The output from the flag -XX:+PrintHeapAtGC is intermixed with the output
from -XX:+PrintGCDetails and -XX:+PrintGC (the latter is equivalent to
-verbose:gc) with the "framework" collectors (serial gc, parnew gc, cms).
This makes it hard to read and analyze.

[GC {Heap before GC invocations=0 (full 0):
 par new generation total 4032K, used 3793K [0xe0000000, 0xe0400000, 0xe0800000)
  eden space 3968K, 95% used [0xe0000000, 0xe03b47c8, 0xe03e0000)
  from space 64K, 0% used [0xe03e0000, 0xe03e0000, 0xe03f0000)
  to space 64K, 0% used [0xe03f0000, 0xe03f0000, 0xe0400000)
 concurrent mark-sweep generation total 12288K, used 0K [0xe0800000, 0xe1400000, 0xf8000000)
 concurrent-mark-sweep perm gen total 12288K, used 1603K [0xf8000000, 0xf8c00000, 0xfc000000)
[ParNew: 3793K->0K(4032K), 0.0097281 secs] 3793K->3623K(16320K) {Heap after GC invocations=1 (full 0):
 par new generation total 4032K, used 0K [0xe0000000, 0xe0400000, 0xe0800000)
  eden space 3968K, 0% used [0xe0000000, 0xe0000000, 0xe03e0000)
  from space 64K, 0% used [0xe03f0000, 0xe03f0000, 0xe0400000)
  to space 64K, 0% used [0xe03e0000, 0xe03e0000, 0xe03f0000)
 concurrent mark-sweep generation total 12288K, used 3623K [0xe0800000, 0xe1400000, 0xf8000000)
 concurrent-mark-sweep perm gen total 12288K, used 1603K [0xf8000000, 0xf8c00000, 0xfc000000)
}
, 0.0142133 secs]