Optimize indentation in the pretty printer.
Currently the pretty-printer calls `write!` for every space of indentation. On some workloads the indentation level can exceed 100, and a faster implementation reduces instruction counts by up to 7% on a few workloads.
This commit is contained in:
parent
4c27fb19ba
commit
606f3158bf
@ -300,6 +300,8 @@ impl Default for BufEntry {
|
||||
}
|
||||
}
|
||||
|
||||
const SPACES: [u8; 128] = [b' '; 128];
|
||||
|
||||
impl<'a> Printer<'a> {
|
||||
pub fn last_token(&mut self) -> Token {
|
||||
self.buf[self.right].token.clone()
|
||||
@ -580,10 +582,24 @@ impl<'a> Printer<'a> {
|
||||
debug!("print String({})", s);
|
||||
// assert!(len <= space);
|
||||
self.space -= len;
|
||||
while self.pending_indentation > 0 {
|
||||
write!(self.out, " ")?;
|
||||
self.pending_indentation -= 1;
|
||||
|
||||
// Write the pending indent. A more concise way of doing this would be:
|
||||
//
|
||||
// write!(self.out, "{: >n$}", "", n = self.pending_indentation as usize)?;
|
||||
//
|
||||
// But that is significantly slower than using `SPACES`. This code is
|
||||
// sufficiently hot, and indents can get sufficiently large, that the
|
||||
// difference is significant on some workloads.
|
||||
let spaces_len = SPACES.len() as isize;
|
||||
while self.pending_indentation >= spaces_len {
|
||||
self.out.write_all(&SPACES)?;
|
||||
self.pending_indentation -= spaces_len;
|
||||
}
|
||||
if self.pending_indentation > 0 {
|
||||
self.out.write_all(&SPACES[0..self.pending_indentation as usize])?;
|
||||
self.pending_indentation = 0;
|
||||
}
|
||||
|
||||
write!(self.out, "{}", s)
|
||||
}
|
||||
|
||||
|
Loading…
Reference in New Issue
Block a user