Crate pulp

Expand description

pulp is a safe abstraction over SIMD instructions, that allows you to write a function once and dispatch to equivalent vectorized versions based on the features detected at runtime.

§Autovectorization example

use pulp::Arch;

let mut v = (0..1000).map(|i| i as f64).collect::<Vec<_>>();
let arch = Arch::new();

arch.dispatch(|| {
	for x in &mut v {
		*x *= 2.0;
	}
});

for (i, x) in v.into_iter().enumerate() {
	assert_eq!(x, 2.0 * i as f64);
}

§Manual vectorization example

use pulp::{Arch, Simd, WithSimd};

struct TimesThree<'a>(&'a mut [f64]);
impl<'a> WithSimd for TimesThree<'a> {
	type Output = ();

	#[inline(always)]
	fn with_simd<S: Simd>(self, simd: S) -> Self::Output {
		let v = self.0;
		let (head, tail) = S::as_mut_simd_f64s(v);

		let three = simd.splat_f64s(3.0);
		for x in head {
			*x = simd.mul_f64s(three, *x);
		}

		for x in tail {
			*x = *x * 3.0;
		}
	}
}

let mut v = (0..1000).map(|i| i as f64).collect::<Vec<_>>();
let arch = Arch::new();

arch.dispatch(TimesThree(&mut v));

for (i, x) in v.into_iter().enumerate() {
	assert_eq!(x, 3.0 * i as f64);
}

Re-exports§

pub use x86::Arch;
pub use bytemuck;
pub use num_complex;

Modules§

core_arch: Platform dependent intrinsics.
x86x86 or x86-64: Low level x86 API.

Macros§

cast: Safe transmute macro.
feature_detected
simd_type
static_assert_same_size
static_assert_size_less_than_or_equal
try_const

Structs§

MemMask
Scalar
Scalar128b
Scalar256b
Scalar512b
b8: Bitmask type for 8 elements, used for mask operations on AVX512.
b16: Bitmask type for 16 elements, used for mask operations on AVX512.
b32: Bitmask type for 32 elements, used for mask operations on AVX512.
b64: Bitmask type for 64 elements, used for mask operations on AVX512.
f32x4: A 128-bit SIMD vector with 4 elements of type f32.
f32x8: A 256-bit SIMD vector with 8 elements of type f32.
f32x16: A 512-bit SIMD vector with 16 elements of type f32.
f64x2: A 128-bit SIMD vector with 2 elements of type f64.
f64x4: A 256-bit SIMD vector with 4 elements of type f64.
f64x8: A 512-bit SIMD vector with 8 elements of type f64.
i8x16: A 128-bit SIMD vector with 16 elements of type i8.
i8x32: A 256-bit SIMD vector with 32 elements of type i8.
i8x64: A 512-bit SIMD vector with 64 elements of type i8.
i16x8: A 128-bit SIMD vector with 8 elements of type i16.
i16x16: A 256-bit SIMD vector with 16 elements of type i16.
i16x32: A 512-bit SIMD vector with 32 elements of type i16.
i32x4: A 128-bit SIMD vector with 4 elements of type i32.
i32x8: A 256-bit SIMD vector with 8 elements of type i32.
i32x16: A 512-bit SIMD vector with 16 elements of type i32.
i64x2: A 128-bit SIMD vector with 2 elements of type i64.
i64x4: A 256-bit SIMD vector with 4 elements of type i64.
i64x8: A 512-bit SIMD vector with 8 elements of type i64.
m8: Mask type with 8 bits. Its bit pattern is either all ones or all zeros. Unsafe code must not depend on this, however.
m8x16: A 128-bit SIMD vector with 16 elements of type m8.
m8x32: A 256-bit SIMD vector with 32 elements of type m8.
m16: Mask type with 16 bits. Its bit pattern is either all ones or all zeros. Unsafe code must not depend on this, however.
m32: Mask type with 32 bits. Its bit pattern is either all ones or all zeros. Unsafe code must not depend on this, however.
m64: Mask type with 64 bits. Its bit pattern is either all ones or all zeros. Unsafe code must not depend on this, however.
m16x8: A 128-bit SIMD vector with 8 elements of type m16.
m16x16: A 256-bit SIMD vector with 16 elements of type m16.
m32x4: A 128-bit SIMD vector with 4 elements of type m32.
m32x8: A 256-bit SIMD vector with 8 elements of type m32.
m32x16: A 512-bit SIMD vector with 16 elements of type m32.
m64x2: A 128-bit SIMD vector with 2 elements of type m64.
m64x4: A 256-bit SIMD vector with 4 elements of type m64.
m64x8: A 512-bit SIMD vector with 8 elements of type m64.
u8x16: A 128-bit SIMD vector with 16 elements of type u8.
u8x32: A 256-bit SIMD vector with 32 elements of type u8.
u8x64: A 512-bit SIMD vector with 64 elements of type u8.
u16x8: A 128-bit SIMD vector with 8 elements of type u16.
u16x16: A 256-bit SIMD vector with 16 elements of type u16.
u16x32: A 512-bit SIMD vector with 32 elements of type u16.
u32x4: A 128-bit SIMD vector with 4 elements of type u32.
u32x8: A 256-bit SIMD vector with 8 elements of type u32.
u32x16: A 512-bit SIMD vector with 16 elements of type u32.
u64x2: A 128-bit SIMD vector with 2 elements of type u64.
u64x4: A 256-bit SIMD vector with 4 elements of type u64.
u64x8: A 512-bit SIMD vector with 8 elements of type u64.

Traits§

Interleave: Types that allow [de]interleaving.
NullaryFnOnce
PortableSimd
Simd
WithSimd

Functions§

as_arrays: Splits a slice into chunks of equal size (known at compile time).
as_arrays_mut: Splits a slice into chunks of equal size (known at compile time).
cast: Safe transmute function.
cast_lossy: Safe lossy transmute function, where the destination type may be smaller than the source type.
iota_32
iota_64

Type Aliases§

c32
c64

Attribute Macros§

with_simdmacro: Requires the first non-lifetime generic parameter, as well as the function’s first input parameter to be the SIMD type. Also currently requires that all the lifetimes be explicitly specified.

Crate pulpCopy item path