AutoMovie API
    Preparing search index...
    • The actual posed composite tensor about its own centre, kg m squared. Only the intrinsic rotation prerequisites share exact numeric work; every current mass shift and segment-order sum is performed again.

      States tensor units, centre and which current contributions remain evaluated. Delegates R diag(I) R transpose and the parallel-axis sum to the physical tensor owner's prepared reader; translated centres never enter an intrinsic tensor's equivalence key. A live lookup lazily acquires the operation's tensor reader at its first actual tensor query, then supplies the existing posed parts and centre. No inertia term, shift or physical check is omitted, and no symmetric-entry copy changes the original arithmetic. O(S) comparisons and ordered sums for S parts; unchanged principal moments and rotations avoid repeated intrinsic axis rotations and products. The physical owner compares copied moment/quaternion components at each segment index; the same token fixes model order without retaining a target-dependent tensor result. Three fresh tensor rows escape. The token retains at most one fixed-size intrinsic record per current segment and releases the reader with all model state.

      Returns number[][]