Quaternions #1796
Quaternions #1796
Conversation
|
Guess you already checked this and not the most critical thing atm, but since I looked it up, so it can serve as reference for later on: Some possible mathematical simplifications for 4x4 det: blender code (not sure where their method comes from but seems to do less operations). Also for reference in case it can be useful, their quaternion code. Also, from wikipedia: Being A B C D the 2x2 blocks. this would also possibly result in less operations. There's lot of literature on the subject, specially re simd optimizations, but anyways those seem straighforward simplifications. Looking great, maybe some tests would be in order? |
I'm going to switch most of the engine internals to quaternions, so there will be lots of manual tests :) |
lhog
force-pushed
the
BAR-quat
branch
5 times, most recently
from
0bb4f3f to
cb56e17
Compare
rts/System/float3.cpp
Outdated
| auto mi = std::min_element(std::begin(xyz), std::end(xyz), [](const auto& a, const auto& b) { return math::fabs(a) < math::fabs(b); }); | ||
| auto Mi = std::max_element(std::begin(xyz), std::end(xyz), [](const auto& a, const auto& b) { return math::fabs(a) > math::fabs(b); }); |
There was a problem hiding this comment.
Why the opposite comparator? Won't it this result in mi and Mi being the same? If there are two minimums then it will still return the same one in both. See https://godbolt.org/z/cdPd4h7Pe
If this really is just looking for min and max then consider auto [mi, Mi] = std::minmax_element(...) to look for both at the same time.
There was a problem hiding this comment.
Oh lol, you're right, nice catch!
…ns to employ SIMD
…del pieces.
Non-uniform scaling doesn't produce affine transformation matrices upon their multiplication,
nor could be represented by {rotation, translation, scaling} tuple,
shouldn't be a big deal as non-uniform scaling was never supported by the engine in the first place
No description provided.