPU are identified based on the representation of the one or more render targets, disable the analysis of the command stream, disable the identification of the one or more rendering inefficiencies and disable the re-ordering of the one or more commands in the command stream.

16. The device of claim 15,

wherein the processor is further configured to, upon determining that no inefficiencies will occur upon execution of the command stream by the GPU, store one of a plurality of application profiles that includes data indicating that the analysis of the command stream has been disabled for a corresponding application that generated the command stream, wherein the application profile is associated with the application that generated the command stream, and wherein the command stream comprises a first command stream generated by the application,

wherein the processor is further configured to receive a second command stream that is also generated by the application, access the one of the plurality of application profiles associated with the application to determine whether to perform the analysis of the second command stream, determine, based on the one of the plurality of application profiles, that the analysis of the command stream has been disabled, and forward the second command stream to the GPU without performing the analysis of the second command stream, the identification of the one or more rendering inefficiencies and the re-ordering of the one or more commands in the second command stream.

17. The device of claim 11,

wherein the processor is further configured to forward the re-ordered command stream to the GPU, and

wherein the device further includes the GPU that is configured to generate rendered image data rendered in accordance with the re-ordered command stream and present the rendered image data.

18. The device of claim 11,

wherein the processor is further configured to, when analyzing the command stream, analyze the command stream with a graphics driver to determine the representation of the one or more render targets defined by the command stream during execution of an application,

wherein the processor is further configured to, when identifying one or more rendering inefficiencies, identify, based on the representation of the render targets, the one or more rendering inefficiencies with the graphics driver that will occur upon execution of the command stream by the GPU, and

wherein the processor is further configured to, when re-ordering the one or more commands, re-order the one or more commands in the command stream with the graphics driver so as to reduce the identified inefficiencies that will occur upon execution of the command stream by the GPU.

19. The device of claim 11,

wherein the processor is further configured to, when analyzing the command stream, analyze the command stream with a graphics library operating in a developer environment to determine the representation of the one or more render targets defined by the command stream during development of an application,

wherein the processor is further configured to, when identifying one or more rendering inefficiencies, identify, based on the representation of the render targets, the one or more rendering inefficiencies with the graphics library that will occur upon execution of the command stream by the GPU, and

wherein the processor is further configured to, when re-ordering the one or more commands, re-order the one or more commands in the command stream with the graphics library so as to reduce the identified inefficiencies that will occur upon execution of the command stream by the GPU.

20. The device of claim 11, the processor is further configured to generate one or more notifications alerting a developer of the identified one or more rendering inefficiencies.

21. A device configured to analyze a command stream that configures a graphics processing unit (GPU) to render one or more render targets, the device comprising:

means for analyzing the command stream to determine a representation of the one or more render targets defined by the command stream;

means for, based on the representation of the render targets, identifying one or more rendering inefficiencies that will occur upon execution of the command stream by the GPU; and

means for re-ordering one or more commands in the command stream so as to reduce the identified rendering inefficiencies that will occur upon execution of the command stream by the GPU.

22. The device of claim 21,

wherein the means for analyzing the command stream comprises:

means for buffering the command stream; and

means for generating a graph data structure that represents the one or more render targets as nodes of the graph data structure and any dependencies between the render targets as edges of the graph data structure that interconnect two or more of the nodes of the graph data structure, and

wherein the means for identifying the one or more rendering inefficiencies comprises means for traversing the graph data structure to identify the one or more rendering inefficiencies that will occur upon execution of the command stream by the GPU.

23. The device of claim 21, further comprising:

means for determining, based on the representation of the render targets, a rendering mode by which to render the render targets with the GPU;

means for configuring the GPU to operate in the determined rendering mode; and

means for executing the command stream in the determined rendering mode with the GPU.

24. The device of claim 23, wherein the means for determining a rendering mode comprises means for determining whether to execute the command stream in a direct rendering mode or a tile-based rendering mode.

25. The device of claim 21, further comprising means for disabling, when no inefficiencies that will occur upon execution of the command stream by the GPU are identified based on the representation of the one or more render targets, the analysis of the command stream, the identification of the one or more rendering inefficiencies and disabling the re-ordering of the one or more commands in the command stream.

26. The device of claim 25, further comprising:

upon determining that no inefficiencies will occur upon execution of the command stream by the GPU, means for storing one of a plurality of application profiles that includes data indicating that the analysis of the command stream has been disabled for a corresponding application that generated the command stream, wherein the application profile is associated with the application that generated the command stream, and wherein the command stream comprises a first command stream generated by the application;

means for receiving a second command stream that is also generated by the application;

means for accessing the one of the plurality of application profiles associated with the application to determine whether to perform the analysis of the second command stream;

means for determining, based on the one of the plurality of application profiles, that the analysis of the command stream has been disabled; and

means for forwarding the second command stream to the GPU without performing the analysis of the second command stream, the identification of the one or more rendering inefficiencies and the re-ordering of the one or more commands in the second command stream.

27. The device of claim 21, further comprising:

means for forwarding the re-ordered command stream to the GPU;

means for generating rendered image data rendered in accordance with the re-ordered command stream; and

means for presenting the rendered image data.

28. The device of claim 21,

wherein the means for analyzing the command stream comprises means for analyzing the command stream with a graphics driver to determine the representation of the one or more render targets defined by the command stream during execution of an application,

wherein the means for identifying one or more rendering inefficiencies comprises means for identifying, based on the representation of the render targets, the one or more rendering inefficiencies with the graphics driver that will occur upon execution of the command stream by the GPU, and

wherein the means for re-ordering the one or more commands comprises means for re-ordering the one or more commands in the command stream with the graphics driver so as to reduce the identified inefficiencies that will occur upon execution of the command stream by the GPU.

29. The device of claim 21,

wherein the means for analyzing the command stream comprises means for analyzing the command stream with a graphics library operating in a developer environment to determine the representation of the one or more render targets defined by the command stream during development of an application,

wherein the means for identifying one or more rendering inefficiencies comprises means for identifying, based on the representation of the render targets, the one or more rendering inefficiencies with the graphics library that will occur upon execution of the command stream by the GPU, and

wherein the means for re-ordering the one or more commands comprises means for re-ordering the one or more commands in the command stream with the graphics library so as to reduce the identified inefficiencies that will occur upon execution of the command stream by the GPU.

30. The device of claim 21, further comprising means for generating one or more notifications alerting a developer of the identified one or more rendering inefficiencies.

31. A non-transitory computer-readable storage medium having stored thereon instructions that, when executed cause, one or more processors to:

analyze the command stream to determine a representation of the one or more render targets defined by the command stream;

based on the representation of the render targets, identify one or more rendering inefficiencies that will occur upon execution of the command stream by the GPU; and

re-order one or more commands in the command stream so as to reduce the identified rendering inefficiencies that will occur upon execution of the command stream by the GPU.

32. The non-transitory computer-readable storage medium of claim 31,

wherein the instructions that, when executed, cause the one or more processors to analyze the command stream comprises instructions that, when executed, cause the one or processors to:

buffer the command stream; and

generate a graph data structure that represents the one or more render targets as nodes of the graph data structure and any dependencies between the render targets as edges of the graph data structure that interconnect two or more of the nodes of the graph data structure, and

wherein the instructions that, when executed, cause the one or more processors to identify the one or more rendering inefficiencies comprises instructions that, when executed, cause the one or processors to traverse the graph data structure to identify the one or more rendering inefficiencies that will occur upon execution of the command stream by the GPU.

33. The non-transitory computer-readable storage medium of claim 31, further comprising instructions that, when executed, cause the one or more processors to:

based on the representation of the render targets, determine a rendering mode by which to render the render targets with the GPU;

configure the GPU to operate in the determined rendering mode; and

execute the command stream in the determined rendering mode with the GPU.

34. The non-transitory computer-readable storage medium of claim 33, wherein the instructions that, when executed, cause the one or more processors to determine a rendering mode comprises instructions that, when executed, cause the one or processors to determine whether to execute the command stream in a direct rendering mode or a tile-based rendering mode.

35. The non-transitory computer-readable storage medium of claim 31, further comprising instructions that, when executed, cause the one or more processors to, when no inefficiencies that will occur upon execution of the command stream by the GPU are identified based on the representation of the one or more render targets, disable the analysis of the command stream, disable the identification of the one or more rendering inefficiencies and disable the re-ordering of the one or more commands in the command stream.

36. The non-transitory computer-readable storage medium of claim 35, further comprising instructions that, when executed, cause the one or more processors to:

upon determining that no inefficiencies will occur upon execution of the command stream by the GPU, store one of a plurality of application profiles that includes data indicating that the analysis of the command stream has been disabled for a corresponding application that generated the command stream, wherein the application profile is associated with the application that generated the command stream, and wherein the command stream comprises a first command stream generated by the application;

receive a second command stream that is also generated by the application;

access the one of the plurality of application profiles associated with the application to determine whether to perform the analysis of the second command stream;

determine, based on the one of the plurality of application profiles, that the analysis of the command stream has been disabled; and

forward the second command stream to the GPU without performing the analysis of the second command stream, the identification of the one or more rendering inefficiencies and the re-ordering of the one or more commands in the second command stream.

37. The non-transitory computer-readable storage medium of claim 31, further comprising instructions that, when executed, cause the one or more processors to:

forward the re-ordered command stream to the GPU;

generate, with the GPU, rendered image data rendered in accordance with the re-ordered command stream; and

present the rendered image data.

38. The non-transitory computer-readable storage medium of claim 31,

wherein the instructions that, when executed, cause the one or more processors to analyze the command stream comprises instructions that, when executed, cause the one or processors to analyze the command stream with a graphics driver to determine the representation of the one or more render targets defined by the command stream during execution of an application,

wherein the instructions that, when executed, cause the one or more processors to identify one or more rendering inefficiencies comprises instructions that, when executed, cause the one or processors to identify, based on the representation of the render targets, the one or more rendering inefficiencies with the graphics driver that will occur upon execution of the command stream by the GPU, and

wherein the instructions that, when executed, cause the one or more processors to re-order the one or more commands comprises instructions that, when executed, cause the one or processors to re-order the one or more commands in the command stream with the graphics driver so as to reduce the identified inefficiencies that will occur upon execution of the command stream by the GPU.

39. The non-transitory computer-readable storage medium of claim 31,

wherein the instructions that, when executed, cause the one or more processors to analyze the command stream comprises instructions that, when executed, cause the one or processors to analyze the command stream with a graphics library operating in a developer environment to determine the representation of the one or more render targets defined by the command stream during development of an application,

wherein the instructions that, when executed, cause the one or more processors to identify one or more rendering inefficiencies comprises instructions that, when executed, cause the one or processors to identify, based on the representation of the render targets, the one or more rendering inefficiencies with the graphics library that will occur upon execution of the command stream by the GPU, and

wherein the instructions that, when executed, cause the one or more processors to re-order the one or more commands comprises instructions that, when executed, cause the one or processors to re-order the one or more commands in the command stream with the graphics library so as to reduce the identified inefficiencies that will occur upon execution of the command stream by the GPU.

40. The non-transitory computer-readable storage medium of claim 31, further comprising instructions that, when executed, cause the one or more processors to generate one or more notifications alerting a developer of the identified one or more rendering inefficiencies.