Globe atmosphere + Style specification

[karimnaaji]

Atmosphere style specification

This PR extends the fog system to allow design of an atmosphere effect supported on both globe and mercator projections. It extends the fog style specification to account for three new values: high-color, space-color and star-intensity. The star-intensity allows for a parallax effect while rotating and panning the globe to increase the effect of 3d.

Changelist

• Add an atmosphere debug page with different preset to test the feature
• Add support for fog on g lobe with a fixed range, the fixed range value is chosen arbitrarily to give plausible rendering of a thin atmospheric layer over the globe.
• Consolidate frustum corner calculation and share it between atmosphere rendering and fog uniform upload
• Split atmosphere buffer from GlobeSharedBuffer so that it's not tied to globe rendering
• Remove clear color potentially using fog color, as this shader now covers the entire screen, it is unnecessary to dispatch an expensive clear
• Make sure that globe fixed fog range does not affect rendering of symbols and fog tile culling
• Make sure that heatmaps are not affected by globe fog fixed range, so that this layer can be rendered without being affected by fog opacity, as we get flickering artifacts otherwise
• Update render test baselines to account for the new changes
• Add style specification defaults using complex expressions so that map.setFog({}) results in a nice default for the user
• Add a fake cubemap in the shader in order to create the parallax effect for stars

fake-cubemap.mov

Style specification entries (including old ones):

• color alpha channel

• color rgb channels

• horizon-blend

• high-color alpha channel

• high-color rgb channels

• space-color alpha channel

• space-color rgb channels

• star-intensity

Follow up to address after this PR

• Reduce frequency of globe specific uniforms when updating fog uniform, these values can be 1) calculated once per frame and 2) only be done when used along with globe
• Fix horizon angle line visual artifact, refer https://github.com/mapbox/gl-js-team/issues/410
• Add render tests specific to globe atmosphere
• Fix fill-extrusion fog contribution during transition

Launch Checklist

• briefly describe the changes in this PR
• include before/after visuals or gifs if this PR includes visual changes
• write tests for all new functionality
• manually test the debug page
• apply changelog label ('bug', 'feature', 'docs', etc) or use the label 'skip changelog'

[mpulkki-mapbox]

Looks really good! Are we supposed to have a default preset available when switching to globe view? Rest of the debug pages (e.g. terrain-debug.html) have no atmosphere set.

[karimnaaji]

Are we supposed to have a default preset available when switching to globe view?

No, there's no dependency between globe-view and atmosphere at the spec level. e.g. when setting globe-view we don't automatically enable the atmosphere. To improve UX, this is something to consider fixing at a higher level in Studio/when setting default styling, so that we're not adding dependencies between different style specification entries.

Rest of the debug pages (e.g. terrain-debug.html) have no atmosphere set.

I've updated most of the pages in latest to use the default atmosphere.

There are two follow-ups to this PR that I've captured in other tickets:

• Reduce frequency of operations for Globe atmosphere + Style specification #11590 (comment) (https://github.com/mapbox/gl-js-team/issues/425)
• There's an issue that I've noticed on fill-extrusion on globe during the transition. I'll address that separately (https://github.com/mapbox/gl-js-team/issues/426):

globe-fill-extrusion.mov

Otherwise please take another look as this is ready for another review, I think I've addressed all of the other comments.

[mpulkki-mapbox]

is this an optimization to write pow(x, -1) instead of 1/x?

[karimnaaji]

Not an optimization, that's an overlook from converting the function from desmos, thanks for pointing that out I'll update it!

[karimnaaji]

Also, I switched that function to prefer exp over an sqrt based function, so that the effect is smoother, you can notice that the sqrt based function that we used in main (in red) results in an abrupt change when crossing y=0 due to clamping:

[mpulkki-mapbox]

yes I noticed! That is a better option 👍

[mpulkki-mapbox]

LGTM!