If a thread is created while the parent thread is "Background",
then the default timerslack value gets set to the current
timerslack value of the parent (40ms). The default value is
used when transitioning to "Foreground" -- so the effect is that
the timerslack value becomes 40ms regardless of foreground/background.
This does occur intermittently for systemui when creating its
render thread (pretty often on hammerhead and has been seen on
shamu). If this occurs, then some systemui animations like navbar
ripples can wait for up to 40ms to draw a frame when they intended
to wait 3ms -- jank.
This fix is to explicitly set the foreground timerslack to 50us.
A consequence of setting timerslack behind the process' back is
that any custom values for timerslack get lost whenever the thread
has transition between fg/bg.
See Bug: 19398120
Change-Id: Idc259717f62fa2255f8bafbbf88b68c0043f29cf
All kernel services will now be in the same cgroup as
foreground applications. This will now make kernel threads
not implicitly higher priority than android foreground
services.
Bug 17681097
Change-Id: I28e81c7aade50428d5395df86f00ce01c1e7af02
Adds an apps cgroup which will contain fg tasks. Move the bg group
nested underneath it.
Bug: 6538553
Change-Id: I55279415da625e15a5bb061c2b45fbe7f7a4c047
Signed-off-by: Dima Zavin <dima@android.com>
Also tid == 0 maps to gettid() on platforms that support it.
This is needed by mediaserver code that was formerly using
the old C++ API.
Change-Id: I8baaa3e42315c9b924cec26b6da01e7cb69e0078
This will make it easier to add additional policies (cgroups) if needed.
Also added comments to the sched_policy APIs.
Change-Id: I33ce1cc4deae10983241f7391294b7a512d2c47c
Re-use the file descriptor to change the cgroups. There's a fast path
in the kernel for the case where the value's already uncahnged. Also,
this is okay because the kernel ignores the write offset. This is
essentially a ioctl-ish/datagram interface, despite being a fd we're
writing to.
This reduces Dalvik->Dalvik void/void Binder calls on Sapphire from
~1.2 to ~0.8 ms. This is a bit slower than the ~0.7 ms we get
avoiding the write altogether.
strace confirms:
[pid 1557] ioctl(21, 0xc0186201 <unfinished ...>
[pid 1556] <... ioctl resumed> , 0x45399cd0) = 0
[pid 1556] write(22, "1556", 4) = 4
[pid 1556] ioctl(21, 0xc0186201 <unfinished ...>
[pid 1555] <... ioctl resumed> , 0x45299cd0) = 0
[pid 1555] write(22, "1555", 4) = 4
[pid 1555] ioctl(21, 0xc0186201 <unfinished ...>
[pid 1554] <... ioctl resumed> , 0x45199cd0) = 0
[pid 1554] write(22, "1554", 4) = 4
BUG=2660235
Change-Id: Ia9f9e1d5b792eaebc9560f89931faf6df0cf9c0d
- use static /dev/cpuctl filenames, rather than sprintf() on the fly
- use a custom formatter instead of sprintf() for the tid number
This is a simplified version of Dan's original
Ifc9c81f74fe65f0695ead53b30194bc6adf00da0 to be less risky for Froyo.
Bug: 2660235
Change-Id: I744bbc2ec01fd2569612c814a497df40a933b622
This introduces a new HAVE_SCHED_H arch define, which is used
to prevent from building sched_policy.c during the Windows SDK.
Change-Id: I3667857a4ae7d6baaf1efd1cd187a5baf91419d8
