Defining get() in the declaration of nsThreadManager implicitly sticks
an "inline" on the function, which is not what we want: inlining it
spreads around a lot of static initialization code. Providing an
out-of-line definition is much better in terms of code size.
By doing this we avoid triggering assertions in the Servo code that ensure
we have registered the thread with Servo and set the proper state on it.
MozReview-Commit-ID: K6qHrYoQDLm
--HG--
extra : rebase_source : d01b0aad42273f6b92b7cfd5f5fe17ffe7b4cda0
Unfortunately, this needed some additional trickery in order to keep its
constructor "private". I stole this trick from [1]. With this patch, we tear
down the statistics object during XPCOM shutdown intead of after it. I don't
believe that we need the object to live past the ClearOnShutdown destructors.
[1] http://rienajouter.blogspot.com/2014/10/makeshared-and-makeunique-for-classes.html
MozReview-Commit-ID: JsiN6Bq9Yp4
--HG--
extra : rebase_source : dd26c8e6906a6c9fd500c28379f8c63fd7c3ad6a
Properly enclose all relevant details of CPUUsageWatcher in ifdefs
which control whether it should be active or not. Additionally,
apparently clock_gettime is not defined on OSX prior to 10.12, so
this is failing to compile for OSX on the build server, but not
locally. However, clock_get_time and getrusage should cover our
use cases sufficiently.
MozReview-Commit-ID: Ffi6yXLb9gO
--HG--
extra : rebase_source : 84f9cf3b2074883dc6fe6d5a50ff27ffdb008a4f
We would like to be able to see if a given hang in BHR occurred
under high CPU load, as this is an indication that the hang is
of less use to us, since it's likely that the external CPU use
is more responsible for it.
The way this works is fairly simple. We get the system CPU usage
on a scale from 0 to 1, and we get the current process's CPU
usage, also on a scale from 0 to 1, and we subtract the latter
from the former. We then compare this value to a threshold, which
is 1 - (1 / p), where p is the number of (virtual) cores on the
machine. This threshold might need to be tuned, so that we
require an entire physical core in order to not annotate the hang,
but for now it seemed the most reasonable line in the sand.
I should note that this considers CPU usage in child or parent
processes as external. While we are responsible for that CPU usage,
it still indicates that the stack we receive from BHR is of little
value to us, since the source of the actual hang is external to
that stack.
MozReview-Commit-ID: JkG53zq1MdY
--HG--
extra : rebase_source : 16553a9b5eac0a73cd1619c6ee01fa177ca60e58
In some unknown circumstance, possibly if the parent process has a
different version than the child process, the child does not receive
scheduler prefs, which makes it read out of an uninitialized local
variable. This can probably happen because the scheduler prefs are
checked before we do the ContentChild::Init version check. Bill also
suggested that the pref env var might be getting truncated.
This patch improves on the situation by using null for the prefs array
if none was sent, and falling back on the default values, which leave
the scheduler disabled.
This also checks that the pref string is at least long enough to avoid
a buffer overflow. Note that if the end of the string isn't an integer
we'll end up with an sPrefThreadCount of zero, which can't be good.
MozReview-Commit-ID: ByHLFMEpgyZ
--HG--
extra : rebase_source : 8f6368b88ec3746f4d1c7716a962bb2ac3c2f3b5
The class, in practice, was already doing nothing in that case, but with
it being half #ifdef'ed on EARLY_BETA_OR_EARLIER, that led to build
failures for unused code on late beta. Just remove the class completely
on late beta.
--HG--
extra : rebase_source : 801b0b9bb9faf41270f72f616e720d5725e8b25c
The return of these functions is actually (DWORD) –1
MozReview-Commit-ID: 112d6BTBt8O
--HG--
extra : rebase_source : f36ec05d9a1e85d4d2dd844d8024189971aaeb46
This patch refactors the nsThread event queue to clean it up and to make it easier to restructure. The fundamental concepts are as follows:
Each nsThread will have a pointer to a refcounted SynchronizedEventQueue. A SynchronizedEQ takes care of doing the locking and condition variable work when posting and popping events. For the actual storage of events, it delegates to an AbstractEventQueue data structure. It keeps a UniquePtr to the AbstractEventQueue that it uses for storage.
Both SynchronizedEQ and AbstractEventQueue are abstract classes. There is only one concrete implementation of SynchronizedEQ in this patch, which is called ThreadEventQueue. ThreadEventQueue uses locks and condition variables to post and pop events the same way nsThread does. It also encapsulates the functionality that DOM workers need to implement their special event loops (PushEventQueue and PopEventQueue). In later Quantum DOM work, I plan to have another SynchronizedEQ implementation for the main thread, called SchedulerEventQueue. It will have special code for the cooperatively scheduling threads in Quantum DOM.
There are two concrete implementations of AbstractEventQueue in this patch: EventQueue and PrioritizedEventQueue. EventQueue replaces the old nsEventQueue. The other AbstractEventQueue implementation is PrioritizedEventQueue, which uses multiple queues for different event priorities.
The final major piece here is ThreadEventTarget, which splits some of the code for posting events out of nsThread. Eventually, my plan is for multiple cooperatively scheduled nsThreads to be able to share a ThreadEventTarget. In this patch, though, each nsThread has its own ThreadEventTarget. The class's purpose is just to collect some related code together.
One final note: I tried to avoid virtual dispatch overhead as much as possible. Calls to SynchronizedEQ methods do use virtual dispatch, since I plan to use different implementations for different threads with Quantum DOM. But all the calls to EventQueue methods should be non-virtual. Although the methods are declared virtual, all the classes used are final and the concrete classes involved should all be known through templatization.
MozReview-Commit-ID: 9Evtr9oIJvx
HangAnnotations was very complex, required a separate allocation, and used this
unfortunate virtual interface implementation which made it harder to do
interesting things with it (such as serialize it over IPC).
This new implementation is much simpler and more concrete, making
HangAnnotations simply be a nsTArray<Annotation>. This also simplifies some of
the IPC code which was added in part 7.
MozReview-Commit-ID: EzaaxdHpW1t
These will be used to implement IPC serialization and deserialization of the
HangDetails object to send over IPC. This is a temporary measure as
HangAnnotations is rewritten in part 11.
MozReview-Commit-ID: 1WHNvhDrMF5
The test helper_touch_action_regions.html uses nsDOMWindowUtils to synthesize native input events and creates some runnables to trigger the test. It expects the runnables which synthesize native input events are processed first, then the runnables to continue the test, and finally the input events are forwarded from chrome process to content process. Enabling event prioritization may change the execution order.
Wraps those runnables to synthesize native input events as priority=input and dispatches those runnables to continue the test with priority=input to make sure the execution order is as expected.
MozReview-Commit-ID: 8hkaB1FRW9T