Device and method for achieving triangle antialiasing in GPU

Abstract

The invention discloses a device and method for achieving triangle antialiasing in a GPU. The device comprises a triangle horizontal scanning line generation module, a scanning line boundary extensionmodule, an area ratio calculation and processing module and a fragment fusion module. The implementation method comprises the steps that the boundary of a horizontal scanning line in a triangle is calculated and extended, the area ratio of fragments on the scanning line is calculated and processed, finally, the area ratio is used as a source blending factor of fusion operation, the result that the area ratio is subtracted from one serves as a target blending factor, and corresponding pixels in a source fragment and a frame memory are fused so as to achieve the triangle primitive antialiasingeffect.

Description

technical field [0001] The invention mainly relates to the field of GPU design, in particular to the field of anti-aliasing realization of triangles in GPU. Background technique [0002] When drawing a triangle on a raster graphics display, since the raster is composed of discrete points, it must be sampled at discrete positions on the raster display device. The information distortion caused by insufficient sampling information after reconstruction is called aliasing, which is used to reduce or eliminate this. The technique of this effect is called anti-aliasing. Since the aliasing phenomenon of the graphics has a great impact on the graphics quality, it is necessary to perform anti-aliasing processing on the basic primitives, and triangle is one of the basic primitives. [0003] There are two main methods of anti-aliasing: SSAA (Super Sampling Anti-Aliasing, super sampling anti-aliasing), by rendering a large image and then shrinking, this is equivalent to making a uniform...

AI Technical Summary

Problems solved by technology

[0003] There are two main methods of anti-aliasing: SSAA (Super Sampling Anti-Aliasing, super sampling anti-aliasing), by rendering a large image and then shrinking, this is equivalent to making a uniform distribution sampling inside each final pixel (which can be evenly distributed distribution, random distribution, jitter distribution, etc.), each sampling point has independent color and depth values, fragment processing is performed once at each sampling point, SSAA anti-aliasing effect is good, but the problem is that it leads to storing color and depth values The storage space is doubled; SSAA needs to execute PS once at each sampling point and save color and depth, and the time and space overhead are astonishing; MSAA (Multi-Sampling Anti-Aliasing, multi-sampling anti-aliasing) appears extremely This has been greatly improved. MSAA only performs fragment processing once for each fragment, and the output color is written to all sample books that pass the depth-stencil test. This method reduces the amount of calculation, but the disadvantage of high storage space is not improve

Images