maszyna/docs/opengl46-renderer.md

OpenGL 4.6 renderer

The default renderer (gfxrenderer default, the modern backend, rendering/opengl46/) runs on an OpenGL 4.6 core profile context. There is no fallback to an older version, and the OpenGL 2.1 (legacy) renderer and the OpenGL ES / ANGLE path no longer exist.

For how the lights are handled, see clustered-lighting.md.

Requirements

• A graphics card and driver that provide OpenGL 4.6 (core profile). What the driver offers is written to the log at start: Renderer:, Vendor: and GL version:.
• A CPU with AVX2, as before.
• Windows and Linux. macOS provides OpenGL only up to 4.1, so the renderer does not run there.
• The experimental Better Renderer (gfxrenderer experimental) is a separate renderer and is not affected.

On Linux, glxinfo -B should report an OpenGL core profile version string of 4.6 or newer.

If the simulator does not start

When the context cannot be created, the log (log.txt) contains a line like

Bad init: failed to create OpenGL 4.6 core profile context (<GLFW error>); the graphics card/driver provides: OpenGL X.Y ..., <renderer> (<vendor>)

On Windows the same text is shown in a message box. Updating the graphics driver is the usual fix; a card which does not support OpenGL 4.6 cannot run this renderer. The process ends with a non-zero exit code.

eu07.ini

Removed options

Old eu07.ini files still load. These entries are read and ignored, with a line in the log:

Entry	What happens now
gfxrenderer legacy, simple or any other value except default, full and experimental	the default renderer is used (eu07.ini: gfxrenderer "X" is no longer supported, using "default")
usevbo	ignored (eu07.ini: "usevbo" is no longer supported and was ignored)
gfx.usegles	ignored, same message
gfx.angleplatform	ignored, same message

gfxrenderer full is still taken for default.

New options

Entry	Default	Meaning
gfx.lights.clustered yes/no	yes	yes: every lit vehicle headlight within 1 km lights the surroundings (see below). no: the shading goes through the first dynamiclights headlights only, as the old renderer did; useful to compare images or to work around a problem
gfx.lights.clusterheatmap yes/no	no	debug view: the number of lights in each cluster, drawn over the image

Sky light and exposure

All of these are optional; a missing entry takes the default. The switches can also be changed while the simulator runs, in the Sky and exposure debug window. With every switch set to no the image is as before these options existed.

Entry	Default	Meaning
gfx.ambient.sh yes/no	yes	the ambient light depends on where a surface faces: more from the bright part of the sky, less from the ground below. The average stays the ambient light the simulation works out
gfx.ambient.sh.strength	1.0	0..2: how much the ambient light differs between directions (0 = uniform, as before)
gfx.ambient.groundalbedo	0.1	0..1: how much light the ground below the horizon gives back (to the directional ambient light and to the atmosphere)
gfx.envmap.ggx yes/no	yes	rough reflections read the environment map prefiltered with the GGX lobe (made in a compute pass after each update of the map) instead of its plain mipmaps
gfx.atmosphere yes/no	yes	the sky, the colour of the sunlight, the ambient light and the fog colour come from a physical model of the atmosphere (Hillaire 2020) during the day; the old sky takes over between 10 degrees of sun elevation and 6 degrees below the horizon, and at night. Render only: the simulation keeps its own values
gfx.atmosphere.sky	1.0	brightness of the drawn sky and of the fog colour (1 = calibrated against the old sky)
gfx.atmosphere.ambient	1.0	brightness of the ambient light from the atmosphere
gfx.atmosphere.sun	1.0	brightness of the sunlight from the atmosphere
gfx.atmosphere.haze	1.0	amount of aerosols on a clear day (whiter, hazier sky); the weather adds to it with the overcast
gfx.atmosphere.altitude	200	height of the scenery above the sea level, metres
gfx.postfx.autoexposure yes/no	yes	the exposure follows the average brightness of the image (luminance histogram, compute pass)
gfx.postfx.autoexposure.key	0.3	average luminance of the image the exposure is set for
gfx.postfx.autoexposure.min	-1.0	the most the exposure darkens a bright image, EV
gfx.postfx.autoexposure.max	0.5	the most the exposure brightens a dark image, EV. Kept low on purpose: tunnels and the night are not meant to be brightened much
gfx.postfx.autoexposure.speed.dark	1.0	rate of adaptation to a darker image, 1/s
gfx.postfx.autoexposure.speed.bright	3.0	rate of adaptation to a brighter image, 1/s
gfx.postfx.autoexposure.compensation	0.0	fixed exposure correction, EV; applies also with the automatic exposure off
gfx.postfx.bloom yes/no	yes	glow around bright parts of the image
gfx.postfx.bloom.intensity	0.04	share of the glow in the image
gfx.postfx.bloom.threshold	0.0	luminance below which nothing glows; 0 = everything contributes a little (physically based)

The automatic exposure and the bloom need the post processing pipeline (gfx.skippipeline no, the default). Model previews of the editor use none of these.

Ambient occlusion

Darkening of creases, corners and the ground under objects, worked out from the depth of the image (screen space) and multiplied into the image before the motion blur and the tonemapping. It works with and without motion blur. It darkens all light alike, the sun and the headlights included; limiting it to the ambient light needs the depth before the shading (a depth prepass), which is planned later.

Entry	Default	Meaning
gfx.postfx.ao gtao/ssao/no	gtao	gtao (or yes): ground truth ambient occlusion (Jimenez et al. 2016), at half the resolution with a depth aware smoothing and upsampling. ssao: the older, experimental ssao. no: none, and no targets are made for it; the mode can then not be turned on in the debug window
gfx.postfx.ao.radius	1.5	0.1..10: reach of the occluders, metres
gfx.postfx.ao.strength	1.0	0..4: power the occlusion is raised to; 1 = as computed, 0 = none
gfx.postfx.ao.bounce	0.3	0..1: albedo of the surfaces assumed for the light that comes back after more bounces (brightens the occlusion a little)
gfx.postfx.ao.quality slices steps	3 4	directions searched around each pixel (1..8) and samples on each side of a direction (1..16)

Lamps

With the clustered lights on, lamps light their surroundings, besides the sun and the vehicle headlights. Without the clustered lights (gfx.lights.clustered no) none of this happens and the cab is lit as before. The switches and scales can also be changed in the Light clusters debug window.

• Light points of models (freespotlight submodels: street and platform lamps, signals, lamps of vehicles) are taken the way the experimental renderer takes them. A point which is visible, has self-illumination (SelfIllum true, or a threshold the light level is below) shines as a spot light along the -z axis of its submodel, in the cone of its glare (FalloffAngle, full brightness within HotspotAngle), in its colour; hotspotpower scales it. Lamps of scenery models switched with Light_On / Light_Off (lights of the model node, events) turn their light points on and off with them. Nothing in the models has to change; a light point without self-illumination stays a glare only.
• Lit compartments of vehicles (the cab, corridor/korytarz and compartment/przedzial sections of the low poly interior) give a point light in the middle of the section, in the colour of the interior light, unless the section holds light points of its own (these light it then, through its self-illumination).
• The cab lamp of the occupied cab is a point light under the top of the cab model. Before, it added to the ambient light of the whole cab and weakened the shadows of the sun in it; now a share of that ambient light is kept for the light which bounces around the cab (gfx.lights.cab.bounce), and the sun shadows stay.

The lamps of models and compartments fade in from the light level at which scenery lamps switch on (ls_Dark) to full darkness; in daylight they are not gathered at all. The cab lamp and the lamps of scenery models nearest to the camera cast shadows (see below); the others (lamps of vehicles, compartments, farther scenery lamps) also light what is behind walls within their reach, which is kept short for compartments.

Shadows of the cab lamp (gfx.lights.shadows): the cab casts shadows in the light of its lamp, the desk, the seats, the levers and the walls (light leaves the cab only through the windows, glass doesn't stop it). They come from a cube shadow map around the lamp, drawn with the opaque parts of the cab model only. The map is kept in the space of the cab, so it stays right as the vehicle moves; it is drawn anew every few frames, for the moving parts of the cab, and at once when the lamp turns on or the cab changes. All six faces are drawn in one go, so drawing it costs about as much as drawing the cab once more. Nothing is drawn with the lamp off (in the daylight it is). The level:

Level	Map	Drawn	Edges
no (0)	none	never	lamp without shadows, as before
low (1)	512 x 512 per face, 6 MB	every 4th frame	hard (2x2 filter)
medium (2, default)	1024 x 1024 per face, 24 MB	every 2nd frame	soft, 4 samples
high (3)	1024 x 1024 per face, 24 MB	every frame	softer, 8 samples

Shadows of scenery lamps (with gfx.lights.shadows above no): the street, platform and other lamps of scenery models nearest to the camera (gfx.lights.shadows.lamps of them, within gfx.lights.shadows.lamps.range of the camera) cast the shadows of the scenery around them, of the vehicles near the camera and of the occupied cab: light from a lamp outside comes into the cab only through the windows. Each such lamp has a cube shadow map in world axes around it, in two parts:

• the scenery (terrain, tracks, models; not the model of the lamp itself, which would shadow everything), drawn once when the lamp gets its map and again only when the scene changes (events showing or hiding models, the editor). At most one map is drawn per frame (two at high), so lamps coming into reach get their shadows over a few frames;
• a copy of the scenery with the vehicles within gfx.lights.shadows.vehicles.range of the camera (and the cab) added, made every few frames (as for the cab lamp) while any of them is within the reach of the lamp. A vehicle farther from the camera casts no shadow of the lamps; with no vehicle near a lamp nothing is drawn for it.

Lamps of vehicles and lit compartments cast no shadows. The memory taken is 2 maps per lamp: with the defaults (8 lamps, 512 x 512 per face) about 100 MB of video memory.

Entry	Default	Meaning
gfx.lights.scenery yes/no	yes	light points of scenery models light their surroundings
gfx.lights.vehicles yes/no	yes	light points of vehicles and their lit compartments light their surroundings
gfx.lights.cab yes/no	yes	the cab lamp is a point light; no: it adds to the ambient light of the cab, as before
gfx.lights.freespot.intensity	10	brightness of a light point (the experimental renderer uses the same default)
gfx.lights.compartment.intensity	1.0	brightness of the light of a lit compartment
gfx.lights.cab.intensity	1.5	brightness of the cab lamp
gfx.lights.cab.bounce	0.3	0..1: share of the former ambient light of the cab lamp kept with the lamp on
gfx.lights.range	300	distance from the camera the lamps of models and vehicles are gathered from, metres
gfx.lights.shadows	medium	no/low/medium/high (or 0..3): shadows of the cab lamp and the scenery lamps, see above
gfx.lights.shadows.lamps	8	0..32: how many scenery lamps nearest to the camera cast shadows; 0: none
gfx.lights.shadows.lamps.size	512	128..2048: size of a face of the shadow map of a scenery lamp
gfx.lights.shadows.lamps.range	100	distance from the camera within which scenery lamps cast shadows, metres
gfx.lights.shadows.vehicles.range	50	distance from the camera within which vehicles (and the occupied cab) cast the shadows of scenery lamps, metres; 0: only the scenery does

Models drawn from the GPU

Many models of the scenery (trees, bushes, poles, fences: models without lights and animations) are drawn in batches, one batch per model and level of detail. Which of them are drawn, and at which level of detail, can be worked out on the processor (as before) or on the graphics card. On the graphics card the processor no longer goes through the models one by one, so with dense vegetation it is much less loaded. Only the main view is drawn this way; shadows, reflections and other views go as before.

Entry	Default	Meaning
gfx.gpucull	no	no: on the processor, as before. yes: on the graphics card, and the models are drawn from its result. compare: on the processor, which draws, and on the graphics card too, for the results to be compared (debugging)
gfx.gpucull.occlusion yes/no	yes	with gfx.gpucull yes: models hidden whole behind what was drawn before them (the cab, the terrain, buildings and other models drawn on their own, vehicles) are left out
gfx.depthprepass yes/no	yes	with gfx.gpucull yes: the depth of the models is drawn first, so the colour is worked out only for what stays in view (dense foliage)

The occlusion test does not take models drawn this way as covering each other: a tree hidden behind another tree is still drawn (its colour is not worked out with the depth prepass on). Models whose material is not one of the usual ones (parallax, water and the like) are left out of the depth prepass and drawn as before. All three can be changed while the simulator runs, in the Instance culling debug window.

Changed option

dynamiclights (0..7) used to limit the number of headlight sets that light the surroundings. It is still read, but it only takes effect with gfx.lights.clustered no.

What looks different from the previous renderer

• No limit of 7 lit headlights. All lit headlights within 1 km take part in the lighting (each lit end of a vehicle counts separately), up to 1023 lights in the buffer (headlights, lamps and the cab lamp) and 64 reaching any one part of the view. Beyond these capacities the farthest and dimmest lights are dropped, and the statistics show how many.
• Outside the beam texture a headlight adds nothing. Before, the edge of the texture was stretched beyond the beam, which could leave a faint glow around vehicles. This also applies with gfx.lights.clustered no.
• Reflections (environment map) are lit with the sun as seen from the reflection itself, and vehicle headlights no longer light the scene in them.
• Motion blur and ambient occlusion read the real depth of the scene. Before, the depth texture was read in a way the specification leaves undefined, and the result depended on the driver.
• Lines are 1 pixel wide, as they always were on this kind of context. The widths the code asked for (catenary wires, map tracks) never took effect.

The editor (F11, orthographic view), the reflection pass and the model previews do not use clusters. The editor goes through the first dynamiclights headlights, the reflections through the sun only.

Statistics and debugging

• Shift+F12, Gfx Renderer: the lights: row gives the number of lit headlight sets, how many did not fit, and clustered or first N. With clusters on, clusters: (rebuilding of the grid, which happens when the projection or the draw range change) and cull: (assigning lights to clusters, every frame) show the GPU time, and cluster lists: the size of the lists and the number of dropped (cluster, light) pairs (overflow, with its peak). A non-zero overflow means some cluster held more than 64 lights.
• Light clusters window (in debug mode, menu Windows, item Headlight config opens the renderer debug windows): the switch for clustered lights, the heat map, the grid state and the same counters, and the switches and scales of the lamps with their counts. Shift+F12 shows the lamps in the local lights: row: how many were gathered (from scenery models, vehicles and compartments, and the cab lamp), how many were placed in the light buffer of the main view (those that may light something in view) and did not fit, and the cpu time of gathering them. The lamp shadows: row gives the size of the cube map, whether it holds the cab lamp, the draw calls and cpu time of its last drawing, and the timings the GPU time of drawing it (lamp shadows:); the level can be changed in the window. The scenery lamp shadows: row gives the lamps with maps (of the slots, how many ready, how many with vehicles), how many scenery parts and vehicle copies were drawn so far, and the draw calls and cpu time of the last frame which drew any; the timings show their GPU time (scenery lamp shadows:). The count and both ranges can be changed in the window too.
• Sky and exposure window (next to the Light clusters window): the switches and scales above, the share of the atmosphere, the colours of the sun, ambient light and fog in use, and the measured luminance and exposure. Shift+F12 shows the GPU time of atmosphere: (look-up tables, only when the sun moved or the weather changed), envmap ggx: (prefiltering, with each update of the environment map), exposure: and bloom:. The same window holds the ambient occlusion mode and its settings; Shift+F12 shows its time as ambient occlusion: (gtao only).
• Instance culling window: the mode of gfx.gpucull and the two switches above. With gfx.gpucull yes Shift+F12 shows gpu cull: (models drawn and hidden by the occlusion test, as counted on the graphics card a few frames before, the levels of detail, the draw calls with those of the depth prepass, and the processor time of the whole step) and the timings instance cull:, occlusion depth: (the depth pyramid of the test) and depth prepass:. With compare it shows how many models both counted and how many levels of detail differ.
• gfx.gldebug yes turns on the OpenGL debug output (KHR_debug); messages go to the log.

Building

-DWITH_OPENGL_MODERN=ON (default) builds this renderer; WITH_OPENGL_LEGACY and the GLES / ANGLE options are gone. Linux builds still use -DWITH_BETTER_RENDERER=OFF.

All shaders in shaders/ can be checked without a graphics card against GLSL 4.60, compute stage included:

ci_shadervalidator/build.sh
cd ci_shadervalidator && ./validateshaders

It needs g++ with C++20, glslangValidator (Debian/Ubuntu: glslang-tools) and the ref/glm submodule. The same check runs in CI for changes to shaders and the gl/ wrappers.