magnum

Commit Graph

Author	SHA1	Message	Date
Pablo Escobar	daa6d9125e	Math: add fmod function	6 years ago
Vladimír Vondruš	ca8172311d	Math: FFS, GCC.	6 years ago
Vladimír Vondruš	12577ce07f	Math: verify normal matrix calculation against the ground truth code. It gives the same result, nevertheless something is not right when it comes to negatively scaled meshes. Postponing the rest of the investigation to later.	6 years ago
Vladimír Vondruš	1026aee80e	Moved the NoInit tag from Math to the root namespace. 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.	6 years ago
Vladimír Vondruš	19e0e96d74	Updated copyright year.	6 years ago
Vladimír Vondruš	5be7fbe1e8	Math: make Range2D and Range3D properly convertible from std::pair. Not sure what was I doing in `64f8a90095`.	6 years ago
Vladimír Vondruš	060d8548c6	Math: benchmark some sqrt variants.	6 years ago
Vladimír Vondruš	e05fda0db4	Math: this got forgotten.	6 years ago
Vladimír Vondruš	573125d0a9	Math: make cross() 10x faster in Debug. And the Vector3 version 5% slower in Release, on GCC at least. FFS, what was I thinking with the gather() things. Nice in user code, extremely bad in library code.	6 years ago
Vladimír Vondruš	341a497932	Math: benchmark vector dot() and cross(). Have to do some precision improvements, so a baseline is needed. The debug perf is beyond awful, actually.	6 years ago
Vladimír Vondruš	65a2ced2a7	Make this compile and test cleanly with CORRADE_NO_ASSERT defined.	6 years ago
Vladimír Vondruš	99a5e1d246	Math: test that slerp() is immune against too long quats as well.	6 years ago
Vladimír Vondruš	498f4c59e1	Math: fix angle() returning NaN for close arguments.	6 years ago
Vladimír Vondruš	94cba139ff	Math: improve and clean up angle() tests.	6 years ago
Vladimír Vondruš	f2f66d764f	Math: make TypeTraits::epsilon() consistent with Corrade. The precision stays the same, but the long double variant is now exposed on Emscripten as well, following a similar change in Corrade. Additionally, the alien-looking _EQUALITY_PRECISION macros are now unused and deprecated. For some reason these weren't ever prefixed with MAGNUM_, and the ability to override those is an extremely rare use case that would break half of the assumptions everywhere, so better not allow that at all. The TypeTraits test is further extended to compare directly the epsilons between Magnum and Corrade, in addition to verifying that TestSuite and TypeTraits have the same comparison results.	6 years ago
Vladimír Vondruš	f3566f637f	Math: add integer-to-integer castInto() variants also. Currently one of the types is always 32-bit and preserving signedness, because that's what I need right now. Other variants might get added later.	6 years ago
Vladimír Vondruš	881003d90b	Test: adapt tests to change in TestSuite float epsilon precisions.	6 years ago
Vladimír Vondruš	ab0cba2573	Math: minor cleanup.	6 years ago
Vladimír Vondruš	53151f606a	Math: ensure TypeTraits fuzzy compare is consistent with TestSuite. It wasn't until now (TestSuite was wrong), so prevent similar issues from happening in the future.	6 years ago
Vladimír Vondruš	526254912e	doc: updated credits and changelog.	6 years ago
Squareys	a0e6ac1ed0	Math: Non-const overloads for Frustum::(left\|right\|top\|bottom\|near\|far\|[]) Signed-off-by: Squareys <squareys@googlemail.com>	6 years ago
Vladimír Vondruš	d2513cac3c	Math: make Color default alpha values work with Half. There's some fugliness needed for the MSVC 2017 crash workaround but otherwise all good.	6 years ago
Vladimír Vondruš	c8afe0f732	Math: make Vector::pad() default argument work with Half. Got lucky enough that the default zero-initialization constructor does exactly what I need. Otherwise ugh I don't know what I would do.	6 years ago
Vladimír Vondruš	50e21015eb	Math: make it possible to construct Half from a double. A lot of generic code uses T(1.0) and right now that caused an ambiguity between the Float and UnsignedShort constructors.	6 years ago
Vladimír Vondruš	10dd468eca	Math: minimal support for Half in TypeTraits.	6 years ago
Vladimír Vondruš	f5a5e10c50	Math: batch table-based half-float packing/unpacking. It's a lot of code, but it still seems to be the fastest option of all we have. This was the original idea when implementing half-float support in 2016 but then I shelved it in favor of a simpler (but slower) code, keeping the table only for the benchmark, calculated at runtime. But now we need a batch version of this, so this comes handy.	6 years ago
Vladimír Vondruš	69e370c4c7	GCC 4.8, happy to see you're still around, causing problems.	6 years ago
Vladimír Vondruš	88fabb1ef6	Math: reduce header dependencies a bit.	6 years ago
Vladimír Vondruš	daf471381e	Math: add batch integer packing/unpacking and casting functions. To be used for conversion of the upcoming mesh attribute types.	6 years ago
Vladimír Vondruš	c85b537937	Revert "Math: preserve signs in the Matrix[34]::scaling() getter." Turns out this seriously broke properites that should always hold, such as that RS = M for a R and S extracted out of M. Another way to fix this would be flipping signs in the rotation() / rotationShear() / ... queries as well, but that adds extra operations to both rotation() and scaling() and at that point it's just better to do what was done the whole decade before. To ensure this doesn't happen again and doesn't cost me a whole day of debugging in an end-user app (SceneGraph TRS transformations got broken because those assume that rotation() and scaling() do* combine back, which is a reasonable thing to assume), the docs are extended to note this and there's an explicit test for both negative scale and large angle corner cases as well. This reverts commit `9aa68715db`.	6 years ago
Nghia Truong	1ef813fdad	Math: add reflect() and refract() functions. Co-authored-by: Vladimír Vondruš <mosra@centrum.cz>	6 years ago
Vladimír Vondruš	b2328f57cd	Math: fall back to lerp when slerp'd quaternions are too close. Returning just the first argument had several issues, one of them was causing iterative interpolations to get stuck at the initial position forever.	6 years ago
Vladimír Vondruš	c4a3b273e6	Math: provide transformVector() also on the DualThing classes. To make them more consistent with Matrix3/4, making switching from one to the other easier.	6 years ago
Vladimír Vondruš	ea92c5b36f	Math: clarify that implicit conversion of Foo to a DualFoo is rotation. Had to remind myself about that.	6 years ago
Vladimír Vondruš	a6dd63af0f	Math: make batch functions actually usable. No longer it's needed to add ugly explicit template types or casts.	6 years ago
Vladimír Vondruš	d69b6f8c40	Math: test and fix array-taking isInf()/isNan() also with vectors. Next time I need to test properly.	6 years ago
Vladimír Vondruš	71b775aeb1	Math: minor test cleanup.	6 years ago
Vladimír Vondruš	9aa68715db	Math: preserve signs in the Matrix[34]::scaling() getter. This is a breaking change, but I think it is worth doing. Because now Matrix4::scaling(vec).scaling() == vec which was true also for translation and other.	6 years ago
Vladimír Vondruš	15e5bfa09e	Update credits and changelog, minor cleanup.	7 years ago
Marco Melorio	097289850a	Add toEuler() to Quaternion	7 years ago
Vladimír Vondruš	515637c76a	Math: implement a robust and faster Matrix4::normalMatrix(). This change makes the operations split into a bunch of separate functions, making the parts easier to document, however with a slight negative effect on debug performance: Starting Magnum::Math::Test::MatrixBenchmark with 16 test cases... INFO Benchmarking a debug build. BENCH [01] 94.54 ± 3.62 ns multiply3()@499x10000 (wall time) BENCH [02] 183.47 ± 7.50 ns multiply4()@499x10000 (wall time) BENCH [03] 318.11 ± 11.59 ns comatrix3()@49x10000 (wall time) BENCH [04] 379.51 ± 12.17 ns invert3()@49x10000 (wall time) BENCH [05] 448.23 ± 17.61 ns invert3GaussJordan()@49x10000 (wall time) BENCH [06] 338.96 ± 11.61 ns invert3Rigid()@49x10000 (wall time) BENCH [07] 206.37 ± 10.59 ns invert3Orthogonal()@49x10000 (wall time) BENCH [08] 879.40 ± 20.03 ns comatrix4()@49x10000 (wall time) BENCH [09] 1.16 ± 0.03 µs invert4()@49x10000 (wall time) BENCH [10] 825.40 ± 17.34 ns invert4GaussJordan()@49x10000 (wall time) BENCH [11] 534.86 ± 15.73 ns invert4Rigid()@49x10000 (wall time) BENCH [12] 347.36 ± 11.62 ns invert4Orthogonal()@49x10000 (wall time) BENCH [13] 65.70 ± 6.73 ns transformVector3()@999x10000 (wall time) BENCH [14] 62.56 ± 3.09 ns transformPoint3()@999x10000 (wall time) BENCH [15] 81.25 ± 2.78 ns transformVector4()@999x10000 (wall time) BENCH [16] 82.14 ± 4.26 ns transformPoint4()@999x10000 (wall time) Finished Magnum::Math::Test::MatrixBenchmark with 0 errors out of 5500 checks.	7 years ago
Vladimír Vondruš	e3a820f468	Math: benchmark basic Matrix operations. So I'm sure the next additions don't make things worse -- in the debug build, in particular. Current state: Starting Magnum::Math::Test::MatrixBenchmark with 14 test cases... INFO Benchmarking a debug build. BENCH [01] 102.25 ± 4.58 ns multiply3()@499x10000 (wall time) BENCH [02] 183.84 ± 7.75 ns multiply4()@499x10000 (wall time) BENCH [03] 329.84 ± 20.50 ns invert3()@49x10000 (wall time) BENCH [04] 442.08 ± 13.47 ns invert3GaussJordan()@49x10000 (wall time) BENCH [05] 331.23 ± 13.85 ns invert3Rigid()@49x10000 (wall time) BENCH [06] 206.01 ± 11.13 ns invert3Orthogonal()@49x10000 (wall time) BENCH [07] 1.01 ± 0.03 µs invert4()@49x10000 (wall time) BENCH [08] 826.77 ± 19.33 ns invert4GaussJordan()@49x10000 (wall time) BENCH [09] 520.12 ± 21.97 ns invert4Rigid()@49x10000 (wall time) BENCH [10] 344.20 ± 13.32 ns invert4Orthogonal()@49x10000 (wall time) BENCH [11] 61.00 ± 2.89 ns transformVector3()@999x10000 (wall time) BENCH [12] 61.55 ± 3.35 ns transformPoint3()@999x10000 (wall time) BENCH [13] 80.65 ± 2.99 ns transformVector4()@999x10000 (wall time) BENCH [14] 80.86 ± 3.32 ns transformPoint4()@999x10000 (wall time) Finished Magnum::Math::Test::MatrixBenchmark with 0 errors out of 5400 checks.	7 years ago
Vladimír Vondruš	7f70dc7385	Math: XFAIL most (all?) NoInit tests in Release builds.	7 years ago
Vladimír Vondruš	17b3c8fac5	Math: support numeric indexing in gather() / scatter() as well.	7 years ago
Vladimír Vondruš	acc8d8994b	Math: added scatter(), renamed swizzle() to gather(). The swizzle() function is still present, only marked as deprecated and scheduled to be removed in some future release.	7 years ago
Vladimír Vondruš	4bc84f6815	Math: expand UnderlyingTypeOf to support vectors and matrices. This allows getting underlying type of those without needing to include them. Yay.	7 years ago
Vladimír Vondruš	1887fe0ab9	Math: added a scalar equal() / notEqual() variant.	7 years ago
Vladimír Vondruš	13b4bf8461	Math: use __builtin_sincos for sincos() on GCC. Makes this thing faster on Debug. But I'm not sure whether it's due to less function calls or this intrinsic being actually faster, heh. Probably a combination of both.	7 years ago
Vladimír Vondruš	a28e14491c	Math: verify that TypeTraits work correctly for signed infinities.	7 years ago
Vladimír Vondruš	3cf9dbf409	Math: added component-wise equal()/notEqual() for vectors.	7 years ago

1 2 3 4 5 ...

342 Commits (5d78d7fe596b2f6d2d90c347b5628ad7262e385b)