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.

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

Speed up tidy #105829

Merged
merged 3 commits into from
Dec 18, 2022
Merged

Speed up tidy #105829

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
f7bfc48
optimize line-by-line style checks in tidy
the8472 Nov 21, 2022
5620afc
poll tidy threads for completion before waiting
the8472 Nov 21, 2022
ab7d769
poll rustfmt child processes
the8472 Nov 21, 2022
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
23 changes: 19 additions & 4 deletions src/bootstrap/format.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ use std::path::{Path, PathBuf};
use std::process::{Command, Stdio};
use std::sync::mpsc::SyncSender;

fn rustfmt(src: &Path, rustfmt: &Path, paths: &[PathBuf], check: bool) -> impl FnMut() {
fn rustfmt(src: &Path, rustfmt: &Path, paths: &[PathBuf], check: bool) -> impl FnMut(bool) -> bool {
let mut cmd = Command::new(&rustfmt);
// avoid the submodule config paths from coming into play,
// we only allow a single global config for the workspace for now
Expand All @@ -23,7 +23,13 @@ fn rustfmt(src: &Path, rustfmt: &Path, paths: &[PathBuf], check: bool) -> impl F
let cmd_debug = format!("{:?}", cmd);
let mut cmd = cmd.spawn().expect("running rustfmt");
// poor man's async: return a closure that'll wait for rustfmt's completion
move || {
move |block: bool| -> bool {
if !block {
match cmd.try_wait() {
Ok(Some(_)) => {}
_ => return false,
}
}
let status = cmd.wait().unwrap();
if !status.success() {
eprintln!(
Expand All @@ -34,6 +40,7 @@ fn rustfmt(src: &Path, rustfmt: &Path, paths: &[PathBuf], check: bool) -> impl F
);
crate::detail_exit(1);
}
true
}
}

Expand Down Expand Up @@ -146,15 +153,23 @@ pub fn format(build: &Builder<'_>, check: bool, paths: &[PathBuf]) {
let child = rustfmt(&src, &rustfmt_path, paths.as_slice(), check);
children.push_back(child);

// poll completion before waiting
for i in (0..children.len()).rev() {
if children[i](false) {
children.swap_remove_back(i);
break;
}
}

if children.len() >= max_processes {
// await oldest child
children.pop_front().unwrap()();
children.pop_front().unwrap()(true);
}
}

// await remaining children
for mut child in children {
child();
child(true);
}
});

Expand Down
22 changes: 16 additions & 6 deletions src/tools/tidy/src/main.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,15 +35,26 @@ fn main() {

let bad = std::sync::Arc::new(AtomicBool::new(false));

let drain_handles = |handles: &mut VecDeque<ScopedJoinHandle<'_, ()>>| {
// poll all threads for completion before awaiting the oldest one
for i in (0..handles.len()).rev() {
if handles[i].is_finished() {
handles.swap_remove_back(i).unwrap().join().unwrap();
}
}
Comment on lines +38 to +44
Copy link
Member

Choose a reason for hiding this comment

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

Why is this useful? Does it run any faster or is it just releasing the resources earlier?

Oh, I guess removing a handle means you have more space in the queue to add more jobs? I'm a little confused why you would only run this once instead of putting it in the loop below ... maybe we should use rayon::ParallelIterator instead so we always join the first job to finish? All the custom concurrency code is a little hard to maintain in general.

Copy link
Member Author

Choose a reason for hiding this comment

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

This is called every time in the check macro when a task is added to make room for the next task.

When I originally wrote this code (before is_finished existed) I think I wanted to add rayon but someone said it's better to keep the number of dependencies low so it's faster when one doesn't have tidy compiled.

Copy link
Member

Choose a reason for hiding this comment

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

Hmm, the only place I see rayon mentioned in the original PR is #81833 (comment) - do you remember why you made that change?

Copy link
Member

Choose a reason for hiding this comment

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

ah, looks like #81833 (comment). Ok, we don't need to revisit that here.


while handles.len() >= concurrency.get() {
handles.pop_front().unwrap().join().unwrap();
}
};

scope(|s| {
let mut handles: VecDeque<ScopedJoinHandle<'_, ()>> =
VecDeque::with_capacity(concurrency.get());

macro_rules! check {
($p:ident $(, $args:expr)* ) => {
while handles.len() >= concurrency.get() {
handles.pop_front().unwrap().join().unwrap();
}
drain_handles(&mut handles);

let handle = s.spawn(|| {
let mut flag = false;
Expand Down Expand Up @@ -97,9 +108,8 @@ fn main() {
check!(alphabetical, &library_path);

let collected = {
while handles.len() >= concurrency.get() {
handles.pop_front().unwrap().join().unwrap();
}
drain_handles(&mut handles);

let mut flag = false;
let r = features::check(&src_path, &compiler_path, &library_path, &mut flag, verbose);
if flag {
Expand Down
59 changes: 35 additions & 24 deletions src/tools/tidy/src/style.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,7 @@
//! `// ignore-tidy-CHECK-NAME`.

use crate::walk::{filter_dirs, walk};
use regex::Regex;
use regex::{Regex, RegexSet};
use std::path::Path;

/// Error code markdown is restricted to 80 columns because they can be
Expand Down Expand Up @@ -225,6 +225,7 @@ pub fn check(path: &Path, bad: &mut bool) {
.chain(PROBLEMATIC_CONSTS.iter().map(|v| format!("{:x}", v)))
.chain(PROBLEMATIC_CONSTS.iter().map(|v| format!("{:X}", v)))
.collect();
let problematic_regex = RegexSet::new(problematic_consts_strings.as_slice()).unwrap();
walk(path, &mut skip, &mut |entry, contents| {
let file = entry.path();
let filename = file.file_name().unwrap().to_string_lossy();
Expand Down Expand Up @@ -281,7 +282,27 @@ pub fn check(path: &Path, bad: &mut bool) {
let mut trailing_new_lines = 0;
let mut lines = 0;
let mut last_safety_comment = false;
let is_test = file.components().any(|c| c.as_os_str() == "tests");
// scanning the whole file for multiple needles at once is more efficient than
// executing lines times needles separate searches.
let any_problematic_line = problematic_regex.is_match(contents);
for (i, line) in contents.split('\n').enumerate() {
if line.is_empty() {
if i == 0 {
leading_new_lines = true;
}
trailing_new_lines += 1;
continue;
} else {
trailing_new_lines = 0;
}

let trimmed = line.trim();

if !trimmed.starts_with("//") {
lines += 1;
}
Comment on lines +290 to +304
Copy link
Member

Choose a reason for hiding this comment

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

Why did this move? I don't see lines / trailing_new_lines / leading_new_lines used until after where it was before.

Copy link
Member Author

Choose a reason for hiding this comment

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

So we can continue earlier and skip attempting all the string processing stuff below on empty lines.


let mut err = |msg: &str| {
tidy_error!(bad, "{}:{}: {}", file.display(), i + 1, msg);
};
Expand All @@ -308,36 +329,38 @@ pub fn check(path: &Path, bad: &mut bool) {
suppressible_tidy_err!(err, skip_cr, "CR character");
}
if filename != "style.rs" {
if line.contains("TODO") {
if trimmed.contains("TODO") {
err("TODO is deprecated; use FIXME")
}
if line.contains("//") && line.contains(" XXX") {
if trimmed.contains("//") && trimmed.contains(" XXX") {
err("XXX is deprecated; use FIXME")
}
for s in problematic_consts_strings.iter() {
if line.contains(s) {
err("Don't use magic numbers that spell things (consider 0x12345678)");
if any_problematic_line {
for s in problematic_consts_strings.iter() {
if trimmed.contains(s) {
err("Don't use magic numbers that spell things (consider 0x12345678)");
}
}
}
}
let is_test = || file.components().any(|c| c.as_os_str() == "tests");
// for now we just check libcore
if line.contains("unsafe {") && !line.trim().starts_with("//") && !last_safety_comment {
if file.components().any(|c| c.as_os_str() == "core") && !is_test() {
if trimmed.contains("unsafe {") && !trimmed.starts_with("//") && !last_safety_comment {
if file.components().any(|c| c.as_os_str() == "core") && !is_test {
suppressible_tidy_err!(err, skip_undocumented_unsafe, "undocumented unsafe");
}
}
if line.contains("// SAFETY:") {
if trimmed.contains("// SAFETY:") {
last_safety_comment = true;
} else if line.trim().starts_with("//") || line.trim().is_empty() {
} else if trimmed.starts_with("//") || trimmed.is_empty() {
// keep previous value
} else {
last_safety_comment = false;
}
if (line.starts_with("// Copyright")
|| line.starts_with("# Copyright")
|| line.starts_with("Copyright"))
&& (line.contains("Rust Developers") || line.contains("Rust Project Developers"))
&& (trimmed.contains("Rust Developers")
|| trimmed.contains("Rust Project Developers"))
{
suppressible_tidy_err!(
err,
Expand All @@ -351,18 +374,6 @@ pub fn check(path: &Path, bad: &mut bool) {
if filename.ends_with(".cpp") && line.contains("llvm_unreachable") {
err(LLVM_UNREACHABLE_INFO);
}
if line.is_empty() {
if i == 0 {
leading_new_lines = true;
}
trailing_new_lines += 1;
} else {
trailing_new_lines = 0;
}

if !line.trim().starts_with("//") {
lines += 1;
}
}
if leading_new_lines {
let mut err = |_| {
Expand Down