This is an article I originally published on everything2.com in 2003; it subsequently became the basis (with my permission) for articles on Wikipedia and the Doom Wiki. The original text is below.
The Doom rendering engine is an interesting study in software rendering. It is not a true “3D” engine (as it is not possible look up and down properly), but is however a fairly elegant system that allows pseudo-3D rendering. In its time, Doom was revolutionary and almost unique in its providing a fast texture-mapped 3D environment.
Viewed from the top down, all Doom levels are really 2D, demonstrating one of the key limitations of the Doom engine: it is not possible to have “rooms above rooms”.
The base unit is the vertex, which represents a single 2D point. Vertices (or “vertexes” as they are referred to internally) are then joined to form lines, known as “linedefs”. Each linedef can have either one or two sides, these are known as “sidedefs”. Sidedefs are then grouped together to form polygons; these are called “sectors”. Sectors represent particular areas of the level.
Each sector contains a number of properties: a floor height, ceiling height, light level, a floor texture and a ceiling texture. To have a different light level in a particular area, for example, a new sector must be created for that area with a different light level. One-sided linedefs therefore represent solid “walls”, while two-sided linedefs represent “bridge” lines between sectors.
Sidedefs are used to store wall textures; these are totally separate from the floor and ceiling textures. Each sidedef can have three textures; these are called the middle, upper and lower textures. In one-sided linedefs, only the “middle” texture is used for the texture on the wall. In two-sided linedefs, the situation is more complex. The lower and upper textures are used to fill the gaps where adjacent sectors have different floor and ceiling heights: lower textures are used for steps, for example. The sidedefs can have a middle texture as well, although most do not; this is used to make textures “hang” in mid air. For example, when a transparent bar texture is seen forming a cage, this is an example of a middle texture on a two-sided linedef.
Finally, there is a list of objects in the level; objects are known as “things” by Doom. These are used to place players, monsters, powerups, etc. Each thing is given a 2D coordinate, as with the vertices. Things are then automatically placed on the floor or the ceiling depending on their type.
This is only an overview of the basic structure of Doom levels; most of the data structures here carry extra properties, for example, texture offsets, and special properties for gameplay. These are omitted here as they are not particularly relevant to this discussion.
Doom makes use of a system known as Binary Space Partitioning (BSP). A tool is used to generate the BSP data for a level beforehand. Depending on the size of the level, this process can take quite some time. It is because of this that it is not possible to move the walls in Doom; while doors and lifts move up and down, none of them ever move sideways.
The level is divided up into a binary tree: each location in the tree is a “node” which represents a particular area of the level (with the root node representing the entire level). At each branch of the tree there is a dividing line which divides the area of the node into two subnodes. At the same time, the dividing line divides linedefs into line segments called “segs”.
At the leaves of the tree are convex polygons, where it is not useful to divide the level up any further. These convex polygons are called “ssectors” (subsectors), and are bound to a particular sector. Each ssector has an associated list of segs associated with it.
The BSP system is really a very clever way of sorting the ssectors into the right order for rendering. The algorithm is fairly simple:
In this way, it is possible to always draw the far away ssectors before the close up ones. It is also possible to exclude large parts of the level which cannot be seen: each node has a 2D bounding box. If that bounding box is totally outside the field of view, it is not drawn.
All walls in Doom are drawn vertically; it is because of this that it is not possible to properly look up and down. It is possible to perform a form of look up/down via “y-shearing”, and many modern Doom source ports do this. Essentially this works by increasing the vertical resolution, and then providing a “window” on that space. By moving the window up and down, it is possible to give the illusion of looking up and down. However, this tends to distort the view the further up and down the player looks.
The Doom engine renders the walls as it traverses the BSP tree, drawing ssectors by order of distance from the camera so that the closest segs are drawn first. As the segs are drawn, they are stored in a linked list. This is used to clip other segs rendered later on, reducing overdraw. This is also used later to clip the edges of sprites.
Once the engine reaches a solid (1-sided) wall at a particular x ordinate, no more lines need to be drawn at that area. For clipping the engine stores a “map” of areas of the screen where solid walls have been reached. This allows far away parts of the level which are invisible to the player to be clipped completely.
The Doom graphic format stores the wall textures as sets of vertical columns; this is useful to the renderer, which essentially renders the walls by drawing lots of vertical columns of texture.
The system for drawing floors and ceilings (“flats”) is less elegant than that used for the walls. Flats are drawn with a flood-fill like algorithm. Because of this, it is sometimes possible if a bad BSP builder is used to get “holes” where the floor or ceiling bleeds down to the edges of the screen. This is also the reason that if the player travels outside of the level using the noclip cheat, the floors and ceilings appear to stretch out from the level over the empty space.
The floor and ceiling are drawn as “visplanes”. These represent horizontal runs of texture, from a floor or ceiling at a particular height, light level and texture (if two adjacent sectors have the exact same floor, these can get merged into one visplane). Each x position in the visplane has a particular vertical line of texture which is to be drawn.
Because of this limit of drawing one vertical line at each x position, it is sometimes necessary to split visplanes into multiple visplanes. For example, consider viewing a floor with two concentric squares. The inner square will vertically divide the surrounding floor. In that horizontal range where the inner square is drawn, two visplanes are needed for the surrounding floor.
This leads to one of Doom's classic limitations which frustrated many mappers for a long time. Doom contained a static limit on the number of visplanes; if exceeded, it would crash back to DOS with the message, “No more visplanes!”. The easiest way to invoke the visplane limit is a large checkerboard floor pattern; this creates a large number of visplanes.
As the segs are rendered, visplanes are also added, extending from the edges of the segs towards the vertical edges of the screen. These extend until they reach existing visplanes. Because of the way this works, the system is dependent on the fact that segs are rendered in order by the overall engine; it is necessary to draw close up visplanes first, so that they can “cut off” by the further away ones. If unstopped, the floor or ceiling will “bleed out” to the edges of the screen, as previously described. Eventually, the visplanes form a “map” of particular areas of the screen in which to draw particular textures.
While visplanes are constructed essentially from vertical “strips”, the actual low level rendering is performed in the form of horizontal “spans” of texture. After all the visplanes have been constructed, they are converted into spans which are then rendered to the screen. This appears to be a tradeoff: it is easier to construct visplanes as vertical strips, but because of the nature of how the floor and ceiling textures appear it is easier to draw them as horizontal strips. Because of the nature of visplanes, the conversion is fairly trivial, however.
Each sector within the level has a linked list of things stored in that sector. As each sector is drawn the sprites are placed into a list of sprites to be drawn. If not within the field of view these are ignored.
The edges of sprites are clipped by checking the list of segs previously drawn. Sprites in Doom are stored in the same column based format as the walls are, which again is useful for the renderer. The same functions which are used to draw walls are used to draw sprites as well.
While ssectors are guaranteed to be in order, the sprites within them are not. Doom stores a list of sprites to be drawn (“vissprites”) and sorts the list before rendering. Far away sprites are drawn before close ones. This causes some overdraw but usually this is negligible.
There is a final issue of middle textures on 2-sided lines, used in transparent bars for example. These are mixed in and drawn with the sprites at the end of the rendering process, rather than with the other walls.
The Doom Source, as released by Id Software.