This makes the Matrix3/4 and Decomposed `look_at_*` functions consistent
with looking at a center/focus point and `look_to_*` functions consistent
with looking in a direction.
Corresponding functions have been added to Matrix4 and Decomposed
and are now consistent.
Matrix3, Matrix2, and the Rotation trait are only partially updated.
* Now there is only one one
* Rotation no longer has a parameter
* Moved some type parameters to associated types
* Relaxed some bounds and simplified a bound
* Removed unnecessary bound in
* Deduplicated multiplication code
491: Fix Matrix2::look_at, add look_at_stable r=kvark a=blargg
## Changes
1. Fixes `Matrix2::look_at`
2. Adds tests for `look_at`
3. Adds a new function, `look_at_stable`
## Notes
I added a new function for 2d look at rotation. `look_at` is a bit weird in practice for 2d. For example, if you are making a basis matrix to orient a 2d character to look at a point, `look_at` will flip the character as they rotate past `up` or `-up` vectors. This is the best match for what look_at is supposed to do, I think.
`look_at_stable` will not flip based on orientation, you just pass in which way to flip. This is a bit easier to use to rotate 2d characters.
`look_at_stable` could have a better name. I think we can also consider removing the flip param, and just let the user flip the matrix with a transform later.
Co-authored-by: blargg <tomjankauski@gmail.com>
Most users probably don't use the rand impls, so the `rand` crate pulls
a large number of dependencies into the dependency tree which is just
wasted compilation time.
Replace all `impl .. for Struct<<A as Angle>::Unitless>` by
`Struct<A::Unitless>`. This allows -Zsave-analysis to work on nightly, and
I think this is more readable too.
This adds `Sum` trait for the `MatrixN`, `VectorN`, `Quaternion`
structures and the `Product` trait for `MatrixN`, `BasisN` and
`Quaternion`.
It also add constraints on the `Rotation` and `SquareMatrix` to require
the `Product` trait and `VectorSpace` to require `Sum`.
- Add an opt-in SIMD support for the module. The feature requires crate
`simd` and specialization, thus can only be enabled under nightly. Under
the given benchmark certain operations were able to be up to 60% faster.
Currently the supported types as well as operations are highly limited.
- Clean up some deadly tests. Also add new tests for SIMD.
