#![allow(dead_code)]
#[cfg(feature = "rand")]
use rand::{
distributions::{Distribution, Standard},
Rng,
};
use crate::{
f32::{Vec2, Vec4, X_AXIS, Y_AXIS, Z_AXIS},
Align16,
};
#[cfg(target_arch = "x86")]
use std::arch::x86::*;
#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::*;
use std::{cmp::Ordering, f32, fmt, mem::MaybeUninit, ops::*};
#[derive(Clone, Copy, Debug)]
#[repr(C)]
pub struct Vec3(pub(crate) __m128);
impl Vec3 {
#[inline]
pub fn new(x: f32, y: f32, z: f32) -> Self {
unsafe { Self(_mm_set_ps(z, z, y, x)) }
}
#[inline]
pub fn zero() -> Self {
unsafe { Self(_mm_setzero_ps()) }
}
#[inline]
pub fn one() -> Self {
unsafe { Self(_mm_set1_ps(1.0)) }
}
#[inline]
pub fn unit_x() -> Self {
unsafe {
Self(_mm_load_ps(
&X_AXIS as *const Align16<(f32, f32, f32, f32)> as *const f32,
))
}
}
#[inline]
pub fn unit_y() -> Self {
unsafe {
Self(_mm_load_ps(
&Y_AXIS as *const Align16<(f32, f32, f32, f32)> as *const f32,
))
}
}
#[inline]
pub fn unit_z() -> Self {
unsafe {
Self(_mm_load_ps(
&Z_AXIS as *const Align16<(f32, f32, f32, f32)> as *const f32,
))
}
}
#[inline]
pub fn splat(v: f32) -> Self {
unsafe { Self(_mm_set_ps1(v)) }
}
#[inline]
pub fn extend(self, w: f32) -> Vec4 {
let mut temp: Vec4 = self.0.into();
temp.set_w(w);
temp
}
#[inline]
pub fn truncate(self) -> Vec2 {
let (x, y, _) = self.into();
Vec2::new(x, y)
}
#[inline]
pub fn x(self) -> f32 {
unsafe { _mm_cvtss_f32(self.0) }
}
#[inline]
pub fn y(self) -> f32 {
unsafe { _mm_cvtss_f32(_mm_shuffle_ps(self.0, self.0, 0b01_01_01_01)) }
}
#[inline]
pub fn z(self) -> f32 {
unsafe { _mm_cvtss_f32(_mm_shuffle_ps(self.0, self.0, 0b10_10_10_10)) }
}
#[inline]
pub fn set_x(&mut self, x: f32) {
unsafe {
self.0 = _mm_move_ss(self.0, _mm_set_ss(x));
}
}
#[inline]
pub fn set_y(&mut self, y: f32) {
unsafe {
let mut t = _mm_move_ss(self.0, _mm_set_ss(y));
t = _mm_shuffle_ps(t, t, 0b11_10_00_00);
self.0 = _mm_move_ss(t, self.0);
}
}
#[inline]
pub fn set_z(&mut self, z: f32) {
unsafe {
let mut t = _mm_move_ss(self.0, _mm_set_ss(z));
t = _mm_shuffle_ps(t, t, 0b11_00_01_00);
self.0 = _mm_move_ss(t, self.0);
}
}
#[inline]
pub(crate) fn dup_x(self) -> Self {
unsafe { Self(_mm_shuffle_ps(self.0, self.0, 0b00_00_00_00)) }
}
#[inline]
pub(crate) fn dup_y(self) -> Self {
unsafe { Self(_mm_shuffle_ps(self.0, self.0, 0b01_01_01_01)) }
}
#[inline]
pub(crate) fn dup_z(self) -> Self {
unsafe { Self(_mm_shuffle_ps(self.0, self.0, 0b10_10_10_10)) }
}
#[inline]
unsafe fn dot_as_m128(self, other: Self) -> __m128 {
let x2_y2_z2_w2 = _mm_mul_ps(self.0, other.0);
let y2_0_0_0 = _mm_shuffle_ps(x2_y2_z2_w2, x2_y2_z2_w2, 0b00_00_00_01);
let x2y2_0_0_0 = _mm_add_ss(x2_y2_z2_w2, y2_0_0_0);
let z2_0_0_0 = _mm_shuffle_ps(x2_y2_z2_w2, x2_y2_z2_w2, 0b00_00_00_10);
_mm_add_ss(x2y2_0_0_0, z2_0_0_0)
}
#[inline]
pub(crate) fn dot_as_vec3(self, other: Self) -> Self {
unsafe {
let dot_in_x = self.dot_as_m128(other);
Vec3(_mm_shuffle_ps(dot_in_x, dot_in_x, 0b00_00_00_00))
}
}
#[inline]
pub fn dot(self, other: Self) -> f32 {
unsafe { _mm_cvtss_f32(self.dot_as_m128(other)) }
}
#[inline]
pub fn cross(self, other: Self) -> Self {
unsafe {
let lhszxy = _mm_shuffle_ps(self.0, self.0, 0b01_01_00_10);
let rhszxy = _mm_shuffle_ps(other.0, other.0, 0b01_01_00_10);
let lhszxy_rhs = _mm_mul_ps(lhszxy, other.0);
let rhszxy_lhs = _mm_mul_ps(rhszxy, self.0);
let sub = _mm_sub_ps(lhszxy_rhs, rhszxy_lhs);
Self(_mm_shuffle_ps(sub, sub, 0b01_01_00_10))
}
}
#[inline]
pub fn length(self) -> f32 {
let dot = self.dot_as_vec3(self);
unsafe { _mm_cvtss_f32(_mm_sqrt_ps(dot.0)) }
}
#[inline]
pub fn length_squared(self) -> f32 {
self.dot(self)
}
#[inline]
pub fn length_reciprocal(self) -> f32 {
let dot = self.dot_as_vec3(self);
unsafe {
_mm_cvtss_f32(_mm_div_ps(_mm_set_ps1(1.0), _mm_sqrt_ps(dot.0)))
}
}
#[inline]
pub fn normalize(self) -> Self {
let dot = self.dot_as_vec3(self);
unsafe { Self(_mm_div_ps(self.0, _mm_sqrt_ps(dot.0))) }
}
#[inline]
pub fn min(self, other: Self) -> Self {
unsafe { Self(_mm_min_ps(self.0, other.0)) }
}
#[inline]
pub fn max(self, other: Self) -> Self {
unsafe { Self(_mm_max_ps(self.0, other.0)) }
}
#[inline]
pub fn min_element(self) -> f32 {
unsafe {
let v = self.0;
let v = _mm_min_ps(v, _mm_shuffle_ps(v, v, 0b01_01_10_10));
let v = _mm_min_ps(v, _mm_shuffle_ps(v, v, 0b00_00_00_01));
_mm_cvtss_f32(v)
}
}
#[inline]
pub fn max_element(self) -> f32 {
unsafe {
let v = self.0;
let v = _mm_max_ps(v, _mm_shuffle_ps(v, v, 0b00_00_10_10));
let v = _mm_max_ps(v, _mm_shuffle_ps(v, v, 0b00_00_00_01));
_mm_cvtss_f32(v)
}
}
#[inline]
pub fn cmpeq(self, other: Self) -> Vec3Mask {
unsafe { Vec3Mask(_mm_cmpeq_ps(self.0, other.0)) }
}
#[inline]
pub fn cmpne(self, other: Self) -> Vec3Mask {
unsafe { Vec3Mask(_mm_cmpneq_ps(self.0, other.0)) }
}
#[inline]
pub fn cmpge(self, other: Self) -> Vec3Mask {
unsafe { Vec3Mask(_mm_cmpge_ps(self.0, other.0)) }
}
#[inline]
pub fn cmpgt(self, other: Self) -> Vec3Mask {
unsafe { Vec3Mask(_mm_cmpgt_ps(self.0, other.0)) }
}
#[inline]
pub fn cmple(self, other: Self) -> Vec3Mask {
unsafe { Vec3Mask(_mm_cmple_ps(self.0, other.0)) }
}
#[inline]
pub fn cmplt(self, other: Self) -> Vec3Mask {
unsafe { Vec3Mask(_mm_cmplt_ps(self.0, other.0)) }
}
#[inline]
pub(crate) fn mul_add(self, a: Self, b: Self) -> Self {
unsafe { Self(_mm_add_ps(_mm_mul_ps(self.0, a.0), b.0)) }
}
#[inline]
pub(crate) fn neg_mul_sub(self, a: Self, b: Self) -> Self {
unsafe { Self(_mm_sub_ps(b.0, _mm_mul_ps(self.0, a.0))) }
}
#[inline]
pub fn abs(self) -> Self {
unsafe {
Self(_mm_and_ps(
self.0,
_mm_castsi128_ps(_mm_set1_epi32(0x7f_ff_ff_ff)),
))
}
}
}
impl fmt::Display for Vec3 {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
let (x, y, z) = (*self).into();
write!(f, "[{}, {}, {}]", x, y, z)
}
}
impl Div<Vec3> for Vec3 {
type Output = Self;
#[inline]
fn div(self, other: Self) -> Self {
unsafe { Self(_mm_div_ps(self.0, other.0)) }
}
}
impl DivAssign<Vec3> for Vec3 {
#[inline]
fn div_assign(&mut self, other: Self) {
unsafe {
self.0 = _mm_div_ps(self.0, other.0);
}
}
}
impl Div<f32> for Vec3 {
type Output = Self;
#[inline]
fn div(self, other: f32) -> Self {
unsafe { Self(_mm_div_ps(self.0, _mm_set1_ps(other))) }
}
}
impl DivAssign<f32> for Vec3 {
#[inline]
fn div_assign(&mut self, other: f32) {
unsafe { self.0 = _mm_div_ps(self.0, _mm_set1_ps(other)) }
}
}
impl Mul<Vec3> for Vec3 {
type Output = Self;
#[inline]
fn mul(self, other: Self) -> Self {
unsafe { Self(_mm_mul_ps(self.0, other.0)) }
}
}
impl MulAssign<Vec3> for Vec3 {
#[inline]
fn mul_assign(&mut self, other: Self) {
unsafe {
self.0 = _mm_mul_ps(self.0, other.0);
}
}
}
impl Mul<f32> for Vec3 {
type Output = Self;
#[inline]
fn mul(self, other: f32) -> Self {
unsafe { Self(_mm_mul_ps(self.0, _mm_set1_ps(other))) }
}
}
impl MulAssign<f32> for Vec3 {
#[inline]
fn mul_assign(&mut self, other: f32) {
unsafe { self.0 = _mm_mul_ps(self.0, _mm_set1_ps(other)) }
}
}
impl Mul<Vec3> for f32 {
type Output = Vec3;
#[inline]
fn mul(self, other: Vec3) -> Vec3 {
unsafe { Vec3(_mm_mul_ps(_mm_set1_ps(self), other.0)) }
}
}
impl Add for Vec3 {
type Output = Self;
#[inline]
fn add(self, other: Self) -> Self {
unsafe { Self(_mm_add_ps(self.0, other.0)) }
}
}
impl AddAssign for Vec3 {
#[inline]
fn add_assign(&mut self, other: Self) {
unsafe { self.0 = _mm_add_ps(self.0, other.0) }
}
}
impl Sub for Vec3 {
type Output = Self;
#[inline]
fn sub(self, other: Self) -> Self {
unsafe { Self(_mm_sub_ps(self.0, other.0)) }
}
}
impl SubAssign for Vec3 {
#[inline]
fn sub_assign(&mut self, other: Self) {
unsafe { self.0 = _mm_sub_ps(self.0, other.0) }
}
}
impl Neg for Vec3 {
type Output = Self;
#[inline]
fn neg(self) -> Self {
unsafe { Self(_mm_sub_ps(_mm_set1_ps(0.0), self.0)) }
}
}
impl Default for Vec3 {
#[inline]
fn default() -> Self {
Vec3::zero()
}
}
impl PartialEq for Vec3 {
#[inline]
fn eq(&self, other: &Self) -> bool {
self.cmpeq(*other).all()
}
}
impl PartialOrd for Vec3 {
#[inline]
fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
self.as_ref().partial_cmp(other.as_ref())
}
}
impl From<Vec3> for __m128 {
#[cfg_attr(tarpaulin, skip)]
#[inline]
fn from(t: Vec3) -> Self {
t.0
}
}
impl From<__m128> for Vec3 {
#[inline]
fn from(t: __m128) -> Self {
Self(t)
}
}
impl From<(f32, f32, f32)> for Vec3 {
#[inline]
fn from(t: (f32, f32, f32)) -> Self {
Self::new(t.0, t.1, t.2)
}
}
impl From<Vec3> for (f32, f32, f32) {
#[inline]
fn from(v: Vec3) -> Self {
let mut out: MaybeUninit<Align16<(f32, f32, f32)>> = MaybeUninit::uninit();
unsafe {
_mm_store_ps(out.as_mut_ptr() as *mut f32, v.0);
out.assume_init().0
}
}
}
impl From<[f32; 3]> for Vec3 {
#[inline]
fn from(a: [f32; 3]) -> Self {
Self::new(a[0], a[1], a[2])
}
}
impl From<Vec3> for [f32; 3] {
#[inline]
fn from(v: Vec3) -> Self {
let mut out: MaybeUninit<Align16<[f32; 3]>> = MaybeUninit::uninit();
unsafe {
_mm_store_ps(out.as_mut_ptr() as *mut f32, v.0);
out.assume_init().0
}
}
}
#[cfg(feature = "rand")]
impl Distribution<Vec3> for Standard {
#[inline]
fn sample<R: Rng + ?Sized>(&self, rng: &mut R) -> Vec3 {
rng.gen::<(f32, f32, f32)>().into()
}
}
#[derive(Clone, Copy)]
#[repr(C)]
pub struct Vec3Mask(__m128);
impl Vec3Mask {
#[inline]
pub fn new(x: bool, y: bool, z: bool) -> Self {
const MASK: [u32; 2] = [0, 0xff_ff_ff_ff];
unsafe {
Self(_mm_set_ps(
f32::from_bits(MASK[z as usize]),
f32::from_bits(MASK[z as usize]),
f32::from_bits(MASK[y as usize]),
f32::from_bits(MASK[x as usize]),
))
}
}
#[inline]
pub fn bitmask(&self) -> u32 {
unsafe { (_mm_movemask_ps(self.0) as u32) & 0x7 }
}
#[inline]
pub fn any(&self) -> bool {
unsafe { (_mm_movemask_ps(self.0) & 0x7) != 0 }
}
#[inline]
pub fn all(&self) -> bool {
unsafe { (_mm_movemask_ps(self.0) & 0x7) == 0x7 }
}
#[inline]
pub fn select(self, if_true: Vec3, if_false: Vec3) -> Vec3 {
unsafe {
Vec3(_mm_or_ps(
_mm_andnot_ps(self.0, if_false.0),
_mm_and_ps(if_true.0, self.0),
))
}
}
}
impl Default for Vec3Mask {
#[inline]
fn default() -> Self {
unsafe { Self(_mm_setzero_ps()) }
}
}
impl BitAnd for Vec3Mask {
type Output = Self;
#[inline]
fn bitand(self, other: Self) -> Self {
unsafe { Self(_mm_and_ps(self.0, other.0)) }
}
}
impl BitAndAssign for Vec3Mask {
fn bitand_assign(&mut self, other: Self) {
*self = *self & other
}
}
impl BitOr for Vec3Mask {
type Output = Self;
#[inline]
fn bitor(self, other: Self) -> Self {
unsafe { Self(_mm_or_ps(self.0, other.0)) }
}
}
impl BitOrAssign for Vec3Mask {
fn bitor_assign(&mut self, other: Self) {
*self = *self | other
}
}
impl Not for Vec3Mask {
type Output = Self;
#[inline]
fn not(self) -> Self {
unsafe {
Self(_mm_andnot_ps(
self.0,
_mm_set_ps1(f32::from_bits(0xff_ff_ff_ff)),
))
}
}
}