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

Jump to bottom

Avoid unnecessary allocations in float_lit and integer_lit. #55384

Merged
merged 1 commit into from
Oct 29, 2018
Merged

Avoid unnecessary allocations in float_lit and integer_lit. #55384

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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/libsyntax/parse/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -494,8 +494,17 @@ fn float_lit(s: &str, suffix: Option<Symbol>, diag: Option<(Span, &Handler)>)
-> Option<ast::LitKind> {
debug!("float_lit: {:?}, {:?}", s, suffix);
// FIXME #2252: bounds checking float literals is deferred until trans
let s = s.chars().filter(|&c| c != '_').collect::<String>();
filtered_float_lit(Symbol::intern(&s), suffix, diag)

// Strip underscores without allocating a new String unless necessary.
let s2;
let s = if s.chars().any(|c| c == '_') {
Copy link
Member

Choose a reason for hiding this comment

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

It would be interesting to see if std::slice::memchr::memchr(b'_', s.as_bytes()).is_some() is faster than s.chars().any().

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I tried it. Cachegrind says it's marginally more instructions: 22,526,559,505 up from 22,516,683,388. So I think I'll stick with the original.

s2 = s.chars().filter(|&c| c != '_').collect::<String>();
Copy link
Member

Choose a reason for hiding this comment

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

Hm, since underscores are in ASCII, couldn't we into_bytes, retain, and then String::from_utf8, maybe unchecked?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

That would work, but it's not clear to me that it would be any faster; it might even be slower because there are two allocations involved? (Because it converts to a vector, and then back to a new String?)

Besides, this is the cold path, so I'm satisfied with reusing the existing code, which until now had been considered good enough for the hot path :)

Copy link
Member

Choose a reason for hiding this comment

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

Just FYI, String::from_utf8 consumes the vector given to it, so no additional allocation would happen.

&s2
} else {
s
};

filtered_float_lit(Symbol::intern(s), suffix, diag)
}

/// Parse a string representing a byte literal into its final form. Similar to `char_lit`
Expand Down Expand Up @@ -591,8 +600,14 @@ fn integer_lit(s: &str, suffix: Option<Symbol>, diag: Option<(Span, &Handler)>)
-> Option<ast::LitKind> {
// s can only be ascii, byte indexing is fine

let s2 = s.chars().filter(|&c| c != '_').collect::<String>();
let mut s = &s2[..];
// Strip underscores without allocating a new String unless necessary.
let s2;
let mut s = if s.chars().any(|c| c == '_') {
s2 = s.chars().filter(|&c| c != '_').collect::<String>();
&s2
} else {
s
};

debug!("integer_lit: {}, {:?}", s, suffix);

Expand Down