Graphical processing system, graphical pipeline and method for implementing subpixel shifting to anti-alias texture

• US 7,456,846 B1
• Filed: 06/03/2005
• Issued: 11/25/2008
• Est. Priority Date: 06/03/2005
• Status: Active Grant

• Current Assignee: NVIDIA Corporation
• Original Assignee: NVIDIA Corporation
• Inventors: Toksvig, Michael J. M., King, Gary C.
• Primary Examiner(s): Tung; Kee M
• Assistant Examiner(s): Richer; Aaron M
• Application Number:
  US11/144,920
• Time in Patent Office: 1,271 Days
• Field of Search: 345/611, 345/613
• US Class Current: 345/613
• CPC Class Codes: G06T 15/005 General purpose rendering architectures
• CPC Class Codes: G06T 2200/12 involving antialiasing
• CPC Class Codes: G06T 2210/36 Level of detail
• CPC Class Codes: G09G 5/363 Graphics controllers
• CPC Class Codes: G09G 5/391 Resolution modifying circuits, e.g. variable screen formats
• CPC Class Codes: G09G 5/393 Arrangements for updating the contents of the bit-mapped memory

Abstract

A system, apparatus, and method are disclosed for modifying positions of sample positions for selectably oversampling pixels to anti-alias non-geometric portions of computer-generated images, such as texture, at least in part, by shifting shading sample positions relative to a frame of reference. There is generally no relative motion between the geometries and the coverage sample positions. In one embodiment, an apparatus, such as a graphics pipeline and/or a general purpose graphics processing unit, anti-aliases geometries of a computer-generated object. The apparatus includes at least a texture unit and a pipeline front end unit to determine geometry coverage and a subpixel shifter to shift shading sample positions relative to the frame of reference. The apparatus can receive subpixel shifting masks to select subsets of shading sample positions. Each of the shading sample positions is shifted to a coverage sample position to reduce level of detail (“LOD”) artifacts.

Claims

1. An apparatus for anti-aliasing a computer-generated object in a graphics pipeline configured to execute instructions of an applications program to generate geometry data representing geometries, said apparatus comprising:

a pipeline front end unit configured to at least determine coverage of said geometries over coverage sample positions;

a subpixel shifter configured to shift shading sample positions relative to a frame of reference that includes said geometries and said coverage sample positions such that there is no relative translation of said geometries with respect to said coverage sample positions; and

a texture unit configured to determine shading at each of said subsets of shading sample positions,

wherein said subpixel shifter is configured to shift each of said shading sample positions to a respective coverage sample position in a pixel for anti-aliasing said computer-generated object; and

wherein said subpixel shifter aligns said shading sample positions with said coverage samples to produce symmetrical arrangements of said shading sample positions over groupings of pixels, thereby reducing mismatched values of level of detail (“

LOD”

) that otherwise would introduce LOD artifacts.

2. The apparatus of claim 1 wherein said subpixel shifter further comprises a mask input to receive subpixel shifting masks for selecting subsets of shading sample positions of said shading sample positions.

3. The apparatus of claim 2 wherein one or more instructions of said applications program include an instruction to execute centroid sampling to enable said subpixel shifter to selectably shift each of said subsets of shading sample positions for translation relative to said frame of references wherein each of said subsets of shading sample positions are shifted in series as said applications program submits a scene into said graphics pipeline a number of times equivalent to the number of said subsets being shifted.

4. The apparatus of claim 3 further comprising:

a raster operations (“

ROP”

) unit configured to serially receive information for groups of subpixels associated with each of said subsets of shading sample positions, said information at least including shading and depth values, said raster operations unit being configured to at least perform alpha-testing serially on each of said groups of subpixels; and

a frame buffer for storing said groups of subpixels.

5. The apparatus of claim 2 further comprising:

a subpixel shifting mask selector unit coupled to said mask input to provide each of said subpixel shifting masks; and

an anti-aliasing mode selector unit configured to enable said subpixel shifter to perform oversampling.

6. The apparatus of claim 5 wherein said anti-aliasing mode selector unit is coupled to said pipeline front end unit to generate of multiple copies of said geometries from a single scene, and to provide said multiple copies in parallel from said pipeline front end unit to said subpixel shifter for shifting each of said subset of shading sample positions in approximate synchronicity.

7. The apparatus of claim 5 further comprising a duplicator unit coupled to said subpixel shifter and configured to locally generate multiple copies of said geometries from a single scene, and to provide said multiple copies to said subpixel shifter to shift each of said subset of shading sample positions.

8. The apparatus of claim 7 further comprising a pixel shader that includes said duplicator unit and said subpixel shifter, said pixel shader configured to provide at least shading values for groups of subpixels associated with each of said subsets of shading sample positions, wherein said anti-aliasing mode selector unit is coupled to said subpixel shifter to selectively shift each of said subset of shading sample positions for said geometries if oversampling is selected.

9. The apparatus of claim 1 wherein said texture unit is further configured to determine shading at said coverage sample positions rather than at default sample positions to facilitate accurate interpolation between screen coordinates and texture coordinates, thereby ensuring consistency of values of LOD over adjacent groupings of pixels of said computer-generate object.

10. The apparatus of claim 1 wherein said subpixel shifter further comprises a subpixel centroid sampling unit configured to establish centroids at only coverage sample positions.

11. The apparatus of claim 10 wherein said centroid sampling unit is configured to shift said shading sample positions to locations of said coverage sample positions without said applications program specifying either said locations or subpixel shift amounts, or both, thereby obviating manual determination of both said locations and said subpixel shift amounts.

12. The apparatus of claim 1 wherein said subpixel shifter performs multisampling when in a first state and performs oversampling of at least a portion of said computer-generated object using said subpixel shifter when in a second state, thereby enabling effective use of both multisampling and oversampling during multi-pass rendering.

13. The apparatus of claim 12 wherein said subpixel shifter is selectably operable to either configure said subpixel shifter to perform multisampling or to configure said subpixel shifter to perform oversampling of said non-geometric portions of said computer-generated object,

wherein said subpixel shifter is configured to shift any number of subsets of said shading sample positions to effectuate any degree of anti-aliasing,

wherein said subpixel shifter is configured to maintain substantially constant depth values of said geometries at each of said shading sample positions, thereby dispensing with a requirement to reconcile varying depth values that arise when shifting said geometries.

14. The apparatus of claim 13 wherein a first degree of anti-aliasing includes oversampling both coverage and shading four times per pixel, and a second degree of anti-aliasing includes oversampling coverage two times per pixel and multisampling shading two times per pixel.

15. A method for selectably performing multisampling and oversampling to sample shading for geometries of a computer-generated object, said method comprising:

multisampling a first portion of a scene to determine one shading value per pixel; and

oversampling a second portion of said scene by

shifting shading sample positions relative to a frame of reference that includes said geometries and coverage sample positions such that there is no relative translation between said geometries and said coverage sample positions, said shading sample positions being shifted to provide multiple shading values per pixel,

wherein said shading sample positions are shifted to establish translated shading sample positions coinciding with said coverage sample positions to reduce aliasing of non-geometric portions of said computer-generated object; and

wherein said shading sample positions are aligned with said coverage samples to produce symmetrical arrangements of said shading sample positions over groupings of pixels, thereby reducing mismatched values of level of detail (“

LOD”

) that otherwise would introduce LOD artifacts.

16. The method of claim 15 wherein multisampling said first portion comprises:

determining shading for a background object composed of tessellated geometries associated with a coarse level of detail (“

LOD”

), and oversampling said second portion comprises;

determining shading for a foreground object composed of tessellated geometries associated with a fine LOD, wherein oversampling reduces texture aliasing between said coarse LOD and said fine LOD.

17. The method of claim 15 wherein oversampling said second portion comprises:

determining shading during a first multipass-rendering pass for an object composed of tessellated geometries associated with a fine LOD, and multisampling said object during a second multipass-rendering pass comprises;

applying modified shading,

wherein said first multipass-rendering pass and said second multipass-rendering pass maintain substantially constant depth values for shading sampled at said shading sample positions, thereby reducing computational overhead otherwise used to reconcile mismatched depth values due to shifting said geometries.

18. The method of claim 15 wherein oversampling said second portion of said scene comprises performing for each subpixel constituting said second portion:

selecting one of a number of subpixel shifting masks;

transforming a number of geometries constituting said second portion;

rasterizing said number of geometries;

shifting shading sample positions associated with a subset of said shading sample positions, each subset being associated with different coverage sample positions;

sampling shading to derive shading values at said subset of said shading sample positions;

storing said shading values in a frame buffer; and

repeating for other subpixel shifting masks;

transforming, rasterizing, shifting shading sample positions, sampling shading, and storing shading values.

19. The method of claim 15 wherein oversampling said second portion of said scene comprises:

transforming a number of geometries constituting said second portion;

selecting a number of subpixel shifting masks;

rasterizing said number of geometries a number of times equivalent to said number of subpixel shifting masks;

shifting shading sample positions associated with a subset of said shading sample positions, each subset being associated with unique coverage sample positions;

sampling shading to derive shading values at said subset of said shading sample positions;

storing said shading values in a frame buffer; and

repeating for other subpixel shifting masks;

rasterizing, shifting shading sample positions, sampling shading, and storing shading values.

20. The method of claim 15 wherein oversampling said second portion of said scene comprises:

transforming a number of geometries constituting said second portion;

rasterizing said number of geometries;

selecting a number of subpixel shifting masks;

duplicating said number of geometries for each subpixel shifting mask;

shifting shading sample positions associated with a subset of said shading sample positions, each subset being associated with unique coverage sample positions;

sampling shading to derive shading values at said subset of said shading sample positions;

repeating for other subpixel shifting masks until each pixel of said second portion is oversampled;

shifting shading sample positions and sampling shading; and

storing said shading values as an oversampled pixel in a frame buffer.

21. A general purpose graphics processing unit (“

GPGPU”

) comprising;

a geometry processor configured to generate a geometry in accordance with application program instructions;

a rasterizer configured to convert said geometry into data representing coverage of subpixels at coverage sample positions;

a shader to determine shading for subpixels, said shader including;

a subpixel centroid sampling unit configured to perform centroid multisampling, said subpixel centroid sampling unit being configured to receive a number of subpixel shifting masks to either establish a subpixel centroid that coincides with a coverage sample position or reposition a default centroid from a pixel center to said coverage sample position, and

a texture unit configured to determine shading for said geometry using said coverage sample positions; and

a raster operations (“

ROP”

) unit configured to at least perform alpha-testing against said shading for said geometry;

wherein said subpixel centroid sampling unit comprises;

a coverage sample determinator configured to determine a number of covered sample positions that are covered by said geometry and a number of uncovered sample positions that are not covered by said geometry, both covered sample positions and uncovered sample positions sharing a frame of reference with said geometry such that there is no relative shifting among said geometry, said covered sample positions and said uncovered sample positions when anti-aliasing said computer-generated images;

a shading sample selector configured to select either a covered sample position from said number of covered sample positions or an uncovered sample position from said number of uncovered sample positions; and

a shading sample translator configured to establish said subpixel centroid if said covered sample position is selected or to reposition said default centroid if said uncovered sample position is selected,

wherein said subpixel centroid and said default centroid form symmetrically-spaced texture sample positions over a number of pixels to reduce level of detail (“

LOD”

) artifacts.