Fast reconfiguration of graphics pipeline state

• US 8,111,260 B2
• Filed: 06/28/2006
• Issued: 02/07/2012
• Est. Priority Date: 06/28/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Srinivasan, Ramanujan, Markovic, Relja, Glassenberg, Samuel
• Primary Examiner(s): Tung, Kee M
• Assistant Examiner(s): Perromat, Carlos
• Application Number:
  US11/477,272
• Publication Number:
  US 20080001952A1
• Time in Patent Office: 2,050 Days
• Field of Search: 345/543
• US Class Current: 345/543
• CPC Class Codes: G06T 1/20 Processor architectures; Processor configuration, e.g. pipelining

Abstract

Techniques and technologies are provided for binding resources to particular slots associated with shaders in a graphics pipeline. Resource dependencies between resources being utilized by each shader can be determined, and, based on these resource dependencies, common resource/slot associations can be computed. Each common resource/slot association identifies a particular one of the resources to be associated with a particular one of the slots.

Claims

1. A method for assigning resources used by a plurality of shaders of a graphics pipeline to particular slots of a processing unit, comprising:

identifying a particular resource/slot arrangement of a first shader;

identifying a particular resource/slot arrangement of a second shader;

identifying a particular resource/slot arrangement of a third shader;

determining a common resource/slot arrangement for at least the first, second and third shaders as an arrangement having a cost lower than a predetermined cost metric, the common resource/slot arrangement based upon the first shader associating a first resource with a first slot, the second shader associating the first resource with the first slot, and the third shader associating the first resource with a second slot different than the first slot; and

binding the first resource to the first slot for the plurality of shaders based upon the determined common resource/slot arrangement, such that the third shader is modified to associate the first resource with the first slot instead of the second slot,

at least some of at least one of the identifying a particular resource/slot arrangement of a first shader, identifying a particular resource/slot arrangement of a second shader, identifying a particular resource/slot arrangement of a third shader, determining, binding, and modifying implemented at least in part via a microprocessor.

2. The method of claim 1, a cost of the arrangement comprising a function of a number of rebindings necessary to transition between currently loaded shaders.

3. The method of claim 1, the common resource/slot arrangement binding respective resources used by at least one of the first, second, and third shaders to at least one slot.

4. The method of claim 1, the first, second, and third shaders using a plurality of same resources and the first, second, and third shaders modified according to the common resource/slot arrangement such that the respective shaders associate a same resource with a same slot.

5. The method of claim 1, comprising:

calculating, before determining a common resource/slot arrangement, a cost for binding the first resource to the first slot based upon resource dependencies between remaining resources being utilized by the first, second, and third shaders; and

calculating, before determining a common resource/slot arrangement, a cost for binding the first resource to the second slot based upon the resource dependencies between remaining resources being utilized by the first, second, and third shaders,

determining the common resource/slot arrangement comprising;

determining that the cost for binding the first resource to the first slot is less than the cost for binding the first resource to the second slot.

6. The method of claim 5, calculating the cost for binding the first resource to the first slot comprising:

determining a number of state changes required to render the first, second, and third shaders if the first resource is bound to the first slot;

determining a number of wasted slots required if the first resource is bound to the first slot;

calculating a first cost associated with binding the first resource to the first slot based upon at least one of the number of state changes and the number of wasted slots; and

repeating the acts of determining the number of state changes, determining the number of wasted slots, and calculating a cost until a cost has been calculated for binding respective resources used by at least one of the first, second, and third shaders to the first slot.

7. The method of claim 1, the processing unit comprising a graphics processing unit and the first resource comprising a constant buffer.

8. The method of claim 1, the method performed dynamically at one or more of application run time and load time.

9. A computer readable storage device comprising computer executable instructions for performing the method of claim 1.

10. A computer readable storage device comprising a graphics application programming interface for instructing a graphics processing unit to perform the method of claim 1.

11. A computer readable storage device comprising computer executable instructions that when executed via a processor on a computer perform a method for assigning resources used by at least two shaders of a graphics pipeline to particular slots of a graphics processing unit, comprising:

identifying a first shader assigning a first resource to a first slot;

identifying a second shader assigning the first resource to a second slot, the second slot different than the first slot; and

reassigning, at the second shader, the first resource to the first slot to achieve a cost lower than a predetermined cost metric.

12. The computer readable storage device of claim 11, the first resource reassigned to the first slot at the second shader to reduce a cost of resource reconfiguration between shaders.

13. The computer readable storage device of claim 11, comprising:

calculating a cost for assigning the first resource to the first slot; and

calculating a cost for assigning the first resource to the second slot, the first resource reassigned to the first slot at the second shader because the cost for assigning the first resource to the first slot is less than the cost for assigning the first resource to the second slot.

14. The computer readable storage device of claim 13, calculating the cost for assigning the first resource to the first slot comprising:

determining a number of state changes required to render the first and second shaders if the first resource is assigned to the first slot;

determining a number of wasted slots required if the first resource is assigned to the first slot; and

calculating the cost for assigning the first resource to the first slot based upon at least one of the number of state changes and the number of wasted slots.

15. The computer readable storage device of claim 11, the first shader and second shader configured to run consecutively.

16. A method for modifying a configuration of at least one shader having a particular resource/slot association, comprising:

scanning respective resource dependencies in a dependency tree, respective resources bound to respective slots of a parent object in the dependency tree, the parent object representative of a shader;

computing a cost associated with a number of resource configurations of the dependency tree;

selecting a particular resource configuration of the dependency tree which reduces a number of device reconfigurations when switching between different shaders; and

modifying a configuration of at least one shader based upon the particular resource configuration that is selected such that a resource that was bound to a first slot for the at least one shader is bound to a second slot,

at least some of at least one of the scanning, computing, selecting, and modifying implemented at least in part via a microprocessor.

17. The method of claim 16, comprising:

identifying a plurality of shaders running in parallel;

determining a number of resources associated with respective shaders running in parallel, respective shaders configured to assign respective resources to a slot; and

creating the dependency tree based upon the determined number of resources associated with respective shaders running in parallel.

18. The method of claim 16, the particular resource configuration of the dependency selected configured to reduce the number of device reconfigurations to a minimum.

19. A computer readable storage device comprising a graphics application programming interface for instructing a graphics processing unit to perform the method of claim 16.