PerformanceCounters are currently disabled in two ways:
- a preference that's off by default "dom.performance.enable_scheduler_timing"
- calls made only for nightly using #ifndef RELEASE_OR_BETA
In order to simplify the code, let's remove the #ifndef and rely only on the pref.
That will also allows us to use the feature in every version going forward.
The performance will not be impacted since the current code is already using
the (cached) pref value to determine if the counters are used.
MozReview-Commit-ID: 47t2M1O13aH
--HG--
extra : rebase_source : e129e1829f1dc37c019e50e156474c4876d6d6cb
Adds a PeformanceCounter class that is used in DocGroup and WorkerPrivate
to track runnables execution and dispatch counts.
MozReview-Commit-ID: 51DLj6ORD2O
--HG--
extra : rebase_source : b481c9aa3b735569722bb7472872ec2d22adcb89
The change to RootAccessible.cpp fixes an obvious bug introduced in bug 741707.
The visibility changes in gfx/thebes are because NS_DECL_ISUPPORTS has a
trailing "public:" that those classes were relying on to have public
constructors.
MozReview-Commit-ID: IeB8KIJCGhU
Our Web Worker code uses a thread pool where a single OS thread can be reused
for different worker scripts during its lifetime. Before this patch, we only
registered these threads with the profiler for the duration that they're
running a worker script. So the same OS thread could be registered with the
profiler during multiple disjoint time ranges, and we would expect the profiler
to treat those different registrations as different conceptual threads.
This had multiple advantages:
- The "thread name" of the conceptual thread can include the script URL:
"DOM Worker <scriptURL>". This allowed you to create thread filter which
match a part of the URL, so you had the option of profiling just the worker
threads you were interested in.
- We wouldn't waste time sampling a worker thread while it's idle and has no
script.
But it also had disadvantages:
- The profiler platform doesn't actually know how to deal with different
"conceptual threads" that share the same OS thread. This lead to surprising
breakage in different places. For example, the contents in the profiler
buffer are marked with ThreadId entries which use the OS thread id.
- What we show in the profiler UI didn't not match reality, and might be
confusing to some people.
I don't think the advantages are large enough to warrant teaching the rest of
the profiler platform to deal with conceptual threads. So this change makes us
stop doing the special thing and just register the OS threads for their entire
duration.
MozReview-Commit-ID: 82RtlRlwy3Y
--HG--
extra : rebase_source : 101e144f17718ea8f05e8bef8200b8ed41ee854e
Most cases where the pointer is stored into an already-declared variable can
trivially be changed to MakeNotNull<T*>, as the NotNull raw pointer will end
up in a smart pointer.
In RAII cases, the target type can be specified (e.g.:
`MakeNotNull<RefPtr<imgFrame>>)`), in which case the variable type may just be
`auto`, similar to the common use of MakeUnique.
Except when the target type is a base pointer, in which case it must be
specified in the declaration.
MozReview-Commit-ID: BYaSsvMhiDi
--HG--
extra : rebase_source : 8fe6f2aeaff5f515b7af2276c439004fa3a1f3ab
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
The fact that WorkerThread uses the mLock field from nsThread is a
pretty big hole in the nsThread abstraction. I don't see any reason
for the re-use, either. This locks protects a few fields inside
WorkerThread, but none of them seem in any way related to things
done by nsThread. This patch uses a separate lock for WorkerThreads.
MozReview-Commit-ID: 8z2SYzAjBWp
This parameter isn't used by any implementation of onDispatchedEvent,
and keeping the parameter makes later refactorings in this bug more difficult.
MozReview-Commit-ID: 90VY2vYtwCW
nsThreadManager::get() can return a reference. This lets us remove some
redundant assertions.
nsThreadArray elements can be NotNull<>s.
--HG--
extra : rebase_source : fd49010167101bc15f7f6d01bf95fd63b81d60fb
The bulk of this commit was generated with a script, executed at the top
level of a typical source code checkout. The only non-machine-generated
part was modifying MFBT's moz.build to reflect the new naming.
CLOSED TREE makes big refactorings like this a piece of cake.
# The main substitution.
find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \
xargs perl -p -i -e '
s/nsRefPtr\.h/RefPtr\.h/g; # handle includes
s/nsRefPtr ?</RefPtr</g; # handle declarations and variables
'
# Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h.
perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h
# Handle nsRefPtr.h itself, a couple places that define constructors
# from nsRefPtr, and code generators specially. We do this here, rather
# than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename
# things like nsRefPtrHashtable.
perl -p -i -e 's/nsRefPtr/RefPtr/g' \
mfbt/nsRefPtr.h \
xpcom/glue/nsCOMPtr.h \
xpcom/base/OwningNonNull.h \
ipc/ipdl/ipdl/lower.py \
ipc/ipdl/ipdl/builtin.py \
dom/bindings/Codegen.py \
python/lldbutils/lldbutils/utils.py
# In our indiscriminate substitution above, we renamed
# nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up.
find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \
xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g'
if [ -d .git ]; then
git mv mfbt/nsRefPtr.h mfbt/RefPtr.h
else
hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h
fi
--HG--
rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
This is motivated by three separate but related problems:
1. Our concept of recursion depth is broken for things that run from AfterProcessNextEvent observers (e.g. Promises). We decrement the recursionDepth counter before firing observers, so a Promise callback running at the lowest event loop depth has a recursion depth of 0 (whereas a regular nsIRunnable would be 1). This is a problem because it's impossible to distinguish a Promise running after a sync XHR's onreadystatechange handler from a top-level event (since the former runs with depth 2 - 1 = 1, and the latter runs with just 1).
2. The nsIThreadObserver mechanism that is used by a lot of code to run "after" the current event is a poor fit for anything that runs script. First, the order the observers fire in is the order they were added, not anything fixed by spec. Additionally, running script can cause the event loop to spin, which is a big source of pain here (bholley has some nasty bug caused by this).
3. We run Promises from different points in the code for workers and main thread. The latter runs from XPConnect's nsIThreadObserver callbacks, while the former runs from a hardcoded call to run Promises in the worker event loop. What workers do is particularly problematic because it means we can't get the right recursion depth no matter what we do to nsThread.
The solve this, this patch does the following:
1. Consolidate some handling of microtasks and all handling of stable state from appshell and WorkerPrivate into CycleCollectedJSRuntime.
2. Make the recursionDepth counter only available to CycleCollectedJSRuntime (and its consumers) and remove it from the nsIThreadInternal and nsIThreadObserver APIs.
3. Adjust the recursionDepth counter so that microtasks run with the recursionDepth of the task they are associated with.
4. Introduce the concept of metastable state to replace appshell's RunBeforeNextEvent. Metastable state is reached after every microtask or task is completed. This provides the semantics that bent and I want for IndexedDB, where transactions autocommit at the end of a microtask and do not "spill" from one microtask into a subsequent microtask. This differs from appshell's RunBeforeNextEvent in two ways:
a) It fires between microtasks, which was the motivation for starting this.
b) It no longer ensures that we're at the same event loop depth in the native event queue. bent decided we don't care about this.
5. Reorder stable state to happen after microtasks such as Promises, per HTML. Right now we call the regular thread observers, including appshell, before the main thread observer (XPConnect), so stable state tasks happen before microtasks.
The bulk of this commit was generated by running:
run-clang-tidy.py \
-checks='-*,llvm-namespace-comment' \
-header-filter=^/.../mozilla-central/.* \
-fix
The current name reads to me like a boolean variable, even though it's
actually a counter. Try to make that property more explicit at its uses
by renaming it to something more evocative of counter-ness.
