The perf cost is just too great for these to be enabled always. The only
place where the assertions are kept always is in the batch APIs -- there
it's assumed the function is called on large enough data to offset this
overhead, plus since it's often dealing with large blocks of data the
memory safety is more important than various FP drifts which were the
usual case why other assertions were firing.
This makes it possible to conveniently do things like
Containers::StridedArrayView1D<Float> array = …;
Vector4 vector{NoInit};
Utility::copy(array, vector); // or the other way around
which is especially useful together with the new JSON classes. In some
cases this means the function is no longer constexpr, but those weren't
constexpr because it was useful for anything, they were only because it
was possible. So this breakage shouldn't do any harm I think.
The old one is deprecated, and will be removed in a future release.
Unfortunately, to avoid deprecation warnings, all use of NoInit in the
Math library temporarily have to be Magnum::NoInit This will be cleaned
up when the deprecated alias is removed.
Before neither of the lerp(), slerp() had the shortest path check, while
sclerp() had it. Now, to be consistent, none of them has it and there
are lerpShortestPath(), slerpShortestPath() and sclerpShortestPath()
functions that have the shortest path check.
This is different from other engines, where there's usually only the
shortest path interpolation by default and either an optional
"non-shortest-path" interpolation or no alternative at all. I like to
give the users a choice, so there's both versions and the
non-shortest-path version is the default, because -- at least in case of
lerp() -- this results in a quite significant perf difference (15%
faster), so why not have it. Preprocess your data instead ;)
The fix done with https://github.com/mosra/magnum/pull/122
(0e05c7289e) was not tested properly (see
previous commit) and thus this code path never worked. This properly
lerps the translation part and recombines it with the rotation instead
of interpolating just a part of it.
Also I'm no longer having any "dotResult" that's done only on the
vector part of the rotation, but instead using the full
rotation quaternion dot product. I have no idea why it was done this
way. This branch was also never properly tested -- it'll be with the
introduction of "shortest path" variants in the next commit.
When rotation is identical, the rotation of the first dual quaternion is
returned instead, together with the linearly interpolated translation of
both (lerp of the vectors of the dual part). The additional include is
needed for `Math::lerp(Vector<3, T>, Vector<3, T>, T)`.
Signed-off-by: Squareys <Squareys@googlemail.com>
Making use of sincos() for Dual numbers, constructing DualQuaternion
from dual vector and scalar parts and using
DualQuaternion::isNormalized(). Also updated the math equation to be
consistent with conventions elsewhere.