Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Implement Vec::from_elem specialization for all Copy types #41335

Closed
wants to merge 2 commits into from
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 0 additions & 1 deletion src/libcollections/lib.rs
Original file line number Diff line number Diff line change
Expand Up @@ -35,7 +35,6 @@
#![feature(box_patterns)]
#![feature(box_syntax)]
#![cfg_attr(not(test), feature(char_escape_debug))]
#![cfg_attr(not(test), feature(core_float))]
#![feature(core_intrinsics)]
#![feature(dropck_eyepatch)]
#![feature(exact_size_is_empty)]
Expand Down
90 changes: 45 additions & 45 deletions src/libcollections/vec.rs
Original file line number Diff line number Diff line change
Expand Up @@ -77,8 +77,6 @@ use core::hash::{self, Hash};
use core::intrinsics::{arith_offset, assume};
use core::iter::{FromIterator, FusedIterator, TrustedLen};
use core::mem;
#[cfg(not(test))]
use core::num::Float;
use core::ops::{InPlace, Index, IndexMut, Place, Placer};
use core::ops;
use core::ptr;
Expand Down Expand Up @@ -1388,59 +1386,61 @@ impl<T: Clone> SpecFromElem for T {
}
}

impl SpecFromElem for u8 {
#[inline]
fn from_elem(elem: u8, n: usize) -> Vec<u8> {
if elem == 0 {
// Computes the bitwise OR of the input, reinterpreted as [U].
// Assumes that U is a primitive integer type and that `T` can be
// represented exactly as a slice of elements of type `U`, i.e.
// `mem::size_of::<T>() % mem::size_of::<U>() == 0`
unsafe fn chunked_or<T, U: ops::BitOr<Output = U> + Copy>(x: T) -> U {
let p = &x as *const T as *const U;
let len = mem::size_of::<T>() / mem::size_of::<U>();
slice::from_raw_parts(p, len).iter().fold(mem::zeroed(), |state, &x| state | x)
}

// Checks if the raw representation of the input is only binary zeroes.
// Instead of comparing each byte with 0, the whole memory region is
// OR-ed together and the result is compared to 0.
fn is_zero<T: Copy>(x: T) -> bool {
// Find the greatest alignment that can be used to scan x, as that
// leads to less code and better performance.
// If the alignment is greater than 16, compute the OR using u128,
// as no bigger native integers are available.
// The calls to chunked_or() are safe because mem::size_of::<T>()
// is guaranteed to be a multiple of mem::align_of::<T>().
unsafe {
match mem::align_of::<T>() {
n if n % 16 == 0 => 0u128 == chunked_or(x),
n if n % 8 == 0 => 0u64 == chunked_or(x),
n if n % 4 == 0 => 0u32 == chunked_or(x),
n if n % 2 == 0 => 0u16 == chunked_or(x),
_ => 0u8 == chunked_or(x),
}
}
}
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you write some comments explaining how chunked_or and is_zero work? Are there any pitfalls? What happens when the alignment of a type is bigger than 16?

Copy link
Contributor Author

@ranma42 ranma42 Apr 19, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I added some comments that should answer your questions, with the exception of "are there any pitfalls?".
There should be no correctness issues, but there is certainly a performance tradeoff: we are now checking the contents of the input and we pay for that.
We are making vec![None; 1024*1024] much faster, but vec![Some(3); 1] is going to be a little slower (if the compiler can see the literal value, the checks are going to be optimized away, but in general that might not be possible).


impl<T: Copy> SpecFromElem for T {
default fn from_elem(elem: Self, n: usize) -> Vec<Self> {
if is_zero(elem) {
return Vec {
buf: RawVec::with_capacity_zeroed(n),
len: n,
}
}
unsafe {
let mut v = Vec::with_capacity(n);
ptr::write_bytes(v.as_mut_ptr(), elem, n);
v.set_len(n);
v
}
}
}

macro_rules! impl_spec_from_elem {
($t: ty, $is_zero: expr) => {
impl SpecFromElem for $t {
#[inline]
fn from_elem(elem: $t, n: usize) -> Vec<$t> {
if $is_zero(elem) {
return Vec {
buf: RawVec::with_capacity_zeroed(n),
len: n,
}
}
let mut v = Vec::with_capacity(n);
v.extend_with_element(n, elem);
v
let mut v = Vec::with_capacity(n);
if mem::size_of::<T>() == 1 {
unsafe {
// let elem: u8 = mem::transmute(elem);
let elem: u8 = *(&elem as *const T as *const u8);
ptr::write_bytes(v.as_mut_ptr(), elem, n);
v.set_len(n);
}
} else {
v.extend_with_element(n, elem);
}
};
v
}
}

impl_spec_from_elem!(i8, |x| x == 0);
impl_spec_from_elem!(i16, |x| x == 0);
impl_spec_from_elem!(i32, |x| x == 0);
impl_spec_from_elem!(i64, |x| x == 0);
impl_spec_from_elem!(i128, |x| x == 0);
impl_spec_from_elem!(isize, |x| x == 0);

impl_spec_from_elem!(u16, |x| x == 0);
impl_spec_from_elem!(u32, |x| x == 0);
impl_spec_from_elem!(u64, |x| x == 0);
impl_spec_from_elem!(u128, |x| x == 0);
impl_spec_from_elem!(usize, |x| x == 0);

impl_spec_from_elem!(f32, |x: f32| x == 0. && x.is_sign_positive());
impl_spec_from_elem!(f64, |x: f64| x == 0. && x.is_sign_positive());

////////////////////////////////////////////////////////////////////////////////
// Common trait implementations for Vec
////////////////////////////////////////////////////////////////////////////////
Expand Down