glam/lib.rs
1/*!
2# glam
3
4`glam` is a simple and fast linear algebra library for games and graphics.
5
6## Features
7
8* [`f32`](mod@f32) types
9 * vectors: [`Vec2`], [`Vec3`], [`Vec3A`] and [`Vec4`]
10 * square matrices: [`Mat2`], [`Mat3`], [`Mat3A`] and [`Mat4`]
11 * a quaternion type: [`Quat`]
12 * affine transformation types: [`Affine2`] and [`Affine3A`]
13* [`f64`](mod@f64) types
14 * vectors: [`DVec2`], [`DVec3`] and [`DVec4`]
15 * square matrices: [`DMat2`], [`DMat3`] and [`DMat4`]
16 * a quaternion type: [`DQuat`]
17 * affine transformation types: [`DAffine2`] and [`DAffine3`]
18* [`i8`](mod@i8) types
19 * vectors: [`I8Vec2`], [`I8Vec3`] and [`I8Vec4`]
20* [`u8`](mod@u8) types
21 * vectors: [`U8Vec2`], [`U8Vec3`] and [`U8Vec4`]
22* [`i16`](mod@i16) types
23 * vectors: [`I16Vec2`], [`I16Vec3`] and [`I16Vec4`]
24* [`u16`](mod@u16) types
25 * vectors: [`U16Vec2`], [`U16Vec3`] and [`U16Vec4`]
26* [`i32`](mod@i32) types
27 * vectors: [`IVec2`], [`IVec3`] and [`IVec4`]
28* [`u32`](mod@u32) types
29 * vectors: [`UVec2`], [`UVec3`] and [`UVec4`]
30* [`i64`](mod@i64) types
31 * vectors: [`I64Vec2`], [`I64Vec3`] and [`I64Vec4`]
32* [`u64`](mod@u64) types
33 * vectors: [`U64Vec2`], [`U64Vec3`] and [`U64Vec4`]
34* [`usize`](mod@usize) types
35 * vectors: [`USizeVec2`], [`USizeVec3`] and [`USizeVec4`]
36* [`bool`](mod@bool) types
37 * vectors: [`BVec2`], [`BVec3`] and [`BVec4`]
38
39## SIMD
40
41`glam` is built with SIMD in mind. Many `f32` types use 128-bit SIMD vector types for storage
42and/or implementation. The use of SIMD generally enables better performance than using primitive
43numeric types such as `f32`.
44
45Some `glam` types use SIMD for storage meaning they are 16 byte aligned, these types include
46`Mat2`, `Mat3A`, `Mat4`, `Quat`, `Vec3A`, `Vec4`, `Affine2` an `Affine3A`. Types
47with an `A` suffix are a SIMD alternative to a scalar type, e.g. `Vec3` uses `f32` storage and
48`Vec3A` uses SIMD storage.
49
50When SIMD is not available on the target the types will maintain 16 byte alignment and internal
51padding so that object sizes and layouts will not change between architectures. There are scalar
52math fallback implementations exist when SIMD is not available. It is intended to add support for
53other SIMD architectures once they appear in stable Rust.
54
55Currently only SSE2 on x86/x86_64, NEON on Aarch64, and simd128 on WASM are supported.
56
57## Vec3A and Mat3A
58
59`Vec3A` is a SIMD optimized version of the `Vec3` type, which due to 16 byte alignment results
60in `Vec3A` containing 4 bytes of padding making it 16 bytes in size in total. `Mat3A` is composed
61of three `Vec3A` columns.
62
63| Type | `f32` bytes | Align bytes | Size bytes | Padding |
64|:-----------|------------:|------------:|-----------:|--------:|
65|[`Vec3`] | 12| 4| 12| 0|
66|[`Vec3A`] | 12| 16| 16| 4|
67|[`Mat3`] | 36| 4| 36| 0|
68|[`Mat3A`] | 36| 16| 48| 12|
69
70Despite this wasted space the SIMD implementations tend to outperform `f32` implementations in
71[**mathbench**](https://github.com/bitshifter/mathbench-rs) benchmarks.
72
73`glam` treats [`Vec3`] as the default 3D vector type and [`Vec3A`] a special case for optimization.
74When methods need to return a 3D vector they will generally return [`Vec3`].
75
76There are [`From`] trait implementations for converting from [`Vec4`] to a [`Vec3A`] and between
77[`Vec3`] and [`Vec3A`] (and vice versa).
78
79```
80use glam::{Vec3, Vec3A, Vec4};
81
82let v4 = Vec4::new(1.0, 2.0, 3.0, 4.0);
83
84// Convert from `Vec4` to `Vec3A`, this is a no-op if SIMD is supported.
85// We use an explicit method here instead of a From impl as data is lost in the conversion.
86let v3a = Vec3A::from_vec4(v4);
87assert_eq!(Vec3A::new(1.0, 2.0, 3.0), v3a);
88
89// Convert from `Vec3A` to `Vec3`.
90let v3 = Vec3::from(v3a);
91assert_eq!(Vec3::new(1.0, 2.0, 3.0), v3);
92
93// Convert from `Vec3` to `Vec3A`.
94let v3a = Vec3A::from(v3);
95assert_eq!(Vec3A::new(1.0, 2.0, 3.0), v3a);
96```
97
98## Affine2 and Affine3A
99
100`Affine2` and `Affine3A` are composed of a linear transform matrix and a vector translation. The
101represent 2D and 3D affine transformations which are commonly used in games.
102
103The table below shows the performance advantage of `Affine2` over `Mat3A` and `Mat3A` over `Mat3`.
104
105| operation | `Mat3` | `Mat3A` | `Affine2` |
106|--------------------|-------------|------------|------------|
107| inverse | 11.4±0.09ns | 7.1±0.09ns | 5.4±0.06ns |
108| mul self | 10.5±0.04ns | 5.2±0.05ns | 4.0±0.05ns |
109| transform point2 | 2.7±0.02ns | 2.7±0.03ns | 2.8±0.04ns |
110| transform vector2 | 2.6±0.01ns | 2.6±0.03ns | 2.3±0.02ns |
111
112Performance is much closer between `Mat4` and `Affine3A` with the affine type being faster to
113invert.
114
115| operation | `Mat4` | `Affine3A` |
116|--------------------|-------------|-------------|
117| inverse | 15.9±0.11ns | 10.8±0.06ns |
118| mul self | 7.3±0.05ns | 7.0±0.06ns |
119| transform point3 | 3.6±0.02ns | 4.3±0.04ns |
120| transform point3a | 3.0±0.02ns | 3.0±0.04ns |
121| transform vector3 | 4.1±0.02ns | 3.9±0.04ns |
122| transform vector3a | 2.8±0.02ns | 2.8±0.02ns |
123
124Benchmarks were taken on an Intel Core i7-4710HQ.
125
126## Linear algebra conventions
127
128`glam` interprets vectors as column matrices (also known as column vectors) meaning when
129transforming a vector with a matrix the matrix goes on the left.
130
131```
132use glam::{Mat3, Vec3};
133let m = Mat3::IDENTITY;
134let x = Vec3::X;
135let v = m * x;
136assert_eq!(v, x);
137```
138
139Matrices are stored in memory in column-major order.
140
141All angles are in radians. Rust provides the `f32::to_radians()` and `f64::to_radians()` methods to
142convert from degrees.
143
144## Direct element access
145
146Because some types may internally be implemented using SIMD types, direct access to vector elements
147is supported by implementing the [`Deref`] and [`DerefMut`] traits.
148
149```
150use glam::Vec3A;
151let mut v = Vec3A::new(1.0, 2.0, 3.0);
152assert_eq!(3.0, v.z);
153v.z += 1.0;
154assert_eq!(4.0, v.z);
155```
156
157[`Deref`]: https://doc.rust-lang.org/std/ops/trait.Deref.html
158[`DerefMut`]: https://doc.rust-lang.org/std/ops/trait.DerefMut.html
159
160## glam assertions
161
162`glam` does not enforce validity checks on method parameters at runtime. For example methods that
163require normalized vectors as input such as `Quat::from_axis_angle(axis, angle)` will not check
164that axis is a valid normalized vector. To help catch unintended misuse of `glam` the
165`debug-glam-assert` or `glam-assert` features can be enabled to add checks ensure that inputs to
166are valid.
167
168## Vector swizzles
169
170`glam` vector types have functions allowing elements of vectors to be reordered, this includes
171creating a vector of a different size from the vectors elements.
172
173The swizzle functions are implemented using traits to add them to each vector type. This is
174primarily because there are a lot of swizzle functions which can obfuscate the other vector
175functions in documentation and so on. The traits are [`Vec2Swizzles`], [`Vec3Swizzles`] and
176[`Vec4Swizzles`].
177
178Note that the [`Vec3Swizzles`] implementation for [`Vec3A`] will return a [`Vec3A`] for 3 element
179swizzles, all other implementations will return [`Vec3`].
180
181```
182use glam::{swizzles::*, Vec2, Vec3, Vec3A, Vec4};
183
184let v = Vec4::new(1.0, 2.0, 3.0, 4.0);
185
186// Reverse elements of `v`, if SIMD is supported this will use a vector shuffle.
187let wzyx = v.wzyx();
188assert_eq!(Vec4::new(4.0, 3.0, 2.0, 1.0), wzyx);
189
190// Swizzle the yzw elements of `v` into a `Vec3`
191let yzw = v.yzw();
192assert_eq!(Vec3::new(2.0, 3.0, 4.0), yzw);
193
194// To swizzle a `Vec4` into a `Vec3A` swizzle the `Vec4` first then convert to
195// `Vec3A`. If SIMD is supported this will use a vector shuffle. The last
196// element of the shuffled `Vec4` is ignored by the `Vec3A`.
197let yzw = Vec3A::from_vec4(v.yzwx());
198assert_eq!(Vec3A::new(2.0, 3.0, 4.0), yzw);
199
200// You can swizzle from a `Vec4` to a `Vec2`
201let xy = v.xy();
202assert_eq!(Vec2::new(1.0, 2.0), xy);
203
204// And back again
205let yyxx = xy.yyxx();
206assert_eq!(Vec4::new(2.0, 2.0, 1.0, 1.0), yyxx);
207```
208
209## SIMD and scalar consistency
210
211`glam` types implement `serde` `Serialize` and `Deserialize` traits to ensure
212that they will serialize and deserialize exactly the same whether or not
213SIMD support is being used.
214
215The SIMD versions implement the `core::fmt::Debug` and `core::fmt::Display`
216traits so they print the same as the scalar version.
217
218```
219use glam::Vec4;
220let a = Vec4::new(1.0, 2.0, 3.0, 4.0);
221assert_eq!(format!("{}", a), "[1, 2, 3, 4]");
222```
223
224## Feature gates
225
226All `glam` dependencies are optional, however some are required for tests
227and benchmarks.
228
229* `std` - the default feature, has no dependencies.
230* `approx` - traits and macros for approximate float comparisons
231* `bytemuck` - for casting into slices of bytes
232* `libm` - uses `libm` math functions instead of `std`
233* `nostd-libm` - uses `libm` math functions if `std` is not available
234* `mint` - for interoperating with other 3D math libraries
235* `rand` - implementations of `Distribution` trait for all `glam` types.
236* `rkyv` - implementations of `Archive`, `Serialize` and `Deserialize` for all
237 `glam` types. Note that serialization is not interoperable with and without the
238 `scalar-math` feature. It should work between all other builds of `glam`.
239 Endian conversion is currently not supported
240* `bytecheck` - to perform archive validation when using the `rkyv` feature
241* `encase` - `encase` trait implementations for `glam` types.
242* `serde` - implementations of `Serialize` and `Deserialize` for all `glam`
243 types. Note that serialization should work between builds of `glam` with and without SIMD enabled
244* `speedy` - implementations of `speedy`'s `Readable` and `Writable` for all `glam` types.
245* `scalar-math` - disables SIMD support and uses native alignment for all types.
246* `debug-glam-assert` - adds assertions in debug builds which check the validity of parameters
247 passed to `glam` to help catch runtime errors.
248* `glam-assert` - adds assertions to all builds which check the validity of parameters passed to
249 `glam` to help catch runtime errors.
250* `cuda` - forces `glam` types to match expected cuda alignment
251* `fast-math` - By default, glam attempts to provide bit-for-bit identical
252 results on all platforms. Using this feature will enable platform specific
253 optimizations that may not be identical to other platforms. **Intermediate
254 libraries should not use this feature and defer the decision to the final
255 binary build**.
256* `core-simd` - enables SIMD support via the portable simd module. This is an
257 unstable feature which requires a nightly Rust toolchain and `std` support.
258
259## Minimum Supported Rust Version (MSRV)
260
261The minimum supported Rust version is `1.68.2`.
262
263*/
264#![doc(html_root_url = "https://docs.rs/glam/0.30.8")]
265#![cfg_attr(not(feature = "std"), no_std)]
266#![cfg_attr(target_arch = "spirv", feature(repr_simd))]
267#![deny(
268 rust_2018_compatibility,
269 rust_2018_idioms,
270 future_incompatible,
271 nonstandard_style
272)]
273// clippy doesn't like `to_array(&self)`
274#![allow(clippy::wrong_self_convention)]
275#![cfg_attr(
276 all(feature = "core-simd", not(feature = "scalar-math")),
277 feature(portable_simd)
278)]
279
280#[cfg(all(
281 not(feature = "std"),
282 not(feature = "libm"),
283 not(feature = "nostd-libm")
284))]
285compile_error!(
286 "You must specify a math backend. Consider enabling either `std`, `libm`, or `nostd-libm`."
287);
288
289#[macro_use]
290mod macros;
291
292mod align16;
293mod deref;
294mod euler;
295mod features;
296
297#[cfg(all(
298 target_arch = "aarch64",
299 not(any(feature = "core-simd", feature = "scalar-math"))
300))]
301mod neon;
302
303#[cfg(target_arch = "spirv")]
304mod spirv;
305
306#[cfg(all(
307 target_feature = "sse2",
308 not(any(feature = "core-simd", feature = "scalar-math"))
309))]
310mod sse2;
311
312#[cfg(all(
313 target_feature = "simd128",
314 not(any(feature = "core-simd", feature = "scalar-math"))
315))]
316mod wasm32;
317
318#[cfg(all(feature = "core-simd", not(feature = "scalar-math")))]
319mod coresimd;
320
321#[cfg(all(
322 target_feature = "sse2",
323 not(any(feature = "core-simd", feature = "scalar-math"))
324))]
325use align16::Align16;
326
327/** `bool` vector mask types. */
328pub mod bool;
329pub use self::bool::*;
330
331/** `f32` vector, quaternion and matrix types. */
332pub mod f32;
333pub use self::f32::*;
334
335/** `f64` vector, quaternion and matrix types. */
336pub mod f64;
337pub use self::f64::*;
338
339/** `i8` vector types. */
340pub mod i8;
341pub use self::i8::*;
342
343/** `u8` vector types. */
344pub mod u8;
345pub use self::u8::*;
346
347/** `i16` vector types. */
348pub mod i16;
349pub use self::i16::*;
350
351/** `u16` vector types. */
352pub mod u16;
353pub use self::u16::*;
354
355/** `i32` vector types. */
356pub mod i32;
357pub use self::i32::*;
358
359/** `u32` vector types. */
360pub mod u32;
361pub use self::u32::*;
362
363/** `i64` vector types. */
364pub mod i64;
365pub use self::i64::*;
366
367/** `u64` vector types. */
368pub mod u64;
369pub use self::u64::*;
370
371/** `usize` vector types. */
372pub mod usize;
373pub use self::usize::*;
374
375/** Traits adding swizzle methods to all vector types. */
376pub mod swizzles;
377pub use self::swizzles::{Vec2Swizzles, Vec3Swizzles, Vec4Swizzles};
378
379/** Rotation Helper */
380pub use euler::EulerRot;
381
382/** A trait for extending [`prim@f32`] and [`prim@f64`] with extra methods. */
383mod float;
384pub use float::FloatExt;