hammerhead: update prebuilt kernel

fa9deaf kgsl: allocate shared memory one page at a time

This helps greatly alleviate pressure on the page allocator on devices running
for longer periods of time, with highly-fragmented memory.  Of all allocations
of order > 0, KGSL allocations consistently represent 70%, with another 20%
coming from order-1 allocations due to calls to fork() and clone().

b/18069309 N5 running L can't run as many apps simultaneously as running K

Change-Id: I184ef9c8035ca065dbafe8e672a038867ef85a5f
Signed-off-by: Iliyan Malchev <malchev@google.com>
1 file changed
tree: 6ae7bcc4f17ae45f29c87976d968561b71dd0c1c
  1. zImage-dtb