Optimization in quaternion::rotationFromTo from Robert Eisele

Turns out we can avoid a square root and a division.
Patch comes even with proof: https://raw.org/proof/quaternion-from-two-vectors
(I also tested it a while and indeed got same results)


git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@6511 dfc29bdd-3216-0410-991c-e03cc46cb475
This commit is contained in:
cutealien 2023-07-30 16:34:24 +00:00
parent f7de9091a1
commit 2bfe7506c2
2 changed files with 5 additions and 4 deletions

View File

@ -1,6 +1,7 @@
-------------------------- --------------------------
Changes in 1.9 (not yet released) Changes in 1.9 (not yet released)
- Optimize quaternion::rotationFromTo. Thanks @Robert Eisele for patch and proof (https://raw.org/proof/quaternion-from-two-vectors)
- Shader materials shows now how to pass material values. - Shader materials shows now how to pass material values.
In 1.8 we could still use gl_FrontMaterial, but that is no longer supported in shaders In 1.8 we could still use gl_FrontMaterial, but that is no longer supported in shaders
- Get IMeshBuffer::append functions working for a few more cases and adapt interface so one can prevent the BoundingBox update. - Get IMeshBuffer::append functions working for a few more cases and adapt interface so one can prevent the BoundingBox update.

View File

@ -721,6 +721,8 @@ inline core::quaternion& quaternion::makeIdentity()
inline core::quaternion& quaternion::rotationFromTo(const vector3df& from, const vector3df& to) inline core::quaternion& quaternion::rotationFromTo(const vector3df& from, const vector3df& to)
{ {
// Based on Stan Melax's article in Game Programming Gems // Based on Stan Melax's article in Game Programming Gems
// Optimized by Robert Eisele: https://raw.org/proof/quaternion-from-two-vectors
// Copy, since cannot modify local // Copy, since cannot modify local
vector3df v0 = from; vector3df v0 = from;
vector3df v1 = to; vector3df v1 = to;
@ -745,10 +747,8 @@ inline core::quaternion& quaternion::rotationFromTo(const vector3df& from, const
return set(axis.X, axis.Y, axis.Z, 0).normalize(); return set(axis.X, axis.Y, axis.Z, 0).normalize();
} }
const f32 s = sqrtf( (1+d)*2 ); // optimize inv_sqrt const vector3df c = v0.crossProduct(v1);
const f32 invs = 1.f / s; return set(c.X, c.Y, c.Z, 1 + d).normalize();
const vector3df c = v0.crossProduct(v1)*invs;
return set(c.X, c.Y, c.Z, s * 0.5f).normalize();
} }