KR100269118B1 - Rasterization using quadrangle - Google Patents

Abstract

PURPOSE: A method for reasterization using a rectangle is provided to reduce the amount of input data and a memory accessing time by counting two triangles sharing one edge as one rectangle. CONSTITUTION: A rectangle with formed two triangles sharing one edge is extracted from an object projected on a screen coordinate(300). An edge function value for four edges of the rectangle and the shared edge of two triangles is calculated if there are coordinate values of vertical axises of two fixed points shared by two triangles between coordinate values of vertical axises of two fixed points which are not shared by two triangles(302,306). A span start point is obtained by edge-walking left edges of the triangles(308). An interpolating process is performed by using the edge function value and the span start point(310).

Description

Rasterization using rectangles {Rasterization using quadrangle}

The present invention relates to a rasterization method, and more particularly to a rasterization method using a square.

Modern three-dimensional computer graphics are rapidly spreading to personal computers, and can generate images at high speed as a result of three-dimensional graphics rendering. This requires a dedicated hardware board for 3D rendering, which is usually equipped with a 3D rendering acceleration chipset. The 3D acceleration chipset receives triangular data projected from the screen coordinates from a setup engine and applies a predetermined function to generate pixel values for all the pixels in the triangle. do. This chipset is commonly referred to as a rendering engine or raster engine.

FIG. 1 is a block diagram of a general three-dimensional graphics system. The graphics system according to FIG. 1 includes a geometry engine 100, a setup engine 110, and a rendering engine 120. The geometry engine 100 calculates the geometric change of the object to be rendered and projects the result on screen coordinates, and the setup engine 110 needs to rasterize from the projected polygons. Calculate the parameter. The rendering engine 120 generates the final pixel value using the rasterization parameter calculated by the setup engine 110.

In the setup engine 110, when the rendering engine performs rendering in triangle units, an edge function value calculator which calculates, from a vertex of a triangle, an edge function value required to traverse each triangle. (112), the rendering engine 120 registers the edge function values for the three edges of the triangle according to the pixel position and registers 122 storing the edge function value calculated by the edge function value calculation unit 112. There is an edge function interpolator 124 to calculate.

The amount and type of data to be input to the rendering engine for rasterization of one triangle depends on the triangle scanning method supported by the rendering engine.

In general, the triangular scanning method includes a method using a digital differential analyzer (DDA) algorithm and a Pineda method using an edge function value.

Among them, the Pineda method uses an edge function value along the left edge of a triangle to find the starting point of a span, which is a predetermined number of pulse trains. Here, the edge function value represents the intensity of each pixel constituting the edge. In the screen coordinates, each edge of the triangle is calculated from each vertex of the projected triangle, and the edge function initial values edge0, edge1, and edge2 for the three edges of the triangle and the interpolation parameters dE0dx, dE1dx, dE2dx, dE0dy, dE1dy and dE2dy are calculated by the setup engine as follows and entered into the rendering engine.

Here, sx and sy are the difference between the value in the real world coordinate of the triangle and the value in the screen coordinate, and Vx and Vy are the vertex coordinate values of the triangle in the screen coordinate.

In Edge Walking, which follows the left edge of the triangle and finds the starting point of the span, the edge function is interpolated with the interpolation parameters for each walking step starting from the initial value of the edge function and then added to the sign of the interpolated value. Accordingly, the inside and outside of the triangle are determined, and the next direction of the walking step is determined.

The direction of edge walking proceeds as shown in FIG. 2A. Each time one span is advanced, the edge 0 function interpolation value in the corresponding direction is added to the initial edge 0 function value, and the sign of the edge 0 function value is checked. do. All function value interpolators (color, xy-coordinates, depth value, texture-coords, alpha-value, fog-value) and edge 1, edge Each time 2 degrees edge 0 advances by one span, the corresponding interpolation value in the corresponding direction is added. If the value of edge 0 calculated as the result of interpolation is negative (i.e. outside the triangle), the direction of travel is shifted downward. Edgeworking is performed along edge1 starting from the middle vertex. At this time, the inside and outside of the triangle is checked according to the sign of the function value of edge1. Edge 1 proceeds to the right as opposed to edge 0 and then turns downward if the function value of edge 1 is positive (ie inside the triangle).

As a result, one span is calculated each time the edge worker goes downward, and the span data is input to the span interpolator to generate a function value of each pixel present on the span. In span interpolation, as in edge walking, each edge of a triangle is checked using an edge function value. When the middle vertex is a clockwise triangle to the right of the longest edge as shown in FIG. The span is interpolated by the function value of edge 2 and the lower span is interpolated by the function value of edge 1 and the right side of the triangle is determined by checking the sign of the result value. The arrow indicates the span interpolation direction, and span interpolation is made up to the part where the clipping occurs by the screen s.

However, the conventional triangular scanning method has a large amount of data input from the setup engine to the rendering engine to render a single object, and especially when the multifunction is implemented in a single chip such as a media chipset, the internal data bus has many functions. Because they are shared with the logic block of the bus, the bus mastering time allocated to each other is limited. As a result, large amounts of data require more bus mastering, which degrades the overall system performance. In addition, due to the triangular shape of the triangle, frequent page misses may occur when the memory used as the frame buffer is accessed. When a page loss occurs, the memory needs a precharge time to activate a new row, which lengthens the data access time. In order to minimize the data access time, desired data must exist on the same row, but data may not exist on the same row due to the characteristics of a triangle.

The technical problem to be achieved by the present invention is to consider the triangles sharing one edge as a rectangle and scan the pixel data generated as a result of the rasterization while reducing the overall data that must be input from the setup engine to the rendering engine in order to rasterize an object. To provide a rasterization method using a rectangle to reduce the memory access time for the storage of.

1 is a block diagram of a general three-dimensional graphics system.

2A illustrates an edge walking method using an edge function value.

2B illustrates a span interpolation process using edge function values.

3 is a flowchart illustrating a rasterization method using a rectangle according to the present invention.

4A and 4B illustrate a rectangular shape to which the present invention is applied.

4C and 4D illustrate rectangular shapes to which the present invention is not applied.

According to an aspect of the present invention, there is provided a rasterization method using a quadrangle, comprising: extracting a quadrangle consisting of two triangles sharing one edge from an object image projected on screen coordinates; If each of the vertical coordinate coordinates of the two vertices shared by the two triangles extracted from the screen coordinates is present between the vertical coordinate coordinates of the other two vertices not shared by the two triangles, the four edges of the rectangle and the two triangles are Calculating an edge function value for a shared shared edge; Obtaining a span starting point at the left edge of the quadrangle; And performing a span interpolation in the horizontal direction from the obtained edge function value and the span starting point, and performing a span interpolation for the other triangle when the span passes through the shared edge and proceeds from one triangle to another. It is characterized by including.

Hereinafter, the present invention will be described in more detail with reference to the accompanying drawings. 3 is a flowchart illustrating a rasterization method using a rectangle according to the present invention. The rasterization method using the quadrangle according to FIG. 3 includes extracting two triangles sharing one edge (300), determining whether the rasterization is possible with the square (302), and rasterizing the square (306 to 314). And rasterizing into triangles 304.

The operation will be described in more detail as follows. First, the object to be rendered is composed of triangles having three vertices, projected to screen coordinates, and two triangles that share one edge, that is, four vertices are extracted from the projected object image (step 300).

The y-coordinate values of both endpoints (hereinafter referred to as shared vertices) of the edges shared by the extracted two triangles are compared with the y-coordinate values of each vertex not shared by the two triangles (step 302).

As a result of the comparison, if the two y coordinate values of the shared vertices are larger or smaller than the y coordinate values of the non-shared vertices, it is determined that it is not possible to erase the rectangles, and the two triangles are rasterized separately (step 304). An example of this is shown in FIGS. 4A and 4B. That is, when both y-coordinate values of the shared vertices are larger than the y-coordinate values of the non-shared vertices as shown in FIG. 4A, when span interpolating from the left edge v1s1, the rectangles (v1, s1, v0, s0) Span interpolation occurs inside of the module. Similarly, as shown in FIG. 4B, when both y-coordinate values of the shared vertices are smaller than the y-coordinate values of the non-shared vertices, the inside weight of the rectangle v0, s0, v1, s1 during span interpolation from the left edge v0s0 is shown. A part where span interpolation is not performed occurs. In this case, the two triangles are rasterized separately. That is, rasterization is performed by performing edge walking and span interpolation on each of the two triangles.

As a result of the comparison in step 302, if the two y-coordinate values of the shared vertices exist between the y-coordinate values of the non-shared vertices, it is determined that rasterization is possible in the rectangle, and the rasterization is performed in the rectangle (steps 306 to 314). . This case is illustrated in FIGS. 4C and 4D. In order to rasterize the two triangles into squares, the edge function values for the four edges constituting the square and the shared edges shared by the two triangles are obtained by a known method (step 306). From the obtained edge function value, the span starting point is found while edgeworking the left edge of the rectangle (step 308). When the span starting point is found, the span and edge function values are interpolated with respect to the shared edge s0s1 and the right edge s1v1 of FIG. 4C, and function values of each pixel included in the span are generated (step 310). When span interpolation is described with reference to FIG. 4C, the triangle t0 is the span interpolation of the clockwise triangle as shown in FIG. 2B, and the t1 is the interpolation as shown in FIG. 2A so that the common edge s0s1 and the right edge ( Two function values for s1v1) are interpolated. As a result of span interpolation, when the span passes through the shared edge (s0s1) (step 312), the interpolated edge function value is converted from a positive value to a negative value, from which the function value, e.g., For example, values such as edge function values, colors, depth information, and the like are calculated (step 314).

According to the present invention, since five edge function values are input to the rendering engine to rasterize two triangles, data transmission of about 17% can be reduced. This can reduce bus ownership time when creating chipsets that share an internal data bus with other logic blocks. It also allows efficient memory access. In other words, assuming that the triangle to be rasterized consists of approximately 50 pixels and the span consists of 3 to 5 pixels in length, after 3 to 5 memory read / write operations, page loss of the memory row occurs, driving another row. Although precharge time is required, the square rasterization method according to the present invention reduces the precharge time by about 20 to 50%.

Claims (2)

A first step of extracting a quadrangle consisting of two triangles sharing one edge from an object image projected on screen coordinates; If each of the vertical coordinate coordinates of the two vertices shared by the two triangles extracted from the screen coordinates is present between the vertical coordinate coordinates of the other two vertices not shared by the two triangles, the four edges of the rectangle and the two triangles are Calculating an edge function value for a shared shared edge; Obtaining a span starting point at the left edge of the quadrangle; And And a fourth step of span interpolating in the horizontal direction from the obtained edge function value and the span starting point, wherein the span interpolates with respect to the other triangle when the span passes through the shared edge and proceeds from one triangle to another. Rasterization method using a square characterized in that. The method of claim 1, wherein the second step If each of the vertical coordinates of the two vertices shared by the two triangles does not exist between the vertical coordinates of the other two vertices that the two triangles do not share, the edge function value for each edge for each of the two triangles is determined. And obtaining a span starting point at the left edge of each triangle, and performing a span interpolation in the horizontal direction from the obtained edge function value and the span starting point.

Images