gecko-dev/gfx/ipc/CanvasRenderThread.cpp
Lee Salzman 3df91287ff Bug 1888338 - Use a single SharedContextWebgl. r=aosmond
This shares a global SharedContextWebgl among all instances of CanvasTranslator.
The goal is that regardless of how many windows are open, we only have to pay the
startup costs and shader compilation times for SharedContextWebgl once. In the
event that all CanvasTranslators are gone, the SharedContextWebgl is kept around
while its internal caches and textures are discarded to avoid significant memory
usage when no canvases are in use, while at the same time saving on startup
costs the next time a first live CanvasTranslator is created.

Differential Revision: https://phabricator.services.mozilla.com/D205977
2024-03-28 17:33:58 +00:00

283 lines
9.5 KiB
C++

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* 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/. */
#include "CanvasRenderThread.h"
#include "mozilla/BackgroundHangMonitor.h"
#include "mozilla/SharedThreadPool.h"
#include "mozilla/StaticPrefs_gfx.h"
#include "mozilla/TaskQueue.h"
#include "mozilla/gfx/CanvasManagerParent.h"
#include "mozilla/gfx/gfxVars.h"
#include "mozilla/layers/CanvasTranslator.h"
#include "mozilla/layers/CompositorThread.h"
#include "mozilla/webrender/RenderThread.h"
#include "nsThread.h"
#include "prsystem.h"
#include "transport/runnable_utils.h"
bool NS_IsInCanvasThreadOrWorker() {
return mozilla::gfx::CanvasRenderThread::IsInCanvasRenderOrWorkerThread();
}
namespace mozilla::gfx {
static StaticRefPtr<CanvasRenderThread> sCanvasRenderThread;
static mozilla::BackgroundHangMonitor* sBackgroundHangMonitor;
#ifdef DEBUG
static bool sCanvasRenderThreadEverStarted = false;
#endif
CanvasRenderThread::CanvasRenderThread(nsCOMPtr<nsIThread>&& aThread,
nsCOMPtr<nsIThreadPool>&& aWorkers,
bool aCreatedThread)
: mMutex("CanvasRenderThread::mMutex"),
mThread(std::move(aThread)),
mWorkers(std::move(aWorkers)),
mCreatedThread(aCreatedThread) {}
CanvasRenderThread::~CanvasRenderThread() = default;
// static
void CanvasRenderThread::Start() {
MOZ_ASSERT(NS_IsMainThread());
MOZ_ASSERT(!sCanvasRenderThread);
#ifdef DEBUG
// Check to ensure nobody will try to ever start us more than once during
// the process' lifetime (in particular after Stop).
MOZ_ASSERT(!sCanvasRenderThreadEverStarted);
sCanvasRenderThreadEverStarted = true;
#endif
int32_t threadPref =
StaticPrefs::gfx_canvas_remote_worker_threads_AtStartup();
uint32_t threadLimit;
if (threadPref < 0) {
// Given that the canvas workers are receiving instructions from
// content processes, it probably doesn't make sense to have more than
// half the number of processors doing canvas drawing. We set the
// lower limit to 2, so that even on single processor systems, if
// there is more than one window with canvas drawing, the OS can
// manage the load between them.
threadLimit = std::max(2, PR_GetNumberOfProcessors() / 2);
} else {
threadLimit = uint32_t(threadPref);
}
// We don't spawn any workers if the user set the limit to 0. Instead we will
// use the CanvasRenderThread virtual thread.
nsCOMPtr<nsIThreadPool> workers;
if (threadLimit > 0) {
workers = SharedThreadPool::Get("CanvasWorkers"_ns, threadLimit);
if (NS_WARN_IF(!workers)) {
return;
}
}
nsCOMPtr<nsIThread> thread;
if (!gfxVars::SupportsThreadsafeGL()) {
thread = wr::RenderThread::GetRenderThread();
MOZ_ASSERT(thread);
} else if (!gfxVars::UseCanvasRenderThread()) {
thread = layers::CompositorThread();
MOZ_ASSERT(thread);
}
if (thread) {
sCanvasRenderThread = new CanvasRenderThread(
std::move(thread), std::move(workers), /* aCreatedThread */ false);
return;
}
// This is 4M, which is higher than the default 256K.
// Increased with bug 1753349 to accommodate the `chromium/5359` branch of
// ANGLE, which has large peak stack usage for some pathological shader
// compilations.
//
// Previously increased to 512K to accommodate Mesa in bug 1753340.
//
// Previously increased to 320K to avoid a stack overflow in the
// Intel Vulkan driver initialization in bug 1716120.
//
// Note: we only override it if it's limited already.
const uint32_t stackSize =
nsIThreadManager::DEFAULT_STACK_SIZE ? 4096 << 10 : 0;
nsresult rv = NS_NewNamedThread(
"CanvasRenderer", getter_AddRefs(thread),
NS_NewRunnableFunction(
"CanvasRender::BackgroundHangSetup",
[]() {
sBackgroundHangMonitor = new mozilla::BackgroundHangMonitor(
"CanvasRendererBHM",
/* Timeout values are powers-of-two to enable us get better
data. 128ms is chosen for transient hangs because 8Hz should
be the minimally acceptable goal for Compositor
responsiveness (normal goal is 60Hz). */
128,
/* 2048ms is chosen for permanent hangs because it's longer than
* most Compositor hangs seen in the wild, but is short enough
* to not miss getting native hang stacks. */
2048);
nsCOMPtr<nsIThread> thread = NS_GetCurrentThread();
nsThread* nsthread = static_cast<nsThread*>(thread.get());
nsthread->SetUseHangMonitor(true);
nsthread->SetPriority(nsISupportsPriority::PRIORITY_HIGH);
}),
{.stackSize = stackSize});
if (NS_WARN_IF(NS_FAILED(rv))) {
return;
}
sCanvasRenderThread = new CanvasRenderThread(
std::move(thread), std::move(workers), /* aCreatedThread */ true);
}
// static
void CanvasRenderThread::Shutdown() {
MOZ_ASSERT(NS_IsMainThread());
// It is possible we never initialized this thread in the parent process,
// because we used the GPU process instead.
if (!sCanvasRenderThread) {
MOZ_ASSERT(XRE_IsParentProcess());
return;
}
// This closes all of the IPDL actors with possibly active task queues.
CanvasManagerParent::Shutdown();
// Queue any remaining global cleanup for CanvasTranslator
layers::CanvasTranslator::Shutdown();
// Any task queues that are in the process of shutting down are tracked in
// mPendingShutdownTaskQueues. We need to block on each one until all events
// are flushed so that we can safely teardown RemoteTextureMap afterwards.
while (true) {
RefPtr<TaskQueue> taskQueue;
{
MutexAutoLock lock(sCanvasRenderThread->mMutex);
auto& pendingQueues = sCanvasRenderThread->mPendingShutdownTaskQueues;
if (pendingQueues.IsEmpty()) {
break;
}
taskQueue = pendingQueues.PopLastElement();
}
taskQueue->AwaitShutdownAndIdle();
}
bool createdThread = sCanvasRenderThread->mCreatedThread;
nsCOMPtr<nsIThread> oldThread = sCanvasRenderThread->GetCanvasRenderThread();
nsCOMPtr<nsIThreadPool> oldWorkers = sCanvasRenderThread->mWorkers;
// Ensure that we flush the CanvasRenderThread event queue before clearing our
// singleton.
NS_DispatchAndSpinEventLoopUntilComplete(
"CanvasRenderThread::Shutdown"_ns, oldThread,
NS_NewRunnableFunction("CanvasRenderThread::Shutdown", []() -> void {}));
// Null out sCanvasRenderThread before we enter synchronous Shutdown,
// from here on we are to be considered shut down for our consumers.
sCanvasRenderThread = nullptr;
if (oldWorkers) {
oldWorkers->Shutdown();
}
// We do a synchronous shutdown here while spinning the MT event loop, but
// only if we created a dedicated CanvasRender thread.
if (createdThread) {
oldThread->Shutdown();
}
}
// static
bool CanvasRenderThread::IsInCanvasRenderThread() {
return sCanvasRenderThread &&
sCanvasRenderThread->mThread == NS_GetCurrentThread();
}
/* static */ bool CanvasRenderThread::IsInCanvasWorkerThread() {
// It is possible there are no worker threads, and the worker is the same as
// the CanvasRenderThread itself.
return sCanvasRenderThread &&
((sCanvasRenderThread->mWorkers &&
sCanvasRenderThread->mWorkers->IsOnCurrentThread()) ||
(!sCanvasRenderThread->mWorkers &&
sCanvasRenderThread->mThread == NS_GetCurrentThread()));
}
/* static */ bool CanvasRenderThread::IsInCanvasRenderOrWorkerThread() {
// It is possible there are no worker threads, and the worker is the same as
// the CanvasRenderThread itself.
return sCanvasRenderThread &&
(sCanvasRenderThread->mThread == NS_GetCurrentThread() ||
(sCanvasRenderThread->mWorkers &&
sCanvasRenderThread->mWorkers->IsOnCurrentThread()));
}
// static
already_AddRefed<nsIThread> CanvasRenderThread::GetCanvasRenderThread() {
nsCOMPtr<nsIThread> thread;
if (sCanvasRenderThread) {
thread = sCanvasRenderThread->mThread;
}
return thread.forget();
}
/* static */ already_AddRefed<TaskQueue>
CanvasRenderThread::CreateWorkerTaskQueue() {
if (!sCanvasRenderThread || !sCanvasRenderThread->mWorkers) {
return nullptr;
}
return TaskQueue::Create(do_AddRef(sCanvasRenderThread->mWorkers),
"CanvasWorker")
.forget();
}
/* static */ void CanvasRenderThread::ShutdownWorkerTaskQueue(
TaskQueue* aTaskQueue) {
MOZ_ASSERT(aTaskQueue);
aTaskQueue->BeginShutdown();
if (!sCanvasRenderThread) {
MOZ_ASSERT_UNREACHABLE("No CanvasRenderThread!");
return;
}
MutexAutoLock lock(sCanvasRenderThread->mMutex);
auto& pendingQueues = sCanvasRenderThread->mPendingShutdownTaskQueues;
pendingQueues.AppendElement(aTaskQueue);
}
/* static */ void CanvasRenderThread::FinishShutdownWorkerTaskQueue(
TaskQueue* aTaskQueue) {
if (!sCanvasRenderThread) {
return;
}
MutexAutoLock lock(sCanvasRenderThread->mMutex);
sCanvasRenderThread->mPendingShutdownTaskQueues.RemoveElement(aTaskQueue);
}
/* static */ void CanvasRenderThread::Dispatch(
already_AddRefed<nsIRunnable> aRunnable) {
if (!sCanvasRenderThread) {
MOZ_DIAGNOSTIC_ASSERT(false,
"Dispatching after CanvasRenderThread shutdown!");
return;
}
sCanvasRenderThread->mThread->Dispatch(std::move(aRunnable));
}
} // namespace mozilla::gfx