Method and apparatus implementing high resolution rendition of Z-buffered primitives

• US 6,046,746 A
• Filed: 07/01/1996
• Issued: 04/04/2000
• Est. Priority Date: 07/01/1996
• Status: Expired due to Term

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Deering, Michael F.
• Primary Examiner(s): Trammell, James P.
• Assistant Examiner(s): Young, John Leonard
• Application Number:
  US08/673,117
• Time in Patent Office: 1,373 Days
• Field of Search: 345/139, 345/516, 345/424, 345/431, 345/429, 345/426, 345/419, 345/422, 345/505, 345/421, 345/503, 345/441, 345/502, 345/507, 345/420, 345/418, 345/423, 345/382, 345/285, 364/559, 364/746.1, 364/748, 364/715.04, 364/578, 364/715.03
• US Class Current: 345/422
• CPC Class Codes: G06T 15/405 using Z-buffer

Abstract

An object'"'"'s Z-buffered primitives are determined using a floating point Z=wF/W, where wF is the value the W coordinate achieves at the front clipping plane F. This representation produces Z-values ranging from 1 to ≈0 as W varies from W=wF to W→∞. Z-values for distances near the back clipping plane advantageously have leading zeros and are more amenable to floating point representation and exhibit less information loss. Primitive vertices are examined and the largest vertex floating point exponent is stored and associated with the entire primitive as that triangle'"'"'s floating point Z exponent. The stored exponent is subtracted from all the vertices'"'"' exponents and the results converted to fixed point, which format advantageously typically has few or no leading zeroes. After normal fixed point set-up and scan conversion operations occur, each pixel'"'"'s final Z fixed-point value is converted back to floating point, and the relevant stored exponent for the polygon is added back in before distance comparisons are made.

Claims

1. A method for rendering three-dimensional graphics data, the method comprising:

receiving a set of three-dimensional graphics data representing a plurality of primitives, wherein each primitive comprises one or more vertices;

translating said one or more vertices to view clipping coordinates based on a selected viewpoint, wherein said translating includes generating Z-values for the one or more vertices, wherein the Z-values vary from one at a front clipping plane to near zero at a back clipping plane; and

scan converting the primitives into pixels, wherein said pixels are displayable to form an image, wherein said translating is accomplished by applying a formula substantially equivalent to Z=Wf/w, wherein Z is the Z-value being calculated, wherein w is a scaled distance along a z-axis from the selected viewpoint to the vertex being transformed, and wherein Wf is a value of w at the front clipping plane.

2. The method as recited in claim 1, wherein each primitive comprises two or more vertices, wherein the Z-values are generated in floating point format, and wherein said method further comprises:

comparing the Z-values of said two or more vertices to determine which of the Z-values has the largest exponent value; and

storing the largest exponent value.

3. The method as recited in claim 2, further comprising

subtracting the largest exponent value from each of the Z-values;

converting the resulting Z-values to fixed-point format; and

performing scan line conversion on the fixed-point Z-values to generate pixels having fixed-point Z-values.

4. The method as recited in claim 3, further comprising

converting the fixed point pixel Z-values to floating point format;

adding the corresponding stored floating point exponents to the floating point pixel Z-values; and

storing the floating point pixel Z-values in a frame buffer.

5. The method as recited in claim 4, wherein said storing is performed only in response to a successful Z-distance comparison that compares the floating point pixel Z-values to be stored in the frame buffer with the floating point pixel Z-values already in the frame buffer.

6. The method as recited in claim 5, further comprising normalizing the floating point pixel Z-values.

7. The method as recited in claim 2, further comprising associating the stored exponent with the corresponding primitive.

8. The method as recited in claim 1, wherein said associating includes providing at least one or a (i) special header, and (ii) a tag in data representing values of said primitive.

9. A graphics system configured to render three-dimensional graphics data, the graphics system comprising:

a graphics accelerator configured to receive the three-dimensional graphics data and generate a corresponding plurality of pixels that are displayable to form an image, wherein the three-dimensional graphics data represents a plurality of primitives, wherein each primitive comprises one or more vertices, and wherein the graphics accelerator is configured to translate the vertices to view clipping coordinates based on a selected viewpoint by generating Z-values for the one or more vertices, wherein the Z-values vary from one at a front clipping plane to near zero at a back clipping plane;

a frame buffer configured to receive and store pixels from the graphics processor; and

a video control unit configured to read the contents of the frame buffer and output them to a display device, wherein said graphics accelerator is configured to translate the vertices by applying a formula substantially equivalent to Z=Wf/w, wherein Z is the Z-value being calculated, wherein w is a scaled distance along a z-axis from the selected viewpoint to the vertex being transformed, and wherein Wf is a value of w at the front clipping plane.

10. The graphics system as recited in claim 9, wherein each primitive comprises two or more vertices, wherein the Z-values are generated in floating point format, and wherein the graphics accelerator is further configured to compare the Z-values of said two or more vertices to determine which of the Z-values has the largest exponent value;

and wherein the graphics accelerator is configured with a plurality of storage locations configured to the largest exponent values for a plurality of the primitives being rendered.

11. The graphics system as recited in claim 10, wherein the graphics accelerator is further configured to:

subtract the largest exponent value from each of the Z-values;

convert the resulting Z-values to fixed-point format; and

perform scan line conversion on the fixed-point Z-values to generate pixels having fixed-point Z-values.

12. The graphics system as recited in claim 11, wherein the graphics accelerator is further configured to:

convert the fixed point pixel Z-values to floating point format;

add the corresponding stored floating point exponents to the floating point pixel Z-values; and

store the floating point pixel Z-values in a frame buffer.

13. The graphics system as recited in claim 12, wherein the frame buffer is configured to store only the floating point pixels with Z-values that indicate positions that are closer to the selected viewpoint than any pixels already stored in the frame buffer at corresponding locations.

14. A method for rendering three-dimensional graphics data, the method comprising:

receiving a set of three-dimensional graphics data representing a plurality of primitives, wherein each primitive comprises two or more vertices;

translating said two or more vertices to view clipping coordinates based on a selected viewpoint, wherein said translating includes generating floating point Z-values for each vertex;

comparing the Z-values of said two or more vertices to determine which of the Z-values has the largest exponent value;

subtracting the largest exponent value from each of the Z-values;

converting the resulting Z-values to fixed-point format;

performing scan line conversion on the fixed-point Z-values to generate pixels having fixed-point Z-values, wherein the pixels are displayable to form an image;

converting the fixed point pixel Z-values to floating point format;

adding the corresponding largest exponent to the floating point pixel Z-values; and

storing the pixels and the floating point pixel Z-values in a frame buffer, wherein said translating is accomplished by applying a formula substantially equivalent to Z=Wf/w, wherein Z is the Z-value being calculated, wherein w is a scaled distance along a z-axis from the selected viewpoint to the vertex being transformed, and wherein Wf is a value of w at the front clipping plane.

15. The method as recited in claim 14, wherein said translating includes generating Z-values for the one or more vertices, wherein the Z-values vary from one at a front clipping plane to near zero at a back clipping plane.

16. The method as recited in claim 14, further comprising storing the largest exponent for each primitive that is being processed by the graphics accelerator, wherein the largest exponent for each primitive is assigned a tag associated with the corresponding primitive.

17. The method as recited in claim 14, further comprising normalizing the floating point pixel Z-values.

18. A graphics accelerator configured to render three-dimensional graphics data, the graphics system comprising:

a means for receiving the three-dimensional graphics data;

a graphics processor configured to generate a plurality of pixels corresponding to the three-dimensional graphics data, wherein the three-dimensional graphics data represents a plurality of primitives, wherein each primitive comprises one or more vertices, and wherein the graphics processor is configured to translate the vertices to view clipping coordinates based on a selected viewpoint by generating Z-values for the one or more vertices, wherein the Z-values vary from one at a front clipping plane to near zero at a back clipping plane, wherein said graphics accelerator is configured to translate the vertices by applying a formula substantially equivalent to Z=Wf/w, wherein Z is the Z-value being calculated, wherein w is a scaled distance along a z-axis from the selected viewpoint to the vertex being transformed, and wherein Wf is a value of w at the front clipping plane; and

a means for outputting the pixels to a frame buffer for storage and eventual display as an image.

19. The graphics accelerator as recited in claim 18, further comprising:

a means for translating at least part of each vertex'"'"'s position information into fixed point format in preparation for performing scan line conversion to generate the plurality of pixels.

20. The graphics accelerator as recited in claim 19, further comprising:

a means for translating each pixel'"'"'s Z value from fixed point format into floating point format before outputting the pixels to the frame buffer.

21. The graphics accelerator as recited in claim 19, further comprising:

a means for translating each pixel'"'"'s Z value from fixed point format into floating point format.

22. The graphics accelerator as recited in claim 21, further comprising:

a means for normalizing the floating point format of each pixel'"'"'s Z value before outputting the pixel to the frame buffer.

23. A method for rendering graphics data in a graphics system, the method comprising:

receiving three-dimensional graphics data representing objects to be displayed by the computer graphics system, wherein the graphics data comprises one or more graphics primitives, wherein each graphics primitives comprises one or more vertices, wherein each vertex includes a floating point Z-value;

comparing the exponents of the floating point Z-values for a particular geometric primitive;

storing the largest exponent;

subtracting the largest exponent from the Z-values for the particular geometric primitive;

translating the Z-values for the particular geometric primitive into fixed-point format using the formula Z=Wf/w, wherein Wf is the value of a w-coordinate at a front clipping plane F;

performing set-up calculations on the particular geometric primitive using the fixed-point format Z-values to generate fixed-point results;

converting the fixed-point results of the set-up calculations into converted floating point Z-values;

performing Z comparisons using the converted floating point Z-values;

rendering the geometric primitive into a frame buffer in response to a winning Z comparison; and

displaying the contents of the frame buffer on a display device to form an image.