ORDER-PRESERVING DISTRIBUTED RASTERIZER

• US 20110090220A1
• Filed: 10/15/2009
• Published: 04/21/2011
• Est. Priority Date: 10/15/2009
• Status: Active Grant

• Current Assignee: NVIDIA Corporation
• Original Assignee: NVIDIA Corporation
• Inventors: Purcell, Timothy John, Crow, Franklin C., Rhoades, Johnny S., Molnar, Steven E., Kilgariff, Emmett M., Treichler, Sean J., Bowman, James C., Hakura, Ziyad S.
• Granted Patent: US 8,587,581 B2
• Time in Patent Office: Days
• US Class Current: 345/420
• CPC Class Codes: G06T 15/005 General purpose rendering architectures
• CPC Class Codes: G06T 2210/52 Parallel processing

Abstract

One embodiment of the present invention sets forth a technique for rendering graphics primitives in parallel while maintaining the API primitive ordering. Multiple, independent geometry units perform geometry processing concurrently on different graphics primitives. A primitive distribution scheme delivers primitives concurrently to multiple rasterizers at rates of multiple primitives per clock while maintaining the primitive ordering for each pixel. The multiple, independent rasterizer units perform rasterization concurrently on one or more graphics primitives, enabling the rendering of multiple primitives per system clock.

Claims

1. A method for rendering primitives in parallel, the method comprising:

receiving primitives in a first order, the primitives encoded as a list of vertex indices;

distributing the primitives for parallel processing in object-space to produce multiple streams, each stream including object-space processed primitive descriptors that represent individual primitives;

routing the multiple streams of object-space processed primitive descriptors based on screen-space positions to produce a first stream of processed primitive descriptors representing primitives that intersect a first portion of a screen and a second stream of processed primitive descriptors representing primitives that intersect a second portion of the screen;

reordering the object-space processed primitive descriptors in the first stream to match the first order, producing a reordered first stream of object-space processed primitive descriptors;

reordering the object-space processed primitive descriptors in the second stream to match the first order, producing a reordered second stream of object-space processed primitive descriptors; and

rasterizing the primitives that intersect the first portion of the screen, and are represented by the first stream of object-space processed primitive descriptors, and the primitives that intersect the second portion of the screen, and are represented by the second stream of object-space processed primitive descriptors, in parallel to produce rasterized pixels of an image.

2. The method of claim 1, further comprising, prior to the step of distributing the primitives, grouping the primitives into batches.

3. The method of claim 2, wherein the batches are distributed in round-robin fashion for the parallel processing in object-space.

4. The method of claim 2, wherein the batches encode a number of sequential primitives represented as a set of vertex indices and primitive topology information and the routing of the multiple streams of object-space processed primitive descriptors transmits the vertex indices instead of the vertex data.

5. The method of claim 1, further comprising, prior to the step of routing the multiple streams, the step of computing a bounding box indicating screen-space positions for each primitive represented by an object-space processed primitive descriptor.

6. The method of claim 1, wherein the multiple streams include a merged stream that is produced by merging two or more of the streams of object-space processed primitive descriptors and reordering the object-space processed primitive descriptors in the merged stream to match the first order.

7. The method of claim 1, wherein the step of routing comprises routing the first stream of object-space processed primitive descriptors to a first processing unit that is configured to rasterize portions of the primitives that intersect the first portion of the screen and routing the second stream of object-space processed primitive descriptors to a second processing unit that is configured to rasterize portions of the primitives that intersect the second portion of the screen.

8. The method of claim 1, further comprising the step of storing vertex attributes produced by the parallel processing in object-space in memory.

9. The method of claim 8, wherein the step of rasterizing the primitives that intersect the first portion of the screen and the primitives that intersect the second portion of the screen further comprises reading the vertex attributes from the memory.

10. The method of claim 7, wherein the routing comprises broadcasting a primitive descriptor representing a first primitive that intersects the first portion of the screen and the second portion of the screen to the first processing unit and the second processing unit.

11. A computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to render primitives in parallel, by performing the steps of:

receiving primitives in a first order, the primitives encoded as a list of vertex indices;

distributing primitive descriptors that represent the primitives for parallel processing in object-space to produce multiple streams, each stream including object-space processed primitive descriptors;

routing the multiple streams of object-space processed primitive descriptors based on screen-space positions to produce a first stream of processed primitive descriptors representing primitives that intersect a first portion of a screen and a second stream of processed primitive descriptors representing primitives that intersect a second portion of the screen;

reordering the object-space processed primitive descriptors in the first stream to match the first order, producing a reordered first stream of object-space processed primitive descriptors;

reordering the object-space processed primitive descriptors in the second stream to match the first order, producing a reordered second stream of object-space processed primitive descriptors; and

rasterizing the primitives that intersect the first portion of the screen, and are represented by the first stream of object-space processed primitive descriptors, and the primitives that intersect the second portion of the screen, and are represented by the second stream of object-space processed primitive descriptors, in parallel to produce rasterized pixels of an image.

12. A system for rendering primitives in parallel, the system comprising:

a processor configured to;

receive primitives in a first order, the primitives encoded as a list of vertex indices;

distribute primitive descriptors that represent the primitives for parallel processing in object-space to produce multiple streams, each stream including object-space processed primitive descriptors;

route the multiple streams of object-space processed primitive descriptors based on screen-space positions to produce a first stream of processed primitive descriptors representing primitives that intersect a first portion of a screen and a second stream of processed primitive descriptors representing primitives that intersect a second portion of the screen;

reorder the object-space processed primitive descriptors in the first stream to match the first order, producing a reordered first stream of object-space processed primitive descriptors;

reorder the object-space processed primitive descriptors in the second stream to match the first order, producing a reordered second stream of object-space processed primitive descriptors; and

rasterize the primitives that intersect the first portion of the screen, and are represented by the first stream of object-space processed primitive descriptors, and the primitives that intersect the second portion of the screen, and are represented by the second stream of object-space processed primitive descriptors, in parallel to produce rasterized pixels of an image.

13. The system of claim 12, further comprising a memory storing instructions that, when executed by the processor, configures the processor to:

receive the primitive descriptors in the first order;

distribute the primitive descriptors for parallel processing in object-space;

route the multiple streams of object-space processed primitive descriptors;

reorder the object-space processed primitive descriptors in the first stream;

reorder the object-space processed primitive descriptors in the second stream; and

rasterize the primitives that intersect the first portion of the screen and the primitives that intersect the second portion of the screen.

14. The system of claim 12, wherein the processor is further configured to divide the primitive descriptors into batches and the batches are distributed for the parallel processing in object-space.

15. The system of claim 14, wherein the batches are distributed in round-robin fashion.

16. The system of claim 14, wherein the batches contain pointers to vertex data and the routing of the multiple streams of object-space processed primitive descriptors transmits the pointers instead of the vertex data.

17. The system of claim 12, wherein the processor is further configured to computing a bounding box indicating screen-space positions for each primitive represented by an object-space processed primitive descriptor, prior to routing the multiple streams.

18. The system of claim 12, wherein the processor is further configured to merge two or more of the streams of object-space processed primitive descriptors and reorder the object-space processed primitive descriptors in the merged stream to match the first order, prior to routing the multiple streams.

19. The system of claim 12, wherein the processor is further configured to route the first stream of object-space processed primitive descriptors to a first processing unit that is configured to rasterize portions of the primitives that intersect the first portion of the screen and route the second stream of object-space processed primitive descriptors to a second processing unit that is configured to rasterize portions of the primitives that intersect the second portion of the screen.

20. The system of claim 19, wherein the first processing unit includes a rasterizer that is configured to skip over screen tiles of a tiling pattern that are outside of the first portion of the screen when rasterizing the portions of the primitives that intersect the first portion of the screen.