| Commit message (Collapse) | Author | Age |
|
|
|
|
| |
Change-Id: I3fd23490bcdfa4097dad73161ad226337e93cd18
Signed-off-by: Davide Garberi <dade.garberi@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The maximum heap size for each virtual machine is 1GB. This is too big
and puts too much pressure on both the garbage collector when it tries
to perform collections in older generations. The symptoms are simple,
some applications will allocate objects that live long enough to be
promoted to older generations, escaping the efficient CMS collections.
Eventually, a stop-the-world collection will go through this large block
of memory and cause applications to pause for too long.
Additionally, the out of memory killer follows a LRU approach where it
tries to kill older applciations to make room for new ones. The problem
arises when the least recently used applications don't have large heaps
but the most recent ones do. When this happens the system halts to a
crawl while older apps and services are killed and respawned in a never
ending cycle of jankiness.
These values were obtained experimentally but actually match the
defaults for devices with >2gb of ram and xhdpi.
Change-Id: I73a92269ceac61dd4d4c1b1600f55a21534a27a0
Signed-off-by: Davide Garberi <dade.garberi@gmail.com>
|
|
|
|
| |
Signed-off-by: Davide Garberi <dade.garberi@gmail.com>
|
|
|
|
|
| |
Change-Id: I656cbed7abf351fea4129e7ff3c930eb9d8914ae
Signed-off-by: Davide Garberi <dade.garberi@gmail.com>
|
|
|
|
| |
Signed-off-by: Davide Garberi <dade.garberi@gmail.com>
|
|
|
|
| |
Signed-off-by: dd3boh <dade.garberi@gmail.com>
|
|
|
|
| |
Signed-off-by: dd3boh <dade.garberi@gmail.com>
|
|
|
|
|
|
| |
I prefer it like this :P
Signed-off-by: dd3boh <dade.garberi@gmail.com>
|
|
|
|
|
| |
Change-Id: I656cbed7abf351fea4129e7ff3c930eb9d8914ae
Signed-off-by: dd3boh <dade.garberi@gmail.com>
|
| |
|
| |
|
|
|
|
| |
Change-Id: I4108d946fe4005af64a34b46e68dd94cdef98a6b
|
| |
|
|
|
|
| |
*Lets start
|
|
|
|
| |
Change-Id: Ifd9184c31569d88a21d989e6df59d20e47aed135
|
| |
|
| |
|
| |
|
|
|