WO2009117691A4

WO2009117691A4 - Architectures for parallelized intersection testing and shading for ray-tracing rendering

Info

Publication number: WO2009117691A4
Application number: PCT/US2009/037860
Authority: WO
Inventors: Luke Tilman Peterson; James Alexander Mccombe; Ryan R. Salsbury; Stephen Purcell
Original assignee: Caustic Graphics, Inc
Priority date: 2008-03-21
Filing date: 2009-03-20
Publication date: 2009-12-30
Also published as: CN102037497A; CN104112291B; CN102037497B; JP2014089773A; JP5740704B2; KR101550477B1; JP5485257B2; CN104112291A; JP2011515766A; KR20100128337A; WO2009117691A3; WO2009117691A2

Abstract

In an example, ray tracing scenes includes using a plurality of intersection testing resources coupled with a plurality of shading resources, communicative in the aggregate through links/queues. A queue from testing to shading comprises respective ray/primitive intersection indications, comprising ray identifiers. A shading to testing queue comprises identifiers of new rays for test, wherein data defining the rays is separately stored in memories distributed among the intersection testing resources. Ray definition data can be retained in distributed memories until rays complete intersection testing, being selectable for testing multiple times based on ray identifier. A structure of acceleration shapes can be used. Packets of ray identifiers and shape data can circulate among intersection testing resources, each resource can test rays identified in the packet, and for which definition data is present in its memory. Acceleration shape test results allow collection of rays based on intersected shape, and closest detection ray/primitive intersections are indicated by queuing ray identifiers for shading.

Claims

AMENDED CLAIMS Received by the International Bureau on 09. Nov.2009 (09.11.2009)

1. A system for intersection testing of rays in a 3-D scene, comprising: a plurality of intersection testers, each having access to a respective cache memory to store a respective subset of a master copy of ray definition data that defines a plurality of rays; control logic having access to respective identifiers for at least some of the rays defined by definition data stored in the cache memories of the intersection testers, the control logic operable to control ray intersection testing by associating a plurality of the ray identifiers with shape information that is indicative of one or more shapes located in the 3-D scene to be intersection tested with those identified rays, and making data representative of the association available to the plurality of intersection testers, wherein each of the intersection testers is operable to test, for intersection with the indicated one or more shapes, identified rays for which its cache stores ray definition data; and an output buffer coupled to receive outputs from the intersection testers, the outputs comprising data indicative of results of the intersection testing.

2- The system of claim 1, further comprising a plurality of computation resources for executing shader code routines associated with primitives, which arc identified by the outputs from the intersection testers, retrievable from the output buffer.

3. The system of claim 2, wherein the execution of the shader code routines generates new rays to be intersection tested, and further comprising an input queue to the plurality of intersection testers to receive the new rays, and the control logic is operable to begin intersection testers of the new rays as other rays complete intersection testing.

4. The system of any one of claims 1 -3, wherein each intersection tester is configured to be responsive to receiving an identifier for a ray stored in its respective memory by testing the identified ray for intersection with one or more shapes, for which definition data is provided with the ray identifier or for which memory reference information is provided.

5. The system of any one of claims 1-4, wherein the intersection testing occurs between rays and geometric shapes comprising acceleration structure elements selected from one or more of cutting planes of a fcD-trce, axis-aligned bounding boxes, and spheres.

6. The system of any one of claims 1-5, wherein the intersection testers are configured for sequentially passing, among the intersection testers, the data representative of the association between the plurality of the ray identifiers and the shape information that is indicative of the one or more shapes to be intersection tested with those identified rays, to make that data available to all of the intersection testers.

50

7. The system of claim 6, wherein the shapes that each intersection tester can test for intersection with rays comprise geometry acceleration elements and primitives, and each intersection tester further is lo pass results of testing rays for intersection with geometry acceleration elements with the data representative of the association, and to maintain results of testing rays for intersection with primitives in the cache memories of the intersection testers.

8. The system of any one of claims 1-7, wherein the plurality of intersection testers are implemented as threads of computer executable instructions executing on one or more computational cores.

9. The system of any one of claims 1-8, further comprising a memory storing primitives composing a 3-D scene, the primitive-storing memory also serving as a main memory for a computing system comprising one or more computational cores, the one or more computational cores collectively can concurrently execute a plurality of threads, the threads allocated by the system between implementing the intersection testers and executing shader code routines on a time-varying basis.

10. A method of controlling ray tracing of a scene composed of primitives m a system having a plurality of compute resources, each coupled to a respective local memory and a shared main memory, where the main memory is higher in latency than the local memories, comprising: distributing, among the local memories of the compute resources, data defining respective subsets of the rays to be intersection tested in the scene; determining to test a group of the rays, members of the group collectively stored in multiple of the local memories, for intersection with a geometric shape; providing data for the geometric shape and ray identifiers to the compute resources to cause at least one compute resource whose local memory stores definition data for a ray in the group to receive the geometric shape data and the ray identifiers; and receiving, from the compute resources, indications of detected intersections between rays of the group and the geometric shape, the indications resulting from testing each ray of the group in at least one computing resource whose local memory was storing definition data for that ray.

11. The method of claim 10, further comprising fetching data defining the shape from the main memory and wherein the providing data for the geometric shape causes provision of the shape definition data to the compute resources with the ray identifiers of the group.

12. The method of any one of claims 10-1 1 , wherein the indications comprise data for intersections between geometry acceleration elements and rays, and the group of rays was formed by collecting rays determined to intersect the same geometry acceleration clement and further comprising

51 deferring further testing of geometry acceleration elements related to tbat geometry acceleration element until a sufficient number rays are collected.

13. The method of any one of claims 10-12, wherein the local memories include cache memories and further comprising protecting data defining a given ray from being overwritten in its cache memory until that ray has completed intersection testing.

14. The method of any one of claims 10-13, further comprising retaining in respective local memories respective current closest detected intersection for rays having ray definition data stored in that local memory and generating each intersection indication in response to identifying the closest possible intersection between any primitive and a given ray.

15. The method of any one of claims 10-14, wherein the data for the geometric shape comprises a selection from the set of (1) a reference identifying one or more shapes to be tested and (2) data defining one or more shapes to be tested.

16. The method of any one of claims 10-15, wherein the providing comprises queuing ray identifiers into a first queue, from which the compute resources are coupled to receive, and the receiving comprises receiving the intersection indications from a second queue.

17. The method of any one of claims 10-16, further comprising maintaining a master copy of the rays in the main memory.

18. A computer readable medium storing modules of computer executable code for implementing a method according to any one of claims 10-17 on one or more computation resources.

19. A system for rendering representations of a 3-D scene composed of primitives using ray ^"tracing, comprising: a memory storing primitives composing a 3-D scene; one or more memories storing definition data for a plurality of rays; a plurality of intersection testers, each for testing at least one of the rays with at least one of the primitives, resulting in output of an intersection testing result; a plurality of shadcr computation units, each operable to run shading routines for detected ray /primitive intersections, indicated by the results, the running of the shading routines resulting in new rays to be intersection tested, definition data for the new rays to be stored in the one or more memories storing definition data for the plurality of rays; a first communication link for outputting the intersection testing results from the intersection testers to the shader resources; and a second communication link for sending, to the intersection testers, identifiers for new rays resulting from running of the shading routines, the identifiers to be for retrieving definition data for respective rays from the one or more memories during intersection testing of those rays.

52

20. The system of claim J 9, further comprising channels for passing messages among the plurality of intersection testers, wherein each of the intersection testers is configured to interpret data in a message received by it as containing a plurality of ray identifiers, and to intersection test a selected ray identified in the message.

21. The system of claim 19, wherein the intersection testers are configured in a ring to pass a packet of ray identifiers among the intersection testers.

22. The system of claim 19, wherein each of the intersection testers are operable to select respective rays for test based on determining whether a cache associated with that intersection tester stores definition data for any of the rays identified in a message passed among the intersection testers.

23. The system of claim 19, wherein the plurality of intersection testers are implemented as threads of computer executable instructions executing on one or more computational cores, each having localized cache memory access to a subset of the rays traveling in the scene.

24. The system of claim 19, wherein the memory storing primitives composing a 3-D scene is implemented as a main memory for one or more computational cores, the one or more cores collectively can concurrently execute a number of threads, the number of threads being allocated between implementing the intersection testers and the shader computation units on a time-varying basis.

53