Skip to content

Added compile time optimization for bytewise slice comparison #39422

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

Closed
wants to merge 4 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: 1 addition & 0 deletions src/libcore/Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
authors = ["The Rust Project Developers"]
name = "core"
version = "0.0.0"
build = "build.rs"

[lib]
name = "core"
Expand Down
90 changes: 90 additions & 0 deletions src/libcore/build.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,90 @@
// Copyright 2012-2014 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.

//! Code generation for bytewise slice comparison. Target is to provide a compile time optimization
//! within a single macro which looks like:
//!
//! ```
//! macro_rules! slice_compare (
//! ($a:expr, $b:expr, $len:expr) => {{
//! match $len {
//! 1 => cmp!($a, $b, u8, 0),
//! 2 => cmp!($a, $b, u16, 0),
//! 3 => cmp!($a, $b, u16, 0) && cmp!($a, $b, u8, 2),
//! 4 => cmp!($a, $b, u32, 0),
//! ...
//! }
//! }}
//! );
//! ```
//!
//! The supported slice length can be set by changing the `OPT_LEN` variable.

static OPT_LEN: usize = 256;

use std::{env, iter, io};
use std::fs::File;
use std::io::Write;
use std::path::Path;
use std::error::Error;

// Generates the code for the compile time optimization for bytewise slice comparison
pub fn main() {
run().expect("Could not generate slice comparison source code.");
}

fn fill(indent: usize) -> String {
iter::repeat(' ').take(indent).collect()
}

fn run() -> Result<(), Box<Error>> {
let out_dir = env::var("OUT_DIR")?;
let dest_path = Path::new(&out_dir).join("memcmp_optimization.rs");
let mut f = File::create(&dest_path)?;

// Generate the slice comparison source code
writeln!(f, "macro_rules! slice_compare (")?;
writeln!(f, "{}($a:expr, $b:expr, $len:expr) => {{{{", fill(4))?;
writeln!(f, "{}match $len {{", fill(8))?;

for i in 1..OPT_LEN + 1 {
let mut bits = i * 8 as usize;
let mut sizes = vec![8, 16, 32, 64];
let mut offset = 0;

write!(f, "{}{} => ", fill(12), i)?;
while !sizes.is_empty() {
let size = sizes.last().ok_or(io::Error::from(io::ErrorKind::Other))?.clone();
if bits >= size {
if offset > 0 {
write!(f, " && ")?;
}
write!(f, "cmp!($a, $b, u{}, {})", size, offset)?;
bits = bits.checked_sub(size).ok_or(io::Error::from(io::ErrorKind::Other))?;
offset += size / 8;
} else {
sizes.pop();
}
if bits == 0 {
break;
}
}
writeln!(f, ",")?;
}

writeln!(f,
"{}_ => unsafe {{ memcmp($a, $b, $len) == 0 }},",
fill(12))?;

writeln!(f, "{}}}", fill(8))?;
writeln!(f, "{}}}}}", fill(4))?;
writeln!(f, ");")?;
Ok(())
}
17 changes: 12 additions & 5 deletions src/libcore/slice.rs
Original file line number Diff line number Diff line change
Expand Up @@ -2177,6 +2177,16 @@ pub unsafe fn from_raw_parts_mut<'a, T>(p: *mut T, len: usize) -> &'a mut [T] {
// Comparison traits
//

// A compile time optimization for faster bytewise slice comparison
include!(concat!(env!("OUT_DIR"), "/memcmp_optimization.rs"));

// The compare macro for the bytewise slice comparison
macro_rules! cmp (
($left:expr, $right: expr, $var:ident, $offset:expr) => {{
unsafe {*($left.offset($offset) as *const $var) == *($right.offset($offset) as *const $var)}
Copy link
Member

Choose a reason for hiding this comment

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

This is a possibly unaligned load (for example *const u64 used that is not necessarily well aligned for u64. We can't merge the code like this, because it is not portable and will crash on certain platforms.)

Copy link
Contributor

Choose a reason for hiding this comment

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

Fixing this issue will likely have an impact on the performance. In order to get the best possible performance on platforms which allow unaligned memory access we might need/want to perform this optimisation at a lower level (i.e. replace the implementation of memcmp and/or the strategy used in LLVM to replace memcmp with inline comparisons).

Copy link
Author

Choose a reason for hiding this comment

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

I am currently not sure how to fix this unaligned loads here. Is there something like get_unaligned in the compiler?

}}
);

extern {
/// Call implementation provided memcmp
///
Expand Down Expand Up @@ -2254,11 +2264,8 @@ impl<A> SlicePartialEq<A> for [A]
if self.as_ptr() == other.as_ptr() {
return true;
}
unsafe {
let size = mem::size_of_val(self);
memcmp(self.as_ptr() as *const u8,
other.as_ptr() as *const u8, size) == 0
}
let size = mem::size_of_val(self);
slice_compare!(other.as_ptr() as *const u8, self.as_ptr() as *const u8, size)
}
}

Expand Down