Auto merge of rust-lang#99487 - bmacnaughton:is_whitespace_updates, r=thomcc

is_whitespace() performance improvements

This is my first rust PR, so if I miss anything obvious please let me know and I'll do my best to fix it.

This was a bit more of a challenge than I realized because, while I made working code locally and tested it against the native `is_whitespace()`, this PR required changing `src/tools/unicode-table-generator`, the code that generated the code.

I have benchmarked this locally, using criterion, and have seen meaningful performance improvements. I can add those outputs to this if you'd like, but am guessing that the perf run that `@fmease` recommended is what's needed.

I have run ` ./x.py test --stage 0 library/std` after building it locally after executing `./x.py build library`. I didn't try to build the whole compiler, but maybe I should have - any guidance would be appreciated.

If this general approach makes sense, I'll take a look at some other candidate categories, e.g., `Cc`, in the future.

Oh, and I wasn't sure whether the generated code should be included in this PR or not. I did include it.

Author: bors

library/core/src/unicode/unicode_data.rs

@@ -544,18 +544,26 @@ pub mod uppercase {
 
 #[rustfmt::skip]
 pub mod white_space {
-    static SHORT_OFFSET_RUNS: [u32; 4] = [
-        5760, 18882560, 23080960, 40972289,
-    ];
-    static OFFSETS: [u8; 21] = [
-        9, 5, 18, 1, 100, 1, 26, 1, 0, 1, 0, 11, 29, 2, 5, 1, 47, 1, 0, 1, 0,
+    static WHITESPACE_MAP: [u8; 256] = [
+        2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 2, 2, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        0, 0, 0, 0, 0, 0, 0, 0, 0,
     ];
+    #[inline]
     pub fn lookup(c: char) -> bool {
-        super::skip_search(
-            c as u32,
-            &SHORT_OFFSET_RUNS,
-            &OFFSETS,
-        )
+        match c as u32 >> 8 {
+            0 => WHITESPACE_MAP[c as usize & 0xff] & 1 != 0,
+            22 => c as u32 == 0x1680,
+            32 => WHITESPACE_MAP[c as usize & 0xff] & 2 != 0,
+            48 => c as u32 == 0x3000,
+            _ => false,
+        }
     }
 }
 

src/tools/unicode-table-generator/src/cascading_map.rs

@@ -0,0 +1,78 @@
+use crate::fmt_list;
+use crate::raw_emitter::RawEmitter;
+use std::collections::HashMap;
+use std::fmt::Write as _;
+use std::ops::Range;
+
+impl RawEmitter {
+    pub fn emit_cascading_map(&mut self, ranges: &[Range<u32>]) -> bool {
+        let mut map: [u8; 256] = [
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+        ];
+
+        let points = ranges
+            .iter()
+            .flat_map(|r| (r.start..r.end).into_iter().collect::<Vec<u32>>())
+            .collect::<Vec<u32>>();
+
+        println!("there are {} points", points.len());
+
+        // how many distinct ranges need to be counted?
+        let mut codepoints_by_high_bytes = HashMap::<usize, Vec<u32>>::new();
+        for point in points {
+            // assert that there is no whitespace over the 0x3000 range.
+            assert!(point <= 0x3000, "the highest unicode whitespace value has changed");
+            let high_bytes = point as usize >> 8;
+            let codepoints = codepoints_by_high_bytes.entry(high_bytes).or_insert_with(Vec::new);
+            codepoints.push(point);
+        }
+
+        let mut bit_for_high_byte = 1u8;
+        let mut arms = Vec::<String>::new();
+
+        let mut high_bytes: Vec<usize> =
+            codepoints_by_high_bytes.keys().map(|k| k.clone()).collect();
+        high_bytes.sort();
+        for high_byte in high_bytes {
+            let codepoints = codepoints_by_high_bytes.get_mut(&high_byte).unwrap();
+            if codepoints.len() == 1 {
+                let ch = codepoints.pop().unwrap();
+                arms.push(format!("{} => c as u32 == {:#04x}", high_byte, ch));
+                continue;
+            }
+            // more than 1 codepoint in this arm
+            for codepoint in codepoints {
+                map[(*codepoint & 0xff) as usize] |= bit_for_high_byte;
+            }
+            arms.push(format!(
+                "{} => WHITESPACE_MAP[c as usize & 0xff] & {} != 0",
+                high_byte, bit_for_high_byte
+            ));
+            bit_for_high_byte <<= 1;
+        }
+
+        writeln!(&mut self.file, "static WHITESPACE_MAP: [u8; 256] = [{}];", fmt_list(map.iter()))
+            .unwrap();
+        self.bytes_used += 256;
+
+        writeln!(&mut self.file, "#[inline]").unwrap();
+        writeln!(&mut self.file, "pub fn lookup(c: char) -> bool {{").unwrap();
+        writeln!(&mut self.file, "    match c as u32 >> 8 {{").unwrap();
+        for arm in arms {
+            writeln!(&mut self.file, "        {},", arm).unwrap();
+        }
+        writeln!(&mut self.file, "        _ => false,").unwrap();
+        writeln!(&mut self.file, "    }}").unwrap();
+        writeln!(&mut self.file, "}}").unwrap();
+
+        true
+    }
+}

src/tools/unicode-table-generator/src/main.rs

@@ -75,12 +75,13 @@ use std::collections::{BTreeMap, HashMap};
 use std::ops::Range;
 use ucd_parse::Codepoints;
 
+mod cascading_map;
 mod case_mapping;
 mod raw_emitter;
 mod skiplist;
 mod unicode_download;
 
-use raw_emitter::{emit_codepoints, RawEmitter};
+use raw_emitter::{emit_codepoints, emit_whitespace, RawEmitter};
 
 static PROPERTIES: &[&str] = &[
     "Alphabetic",
@@ -241,8 +242,13 @@ fn main() {
     let mut modules = Vec::new();
     for (property, ranges) in ranges_by_property {
         let datapoints = ranges.iter().map(|r| r.end - r.start).sum::<u32>();
+
         let mut emitter = RawEmitter::new();
-        emit_codepoints(&mut emitter, &ranges);
+        if property == &"White_Space" {
+            emit_whitespace(&mut emitter, &ranges);
+        } else {
+            emit_codepoints(&mut emitter, &ranges);
+        }
 
         modules.push((property.to_lowercase().to_string(), emitter.file));
         println!(

src/tools/unicode-table-generator/src/raw_emitter.rs

@@ -170,6 +170,15 @@ pub fn emit_codepoints(emitter: &mut RawEmitter, ranges: &[Range<u32>]) {
     }
 }
 
+pub fn emit_whitespace(emitter: &mut RawEmitter, ranges: &[Range<u32>]) {
+    emitter.blank_line();
+
+    let mut cascading = emitter.clone();
+    cascading.emit_cascading_map(&ranges);
+    *emitter = cascading;
+    emitter.desc = String::from("cascading");
+}
+
 struct Canonicalized {
     canonical_words: Vec<u64>,
     canonicalized_words: Vec<(u8, u8)>,