Skip to content

Commit

Permalink
[zerocopy] Impl FromBytes/AsByes for SIMD types
Browse files Browse the repository at this point in the history 
Put these impls behind the `simd` feature flag because the layout of
SIMD types is not yet stable. Provide impls of nightly-only SIMD types
behind the `simd-nightly` feature flag.

Fixed: 83299

Change-Id: I112d77bb63dbfd8c84b96445d2ec58efcdc88fda
Reviewed-on: https://fuchsia-review.googlesource.com/c/fuchsia/+/573083
Commit-Queue: Joshua Liebow-Feeser <joshlf@google.com>
Reviewed-by: Adam Perry <adamperry@google.com>
Reviewed-by: Paul Faria <paulfaria@google.com>
  • Loading branch information
@joshlf
joshlf authored and CQ Bot committed Aug 28, 2021
1 parent 2bf6960 commit cfb2a95
Show file tree
Hide file tree
Showing 2 changed files with 169 additions and 29 deletions.
5 changes: 5 additions & 0 deletions Cargo.toml.crates-io
View file
Original file line number Diff line number Diff line change
Expand Up @@ -15,8 +15,13 @@ repository = "https://fuchsia.googlesource.com/fuchsia/+/HEAD/src/lib/zerocopy"

include = ["src/*", "Cargo.toml", "LICENSE"]

[package.metadata.docs.rs]
all-features = true

[features]
alloc = []
simd = []
simd-nightly = ["simd"]

[dependencies]
zerocopy-derive = "0.3.0"
Expand Down
193 changes: 164 additions & 29 deletions src/lib.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,25 @@
//!
//! Note that these traits are ignorant of byte order. For byte order-aware
//! types, see the [`byteorder`] module.
//!
//! # Features
//!
//! `alloc`: By default, `zerocopy` is `no_std`. When the `alloc` feature is
//! enabled, the `alloc` crate is added as a dependency, and some
//! allocation-related functionality is added.
//!
//! `simd`: When the `simd` feature is enabled, `FromBytes` and `AsBytes` impls
//! are emitted for all stable SIMD types which exist on the target platform.
//! Note that the layout of SIMD types is not yet stabilized, so these impls may
//! be removed in the future if layout changes make them invalid. For more
//! information, see the Unsafe Code Guidelines Reference page on the [Layout of
//! packed SIMD vectors][simd-layout].
//!
//! `simd-nightly`: Enables the `simd` feature and adds support for SIMD types
//! which are only available on nightly. Since these types are unstable, support
//! for any type may be removed at any point in the future.
//!
//! [simd-layout]: https://rust-lang.github.io/unsafe-code-guidelines/layout/packed-simd-vectors.html
#![deny(missing_docs)]
#![cfg_attr(not(test), no_std)]
Expand Down Expand Up @@ -80,24 +99,30 @@ macro_rules! impl_for_composite_types {
};
}

// implement an unsafe trait for all signed and unsigned primitive types
macro_rules! impl_for_primitives {
($trait:ident) => (
impl_for_primitives!(@inner $trait, u8, i8, u16, i16, u32, i32, u64, i64, u128, i128, usize, isize, f32, f64);
);
(@inner $trait:ident, $type:ty) => (
/// Implements `$trait` for one or more `$type`s.
macro_rules! impl_for_types {
($trait:ident, $type:ty) => (
unsafe impl $trait for $type {
fn only_derive_is_allowed_to_implement_this_trait() where Self: Sized {}
}
);
(@inner $trait:ident, $type:ty, $($types:ty),*) => (
($trait:ident, $type:ty, $($types:ty),*) => (
unsafe impl $trait for $type {
fn only_derive_is_allowed_to_implement_this_trait() where Self: Sized {}
}
impl_for_primitives!(@inner $trait, $($types),*);
impl_for_types!($trait, $($types),*);
);
}

/// Implements `$trait` for all signed and unsigned primitive types.
macro_rules! impl_for_primitives {
($trait:ident) => {
impl_for_types!(
$trait, u8, i8, u16, i16, u32, i32, u64, i64, u128, i128, usize, isize, f32, f64
);
};
}

/// Types for which any byte pattern is valid.
///
/// WARNING: Do not implement this trait yourself! Instead, use
Expand Down Expand Up @@ -356,14 +381,8 @@ pub unsafe trait AsBytes {
}
}

// Special case for bool
unsafe impl AsBytes for bool {
fn only_derive_is_allowed_to_implement_this_trait()
where
Self: Sized,
{
}
}
// Special case for bool (it is not included in `impl_for_primitives!`).
impl_for_types!(AsBytes, bool);

impl_for_primitives!(FromBytes);
impl_for_primitives!(AsBytes);
Expand Down Expand Up @@ -392,21 +411,137 @@ pub unsafe trait Unaligned {
Self: Sized;
}

unsafe impl Unaligned for u8 {
fn only_derive_is_allowed_to_implement_this_trait()
where
Self: Sized,
{
}
}
unsafe impl Unaligned for i8 {
fn only_derive_is_allowed_to_implement_this_trait()
where
Self: Sized,
{
impl_for_types!(Unaligned, u8, i8);
impl_for_composite_types!(Unaligned);

// SIMD support
//
// Per the Unsafe Code Guidelines Reference [1]:
//
// Packed SIMD vector types are `repr(simd)` homogeneous tuple-structs
// containing `N` elements of type `T` where `N` is a power-of-two and the
// size and alignment requirements of `T` are equal:
//
// ```rust
// #[repr(simd)]
// struct Vector<T, N>(T_0, ..., T_(N - 1));
// ```
//
// ...
//
// The size of `Vector` is `N * size_of::<T>()` and its alignment is an
// implementation-defined function of `T` and `N` greater than or equal to
// `align_of::<T>()`.
//
// ...
//
// Vector elements are laid out in source field order, enabling random access
// to vector elements by reinterpreting the vector as an array:
//
// ```rust
// union U {
// vec: Vector<T, N>,
// arr: [T; N]
// }
//
// assert_eq!(size_of::<Vector<T, N>>(), size_of::<[T; N]>());
// assert!(align_of::<Vector<T, N>>() >= align_of::<[T; N]>());
//
// unsafe {
// let u = U { vec: Vector<T, N>(t_0, ..., t_(N - 1)) };
//
// assert_eq!(u.vec.0, u.arr[0]);
// // ...
// assert_eq!(u.vec.(N - 1), u.arr[N - 1]);
// }
// ```
//
// Given this background, we can observe that:
// - The size and bit pattern requirements of a SIMD type are equivalent to the
// equivalent array type. Thus, for any SIMD type whose primitive `T` is
// `FromBytes`, that SIMD type is also `FromBytes`. The same holds for
// `AsBytes`.
// - Since no upper bound is placed on the alignment, no SIMD type can be
// guaranteed to be `Unaligned`.
//
// Also per [1]:
//
// This chapter represents the consensus from issue #38. The statements in
// here are not (yet) "guaranteed" not to change until an RFC ratifies them.
//
// See issue #38 [2]. While this behavior is not technically guaranteed, the
// likelihood that the behavior will change such that SIMD types are no longer
// `FromBytes` or `AsBytes` is next to zero, as that would defeat the entire
// purpose of SIMD types. Nonetheless, we put this behavior behind the `simd`
// Cargo feature, which requires consumers to opt into this stability hazard.
//
// [1] https://rust-lang.github.io/unsafe-code-guidelines/layout/packed-simd-vectors.html
// [2] https://github.com/rust-lang/unsafe-code-guidelines/issues/38
#[cfg(feature = "simd")]
mod simd {
/// Defines a module which implements `FromBytes` and `AsBytes` for a set of
/// types from a module in `core::arch`.
///
/// `$arch` is both the name of the defined module and the name of the
/// module in `core::arch`, and `$typ` is the list of items from that module
/// to implement `FromBytes` and `AsBytes` for.
macro_rules! simd_arch_mod {
($arch:ident, $($typ:ident),*) => {
mod $arch {
use core::arch::$arch::{$($typ),*};

use crate::*;

impl_for_types!(FromBytes, $($typ),*);
impl_for_types!(AsBytes, $($typ),*);
}
};
}

#[cfg(target_arch = "x86")]
simd_arch_mod!(x86, __m128, __m128d, __m128i, __m256, __m256d, __m256i);
#[cfg(target_arch = "x86_64")]
simd_arch_mod!(x86_64, __m128, __m128d, __m128i, __m256, __m256d, __m256i);
#[cfg(target_arch = "wasm32")]
simd_arch_mod!(wasm32, v128);
#[cfg(all(feature = "simd-nightly", target_arch = "powerpc"))]
simd_arch_mod!(
powerpc,
vector_bool_long,
vector_double,
vector_signed_long,
vector_unsigned_long
);
#[cfg(all(feature = "simd-nightly", target_arch = "powerpc64"))]
simd_arch_mod!(
powerpc64,
vector_bool_long,
vector_double,
vector_signed_long,
vector_unsigned_long
);
#[cfg(all(feature = "simd-nightly", target_arch = "aarch64"))]
#[rustfmt::skip]
simd_arch_mod!(
aarch64, float32x2_t, float32x4_t, float64x1_t, float64x2_t, int8x8_t, int8x8x2_t,
int8x8x3_t, int8x8x4_t, int8x16_t, int8x16x2_t, int8x16x3_t, int8x16x4_t, int16x4_t,
int16x8_t, int32x2_t, int32x4_t, int64x1_t, int64x2_t, poly8x8_t, poly8x8x2_t, poly8x8x3_t,
poly8x8x4_t, poly8x16_t, poly8x16x2_t, poly8x16x3_t, poly8x16x4_t, poly16x4_t, poly16x8_t,
poly64x1_t, poly64x2_t, uint8x8_t, uint8x8x2_t, uint8x8x3_t, uint8x8x4_t, uint8x16_t,
uint8x16x2_t, uint8x16x3_t, uint8x16x4_t, uint16x4_t, uint16x8_t, uint32x2_t, uint32x4_t,
uint64x1_t, uint64x2_t
);
#[cfg(all(feature = "simd-nightly", target_arch = "arm"))]
#[rustfmt::skip]
simd_arch_mod!(
arm, float32x2_t, float32x4_t, int8x4_t, int8x8_t, int8x8x2_t, int8x8x3_t, int8x8x4_t,
int8x16_t, int16x2_t, int16x4_t, int16x8_t, int32x2_t, int32x4_t, int64x1_t, int64x2_t,
poly8x8_t, poly8x8x2_t, poly8x8x3_t, poly8x8x4_t, poly8x16_t, poly16x4_t, poly16x8_t,
poly64x1_t, poly64x2_t, uint8x4_t, uint8x8_t, uint8x8x2_t, uint8x8x3_t, uint8x8x4_t,
uint8x16_t, uint16x2_t, uint16x4_t, uint16x8_t, uint32x2_t, uint32x4_t, uint64x1_t,
uint64x2_t
);
}
impl_for_composite_types!(Unaligned);

// Used in `transmute!` below.
#[doc(hidden)]
Expand Down

0 comments on commit cfb2a95

Please sign in to comment.