Threads Aren't Evil - Speaker Deck

Slide 1

Slide 1 text

Threads Aren’t Evil @schneems

Slide 2

Slide 2 text

Threads

Slide 3

Slide 3 text

Are

Slide 4

Slide 4 text

Not

Slide 5

Slide 5 text

Evil

Slide 6

Slide 6 text

No content

Slide 7

Slide 7 text

No content

Slide 8

Slide 8 text

“Ruby doesn’t have real threads”

Slide 9

Slide 9 text

“Ruby doesn’t have real threads”

Slide 10

Slide 10 text

“Ruby doesn’t have real threads”

Slide 11

Slide 11 text

“Ruby is incapable of using threads because of the GIL

Slide 12

Slide 12 text

“Ruby is incapable of using threads because of the GIL

Slide 13

Slide 13 text

“Ruby is incapable of using threads because of the GIL

Slide 14

Slide 14 text

"Rails is not able to run on threads so we can't use them in Ruby"

Slide 15

Slide 15 text

"Rails is not able to run on threads so we can't use them in Ruby"

Slide 16

Slide 16 text

"My ruby code is not thread safe and will not run in threads"

Slide 17

Slide 17 text

What do you hear?

Slide 18

Slide 18 text

“doesn’t”

Slide 19

Slide 19 text

“incapable”

Slide 20

Slide 20 text

“not able”

Slide 21

Slide 21 text

“will not”

Slide 22

Slide 22 text

2012

Slide 23

Slide 23 text

2013

Slide 24

Slide 24 text

2013

Slide 25

Slide 25 text

2013

Slide 26

Slide 26 text

2013

Slide 27

Slide 27 text

“doesn’t”

Slide 28

Slide 28 text

“incapable”

Slide 29

Slide 29 text

“not able”

Slide 30

Slide 30 text

“will not”

Slide 31

Slide 31 text

No content

Slide 32

Slide 32 text

1490

Slide 33

Slide 33 text

No content

Slide 34

Slide 34 text

No content

Slide 35

Slide 35 text

No content

Slide 36

Slide 36 text

1490

Slide 37

Slide 37 text

1490

Slide 38

Slide 38 text

1490

Slide 39

Slide 39 text

1490

Slide 40

Slide 40 text

1490

Slide 41

Slide 41 text

1490

Slide 42

Slide 42 text

1490

Slide 43

Slide 43 text

No content

Slide 44

Slide 44 text

цинга

Slide 45

Slide 45 text

цинга

Slide 46

Slide 46 text

цинга

Slide 47

Slide 47 text

цинга

Slide 48

Slide 48 text

No content

Slide 49

Slide 49 text

No content

Slide 50

Slide 50 text

No content

Slide 51

Slide 51 text

No content

Slide 52

Slide 52 text

No content

Slide 53

Slide 53 text

No content

Slide 54

Slide 54 text

No content

Slide 55

Slide 55 text

No content

Slide 56

Slide 56 text

Problem Solved?

Slide 57

Slide 57 text

No content

Slide 58

Slide 58 text

No content

Slide 59

Slide 59 text

No content

Slide 60

Slide 60 text

No content

Slide 61

Slide 61 text

No content

Slide 62

Slide 62 text

No content

Slide 63

Slide 63 text

No content

Slide 64

Slide 64 text

No content

Slide 65

Slide 65 text

The Future

Slide 66

Slide 66 text

BTW

Slide 67

Slide 67 text

Find me @schneems

Slide 68

Slide 68 text

https:// www. schneems .com

Slide 69

Slide 69 text

No content

Slide 70

Slide 70 text

No content

Slide 71

Slide 71 text

Who here knows what a thread is?

Slide 72

Slide 72 text

Process Address Space

Slide 73

Slide 73 text

Process Address Space - code

Slide 74

Slide 74 text

Process Address Space - code - data

Slide 75

Slide 75 text

Process Address Space - code - data - register

Slide 76

Slide 76 text

Process Address Space - code - data - register - stack

Slide 77

Slide 77 text

Process Address Space - code - data - register - stack

Slide 78

Slide 78 text

#include int increment(int x) { return x + 1; } int main() { int i = 0; printf("Incremented to: %i\n", increment(i)) }

Slide 79

Slide 79 text

_main: 100000f40: 55 pushq %rbp 100000f41: 48 89 e5 movq %rsp, %rbp 100000f44: 48 83 ec 10 subq $16, %rsp 100000f48: c7 45 fc 00 00 00 00 movl $0, -4(%rbp) 100000f4f: c7 45 f8 00 00 00 00 movl $0, -8(%rbp) 100000f56: 83 7d f8 0a cmpl $10, -8(%rbp) 100000f5a: 0f 8d 1f 00 00 00 jge 31 <_main+3F> 100000f60: 48 8d 3d 43 00 00 00 leaq 67(%rip), %rdi 100000f67: b0 00 movb $0, %al 100000f69: e8 1a 00 00 00 callq 26 100000f6e: 89 45 f4 movl %eax, -12(%rbp) 100000f71: 8b 45 f8 movl -8(%rbp), %eax 100000f74: 83 c0 01 addl $1, %eax 100000f77: 89 45 f8 movl %eax, -8(%rbp) 100000f7a: e9 d7 ff ff ff jmp -41 <_main+16> 100000f7f: 8b 45 fc movl -4(%rbp), %eax 100000f82: 48 83 c4 10 addq $16, %rsp 100000f86: 5d popq %rbp 100000f87: c3 retq

Slide 80

Slide 80 text

Slide 81

Slide 81 text

Slide 82

Slide 82 text

Slide 83

Slide 83 text

#include int increment(int x) { return x + 1; } int main() { int i = 0; printf("Incremented to: %i\n", increment(i)) }

Slide 84

Slide 84 text

#include int increment(int x) { return x + 1; } int main() { int i = 0; printf("Incremented to: %i\n", increment(i)) }

Slide 85

Slide 85 text

#include int increment(int x) { return x + 1; } int main() { int i = 0; printf("Incremented to: %i\n", increment(i)) }

Slide 86

Slide 86 text

#include int increment(int x) { return x + 1; } int main() { int i = 0; printf("Incremented to: %i\n", increment(i)) }

Slide 87

Slide 87 text

#include int increment(int x) { return x + 1; } int main() { int i = 0; printf("Incremented to: %i\n", increment(i)) }

Slide 88

Slide 88 text

Thanks to Julia Evans For the example code https://jvns.ca

Slide 89

Slide 89 text

Process Address Space - code - data - register - stack

Slide 90

Slide 90 text

Process Address Space - code - data - register - stack

Slide 91

Slide 91 text

Process Control Block Process State Process Number Program Counter Registers Memory Limits List of open Files Signal Mask CPU Scheduling info (PCB)

Slide 92

Slide 92 text

No content

Slide 93

Slide 93 text

Process - code - data - register - stack

Slide 94

Slide 94 text

Process - code - data - register - stack CPU

Slide 95

Slide 95 text

Process - code - data - register - stack Input Output (IO)

Slide 96

Slide 96 text

Process - code - data - register - stack Input Output (IO)

Slide 97

Slide 97 text

Process - code - data - register - stack Input Output (IO)

Slide 98

Slide 98 text

No content

Slide 99

Slide 99 text

No content

Slide 100

Slide 100 text

No content

Slide 101

Slide 101 text

Pros?

Slide 102

Slide 102 text

Better CPU utilization

Slide 103

Slide 103 text

Cons?

Slide 104

Slide 104 text

Need lots of processes

Slide 105

Slide 105 text

Processes take up memory which is a finite resource

Slide 106

Slide 106 text

Loading and unloading a PCB is expensive

Slide 107

Slide 107 text

Sharing data between processes is HARD

Slide 108

Slide 108 text

No content

Slide 109

Slide 109 text

No content

Slide 110

Slide 110 text

No content

Slide 111

Slide 111 text

Introducing

Slide 112

Slide 112 text

Threads

Slide 113

Slide 113 text

Process - code - data - register - stack

Slide 114

Slide 114 text

Process - code - data - register - stack

Slide 115

Slide 115 text

Process - code - data - register - stack

Slide 116

Slide 116 text

Process register stack ⚡

Slide 117

Slide 117 text

Process register stack ⚡ ⚡ ⚡ register stack register stack

Slide 118

Slide 118 text

Pros?

Slide 119

Slide 119 text

Better CPU utilization

Slide 120

Slide 120 text

Reuse existing process memory (code and data)

Slide 121

Slide 121 text

Smaller Context Switch Time

Slide 122

Slide 122 text

Cons?

Slide 123

Slide 123 text

Shared code + shared data =

Slide 124

Slide 124 text

Shared code + shared data =

Slide 125

Slide 125 text

Threads are simple

Slide 126

Slide 126 text

Threads are hard

Slide 127

Slide 127 text

In every language *

Slide 128

Slide 128 text

It’s possible to avoid threads

Slide 129

Slide 129 text

Javascript is single threaded

Slide 130

Slide 130 text

Threading is easier in some languages

Slide 131

Slide 131 text

Functional languages (can) make threading easy

Slide 132

Slide 132 text

Functional languages make side effects impossible

Slide 133

Slide 133 text

Which makes coding harder (IMHO)

Slide 134

Slide 134 text

OMSCS

Slide 135

Slide 135 text

Threading sucks (less) in Ruby

Slide 136

Slide 136 text

Are threads just for performance?

Slide 137

Slide 137 text

Nope

Slide 138

Slide 138 text

Thread.new do while @running sleep @timeout @reaper.reap end end

Slide 139

Slide 139 text

Thread.new do while @running sleep @timeout @reaper.reap end end

Slide 140

Slide 140 text

Was that scary?

Slide 141

Slide 141 text

Are threads scary?

Slide 142

Slide 142 text

(hint: no)

Slide 143

Slide 143 text

Any other examples?

Slide 144

Slide 144 text

@background_worker_thread = Thread.new do @background_worker.start { ScoutApm::Debug.instance.call_periodic_hooks ScoutApm::Agent.instance.process_metrics clean_old_percentiles } end

Slide 145

Slide 145 text

What about the GVL?

Slide 146

Slide 146 text

array = Array.new(100) { String.new } t1 = Thread.new do array.each {|x| x.prepend("hello") } end t2 = Thread.new do array.each {|x| x << " world" } end t1.join; t2.join puts array

Slide 147

Slide 147 text

hello world hello world hello world hello world hello world hello world hello world hello world

Slide 148

Slide 148 text

hello world hello world hello world hello world hello world hello world hello world hello world

Slide 149

Slide 149 text

What did the GVL do?

Slide 150

Slide 150 text

array = Array.new(100) { String.new } t1 = Thread.new do array.each {|x| x.prepend("hello") } end t2 = Thread.new do array.each {|x| x << " world" } end t1.join; t2.join puts array

Slide 151

Slide 151 text

array = Array.new(100) { String.new } t1 = Thread.new do array.each {|x| x.prepend("hello") } end t2 = Thread.new do array.each {|x| x << " world" } end t1.join; t2.join puts array

Slide 152

Slide 152 text

GVL means we cannot see speed increase just by using threads

Slide 153

Slide 153 text

No content

Slide 154

Slide 154 text

GVL is released when we do IO!

Slide 155

Slide 155 text

array = Array.new(100) { String.new } t1 = Thread.new do uri = URI.parse(“http://example.com/hello") hi = Net::Http.get(uri) array.each {|x| x.prepend(hi) } end t2 = Thread.new do uri = URI.parse(“http://example.com/hello") world = Net::Http.get(uri) array.each {|x| x << world } end t1.join; t2.join puts array

Slide 156

Slide 156 text

When is the GVL released exactly?

Slide 157

Slide 157 text

void * rb_thread_call_without_gvl2(void *(*func)(void *), void *data1, rb_unblock_function_t *ubf, void *data2) { return call_without_gvl(func, data1, ubf, data2, TRUE); } void * rb_thread_call_without_gvl(void *(*func)(void *data), void *data1, rb_unblock_function_t *ubf, void *data2) { return call_without_gvl(func, data1, ubf, data2, FALSE); }

Slide 158

Slide 158 text

Slide 159

Slide 159 text

Use a debugger to see GVL get released!

Slide 160

Slide 160 text

$ lldb ruby scratch.rb breakpoint set --name rb_thread_call_without_gvl breakpoint set --name rb_thread_call_without_gv2

Slide 161

Slide 161 text

$ lldb ruby scratch.rb breakpoint set --name rb_thread_call_without_gvl breakpoint set --name rb_thread_call_without_gv2

Slide 162

Slide 162 text

$ lldb ruby scratch.rb breakpoint set --name rb_thread_call_without_gvl breakpoint set --name rb_thread_call_without_gv2

Slide 163

Slide 163 text

Now run

Slide 164

Slide 164 text

$ lldb ruby scratch.rb breakpoint set --name rb_thread_call_without_gvl breakpoint set --name rb_thread_call_without_gv2 run

Slide 165

Slide 165 text

When we hit a pause, find the backtrace with

Slide 166

Slide 166 text

call rb_backtrace()

Slide 167

Slide 167 text

require ‘net/http' url = “https://www.schneems.com" uri = URI.parse(url) response = Net::HTTP.get(uri) puts response

Slide 168

Slide 168 text

require ‘net/http' url = “https://www.schneems.com" uri = URI.parse(url) response = Net::HTTP.get(uri) puts response

Slide 169

Slide 169 text

require ‘net/http' url = “https://www.schneems.com" uri = URI.parse(url) response = Net::HTTP.get(uri) puts response

Slide 170

Slide 170 text

require ‘net/http' url = “https://www.schneems.com" uri = URI.parse(url) response = Net::HTTP.get(uri) puts response

Slide 171

Slide 171 text

require ‘net/http' url = “https://www.schneems.com" uri = URI.parse(url) response = Net::HTTP.get(uri) puts response

Slide 172

Slide 172 text

require ‘net/http' url = “https://www.schneems.com" uri = URI.parse(url) response = Net::HTTP.get(uri) puts response

Slide 173

Slide 173 text

None of the above?

Slide 174

Slide 174 text

Use the debugger!

Slide 175

Slide 175 text

1309 { 1310 void *val = 0; 1311 -> 1312 rb_thread_t *th = GET_THREAD(); 1313 int saved_errno = 0; 1314 1315 if (ubf == RUBY_UBF_IO || ubf == RUBY_UBF_PROCESS) { (lldb) call rb_backtrace() from :4:in `' from :4:in `require' from.../rubygems.rb:1363:in `' from.../rubygems/specification.rb:873:in `load_defaults' from.../rubygems/specification.rb:821:in èach_spec' from.../rubygems/specification.rb:743:in èach_gemspec' from.../rubygems/specification.rb:743:in èach' from.../rubygems/specification.rb:744:in `block in each_gemspec' from.../rubygems/specification.rb:744:in èach' from.../rubygems/specification.rb:745:in `block (2 levels) in each_gemspec' from.../rubygems/specification.rb:822:in `block in each_spec' from.../rubygems/specification.rb:1161:in `load' from.../rubygems/specification.rb:1161:in `read' (lldb)

Slide 176

Slide 176 text

require ‘net/http' url = “https://www.schneems.com" uri = URI.parse(url) response = Net::HTTP.get(uri) puts response

Slide 177

Slide 177 text

I told a small lie

Slide 178

Slide 178 text

Slide 179

Slide 179 text

#define GVL_UNLOCK_BEGIN() do { \ rb_thread_t *_th_stored = GET_THREAD(); \ RB_GC_SAVE_MACHINE_CONTEXT(_th_stored); \ gvl_release(_th_stored->vm); #define GVL_UNLOCK_END() \ gvl_acquire(_th_stored->vm, _th_stored); \ rb_thread_set_current(_th_stored); \ } while(0)

Slide 180

Slide 180 text

BTW, you don’t need to know how to do this.

Slide 181

Slide 181 text

I’m just showing you it’s not magic

Slide 182

Slide 182 text

Recap: Threads are not…

Slide 183

Slide 183 text

Recap: Threads are not evil

Slide 184

Slide 184 text

Recap: GVL is not magic

Slide 185

Slide 185 text

How do we start using threads?

Slide 186

Slide 186 text

Make a Toy threading library!

Slide 187

Slide 187 text

It’s fun I promise

Slide 188

Slide 188 text

Make a worker thread

Slide 189

Slide 189 text

require ‘thread my_queue = Queue.new my_thread = Thread.new do loop do job = my_queue.pop job.call end end

Slide 190

Slide 190 text

require ‘thread my_queue = Queue.new my_thread = Thread.new do loop do job = my_queue.pop job.call end end

Slide 191

Slide 191 text

require ‘thread my_queue = Queue.new my_thread = Thread.new do loop do job = my_queue.pop job.call end end

Slide 192

Slide 192 text

require ‘thread my_queue = Queue.new my_thread = Thread.new do loop do job = my_queue.pop job.call end end

Slide 193

Slide 193 text

Use a worker thread via a “boss” i.e. main

Slide 194

Slide 194 text

10.times { job = Proc.new { puts "hello" } my_queue.push(job) }

Slide 195

Slide 195 text

It works!

Slide 196

Slide 196 text

But there are bugs

Slide 197

Slide 197 text

How can we be sure that all jobs are finished?

Slide 198

Slide 198 text

Join the thread

Slide 199

Slide 199 text

require 'thread' my_queue = Queue.new my_thread = Thread.new do loop do job = my_queue.pop job.call end end 10.times { job = Proc.new { puts "hello" } my_queue.push(job) } my_thread.join

Slide 200

Slide 200 text

Infinite loop

Slide 201

Slide 201 text

Why?

Slide 202

Slide 202 text

require ‘thread my_queue = Queue.new my_thread = Thread.new do loop do job = my_queue.pop job.call end end

Slide 203

Slide 203 text

Poison our queue

Slide 204

Slide 204 text

my_queue = Queue.new POISON = Object.new my_thread = Thread.new do loop do job = my_queue.pop break if job == POISON job.call end end

Slide 205

Slide 205 text

require 'thread' my_queue = Queue.new POISON = Object.new my_thread = Thread.new do loop do job = my_queue.pop break if job == POISON job.call end end 10.times { job = Proc.new { puts "hello" } my_queue.push(job) } my_queue.push(POISON) my_thread.join

Slide 206

Slide 206 text

Exercise: Build a thread pool that has multiple threads

Slide 207

Slide 207 text

To poison multiple workers, add multiple poison objects into pool

Slide 208

Slide 208 text

Threads are not…

Slide 209

Slide 209 text

Threads are not evil

Slide 210

Slide 210 text

GVL is not magic

Slide 211

Slide 211 text

There are other thread constructs

Slide 212

Slide 212 text

Mutex

Slide 213

Slide 213 text

Condition Variable

Slide 214

Slide 214 text

You can ignore most of them if you use a queue and boss/ worker pattern.

Slide 215

Slide 215 text

How do I _actually_ use threads?

Slide 216

Slide 216 text

concurrent_ruby

Slide 217

Slide 217 text

require 'concurrent' 10.times { Concurrent::Promise.execute( executor: :fast) do puts "hello" end }

Slide 218

Slide 218 text

require 'concurrent' 10.times { Concurrent::Promise.execute( executor: :fast) do puts "hello" end }

Slide 219

Slide 219 text

require 'concurrent' 10.times { Concurrent::Promise.execute( executor: :fast) do puts "hello" end }

Slide 220

Slide 220 text

Protip!

Slide 221

Slide 221 text

Always call wait!

Slide 222

Slide 222 text

Sprockets Example

Slide 223

Slide 223 text

After we process an asset we have to write it to disk

Slide 224

Slide 224 text

We also want a gzip file etc.

Slide 225

Slide 225 text

File writes release the GVL

Slide 226

Slide 226 text

Use Concurrent Ruby!

Slide 227

Slide 227 text

promise = Concurrent::Promise.execute( executor: executor) do exporter.call end concurrent_exporters << promise

Slide 228

Slide 228 text

promise = Concurrent::Promise.execute( executor: executor) do exporter.call end concurrent_exporters << promise