/* This Source Code Form is subject to the terms of the Mozilla Public * License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
/* * GC Scheduling from Firefox * ========================== * * See also GC Scheduling from SpiderMonkey's perspective here: * https://searchfox.org/mozilla-central/source/js/src/gc/Scheduling.h * * From Firefox's perspective GCs can start in 5 different ways: * * * The JS engine just starts doing a GC for its own reasons (see above). * Firefox finds out about these via a callback in nsJSEnvironment.cpp * * PokeGC() * * PokeFullGC() * * PokeShrinkingGC() * * memory-pressure GCs (via a listener in nsJSEnvironment.cpp). * * PokeGC * ------ * * void CCGCScheduler::PokeGC(JS::GCReason aReason, JSObject* aObj, * TimeDuration aDelay) * * PokeGC provides a way for callers to say "Hey, there may be some memory * associated with this object (via Zone) you can collect." PokeGC will: * * add the zone to a set, * * set flags including what kind of GC to run (SetWantMajorGC), * * then creates the mGCRunner with a short delay. * * The delay can allow other calls to PokeGC to add their zones so they can * be collected together. * * See below for what happens when mGCRunner fires. * * PokeFullGC * ---------- * * void CCGCScheduler::PokeFullGC() * * PokeFullGC will create a timer that will initiate a "full" (all zones) * collection. This is usually used after a regular collection if a full GC * seems like a good idea (to collect inter-zone references). * * When the timer fires it will: * * set flags (SetWantMajorGC), * * start the mGCRunner with zero delay. * * See below for when mGCRunner fires. * * PokeShrinkingGC * --------------- * * void CCGCScheduler::PokeShrinkingGC() * * PokeShrinkingGC is called when Firefox's user is inactive. * Like PokeFullGC, PokeShrinkingGC uses a timer, but the timeout is longer * which should prevent the ShrinkingGC from starting if the user only * glances away for a brief time. When the timer fires it will: * * * set flags (SetWantMajorGC), * * create the mGCRunner. * * There is a check if the user is still inactive in GCRunnerFired), if the * user has become active the shrinking GC is canceled and either a regular * GC (if requested, see mWantAtLeastRegularGC) or no GC is run. * * When mGCRunner fires * -------------------- * * When mGCRunner fires it calls GCRunnerFired. This starts in the * WaitToMajorGC state: * * * If this is a parent process it jumps to the next state * * If this is a content process it will ask the parent if now is a good * time to do a GC. (MayGCNow) * * kill the mGCRunner * * Exit * * Meanwhile the parent process will queue GC requests so that not too many * are running in parallel overwhelming the CPU cores (see * IdleSchedulerParent). * * When the promise from MayGCNow is resolved it will set some * state (NoteReadyForMajorGC) and restore the mGCRunner. * * When the mGCRunner runs a second time (or this is the parent process and * which jumped over the above logic. It will be in the StartMajorGC state. * It will initiate the GC for real, usually. If it's a shrinking GC and the * user is now active again it may abort. See GCRunnerFiredDoGC(). * * The runner will then run the first slice of the garbage collection. * Later slices are also run by the runner, the final slice kills the runner * from the GC callback in nsJSEnvironment.cpp. * * There is additional logic in the code to handle concurrent requests of * various kinds.
*/
namespace mozilla {
void CCGCScheduler::NoteGCBegin(JS::GCReason aReason) { // Treat all GC as incremental here; non-incremental GC will just appear to // be one slice.
mInIncrementalGC = true;
mReadyForMajorGC = !mAskParentBeforeMajorGC;
// Tell the parent process that we've started a GC (it might not know if // we hit a threshold in the JS engine). using mozilla::ipc::IdleSchedulerChild;
IdleSchedulerChild* child = IdleSchedulerChild::GetMainThreadIdleScheduler(); if (child) {
child->StartedGC();
}
// The reason might have come from mMajorReason, mEagerMajorGCReason, or // in the case of an internally-generated GC, it might come from the // internal logic (and be passed in here). It's easier to manage a single // reason state variable, so merge all sources into mMajorGCReason.
MOZ_ASSERT(aReason != JS::GCReason::NO_REASON);
mMajorGCReason = aReason;
mEagerMajorGCReason = JS::GCReason::NO_REASON;
}
using mozilla::ipc::IdleSchedulerChild;
IdleSchedulerChild* child = IdleSchedulerChild::GetMainThreadIdleScheduler(); if (child) {
child->DoneGC();
}
}
void CCGCScheduler::NoteGCSliceEnd(TimeStamp aStart, TimeStamp aEnd) { if (mMajorGCReason == JS::GCReason::NO_REASON) { // Internally-triggered GCs do not wait for the parent's permission to // proceed. This flag won't be checked during an incremental GC anyway, // but it better reflects reality.
mReadyForMajorGC = true;
}
// Subsequent slices should be INTER_SLICE_GC unless they are triggered by // something else that provides its own reason.
mMajorGCReason = JS::GCReason::INTER_SLICE_GC;
// Compute how much GC time was spent in predicted-to-be-idle time. In the // unlikely event that the slice started after the deadline had already // passed, treat the entire slice as non-idle.
TimeDuration nonIdleDuration; bool startedIdle = mTriggeredGCDeadline.isSome() &&
!mTriggeredGCDeadline->IsNull() &&
*mTriggeredGCDeadline > aStart; if (!startedIdle) {
nonIdleDuration = sliceDuration;
} else { if (*mTriggeredGCDeadline < aEnd) { // Overran the idle deadline.
nonIdleDuration = aEnd - *mTriggeredGCDeadline;
}
}
// Note the GC_SLICE_DURING_IDLE previously had a different definition: it was // a histogram of percentages of externally-triggered slices. It is now a // histogram of percentages of all slices. That means that now you might have // a 4ms internal slice (0% during idle) followed by a 16ms external slice // (15ms during idle), whereas before this would show up as a single record of // a single slice with 75% of its time during idle (15 of 20ms).
TimeDuration idleDuration = sliceDuration - nonIdleDuration;
uint32_t percent =
uint32_t(idleDuration.ToSeconds() / sliceDuration.ToSeconds() * 100);
Telemetry::Accumulate(Telemetry::GC_SLICE_DURING_IDLE, percent);
void CCGCScheduler::NoteWontGC() {
mReadyForMajorGC = !mAskParentBeforeMajorGC;
mMajorGCReason = JS::GCReason::NO_REASON;
mEagerMajorGCReason = JS::GCReason::NO_REASON;
mWantAtLeastRegularGC = false; // Don't clear the WantFullGC state, we will do a full GC the next time a // GC happens for any other reason.
}
bool CCGCScheduler::GCRunnerFired(TimeStamp aDeadline) {
MOZ_ASSERT(!mDidShutdown, "GCRunner still alive during shutdown");
if (!aDeadline) {
mCurrentCollectionHasSeenNonIdle = true;
} elseif (mPreferFasterCollection) { // We found some idle time, try to utilize that a bit more given that // we're in a mode where idle time is rare.
aDeadline = aDeadline + TimeDuration::FromMilliseconds(5.0);
}
case GCRunnerAction::MinorGC:
JS::MaybeRunNurseryCollection(CycleCollectedJSRuntime::Get()->Runtime(),
step.mReason);
NoteMinorGCEnd(); return HasMoreIdleGCRunnerWork();
case GCRunnerAction::WaitToMajorGC: {
MOZ_ASSERT(!mHaveAskedParent, "GCRunner alive after asking the parent");
RefPtr<CCGCScheduler::MayGCPromise> mbPromise =
CCGCScheduler::MayGCNow(step.mReason); if (!mbPromise) { // We can GC now. break;
}
mHaveAskedParent = true;
KillGCRunner();
mbPromise->Then(
GetMainThreadSerialEventTarget(), __func__,
[this](bool aMayGC) {
mHaveAskedParent = false; if (aMayGC) { if (!NoteReadyForMajorGC()) { // Another GC started and maybe completed while waiting. return;
} // Recreate the GC runner with a 0 delay. The new runner will // continue in idle time.
KillGCRunner();
EnsureGCRunner(0);
} elseif (!InIncrementalGC()) { // We should kill the GC runner since we're done with it, but // only if there's no incremental GC.
KillGCRunner();
NoteWontGC();
}
},
[this](mozilla::ipc::ResponseRejectReason r) {
mHaveAskedParent = false; if (!InIncrementalGC()) {
KillGCRunner();
NoteWontGC();
}
});
returntrue;
}
case GCRunnerAction::StartMajorGC: case GCRunnerAction::GCSlice: break;
}
return GCRunnerFiredDoGC(aDeadline, step);
}
bool CCGCScheduler::GCRunnerFiredDoGC(TimeStamp aDeadline, const GCRunnerStep& aStep) { // Run a GC slice, possibly the first one of a major GC.
nsJSContext::IsShrinking is_shrinking = nsJSContext::NonShrinkingGC; if (!InIncrementalGC() && aStep.mReason == JS::GCReason::USER_INACTIVE) { bool do_gc = mWantAtLeastRegularGC;
if (!mUserIsActive) { if (!nsRefreshDriver::IsRegularRateTimerTicking()) {
mIsCompactingOnUserInactive = true;
is_shrinking = nsJSContext::ShrinkingGC;
do_gc = true;
} else { // Poke again to restart the timer.
PokeShrinkingGC();
}
}
if (!do_gc) { using mozilla::ipc::IdleSchedulerChild;
IdleSchedulerChild* child =
IdleSchedulerChild::GetMainThreadIdleScheduler(); if (child) {
child->DoneGC();
}
NoteWontGC();
KillGCRunner(); returntrue;
}
}
// Note that we are triggering the following GC slice and recording whether // it started in idle time, for use in the callback at the end of the slice.
mTriggeredGCDeadline = Some(aDeadline);
// If the GC doesn't have any more work to do on the foreground thread (and // e.g. is waiting for background sweeping to finish) then return false to // make IdleTaskRunner postpone the next call a bit.
JSContext* cx = dom::danger::GetJSContext(); return JS::IncrementalGCHasForegroundWork(cx);
}
// We ask the parent if we should GC for GCs that aren't too timely, // with the exception of MEM_PRESSURE, in that case we ask the parent // because GCing on too many processes at the same time when under // memory pressure could be a very bad experience for the user. switch (reason) { case JS::GCReason::PAGE_HIDE: case JS::GCReason::MEM_PRESSURE: case JS::GCReason::USER_INACTIVE: case JS::GCReason::FULL_GC_TIMER: case JS::GCReason::CC_FINISHED: { if (XRE_IsContentProcess()) {
IdleSchedulerChild* child =
IdleSchedulerChild::GetMainThreadIdleScheduler(); if (child) { return child->MayGCNow();
}
} // The parent process doesn't ask IdleSchedulerParent if it can GC. break;
} default: break;
}
// We use synchronous task dispatch here to avoid a trip through the event // loop if we're on the parent process or it's a GC reason that does not // require permission to GC.
RefPtr<MayGCPromise::Private> p = MakeRefPtr<MayGCPromise::Private>(__func__);
p->UseSynchronousTaskDispatch(__func__);
p->Resolve(true, __func__); return p;
}
// When we're in an incremental GC, we should always have an sGCRunner, so do // not check CC timers. The CC timers won't do anything during a GC.
MOZ_ASSERT_IF(InIncrementalGC(), mGCRunner);
if (runner) {
runner->SetIdleDeadline(aDeadline);
runner->Run();
}
}
void CCGCScheduler::PokeShrinkingGC() { if (mShrinkingGCTimer || mDidShutdown) { return;
}
NS_NewTimerWithFuncCallback(
&mShrinkingGCTimer,
[](nsITimer* aTimer, void* aClosure) {
CCGCScheduler* s = static_cast<CCGCScheduler*>(aClosure);
s->KillShrinkingGCTimer(); if (!s->mUserIsActive) { if (!nsRefreshDriver::IsRegularRateTimerTicking()) {
s->SetWantMajorGC(JS::GCReason::USER_INACTIVE); if (!s->mHaveAskedParent) {
s->EnsureGCRunner(0);
}
} else {
s->PokeShrinkingGC();
}
}
}, this, StaticPrefs::javascript_options_compact_on_user_inactive_delay(),
nsITimer::TYPE_ONE_SHOT_LOW_PRIORITY, "ShrinkingGCTimerFired");
}
void CCGCScheduler::PokeFullGC() { if (!mFullGCTimer && !mDidShutdown) {
NS_NewTimerWithFuncCallback(
&mFullGCTimer,
[](nsITimer* aTimer, void* aClosure) {
CCGCScheduler* s = static_cast<CCGCScheduler*>(aClosure);
s->KillFullGCTimer();
// Even if the GC is denied by the parent process, because we've // set that we want a full GC we will get one eventually.
s->SetNeedsFullGC();
s->SetWantMajorGC(JS::GCReason::FULL_GC_TIMER); if (s->mCCRunner) {
s->EnsureCCThenGC(CCReason::GC_WAITING);
} elseif (!s->mHaveAskedParent) {
s->EnsureGCRunner(0);
}
}, this, StaticPrefs::javascript_options_gc_delay_full(),
nsITimer::TYPE_ONE_SHOT_LOW_PRIORITY, "FullGCTimerFired");
}
}
// If a post-CC GC was pending, then we'll make sure one is happening.
mNeedsGCAfterCC = false;
if (aObj) {
JS::Zone* zone = JS::GetObjectZone(aObj);
CycleCollectedJSRuntime::Get()->AddZoneWaitingForGC(zone);
} elseif (aReason != JS::GCReason::CC_FINISHED) {
SetNeedsFullGC();
}
if (mGCRunner || mHaveAskedParent) { // There's already a GC runner, or there will be, so just return. return;
}
SetWantMajorGC(aReason);
if (mCCRunner) { // Make sure CC is called regardless of the size of the purple buffer, and // GC after it.
EnsureCCThenGC(CCReason::GC_WAITING); return;
}
// Wait for javascript.options.gc_delay (or delay_first) then start // looking for idle time to run the initial GC slice. staticbool first = true;
TimeDuration delay =
aDelay ? aDelay
: TimeDuration::FromMilliseconds(
first ? StaticPrefs::javascript_options_gc_delay_first()
: StaticPrefs::javascript_options_gc_delay());
first = false;
EnsureGCRunner(delay);
}
TimeDuration CCGCScheduler::ComputeMinimumBudgetForRunner(
TimeDuration aBaseValue) { // If the main thread was too busy to find idle for the whole last collection, // allow a very short budget this time. return mPreferFasterCollection ? TimeDuration::FromMilliseconds(1.0)
: TimeDuration::FromMilliseconds(std::max(
nsRefreshDriver::HighRateMultiplier() *
aBaseValue.ToMilliseconds(),
1.0));
}
void CCGCScheduler::EnsureGCRunner(TimeDuration aDelay) { if (mGCRunner) { return;
}
// Wait at most the interslice GC delay before forcing a run.
mGCRunner = IdleTaskRunner::Create(
[this](TimeStamp aDeadline) { return GCRunnerFired(aDeadline); }, "CCGCScheduler::EnsureGCRunner", aDelay,
TimeDuration::FromMilliseconds(
StaticPrefs::javascript_options_gc_delay_interslice()),
minimumBudget, true, [this] { return mDidShutdown; },
[this](uint32_t) {
PROFILER_MARKER_UNTYPED("GC Interrupt", GCCC);
mInterruptRequested = true;
});
}
// nsJSEnvironmentObserver observes the user-interaction-inactive notifications // and triggers a shrinking a garbage collection if the user is still inactive // after NS_SHRINKING_GC_DELAY ms later, if the appropriate pref is set. void CCGCScheduler::UserIsInactive() {
mUserIsActive = false; if (StaticPrefs::javascript_options_compact_on_user_inactive()) {
PokeShrinkingGC();
}
}
void CCGCScheduler::KillShrinkingGCTimer() { if (mShrinkingGCTimer) {
mShrinkingGCTimer->Cancel();
NS_RELEASE(mShrinkingGCTimer);
}
}
void CCGCScheduler::KillFullGCTimer() { if (mFullGCTimer) {
mFullGCTimer->Cancel();
NS_RELEASE(mFullGCTimer);
}
}
void CCGCScheduler::KillGCRunner() { // If we're in an incremental GC then killing the timer is only okay if // we're shutting down.
MOZ_ASSERT(!(InIncrementalGC() && !mDidShutdown)); if (mGCRunner) {
mGCRunner->Cancel();
mGCRunner = nullptr;
}
}
CCReason reason = ShouldScheduleCC(aNow, aSuspectedCCObjects); if (reason != CCReason::NO_REASON) { // We can kill some objects before running forgetSkippable.
nsCycleCollector_dispatchDeferredDeletion();
if (!mCCRunner) {
InitCCRunnerStateMachine(CCRunnerState::ReducePurple, reason);
}
EnsureCCRunner(kCCSkippableDelay, kForgetSkippableSliceDuration);
}
}
if (aPrevSliceEndTime.IsNull()) { // The first slice gets the standard slice time. return JS::SliceBudget(JS::TimeBudget(baseBudget));
}
// Only run a limited slice if we're within the max running time.
MOZ_ASSERT(aNow >= aCCBeginTime);
TimeDuration runningTime = aNow - aCCBeginTime; if (runningTime >= kMaxICCDuration) { return JS::SliceBudget::unlimited();
}
// Try to make up for a delay in running this slice.
MOZ_ASSERT(aNow >= aPrevSliceEndTime); double sliceDelayMultiplier =
(aNow - aPrevSliceEndTime) / kICCIntersliceDelay;
TimeDuration delaySliceBudget =
std::min(baseBudget.MultDouble(sliceDelayMultiplier), maxSlice);
// Increase slice budgets up to |maxSlice| as we approach // half way through the ICC, to avoid large sync CCs. double percentToHalfDone =
std::min(2.0 * (runningTime / kMaxICCDuration), 1.0);
TimeDuration laterSliceBudget = maxSlice.MultDouble(percentToHalfDone);
// Note: We may have already overshot the deadline, in which case // baseBudget will be negative and we will end up returning // laterSliceBudget. return JS::SliceBudget(JS::TimeBudget(
std::max({delaySliceBudget, laterSliceBudget, baseBudget})));
}
JS::SliceBudget CCGCScheduler::ComputeInterSliceGCBudget(TimeStamp aDeadline,
TimeStamp aNow) { // We use longer budgets when the CC has been locked out but the CC has // tried to run since that means we may have a significant amount of // garbage to collect and it's better to GC in several longer slices than // in a very long one.
TimeDuration budget =
aDeadline.IsNull() ? mActiveIntersliceGCBudget * 2 : aDeadline - aNow; if (!mCCBlockStart) { return CreateGCSliceBudget(budget, !aDeadline.IsNull(), false);
}
// If the budget is being extended, do not allow it to be interrupted. auto result = JS::SliceBudget(JS::TimeBudget(extendedBudget), nullptr);
result.idle = !aDeadline.IsNull();
result.extended = true; return result;
}
// Don't run consecutive CCs too often. if (mCleanupsSinceLastGC && !mLastCCEndTime.IsNull()) { if (aNow - mLastCCEndTime < kCCDelay) { return CCReason::NO_REASON;
}
}
// If GC hasn't run recently and forget skippable only cycle was run, // don't start a new cycle too soon. if ((mCleanupsSinceLastGC > kMajorForgetSkippableCalls) &&
!mLastForgetSkippableCycleEndTime.IsNull()) { if (aNow - mLastForgetSkippableCycleEndTime <
kTimeBetweenForgetSkippableCycles) { return CCReason::NO_REASON;
}
}
return IsCCNeeded(aNow, aSuspectedCCObjects);
}
CCRunnerStep CCGCScheduler::AdvanceCCRunner(TimeStamp aDeadline, TimeStamp aNow,
uint32_t aSuspectedCCObjects) { struct StateDescriptor { // When in this state, should we first check to see if we still have // enough reason to CC? bool mCanAbortCC;
// If we do decide to abort the CC, should we still try to forget // skippables one more time? bool mTryFinalForgetSkippable;
};
// The state descriptors for Inactive and Canceled will never actually be // used. We will never call this function while Inactive, and Canceled is // handled specially at the beginning.
constexpr StateDescriptor stateDescriptors[] = {
{false, false}, /* CCRunnerState::Inactive */
{false, false}, /* CCRunnerState::ReducePurple */
{true, true}, /* CCRunnerState::CleanupChildless */
{true, false}, /* CCRunnerState::CleanupContentUnbinder */
{false, false}, /* CCRunnerState::CleanupDeferred */
{false, false}, /* CCRunnerState::StartCycleCollection */
{false, false}, /* CCRunnerState::CycleCollecting */
{false, false}}; /* CCRunnerState::Canceled */
static_assert(std::size(stateDescriptors) == size_t(CCRunnerState::NumStates), "need one state descriptor per state"); const StateDescriptor& desc = stateDescriptors[int(mCCRunnerState)];
// Make sure we initialized the state machine.
MOZ_ASSERT(mCCRunnerState != CCRunnerState::Inactive);
if (mDidShutdown) { return {CCRunnerAction::StopRunning, Yield};
}
if (mCCRunnerState == CCRunnerState::Canceled) { // When we cancel a cycle, there may have been a final ForgetSkippable. return {CCRunnerAction::StopRunning, Yield};
}
if (InIncrementalGC()) { if (mCCBlockStart.IsNull()) {
BlockCC(aNow);
// If we have reached the CycleCollecting state, then ignore CC timer // fires while incremental GC is running. (Running ICC during an IGC // would cause us to synchronously finish the GC, which is bad.) // // If we have not yet started cycle collecting, then reset our state so // that we run forgetSkippable often enough before CC. Because of reduced // mCCDelay, forgetSkippable will be called just a few times. // // The kMaxCCLockedoutTime limit guarantees that we end up calling // forgetSkippable and CycleCollectNow eventually.
if (GetCCBlockedTime(aNow) < kMaxCCLockedoutTime) { return {CCRunnerAction::None, Yield};
}
// Locked out for too long, so proceed and finish the incremental GC // synchronously.
}
// For states that aren't just continuations of previous states, check // whether a CC is still needed (after doing various things to reduce the // purple buffer). if (desc.mCanAbortCC &&
IsCCNeeded(aNow, aSuspectedCCObjects) == CCReason::NO_REASON) { // If we don't pass the threshold for wanting to cycle collect, stop now // (after possibly doing a final ForgetSkippable).
mCCRunnerState = CCRunnerState::Canceled;
NoteForgetSkippableOnlyCycle(aNow);
// Preserve the previous code's idea of when to check whether a // ForgetSkippable should be fired. if (desc.mTryFinalForgetSkippable &&
ShouldForgetSkippable(aSuspectedCCObjects)) { // The Canceled state will make us StopRunning after this action is // performed (see conditional at top of function). return {CCRunnerAction::ForgetSkippable, Yield, KeepChildless};
}
switch (mCCRunnerState) { // ReducePurple: a GC ran (or we otherwise decided to try CC'ing). Wait // for some amount of time (kCCDelay, or less if incremental GC blocked // this CC) while firing regular ForgetSkippable actions before continuing // on. case CCRunnerState::ReducePurple:
++mCCRunnerEarlyFireCount; if (IsLastEarlyCCTimer(mCCRunnerEarlyFireCount)) {
mCCRunnerState = CCRunnerState::CleanupChildless;
}
if (ShouldForgetSkippable(aSuspectedCCObjects)) { return {CCRunnerAction::ForgetSkippable, Yield, KeepChildless};
}
if (aDeadline.IsNull()) { return {CCRunnerAction::None, Yield};
}
// If we're called during idle time, try to find some work to do by // advancing to the next state, effectively bypassing some possible forget // skippable calls.
mCCRunnerState = CCRunnerState::CleanupChildless;
// Continue on to CleanupChildless, but only after checking IsCCNeeded // again. return {CCRunnerAction::None, Continue};
// CleanupChildless: do a stronger ForgetSkippable that removes nodes with // no children in the cycle collector graph. This state is split into 3 // parts; the other Cleanup* actions will happen within the same callback // (unless the ForgetSkippable shrinks the purple buffer enough for the CC // to be skipped entirely.) case CCRunnerState::CleanupChildless:
mCCRunnerState = CCRunnerState::CleanupContentUnbinder; return {CCRunnerAction::ForgetSkippable, Yield, RemoveChildless};
// CleanupContentUnbinder: continuing cleanup, clear out the content // unbinder. case CCRunnerState::CleanupContentUnbinder: if (aDeadline.IsNull()) { // Non-idle (waiting) callbacks skip the rest of the cleanup, but still // wait for another fire before the actual CC.
mCCRunnerState = CCRunnerState::StartCycleCollection; return {CCRunnerAction::None, Yield};
}
// Running in an idle callback.
// The deadline passed, so go straight to CC in the next slice. if (aNow >= aDeadline) {
mCCRunnerState = CCRunnerState::StartCycleCollection; return {CCRunnerAction::None, Yield};
}
// CleanupDeferred: continuing cleanup, do deferred deletion. case CCRunnerState::CleanupDeferred:
MOZ_ASSERT(!aDeadline.IsNull(), "Should only be in CleanupDeferred state when idle");
// Our efforts to avoid a CC have failed. Let the timer fire once more // to trigger a CC.
mCCRunnerState = CCRunnerState::StartCycleCollection; if (aNow >= aDeadline) { // The deadline passed, go straight to CC in the next slice. return {CCRunnerAction::None, Yield};
}
return {CCRunnerAction::CleanupDeferred, Yield};
// StartCycleCollection: start actually doing cycle collection slices. case CCRunnerState::StartCycleCollection: // We are in the final timer fire and still meet the conditions for // triggering a CC. Let RunCycleCollectorSlice finish the current IGC if // any, because that will allow us to include the GC time in the CC pause.
mCCRunnerState = CCRunnerState::CycleCollecting;
[[fallthrough]];
// CycleCollecting: continue running slices until done. case CCRunnerState::CycleCollecting: {
CCRunnerStep step{CCRunnerAction::CycleCollect, Yield};
step.mParam.mCCReason = mCCReason;
mCCReason = CCReason::SLICE; // Set reason for following slices. return step;
}
// Service a non-eager GC request first, even if it requires waiting. if (mMajorGCReason != JS::GCReason::NO_REASON) { return {mReadyForMajorGC ? GCRunnerAction::StartMajorGC
: GCRunnerAction::WaitToMajorGC,
mMajorGCReason};
}
// Now for eager requests, which are ignored unless we're idle. if (!aDeadline.IsNull()) { if (mEagerMajorGCReason != JS::GCReason::NO_REASON) { return {mReadyForMajorGC ? GCRunnerAction::StartMajorGC
: GCRunnerAction::WaitToMajorGC,
mEagerMajorGCReason};
}
if (mEagerMinorGCReason != JS::GCReason::NO_REASON) { return {GCRunnerAction::MinorGC, mEagerMinorGCReason};
}
}
// If we had forget skippables only at the beginning of the interval, we // still want to use the whole time, minute or more, for frequency // calculation. mLastForgetSkippableEndTime is needed if forget skippable // takes enough time to push the interval to be over a minute.
TimeStamp endPoint = std::max(startPlusMinute, mLastForgetSkippableEndTime);
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung ist noch experimentell.