Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Speaker Deck
PRO
Sign in
Sign up for free
Ruby 3.0 JIT on Rails
Takashi Kokubun
July 03, 2020
Programming
8
8.1k
Ruby 3.0 JIT on Rails
https://ginza-rails.connpass.com/event/178203/
Takashi Kokubun
July 03, 2020
Tweet
Share
More Decks by Takashi Kokubun
See All by Takashi Kokubun
Optimizing Production Performance with MRI JIT / RubyConf 2021
k0kubun
1
130
Why Ruby's JIT was slow / RubyKaigi Takeout 2021
k0kubun
2
1.3k
数時間かかる週一リリースを毎日何度も爆速でできるようにするまで / CI/CD Conference 2021
k0kubun
20
13k
Ruby 3 JIT's roadmap / RubyConf China 2020
k0kubun
0
440
JIT ロードマップ / Ruby 3 さみっと
k0kubun
2
960
JIT compiler improvements in Ruby 2.7 / RubyRussia 2019
k0kubun
0
640
Optimizing Ruby with JIT
k0kubun
13
11k
Ruby 2.7 JIT on Rails / RailsConf 2019
k0kubun
1
610
Performance Improvement of Ruby 2.7 JIT in Real World / RubyKaigi 2019
k0kubun
8
18k
Other Decks in Programming
See All in Programming
Register-based calling convention for Go functions
cjamhe01385
0
390
SwiftUIで「意図」を伝える / swiftui_intention
uhooi
2
140
Enzyme から React Native Testing Library に移行した経緯 / 2022-07-20
tamago3keran
1
160
僕が便利だと感じる Snow Monkey の特徴/20220723_Gifu_WordPress_Meetup
oleindesign
0
100
SRE NEXT 2022に学ぶこれからのSREキャリア
fukubaka0825
2
390
Lookerとdbtの共存
ttccddtoki
0
620
Haskellでオブジェクト指向プログラミング
koheisakata
0
180
NestJS_meetup_atamaplus
atamaplus
0
200
リーダブルテストコード / #vstat
jnchito
46
34k
Reactは何を提供するLibraryなのか?
taro28
6
910
ESM移行は無理だけどおれもSindreのライブラリが使いたい!
sosukesuzuki
2
530
設計の考え方とやり方
masuda220
PRO
46
26k
Featured
See All Featured
Code Review Best Practice
trishagee
44
9.7k
Intergalactic Javascript Robots from Outer Space
tanoku
260
25k
BBQ
matthewcrist
74
7.9k
StorybookのUI Testing Handbookを読んだ
zakiyama
5
2.5k
Become a Pro
speakerdeck
PRO
3
900
GraphQLとの向き合い方2022年版
quramy
16
8.4k
Git: the NoSQL Database
bkeepers
PRO
415
59k
Building an army of robots
kneath
299
40k
The MySQL Ecosystem @ GitHub 2015
samlambert
239
11k
jQuery: Nuts, Bolts and Bling
dougneiner
56
6.4k
How STYLIGHT went responsive
nonsquared
85
4k
Designing for Performance
lara
597
63k
Transcript
Ruby 3.0 JIT on Rails ۜ࠲Rails#23 Takashi Kokubun / @k0kubun
ࣗݾհ • GitHub, Twitter: @k0kubun • Arm Treasure Data •
Backend, SRE, API • Ruby ίϛολ • JIT, VM, ERB, Πϯϑϥ, ϕϯνϚʔΫ
None
એ
એ
એ
ࠓ͢͜ͱ • Ruby 3.0 JIT ͷݱࡏͷੑೳ • Rails ͰͷϘτϧωοΫͷվળ •
C Λ Ruby ʹॻ͖͑ͯΔ • ࠓޙͷ JIT ΞʔΩςΫνϟ
Ruby 3.0 JIT ͷݱࡏͷੑೳ
mame/optcarrot 3VCZ 3VCZ 3VCZ GSBNFTTFD
7. +*5
benchmark-driver/sinatra 3VCZ 3VCZ 3VCZ SFRVFTUTTFD
7. +*5
k0kubun/railsbench 3VCZ 3VCZ 3VCZ SFRVFTUTTFD
7. +*5
͜Ε·Ͱͷੑೳվળͷ֓ཁ • 2.6: Optcarrot্ݱঢ়͜ͷόʔδϣϯͷΈɻJIT compactionRailsʹޮ͘ɻ • 2.7: Ұ෦ϝιουͷΠϯϥΠϯԽɻmax cacheσϑΥϧτมߋҰԠRailsʹޮ͘ɻ •
3.0: builtinϝιουͷΠϯϥΠϯԽɻRails͚ͷ࠷దԽ͕͍͔ͭ͘ (͜Ε͔Βઆ໌)
Rails ͰͷϘτϧωοΫͷվળ
Optcarrot ͱ Sinatra, Rails ͷҧ͍? • طͷࣄ࣮: Optcarrot͕͘Ͱ͖ΔͷΠϯελϯεมͷ࠷దԽ͕ Α͘ޮͨ͘Ί •
͔͠͠ɺSinatraRails͕͘ͳΔͷԿނ͔? • CPUͷͲͷϨΠϠʔͰϘτϧωοΫʹͳ͍ͬͯΔ?
VTune: mame/optcarrot - VM • VTune: VM, JIT
VTune: mame/optcarrot - JIT • VTune: VM, JIT
VTune: benchmark-driver/sinatra - VM
VTune: benchmark-driver/sinatra - JIT
VTune: k0kubun/railsbench - VM
VTune: k0kubun/railsbench - JIT
ϑϩϯτΤϯυͷׂ߹͕૿͍͑ͯΔ • ໋ྩͷ࣮ߦͰͳ͘ɺ໋ྩͷfetchͰstall͍ͯ͠Δ • CPU 1αΠΫϧ͋ͨΓʹ࣮ߦ͞ΕΔ໋ྩ͕গͳ͍ • VMͷ࣌ͰOptcarrotͱ͕ࠩ͋Δ͕ɺJITͰѱԽ͍ͯ͠Δͷ ICache Miss
Instruction Cache (ICache) Miss ͱ • Α͋͘ΔCPUͷΩϟογϡ: L1i, L1d, L2,
L3 (shared) • L1i: Level 1 Instruction Ωϟογϡ • CPU໋ྩͷΩϟογϡϛε • ಉ࣌ʹΞΫηε͢Δίʔυ͕ผͷcachelineϖʔδʹ͋ΔͳͲ (ίʔυͷαΠζ͔ϨΠΞτͷ)
ICache Miss ΛͲ͏ճආ͢Δ͔ • ίʔυαΠζͷݮ • ॏෳΛഉআ͢Δ • ࣮ߦ࣌ใʹج͍ͮͯෆཁͳύεΛΔ •
ϨΠΞτͷվળ • ϝιου͝ͱʹผϖʔδʹόϥόϥͳͷ2.6 JIT compactionͰղܾ • hot path ͷہॴੑΛ্͛Δ
ॏෳΛഉআ͢Δ • 3.0 ʹೖͬͨ࠷దԽ: • JIT compactionͰɺ.o Ͱͳ͘.c͔ΒίϯύΠϧ͢͠ (static͕ؔ dedup͞ΕΔ)
• ivar͕࠷దԽ͞Εͳ͔ͬͨ࣌ʹΘΕΔؔͷΠϯϥΠϯԽΛΊΔ • JIT cancel࣌ͷෳͷؔݺͼग़͠Λ1ͭʹ·ͱΊΔ
࣮ߦ࣌ใʹج͍ͮͯෆཁͳύεΛΔ • 3.0 ʹೖͬͨ࠷దԽ: • ϝιουݺͼग़͠ͰϨγʔόͷΫϥε͕ඇଈͳΒɺଈ͚ͷ ذΛআڈ͢Δ
hot path ͷہॴੑΛ্͛Δ • 3.0 ʹೖͬͨ࠷దԽ: • ϝιου୳ࡧͳͲͷύεʹcold attributeΛ͚ͭΔ •
stackoverflow࣌coldԽ • JITίϯύΠϧͷenqueuecoldԽ
͋ͱͲ͏͢Δ͔ • ·ͩ͘ͳΔݪҼʹͳͬͯͦ͏ͩ͠ɺॏෳ͋Γͦ͏ • ͔͠͠୯ʹΠϯϥΠϯԽΛແޮԽ͢ΔͱͦΕͰ͘ͳΔ͜ͱ͋Δ ͷͰɺ͜ΕҎ্৻ॏʹΔඞཁ͕͋Δ • ϝιου͋ͨΓͷίʔυαΠζ͕ҰؾʹݮΕɺJIT͍͍ͯ͠࠷େϝ ιου૿ͤΔ͔
C Λ Ruby ʹॻ͖͑ͯΔ
RubyKaigi 2018
RubyKaigi 2019
None
None
Builtin ϝιουͱ • RubyίΞͷ࣮Ͱ͔͑͠ͳ͍ػೳ • ISeqͷόΠφϦঢ়ଶͰอଘ͞Ε͓ͯΓɺϩʔυίετ͕͍ • ΩʔϫʔυҾΛऔΔϝιουɺRubyͰ࣮ͨ͠ํ͕ߴ • RubyͰϝιουΛఆٛ͢Δ͕ɺCͷ͕ؔίετͰݺͼग़ͤΔ
Builtin ϝιουͱ
Builtin attr ͱ • LLVMͳͲͳ͠ͰCͷؔͷڍಈΛղੳ͢Δͷ͍͠ • ͦ͜ͰɺϝιουͷಛੑΛΞϊςʔγϣϯ͓͖ͯ͠ɺJITͦΕΛ৴͡Δ • RubyͷCIͳͲͰΞϊςʔγϣϯ͕ਖ਼͍͔࣮͠ߦ࣌ʹݕ͍ࠪͯ͠Δ
Builtin attr ͷҰཡ • inline: ϝιουϑϨʔϜΛpush͠ͳͯ͘ྑ͍ • ݱঢ়͜Ε͚ͩɻJIT͕ϝιουϑϨʔϜͷpushΛεΩοϓ͍ͯ͠Δ
Builtin attr inline Λ͚ͭΔྫ
ϝιουϑϨʔϜεΩοϓͷޮՌ ,FSOFMDMBTT JT 7. +*5
+*5 GSBNFTLJQ 1.3x 1.7x
ϝιουϑϨʔϜεΩοϓʹҙຯ͋Δͷ͔? • Railsϝιουݺͼग़͕͠ଟͦ͏ • ͨͩ͠ɺ͋Δϝιου͋ͨΓͷݺͼग़͠ճͦΕ΄Ͳଟ͘ͳ͘ɺ ෆར • Cͷؔఆٛ͝ͱΠϯϥΠϯԽͨ࣌͠ʹ࠷దԽ࣌ͷϊΠζ͕ݮΔ • 1ճ͋ͨΓ
4ns / 16 clocks ఔޮՌ͕ݟΒΕΔ
ϝιουϑϨʔϜεΩοϓͷ݅ • ϝιουݺͼग़͠Λ͠ͳ͍ • ྫ֎Λ͛ͳ͍ • TypeError, ArgumentError, NoMemoryError ֘͢Δ
• όοΫτϨʔεΛݟͳ͍
ࠓޙͷ JIT ΞʔΩςΫνϟ
C2 Graal ͕ѻ͍ͬͯΔάϥϑ
MJIT͕͍ͬͯΔ͜ͱ • VMಛ༗ͷΦʔόʔϔουͷআڈ • VM໋ྩ୯ҐͰͷ࣮ߦ࣌ใΛͬͨ࠷దԽ • ϝιουͷΠϯϥΠϯԽ • Γͷ࠷దԽCίϯύΠϥͤ
JITʹظ͢Δ͜ͱ • ίϯύΠϧ࣌ܭࢉ • Constant folding • Dead code elimination
• Loop invariant motion • ϝϞϦΞΫηεͷ࠷దԽ
MJIT ͷݶք • CίϯύΠϥ Ruby VM ಛ༗ͷࣄΛߟྀͰ͖ͳ͍ • ΠϯϥΠϯԽͷൣғʹݶք͕͋Δ (C
ϝιου) • Γ͗͢ΔͱίʔυαΠζ͕૿͑ٯޮՌ • ਖ਼ৗܥΛ༏ઌͯ͠࠷దԽͤ͞Δͱ͍ͬͨࢦ͕͍ࣔ͠
ͱ͍͑ • JIT ແޮͷ࣌ʹ͘ͳΔखஈऔΓͨ͘ͳ͍ • JITͰ͍ͷ؆୯ͳίϚϯυͳͲͰศར • ϝϯς͕େมʹͳΔखஈऔΓͨ͘ͳ͍ • VM໋ྩ͕૿͑ͯجຊతʹJITಈ͖ଓ͚ͯཉ͍͠
• VMͷڍಈͱશ͘ಠཱͨ͠ίʔυੜͨ͘͠ͳ͍
͡Ό͋Ͳ͏͢Δ͔ 1. ίϯύΠϧ࣌ܭࢉʹඞཁͳͷΛἧ͑Δ 2. ୯७ͳಠࣗIR͔ΒVMʹ࠷దԽ͢Δج൫Λ͑Δ 3. ΑΓߴͳ࠷దԽʹదͨ͠IRߏʹม͑Δ
1. ίϯύΠϧ࣌ܭࢉʹඞཁͳͷΛἧ͑Δ • CϝιουͷΞϊςʔγϣϯΛ૿͢ (Τεέʔϓ͢Δ͔ɺ७ਮ͔) • ࣮ߦ࣌ͷܕใ͕ΑΓଟ͘ͷॴͰऔΕΔΑ͏ʹ͢Δ (opt໋ྩͳͲ) • Cϝιου,
super, yield, ఆ ͳͲͷΠϯϥΠϯԽରԠΛਐΊΔ
2. ୯७ͳಠࣗIR͔ΒVMʹ࠷దԽ͢Δج൫ • YARV໋ྩΛJITಠࣗͷIRʹղՄೳʹ͠ɺ໋ྩ/IRྻʹ࠷దԽΛ͔͚Δ • ࠷దԽͨ͠ޙͷ໋ྩ/IRྻͰ࠷దԽ͕ൃੜ͢Δҙͷॴ͔ΒVM ࣮ߦʹͤΔঢ়ଶΛҡ࣋͢Δ • On-Stack Replacement
(JIT → VM) ͦͷ͏࣮ͪ
3. ΑΓߴͳ࠷దԽʹదͨ͠IRߏʹม͑Δ • ࠷దԽͷج൫͕ͬͨΒɺͦΕΛҡ࣋͠ͳ͕Β࠷దԽʹదͨ͠IRά ϥϑʹม͍͑ͯ͘ • ίϯύΠϥRuby (on Ractor) Ͱ࣮͍ͨ͠
ߴͳ࠷దԽͷྫ (ࠓCͰॻ͍͍ͯΔ) • ෳ໋ྩؒͰॏෳ͍ͯ͠ΔܕνΣοΫͷϚʔδ (ࠓॻ͍ͨ) • Optcarrot fps: 83.3 →
93.0
ଞʹΓ͍ͨͱࢥ͍ͬͯΔ͜ͱ • ෭࡞༻ͷͳ͍ϝιουͷίϯύΠϧ࣌ݺͼग़͠ • [1, 2].min ͰΞϊςʔγϣϯʹج͍ͯ rb_ary_min ΛίϯύΠϧ࣌ʹݺͼग़͢ •
Τεέʔϓ͠ͳ͍ΦϒδΣΫτͷࣄલׂΓͯ • ҎલStringΛελοΫׂ͢ΔPoCΛ࡞͕ͬͨɺͦͦΞϩέʔγϣϯ͕ͳ ͯ͘ࡁΉ߹ͦΕ͕࠷ • Frozen string literal ͳͲΛෆཁʹ͍ͨ͠
·ͱΊ • ࠓJITʹΑΔ࠷దԽ͕ࠔͳͱ͜ΖͷճΓΛ͍͑ͯΔஈ֊ • Rails͕͘Ͱ͖Δ·ͰͷಓͷΓ͍͕ɺ3.0Ͱண࣮ʹਐḿ͍ͯ͠Δ