Rustで正規表現エンジンっぽいのをアレした話

https://rust.connpass.com/event/48826/ でやったやつ

Transcript

1. RustͰVMܕਖ਼نදݱ

   ΤϯδϯͬΆ͍ͷΛ

   ΞϨͨ͠࿩
   KOBA789
   

   View Slide

2. KOBA789
   • େֶ΍ΊͯྉཧΛָ͠Έʹ͢ΔձࣾͰࣾһ

   ΍ͬͯΔ
   • ࣾͰ͸δϟόॻ͍ͯΔ͠಺༰ͱ͸ؔ܎ͳ͍
   • େੲʹNode.jsͷຊॻ͍ͨؾ͕͢Δ
   • Node.jsྲྀߦΒͤͯ͝ΊΜͳ͍͞
   

   View Slide

3. KOBA789
   ͪΐ͏Ͳ2೥લʹ৮ͬͯͨΒ͍͠
   

   View Slide

4. MeCab(w/ IPA dic)
   Ͱղੳͨ͠ܗଶૉྻʹ
   ύλʔϯϚον΍Γ͍ͨ
   KOBA789
   

   View Slide

5. KOBA789
   

   View Slide

6. 10ྻxNߦ
   N͸ઍສΦʔμʔ
   KOBA789
   

   View Slide

7. ϕΫτϧྻʹਖ਼نදݱ
   KOBA789
   

   View Slide

8. ͳΜͱͳ͘VMܕ
   KOBA789
   

   View Slide

9. Rust͍͍ͱ͜Ζ
   • ྻڍܕ͕·ͱ΋
   • ྻڍܕͷύλʔϯϚον͕·ͱ΋
   • ໢ཏੑνΣοΫ͸ࠓ࣌ৗࣝͩΑͶ
   • Unicodeจࣈྻͷѻ͍͕·ͱ΋
   • Ͱ΋ͬͯ଎͍
   KOBA789
   

   View Slide

10. ॻ͍ͨ
    KOBA789
    

    View Slide

11. χʔϞχοΫ
    Jump:2
    Fail
    Expect:1:໊ࢺ:1
    Expect:2:αม઀ଓ:1
    Next
    Expect:1:ಈࢺ:1
    Expect:5:αมɾεϧ:1
    Match:OK
    Noop
    KOBA789
    

    View Slide

12. ύϑΥʔϚϯε
    • Ϛγϯ: Xeon E3-1220@3.2GHz/16GB RAM
    • ର৅ϑΝΠϧ: 1.7GB 3400ສߦ
    • ࣮ߦ࣌ؒ: 56.18sec (͖ͬ͞ଌͬͨ)
    • 30.26 MB/sec
    KOBA789
    

    View Slide

13. νϡʔχϯά
    • Xcode ͷ Instruments Λ࢖͏ͱͳΜͱͳ͘

    ϓϩϑΝΠϧͰ͖Δ
    • εΫϦϓτݴޠͬΆ͘ࡶʹॻ͍ͨͷͰ

    jemalloc͕ͨ͘͞Μ
    KOBA789
    

    View Slide

14. ͦͷ͏ͪ
    • filebufferͱ͍͏crate͕͋ͬͯศརͦ͏
    • mmap ͬΆ͍͜ͱ͢Δ΍ͭ
    • zero(less) allocation Ͱ͍͍ײ͡ʹϑΝΠϧ

    ॲཧ͢ΔςΫχοΫ஌Γ͍ͨ
    KOBA789
    

    View Slide

15. ͔Μͦ͏
    • ࠷ॳ͸lifetimeʹ຋࿔͞Ε·ͬͨ͘
    • ϝϞϦͷؾ࣋ͪʹͳΔͱॻ͖΍͍͢
    • Boxͱ͔Arcͱ͔Ұճ΋࢖Θͳͯ͘΋

    ͳΜͱ͔ͳΔ
    • ฒߦॲཧ΋ωοτϫʔΫ΋΍ͬͯͶ͐
    • C++͋·Γॻ͔ͳ͍ͷͰmoveͱ͔Α͘Θ͔ͬͯͶ͐
    KOBA789
    

    View Slide