Trait cubecl_core::frontend::Numeric

source ·

pub trait Numeric: Copy + Abs + Max + Min + Clamp + Remainder + ExpandElementBaseInit + CubePrimitive + LaunchArgExpand + AddAssign + SubAssign + MulAssign + DivAssign + Add<Output = Self> + Sub<Output = Self> + Mul<Output = Self> + Div<Output = Self> + PartialOrd + Index<UInt, Output = Self> + IndexMut<UInt, Output = Self> + Index<u32, Output = Self> + IndexMut<u32, Output = Self> + From<u32> + Add<u32, Output = Self> + Sub<u32, Output = Self> + Mul<u32, Output = Self> + Div<u32, Output = Self> + AddAssign<u32> + SubAssign<u32> + MulAssign<u32> + DivAssign<u32> + PartialOrd<u32> + PartialEq<u32> {
    type Primitive: ScalarArgSettings;

    // Provided methods
    fn from_int(_val: u32) -> Self { ... }
    fn from_vec<const D: usize>(_vec: [u32; D]) -> Self { ... }
    fn __expand_from_int(
        _context: &mut CubeContext,
        val: ExpandElementTyped<I64>,
    ) -> <Self as CubeType>::ExpandType { ... }
    fn __expand_from_vec<const D: usize>(
        context: &mut CubeContext,
        vec: [ExpandElementTyped<UInt>; D],
    ) -> <Self as CubeType>::ExpandType { ... }
}

Expand description

Type that encompasses both (unsigned or signed) integers and floats Used in kernels that should work for both.

Required Associated Types§

source

type Primitive: ScalarArgSettings

Provided Methods§

source

fn from_int(_val: u32) -> Self

Create a new constant numeric.

Note: since this must work for both integer and float only the less expressive of both can be created (int) If a number with decimals is needed, use Float::new.

This method panics when unexpanded. For creating an element with a val, use the new method of the sub type.

source