WIP: IK: distinguish 'off' from '0% influence' in IK constraints #117178

Sybren A. Stüvel · 2024-01-16T17:42:14+01:00

Sybren A. Stüvel commented

2024-01-16 17:42:14 +01:00

Context (as in, main Blender behaviour): depending on the situation, the IK solver will use the influence slider in different ways. If there is no pole vector, it will interpolate between the target position and the bone's current position. This means that an influence of 0% will still fully run the IK solver, which overwrites the bone's transform matrix.

This PR introduces a distinction between '0% influence' and 'muted'. The former still behaves as above, and the latter now actually skips the IK solver for that bone chain.

This PR doesn't quite work, though. A simple test case seems to work fine, but when I test with an actual production rig it causes glitching. The glitching disappears when running Blender with -t 1, suggesting that it might be due to threading and missing relations in the dependency graph. From what I can tell, the semantic change of having "OFF" as "completely ignored by Blender" requires deeper code changes so that it's really properly ignored by the depsgraph. Although that might be the correct approach for an always-OFF constraint, things get more complex when actually animating the 'enabled' property.

This is aiming at fixing #115915.

Context (as in, `main` Blender behaviour): depending on the situation, the IK solver will use the influence slider in different ways. If there is no pole vector, it will interpolate between the target position and the bone's current position. This means that an influence of 0% will still fully run the IK solver, which overwrites the bone's transform matrix. This PR introduces a distinction between '0% influence' and 'muted'. The former still behaves as above, and the latter now actually skips the IK solver for that bone chain. This PR doesn't quite work, though. A simple test case seems to work fine, but when I test with [an actual production rig](https://studio.blender.org/characters/5f1ed640e9115ed35ea4b3fb/showcase/1/) it causes glitching. The glitching disappears when running Blender with `-t 1`, suggesting that it might be due to threading and missing relations in the dependency graph. From what I can tell, the semantic change of having "OFF" as "completely ignored by Blender" requires deeper code changes so that it's really properly ignored by the depsgraph. Although that might be the correct approach for an always-OFF constraint, things get more complex when actually animating the 'enabled' property. This is aiming at fixing #115915.

Sybren A. Stüvel added the

Module

Animation & Rigging

label 2024-01-16 17:42:14 +01:00

Sybren A. Stüvel added 5 commits 2024-01-16 17:42:24 +01:00

c0115450ad IK: distinguish 'off' from '0% influence' in IK constraints

Depending on the situation, the IK solver will use the influence slider
in different ways. If there is no pole vector, it will interpolate between
the target position and the bone's current position. This means that an
influence of 0% will still fully run the IK solver, which overwrites the
bone's transform matrix.

This commit introduces a distinction between '0% influence' and 'muted'.
The former still behaves as above, and the latter now actually skips the
IK solver for that bone chain.

There is a difference in target positions between '0% influence' and 'muted'
when it comes to tree IK (i.e. multiple bones with IK constraints, where the
IK chains at some point overlap at a common parent bone).

This should fix #115915.

0cb872ba52 IK: draw relationship line when influence is 0%, hide when OFF

When IK constraints are enabled was recently changed. This commit updates
the overlay drawing code to match.

The relationship line, between the start and the end of the IK chain, is
now drawn whenever the IK system is active for that chain. This means that
it's drawn even at 0% influence, but not when the constraint is OFF/muted.

85e48d68c1 Refactor: move code into a function of its own

19c3f99b3d Stop at the end of the IK chain

53cabd519d Add some debug prints

Brecht Van Lommel commented

2024-01-29 17:52:17 +01:00

iksolver_execute_tree is meant to solve all IK constraint from the pchan_root. But here position_skipped_bones handles all bones in the armature, which indeed is likely not thread safe.

For example the dependency graph might figure out that the left and right arm can be handled in parallel, but iksolver_execute_tree should then only affect the bones on one arm.

I would expect position_skipped_bones to only work on bones in tree->pchan.

`iksolver_execute_tree` is meant to solve all IK constraint from the `pchan_root`. But here `position_skipped_bones` handles all bones in the armature, which indeed is likely not thread safe. For example the dependency graph might figure out that the left and right arm can be handled in parallel, but `iksolver_execute_tree` should then only affect the bones on one arm. I would expect `position_skipped_bones` to only work on bones in `tree->pchan`.

Sybren A. Stüvel added 2 commits 2024-02-08 16:23:49 +01:00

90ed5556dd Merge remote-tracking branch 'origin/main' into pr/anim-ik-off-means-off

e9a585169e Run `make format`

Sybren A. Stüvel added 1 commit 2024-02-08 17:13:31 +01:00

17e6619983 Move the "cleaup" phase to a separate depsgraph operation

This is just for testing, not sure if it's the right way to approach this.

This pull request is marked as a work in progress.

View command line instructions.

Checkout

From your project repository, check out a new branch and test the changes.

git fetch -u pr/anim-ik-off-means-off:dr.sybren-pr/anim-ik-off-means-off

git checkout dr.sybren-pr/anim-ik-off-means-off

Sign in to join this conversation.

No reviewers

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

WIP: IK: distinguish 'off' from '0% influence' in IK constraints #117178

Checkout