Implement multiple lights and more Light types #3246
Conversation
ffreyer
added
the
GLMakie
ffreyer
force-pushed
the
ff/lighting
branch
from
0e81e35
to
f68ca35
Compare
|
I'm following the attenuation from learnopengl/ogre3d for now, which is effectively fig = Figure(backgroundcolor = :black)
sl = Slider(fig[1, 1], range = 0.05:0.05:5.0)
lights = Makie.AbstractLight[
AmbientLight(RGBf(0.1, 0.1, 0.1)),
PointLight(RGBf(1,0,0), Point3f(2, 2, 0), 1.0),
PointLight(RGBf(0,1,0), Point3f(-2, 2, 0), 10.0),
PointLight(RGBf(0,0,1), Point3f(-2, -2, 0), 100.0),
PointLight(RGBf(1,1,1), Point3f(2, -2, 0), 10.0),
]
scene = LScene(
fig[2, 1], show_axis = false,
scenekw=(camera = cam3d!, lights = lights, backgroundcolor = :black, center = false),
)
p = mesh!(scene, Rect3f(Point3f(-10, -10, 0.01), Vec3f(20, 20, 0.02)), color = :white)
map(z -> translate!(p, 0, 0, -z), sl.value)
update_cam!(scene.scene, Vec3f(0, 0, 10), Vec3f(0, 0, 0), Vec3f(0, 1, 0))
idxs = vcat(1:length(sl.range[]), length(sl.range[])-1:-1:2)
record(fig, "attenuation.webm", idxs, framerate = 30) do idx
sl.selected_index[] = idx
endattenuation.webmI also tried just lights = [
AmbientLight(RGBf(0.1, 0.1, 0.1)),
PointLight(RGBf(0,1,0), Point3f(0, -3, 1)),
PointLight(RGBf(1,0,0), Point3f(-3, 0, 1)),
PointLight(RGBf(0,1,0), Point3f(0, 3, 1)),
PointLight(RGBf(1,0,0), Point3f( 3, 0, 1)),
Makie.DirectionalLight(RGBf(0, 0, 0.5), Vec3f(0, 0, 1)),
]
scene = Scene(camera = cam3d!, lights = lights)
mesh!(scene, Rect3f(Point3f(-1), Vec3f(2)), color = :white)
r = range(-10, 10, length=101)
surface!(scene, r, r, [0.5 * (sin(x+y) * sin(2x) * sin(3y)) - 1 for x in r, y in r], colormap = [:white, :white])
ps = [light.position[] for light in lights if light isa PointLight]
scatter!(scene, ps, color = :yellow, strokewidth = 1, strokecolor = :black)
screen = display(scene)
|
Compile Times benchmarkNote, that these numbers may fluctuate on the CI servers, so take them with a grain of salt. All benchmark results are based on the mean time and negative percent mean faster than the base branch. Note, that GLMakie + WGLMakie run on an emulated GPU, so the runtime benchmark is much slower. Results are from running: using_time = @ctime using Backend
# Compile time
create_time = @ctime fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @ctime Makie.colorbuffer(display(fig))
# Runtime
create_time = @benchmark fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @benchmark Makie.colorbuffer(display(fig))
|
GLMakie/src/drawing_primitives.jl
Outdated
| updater = screen.render_tick | ||
| MAX_PRIORITY = typemax(Int) | ||
|
|
||
| attr[:lights_length] = map(updater, priority = -1000) do _ |
There was a problem hiding this comment.
Updating on render_tick doesn't work with record.
Updating (many) lights is kind of complicated in general, because we want to react to updates of each field for each light and also when lights are added or deleted to the lights vector.
Listening to each field individually would require attaching and removing listeners when the lights vector changes. Detecting that would either require some kind of Vector wrapper or an update on draw.
Instead of bothering with all of that I thought it would be easier to just check the lights vector once per frame and update it then. I'm using render_tick for that atm, but I guess we don't want that to update in record to avoid triggering events?
There was a problem hiding this comment.
I adjusted colorbuffer to trigger render_tick now. That seems to be fine for rendering beyond screen resolution and will be useful if we add an on_render_start event in the future.
Moving, colored `PointLight`s with different attenuation + white SpotLight.begin
fig = Figure(backgroundcolor = :black)
sl = Slider(fig[1, 1], range = 0.05:0.05:5.0)
# Prepare lights
angle = Observable(0)
angle2pos(phi) = Point3f(3cosd(phi), 3sind(phi), 0)
lights = Makie.AbstractLight[
AmbientLight(RGBf(0.1, 0.1, 0.1)),
PointLight(RGBf(1,0,0), map(phi -> angle2pos(phi + 0), angle), 10.0),
PointLight(RGBf(0,1,0), map(phi -> angle2pos(phi + 120), angle), 100.0),
PointLight(RGBf(0,0,1), map(phi -> angle2pos(phi + 240), angle), 1000.0),
SpotLight(RGBf(1,1,1), Point3f(0, 0, 1), Vec3f(0, 0, -1), pi/6),
]
# Set scene
scene = LScene(
fig[2, 1], show_axis = false,
scenekw=(camera = cam3d!, lights = lights, backgroundcolor = :black, center = false),
)
# floor
p = mesh!(scene, Rect3f(Point3f(-10, -10, 0.01), Vec3f(20, 20, 0.02)), color = :white)
map(z -> translate!(p, 0, 0, -z), sl.value)
# Random balls
N_balls = 2_000
random_colors = !true
meshscatter!(
scene, [Point3f(20rand()-10, 20rand()-10, 10rand()-5) for _ in 1:N_balls],
color = random_colors ? rand(RGBf, N_balls) : (:white)
)
# Visualize light source positions
ps = Observable([light.position[] for light in lights if light isa Union{PointLight, SpotLight}])
cs = [light.color[] for light in lights if light isa Union{PointLight, SpotLight}]
scatter!(scene, ps, color = cs, strokewidth = 1, strokecolor = :white)
# Place camera above
# update_cam!(scene.scene, Vec3f(0, 0, 10), Vec3f(0, 0, 0), Vec3f(0, 1, 0))
# place camera at angle nearby
update_cam!(scene.scene, Vec3f(8, 1, 4), Vec3f(0, 0, 0), Vec3f(0, 0, 1))
# record
idxs = vcat(1:length(sl.range[]), length(sl.range[])-1:-1:2)
record(fig, "point_lights_and_spot_light.webm", idxs, framerate = 30) do idx
sl.selected_index[] = idx
angle[] = angle[] + 4
ps[] = [light.position[] for light in lights if light isa Union{PointLight, SpotLight}]
notify(fig.scene.current_screens[1].render_tick)
end
# interactive
# display(fig)
# @async while isopen(fig.scene)
# angle[] = angle[] + 2
# ps[] = [light.position[] for light in lights if light isa Union{PointLight, SpotLight}]
# sleep(1/60)
# end
endpoint_lights_and_spot_light.webm |
|
Added refimgs: Ambient + SpotLights + PointLights w/o specular reflections: DirectionalLights + specular reflections: |
| # polling may change window size, when its bigger than monitor! | ||
| # we still need to poll though, to get all the newest events! | ||
| # GLFW.PollEvents() | ||
| pollevents(screen) | ||
| # keep current buffer size to allows larger-than-window renders | ||
| render_frame(screen, resize_buffers=false) # let it render |
There was a problem hiding this comment.
I added pollevents() here so lighting updates from render_tick can take effect before drawing the frame. Given the note I would assume this breaks output sizes with save and record, however I have no problems saving/recording at resolution = (4000, 4000).
ffreyer
changed the title
GLMakie/assets/shader/util.vert
Outdated
| // direction to light for :fast shading | ||
| // (normalizing this will result in incorrect lighting) | ||
| vec4 view_light_pos = view*vec4(lightposition, 1.0); | ||
| o_lightdir = view_light_pos.xyz / view_light_pos.w - view_pos.xyz; |
There was a problem hiding this comment.
As noted normalizing here results in incorrect light directions/perceived positions. To illustrate this, consider a light source shining on a rect viewed from the side:
If we don't normalize o_lightdir it will be a vector pointing from the light source to the vertex handled by the vertex shader. Since I placed the light source at (0, 0) here they'll match p1 and p2 in the example. On the way to the fragment shader o_lightdir will then get interpolated between different vertices. For example, halfway between the two vertices the vector will point to the green "Vertex center". Normalizing in the fragment shader will then give correct light-to-vertex-center direction.
If we do normalize o_lightdir in the vertex shader as before, we will be interpolating the normalized o_lightdir vectors. In the sketch above those point to n1 (equal to p1) and n2. Doing the same half-way interpolation for vertices as above, we end up with a o_lightdir halfway between n1 and n2, which is marked as "Normalized center". This is not a vector passing through the vertex center position and thus the wrong light direction.
I'm pretty sure the test failure in Order Independent transparency is a result of this change. There are likely quite a few more failures that didn't get caught.
There was a problem hiding this comment.
I switched to doing lighting in world space so this is not really relevant to lightdir anymore. Though it probably still applies to camdir. (World vs view/camera space shouldn't matter for lighting. It was using view/camera space because SSAO needed it and lighting could reuse it. Merging lighting between volume and the rest of Makie requires world space though.)
|
Volume with multiple lights: volume.webmVolume plots don't differentiate inwards facing vs outwards facing surfaces, so lighting is a bit weird here. I'm not going to try to change this here though. |
|
Fixed this in WGLMakie (even though it should be unrelated to this pr) |
|
Seems like this too 🤷 There also seem to be some issues with line (and maybe scatter) depth order: |
|
As far as I can tell the only issue left here that isn't also on #3113 is with "contour labels 3D". With Axis3 scaling being moved to |
Continues #2831 ! Still needs to check, if I rebased correctly and didn't incorrectly apply some of the changes! ## Merged PRs - #2598 - #2746 - #2346 - #2544 - #3082 - #2868 - #3062 - #3106 - #3281 - #3246 ## TODOS - [x] fix flaky test `@test GLMakie.window_size(screen.glscreen) == scaled(screen, (W, H))` - [x] Merge axis type inferences from #2220 - [x] Test on different resolution screens, IJulia, Pluto, VSCode, Windowed - [x] rebase to only have merge commits from the PRs - [x] investigate unexpected speed ups - [x] reset camera settings from tests - [ ] check doc image generation - [x] rethink default near/far in Camera3D (compatability with OIT) - [x] merge #3246 - [x] fix WGLMakie issues/tests: - [x] fix line depth issues (see tests: ~~hexbin colorrange~~ (not new), LaTeXStrings in Axis3, Axis3 axis reversal) - [x] fix lighting of surface with nan points (fixed in #3246) - ~~volume/3D contour artifacts (see 3D Contour with 2D contour slices)~~ not new - ~~artifacting in "colorscale (lines)"~~ not new - [x] GLMakie: - [x] slight outline in "scatter image markers" test - ~~clipping/z-fighting in "volume translated"~~ not new - [x] CairoMakie: - ~~Artfiacting in `colorscale (lines)"~~ not new - ~~markersize in "scatter rotations" changed?~~ not new - ~~color change in "colorscale (poly)"~~ not new - ~~transparency/render order of "OldAxis + Surface"~~ not new - ~~render order in "Merged color mesh"~~ not new - ~~render order of "Surface + wireframe + contour"~~ not new - [x] Check "SpecApi in convert_arguments" (colors swapped?) ## Fixes the following errors - fixes #2721 via #2746 - fixes #1600 via #2746 - fixes #1236 via #2746 - fixes MakieOrg/GeoMakie.jl#133 via #2598 - closes #2522 - closes #3239 via #3246 - fixes #3238 via #3246 - fixes #2985 via #3246 - fixes #3307 via #3281
Description
Change Summary
This pr adds/changes:
backlightnow affectsnormalinstead oflightdir. This means specular reflections show up with backlightDirectionalLight(closes Directional light as new default #3239)diff_coeffaround 0. This softens the edge where normals are perpendicular to the light direction)shadingattribute is now:none/:fast/:verboseinstead of true or falsePointLightnow has attenuation parameters (distance based intensity falloff, off by default)DirectionalLight(parallel light rays)SpotLight(light source illuminating a cone)PointLight(and all others) is now connected (fixes Point light color is ignored in GLMakie #3238)colorbuffer()now polls events to triggerrender_tickso that it can cause lighting updateslightsvector (fix Issue with typing on LScene with lights #2985)Type of change
TODO
shading