Skip to content

Make the matches field smaller to improve cache locality #293

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.

Sign up for GitHub

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

Merged
folkertdev merged 2 commits into from
Feb 13, 2025
Merged

Make the matches field smaller to improve cache locality #293

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3cc70aa
Make the matches field smaller to improve cache locality
brianpane Feb 13, 2025
2f0b0a3
Update comments to clarify how matches count is used
brianpane Feb 13, 2025
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
15 changes: 11 additions & 4 deletions zlib-rs/src/deflate.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -374,6 +374,7 @@ pub fn init(stream: &mut z_stream, config: DeflateConfig) -> ReturnCode {
_cache_line_1: (),
_cache_line_2: (),
_cache_line_3: (),
_padding_0: 0,
};

unsafe { state_allocation.as_ptr().write(state) }; // FIXME: write is stable for NonNull since 1.80.0
Expand Down Expand Up @@ -676,6 +677,7 @@ pub fn copy<'a>(
_cache_line_1: (),
_cache_line_2: (),
_cache_line_3: (),
_padding_0: source_state._padding_0,
};

// write the cloned state into state_ptr
Expand Down Expand Up @@ -1281,6 +1283,12 @@ pub(crate) struct State<'a> {
/// than this value. This mechanism is used only for compression levels >= 4.
pub(crate) max_lazy_match: u16,

/// number of string matches in current block
/// NOTE: this is a saturating 8-bit counter, to help keep the struct compact. The code that
/// makes decisions based on this field only cares whether the count is greater than 2, so
/// an 8-bit counter is sufficient.
pub(crate) matches: u8,

/// Window position at the beginning of the current output block. Gets
/// negative when the window is moved backwards.
pub(crate) block_start: isize,
Expand Down Expand Up @@ -1315,9 +1323,6 @@ pub(crate) struct State<'a> {

_cache_line_2: (),

/// number of string matches in current block
pub(crate) matches: usize,

/// bit length of current block with optimal trees
opt_len: usize,
/// bit length of current block with static trees
Expand All @@ -1334,6 +1339,8 @@ pub(crate) struct State<'a> {
gzhead: Option<&'a mut gz_header>,
gzindex: usize,

_padding_0: usize,

_cache_line_3: (),

crc_fold: crate::crc32::Crc32Fold,
Expand Down Expand Up @@ -1457,7 +1464,7 @@ impl<'a> State<'a> {
pub(crate) fn tally_dist(&mut self, mut dist: usize, len: usize) -> bool {
self.sym_buf.push_dist(dist as u16, len as u8);

self.matches += 1;
self.matches = self.matches.saturating_add(1);
dist -= 1;

assert!(
Expand Down
Loading