Separate Appearance on z-fail #5160

bagnell · 2017-03-28T00:09:03Z

Adds support for a separate appearance when the primitive fails the depth test. Only polylines are supported at the moment. Example:

viewer.entities.add({
    polyline : {
        positions : cartesianSamples,
        followSurface : false,
        width : 5,
        material : new Cesium.PolylineOutlineMaterialProperty({
            color : Cesium.Color.ORANGE,
            outlineWidth : 2,
            outlineColor : Cesium.Color.BLACK
        }),
        depthFailMaterial : new Cesium.PolylineOutlineMaterialProperty({
            color : Cesium.Color.RED,
            outlineWidth : 2,
            outlineColor : Cesium.Color.BLACK
        })
    }
});

pjcozzi · 2017-03-28T00:24:54Z

Add support to other primitives where it makes sense.

I could see this being useful for points/labels/billboards (think waypoints), but can you submit a new issue to consider these later?

pjcozzi · 2017-03-28T00:32:44Z

At quick glance, the code looks good to me.

@lilleyse can you please test and review this? Once ready, we'll also merge this into the 3d-tiles branch.

lilleyse · 2017-03-28T13:23:39Z

I'm noticing overlap of the two materials for z pass:

Second screenshot is with this code:

var viewer = new Cesium.Viewer('cesiumContainer');
viewer.scene.globe.depthTestAgainstTerrain = true;
var position1 = Cesium.Cartesian3.fromRadians(0.0, 0.0, 400000.0);
var position2 = new Cesium.Cartesian3(0.0, 0.0, 0.0);
var cartesianSamples = [position1, position2];
viewer.entities.add({
    polyline : {
        positions : cartesianSamples,
        followSurface : false,
        width : 5,
        material : new Cesium.PolylineOutlineMaterialProperty({
            color : Cesium.Color.ORANGE,
            outlineWidth : 2,
            outlineColor : Cesium.Color.BLACK
        }),
        depthFailMaterial : new Cesium.PolylineOutlineMaterialProperty({
            color : Cesium.Color.RED,
            outlineWidth : 2,
            outlineColor : Cesium.Color.BLACK
        })
    }
});

lilleyse · 2017-03-28T14:52:55Z

Source/Scene/Primitive.js

+        modifiedFS += 'varying float v_WindowZ;\n' +
+              'void main() {\n' +
+              '    czm_non_depth_clamp_main();\n' +
+              '    gl_FragDepthEXT = min(v_WindowZ * gl_FragCoord.w, 1.0);\n' +


Why is depth clamping used?

Still some overlap, but it seems to work ok without clamping. Maybe I'm missing some edge case.

Why is depth clamping used?

To prevent multi-frustum artifacts. If you turn it off and zoom in close, the line will be clipped.

Still some overlap, but it seems to work ok without clamping.

Maybe I didn't subsample the line enough or the lower res terrain LOD is rendered when zooming out. This isn't the best example.

Could you get a screenshot of that happening?

It may be worth supporting a fallback shader when the extension isn't supported, since it still mostly works without.

The shader now doesn't write fragment depth without the extension. Here is a gif showing the artifact for the sample you provided:

lilleyse · 2017-03-28T15:05:19Z

Source/DataSources/GeometryVisualizer.js

+        this._closedColorBatches = new Array(numberOfShadowModes * 3);
+        this._closedMaterialBatches = new Array(numberOfShadowModes * 3);
+        this._openColorBatches = new Array(numberOfShadowModes * 3);
+        this._openMaterialBatches = new Array(numberOfShadowModes * 3);


I feel like this batching is starting to get out of hand (even when it was just shadows), but probably worth addressing later.

lilleyse · 2017-03-28T15:16:15Z

Are there situations where the depth-fail won't work due the ordering of entities? Like I could imagine a large sphere occluding a polyline but the depth material not showing because of the order in which things are rendered. Is this generally ok because the depth-fail batch is last?

pjcozzi · 2017-03-28T15:41:45Z

I feel like this batching is starting to get out of hand (even when it was just shadows), but probably worth addressing later.

Agreed, later.

Are there situations where the depth-fail won't work due the ordering of entities?

Yes. The main use case is depth fail vs. terrain or 3D Tiles. We are also adding an environment pass so that 3D Tiles can be rendered before these.

lilleyse · 2017-03-28T16:28:08Z

In overlapping situations, could something like polygon offset help? Or stenciling?

pjcozzi · 2017-03-29T22:46:57Z

On my integrated Intel MacBook Pro, the z-fail looks right, but z-pass does not:

Note the frustum boundary between the cyan and blue on the left.

EXT_frag_depth is supported.

http://cesium-dev.s3-website-us-east-1.amazonaws.com/cesium/3d-tiles-depth-fail/Apps/Sandcastle/index.html?src=Hello%20World.html&label=Showcases

denverpierce · 2017-03-30T16:29:38Z

+1 for billboards/points. Sometimes I've got cases where I'll need to render a line's vertices along with the line itself.

bagnell · 2017-04-06T18:59:31Z

Second screenshot is with this code:

Interpolating depth across a triangle that long was causing the issue. If you add more samples to the line, it's fine. Here is the updated code:

var viewer = new Cesium.Viewer('cesiumContainer');
viewer.scene.globe.depthTestAgainstTerrain = true;
var cartesianSamples = [];
var step = 1000.0;
var end = (Cesium.Ellipsoid.WGS84.radii.x + 400000.0) / step;
for (var i = 0; i < end; ++i) {
    cartesianSamples.push(new Cesium.Cartesian3(i * step, 0.0, 0.0));
}
viewer.entities.add({
    polyline : {
        positions : cartesianSamples,
        followSurface : false,
        width : 5,
        material : new Cesium.PolylineOutlineMaterialProperty({
            color : Cesium.Color.ORANGE,
            outlineWidth : 2,
            outlineColor : Cesium.Color.BLACK
        }),
        depthFailMaterial : new Cesium.PolylineOutlineMaterialProperty({
            color : Cesium.Color.RED,
            outlineWidth : 2,
            outlineColor : Cesium.Color.BLACK
        })
    }
});

denverpierce · 2017-04-06T21:26:21Z

I'm getting inconsistent results when toggling terrain on then off again in the same session:

bagnell · 2017-04-06T21:53:32Z

@denverpierce Is that first screen shot when the example first loads? In the code example we set, depthTestAgainstTerrain to true. When you select terrain in the imagery/terrian picker, it will also be set to true. Then, when selecting the ellipsoid terrain, it will be set to false. From there it should alternate from true to false.

You can ensure it is always true with:

viewer.scene.terrainProviderChanged.addEventListener(function() {
    viewer.scene.globe.depthTestAgainstTerrain = true;
});

bagnell · 2017-04-06T23:04:49Z

@pjcozzi @lilleyse This is ready for another look.

lilleyse · 2017-04-07T18:03:42Z

I'm still seeing z-fighting with the higher-samples and in the sandcastle demo:

bagnell · 2017-04-07T18:40:13Z

@lilleyse I can't reproduce that. Maybe it's the same issue here #5160 (comment)

Can you try it with EXT_frag_depth disabled? Maybe that's the issue.

lilleyse · 2017-04-07T19:06:56Z

Yeah it works without - this screenshot from before disabled the extension: #5160 (comment)

pjcozzi · 2017-04-09T14:24:43Z

The stock Sandcastle example on my MacBook Pro is the same as #5160 (comment). It's not clear to me what the solution is based on the discussion above: does the example need to be updated or there is a bug when EXT_frag_depth is enabled?

pjcozzi · 2017-04-09T14:25:38Z

@bagnell can you also update the 3d-tiles-depth-fail branch when ready for final testing?

bagnell · 2017-04-11T19:28:44Z

Yeah it works without - this screenshot from before disabled the extension: #5160 (comment)

That isn't the best example. I subsampled the line at the highest terrain resolution. It'll fail depth for lower resolution LODs. Also, I didn't subsample at too fine a granularity.

@pjcozzi I cant reproduce #5160 (comment) or #5160 (comment) . Perhaps its a problem with EXT_frag_depth on Macs? Do you also have issues when the camera enters the shadow volume of GroundPrimitives and the volume intersects multiple frustums?

bagnell · 2017-04-11T20:24:16Z

@lilleyse I swapped the draw order like we discussed offline. Can you see if that fixed the issue?

lilleyse · 2017-04-11T20:30:36Z

Awesome that fixed it for me. @pjcozzi can you also confirm?

pjcozzi · 2017-04-11T23:50:13Z

Works for me now!

Please merge to master and 3d-tiles if this is ready.

lilleyse · 2017-04-12T14:40:23Z

Unfortunately I noticed this right after merging after testing with 3d-tiles.

If either material or depthFailMaterial is a plain color I notice z-fighting like before but for the depth fail. All other polyline material combinations work fine. Also the z-fighting is gone when disabling EXT_frag_depth like before.

material : Cesium.Color.ORANGE,
depthFailMaterial : Cesium.Color.RED

pjcozzi · 2017-04-13T12:24:40Z

@bagnell can you please look at this one soon?

bagnell · 2017-04-13T18:41:52Z

@lilleyse Can you try the depth-fail-stencil branch to see if it fixed the issue?

lilleyse · 2017-04-13T18:54:45Z

I see z-fighting for the pass portion, regardless of material.

pjcozzi · 2017-04-18T17:19:33Z

What is the plan here? To roll back to a previous version that @bagnell mentioned offline?

pjcozzi · 2017-04-20T22:26:46Z

Bump @bagnell - #5160 (comment)

bagnell added 3 commits March 24, 2017 19:12

Add depth fail material to primitives.

26ba43a

Clean up depth fail appearance.

28f6a94

Add depth fail appearance entity support to polylines only.

88da9ca

pjcozzi mentioned this pull request Mar 28, 2017

BIM roadmap #4678

Open

53 tasks

lilleyse reviewed Mar 28, 2017

View reviewed changes

emackey mentioned this pull request Mar 28, 2017

Support for "dashed" polylines #5159

Merged

bagnell added 2 commits April 6, 2017 14:04

Remove writing framgment depth.

69e45d2

Re-add writing fragment depth but disabled if unsupported.

e3a885c

bagnell added 5 commits April 6, 2017 15:37

Fix existing tests.

82f54ca

Add depth fail primitive appearance test.

d7869d4

Add data source tests and doc.

0169370

Revert Hello World and add Sandcastle example.

568d696

Fix jsHint error.

2774f0f

bagnell added 3 commits April 6, 2017 18:51

Fix WebGL stub tests.

880c03f

Merge branch 'master' into depth-fail

7ff72ea

Update CHANGES.md.

7965baa

pjcozzi mentioned this pull request Apr 9, 2017

Billboard/label/point depthFail appearance #5190

Open

Merge branch 'master' into depth-fail

49edb91

Swap depth fail draw order.

8a29014

lilleyse merged commit 9309df8 into master Apr 12, 2017

lilleyse deleted the depth-fail branch April 12, 2017 14:11

bagnell mentioned this pull request Apr 26, 2017

Update depth fail materials #5243

Merged

shunter mentioned this pull request Nov 17, 2017

Add missing entity properties to CZML. #5986

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Separate Appearance on z-fail #5160

Separate Appearance on z-fail #5160

bagnell commented Mar 28, 2017 •

edited

Loading

pjcozzi commented Mar 28, 2017

pjcozzi commented Mar 28, 2017

lilleyse commented Mar 28, 2017

lilleyse Mar 28, 2017

lilleyse Mar 28, 2017

bagnell Mar 28, 2017

lilleyse Mar 28, 2017

bagnell Apr 6, 2017

lilleyse Mar 28, 2017

lilleyse commented Mar 28, 2017

pjcozzi commented Mar 28, 2017

lilleyse commented Mar 28, 2017

pjcozzi commented Mar 29, 2017

denverpierce commented Mar 30, 2017

bagnell commented Apr 6, 2017

denverpierce commented Apr 6, 2017

bagnell commented Apr 6, 2017

bagnell commented Apr 6, 2017

lilleyse commented Apr 7, 2017

bagnell commented Apr 7, 2017

lilleyse commented Apr 7, 2017

pjcozzi commented Apr 9, 2017

pjcozzi commented Apr 9, 2017

bagnell commented Apr 11, 2017 •

edited

Loading

bagnell commented Apr 11, 2017

lilleyse commented Apr 11, 2017

pjcozzi commented Apr 11, 2017

lilleyse commented Apr 12, 2017

pjcozzi commented Apr 13, 2017

bagnell commented Apr 13, 2017

lilleyse commented Apr 13, 2017

pjcozzi commented Apr 18, 2017

pjcozzi commented Apr 20, 2017

Separate Appearance on z-fail #5160

Separate Appearance on z-fail #5160

Conversation

bagnell commented Mar 28, 2017 • edited Loading

pjcozzi commented Mar 28, 2017

pjcozzi commented Mar 28, 2017

lilleyse commented Mar 28, 2017

lilleyse Mar 28, 2017

Choose a reason for hiding this comment

lilleyse Mar 28, 2017

Choose a reason for hiding this comment

bagnell Mar 28, 2017

Choose a reason for hiding this comment

lilleyse Mar 28, 2017

Choose a reason for hiding this comment

bagnell Apr 6, 2017

Choose a reason for hiding this comment

lilleyse Mar 28, 2017

Choose a reason for hiding this comment

lilleyse commented Mar 28, 2017

pjcozzi commented Mar 28, 2017

lilleyse commented Mar 28, 2017

pjcozzi commented Mar 29, 2017

denverpierce commented Mar 30, 2017

bagnell commented Apr 6, 2017

denverpierce commented Apr 6, 2017

bagnell commented Apr 6, 2017

bagnell commented Apr 6, 2017

lilleyse commented Apr 7, 2017

bagnell commented Apr 7, 2017

lilleyse commented Apr 7, 2017

pjcozzi commented Apr 9, 2017

pjcozzi commented Apr 9, 2017

bagnell commented Apr 11, 2017 • edited Loading

bagnell commented Apr 11, 2017

lilleyse commented Apr 11, 2017

pjcozzi commented Apr 11, 2017

lilleyse commented Apr 12, 2017

pjcozzi commented Apr 13, 2017

bagnell commented Apr 13, 2017

lilleyse commented Apr 13, 2017

pjcozzi commented Apr 18, 2017

pjcozzi commented Apr 20, 2017

bagnell commented Mar 28, 2017 •

edited

Loading

bagnell commented Apr 11, 2017 •

edited

Loading