Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
The future of Ruby is faster
Search
Dirkjan Bussink
June 29, 2013
Technology
3
530
The future of Ruby is faster
Talk given at Euruko 2013
Dirkjan Bussink
June 29, 2013
Tweet
Share
More Decks by Dirkjan Bussink
See All by Dirkjan Bussink
Managing a widely distributed team
dbussink
1
170
Time
dbussink
0
94
The tricky truth about parallel execution and modern hardware
dbussink
0
330
Security for dummies
dbussink
1
140
The myth of dynamic language performance
dbussink
3
430
Rubinius - Tales from the trenches @ Railsclub.ru 2012
dbussink
2
200
Rubinius - Tales from the trenches @ Baruco 2012
dbussink
1
250
Rubinius Eurucamp 2012 Workshop
dbussink
2
110
Rubinius Euruko 2012 Lightning talk
dbussink
2
240
Other Decks in Technology
See All in Technology
完全自律型AIエージェントとAgentic Workflow〜ワークフロー構築という現実解
pharma_x_tech
0
350
Cloudflareで実現する AIエージェント ワークフロー基盤
kmd09
0
290
30分でわかる「リスクから学ぶKubernetesコンテナセキュリティ」/30min-k8s-container-sec
mochizuki875
3
450
データ基盤におけるIaCの重要性とその運用
mtpooh
4
540
[IBM TechXchange Dojo]Watson Discoveryとwatsonx.aiでRAGを実現!座学①
siyuanzh09
0
110
シフトライトなテスト活動を適切に行うことで、無理な開発をせず、過剰にテストせず、顧客をビックリさせないプロダクトを作り上げているお話 #RSGT2025 / Shift Right
nihonbuson
3
2.2k
Oracle Base Database Service:サービス概要のご紹介
oracle4engineer
PRO
1
16k
Copilotの力を実感!3ヶ月間の生成AI研修の試行錯誤&成功事例をご紹介。果たして得たものとは・・?
ktc_shiori
0
360
JuliaTokaiとJuliaLangJaの紹介 for NGK2025S
antimon2
1
130
JAWS-UG20250116_iOSアプリエンジニアがAWSreInventに行ってきた(真面目編)
totokit4
0
140
Building Scalable Backend Services with Firebase
wisdommatt
0
110
Azureの開発で辛いところ
re3turn
0
240
Featured
See All Featured
Faster Mobile Websites
deanohume
305
30k
Code Reviewing Like a Champion
maltzj
521
39k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
98
18k
Fireside Chat
paigeccino
34
3.1k
It's Worth the Effort
3n
183
28k
The Psychology of Web Performance [Beyond Tellerrand 2023]
tammyeverts
45
2.3k
The Cult of Friendly URLs
andyhume
78
6.1k
Done Done
chrislema
182
16k
The World Runs on Bad Software
bkeepers
PRO
66
11k
Adopting Sorbet at Scale
ufuk
74
9.2k
GitHub's CSS Performance
jonrohan
1030
460k
Bash Introduction
62gerente
610
210k
Transcript
The future of Ruby is faster Euruko 2013 Saturday, June
29, 13
Dirkjan Bussink @dbussink Saturday, June 29, 13
Rubinius Use Ruby Saturday, June 29, 13
Saturday, June 29, 13
Saturday, June 29, 13
“Everything you heard about Ruby being slow is not true.
It’s about twice as slow as that” Saturday, June 29, 13
Saturday, June 29, 13
The lure of the hash table Saturday, June 29, 13
class MethodTable def initialize @methods = {} end def store_method(name,
code) @methods[name] = code end def find_method(name) @methods[name] end end Saturday, June 29, 13
class Address def initialize @instance_variables[:street] = "Pantheon" @instance_variables[:number] = 1
@instance_variables[:city] = "Athens" end end Saturday, June 29, 13
Hash tables everywhere! Saturday, June 29, 13
Inline caching Saturday, June 29, 13
p = Person.new p.name Saturday, June 29, 13
p = Person.new p.name <receiver_class_id, code> Saturday, June 29, 13
“There are 2 hard problems in computer science: caching, naming,
and off-by-1 errors” Saturday, June 29, 13
Global serial number Saturday, June 29, 13
http://jamesgolick.com/2013/4/14/mris-method-caches.html https://charlie.bz/blog/things-that-clear-rubys-method-cache Saturday, June 29, 13
Per class serial number Saturday, June 29, 13
Probably in 2.1 Saturday, June 29, 13
Instance variable packing Saturday, June 29, 13
class Address def initialize @street = "Pantheon" @number = 1
@city = "Athens" end end Address.instance_variable_get("@seen_ivars") => [:@street, :@number, :@city] Saturday, June 29, 13
0001: push_ivar :@number self[1] Removes hash table lookup Saturday, June
29, 13
Just in time compilation Saturday, June 29, 13
def method1 1 + method2 end def method2 2 +
1 end 100000.times do method1 end Saturday, June 29, 13
def method1 1 + method2 end def method2 2 +
1 end 100000.times do method1 end ... 0x110ed90e7 mov $0x9, %eax 0x110ed90ec jmp 0x119 ; 0x110ed9129 ... 0x110ed9129 addq $0xa0, %rsp 0x110ed9130 pop %rbp 0x110ed9131 ret Saturday, June 29, 13
def method1 1 + method2 end def method2 2 +
1 end 100000.times do method1 end ... 0x110ed90e7 mov $0x9, %eax 0x110ed90ec jmp 0x119 ; 0x110ed9129 ... 0x110ed9129 addq $0xa0, %rsp 0x110ed9130 pop %rbp 0x110ed9131 ret b0100 (4) b1000 (8) b1001 (9) Saturday, June 29, 13
def method1 1 + method2 end def method2 2 +
1 end 100000.times do method1 end ... 0x110ed90e7 mov $0x9, %eax 0x110ed90ec jmp 0x119 ; 0x110ed9129 ... 0x110ed9129 addq $0xa0, %rsp 0x110ed9130 pop %rbp 0x110ed9131 ret b0100 (4) b1000 (8) b1001 (9) Saturday, June 29, 13
Saturday, June 29, 13
Creating less garbage Saturday, June 29, 13
a = Address.new a.street = "Pantheon" a.number = "1" a.city
= "Athens" Rubinius.memory_size(a) => 56 VS Rubinius.memory_size(a) => 160 Saturday, June 29, 13
Saturday, June 29, 13
2.1 Saturday, June 29, 13
Auto tuning Saturday, June 29, 13
Parallelism Saturday, June 29, 13
Higher throughput Saturday, June 29, 13
Concurrency Saturday, June 29, 13
Shorter stop the world Saturday, June 29, 13
thread 1 thread 2 thread 3 Saturday, June 29, 13
thread 1 thread 2 thread 3 GC thread Saturday, June
29, 13
thread 1 thread 2 thread 3 GC thread Long GC
pauses be gone! Saturday, June 29, 13
http://rubini.us/2013/06/22/concurrent-garbage-collection/ Saturday, June 29, 13
How much does it help? Saturday, June 29, 13
Before Saturday, June 29, 13
Lifting the server siege... done. Transactions: 501 hits Response time:
0.02 secs Transaction rate: 51.70 trans/sec Throughput: 1.46 MB/sec Concurrency: 1.00 Longest transaction: 0.15 Shortest transaction: 0.01 Saturday, June 29, 13
Lifting the server siege... done. Transactions: 501 hits Response time:
0.02 secs Transaction rate: 51.70 trans/sec Throughput: 1.46 MB/sec Concurrency: 1.00 Longest transaction: 0.15 Shortest transaction: 0.01 Sad long wait time :( Saturday, June 29, 13
Lifting the server siege... done. Transactions: 1032 hits Response time:
0.04 secs Transaction rate: 102.79 trans/sec Throughput: 2.91 MB/sec Concurrency: 4.00 Longest transaction: 0.28 Shortest transaction: 0.02 Saturday, June 29, 13
Lifting the server siege... done. Transactions: 1032 hits Response time:
0.04 secs Transaction rate: 102.79 trans/sec Throughput: 2.91 MB/sec Concurrency: 4.00 Longest transaction: 0.28 Shortest transaction: 0.02 Not so good concurrency Saturday, June 29, 13
After Saturday, June 29, 13
Lifting the server siege... done. Transactions: 599 hits Response time:
0.02 secs Transaction rate: 61.06 trans/sec Throughput: 1.73 MB/sec Concurrency: 1.00 Longest transaction: 0.03 Shortest transaction: 0.01 Saturday, June 29, 13
Lifting the server siege... done. Transactions: 599 hits Response time:
0.02 secs Transaction rate: 61.06 trans/sec Throughput: 1.73 MB/sec Concurrency: 1.00 Longest transaction: 0.03 Shortest transaction: 0.01 No more long pause! Saturday, June 29, 13
Lifting the server siege... done. Transactions: 1374 hits Response time:
0.02 secs Transaction rate: 176.38 trans/sec Throughput: 5.00 MB/sec Concurrency: 3.99 Longest transaction: 0.04 Shortest transaction: 0.01 Saturday, June 29, 13
Lifting the server siege... done. Transactions: 1374 hits Response time:
0.02 secs Transaction rate: 176.38 trans/sec Throughput: 5.00 MB/sec Concurrency: 3.99 Longest transaction: 0.04 Shortest transaction: 0.01 More req/s! Saturday, June 29, 13
1 client: 51 => 61 req/s 4 clients: 102 =>
176 req/s Saturday, June 29, 13
Concurrency & parallelism Saturday, June 29, 13
No GIL/GVL! Saturday, June 29, 13
Future is multi-core Saturday, June 29, 13
We need new API’s Saturday, June 29, 13
Parallel each Saturday, June 29, 13
Thread safe collections Saturday, June 29, 13
What about you as a developer? Saturday, June 29, 13
Be nice Saturday, June 29, 13
Write type stable code Saturday, June 29, 13
Small and simple methods Saturday, June 29, 13
Benchmark! Saturday, June 29, 13
? Saturday, June 29, 13