Android垃圾回收运行amok

Question

我在活动中有以下代码。最终，我试图做框架动画，但GC正在杀死性能。我试过显式回收，但没有区别。

我也知道我可以重新使用位图与BitmapFactor.Options.inBitmap，并且确实解决了问题，但限制我到API> = 11这不是一个选项。

这段代码只是说明了这个问题。这不是我的应用程序的一部分：

 Log.d("GC_badness", "0: " + SystemClock.uptimeMillis()); 
     Bitmap image = BitmapFactory.decodeResource(getResources(), R.drawable.my_image); 
     Log.d("GC_badness", "1: " + SystemClock.uptimeMillis()); 
     image = BitmapFactory.decodeResource(getResources(), R.drawable.my_image); 
     Log.d("GC_badness", "2: " + SystemClock.uptimeMillis()); 
     image = BitmapFactory.decodeResource(getResources(), R.drawable.my_image); 
     Log.d("GC_badness", "3: " + SystemClock.uptimeMillis()); 
     image = BitmapFactory.decodeResource(getResources(), R.drawable.my_image); 
     Log.d("GC_badness", "4: " + SystemClock.uptimeMillis()); 
     image = BitmapFactory.decodeResource(getResources(), R.drawable.my_image); 
     Log.d("GC_badness", "5: " + SystemClock.uptimeMillis()); 
     image = BitmapFactory.decodeResource(getResources(), R.drawable.my_image); 
     Log.d("GC_badness", "6: " + SystemClock.uptimeMillis()); 

这会导致以下日志条目。每个来自jpg资源的位图加载至少导致3个垃圾收集，多达8个垃圾收集。大多数情况下，它们什么都收不到或者什么都收不到。即使有多条消息说明它是堆，堆也没有增长。这也很令人困惑。

我很难过。我无法找到一种方法来重新使用我的位图API-11，我无法忍受所有这些无用的垃圾回收所带来的巨大延迟。

04-07 18:17:51.860: D/GC_badness(7510): 0: 360583998 
04-07 18:17:51.900: D/dalvikvm(7510): GC_FOR_ALLOC freed 34K, 5% free 6320K/6595K, paused 32ms 
04-07 18:17:51.900: I/dalvikvm-heap(7510): Grow heap (frag case) to 7.744MB for 1584016-byte allocation 
04-07 18:17:51.950: D/dalvikvm(7510): GC_FOR_ALLOC freed <1K, 5% free 7867K/8199K, paused 27ms 
04-07 18:17:52.000: D/dalvikvm(7510): GC_CONCURRENT freed <1K, 5% free 7868K/8199K, paused 2ms+3ms 
04-07 18:17:52.030: D/dalvikvm(7510): GC_FOR_ALLOC freed <1K, 5% free 7868K/8199K, paused 31ms 
04-07 18:17:52.030: I/dalvikvm-heap(7510): Grow heap (frag case) to 8.416MB for 704016-byte allocation 
04-07 18:17:52.070: D/dalvikvm(7510): GC_FOR_ALLOC freed <1K, 4% free 8555K/8903K, paused 37ms 
04-07 18:17:52.070: D/dalvikvm(8107): GC_FOR_ALLOC freed 1464K, 39% free 17183K/27911K, paused 62ms 
04-07 18:17:52.100: D/GC_badness(7510): 1: 360584238 
04-07 18:17:52.100: D/dalvikvm(8189): GC_CONCURRENT freed 286K, 8% free 6917K/7495K, paused 17ms+3ms 
04-07 18:17:52.110: D/dalvikvm(7510): GC_CONCURRENT freed 1546K, 22% free 7009K/8903K, paused 2ms+2ms 
04-07 18:17:52.150: D/dalvikvm(7510): GC_FOR_ALLOC freed <1K, 22% free 7008K/8903K, paused 32ms 
04-07 18:17:52.150: I/dalvikvm-heap(7510): Grow heap (frag case) to 8.416MB for 1584016-byte allocation 
04-07 18:17:52.200: D/dalvikvm(7510): GC_FOR_ALLOC freed 0K, 4% free 8555K/8903K, paused 37ms 
04-07 18:17:52.250: D/dalvikvm(7510): GC_CONCURRENT freed <1K, 4% free 8555K/8903K, paused 2ms+2ms 
04-07 18:17:52.280: D/dalvikvm(7510): GC_FOR_ALLOC freed <1K, 4% free 8555K/8903K, paused 30ms 
04-07 18:17:52.280: I/dalvikvm-heap(7510): Grow heap (frag case) to 9.086MB for 704016-byte allocation 
04-07 18:17:52.310: D/dalvikvm(7510): GC_FOR_ALLOC freed 0K, 4% free 9243K/9607K, paused 23ms 
04-07 18:17:52.330: D/dalvikvm(8107): GC_CONCURRENT freed 588K, 34% free 18618K/27911K, paused 4ms+5ms 
04-07 18:17:52.350: D/GC_badness(7510): 2: 360584482 
04-07 18:17:52.350: D/dalvikvm(7510): GC_CONCURRENT freed 1546K, 20% free 7696K/9607K, paused 1ms+2ms 
04-07 18:17:52.380: D/dalvikvm(7510): GC_FOR_ALLOC freed 688K, 28% free 7008K/9607K, paused 22ms 
04-07 18:17:52.380: I/dalvikvm-heap(7510): Grow heap (frag case) to 8.416MB for 1584016-byte allocation 
04-07 18:17:52.400: D/dalvikvm(7510): GC_FOR_ALLOC freed <1K, 11% free 8555K/9607K, paused 21ms 
04-07 18:17:52.440: D/GC_badness(7510): 3: 360584577 
04-07 18:17:52.450: D/dalvikvm(7510): GC_CONCURRENT freed <1K, 4% free 9243K/9607K, paused 1ms+2ms 
04-07 18:17:52.470: D/dalvikvm(7510): GC_FOR_ALLOC freed 2234K, 28% free 7008K/9607K, paused 20ms 
04-07 18:17:52.470: I/dalvikvm-heap(7510): Grow heap (frag case) to 8.416MB for 1584016-byte allocation 
04-07 18:17:52.500: D/dalvikvm(7510): GC_FOR_ALLOC freed 0K, 11% free 8555K/9607K, paused 29ms 
04-07 18:17:52.510: D/dalvikvm(8107): GC_CONCURRENT freed 1785K, 33% free 18832K/27911K, paused 2ms+5ms 
04-07 18:17:52.530: D/GC_badness(7510): 4: 360584668 
04-07 18:17:52.540: D/dalvikvm(7510): GC_CONCURRENT freed <1K, 4% free 9243K/9607K, paused 1ms+2ms 
04-07 18:17:52.570: D/dalvikvm(7510): GC_FOR_ALLOC freed 2234K, 28% free 7008K/9607K, paused 28ms 
04-07 18:17:52.570: I/dalvikvm-heap(7510): Grow heap (frag case) to 8.416MB for 1584016-byte allocation 
04-07 18:17:52.590: D/dalvikvm(7510): GC_FOR_ALLOC freed <1K, 11% free 8555K/9607K, paused 22ms 
04-07 18:17:52.620: D/dalvikvm(7510): GC_CONCURRENT freed <1K, 4% free 9243K/9607K, paused 3ms+2ms 
04-07 18:17:52.630: D/GC_badness(7510): 5: 360584765 
04-07 18:17:52.650: D/dalvikvm(7510): GC_FOR_ALLOC freed 2235K, 28% free 7008K/9607K, paused 21ms 
04-07 18:17:52.650: I/dalvikvm-heap(7510): Grow heap (frag case) to 8.416MB for 1584016-byte allocation 
04-07 18:17:52.680: D/dalvikvm(7510): GC_FOR_ALLOC freed 0K, 11% free 8555K/9607K, paused 27ms 
04-07 18:17:52.710: D/GC_badness(7510): 6: 360584844 

Answer 1

为什么要重用图像？

不要给垃圾收集器一个机会去收集东西。基本上，您在本地定义的每个对象迟早都会被收集。

不要重复使用图像变量。在动画代码之前制作图像阵列和图像（最好是在创建活动期间加载）。 或考虑为您的图片使用ObjectPool。

这是一个很好的SO answer关于如何遵守垃圾收集。

最后你可以找到大量有关如何避免垃圾收集的信息，当你谷歌例如。 “如何避免垃圾回收机器人”。