CondVar already calls `AUTO_PROFILER_THREAD_SLEEP`, which shouldn't be called recursively.
mozilla::Monitor also uses CondVar, so it shouldn't be surrounded by `AUTO_PROFILER_THREAD_SLEEP` either.
Depends on D72850
Differential Revision: https://phabricator.services.mozilla.com/D72851
IDLE markers help with categorizing threads in the profiler UI.
_SLEEP makes the profiler spend less time sampling threads that haven't changed.
Differential Revision: https://phabricator.services.mozilla.com/D10815
Most of the times when we automatically create nsThread wrappers for threads
that don't already have them, we don't actually need the event targets, since
those threads don't run XPCOM event loops. Aside from wasting memory, actually
creating these event loops can lead to leaks if a thread tries to dispatch a
runnable to the queue which creates a reference cycle with the thread.
Not creating the event queues for threads that don't actually need them helps
avoid those foot guns, and also makes it easier to figure out which treads
actually run XPCOM event loops.
MozReview-Commit-ID: Arck4VQqdne
Most of the times when we automatically create nsThread wrappers for threads
that don't already have them, we don't actually need the event targets, since
those threads don't run XPCOM event loops. Aside from wasting memory, actually
creating these event loops can lead to leaks if a thread tries to dispatch a
runnable to the queue which creates a reference cycle with the thread.
Not creating the event queues for threads that don't actually need them helps
avoid those foot guns, and also makes it easier to figure out which treads
actually run XPCOM event loops.
MozReview-Commit-ID: Arck4VQqdne
Most of the times when we automatically create nsThread wrappers for threads
that don't already have them, we don't actually need the event targets, since
those threads don't run XPCOM event loops. Aside from wasting memory, actually
creating these event loops can lead to leaks if a thread tries to dispatch a
runnable to the queue which creates a reference cycle with the thread.
Not creating the event queues for threads that don't actually need them helps
avoid those foot guns, and also makes it easier to figure out which treads
actually run XPCOM event loops.
MozReview-Commit-ID: Arck4VQqdne
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
CachePerfStats gathers performance data for single open, read and write operations as well as the whole cache entry opening. It maintains long term and short term average. The long term average filters out excessive values and it represents and average time for a given operation when the cache is not busy. The short term average represents the current cache speed. By comparing these two stats we know pretty quickly that the cache is getting slower and then we race the cache with network immediately without a delay. Otherwise the delay is based on the average cache entry open time.
CachePerfStats gathers performance data for single open, read and write operations as well as the whole cache entry opening. It maintains long term and short term average. The long term average filters out excessive values and it represents and average time for a given operation when the cache is not busy. The short term average represents the current cache speed. By comparing these two stats we know pretty quickly that the cache is getting slower and then we race the cache with network immediately without a delay. Otherwise the delay is based on the average cache entry open time.
CachePerfStats gathers performance data for single open, read and write operations as well as the whole cache entry opening. It maintains long term and short term average. The long term average filters out excessive values and it represents and average time for a given operation when the cache is not busy. The short term average represents the current cache speed. By comparing these two stats we know pretty quickly that the cache is getting slower and then we race the cache with network immediately without a delay. Otherwise the delay is based on the average cache entry open time.
NS_SetCurrentThreadName() is added as an alternative to PR_SetCurrentThreadName()
inside libxul. The thread names are collected in the form of crash annotation to
be processed on socorro.
MozReview-Commit-ID: 4RpAWzTuvPs
Changed |print("enum ID : uint32_t {", file=output)| to |print("enum HistogramID : uint32_t {", file=output)| at line 53 of the file |toolkit/components/telemetry/gen-histogram-enum.py|, and then replaced all the textual occurrences of |Telemetry::ID| to |Telemetry::HistogramID| and |ID| to |HistogramID| in 43 other files.
As far as I can tell, this covers all the remaining threads which we start
using PR_CreateThread, except the ones that are created inside NSPR or NSS,
and except for the Shutdown Watchdog thread in nsTerminator.cpp and the
CacheIO thread. The Shutdown Watchdog thread stays alive past leak detection
during shutdown (by design), so we'd report leaks if we profiled it. The
CacheIO thread seems to stay alive past shutdown leak detection sometimes as
well.
This adds a AutoProfilerRegister stack class for easy registering and
unregistering. There are a few places where we still call
profiler_register_thread() and profiler_unregister_thread() manually, either
because registration happens conditionally, or because there is a variable that
gets put on the stack before the AutoProfilerRegister (e.g. a dynamically
generated thread name). AutoProfilerRegister needs to be the first object on
the stack because it uses its own `this` pointer as the stack top address.
MozReview-Commit-ID: 3vwhS55Yzt
All of these things are called with the result of
NS_NewRunnableFunction, so we need to transition them over to a world
where NS_NewRunnableFunction returns something different.
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.
This makes it clearer that, unlike how SizeOf*() functions usually work, this
doesn't measure any children hanging off the array.
And do likewise for nsTObserverArray.
