Object data bounds improvements proposal #96968

Open
opened 2022-04-02 00:41:45 +02:00 by Hans Goudey · 1 comment
Member

Problems

There are issues with the handling of bounding boxes currently that
result in performance problems, inconsistencies, and overly complicated code:

  • They are stored in objects, but they really have to do with geometry.
    • This means that geometry instances (without objects) have to recalculate their bounds on every redraw.
    • There are other unnecessary recalculations because code is messy and the situation is hard to reason about.
  • All eight corners of the box are stored, when anything besides the min and max are redundant.
    • This often results in complex code for retrieving the min and max.
  • Behavior between different object types is often inconsistent due to the lack of clear design.
  • There are two many ways to retrieve the same data. The uses of "object" bounds vs "object-data" bounds are often ambiguous.

Design

Bounding Box Definition

  • Bounds describe the local space of the geometry/data
    • Anything else, like bounds with a transformation applied, is usually unnecessary, and is too specific or expensive to cache.
  • Bounds should usually be retrieved from geometry directly
  • For original objects, the bounding box should only include the original data
  • For evaluated objects, the bounding box includes all evaluated geometry besides instances

This implies that the bounding box wouldn't be copied from the evaluated object
back to the original. While this might require some changes, it should make for
clearer code in the end.

Caching

Since the existing bounds cache really refers to the geometry, it should be cached in
the geometry. Object bounds can be built out of the geometry bounds as necessary.
This should significantly speed up geometry instancing, where the bounds currently
need to be recalculated on every redraw.

Return By Value

Returning BoundBox by value when accessing object bounds can make it clear that
it's just built from the geometry bounds. It also makes it possible to remove the allocated
bounding box from Object.runtime.

Implementation

Step 1: Geometry caching, internal changes

  • Add caching of bounds per geometry data-block e8f4010611
    • Share bounds caches between equivalent copied & original data-blocks e8f4010611
  • Apply to all relevant object data types
  • Make accessing bounds on objects use the geometry bounds
    • Return BoundBox by value in BKE_object_boundbox_get
    • Use min and max access (Bounds<float3>) instead the full 8 corners
    • Remove redundant functions like BKE_pointcloud_boundbox_get and BKE_mesh_boundbox_get
  • Change some code to access bounds from geometry instead of objects
  • Make sure uses of object bounding boxes (BKE_object_boundbox_get) use the evaluated object
  • Remove cache of bounds on objects (Object.runtime.bb)

Step 2: Python/RNA API

  • Make changing positions in Python tag bounds dirty
  • Somehow make all python access of Object.bound_box use evaluated objects
    • Add accessors for geometry bounds to Python
    • Make depsgraph necessary for original object evaluated bounds access
    • Handle backwards compatibility?

Step 3: Remove Syncing

  • Remove syncing the evaluated object's bounding box to the original object (original object bounds would only use original geometry).

Step 4: Potential Future Improvements

  • Explore whether bounds calculation can ever be skipped entirely, if render engines don't need it and snapping is turned off, for example

References

# Problems There are issues with the handling of bounding boxes currently that result in performance problems, inconsistencies, and overly complicated code: * They are stored in objects, but they really have to do with geometry. * This means that geometry instances (without objects) have to recalculate their bounds on every redraw. * There are other unnecessary recalculations because code is messy and the situation is hard to reason about. * All eight corners of the box are stored, when anything besides the min and max are redundant. * This often results in complex code for retrieving the min and max. * Behavior between different object types is often inconsistent due to the lack of clear design. * There are two many ways to retrieve the same data. The uses of "object" bounds vs "object-*data*" bounds are often ambiguous. # Design **Bounding Box Definition** * Bounds describe the local space of the geometry/data * Anything else, like bounds with a transformation applied, is usually unnecessary, and is too specific or expensive to cache. * Bounds should usually be retrieved from geometry directly * For original objects, the bounding box should only include the original data * For evaluated objects, the bounding box includes all evaluated geometry besides instances This implies that the bounding box wouldn't be copied from the evaluated object back to the original. While this might require some changes, it should make for clearer code in the end. **Caching** Since the existing bounds cache really refers to the geometry, it should be cached in the geometry. Object bounds can be built out of the geometry bounds as necessary. This should significantly speed up geometry instancing, where the bounds currently need to be recalculated on every redraw. **Return By Value** Returning `BoundBox` by value when accessing object bounds can make it clear that it's just built from the geometry bounds. It also makes it possible to remove the allocated bounding box from `Object.runtime`. # Implementation **Step 1: Geometry caching, internal changes** - [x] Add caching of bounds per geometry data-block e8f4010611e7 - [x] Share bounds caches between equivalent copied & original data-blocks e8f4010611e7 - [ ] Apply to all relevant object data types - [x] Make accessing bounds on objects use the geometry bounds - [x] Return `BoundBox` by value in `BKE_object_boundbox_get` - [x] Use min and max access (`Bounds<float3>`) instead the full 8 corners - [x] Remove redundant functions like `BKE_pointcloud_boundbox_get` and `BKE_mesh_boundbox_get` - [x] Change some code to access bounds from geometry instead of objects - [x] Make sure uses of object bounding boxes (`BKE_object_boundbox_get`) use the evaluated object - [x] Remove cache of bounds on objects (`Object.runtime.bb`) **Step 2: Python/RNA API** - [ ] Make changing positions in Python tag bounds dirty - [ ] Somehow make all python access of Object.bound_box use evaluated objects - [ ] Add accessors for geometry bounds to Python - [ ] Make depsgraph necessary for original object evaluated bounds access - [ ] Handle backwards compatibility? **Step 3: Remove Syncing** - [ ] Remove syncing the evaluated object's bounding box to the original object (original object bounds would only use original geometry). **Step 4: Potential Future Improvements** - [ ] Explore whether bounds calculation can ever be skipped entirely, if render engines don't need it and snapping is turned off, for example # References - #92963 (Instancing Performance Improvements) - #92206 (Bounding Box: compute during depsgraph evaluation) *A related design task for bounding boxes. Proposes some changes but doesn't go as far as this task.* - [D13598: Refactor: Clarify object bounding boxes, cache bounds in data (WIP)](https://archive.blender.org/developer/D13598) *An old WIP patch exploring these ideas. Some of the changes there aren't current with this proposal.*

This issue was referenced by e8f4010611

This issue was referenced by e8f4010611e76e0b88cd8a706b630ad843c09941
Bastien Montagne added this to the Core project 2023-02-09 15:42:54 +01:00
Bastien Montagne removed this from the Core project 2023-02-09 18:20:30 +01:00
Sign in to join this conversation.
No Label
Interest
Alembic
Interest
Animation & Rigging
Interest
Asset Browser
Interest
Asset Browser Project Overview
Interest
Audio
Interest
Automated Testing
Interest
Blender Asset Bundle
Interest
BlendFile
Interest
Collada
Interest
Compatibility
Interest
Compositing
Interest
Core
Interest
Cycles
Interest
Dependency Graph
Interest
Development Management
Interest
EEVEE
Interest
EEVEE & Viewport
Interest
Freestyle
Interest
Geometry Nodes
Interest
Grease Pencil
Interest
ID Management
Interest
Images & Movies
Interest
Import Export
Interest
Line Art
Interest
Masking
Interest
Metal
Interest
Modeling
Interest
Modifiers
Interest
Motion Tracking
Interest
Nodes & Physics
Interest
OpenGL
Interest
Overlay
Interest
Overrides
Interest
Performance
Interest
Physics
Interest
Pipeline, Assets & IO
Interest
Platforms, Builds & Tests
Interest
Python API
Interest
Render & Cycles
Interest
Render Pipeline
Interest
Sculpt, Paint & Texture
Interest
Text Editor
Interest
Translations
Interest
Triaging
Interest
Undo
Interest
USD
Interest
User Interface
Interest
UV Editing
Interest
VFX & Video
Interest
Video Sequencer
Interest
Virtual Reality
Interest
Vulkan
Interest
Wayland
Interest
Workbench
Interest: X11
Legacy
Blender 2.8 Project
Legacy
Milestone 1: Basic, Local Asset Browser
Legacy
OpenGL Error
Meta
Good First Issue
Meta
Papercut
Meta
Retrospective
Meta
Security
Module
Animation & Rigging
Module
Core
Module
Development Management
Module
EEVEE & Viewport
Module
Grease Pencil
Module
Modeling
Module
Nodes & Physics
Module
Pipeline, Assets & IO
Module
Platforms, Builds & Tests
Module
Python API
Module
Render & Cycles
Module
Sculpt, Paint & Texture
Module
Triaging
Module
User Interface
Module
VFX & Video
Platform
FreeBSD
Platform
Linux
Platform
macOS
Platform
Windows
Priority
High
Priority
Low
Priority
Normal
Priority
Unbreak Now!
Status
Archived
Status
Confirmed
Status
Duplicate
Status
Needs Info from Developers
Status
Needs Information from User
Status
Needs Triage
Status
Resolved
Type
Bug
Type
Design
Type
Known Issue
Type
Patch
Type
Report
Type
To Do
No Milestone
No project
No Assignees
2 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: blender/blender#96968
No description provided.