You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Only update the joint matrices if the nodes were targeted by an animation that frame, instead of updating the joint matrices for all skinned nodes.
Don't animation when culled. Will still need the right bounding volume: perhaps we can compute it or the user can provide a maximum radius percentage increase.
updateNodeHierarchyModelMatrix performance ideas with parent pointers. See @pjcozzi's notes.
Loading Performance
Fine-grained dependency requests, e.g., compile shader program once required shaders are fetched, not once all shaders are fetched. See functions called from createResources.
In general, what can be done on a web worker?
Rendering Performance
Auto instance when multiple nodes have the same mesh. This will require copy-on-write if the material for one instance changes, e.g., because it is no longer an instance.
Compute tighter bounding sphere than using the bounding box's corner points. Do this as a preprocess in an extra property and fallback to client-side.
Schedule animations via CZML, i.e., CZML for calling model.activeAnimations.add.
Drive model pointing.
Embed glTF as a CZML extension? Package together as zip?
Model caching
Reference counting the animation cache so it is also automatically deleted.
Ability to invalidate the whole cache or the cache for one url/cache-key
Avoid potential ping-ponging by delaying actually removing from the cache when the count reaches zero for several frames/seconds. This may also move all the GL delete functions into the render loop, which would be better for requestAnimationFrame timing (however, perhaps not noticeable).
Zero out the embedded resources in the glTF to save even more memory?
Viewer
Drag and drop convert COLLADA to glTF via Rest3D web services?
Ecosystem
glTF tutorial series
Make gltfStatistics a separate library
Potential Ideas
maximumSizeInPixels so the min/max could be the same and used as a billboard
For performance, render to texture with depth map when pixel size is constant and there are no animation/light changes (billboarding)
Older
b26
General
Unit tests
Sandcastle example with sample models
Tutorial
Animation
Add offset (in seconds) to ModelAnimation to allow starting an animation "in progress?"
Soonish
parameter[count -1] === parameter[0]
.computeWorldBoundingSphere
?id
andname
in Cesium API to match glTF spec.ById
versions of functions likegetNode
?Later
glTF Features
glTF Extensions
Animation
Animation Performance
updateNodeHierarchyModelMatrix
performance ideas with parent pointers. See @pjcozzi's notes.Loading Performance
createResources
.Rendering Performance
extra
property and fallback to client-side.allowPicking
isfalse
, we should still render depth.Cesium-Specific Features
ModelNode
to, for example, easily change the scale transform and interop with TRS animations.time[0]
and the duration.ModelMesh
.CZML
model.activeAnimations.add
.Model caching
delete
functions into the render loop, which would be better forrequestAnimationFrame
timing (however, perhaps not noticeable).Viewer
Ecosystem
gltfStatistics
a separate libraryPotential Ideas
maximumSizeInPixels
so the min/max could be the same and used as a billboardOlder
b26
General
Animation
offset
(in seconds) toModelAnimation
to allow starting an animation "in progress?"Animation - Cesium Splines - @bagnell
result
parametersAOS
toSOA
to share time array and for less cache pollutionAnimation Performance
Matrix4.fromTranslationQuaternionRotationScale
- @bagnellRendering Performance
DeveloperError
, Remove DeveloperError in release build #1240 - @bagnellDrawCommand
- @bagnellCesium Cleanup
attributeIndices
toattributeLocations
throughout.UniformDatatype
to WebGL.IndexDatatype
,TextureWrap
,TextureMinificationFilter
,TextureMagnificationFilter
, andPrimitiveType
.The text was updated successfully, but these errors were encountered: