Why is my rust program slower than ruby equivalent

2019-06-04 05:28发布

This question already has an answer here:

I wrote an anagram finder in Ruby and Rust and was very surprised to find that Rust program is almost 2 times slower than Ruby version.

Ruby version:

source = ARGV.first
sorted_source = source.chars.sort.join
anagrams = Hash.new
File.open('/usr/share/dict/words') do |f|
  f.each_line do |l|
    word = l.chomp
    sorted_word = word.chars.sort.join
    if anagrams[sorted_word]
      anagrams[sorted_word] << word
    else
      anagrams[sorted_word] = [word]
    end
  end
end
found = anagrams[sorted_source]
puts  found

Rust version:

use std::os;
use std::io::{File, BufferedReader};
use std::collections::HashMap;
fn main(){

    let path = Path::new("/usr/share/dict/words");
    match File::open(&path) {
        Err(e) => println!("Error opening file: {}", e.desc),
        Ok(f) => {
            let mut anagrams: HashMap<String, Vec<String>> = HashMap::new();
            let mut reader = BufferedReader::new(f);
            for maybe_line in reader.lines() {
                let word = maybe_line.unwrap().as_slice().trim_chars('\n').to_string();
                let mut chars: Vec<char> = word.as_slice().chars().collect();
                chars.sort();
                let sorted_word = String::from_chars(chars.as_slice());
                if anagrams.contains_key(&sorted_word) {
                    anagrams.get_mut(&sorted_word).push(word);
                } else {
                    anagrams.insert(sorted_word, vec!(word));
                }
            }

            let args = os::args();
            if args.len() == 2 {
                let source = args[1].clone();
                let mut chars: Vec<char> = source.as_slice().chars().collect();
                chars.sort();
                let sorted_word = String::from_chars(chars.as_slice());
                match anagrams.find(&sorted_word) {
                    Some(anagrams) => println!("{}", anagrams),
                    None => println!("No anagrams found")
                }
            } else {
                println!("Call the app with exactly 1 argument, the word to find anagrams for");
            }
        }
    }
}

Results:

 time ruby anagram.rb horse                                                                                                              
horse
shoer
shore
ruby anagram.rb horse  1.69s user 0.12s system 99% cpu 1.812 total



time ./anagram horse                                                                                                                    
[horse, shoer, shore]
./anagram horse  3.02s user 0.05s system 99% cpu 3.080 total

ruby -v
ruby 2.1.3p242 (2014-09-19 revision 47630) [x86_64-darwin13.0]

rustc --version
rustc 0.13.0-nightly (172b59abe 2014-10-25 00:32:07 +0000)

Ruby gist: https://gist.github.com/Valve/533e0e22ae427d9ce440

Rust gist: https://gist.github.com/Valve/834917941b00668478f2

UPDATE:

As Francis Gagne suggested, I compiled it with -O flag:

 time ./anagram horse                                                                                                                    
[horse, shoer, shore]
./anagram horse  0.37s user 0.05s system 96% cpu 0.429 total

This did a 8x increase in speed, but still only 4x times faster than ruby version. I guess this is a file system limitation now.

The size of the file is: 235886 lines on my machine.

1条回答
小情绪 Triste *
2楼-- · 2019-06-04 05:57

If you're using Cargo you can set the optimization higher than -O.

In your Cargo.toml.

[package]
name = "anagram"
version = "0.1.0"
authors = ["Your name here <Your email here>"]

[profile.release]
opt-level = 3

And run Cargo build --release. Once that is finished run time ./anagram horse

查看更多
登录 后发表回答