public abstract class GraphicsDevice

Sets the active camera supplying the view and projection matrices used by subsequent draws.

Setting the camera starts a new frame as far as shadows go: a shadow map rendered before it is not used again, so set the camera first and then run beginShadowPass.

Starts rendering the light’s shadow map. Until endShadowPass, draw writes only depth as seen from the light, and skips materials that do not cast shadows. Draws after endShadowPass that receive shadows sample the map.

The default implementation supports no shadows and returns false; the caller then skips the pass and draws as usual.

Render the shadow pass before beginPostProcess, never inside it: both passes return to the target that was active when they began, so a shadow pass started offscreen would hand the post process its own image to draw into.

Creates an offscreen image the scene can be drawn into with beginRenderTarget and then sampled like any other texture: a mirror, a security camera screen, a minimap, the refraction behind glass.

The result is an ordinary Texture; hand it to a material or to a ShaderMaterial texture slot, and release it with dispose(Texture). It has a depth buffer of its own while it is drawn into.

The default implementation cannot draw offscreen and returns null, as the software simulator device does.

Sends the draws that follow into target until endRenderTarget, clearing its colour to clearArgb and its depth first. The viewport covers the whole target, and the camera, light and environment are the device’s as usual, so set a camera for the view the target shows.

Draw render targets first in a frame: before the shadow pass, before beginPostProcess and before the scene itself. They do not nest, and on some devices the depth of scene draws made earlier in the frame is not kept across one.

Redirects drawing into an offscreen image so endPostProcess can apply effects to it on its way to the screen. When effects has custom passes (PostProcess.addPass) and this device can draw into a render target, the scene goes into one, and endPostProcess runs the passes before the built-in effects.

Draw the shadow pass and any render targets before this call.

The device’s own post processing: redirects drawing into an offscreen image so endPostEffects can apply the built-in effects of effects. Custom passes are handled by beginPostProcess and never reach this.

The default implementation supports no post processing and returns false; the caller then draws straight to the screen.

Draws count copies of a mesh, each with its own transform and, optionally, colour: crates along a quay, a field of grass, a flock. The material’s shader tells the copies apart with ShaderNode.instanceId, instanceRandom and instanceColor, so one material varies size, tint and motion across thousands of copies.

Every device draws the copies one after another from a single compiled pipeline, changing only the transform and the per copy uniforms between them, which is what makes it cheaper than a model per copy: no state is looked up or rebound and the scene walks one node.