Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Wanna Go So You Fast? Strange Loop 2017 @tyler_treat
@tyler_treat@tyler_treat
@tyler_treat this one weird trick Make your code faster with
@tyler_treat this one weird trick Make your code faster with
@tyler_treat So You Wanna Subvert Go?
@tyler_treat Spoiler Alert:
 Go is not a
 systems language…
@tyler_treat but that doesn’t mean you can’t build internet-scale systems with it.
@tyler_treat
@tyler_treat This is a talk about how to write terrible Go code.
@tyler_treat@tyler_treat
@tyler_treat Because this is a talk about trade-offs.
@tyler_treat - Messaging Nerd @ Apcera - Working on nats.io - Distributed systems - bravenewgeek.com Tyler Treat
@tyler_treat@tyler_treat
@tyler_treat matter? Why does this talk
@tyler_treat The compiler isn’t magic.
@tyler_treat The compiler isn’t magic.
@tyler_treat You have to be
 mindful of performance
 when it matters.
@tyler_treat@tyler_treat Where bad things hide
@tyler_treat@tyler_treat Where bad things hideWhere we’re usually looking
@tyler_treat Tire fires at scale @tyler_treat
@tyler_treat@tyler_treat@tyler_treat
@tyler_treat@tyler_treat@tyler_treat
@tyler_treat@tyler_treat@tyler_treat
@tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
@tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
@tyler_treat Disclaimer:
 Don’t blindly apply optimizations presented.
@tyler_treat tl;dr of this talk is
 “IT DEPENDS!”
@tyler_treat Measure Optimize
@tyler_treat Measurement Techniques - pprof
 - memory
 - cpu
 - blocking - GODEBUG
 - gctrace
 - schedtrace
 - allocfreet...
@tyler_treat@tyler_treat
@tyler_treat The only way to get good at something is to be really fucking bad at it
 for a long time.
@tyler_treat Benchmarking… a great way to rattle the
 Hacker News fart chamber.
@tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
@tyler_treat channels
@tyler_treat “Instead of explicitly using locks to mediate access to shared data, Go encourages the use of channels to pas...
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat USE CHANNELS TO COORDINATE, NOT SYNCHRONIZE.
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat defer
@tyler_treat@tyler_treat
@tyler_treat Is defer still slow?
@tyler_treat@tyler_treat
@tyler_treat The Secret Life of interface{}
@tyler_treat type Stringer interface {
 String() string
 } https://research.swtch.com/interfaces
@tyler_treat type Stringer interface {
 String() string
 }
 type Binary uint64 https://research.swtch.com/interfaces
@tyler_treat type Stringer interface {
 String() string
 }
 type Binary uint64 200 b := Binary(200) https://research.swtch...
@tyler_treat type Stringer interface {
 String() string
 }
 type Binary uint64
 func (i Binary) String() string { return s...
@tyler_treat type Stringer interface {
 String() string
 } https://research.swtch.com/interfaces s := Stringer(b) Stringer...
@tyler_treat s := Stringer(b) Stringer tab data .
 .
 . itable(Stringer, Binary) type fun[0] type(Binary) (*Binary).String...
@tyler_treat tab data 200 Binary s := Stringer(b) Stringer .
 .
 . itable(Stringer, Binary) type fun[0] type(Binary) (*Bin...
@tyler_treat
@tyler_treat So what?
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat Sorting 100M Interfaces
@tyler_treat@tyler_treat Sorting 100M Interfaces
@tyler_treat@tyler_treat Sorting 100M Structs
@tyler_treat@tyler_treat Sorting 100M Structs
@tyler_treat $ go test -bench=. -gcflags="-m"
@tyler_treat $ go test -bench=. -gcflags="-m"
@tyler_treat@tyler_treat
@tyler_treat $ go test -bench=. -gcflags="-l"
@tyler_treat@tyler_treat Struct
 No Inlining Interface
 No Inlining
@tyler_treat@tyler_treat Struct
 No Inlining Interface
 No Inlining
@tyler_treat@tyler_treat Struct
 No Inlining Interface
 No Inlining
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat x.(*T) inlined
@tyler_treat@tyler_treat SSA backend &
 remaining type
 conversions inlined x.(*T) inlined
@tyler_treat@tyler_treat
@tyler_treat
@tyler_treat@tyler_treat Struct Interface
@tyler_treat@tyler_treat Struct Interface
@tyler_treat@tyler_treat
@tyler_treat $ go test -bench=. -gcflags="-S"
@tyler_treat $ go test -bench=. -gcflags="-S"
@tyler_treat $ go test -bench=. -gcflags="-S"
@tyler_treat Key Insight: If performance matters,
 write type-specific code.
@tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
@tyler_treat []byte to string
 conversions
@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat What’s going on here?
@tyler_treat@tyler_treat
@tyler_treat memory allocation
@tyler_treat@tyler_treat
@tyler_treat How is sync.Pool so fast?
@tyler_treat Per-CPU storage!
@tyler_treat@tyler_treat https://golang.org/src/sync/pool.go
@tyler_treat@tyler_treat https://golang.org/src/sync/pool.go
@tyler_treat@tyler_treat
@tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
@tyler_treat “We generally don’t want sync/atomic to be used at all…Experience has shown us again and again that very very...
@tyler_treat
@tyler_treat@tyler_treat Subscribers Messages Fast Topic Matching http://bravenewgeek.com/fast-topic-matching/
@tyler_treat@tyler_treat Subscribers Messages Fast Topic Matching http://bravenewgeek.com/fast-topic-matching/
@tyler_treat@tyler_treat Fast Topic Matching
@tyler_treat@tyler_treat Fast Topic Matching
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat Fast Topic Matching
@tyler_treat@tyler_treat Concurrent
 80,000 inserts
 80,000 lookups

@tyler_treat@tyler_treat Ctrie
@tyler_treat@tyler_treat G1 G1 1. Assign a generation, G1, to each
 I-node (empty struct). Ctrie
@tyler_treat 1. Assign a generation, G1, to each
 I-node (empty struct). 2. Add new node by copying I-node with updated b...
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat The Go race detector
 doesn’t protect you from
 doing dumb stuff.
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat Side note:
 unsafe is, in fact, unsafe.
@tyler_treat “Packages that import unsafe may depend on internal properties of the Go implementation. We reserve the right...
@tyler_treat
@tyler_treat Key Insight: Struct layout can make
 a big difference.
@tyler_treat@tyler_treat Mechanical Sympathy
@tyler_treat https://github.com/Workiva/go-datastructures/blob/master/queue/ring.go @tyler_treat
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat https://golang.org/src/sync/rwmutex.go
@tyler_treat@tyler_treat https://golang.org/src/sync/rwmutex.go
@tyler_treat CPU reader reader reader RWMutex
@tyler_treat CPU reader reader CPU readerreader reader RWMutex
@tyler_treat CPU reader reader CPU reader readerreader reader CPU readerreader reader RWMutex
@tyler_treat CPU reader reader CPU reader readerreader reader CPU readerreader CPU reader reader reader RWMutex
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
@tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU read...
CPU reader CPU readerreaderreader CPU readerreader CPU readerreader U writer CPU reader reader CPU reader readerreader rea...
@tyler_treat@tyler_treat
@tyler_treat How to create
 CPU->RWMutex
 mapping?
@tyler_treat@tyler_treat https://github.com/jonhoo/drwmutex/blob/master/cpu_amd64.s
@tyler_treat /proc/cpuinfo
@tyler_treat@tyler_treat
@tyler_treat memory RWMutex1 24 bytes
@tyler_treat RWMutex1 RWMutex2memory 24 bytes
@tyler_treat RWMutex1 RWMutex2 RWMutex3memory 24 bytes
@tyler_treat RWMutex1 RWMutex2 RWMutex3 RWMutexN…memory 24 bytes
@tyler_treat RWMutex1 RWMutex2 RWMutex3 RWMutexN…memory 24 bytes 64 bytes (cache line size)
@tyler_treat RWMutex1 RWMutex2 RWMutex3 RWMutexN…memory 24 bytes 64 bytes (cache line size) Cache rules everything around ...
@tyler_treat https://github.com/jonhoo/drwmutex/blob/master/drwmutex.go @tyler_treat
@tyler_treat https://github.com/jonhoo/drwmutex/blob/master/drwmutex.go @tyler_treat
@tyler_treat padding … 64 bytes (cache line size) memory 24 bytes RWMutex1 Cache rules everything around me
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat@tyler_treat
@tyler_treat Go makes concurrency
 easy enough to be dangerous.
@tyler_treat Conclusions
@tyler_treat The standard library provides
 general solutions (and they’re
 generally what you should use). 1
@tyler_treat Seemingly small, idiomatic
 decisions can have profound
 performance implications. 2
@tyler_treat The Go toolchain has lots
 of tools for analyzing your
 code—learn them. 3
@tyler_treat Go’s compiler and runtime
 continue to improve. 4
@tyler_treat Performance profile can
 change dramatically
 between releases. 5
@tyler_treat Relying on assumptions
 can be fatal. 6
@tyler_treat Code is marginal,
 architecture is material. 7
@tyler_treat Peeking behind the curtains
 can pay dividends. 8
@tyler_treat Above all, optimize for the
 right trade-off. 9
@tyler_treat Thanks!
Upcoming SlideShare
Loading in …5
×

So You Wanna Go Fast?

4,789 views

Published on

Go's simplicity and concurrency model make it an appealing choice for backend systems, but how does it fare for latency-sensitive applications? In this talk, we explore the other side of the coin by providing some tips on writing high-performance Go and lessons learned in the process. We do a deep dive on low-level performance optimizations in order to make Go a more compelling option in the world of systems programming, but we also consider the trade-offs involved.

Published in: Software
no profile picture user
  • Hello! Get Your Professional Job-Winning Resume Here - Check our website! https://vk.cc/818RFv
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here
  • Nice !! Download 100 % Free Ebooks, PPts, Study Notes, Novels, etc @ https://www.ThesisScientist.com
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here
Show More

So You Wanna Go Fast?

  1. 1. Wanna Go So You Fast? Strange Loop 2017 @tyler_treat
  2. 2. @tyler_treat@tyler_treat
  3. 3. @tyler_treat this one weird trick Make your code faster with
  4. 4. @tyler_treat this one weird trick Make your code faster with
  5. 5. @tyler_treat So You Wanna Subvert Go?
  6. 6. @tyler_treat Spoiler Alert:
 Go is not a
 systems language…
  7. 7. @tyler_treat but that doesn’t mean you can’t build internet-scale systems with it.
  8. 8. @tyler_treat
  9. 9. @tyler_treat This is a talk about how to write terrible Go code.
  10. 10. @tyler_treat@tyler_treat
  11. 11. @tyler_treat Because this is a talk about trade-offs.
  12. 12. @tyler_treat - Messaging Nerd @ Apcera - Working on nats.io - Distributed systems - bravenewgeek.com Tyler Treat
  13. 13. @tyler_treat@tyler_treat
  14. 14. @tyler_treat matter? Why does this talk
  15. 15. @tyler_treat The compiler isn’t magic.
  16. 16. @tyler_treat The compiler isn’t magic.
  17. 17. @tyler_treat You have to be
 mindful of performance
 when it matters.
  18. 18. @tyler_treat@tyler_treat Where bad things hide
  19. 19. @tyler_treat@tyler_treat Where bad things hideWhere we’re usually looking
  20. 20. @tyler_treat Tire fires at scale @tyler_treat
  21. 21. @tyler_treat@tyler_treat@tyler_treat
  22. 22. @tyler_treat@tyler_treat@tyler_treat
  23. 23. @tyler_treat@tyler_treat@tyler_treat
  24. 24. @tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
  25. 25. @tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
  26. 26. @tyler_treat Disclaimer:
 Don’t blindly apply optimizations presented.
  27. 27. @tyler_treat tl;dr of this talk is
 “IT DEPENDS!”
  28. 28. @tyler_treat Measure Optimize
  29. 29. @tyler_treat Measurement Techniques - pprof
 - memory
 - cpu
 - blocking - GODEBUG
 - gctrace
 - schedtrace
 - allocfreetrace - Benchmarking
 - Code-level: testing.B
 - System-level: HdrHistogram (https://github.com/codahale/hdrhistogram)
 bench (https://github.com/tylertreat/bench)
  30. 30. @tyler_treat@tyler_treat
  31. 31. @tyler_treat The only way to get good at something is to be really fucking bad at it
 for a long time.
  32. 32. @tyler_treat Benchmarking… a great way to rattle the
 Hacker News fart chamber.
  33. 33. @tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
  34. 34. @tyler_treat channels
  35. 35. @tyler_treat “Instead of explicitly using locks to mediate access to shared data, Go encourages the use of channels to pass references to data between goroutines.” https://blog.golang.org/share-memory-by-communicating
  36. 36. @tyler_treat@tyler_treat
  37. 37. @tyler_treat@tyler_treat USE CHANNELS TO COORDINATE, NOT SYNCHRONIZE.
  38. 38. @tyler_treat@tyler_treat
  39. 39. @tyler_treat@tyler_treat
  40. 40. @tyler_treat defer
  41. 41. @tyler_treat@tyler_treat
  42. 42. @tyler_treat Is defer still slow?
  43. 43. @tyler_treat@tyler_treat
  44. 44. @tyler_treat The Secret Life of interface{}
  45. 45. @tyler_treat type Stringer interface {
 String() string
 } https://research.swtch.com/interfaces
  46. 46. @tyler_treat type Stringer interface {
 String() string
 }
 type Binary uint64 https://research.swtch.com/interfaces
  47. 47. @tyler_treat type Stringer interface {
 String() string
 }
 type Binary uint64 200 b := Binary(200) https://research.swtch.com/interfaces
  48. 48. @tyler_treat type Stringer interface {
 String() string
 }
 type Binary uint64
 func (i Binary) String() string { return strconv.FormatUint(uint64(i), 2) } 200 b := Binary(200) https://research.swtch.com/interfaces
  49. 49. @tyler_treat type Stringer interface {
 String() string
 } https://research.swtch.com/interfaces s := Stringer(b) Stringer tab data
  50. 50. @tyler_treat s := Stringer(b) Stringer tab data .
 .
 . itable(Stringer, Binary) type fun[0] type(Binary) (*Binary).String type Stringer interface {
 String() string
 } https://research.swtch.com/interfaces
  51. 51. @tyler_treat tab data 200 Binary s := Stringer(b) Stringer .
 .
 . itable(Stringer, Binary) type fun[0] type(Binary) (*Binary).String type Stringer interface {
 String() string
 } https://research.swtch.com/interfaces
  52. 52. @tyler_treat
  53. 53. @tyler_treat So what?
  54. 54. @tyler_treat@tyler_treat
  55. 55. @tyler_treat@tyler_treat
  56. 56. @tyler_treat@tyler_treat Sorting 100M Interfaces
  57. 57. @tyler_treat@tyler_treat Sorting 100M Interfaces
  58. 58. @tyler_treat@tyler_treat Sorting 100M Structs
  59. 59. @tyler_treat@tyler_treat Sorting 100M Structs
  60. 60. @tyler_treat $ go test -bench=. -gcflags="-m"
  61. 61. @tyler_treat $ go test -bench=. -gcflags="-m"
  62. 62. @tyler_treat@tyler_treat
  63. 63. @tyler_treat $ go test -bench=. -gcflags="-l"
  64. 64. @tyler_treat@tyler_treat Struct
 No Inlining Interface
 No Inlining
  65. 65. @tyler_treat@tyler_treat Struct
 No Inlining Interface
 No Inlining
  66. 66. @tyler_treat@tyler_treat Struct
 No Inlining Interface
 No Inlining
  67. 67. @tyler_treat@tyler_treat
  68. 68. @tyler_treat@tyler_treat x.(*T) inlined
  69. 69. @tyler_treat@tyler_treat SSA backend &
 remaining type
 conversions inlined x.(*T) inlined
  70. 70. @tyler_treat@tyler_treat
  71. 71. @tyler_treat
  72. 72. @tyler_treat@tyler_treat Struct Interface
  73. 73. @tyler_treat@tyler_treat Struct Interface
  74. 74. @tyler_treat@tyler_treat
  75. 75. @tyler_treat $ go test -bench=. -gcflags="-S"
  76. 76. @tyler_treat $ go test -bench=. -gcflags="-S"
  77. 77. @tyler_treat $ go test -bench=. -gcflags="-S"
  78. 78. @tyler_treat Key Insight: If performance matters,
 write type-specific code.
  79. 79. @tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
  80. 80. @tyler_treat []byte to string
 conversions
  81. 81. @tyler_treat
  82. 82. @tyler_treat@tyler_treat
  83. 83. @tyler_treat@tyler_treat
  84. 84. @tyler_treat What’s going on here?
  85. 85. @tyler_treat@tyler_treat
  86. 86. @tyler_treat memory allocation
  87. 87. @tyler_treat@tyler_treat
  88. 88. @tyler_treat How is sync.Pool so fast?
  89. 89. @tyler_treat Per-CPU storage!
  90. 90. @tyler_treat@tyler_treat https://golang.org/src/sync/pool.go
  91. 91. @tyler_treat@tyler_treat https://golang.org/src/sync/pool.go
  92. 92. @tyler_treat@tyler_treat
  93. 93. @tyler_treat Overview - Measuring performance - Language features - Memory management - Concurrency and multi-core
  94. 94. @tyler_treat “We generally don’t want sync/atomic to be used at all…Experience has shown us again and again that very very few people are capable of writing correct code that uses atomic operations…” —Ian Lance Taylor
  95. 95. @tyler_treat
  96. 96. @tyler_treat@tyler_treat Subscribers Messages Fast Topic Matching http://bravenewgeek.com/fast-topic-matching/
  97. 97. @tyler_treat@tyler_treat Subscribers Messages Fast Topic Matching http://bravenewgeek.com/fast-topic-matching/
  98. 98. @tyler_treat@tyler_treat Fast Topic Matching
  99. 99. @tyler_treat@tyler_treat Fast Topic Matching
  100. 100. @tyler_treat@tyler_treat
  101. 101. @tyler_treat@tyler_treat Fast Topic Matching
  102. 102. @tyler_treat@tyler_treat Concurrent
 80,000 inserts
 80,000 lookups

  103. 103. @tyler_treat@tyler_treat Ctrie
  104. 104. @tyler_treat@tyler_treat G1 G1 1. Assign a generation, G1, to each
 I-node (empty struct). Ctrie
  105. 105. @tyler_treat 1. Assign a generation, G1, to each
 I-node (empty struct). 2. Add new node by copying I-node with updated branch and generation then GCAS, i.e. atomically:
 - compare I-nodes to detect tree
 mutations.
 - compare root generations to detect
 snapshots. @tyler_treat G2 G1 Ctrie
  106. 106. @tyler_treat@tyler_treat
  107. 107. @tyler_treat@tyler_treat
  108. 108. @tyler_treat The Go race detector
 doesn’t protect you from
 doing dumb stuff.
  109. 109. @tyler_treat@tyler_treat
  110. 110. @tyler_treat@tyler_treat
  111. 111. @tyler_treat@tyler_treat
  112. 112. @tyler_treat Side note:
 unsafe is, in fact, unsafe.
  113. 113. @tyler_treat “Packages that import unsafe may depend on internal properties of the Go implementation. We reserve the right to make changes to the implementation that may break such programs.” https://golang.org/doc/go1compat
  114. 114. @tyler_treat
  115. 115. @tyler_treat Key Insight: Struct layout can make
 a big difference.
  116. 116. @tyler_treat@tyler_treat Mechanical Sympathy
  117. 117. @tyler_treat https://github.com/Workiva/go-datastructures/blob/master/queue/ring.go @tyler_treat
  118. 118. @tyler_treat@tyler_treat
  119. 119. @tyler_treat@tyler_treat
  120. 120. @tyler_treat@tyler_treat
  121. 121. @tyler_treat@tyler_treat https://golang.org/src/sync/rwmutex.go
  122. 122. @tyler_treat@tyler_treat https://golang.org/src/sync/rwmutex.go
  123. 123. @tyler_treat CPU reader reader reader RWMutex
  124. 124. @tyler_treat CPU reader reader CPU readerreader reader RWMutex
  125. 125. @tyler_treat CPU reader reader CPU reader readerreader reader CPU readerreader reader RWMutex
  126. 126. @tyler_treat CPU reader reader CPU reader readerreader reader CPU readerreader CPU reader reader reader RWMutex
  127. 127. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader
  128. 128. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader
  129. 129. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader
  130. 130. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader
  131. 131. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader
  132. 132. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader
  133. 133. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader RWMutex RWMutex RWMutex
  134. 134. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader RWMutex RWMutex RWMutex
  135. 135. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader RWMutex RWMutex RWMutex
  136. 136. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader RWMutex RWMutex RWMutex
  137. 137. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader RWMutex RWMutex RWMutex
  138. 138. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader RWMutex RWMutex RWMutex
  139. 139. @tyler_treat RWMutex CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader RWMutex RWMutex RWMutex
  140. 140. CPU reader CPU readerreaderreader CPU readerreader CPU readerreader U writer CPU reader reader CPU reader readerreader readerreader writer CPU reader readerreader reader CPU reader writerreader reader CPU reader reader CPU reader readerreader writerreader reader CPU reader readerreader reader CPU reader readerreader reader reader readerreaderreader readerreaderreaderreader U reader reader ader ader U reader reader ader ader ader readerader CPU read readreader reader CPU read readreader reader CPU readreader readreader @tyler_treat
  141. 141. @tyler_treat@tyler_treat
  142. 142. @tyler_treat How to create
 CPU->RWMutex
 mapping?
  143. 143. @tyler_treat@tyler_treat https://github.com/jonhoo/drwmutex/blob/master/cpu_amd64.s
  144. 144. @tyler_treat /proc/cpuinfo
  145. 145. @tyler_treat@tyler_treat
  146. 146. @tyler_treat memory RWMutex1 24 bytes
  147. 147. @tyler_treat RWMutex1 RWMutex2memory 24 bytes
  148. 148. @tyler_treat RWMutex1 RWMutex2 RWMutex3memory 24 bytes
  149. 149. @tyler_treat RWMutex1 RWMutex2 RWMutex3 RWMutexN…memory 24 bytes
  150. 150. @tyler_treat RWMutex1 RWMutex2 RWMutex3 RWMutexN…memory 24 bytes 64 bytes (cache line size)
  151. 151. @tyler_treat RWMutex1 RWMutex2 RWMutex3 RWMutexN…memory 24 bytes 64 bytes (cache line size) Cache rules everything around me
  152. 152. @tyler_treat https://github.com/jonhoo/drwmutex/blob/master/drwmutex.go @tyler_treat
  153. 153. @tyler_treat https://github.com/jonhoo/drwmutex/blob/master/drwmutex.go @tyler_treat
  154. 154. @tyler_treat padding … 64 bytes (cache line size) memory 24 bytes RWMutex1 Cache rules everything around me
  155. 155. @tyler_treat@tyler_treat
  156. 156. @tyler_treat@tyler_treat
  157. 157. @tyler_treat@tyler_treat
  158. 158. @tyler_treat Go makes concurrency
 easy enough to be dangerous.
  159. 159. @tyler_treat Conclusions
  160. 160. @tyler_treat The standard library provides
 general solutions (and they’re
 generally what you should use). 1
  161. 161. @tyler_treat Seemingly small, idiomatic
 decisions can have profound
 performance implications. 2
  162. 162. @tyler_treat The Go toolchain has lots
 of tools for analyzing your
 code—learn them. 3
  163. 163. @tyler_treat Go’s compiler and runtime
 continue to improve. 4
  164. 164. @tyler_treat Performance profile can
 change dramatically
 between releases. 5
  165. 165. @tyler_treat Relying on assumptions
 can be fatal. 6
  166. 166. @tyler_treat Code is marginal,
 architecture is material. 7
  167. 167. @tyler_treat Peeking behind the curtains
 can pay dividends. 8
  168. 168. @tyler_treat Above all, optimize for the
 right trade-off. 9
  169. 169. @tyler_treat Thanks!

×