RGBCube/uutils-coreutils
1
Fork 0
mirror of https://github.com/RGBCube/uutils-coreutils synced 2025-07-27 11:07:44 +00:00
Code Issues Activity

wc: Align buffer to 32-byte boundary

Browse source 
bytecount uses vector operations to speed up line counting.
At least on x86 with AVX2 support, the vectors are 256-byte wide,
and operations are much faster if the data is aligned.

Saves about 4% of total performance, matching wc's performance.
This commit is contained in:
Nicolas Boichat 2025-05-14 20:24:06 +08:00
parent 1fe539f1d4
commit 1fc14d837c
1 changed files with 19 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

21
src/uu/wc/src/count_fast.rs
View file

@ -197,6 +197,23 @@ pub(crate) fn count_bytes_fast<T: WordCountable>(handle: &mut T) -> (usize, Opti
}
}
/// A simple structure used to align a BUF_SIZE buffer to 32-byte boundary.
///
/// This is useful as bytecount uses 256-bit wide vector operations that run much
/// faster on aligned data (at least on x86 with AVX2 support).
#[repr(align(32))]
struct AlignedBuffer {
data: [u8; BUF_SIZE],
}
impl Default for AlignedBuffer {
fn default() -> Self {
Self {
data: [0; BUF_SIZE],
}
}
}
/// Returns a WordCount that counts the number of bytes, lines, and/or the number of Unicode characters encoded in UTF-8 read via a Reader.
///
/// This corresponds to the `-c`, `-l` and `-m` command line flags to wc.
@ -213,9 +230,9 @@ pub(crate) fn count_bytes_chars_and_lines_fast<
handle: &mut R,
) -> (WordCount, Option<io::Error>) {
let mut total = WordCount::default();
let mut buf = [0; BUF_SIZE];
let buf: &mut [u8] = &mut AlignedBuffer::default().data;
loop {
match handle.read(&mut buf) {
match handle.read(buf) {
Ok(0) => return (total, None),
Ok(n) => {
if COUNT_BYTES {