Context-aware compression with quantization of hierarchical transform matrices

Patent metadata

Field	Value
Publication number	US-11335035-B2
Application number	US-202017003040-A
Country	US
Kind code	B2
Filing date	Aug 26, 2020
Priority date	Dec 28, 2018
Publication date	May 17, 2022
Grant date	May 17, 2022

Abstract

Apparatus and method for context-aware compression. For example, one embodiment of an apparatus comprises: ray traversal/intersection circuitry to traverse rays through a hierarchical acceleration data structure to identify intersections between rays and primitives of a graphics scene; matrix compression circuitry/logic to compress hierarchical transformation matrices to generate compressed hierarchical transformation matrices by quantizing N-bit floating point data elements associated with child transforms of the hierarchical transformation matrices to variable-bit floating point numbers or integers comprising offsets from a parent transform of the child transform; and an instance processor to generate a plurality of instances of one or more base geometric objects in accordance with the compressed hierarchical transformation matrices.

First claim

What is claimed is:

1. An apparatus comprising:
a graphics processing unit (GPU) comprising a non-transitory machine-readable medium containing code, which when executed by the GPU, capable of causing the GPU to perform:
generating initial hierarchical transformation matrices to transform one or more base geometric objects of a graphics scene;
compressing the initial hierarchical transformation matrices to generate compressed hierarchical transformation matrices by quantizing N-bit floating point data elements associated with a child transform to variable-bit floating point numbers or integers comprising offsets from a parent transform of the child transform; and
generating a plurality of instances of the one or more base geometric objects based on the compressed hierarchical transformation matrices, wherein the plurality of instances of the one or more base geometric objects are used to render the graphics scene.

2. The apparatus of claim 1 , wherein a level of precision with which to quantize one or more of the N-bit floating point data elements is based on an error tolerance of the graphics scene.

3. The apparatus of claim 2 , wherein the GPU is to be caused to further perform:
traversing rays through a hierarchical acceleration data structure to identify intersections between rays and primitives of a graphics scene.

4. The apparatus of claim 3 , wherein current context parameters associated with the ray traversal are analyzed to determine the level of precision with which to quantize one or more of the N-bit floating point data elements.

5. The apparatus of claim 4 , wherein the current context parameters are to be evaluated to determine the error tolerance, the error tolerance based on at least one of a distance from a camera to the primitives of the graphics scene or a type of ray being traversed.

6. The apparatus of claim 5 , wherein the current context parameters are to be evaluated at one of creation time of the graphics scene and runtime.

7. The apparatus of claim 6 , wherein when the current context parameters are evaluated at the graphics scene creation time, the evaluation is associated with a particular one or more objects in a region of interest and wherein when the current context parameters are evaluated at the runtime, the evaluation is based on the distance.

8. The apparatus of claim 1 , wherein compressing the initial hierarchical transformation matrices comprises procedurally generating the compressed hierarchical transformation matrices based on the initial hierarchical transformation matrices.

9. The apparatus of claim 1 , wherein compressing the initial hierarchical transformation matrices comprises selecting at least one of:
only a subset of the initial hierarchical transformation matrices and
only a subset of data elements of the initial hierarchical transformation matrices.

10. A method comprising:
generating initial hierarchical transformation matrices to transform one or more base geometric objects of a graphics scene;
compressing the initial hierarchical transformation matrices to generate compressed hierarchical transformation matrices by quantizing N-bit floating point data elements associated with a child transform to variable-bit floating point numbers or integers comprising offsets from a parent transform of the child transform; and
generating a plurality of instances of the one or more base geometric objects based on the compressed hierarchical transformation matrices, wherein the plurality of instances of the one or more base geometric objects are used to render the graphics scene.

11. The method of claim 10 , wherein a level of precision with which to quantize one or more of the N-bit floating point data elements is based on an error tolerance of the graphics scene.

12. The method of claim 11 , wherein the method further comprises:
traversing rays through a hierarchical acceleration data structure to identify intersections between rays and primitives of a graphics scene.

13. The method of claim 12 , wherein current context parameters associated with the ray traversal are analyzed to determine the level of precision with which to quantize one or more of the N-bit floating point data elements.

14. The method of claim 13 , wherein the current context parameters are to be evaluated to determine the error tolerance, the error tolerance based on at least one of a distance from a camera to the primitives of the graphics scene or a type of ray being traversed.

15. The method of claim 14 , wherein the current context parameters are to be evaluated at one of creation time of the graphics scene and runtime.

16. The method of claim 15 , wherein when the current context parameters are evaluated at the graphics scene creation time, the evaluation is associated with a particular one or more objects in a region of interest and wherein when the current context parameters are evaluated at the runtime, the evaluation is based on the distance.

17. The method of claim 10 , wherein compressing the initial hierarchical transformation matrices comprises procedurally generating the compressed hierarchical transformation matrices based on the initial hierarchical transformation matrices.

18. The method of claim 10 , wherein compressing the initial hierarchical transformation matrices comprises selecting at least one of:
only a subset of the initial hierarchical transformation matrices and
only a subset of data elements of the initial hierarchical transformation matrices.

19. A non-transitory machine-readable medium having program code stored thereon which, when executed by a machine, causes the machine to perform:
generating initial hierarchical transformation matrices to transform one or more base geometric objects of a graphics scene;
compressing the initial hierarchical transformation matrices to generate compressed hierarchical transformation matrices by quantizing N-bit floating point data elements associated with a child transform to variable-bit floating point numbers or integers comprising offsets from a parent transform of the child transform; and
generating a plurality of instances of the one or more base geometric objects based on the compressed hierarchical transformation matrices, wherein the plurality of instances of the one or more base geometric objects are used to render the graphics scene.

20. The non-transitory machine-readable medium of claim 19 , wherein a level of precision with which to quantize one or more of the N-bit floating point data elements is based on an error tolerance of the graphics scene.

Assignees

• Intel Corp

Inventors

• Brownlee Carson
• Benthin Carsten
• Barczak Joshua
• Xiao Kai
• Apodaca Michael
• Surti Prasoonkumar
• Raoux Thomas

Classifications

• G06T9/40 Primary

  Tree coding, e.g. quadtree, octree · CPC title

• G06T1/60 Primary

  Memory management · CPC title

• G06T2210/21

  Collision detection, intersection · CPC title

• G06T15/20

  Perspective computation · CPC title

• G06T9/002

  using neural networks · CPC title