wide/

f32x16_.rs

use super::*;

pick! {
  if #[cfg(target_feature="avx512f")] {
    #[derive(Default, Clone, Copy, PartialEq)]
    #[repr(C, align(64))]
    pub struct f32x16 { pub(crate) avx512: m512 }
  } else {
    #[derive(Default, Clone, Copy, PartialEq)]
    #[repr(C, align(64))]
    pub struct f32x16 { pub(crate) a : f32x8, pub(crate) b : f32x8 }
  }
}

macro_rules! const_f32_as_f32x16 {
  ($i:ident, $f:expr) => {
    #[allow(non_upper_case_globals)]
    pub const $i: f32x16 = f32x16::new([$f; 16]);
  };
}

impl f32x16 {
  const_f32_as_f32x16!(ONE, 1.0);
  const_f32_as_f32x16!(HALF, 0.5);
  const_f32_as_f32x16!(ZERO, 0.0);
  const_f32_as_f32x16!(EPSILON, f32::EPSILON);
  const_f32_as_f32x16!(MIN, f32::MIN);
  const_f32_as_f32x16!(MIN_POSITIVE, f32::MIN_POSITIVE);
  const_f32_as_f32x16!(MAX, f32::MAX);
  const_f32_as_f32x16!(NAN, f32::NAN);
  const_f32_as_f32x16!(INFINITY, f32::INFINITY);
  const_f32_as_f32x16!(NEG_INFINITY, f32::NEG_INFINITY);
  const_f32_as_f32x16!(E, core::f32::consts::E);
  const_f32_as_f32x16!(FRAC_1_PI, core::f32::consts::FRAC_1_PI);
  const_f32_as_f32x16!(FRAC_2_PI, core::f32::consts::FRAC_2_PI);
  const_f32_as_f32x16!(FRAC_2_SQRT_PI, core::f32::consts::FRAC_2_SQRT_PI);
  const_f32_as_f32x16!(FRAC_1_SQRT_2, core::f32::consts::FRAC_1_SQRT_2);
  const_f32_as_f32x16!(FRAC_PI_2, core::f32::consts::FRAC_PI_2);
  const_f32_as_f32x16!(FRAC_PI_3, core::f32::consts::FRAC_PI_3);
  const_f32_as_f32x16!(FRAC_PI_4, core::f32::consts::FRAC_PI_4);
  const_f32_as_f32x16!(FRAC_PI_6, core::f32::consts::FRAC_PI_6);
  const_f32_as_f32x16!(FRAC_PI_8, core::f32::consts::FRAC_PI_8);
  const_f32_as_f32x16!(LN_2, core::f32::consts::LN_2);
  const_f32_as_f32x16!(LN_10, core::f32::consts::LN_10);
  const_f32_as_f32x16!(LOG2_E, core::f32::consts::LOG2_E);
  const_f32_as_f32x16!(LOG10_E, core::f32::consts::LOG10_E);
  const_f32_as_f32x16!(LOG10_2, core::f32::consts::LOG10_2);
  const_f32_as_f32x16!(LOG2_10, core::f32::consts::LOG2_10);
  const_f32_as_f32x16!(PI, core::f32::consts::PI);
  const_f32_as_f32x16!(SQRT_2, core::f32::consts::SQRT_2);
  const_f32_as_f32x16!(TAU, core::f32::consts::TAU);
}

unsafe impl Zeroable for f32x16 {}
unsafe impl Pod for f32x16 {}

impl AlignTo for f32x16 {
  type Elem = f32;
}

impl Add for f32x16 {
  type Output = Self;
  #[inline]
  fn add(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: add_m512(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.add(rhs.a),
          b : self.b.add(rhs.b),
        }
      }
    }
  }
}

impl Sub for f32x16 {
  type Output = Self;
  #[inline]
  fn sub(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: sub_m512(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.sub(rhs.a),
          b : self.b.sub(rhs.b),
        }
      }
    }
  }
}

impl Mul for f32x16 {
  type Output = Self;
  #[inline]
  fn mul(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: mul_m512(self.avx512, rhs.avx512) }
      } else {
        Self { a: self.a.mul(rhs.a), b: self.b.mul(rhs.b) }
      }
    }
  }
}

impl Div for f32x16 {
  type Output = Self;
  #[inline]
  fn div(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: div_m512(self.avx512, rhs.avx512) }
      } else {
        Self { a: self.a.div(rhs.a), b: self.b.div(rhs.b) }
      }
    }
  }
}

impl Rem for f32x16 {
  type Output = Self;
  #[inline]
  fn rem(self, rhs: Self) -> Self::Output {
    Self::new([
      self.to_array()[0] % rhs.to_array()[0],
      self.to_array()[1] % rhs.to_array()[1],
      self.to_array()[2] % rhs.to_array()[2],
      self.to_array()[3] % rhs.to_array()[3],
      self.to_array()[4] % rhs.to_array()[4],
      self.to_array()[5] % rhs.to_array()[5],
      self.to_array()[6] % rhs.to_array()[6],
      self.to_array()[7] % rhs.to_array()[7],
      self.to_array()[8] % rhs.to_array()[8],
      self.to_array()[9] % rhs.to_array()[9],
      self.to_array()[10] % rhs.to_array()[10],
      self.to_array()[11] % rhs.to_array()[11],
      self.to_array()[12] % rhs.to_array()[12],
      self.to_array()[13] % rhs.to_array()[13],
      self.to_array()[14] % rhs.to_array()[14],
      self.to_array()[15] % rhs.to_array()[15],
    ])
  }
}

impl Neg for f32x16 {
  type Output = Self;
  #[inline]
  fn neg(self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: bitxor_m512(self.avx512, Self::splat(-0.0).avx512) }
      } else {
        Self {
          a : self.a.neg(),
          b : self.b.neg(),
        }
      }
    }
  }
}

impl Add<f32> for f32x16 {
  type Output = Self;
  #[inline]
  fn add(self, rhs: f32) -> Self::Output {
    self.add(Self::splat(rhs))
  }
}

impl Sub<f32> for f32x16 {
  type Output = Self;
  #[inline]
  fn sub(self, rhs: f32) -> Self::Output {
    self.sub(Self::splat(rhs))
  }
}

impl Mul<f32> for f32x16 {
  type Output = Self;
  #[inline]
  fn mul(self, rhs: f32) -> Self::Output {
    self.mul(Self::splat(rhs))
  }
}

impl Div<f32> for f32x16 {
  type Output = Self;
  #[inline]
  fn div(self, rhs: f32) -> Self::Output {
    self.div(Self::splat(rhs))
  }
}

impl Rem<f32> for f32x16 {
  type Output = Self;
  #[inline]
  fn rem(self, rhs: f32) -> Self::Output {
    self.rem(Self::splat(rhs))
  }
}

impl Add<f32x16> for f32 {
  type Output = f32x16;
  #[inline]
  fn add(self, rhs: f32x16) -> Self::Output {
    f32x16::splat(self).add(rhs)
  }
}

impl Sub<f32x16> for f32 {
  type Output = f32x16;
  #[inline]
  fn sub(self, rhs: f32x16) -> Self::Output {
    f32x16::splat(self).sub(rhs)
  }
}

impl Mul<f32x16> for f32 {
  type Output = f32x16;
  #[inline]
  fn mul(self, rhs: f32x16) -> Self::Output {
    f32x16::splat(self).mul(rhs)
  }
}

impl Div<f32x16> for f32 {
  type Output = f32x16;
  #[inline]
  fn div(self, rhs: f32x16) -> Self::Output {
    f32x16::splat(self).div(rhs)
  }
}

impl Rem<f32x16> for f32 {
  type Output = f32x16;
  #[inline]
  fn rem(self, rhs: f32x16) -> Self::Output {
    f32x16::splat(self).rem(rhs)
  }
}

impl BitAnd for f32x16 {
  type Output = Self;
  #[inline]
  fn bitand(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: bitand_m512(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.bitand(rhs.a),
          b : self.b.bitand(rhs.b),
        }
      }
    }
  }
}

impl BitOr for f32x16 {
  type Output = Self;
  #[inline]
  fn bitor(self, rhs: Self) -> Self::Output {
    pick! {
    if #[cfg(target_feature="avx512f")] {
        Self { avx512: bitor_m512(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.bitor(rhs.a),
          b : self.b.bitor(rhs.b),
        }
      }
    }
  }
}

impl BitXor for f32x16 {
  type Output = Self;
  #[inline]
  fn bitxor(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: bitxor_m512(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.bitxor(rhs.a),
          b : self.b.bitxor(rhs.b),
        }
      }
    }
  }
}

#[expect(deprecated)]
impl CmpEq for f32x16 {
  type Output = Self;
  #[inline]
  fn simd_eq(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: cmp_op_mask_m512::<{cmp_op!(EqualOrdered)}>(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.simd_eq(rhs.a),
          b : self.b.simd_eq(rhs.b),
        }
      }
    }
  }
}

#[expect(deprecated)]
impl CmpGt for f32x16 {
  type Output = Self;
  #[inline]
  fn simd_gt(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: cmp_op_mask_m512::<{cmp_op!(GreaterThanOrdered)}>(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.simd_gt(rhs.a),
          b : self.b.simd_gt(rhs.b),
        }
      }
    }
  }
}

#[expect(deprecated)]
impl CmpGe for f32x16 {
  type Output = Self;
  #[inline]
  fn simd_ge(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: cmp_op_mask_m512::<{cmp_op!(GreaterEqualOrdered)}>(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.simd_ge(rhs.a),
          b : self.b.simd_ge(rhs.b),
        }
      }
    }
  }
}

#[expect(deprecated)]
impl CmpLt for f32x16 {
  type Output = Self;
  #[inline]
  fn simd_lt(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: cmp_op_mask_m512::<{cmp_op!(LessThanOrdered)}>(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.simd_lt(rhs.a),
          b : self.b.simd_lt(rhs.b),
        }
      }
    }
  }
}

#[expect(deprecated)]
impl CmpLe for f32x16 {
  type Output = Self;
  #[inline]
  fn simd_le(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: cmp_op_mask_m512::<{cmp_op!(LessEqualOrdered)}>(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.simd_le(rhs.a),
          b : self.b.simd_le(rhs.b),
        }
      }
    }
  }
}

#[expect(deprecated)]
impl CmpNe for f32x16 {
  type Output = Self;
  #[inline]
  fn simd_ne(self, rhs: Self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: cmp_op_mask_m512::<{cmp_op!(NotEqualUnordered)}>(self.avx512, rhs.avx512) }
      } else {
        Self {
          a : self.a.simd_ne(rhs.a),
          b : self.b.simd_ne(rhs.b),
        }
      }
    }
  }
}

impl f32x16 {
  #[inline]
  #[must_use]
  pub const fn new(array: [f32; 16]) -> Self {
    unsafe { core::mem::transmute(array) }
  }

  simd_comparison_fns!();

  #[inline]
  #[must_use]
  pub fn blend(self, t: Self, f: Self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: blend_varying_m512(f.avx512, t.avx512, movepi32_mask_m512(self.avx512)) }
      } else {
        Self {
          a : self.a.blend(t.a, f.a),
          b : self.b.blend(t.b, f.b),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn abs(self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        let non_sign_bits = f32x16::from(f32::from_bits(i32::MAX as u32));
        self & non_sign_bits
      } else {
        Self {
          a : self.a.abs(),
          b : self.b.abs(),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn signum(self) -> Self {
    let result = Self::ONE | self & -Self::ZERO;

    self.is_nan().blend(self, result)
  }

  #[inline]
  #[must_use]
  pub fn floor(self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: round_m512::<{round_op!(NegInf)}>(self.avx512) }
      } else {
        Self {
          a : self.a.floor(),
          b : self.b.floor(),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn ceil(self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: round_m512::<{round_op!(PosInf)}>(self.avx512) }
      } else {
        Self {
          a : self.a.ceil(),
          b : self.b.ceil(),
        }
      }
    }
  }

  /// Calculates the lanewise maximum of both vectors. This is a faster
  /// implementation than `max`, but it doesn't specify any behavior if NaNs are
  /// involved.
  #[inline]
  #[must_use]
  pub fn fast_max(self, rhs: Self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: max_m512(self.avx512, rhs.avx512) }
      } else {
        Self {
          a: self.a.fast_max(rhs.a),
          b: self.b.fast_max(rhs.b),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn max(self, rhs: Self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        // max_m512 seems to do rhs < self ? self : rhs. So if there's any NaN
        // involved, it chooses rhs, so we need to specifically check rhs for
        // NaN.
        rhs.is_nan().blend(self, Self { avx512: max_m512(self.avx512, rhs.avx512) })
      } else {
        Self {
          a: self.a.max(rhs.a),
          b: self.b.max(rhs.b),
        }
      }
    }
  }

  /// Calculates the lanewise minimum of both vectors. This is a faster
  /// implementation than `min`, but it doesn't specify any behavior if NaNs are
  /// involved.
  #[inline]
  #[must_use]
  pub fn fast_min(self, rhs: Self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: min_m512(self.avx512, rhs.avx512) }
      } else {
        Self {
          a: self.a.fast_min(rhs.a),
          b: self.b.fast_min(rhs.b),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn min(self, rhs: Self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        // min_m512 seems to do rhs > self ? self : rhs. So if there's any NaN
        // involved, it chooses rhs, so we need to specifically check rhs for
        // NaN.
        rhs.is_nan().blend(self, Self { avx512: min_m512(self.avx512, rhs.avx512) })
      } else {
        Self {
          a: self.a.min(rhs.a),
          b: self.b.min(rhs.b),
        }
      }
    }
  }

  /// Restrict a value to a certain interval unless it is NaN.
  ///
  /// If `self` is NaN, or `min` is NaN, or `max` is NaN, the result is NaN.
  /// If `min > max`, the result is `min`, since `fast_max(min)` dominates.
  #[inline]
  #[must_use]
  pub fn clamp(self, min: Self, max: Self) -> Self {
    let is_nan = self.is_nan() | min.is_nan() | max.is_nan();
    let clamped = self.fast_min(max).fast_max(min);
    is_nan.blend(Self::splat(f32::NAN), clamped)
  }

  /// Restrict a value to a certain interval unless it is NaN.
  ///
  /// Avoids NaN detection; same speed as the old `clamp` prior to IEEE 754-2019
  /// compliance. Does not specify any
  /// behavior if NaNs are involved, and if `min > max` the result is
  /// unspecified.
  #[inline]
  #[must_use]
  pub fn fast_clamp(self, min: Self, max: Self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        // For both `min_m512` and `max_m512` if any input is NaN, `rhs` gets
        // chosen. For `self` to be chosen, `self` must be the second argument.
        Self { avx512: min_m512(max.avx512, max_m512(min.avx512, self.avx512)) }
      } else {
        Self {
          a: self.a.fast_clamp(min.a, max.a),
          b: self.b.fast_clamp(min.b, max.b),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn midpoint(self, other: Self) -> Self {
    (self + other) * 0.5
  }

  #[inline]
  #[must_use]
  pub fn is_nan(self) -> Self {
    pick! {
      if #[cfg(target_feature = "avx512f")] {
        Self { avx512: cmp_op_mask_m512::<{cmp_op!(Unordered)}>(self.avx512, self.avx512) }
      } else {
        Self {
          a: self.a.is_nan(),
          b: self.b.is_nan(),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn is_finite(self) -> Self {
    let shifted_exp_mask = u32x16::splat(0xFF000000);
    let u: u32x16 = cast(self);
    let shift_u = u << 1_u32;
    let out = !(shift_u & shifted_exp_mask).simd_eq(shifted_exp_mask);
    cast(out)
  }

  #[inline]
  #[must_use]
  pub fn is_inf(self) -> Self {
    let shifted_inf = u32x16::from(0xFF000000);
    let u: u32x16 = cast(self);
    let shift_u = u << 1_u64;
    let out = (shift_u).simd_eq(shifted_inf);
    cast(out)
  }

  #[inline]
  #[must_use]
  pub fn round(self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: round_m512::<{round_op!(Nearest)}>(self.avx512) }
      } else {
        Self {
          a: self.a.round(),
          b: self.b.round(),
        }
      }
    }
  }

  /// Rounds each lane into an integer. This is a faster implementation than
  /// `round_int`, but it doesn't handle out of range values or NaNs. For those
  /// values you get implementation defined behavior.
  #[inline]
  #[must_use]
  pub fn fast_round_int(self) -> i32x16 {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        cast(convert_to_i32_m512i_from_m512(self.avx512))
      } else {
        i32x16 {
          a: self.a.fast_round_int(),
          b: self.b.fast_round_int(),
        }
      }
    }
  }

  /// Rounds each lane into an integer. This saturates out of range values and
  /// turns NaNs into 0. Use `fast_round_int` for a faster implementation that
  /// doesn't handle out of range values or NaNs.
  #[inline]
  #[must_use]
  pub fn round_int(self) -> i32x16 {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        // Based on: https://github.com/v8/v8/blob/210987a552a2bf2a854b0baa9588a5959ff3979d/src/codegen/shared-ia32-x64/macro-assembler-shared-ia32-x64.h#L489-L504
        let non_nan_mask = self.simd_eq(self);
        let non_nan = self & non_nan_mask;
        let flip_to_max: i32x16 = cast(self.simd_ge(Self::splat(2147483648.0)));
        let cast: i32x16 = cast(convert_to_i32_m512i_from_m512(non_nan.avx512));
        flip_to_max ^ cast
      } else {
        i32x16 {
          a: self.a.round_int(),
          b: self.b.round_int(),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn trunc(self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: round_m512::<{round_op!(Zero)}>(self.avx512) }
      } else {
        Self {
          a: self.a.trunc(),
          b: self.b.trunc(),
        }
      }
    }
  }

  /// Truncates each lane into an integer. This is a faster implementation than
  /// `trunc_int`, but it doesn't handle out of range values or NaNs. For those
  /// values you get implementation defined behavior.
  #[inline]
  #[must_use]
  pub fn fast_trunc_int(self) -> i32x16 {
    pick! {
      if #[cfg(all(target_feature="avx512f"))] {
        cast(convert_truncate_m512_i32_m512i(self.avx512))
      } else {
        cast([
          self.a.fast_trunc_int(),
          self.b.fast_trunc_int(),
        ])
      }
    }
  }

  /// Truncates each lane into an integer. This saturates out of range values
  /// and turns NaNs into 0. Use `fast_trunc_int` for a faster implementation
  /// that doesn't handle out of range values or NaNs.
  #[inline]
  #[must_use]
  pub fn trunc_int(self) -> i32x16 {
    pick! {
        if #[cfg(target_feature="avx512f")] {
        // Based on: https://github.com/v8/v8/blob/210987a552a2bf2a854b0baa9588a5959ff3979d/src/codegen/shared-ia32-x64/macro-assembler-shared-ia32-x64.h#L489-L504
        let non_nan_mask = self.simd_eq(self);
        let non_nan = self & non_nan_mask;
        let flip_to_max: i32x16 = cast(self.simd_ge(Self::splat(2147483648.0)));
        let cast: i32x16 = cast(convert_truncate_m512_i32_m512i(non_nan.avx512));
        flip_to_max ^ cast
      } else {
        cast([
          self.a.trunc_int(),
          self.b.trunc_int(),
        ])
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn fract(self) -> Self {
    self - self.trunc()
  }

  /// Performs a multiply-add operation: `self * m + a`
  ///
  /// When hardware FMA support is available, this computes the result with a
  /// single rounding operation. Without FMA support, it falls back to separate
  /// multiply and add operations with two roundings.
  ///
  /// # Platform-specific behavior
  /// - On `x86`/`x86_64` with AVX-512F+FMA: Uses 512-bit `vfmadd` (single
  ///   rounding, best accuracy)
  /// - On `x86`/`x86_64` with AVX-512F only: Uses `(self * m) + a` (two
  ///   roundings)
  /// - Other platforms: Delegates to [`f32x8`] (inherits its FMA behavior)
  ///
  /// # Examples
  /// ```
  /// # use wide::f32x16;
  /// let a = f32x16::from([1.0; 16]);
  /// let b = f32x16::from([2.0; 16]);
  /// let c = f32x16::from([10.0; 16]);
  ///
  /// let result = a.mul_add(b, c);
  ///
  /// let expected = f32x16::from([12.0; 16]);
  /// assert_eq!(result, expected);
  /// ```
  #[inline]
  #[must_use]
  pub fn mul_add(self, m: Self, a: Self) -> Self {
    pick! {
      if #[cfg(all(target_feature="avx512f",target_feature="fma"))] {
        Self { avx512: fused_mul_add_m512(self.avx512, m.avx512, a.avx512) }
      } else if #[cfg(target_feature="avx512f")] {
        // still want to use 512 bit ops
        (self * m) + a
      } else {
        Self {
          a: self.a.mul_add(m.a, a.a),
          b: self.b.mul_add(m.b, a.b),
        }
      }
    }
  }

  /// Performs a multiply-subtract operation: `self * m - s`
  ///
  /// When hardware FMA support is available, this computes the result with a
  /// single rounding operation. Without FMA support, it falls back to separate
  /// multiply and subtract operations with two roundings.
  ///
  /// # Platform-specific behavior
  /// - On `x86`/`x86_64` with AVX-512F+FMA: Uses 512-bit `vfmsub` (single
  ///   rounding, best accuracy)
  /// - On `x86`/`x86_64` with AVX-512F only: Uses `(self * m) - s` (two
  ///   roundings)
  /// - Other platforms: Delegates to [`f32x8`] (inherits its FMA behavior)
  ///
  /// # Examples
  /// ```
  /// # use wide::f32x16;
  /// let a = f32x16::from([10.0; 16]);
  /// let b = f32x16::from([3.0; 16]);
  /// let c = f32x16::from([5.0; 16]);
  ///
  /// let result = a.mul_sub(b, c);
  ///
  /// let expected = f32x16::from([25.0; 16]);
  /// assert_eq!(result, expected);
  /// ```
  #[inline]
  #[must_use]
  pub fn mul_sub(self, m: Self, s: Self) -> Self {
    pick! {
      if #[cfg(all(target_feature="avx512f",target_feature="fma"))] {
        Self { avx512: fused_mul_sub_m512(self.avx512, m.avx512, s.avx512) }
      } else if #[cfg(target_feature="avx512f")] {
        // still want to use 512 bit ops
        (self * m) - s
      } else {
        Self {
          a: self.a.mul_sub(m.a, s.a),
          b: self.b.mul_sub(m.b, s.b),
        }
      }
    }
  }

  /// Performs a negative multiply-add operation: `a - (self * m)`
  ///
  /// When hardware FMA support is available, this computes the result with a
  /// single rounding operation. Without FMA support, it falls back to separate
  /// operations with two roundings.
  ///
  /// # Platform-specific behavior
  /// - On `x86`/`x86_64` with AVX-512F+FMA: Uses 512-bit `vfnmadd` (single
  ///   rounding, best accuracy)
  /// - On `x86`/`x86_64` with AVX-512F only: Uses `a - (self * m)` (two
  ///   roundings)
  /// - Other platforms: Delegates to [`f32x8`] (inherits its FMA behavior)
  ///
  /// # Examples
  /// ```
  /// # use wide::f32x16;
  /// let a = f32x16::from([4.0; 16]);
  /// let b = f32x16::from([2.0; 16]);
  /// let c = f32x16::from([10.0; 16]);
  ///
  /// let result = a.mul_neg_add(b, c);
  ///
  /// let expected = f32x16::from([2.0; 16]);
  /// assert_eq!(result, expected);
  /// ```
  #[inline]
  #[must_use]
  pub fn mul_neg_add(self, m: Self, a: Self) -> Self {
    pick! {
      if #[cfg(all(target_feature="avx512f",target_feature="fma"))] {
        Self { avx512: fused_mul_neg_add_m512(self.avx512, m.avx512, a.avx512) }
      } else if #[cfg(target_feature="avx512f")] {
        // still want to use 512 bit ops
        a - (self * m)
      } else {
        Self {
          a: self.a.mul_neg_add(m.a, a.a),
          b: self.b.mul_neg_add(m.b, a.b),
        }
      }
    }
  }

  /// Performs a negative multiply-subtract operation: `-(self * m) - s`
  ///
  /// When hardware FMA support is available, this computes the result with a
  /// single rounding operation. Without FMA support, it falls back to separate
  /// operations with two roundings.
  ///
  /// # Platform-specific behavior
  /// - On `x86`/`x86_64` with AVX-512F+FMA: Uses 512-bit `vfnmsub` (single
  ///   rounding, best accuracy)
  /// - On `x86`/`x86_64` with AVX-512F only: Uses `-(self * m) - s` (two
  ///   roundings)
  /// - Other platforms: Delegates to [`f32x8`] (inherits its FMA behavior)
  ///
  /// # Examples
  /// ```
  /// # use wide::f32x16;
  /// let a = f32x16::from([4.0; 16]);
  /// let b = f32x16::from([2.0; 16]);
  /// let c = f32x16::from([1.0; 16]);
  ///
  /// let result = a.mul_neg_sub(b, c);
  ///
  /// let expected = f32x16::from([-9.0; 16]);
  /// assert_eq!(result, expected);
  /// ```
  #[inline]
  #[must_use]
  pub fn mul_neg_sub(self, m: Self, s: Self) -> Self {
    pick! {
      if #[cfg(all(target_feature="avx512f",target_feature="fma"))] {
        Self { avx512: fused_mul_neg_sub_m512(self.avx512, m.avx512, s.avx512) }
      } else if #[cfg(target_feature="avx512f")] {
        // still want to use 512 bit ops
        -(self * m) - s
      } else {
        Self {
          a: self.a.mul_neg_sub(m.a, s.a),
          b: self.b.mul_neg_sub(m.b, s.b),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn div_euclid(self, rhs: Self) -> Self {
    let q = (self / rhs).trunc();
    (self % rhs)
      .simd_lt(Self::ZERO)
      .blend(rhs.simd_gt(Self::ZERO).blend(q - Self::ONE, q + Self::ONE), q)
  }

  #[inline]
  #[must_use]
  pub fn rem_euclid(self, rhs: Self) -> Self {
    let r = self % rhs;
    r.simd_lt(Self::ZERO).blend(r + rhs.abs(), r)
  }

  #[inline]
  #[must_use]
  pub fn flip_signs(self, signs: Self) -> Self {
    self ^ (signs & Self::from(-0.0))
  }

  #[inline]
  #[must_use]
  pub fn copysign(self, sign: Self) -> Self {
    let magnitude_mask = Self::from(f32::from_bits(u32::MAX >> 1));
    (self & magnitude_mask) | (sign & Self::from(-0.0))
  }

  #[inline]
  pub fn asin_acos(self) -> (Self, Self) {
    // Based on the Agner Fog "vector class library":
    // https://github.com/vectorclass/version2/blob/master/vectormath_trig.h
    const_f32_as_f32x16!(P4asinf, 4.2163199048E-2);
    const_f32_as_f32x16!(P3asinf, 2.4181311049E-2);
    const_f32_as_f32x16!(P2asinf, 4.5470025998E-2);
    const_f32_as_f32x16!(P1asinf, 7.4953002686E-2);
    const_f32_as_f32x16!(P0asinf, 1.6666752422E-1);

    let xa = self.abs();
    let big = xa.simd_ge(f32x16::splat(0.5));

    let x1 = f32x16::splat(0.5) * (f32x16::ONE - xa);
    let x2 = xa * xa;
    let x3 = big.blend(x1, x2);

    let xb = x1.sqrt();

    let x4 = big.blend(xb, xa);

    let z = polynomial_4!(x3, P0asinf, P1asinf, P2asinf, P3asinf, P4asinf);
    let z = z.mul_add(x3 * x4, x4);

    let z1 = z + z;

    // acos
    let z3 = self.simd_lt(f32x16::ZERO).blend(f32x16::PI - z1, z1);
    let z4 = f32x16::FRAC_PI_2 - z.flip_signs(self);
    let acos = big.blend(z3, z4);

    // asin
    let z3 = f32x16::FRAC_PI_2 - z1;
    let asin = big.blend(z3, z);
    let asin = asin.flip_signs(self);

    (asin, acos)
  }

  #[inline]
  #[must_use]
  pub fn asin(self) -> Self {
    // Based on the Agner Fog "vector class library":
    // https://github.com/vectorclass/version2/blob/master/vectormath_trig.h
    const_f32_as_f32x16!(P4asinf, 4.2163199048E-2);
    const_f32_as_f32x16!(P3asinf, 2.4181311049E-2);
    const_f32_as_f32x16!(P2asinf, 4.5470025998E-2);
    const_f32_as_f32x16!(P1asinf, 7.4953002686E-2);
    const_f32_as_f32x16!(P0asinf, 1.6666752422E-1);

    let xa = self.abs();
    let big = xa.simd_ge(f32x16::splat(0.5));

    let x1 = f32x16::splat(0.5) * (f32x16::ONE - xa);
    let x2 = xa * xa;
    let x3 = big.blend(x1, x2);

    let xb = x1.sqrt();

    let x4 = big.blend(xb, xa);

    let z = polynomial_4!(x3, P0asinf, P1asinf, P2asinf, P3asinf, P4asinf);
    let z = z.mul_add(x3 * x4, x4);

    let z1 = z + z;

    // asin
    let z3 = f32x16::FRAC_PI_2 - z1;
    let asin = big.blend(z3, z);
    let asin = asin.flip_signs(self);

    asin
  }

  #[inline]
  #[must_use]
  pub fn acos(self) -> Self {
    // Based on the Agner Fog "vector class library":
    // https://github.com/vectorclass/version2/blob/master/vectormath_trig.h
    const_f32_as_f32x16!(P4asinf, 4.2163199048E-2);
    const_f32_as_f32x16!(P3asinf, 2.4181311049E-2);
    const_f32_as_f32x16!(P2asinf, 4.5470025998E-2);
    const_f32_as_f32x16!(P1asinf, 7.4953002686E-2);
    const_f32_as_f32x16!(P0asinf, 1.6666752422E-1);

    let xa = self.abs();
    let big = xa.simd_ge(f32x16::splat(0.5));

    let x1 = f32x16::splat(0.5) * (f32x16::ONE - xa);
    let x2 = xa * xa;
    let x3 = big.blend(x1, x2);

    let xb = x1.sqrt();

    let x4 = big.blend(xb, xa);

    let z = polynomial_4!(x3, P0asinf, P1asinf, P2asinf, P3asinf, P4asinf);
    let z = z.mul_add(x3 * x4, x4);

    let z1 = z + z;

    // acos
    let z3 = self.simd_lt(f32x16::ZERO).blend(f32x16::PI - z1, z1);
    let z4 = f32x16::FRAC_PI_2 - z.flip_signs(self);
    let acos = big.blend(z3, z4);

    acos
  }

  #[inline]
  pub fn atan(self) -> Self {
    // Based on the Agner Fog "vector class library":
    // https://github.com/vectorclass/version2/blob/master/vectormath_trig.h
    const_f32_as_f32x16!(P3atanf, 8.05374449538E-2);
    const_f32_as_f32x16!(P2atanf, -1.38776856032E-1);
    const_f32_as_f32x16!(P1atanf, 1.99777106478E-1);
    const_f32_as_f32x16!(P0atanf, -3.33329491539E-1);

    let t = self.abs();

    // small:  z = t / 1.0;
    // medium: z = (t-1.0) / (t+1.0);
    // big:    z = -1.0 / t;
    let notsmal = t.simd_ge(Self::SQRT_2 - Self::ONE);
    let notbig = t.simd_le(Self::SQRT_2 + Self::ONE);

    let mut s = notbig.blend(Self::FRAC_PI_4, Self::FRAC_PI_2);
    s = notsmal & s;

    let mut a = notbig & t;
    a = notsmal.blend(a - Self::ONE, a);
    let mut b = notbig & Self::ONE;
    b = notsmal.blend(b + t, b);
    let z = a / b;

    let zz = z * z;

    // Taylor expansion
    let mut re = polynomial_3!(zz, P0atanf, P1atanf, P2atanf, P3atanf);
    re = re.mul_add(zz * z, z) + s;

    // get sign bit
    re = (self.is_sign_negative()).blend(-re, re);

    re
  }

  #[inline]
  pub fn atan2(self, x: Self) -> Self {
    // Based on the Agner Fog "vector class library":
    // https://github.com/vectorclass/version2/blob/master/vectormath_trig.h
    const_f32_as_f32x16!(P3atanf, 8.05374449538E-2);
    const_f32_as_f32x16!(P2atanf, -1.38776856032E-1);
    const_f32_as_f32x16!(P1atanf, 1.99777106478E-1);
    const_f32_as_f32x16!(P0atanf, -3.33329491539E-1);

    let y = self;

    // move in first octant
    let x1 = x.abs();
    let y1 = y.abs();
    let swapxy = y1.simd_gt(x1);
    // swap x and y if y1 > x1
    let mut x2 = swapxy.blend(y1, x1);
    let mut y2 = swapxy.blend(x1, y1);

    // check for special case: x and y are both +/- INF
    let both_infinite = x.is_inf() & y.is_inf();
    if both_infinite.any() {
      let minus_one = -Self::ONE;
      x2 = both_infinite.blend(x2 & minus_one, x2);
      y2 = both_infinite.blend(y2 & minus_one, y2);
    }

    // x = y = 0 will produce NAN. No problem, fixed below
    let t = y2 / x2;

    // small:  z = t / 1.0;
    // medium: z = (t-1.0) / (t+1.0);
    let notsmal = t.simd_ge(Self::SQRT_2 - Self::ONE);

    let a = notsmal.blend(t - Self::ONE, t);
    let b = notsmal.blend(t + Self::ONE, Self::ONE);
    let s = notsmal & Self::FRAC_PI_4;
    let z = a / b;

    let zz = z * z;

    // Taylor expansion
    let mut re = polynomial_3!(zz, P0atanf, P1atanf, P2atanf, P3atanf);
    re = re.mul_add(zz * z, z) + s;

    // move back in place
    re = swapxy.blend(Self::FRAC_PI_2 - re, re);
    re = ((x | y).simd_eq(Self::ZERO)).blend(Self::ZERO, re);
    re = (x.is_sign_negative()).blend(Self::PI - re, re);

    // get sign bit
    re = (y.is_sign_negative()).blend(-re, re);

    re
  }

  #[inline]
  #[must_use]
  pub fn sin_cos(self) -> (Self, Self) {
    // Based on the Agner Fog "vector class library":
    // https://github.com/vectorclass/version2/blob/master/vectormath_trig.h

    const_f32_as_f32x16!(DP1F, 0.78515625_f32 * 2.0);
    const_f32_as_f32x16!(DP2F, 2.4187564849853515625E-4_f32 * 2.0);
    const_f32_as_f32x16!(DP3F, 3.77489497744594108E-8_f32 * 2.0);

    const_f32_as_f32x16!(P0sinf, -1.6666654611E-1);
    const_f32_as_f32x16!(P1sinf, 8.3321608736E-3);
    const_f32_as_f32x16!(P2sinf, -1.9515295891E-4);

    const_f32_as_f32x16!(P0cosf, 4.166664568298827E-2);
    const_f32_as_f32x16!(P1cosf, -1.388731625493765E-3);
    const_f32_as_f32x16!(P2cosf, 2.443315711809948E-5);

    const_f32_as_f32x16!(TWO_OVER_PI, 2.0 / core::f32::consts::PI);

    let xa = self.abs();

    // Find quadrant
    let y = (xa * TWO_OVER_PI).round();
    let q: i32x16 = y.round_int();

    let x = y.mul_neg_add(DP3F, y.mul_neg_add(DP2F, y.mul_neg_add(DP1F, xa)));

    let x2 = x * x;
    let mut s = polynomial_2!(x2, P0sinf, P1sinf, P2sinf) * (x * x2) + x;
    let mut c = polynomial_2!(x2, P0cosf, P1cosf, P2cosf) * (x2 * x2)
      + f32x16::from(0.5).mul_neg_add(x2, f32x16::from(1.0));

    let swap = !(q & i32x16::from(1)).simd_eq(i32x16::from(0));

    let mut overflow: f32x16 = cast(q.simd_gt(i32x16::from(0x2000000)));
    overflow &= xa.is_finite();
    s = overflow.blend(f32x16::from(0.0), s);
    c = overflow.blend(f32x16::from(1.0), c);

    // calc sin
    let mut sin1 = cast::<_, f32x16>(swap).blend(c, s);
    let sign_sin: i32x16 = (q << 30) ^ cast::<_, i32x16>(self);
    sin1 = sin1.flip_signs(cast(sign_sin));

    // calc cos
    let mut cos1 = cast::<_, f32x16>(swap).blend(s, c);
    let sign_cos: i32x16 = ((q + i32x16::from(1)) & i32x16::from(2)) << 30;
    cos1 ^= cast::<_, f32x16>(sign_cos);

    // IEEE 754: sin/cos(±∞) = NaN, sin/cos(NaN) = NaN
    let finite = self.is_finite();
    let nan = Self::splat(f32::NAN);
    let sin_final = finite.blend(sin1, nan);
    let cos_final = finite.blend(cos1, nan);

    (sin_final, cos_final)
  }

  #[inline]
  #[must_use]
  pub fn sin(self) -> Self {
    let (s, _) = self.sin_cos();
    s
  }

  #[inline]
  #[must_use]
  pub fn cos(self) -> Self {
    let (_, c) = self.sin_cos();
    c
  }

  #[inline]
  #[must_use]
  pub fn tan(self) -> Self {
    let (s, c) = self.sin_cos();
    s / c
  }

  /// Calculates hyperbolic sine: `(e^self - e^(-self))/2`.
  #[inline]
  #[must_use]
  pub fn sinh(self) -> Self {
    const_f32_as_f32x16!(P0, 1.0);
    const_f32_as_f32x16!(P1, 1.0 / 6.0);
    const_f32_as_f32x16!(P2, 1.0 / 120.0);
    const_f32_as_f32x16!(P3, 1.0 / 5040.0);
    let a = self.abs();
    // |x| < 0.5: Taylor poly; last truncation term < 1 ULP at x=0.5 for both types
    let small = a.simd_lt(f32x16::from(0.5));
    let t = a * a;
    let poly = a * polynomial_3!(t, P0, P1, P2, P3);
    let exp_based = {
      let e = a.exp();
      (e - Self::ONE / e) * Self::HALF
    };
    let result = small.blend(poly, exp_based);
    result.flip_signs(self)
  }

  /// Calculates hyperbolic cosine: `(e^self + e^(-self))/2`.
  #[inline]
  #[must_use]
  pub fn cosh(self) -> Self {
    const_f32_as_f32x16!(P0, 1.0);
    const_f32_as_f32x16!(P1, 1.0 / 2.0);
    const_f32_as_f32x16!(P2, 1.0 / 24.0);
    const_f32_as_f32x16!(P3, 1.0 / 720.0);
    let a = self.abs();
    // |x| < 0.5: Taylor poly; last truncation term < 1 ULP at x=0.5 for both types
    let small = a.simd_lt(f32x16::from(0.5));
    let t = a * a;
    let poly = polynomial_3!(t, P0, P1, P2, P3);
    let exp_based = {
      let e = a.exp();
      (e + Self::ONE / e) * Self::HALF
    };
    small.blend(poly, exp_based)
  }

  /// Calculates hyperbolic tangent: `sinh(self)/cosh(self)`.
  #[inline]
  #[must_use]
  pub fn tanh(self) -> Self {
    // |x| < 2e-4: tanh(x) ≈ x, error x³/3 < 16·ULP(x)
    // bound: x² < 48·2⁻²³ → x < 2.39e-3; 2e-4 has 10× margin
    // |x| > 9.011: tanh(x) = ±1 to f32 precision (e⁻²ˣ < 2⁻²⁴)
    let a = self.abs();
    let large = a.simd_gt(f32x16::from(9.011));
    if large.all() {
      return Self::ONE.flip_signs(self);
    }
    let small = a.simd_lt(f32x16::from(2e-4));
    let exp_based = {
      let t = (Self::from(-2.0) * a).exp_m1();
      let pos = -t / (t + Self::from(2.0));
      pos.flip_signs(self)
    };
    let result = small.blend(self, exp_based);
    large.blend(Self::ONE.flip_signs(self), result)
  }

  /// Calculates the cube root: `self^(1/3)`.
  #[inline]
  #[must_use]
  pub fn cbrt(self) -> Self {
    let a = self.abs();
    let zero = a.simd_eq(Self::ZERO);
    if zero.all() {
      return self; // preserves -0.0
    }
    let inf = a.is_inf();
    let nan = self.is_nan();

    let tiny = a.simd_lt(Self::from(f32::MIN_POSITIVE));
    let a_work = tiny.blend(a * Self::from(16777216.0), a);

    let e = Self::exponent(a_work) + Self::ONE;
    let d = Self::fraction_2(a_work);

    // C0..C5 from SLEEF's minimax polynomial for 1/cbrt(d) on [0.5, 1.0)
    // Naoki Shibata et al., "SLEEF: A Portable Vectorized Library of C99
    // Mathematical Functions", https://sleef.org / https://github.com/shibatch/sleef
    // Licensed under the Boost Software License 1.0.
    // These are the f32-precision coefficients; our f64 variants use the f64
    // set.
    const_f32_as_f32x16!(C0, 2.2241257);
    const_f32_as_f32x16!(C1, -3.8095417);
    const_f32_as_f32x16!(C2, 5.8982625);
    const_f32_as_f32x16!(C3, -5.532182);
    const_f32_as_f32x16!(C4, 2.8208892);
    const_f32_as_f32x16!(C5, -0.60156447);
    let mut x = polynomial_5!(d, C0, C1, C2, C3, C4, C5);

    let x2 = x * x;
    let x4 = x2 * x2;
    x = x - d.mul_add(x4, -x) * Self::from(1.0 / 3.0);
    // cbrt(d) = d * x² with refinement
    let mut y = (d * x) * x;
    let yx = y * x;
    let t = Self::from(2.0 / 3.0);
    y = y - t * y * (yx - Self::ONE);

    // Scale by 2^(e/3)
    let three = Self::from(3.0);
    let two = Self::from(2.0);
    let neg = e.simd_lt(Self::ZERO);
    let e_adj = neg.blend(e - two, e);
    let k = (e_adj / three).trunc();
    let r = e - three * k;
    const_f32_as_f32x16!(CBRT2, 1.259921);
    const_f32_as_f32x16!(CBRT4, 1.587401);
    y = r.simd_eq(Self::ONE).blend(y * CBRT2, y);
    y = r.simd_eq(two).blend(y * CBRT4, y);
    y *= Self::vm_pow2n(k);
    y = tiny.blend(y / Self::from(256.0_f32), y);

    let result = y.flip_signs(self);
    let result = nan.blend(self, result);
    let result = zero.blend(self, result);
    let result = inf.blend(self, result);
    result
  }

  #[inline]
  #[must_use]
  pub fn to_degrees(self) -> Self {
    const_f32_as_f32x16!(RAD_TO_DEG_RATIO, 180.0_f32 / core::f32::consts::PI);
    self * RAD_TO_DEG_RATIO
  }

  #[inline]
  #[must_use]
  pub fn to_radians(self) -> Self {
    const_f32_as_f32x16!(DEG_TO_RAD_RATIO, core::f32::consts::PI / 180.0_f32);
    self * DEG_TO_RAD_RATIO
  }

  #[inline]
  #[must_use]
  pub fn recip(self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        // TODO: Add `_mm512_rcp14_ps` to `safe_arch`, looks like it is missing,
        // then consider updating this implementation if the relative error is
        // acceptable.
        1.0 / self
      } else {
        Self {
          a : self.a.recip(),
          b : self.b.recip(),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn recip_sqrt(self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        // TODO: Add `_mm512_rsqrt14_ps` to `safe_arch`, looks like it is
        // missing, then consider updating this implementation if the relative
        // error is acceptable.
        self.sqrt().recip()
      } else {
        Self {
          a : self.a.recip_sqrt(),
          b : self.b.recip_sqrt(),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn sqrt(self) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: sqrt_m512(self.avx512) }
      } else {
        Self {
          a : self.a.sqrt(),
          b : self.b.sqrt(),
        }
      }
    }
  }

  #[inline]
  #[must_use]
  #[doc(alias("movemask", "move_mask"))]
  pub fn to_bitmask(self) -> u32 {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        movepi32_mask_m512(self.avx512) as u32
      } else {
        (self.b.to_bitmask() << 8) | self.a.to_bitmask()
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn any(self) -> bool {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        movepi32_mask_m512(self.avx512) != 0
      } else {
        self.a.any() || self.b.any()
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn all(self) -> bool {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        movepi32_mask_m512(self.avx512) == !0_u16
      } else {
        self.a.all() && self.b.all()
      }
    }
  }

  #[inline]
  #[must_use]
  pub fn none(self) -> bool {
    !self.any()
  }

  #[inline]
  fn vm_pow2n(self) -> Self {
    const_f32_as_f32x16!(pow2_23, 8388608.0);
    const_f32_as_f32x16!(bias, 127.0);
    let a = self + (bias + pow2_23);
    let c = cast::<_, i32x16>(a) << 23;
    let std_result = cast::<_, f32x16>(c);

    let min_exp = f32x16::from(-126.0);
    let is_sub = self.simd_lt(min_exp);
    if is_sub.any() {
      let valid = self.simd_ge(f32x16::from(-149.0));
      let shift_f = self + f32x16::from(149.0);
      let mut shift_i = shift_f.trunc_int();
      shift_i = cast::<_, i32x16>(valid).blend(shift_i, i32x16::ZERO);
      let mantissa = i32x16::ONE << shift_i;
      let sub_result = cast::<_, f32x16>(mantissa);
      let sub_result = valid.blend(sub_result, f32x16::ZERO);
      is_sub.blend(sub_result, std_result)
    } else {
      std_result
    }
  }

  /// Calculate the exponent of a packed `f32x16`
  #[inline]
  #[must_use]
  pub fn exp(self) -> Self {
    const_f32_as_f32x16!(P0, 1.0 / 2.0);
    const_f32_as_f32x16!(P1, 1.0 / 6.0);
    const_f32_as_f32x16!(P2, 1.0 / 24.0);
    const_f32_as_f32x16!(P3, 1.0 / 120.0);
    const_f32_as_f32x16!(P4, 1.0 / 720.0);
    const_f32_as_f32x16!(P5, 1.0 / 5040.0);
    // LN2D_HI/LO: double-double decomposition of ln(2) for exp range reduction,
    // following the approach from fdlibm's e_exp.c (Sun Microsystems,
    // https://www.netlib.org/fdlibm/). The f32 split uses f32-precision constants
    // (0.693359375, -2.12194440e-4) summing to ln(2) with single-precision
    // accuracy; the f64 variants use a full f64 double-double
    // decomposition.
    const_f32_as_f32x16!(LN2D_HI, 0.693359375);
    const_f32_as_f32x16!(LN2D_LO, -2.12194440e-4);
    // max_x = ln(f32::MAX) ≈ 88.7229, max_r = 127 (IEEE max normal exponent)
    // min_x = -149.5 ln(2) ≈ -103.63: min r for vm_pow2n subnormal
    let max_x = f32x16::from(88.723);
    let min_x = f32x16::from(-103.63);
    // x < min_x: e^x underflows to 0 -- skip the entire pipeline
    let finite = self.is_finite();
    let neg_underflow = self.simd_lt(min_x) & finite;
    if neg_underflow.all() {
      return Self::ZERO;
    }
    let max_r = f32x16::from(127.0);
    let r = (self * Self::LOG2_E).round();
    let big = r.simd_gt(max_r);
    let r_safe = big.blend(max_r, r);
    let excess = r - max_r;
    let excess = big.blend(excess, Self::ZERO);
    let scale = Self::vm_pow2n(excess);
    let x = r.mul_neg_add(LN2D_HI, self);
    let x = r.mul_neg_add(LN2D_LO, x);
    let z = polynomial_5!(x, P0, P1, P2, P3, P4, P5);
    let x2 = x * x;
    let z = z.mul_add(x2, x);
    let n2 = Self::vm_pow2n(r_safe);
    let z = (z + Self::ONE) * scale * n2;
    let nan_mask = self.is_nan();
    let mut result = nan_mask.blend(Self::nan_pow(), z);
    let pos_overflow = self.simd_gt(max_x) & finite;
    result = pos_overflow.blend(Self::infinity(), result);
    result = neg_underflow.blend(Self::ZERO, result);
    let pos_inf = !finite & !self.is_sign_negative() & !nan_mask;
    result = pos_inf.blend(Self::infinity(), result);
    let neg_inf = !finite & self.is_sign_negative() & !nan_mask;
    result = neg_inf.blend(Self::ZERO, result);
    result
  }

  /// Calculate `e^self - 1` for each lane.
  /// Accurate even for very small values.
  #[inline]
  #[must_use]
  pub fn exp_m1(self) -> Self {
    // x < -17.329: e^x < 2⁻²⁵, exp_m1(x) = -1.0 exactly (mantissa exhaustion)
    // IEEE simd_lt returns false for NaN, so NaN lanes can't reach here.
    // -inf is < -17.329, and exp_m1(-inf) = -1.0, also correct.
    if self.simd_lt(f32x16::from(-17.329)).all() {
      return f32x16::from(-1.0);
    }
    const_f32_as_f32x16!(P0, 1.0 / 2.0);
    const_f32_as_f32x16!(P1, 1.0 / 6.0);
    const_f32_as_f32x16!(P2, 1.0 / 24.0);
    const_f32_as_f32x16!(P3, 1.0 / 120.0);
    const_f32_as_f32x16!(P4, 1.0 / 720.0);
    const_f32_as_f32x16!(P5, 1.0 / 5040.0);
    // LN2D_HI/LO: double-double decomposition of ln(2) for exp range reduction,
    // following the approach from fdlibm's e_exp.c (Sun Microsystems,
    // https://www.netlib.org/fdlibm/). The f32 split uses f32-precision constants
    // (0.693359375, -2.12194440e-4) summing to ln(2) with single-precision
    // accuracy; the f64 variants use a full f64 double-double
    // decomposition.
    const_f32_as_f32x16!(LN2D_HI, 0.693359375);
    const_f32_as_f32x16!(LN2D_LO, -2.12194440e-4);
    // max_x = ln(f32::MAX) ≈ 88.7229, max_r = 127 (IEEE max normal exponent)
    // min_x = -149.5 ln(2) ≈ -103.63: min r for vm_pow2n subnormal
    let max_x = f32x16::from(88.723);
    let min_x = f32x16::from(-103.63);
    let max_r = f32x16::from(127.0);
    let r = (self * Self::LOG2_E).round();
    let big = r.simd_gt(max_r);
    let r_safe = big.blend(max_r, r);
    let excess = r - max_r;
    let excess = big.blend(excess, Self::ZERO);
    let scale = Self::vm_pow2n(excess);
    let x = r.mul_neg_add(LN2D_HI, self);
    let x = r.mul_neg_add(LN2D_LO, x);
    let z = polynomial_5!(x, P0, P1, P2, P3, P4, P5);
    let x2 = x * x;
    let z = z.mul_add(x2, x);
    let n2 = Self::vm_pow2n(r_safe);
    let exp_val = (z + Self::ONE) * scale * n2;
    let r_is_zero = r.simd_eq(Self::ZERO);
    let z = r_is_zero.blend(z, exp_val - Self::ONE);
    let nan_mask = self.is_nan();
    let finite = self.is_finite();
    let mut result = nan_mask.blend(Self::nan_pow(), z);
    let pos_overflow = self.simd_gt(max_x) & finite;
    result = pos_overflow.blend(Self::infinity(), result);
    let neg_underflow = self.simd_lt(min_x) & finite;
    result = neg_underflow.blend(-Self::ONE, result);
    let pos_inf = !finite & !self.is_sign_negative() & !nan_mask;
    result = pos_inf.blend(Self::infinity(), result);
    let neg_inf = !finite & self.is_sign_negative() & !nan_mask;
    result = neg_inf.blend(-Self::ONE, result);
    let is_zero = self.simd_eq(Self::ZERO);
    result = is_zero.blend(self, result);
    result
  }

  /// Returns `2^self`.
  #[inline]
  #[must_use]
  pub fn exp2(self) -> Self {
    const_f32_as_f32x16!(P2, 1.0 / 2.0);
    const_f32_as_f32x16!(P3, 1.0 / 6.0);
    const_f32_as_f32x16!(P4, 1.0 / 24.0);
    const_f32_as_f32x16!(P5, 1.0 / 120.0);
    const_f32_as_f32x16!(P6, 1.0 / 720.0);
    const_f32_as_f32x16!(P7, 1.0 / 5040.0);

    // max_x = log2(f32::MAX) ≈ 127.99999
    // min_x = log2(f32::MIN_POSITIVE) - 23 ≈ -126 - 23 = -149
    let max_x = f32x16::from(127.99999);
    let min_x = f32x16::from(-149.5);
    let finite = self.is_finite();
    let neg_underflow = self.simd_lt(min_x) & finite;
    if neg_underflow.all() {
      return Self::ZERO;
    }

    let round = self.round();
    let max_r = f32x16::from(127.0);
    let big = round.simd_gt(max_r);
    let r_safe = big.blend(max_r, round);
    let excess = round - max_r;
    let excess = big.blend(excess, Self::ZERO);
    let scale = Self::vm_pow2n(excess);

    let fract = (self - round) * Self::LN_2;
    let fract_partial_exp2 = polynomial_5!(fract, P2, P3, P4, P5, P6, P7);
    let fract2 = fract * fract;
    let fract_exp2 = fract_partial_exp2.mul_add(fract2, fract) + Self::ONE;

    let n2 = Self::vm_pow2n(r_safe);
    let result = fract_exp2 * scale * n2;

    let nan_mask = self.is_nan();
    let mut result = nan_mask.blend(Self::nan_pow(), result);
    let pos_overflow = self.simd_gt(max_x) & finite;
    result = pos_overflow.blend(Self::infinity(), result);
    result = neg_underflow.blend(Self::ZERO, result);
    let pos_inf = !finite & !self.is_sign_negative() & !nan_mask;
    result = pos_inf.blend(Self::infinity(), result);
    let neg_inf = !finite & self.is_sign_negative() & !nan_mask;
    result = neg_inf.blend(Self::ZERO, result);
    result
  }

  #[inline]
  fn exponent(self) -> Self {
    const_f32_as_f32x16!(pow2_23, 8388608.0);
    const_f32_as_f32x16!(bias, 127.0);
    let a = cast::<_, u32x16>(self);
    let b = a >> 23;
    let c = b | cast::<_, u32x16>(pow2_23);
    let d = cast::<_, f32x16>(c);
    let e = d - (pow2_23 + bias);
    e
  }

  #[inline]
  fn fraction_2(self) -> Self {
    let t1 = cast::<_, u32x16>(self);
    let t2 = cast::<_, u32x16>(
      (t1 & u32x16::from(0x007FFFFF)) | u32x16::from(0x3F000000),
    );
    cast::<_, f32x16>(t2)
  }

  #[inline]
  fn is_zero_or_subnormal(self) -> Self {
    let t = cast::<_, i32x16>(self);
    let t = t & i32x16::splat(0x7F800000);
    let mask = t.simd_eq(i32x16::splat(0));
    cast::<_, f32x16>(mask)
  }

  #[inline]
  fn infinity() -> Self {
    cast::<_, f32x16>(i32x16::splat(0x7F800000))
  }

  #[inline]
  fn nan_log() -> Self {
    cast::<_, f32x16>(i32x16::splat(0x7FC00000 | 0x101 & 0x003FFFFF))
  }

  #[inline]
  fn nan_pow() -> Self {
    cast::<_, f32x16>(i32x16::splat(0x7FC00000 | 0x101 & 0x003FFFFF))
  }

  /// Returns true for each element if it has a positive sign, including `+0.0`,
  /// `NaN`s with positive sign bit and positive infinity.
  #[inline]
  #[must_use]
  pub fn is_sign_positive(self) -> Self {
    const SIGN_MASK: u32x16 = u32x16::splat((-0.0_f32).to_bits());

    let bits = cast::<f32x16, u32x16>(self);
    let sign = bits & SIGN_MASK;
    let result = sign.simd_eq(u32x16::ZERO);
    cast::<u32x16, f32x16>(result)
  }

  /// Returns true for each element if it has a negative sign, including `-0.0`,
  /// `NaN`s with negative sign bit and negative infinity.
  #[inline]
  #[must_use]
  pub fn is_sign_negative(self) -> Self {
    const SIGN_MASK: u32x16 = u32x16::splat((-0.0_f32).to_bits());

    let bits = cast::<f32x16, u32x16>(self);
    let sign = bits & SIGN_MASK;
    let result = sign.simd_eq(SIGN_MASK);
    cast::<u32x16, f32x16>(result)
  }

  /// horizontal add of all the elements of the vector
  #[inline]
  #[must_use]
  pub fn reduce_add(self) -> f32 {
    pick! {
      if #[cfg(target_feature="avx512f")]{
        reduce_add_m512(self.avx512)
      } else {
        self.a.reduce_add() + self.b.reduce_add()
      }
    }
  }

  /// horizontal multiplication of all the elements of the vector
  #[inline]
  #[must_use]
  pub fn reduce_mul(self) -> f32 {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        // TODO: Add `reduce_mul_m512` to `safe_arch` then make this function
        // safe.
        #[cfg(target_arch = "x86")]
        use core::arch::x86::_mm512_reduce_mul_ps;
        #[cfg(target_arch = "x86_64")]
        use core::arch::x86_64::_mm512_reduce_mul_ps;

        unsafe { _mm512_reduce_mul_ps(self.avx512.0) }
      } else {
        self.a.reduce_mul() * self.b.reduce_mul()
      }
    }
  }

  /// Natural log (ln(x))
  #[inline]
  #[must_use]
  pub fn ln(self) -> Self {
    const_f32_as_f32x16!(HALF, 0.5);
    const_f32_as_f32x16!(P0, 3.3333331174E-1);
    const_f32_as_f32x16!(P1, -2.4999993993E-1);
    const_f32_as_f32x16!(P2, 2.0000714765E-1);
    const_f32_as_f32x16!(P3, -1.6668057665E-1);
    const_f32_as_f32x16!(P4, 1.4249322787E-1);
    const_f32_as_f32x16!(P5, -1.2420140846E-1);
    const_f32_as_f32x16!(P6, 1.1676998740E-1);
    const_f32_as_f32x16!(P7, -1.1514610310E-1);
    const_f32_as_f32x16!(P8, 7.0376836292E-2);
    const_f32_as_f32x16!(LN2F_HI, 0.693359375);
    const_f32_as_f32x16!(LN2F_LO, -2.12194440e-4);
    const_f32_as_f32x16!(VM_SMALLEST_NORMAL, 1.17549435E-38);

    let x1 = self;
    let x = Self::fraction_2(x1);
    let e = Self::exponent(x1);
    let mask = x.simd_gt(Self::SQRT_2 * HALF);
    let x = (!mask).blend(x + x, x);
    let fe = mask.blend(e + Self::ONE, e);
    let x = x - Self::ONE;
    let res = polynomial_8!(x, P0, P1, P2, P3, P4, P5, P6, P7, P8);
    let x2 = x * x;
    let res = x2 * x * res;
    let res = fe.mul_add(LN2F_LO, res);
    let res = res + x2.mul_neg_add(HALF, x);
    let res = fe.mul_add(LN2F_HI, res);
    let overflow = !self.is_finite();
    let underflow = x1.simd_lt(VM_SMALLEST_NORMAL);
    let mask = overflow | underflow;
    if !mask.any() {
      res
    } else {
      let is_zero = self.is_zero_or_subnormal();
      let res = underflow.blend(Self::nan_log(), res);
      // Note: is_zero_or_subnormal() lumps subnormals (exponent==0) with zero.
      // Both get -Inf here. True subnormal inputs (~1.4e-45..1.175e-38) should
      // produce a finite negative result, but are vanishingly rare in
      // practice.
      let res = is_zero.blend(-Self::infinity(), res);
      let res = overflow.blend(self, res);
      // This must come *after* overflow.blend to overwrite ln(-∞) = -∞ to NaN
      let res = (!self.is_finite() & self.is_sign_negative())
        .blend(Self::nan_log(), res);
      res
    }
  }

  /// Calculate `ln(1 + self)` for each lane.
  /// Accurate even for very small values.
  #[inline]
  #[must_use]
  pub fn ln_1p(self) -> Self {
    // Based on the identity ln(1+x) = x·ln(1+x)/((1+x)-1), i.e. x·ln(u)/(u-1)
    // where u = 1+x. From MUSL libc (Rich Felker et al., https://musl.libc.org) src/math/log1pf.c
    // and fdlibm (Sun Microsystems, https://www.netlib.org/fdlibm/) s_log1p.c.
    // When 1+x rounds to 1 exactly (subnormal x), return x directly.
    // When 1+x overflows (+inf), return ln(u) without correction.
    // Mathematically exact: compensates for the rounding loss in 1+x without
    // needing a series threshold.
    let u = self + Self::ONE;
    let eq = u.simd_eq(Self::ONE);
    let ln_u = Self::ln(u);
    let correction = self * (ln_u / (u - Self::ONE));
    let result = eq.blend(self, correction);
    let over = u.is_inf();
    over.blend(ln_u, result)
  }

  #[inline]
  #[must_use]
  pub fn log2(self) -> Self {
    Self::ln(self) * Self::LOG2_E
  }

  #[inline]
  #[must_use]
  pub fn log10(self) -> Self {
    Self::ln(self) * Self::LOG10_E
  }

  #[inline]
  #[must_use]
  pub fn pow_f32x16(self, y: Self) -> Self {
    const_f32_as_f32x16!(ln2f_hi, 0.693359375);
    const_f32_as_f32x16!(ln2f_lo, -2.12194440e-4);
    const_f32_as_f32x16!(P0logf, 3.3333331174E-1);
    const_f32_as_f32x16!(P1logf, -2.4999993993E-1);
    const_f32_as_f32x16!(P2logf, 2.0000714765E-1);
    const_f32_as_f32x16!(P3logf, -1.6668057665E-1);
    const_f32_as_f32x16!(P4logf, 1.4249322787E-1);
    const_f32_as_f32x16!(P5logf, -1.2420140846E-1);
    const_f32_as_f32x16!(P6logf, 1.1676998740E-1);
    const_f32_as_f32x16!(P7logf, -1.1514610310E-1);
    const_f32_as_f32x16!(P8logf, 7.0376836292E-2);

    const_f32_as_f32x16!(p2expf, 1.0 / 2.0); // coefficients for Taylor expansion of exp
    const_f32_as_f32x16!(p3expf, 1.0 / 6.0);
    const_f32_as_f32x16!(p4expf, 1.0 / 24.0);
    const_f32_as_f32x16!(p5expf, 1.0 / 120.0);
    const_f32_as_f32x16!(p6expf, 1.0 / 720.0);
    const_f32_as_f32x16!(p7expf, 1.0 / 5040.0);

    let x1 = self.abs();
    let x = x1.fraction_2();
    let mask = x.simd_gt(f32x16::SQRT_2 * f32x16::HALF);
    let x = (!mask).blend(x + x, x);

    let x = x - f32x16::ONE;
    let x2 = x * x;
    let lg1 = polynomial_8!(
      x, P0logf, P1logf, P2logf, P3logf, P4logf, P5logf, P6logf, P7logf, P8logf
    );
    let lg1 = lg1 * x2 * x;

    let ef = x1.exponent();
    let ef = mask.blend(ef + f32x16::ONE, ef);
    let e1 = (ef * y).round();
    let yr = ef.mul_sub(y, e1);

    let lg = f32x16::HALF.mul_neg_add(x2, x) + lg1;
    let x2_err = (f32x16::HALF * x).mul_sub(x, f32x16::HALF * x2);
    let lg_err = f32x16::HALF.mul_add(x2, lg - x) - lg1;

    let e2 = (lg * y * f32x16::LOG2_E).round();
    let v = lg.mul_sub(y, e2 * ln2f_hi);
    let v = e2.mul_neg_add(ln2f_lo, v);
    let v = v - (lg_err + x2_err).mul_sub(y, yr * f32x16::LN_2);

    let x = v;
    let e3 = (x * f32x16::LOG2_E).round();
    let x = e3.mul_neg_add(f32x16::LN_2, x);
    let x2 = x * x;
    let z = x2.mul_add(
      polynomial_5!(x, p2expf, p3expf, p4expf, p5expf, p6expf, p7expf),
      x + f32x16::ONE,
    );

    let ee = e1 + e2 + e3;
    let ei = cast::<_, i32x16>(ee.round_int());
    let ej = cast::<_, i32x16>(ei + (cast::<_, i32x16>(z) >> 23));

    let overflow = cast::<_, f32x16>(ej.simd_gt(i32x16::splat(0x0FF)))
      | (ee.simd_gt(f32x16::splat(300.0)));
    let underflow = cast::<_, f32x16>(ej.simd_lt(i32x16::splat(0x000)))
      | (ee.simd_lt(f32x16::splat(-300.0)));

    // Add exponent by integer addition
    let z = cast::<_, f32x16>(cast::<_, i32x16>(z) + (ei << 23));
    // Check for overflow/underflow
    let z = underflow.blend(f32x16::ZERO, z);
    let z = overflow.blend(Self::infinity(), z);

    // Check for self == 0
    let x_zero = self.is_zero_or_subnormal();
    let z = x_zero.blend(
      y.simd_lt(f32x16::ZERO).blend(
        Self::infinity(),
        y.simd_eq(f32x16::ZERO).blend(f32x16::ONE, f32x16::ZERO),
      ),
      z,
    );

    let x_sign = self.is_sign_negative();
    let z = if x_sign.any() {
      // Y into an integer
      let yi = y.simd_eq(y.round());

      // Is y odd?
      let y_odd = cast::<_, i32x16>(y.round_int() << 31).round_float();

      let z1 =
        yi.blend(z | y_odd, self.simd_eq(Self::ZERO).blend(z, Self::nan_pow()));

      x_sign.blend(z1, z)
    } else {
      z
    };

    let x_finite = self.is_finite();
    let y_finite = y.is_finite();
    let e_finite = ee.is_finite();
    if (x_finite & y_finite & (e_finite | x_zero)).all() {
      return z;
    }

    (self.is_nan() | y.is_nan()).blend(self + y, z)
  }

  #[inline]
  pub fn powf(self, y: f32) -> Self {
    Self::pow_f32x16(self, f32x16::splat(y))
  }

  /// Transpose matrix of 16x16 `f32` matrix. Currently not accelerated.
  #[must_use]
  #[inline]
  pub fn transpose(data: [f32x16; 16]) -> [f32x16; 16] {
    // TODO: Add `_mm512_unpackhi_ps` to `safe_arch`, looks like it is missing,
    // then try adding an optimized `avx512f` implementation.

    #[inline(always)]
    fn transpose_column(data: &[f32x16; 16], index: usize) -> f32x16 {
      f32x16::new([
        data[0].as_array()[index],
        data[1].as_array()[index],
        data[2].as_array()[index],
        data[3].as_array()[index],
        data[4].as_array()[index],
        data[5].as_array()[index],
        data[6].as_array()[index],
        data[7].as_array()[index],
        data[8].as_array()[index],
        data[9].as_array()[index],
        data[10].as_array()[index],
        data[11].as_array()[index],
        data[12].as_array()[index],
        data[13].as_array()[index],
        data[14].as_array()[index],
        data[15].as_array()[index],
      ])
    }

    [
      transpose_column(&data, 0),
      transpose_column(&data, 1),
      transpose_column(&data, 2),
      transpose_column(&data, 3),
      transpose_column(&data, 4),
      transpose_column(&data, 5),
      transpose_column(&data, 6),
      transpose_column(&data, 7),
      transpose_column(&data, 8),
      transpose_column(&data, 9),
      transpose_column(&data, 10),
      transpose_column(&data, 11),
      transpose_column(&data, 12),
      transpose_column(&data, 13),
      transpose_column(&data, 14),
      transpose_column(&data, 15),
    ]
  }

  #[inline]
  #[must_use]
  pub fn to_array(self) -> [f32; 16] {
    cast(self)
  }

  #[inline]
  #[must_use]
  pub fn as_array(&self) -> &[f32; 16] {
    cast_ref(self)
  }

  #[inline]
  #[must_use]
  pub fn as_mut_array(&mut self) -> &mut [f32; 16] {
    cast_mut(self)
  }

  #[inline]
  pub fn from_i32x16(v: i32x16) -> Self {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: convert_to_m512_from_i32_m512i(v.avx512) }
      } else {
        Self {
          a: f32x8::from_i32x8(v.a),
          b: f32x8::from_i32x8(v.b),
        }
      }
    }
  }

  /// Returns true for each element if its sign bit is set.
  ///
  /// If the sign bit is set, the result has all bits set, not just the sign
  /// bit. This has been renamed to [`is_sign_negative`].
  ///
  /// [`is_sign_negative`]: Self::is_sign_negative
  #[inline]
  #[must_use]
  #[deprecated(since = "1.4.0", note = "renamed to `is_sign_negative`")]
  pub fn sign_bit(self) -> Self {
    self.is_sign_negative()
  }
}

impl Not for f32x16 {
  type Output = Self;
  #[inline]
  fn not(self) -> Self::Output {
    pick! {
      if #[cfg(target_feature="avx512f")] {
        Self { avx512: bitxor_m512(self.avx512, set_splat_m512(f32::from_bits(u32::MAX))) }
      } else {
        Self {
          a : self.a.not(),
          b : self.b.not(),
        }
      }
    }
  }
}