Materials on Ground Primitives #6393

likangning93 · 2018-03-30T01:58:52Z

Opening a PR for some early review, and also to say this is "stable" if anyone wants to play with it.
TODO before full review:

This PR adds support for materials on ground primitives and also adds batching for non-overlapping, "similarly sized" ground entities with the same material. Quick rundown of how things work:

Materials on Ground Primitives

Normals and texture coordinates and the like are computed in the fragment shader, almost like a post-process. Normals are easy - sample the depth texture (which has to be enabled for this to work), do some eye space math, lovely.
Texture coordinates are done in either of two ways:

for entities large enough that the curve of the Earth is noticeable, a world position is computed from the depth texture and is then used to approximate the latitude/longitude at the fragment using spherical coordinates and approximate inverse trigonometry. This spherical coordinate is then compared with a spherical extent per-instance attribute to compute [0-1] range texcoords.
for small entities, we compute the distance from the fragment in eye space to a pair of planes, also in eye space, which are computed in the vertex shader from a double-precision position and a pair of offset vectors that form a bounding box in the local ENU coordinate system centered around the entity.

Batching

We couldn't batch ground entities before because of overdraw problems - the shadow volumes used to color terrain often extend well beyond the area they're supposed to cover in screen space and sometimes overlap in the camera plane, leading to artifacts.

This PR adds the ability to use texture coordinates to clip fragments outside a lat/long volume (rectangle) in space, preventing overdraw as long as the batched primitives' rectangles on the globe don't overlap.

Ground entities with per-instance color are no longer batched by color but based on overlap, so it's now possible to render ~10,000 uniquely colored non-overlapping ground entities in a scene on a mid-range mobile GPU like the GT 750M without crashing. More performance anecdotes to follow.

We may also want to discuss if batching everything that's batchable is a little naive, since in some cases we can actually suffer from loss of culling benefits (I think...).

… of asin

…acy but still jittery

…ne and spherical texcoords mode

cesium-concierge · 2018-03-30T01:58:53Z

@likangning93, thanks for the pull request! Maintainers, we have a signed CLA from @likangning93, so you can review this at any time.

⚠️ I noticed that CHANGES.md has not been updated. If this change updates the public API in any way, fixes a bug, or makes any non-trivial update, please add a bullet point to CHANGES.md and comment on this pull request so we know it was updated. For more info, see the Pull Request Guidelines.

⚠️ I noticed that a file in one of our ThirdParty folders (ThirdParty/, Source/ThirdParty/) has been added or modified. Please verify that it has a section in LICENSE.md and that its license information is up to date with this new version. Once you do, please confirm by commenting on this pull request.

I am a bot who helps you make Cesium awesome! Contributions to my configuration are welcome.

🌍 🌎 🌏

likangning93 · 2018-04-02T18:58:10Z

Aha, texture coordinate rotations don't work yet. Added to TODO:

viewer.entities.add({
    rectangle : {
        coordinates : Cesium.Rectangle.fromDegrees(-110.0, 20.0, -80.0, 25.0),
        material : '../images/Cesium_Logo_Color.jpg',
        stRotation : 10.0
    }
});

bagnell · 2018-04-02T20:34:55Z

At a high level, the code looks good. Let me know when this is ready for a full review.

pjcozzi · 2018-04-03T11:06:10Z

Apps/Sandcastle/gallery/Hello World.html

@@ -28,6 +28,76 @@
    'use strict';
 //Sandcastle_Begin
 var viewer = new Cesium.Viewer('cesiumContainer');
+
+var instance = new Cesium.GeometryInstance({


For clean git history, please don't submit temp code to Sandcastle's Hello World (note that I use to do exactly this to the default Cesium app years ago)

pjcozzi · 2018-04-03T11:07:05Z

Apps/Sandcastle/gallery/batchingGroundPrims.html

+    <meta http-equiv="X-UA-Compatible" content="IE=edge">
+    <meta name="viewport" content="width=device-width, initial-scale=1, maximum-scale=1, minimum-scale=1, user-scalable=no">
+    <meta name="description" content="Use Viewer to start building new applications or easily embed Cesium into existing applications.">
+    <meta name="cesium-sandcastle-labels" content="Beginner, Showcases">


Should this be a "developer" Sandcastle example?

Also do not use abbreviations in file names.

pjcozzi · 2018-04-03T11:08:25Z

Apps/Sandcastle/gallery/testDepth.html

@@ -0,0 +1,136 @@
+<!DOCTYPE html>


Overall looks like these Sandcastle examples are all temp so please bump when you have the right set that you want to show off this new features.

pjcozzi · 2018-04-03T11:08:51Z

LICENSE.md

@@ -670,6 +702,20 @@ https://github.com/KhronosGroup/glTF-WebGL-PBR
 >CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE
 >OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

+### ShaderFastLibs (adapted code)


What does "adapted code" mean? Can you point me to it?

So it looks like you just ported to JavaScript and GLSL. All good. Just remove "(adapted code)" here, I think it will cause confusion.

pjcozzi · 2018-04-03T11:10:02Z

Source/Core/Math.js

+     * @returns {Number} An approximation of atan(x)
+     * @private
+     */
+    CesiumMath.fastApproximateAtan = function(x) {


Here and below could be public IMO like much of the rest of this class already is.

pjcozzi · 2018-04-03T11:33:51Z

Source/Shaders/Builtin/Functions/approximateSphericalCoordinates.glsl

+float fastApproximateAtan2(float x, float y) {
+    // atan approximations are usually only reliable over [-1, 1], or, in our case, [0, 1] due to modifications.
+    // So range-reduce using abs and by flipping whether x or y is on top.
+    float opposite, adjacent, t; // t used as swap and atan result.


Remove this, and put the float declaration with the assignment statements below.

pjcozzi · 2018-04-03T11:34:51Z

Source/Shaders/Builtin/Functions/approximateSphericalCoordinates.glsl

+ *
+ * @returns {vec2} Approximate latitude and longitude spherical coordinates.
+ */
+vec2 czm_approximateSphericalCoordinates(vec3 normal) {


Please make sure that somewhere in the reference documentation very clearly states that textured polygons on terrain are for notional patterns not for precisely mapping textures to the ground - for that case, use the single imagery provider.

pjcozzi · 2018-04-03T11:38:34Z

LICENSE.md

+
+https://github.com/mourner/quickselect
+
+No license given, used by rbush


I think it would fall under the rbrush license, but submitted mourner/quickselect#5

Here is the license: https://github.com/mourner/quickselect/blob/master/LICENSE

pjcozzi · 2018-04-03T11:41:55Z

Source/Shaders/Builtin/Functions/approximateSphericalCoordinates.glsl

+}
+
+// Used in place of ternaries to trade some math for branching
+float branchFree(bool comparison, float a, float b) {


If this is generally useful to other shaders, perhaps put in its own file and rename to czm_branchFreeTernary.

pjcozzi · 2018-04-03T11:45:15Z

Source/Scene/ShadowVolumeAppearanceShader.js

+        }
+    });
+
+    function createShadowVolumeAppearanceShader(appearance, extentsCull, planarExtents) {


The shader looks pretty fast all things consider, nice. 🏃

It is, however, spread across a few different files with all sorts of string concatenation and preprocessor defines. Maybe that is the nature of the beast...or perhaps there is a way to make this more cohesive?

pjcozzi · 2018-04-03T11:52:44Z

Is this my video card? Intel/Mac/Chrome. When I zoom in close (perhaps the polygon enters the first frustum), I get this with the initial polygon in http://localhost:8080/Apps/Sandcastle/index.html

pjcozzi · 2018-04-03T11:54:25Z

Not sure if this example is supposed to work, but "switch to color appearance" yields:

pjcozzi · 2018-04-03T11:57:34Z

Good start, @likangning93. The precision issue is important and should be addressed before anything else - maybe it is just my video card which I think has other ground primitive artifacts - we can test.

Otherwise, this just needs a lot of clean up, Sandcastle examples, tests, etc. that you already have in the task list.

Please iterate with @bagnell and I will try to take one last look before you merge if I have time.

Also looking forward to the blog post(s).

likangning93 · 2018-04-03T20:09:51Z

I did something very bad and tried to use rebase to remove unintentional changes to Hello World... I'll fix this soon. Sorry!

likangning93 · 2018-04-11T21:15:43Z

On the subject of picking:

I think ClassificationPrimitive used to handle picking for the limited batchable cases by rendering each geometry instance separately using offsets into the vertex buffers, which seems like something we don't want to do if we can avoid it.
Since we're enforcing non-overlap, I think we could also use the texcoord culling thing to get around this. @bagnell any thoughts?

bagnell · 2018-04-11T21:40:25Z

Yes, that's how it works. When we pick, we create an off-center frustum and scissor test only covering a 3x3 pixel area around the pick position so performance isn't terrible.

Since we're enforcing non-overlap, I think we could also use the texcoord culling thing to get around this.

Sounds good .I'd be interested in the performance comparison to the current approach when there are a lot of volumes.

likangning93 · 2018-05-10T13:55:59Z

This trick might be practical if we wanted to implement KML Ground Overlays though, since those let you do weird things like stretch the image based on the positions of a quad using the gx:latlonquad extension

mramato · 2018-05-10T14:02:11Z

This trick might be practical if we wanted to implement KML Ground Overlays though, since those let you do weird things like stretch the image based on the positions of a quad using the gx:latlonquad extension

We definitely want projected imagery like latlonquad long term, but it's probably not worth worrying about until we actually go to do it. If you have notes that would be relevant to implementation; maybe write up a new issue so we don't forget about them.

hpinkos · 2018-05-10T14:29:26Z

FYI, trapezoidal texturing will be based on this when we get a chance to do it: http://help.agi.com/AGIComponentsJava/html/BlogTrapezoidalTextureProjectionWithOpenGL.htm

likangning93 · 2018-05-10T20:34:21Z

funny/stupid idea

@bagnell: "don't do that"

…ses and Polygons

likangning93 · 2018-05-11T12:43:27Z

@bagnell I think this is finally ready for another look! Sorry for the churn.

likangning93 · 2018-05-11T12:48:03Z

Source/Core/Geometry.js

-        './PrimitiveType'
+        './Matrix2',
+        './Matrix3',
+        './Matrix4',


the gang's all here...

bagnell · 2018-05-11T19:38:10Z

Source/Scene/ShadowVolumeAppearance.js

+     * @private
+     *
+     * @param {Rectangle} boundingRectangle Rectangle object that the points will approximately bound
+     * @param {Rectangle} textureCoordinateRotationPoints TODO: doc


There are a few TODOs left in the doc.

bagnell · 2018-05-11T19:39:16Z

Specs/DataSources/StaticGroundGeometryPerMaterialBatchSpec.js

+    });
+
+    it('handles shared material being invalidated with geometry', function() {
+        if (!GroundPrimitive.isSupported(scene) | !GroundPrimitive.supportsMaterials(scene)) {


Here and throughout, change this to ||

bagnell · 2018-05-11T19:40:11Z

@likangning93 Just those two comments and this is good to go.

likangning93 · 2018-05-11T20:13:40Z

@bagnell ready!

lilleyse · 2018-05-15T02:54:26Z

I'm seeing some problems with ground primitives, textured or not, when overlaying over 3D Tiles. Classification Types demo:

likangning93 · 2018-05-15T04:04:46Z

I'm seeing some problems with ground primitives, textured or not, when overlaying over 3D Tiles. Classification Types demo:

I think what might be going on is that the stencil state takes the tileset into account but the tileset's depth isn't getting into globeDepth aka czm_globeDepthTexture. I'll take a deeper look in the morning.

likangning93 · 2018-05-15T13:55:48Z

Getting this fully working with 3D Tilesets is going to involve messing with passes that I'd like to discuss with @lilleyse and @bagnell, let's put that off for a future PR.

For this PR I'm going to restrict materials on GroundPrimitives to terrain classification and have the Entity layer detect if a material is being used and automatically render assuming TERRAIN when encountering BOTH/CESIUM_3D_TILE.

Color classification BOTH/CESIUM_3D_TILE will fall back to original behavior.

…roundPrimitives by falling back to TERRAIN

likangning93 · 2018-05-15T18:33:55Z

@bagnell ready again

lilleyse · 2018-05-27T17:59:36Z

Getting this fully working with 3D Tilesets is going to involve messing with passes that I'd like to discuss with @lilleyse and @bagnell, let's put that off for a future PR.

@likangning93 can you open an issue for this so we don't forget? I think once people see this working with the globe they're going to want it for 3D Tiles.

likangning93 · 2018-05-29T13:53:55Z

#6636

likangning93 added 16 commits February 13, 2018 17:38

able to get world coordinates from depth tex

bdf302e

texcoords from naive inverse trig

4fe0397

using cg ref approx to compute spherical coordinates on CPU and GPU

956df6c

add separate shader for shadow volume final pass, use a cubic instead…

529b5ab

… of asin

merged in master

7af7d47

WIP material support for ground primitives w/normals from depth tex

3e5c724

change ground primitive normal computation to EC for less noise

0433738

reduce multifrustum artifacts for EC normals from depth

2b47810

bespoke shadow volume appearance shaders

693f7e2

switch spherical coordinates computation to atan

f9ca260

simplified spherical coordinates via atan

d02346b

merged in master

71b6095

added plane-based texcoords, still has precision artifacts

e7c4679

changed texcoord planes to use a model matrix attribute, less inaccur…

23d6dfb

…acy but still jittery

using double-precision points for planes, added switching between pla…

62ab0ba

…ne and spherical texcoords mode

add batching for ground primitives

2cb5132

pjcozzi requested changes Apr 3, 2018

View reviewed changes

likangning93 force-pushed the texcoordsOnTerrain branch from 2cb5132 to a76cfdd Compare April 3, 2018 20:06

likangning93 force-pushed the texcoordsOnTerrain branch from a76cfdd to 2cb5132 Compare April 4, 2018 01:03

likangning93 added 2 commits April 5, 2018 15:59

PR feedback

9212796

add texture coordinate rotation for materials on GroundPrimitives

8e59e2d

bagnell mentioned this pull request May 10, 2018

Per-feature post-processing #6476

Merged

4 tasks

likangning93 added 2 commits May 10, 2018 21:26

compute texture coordinate rotation parameters in local ENU for Ellip…

fe8d1f7

…ses and Polygons

doc tweaks

9643b2c

likangning93 commented May 11, 2018

View reviewed changes

Source/Core/Geometry.js

'./PrimitiveType'

'./Matrix2',

'./Matrix3',

'./Matrix4',

Copy link

Contributor Author

likangning93 May 11, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

the gang's all here...

bagnell reviewed May 11, 2018

View reviewed changes

PR feedback and cleanup

2aaf979

handle ClassificationType BOTH and CESIUM_3D_TILE with materials on G…

16c7305

…roundPrimitives by falling back to TERRAIN

likangning93 added 2 commits May 15, 2018 14:41

format CHANGES.md entry with bullets

fbf05c8

add textured terrain entity example to Ground Clamping Sandcastle

4f580be

bagnell merged commit 01c49d6 into CesiumGS:master May 15, 2018

likangning93 deleted the texcoordsOnTerrain branch May 15, 2018 19:07

bagnell mentioned this pull request May 15, 2018

Fix uniform map name for color copy. #6590

Merged

likangning93 mentioned this pull request Jun 4, 2018

Polylines on terrain via Primitive API #6615

Merged

5 tasks

likangning93 mentioned this pull request Oct 12, 2018

Proj4JS and Custom Projections in Columbus View #6986

Merged

3 tasks

OmarShehata mentioned this pull request Aug 9, 2019

DrillPick on 2D map broken based on zoom level #8052

Open

OmarShehata mentioned this pull request Nov 3, 2019

Make polygons with an image material render as translucent #8349

Closed


		https://github.com/mourner/quickselect

		No license given, used by rbush

Materials on Ground Primitives #6393

Materials on Ground Primitives #6393

Conversation

likangning93 commented Mar 30, 2018 • edited Loading

Materials on Ground Primitives

Batching

cesium-concierge commented Mar 30, 2018

likangning93 commented Apr 2, 2018

bagnell commented Apr 2, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pjcozzi commented Apr 3, 2018

pjcozzi commented Apr 3, 2018

pjcozzi commented Apr 3, 2018

likangning93 commented Apr 3, 2018

likangning93 commented Apr 11, 2018 • edited Loading

bagnell commented Apr 11, 2018

likangning93 commented May 10, 2018 • edited Loading

mramato commented May 10, 2018

hpinkos commented May 10, 2018

likangning93 commented May 10, 2018

likangning93 commented May 11, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bagnell commented May 11, 2018

likangning93 commented May 11, 2018

lilleyse commented May 15, 2018

likangning93 commented May 15, 2018

likangning93 commented May 15, 2018 • edited Loading

likangning93 commented May 15, 2018

lilleyse commented May 27, 2018

likangning93 commented May 29, 2018

likangning93 commented Mar 30, 2018 •

edited

Loading

likangning93 commented Apr 11, 2018 •

edited

Loading

likangning93 commented May 10, 2018 •

edited

Loading

likangning93 commented May 15, 2018 •

edited

Loading