一. GC

  • GC英文全称 garbage collector
  • Go语言GC是相对C/C++语言非常重要的改进
  • 一些常用GC算法
    • 引用计算法.当对象被引用时计算器加一.不被引用计数器减一
      • PHP和Object-C使用
      • 相互引用无法回收
      • 计数增加消耗
    • Mark And Sweep 标记和清除算法.停止程序运行,递归遍历对象,进行标记.标记完成后将所有没有引用的对象进行清除
      • 由于标记需要停止程序(Stop the world),当对象特别多时,标记和清除过程比较耗时(可能几百毫秒),很难接受
    • 三色标记法:是Mark And Sweep的改进版.从逻辑上分为白色区(未搜索),灰色区(正搜索),黑色区(已搜索).灰色区内容是子引用没有进行搜索,黑色区表示子引用存在
    • 分代收集.一般情况都有三代,例如java中新生代,老年代,永久代.当新生代中带有阈值时会把对象放入到老年代,相同道理老年代内容达到阈值会放入到永久代

二.Go语言中的GC

  • Go语言中采用Stop The World方式
  • Golang每个版本基本上都会对GC进行优化,从Golang1.5开始支持并发(concurrent )收集,从1.8版本已经把STW时间优化到了100微妙,通常只需要10微妙以下.且在1.10版本时再次优化减少GC对CPU占用
  • Go语言中GC是自动运行的,在下列情况下会触发GC
    • 当需要申请内存时,发现GC是上次GC两倍时会触发
    • 每2分钟自动运行一次GC
  • GC调优
    • 小对象复用,局部变量尽量少声明,多个小对象可以放入到结构体,方便GC扫描
    • 少用string的”+”
  • 在runtime包下mgc.go中明确的说明了Golang的GC的解释
    1. // Garbage collector (GC).
    2. //
    3. // The GC runs concurrently with mutator threads, is type accurate (aka precise), allows multiple
    4. // GC thread to run in parallel. It is a concurrent mark and sweep that uses a write barrier. It is
    5. // non-generational and non-compacting. Allocation is done using size segregated per P allocation
    6. // areas to minimize fragmentation while eliminating locks in the common case.
    7. //
    8. // The algorithm decomposes into several steps.
    9. // This is a high level description of the algorithm being used. For an overview of GC a good
    10. // place to start is Richard Jones' gchandbook.org.
    11. //
    12. // The algorithm's intellectual heritage includes Dijkstra's on-the-fly algorithm, see
    13. // Edsger W. Dijkstra, Leslie Lamport, A. J. Martin, C. S. Scholten, and E. F. M. Steffens. 1978.
    14. // On-the-fly garbage collection: an exercise in cooperation. Commun. ACM 21, 11 (November 1978),
    15. // 966-975.
    16. // For journal quality proofs that these steps are complete, correct, and terminate see
    17. // Hudson, R., and Moss, J.E.B. Copying Garbage Collection without stopping the world.
    18. // Concurrency and Computation: Practice and Experience 15(3-5), 2003.
    19. //
    20. // 1. GC performs sweep termination.
    21. //
    22. //    a. Stop the world. This causes all Ps to reach a GC safe-point.
    23. //
    24. //    b. Sweep any unswept spans. There will only be unswept spans if
    25. //    this GC cycle was forced before the expected time.
    26. //
    27. // 2. GC performs the "mark 1" sub-phase. In this sub-phase, Ps are
    28. // allowed to locally cache parts of the work queue.
    29. //
    30. //    a. Prepare for the mark phase by setting gcphase to _GCmark
    31. //    (from _GCoff), enabling the write barrier, enabling mutator
    32. //    assists, and enqueueing root mark jobs. No objects may be
    33. //    scanned until all Ps have enabled the write barrier, which is
    34. //    accomplished using STW.
    35. //
    36. //    b. Start the world. From this point, GC work is done by mark
    37. //    workers started by the scheduler and by assists performed as
    38. //    part of allocation. The write barrier shades both the
    39. //    overwritten pointer and the new pointer value for any pointer
    40. //    writes (see mbarrier.go for details). Newly allocated objects
    41. //    are immediately marked black.
    42. //
    43. //    c. GC performs root marking jobs. This includes scanning all
    44. //    stacks, shading all globals, and shading any heap pointers in
    45. //    off-heap runtime data structures. Scanning a stack stops a
    46. //    goroutine, shades any pointers found on its stack, and then
    47. //    resumes the goroutine.
    48. //
    49. //    d. GC drains the work queue of grey objects, scanning each grey
    50. //    object to black and shading all pointers found in the object
    51. //    (which in turn may add those pointers to the work queue).
    52. //
    53. // 3. Once the global work queue is empty (but local work queue caches
    54. // may still contain work), GC performs the "mark 2" sub-phase.
    55. //
    56. //    a. GC stops all workers, disables local work queue caches,
    57. //    flushes each P's local work queue cache to the global work queue
    58. //    cache, and reenables workers.
    59. //
    60. //    b. GC again drains the work queue, as in 2d above.
    61. //
    62. // 4. Once the work queue is empty, GC performs mark termination.
    63. //
    64. //    a. Stop the world.
    65. //
    66. //    b. Set gcphase to _GCmarktermination, and disable workers and
    67. //    assists.
    68. //
    69. //    c. Drain any remaining work from the work queue (typically there
    70. //    will be none).
    71. //
    72. //    d. Perform other housekeeping like flushing mcaches.
    73. //
    74. // 5. GC performs the sweep phase.
    75. //
    76. //    a. Prepare for the sweep phase by setting gcphase to _GCoff,
    77. //    setting up sweep state and disabling the write barrier.
    78. //
    79. //    b. Start the world. From this point on, newly allocated objects
    80. //    are white, and allocating sweeps spans before use if necessary.
    81. //
    82. //    c. GC does concurrent sweeping in the background and in response
    83. //    to allocation. See description below.
    84. //
    85. // 6. When sufficient allocation has taken place, replay the sequence
    86. // starting with 1 above. See discussion of GC rate below.