12345qiupeng
/
cartographer
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Introduce separate mutex for PoseGraph3D work queue access (#1284) 

Introduce separate mutex for PoseGraph3D work queue access.

This reduces lock contention, see issue #1250.
In particular higher frequency call to AddWorkItem
don't need to take the main mutex from the froground
thread anymore.
master
danielsievers 2018-07-16 17:15:28 +02:00 committed by Wally B. Feed
parent 31d0a6acc7
commit 482f912d95
2 changed files with 48 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

68
cartographer/mapping/internal/3d/pose_graph_3d.cc
View File

@ -51,7 +51,7 @@ PoseGraph3D::PoseGraph3D(
PoseGraph3D::~PoseGraph3D() {
WaitForAllComputations();
common::MutexLocker locker(&mutex_);
common::MutexLocker locker(&work_queue_mutex_);
CHECK(work_queue_ == nullptr);
}
@ -153,7 +153,7 @@ NodeId PoseGraph3D::AddNode(
void PoseGraph3D::AddWorkItem(
const std::function<WorkItem::Result()>& work_item) {
{
common::MutexLocker locker(&mutex_);
common::MutexLocker locker(&work_queue_mutex_);
if (work_queue_ != nullptr) {
const auto now = std::chrono::steady_clock::now();
work_queue_->push_back({now, work_item});
@ -164,7 +164,7 @@ void PoseGraph3D::AddWorkItem(
}
if (work_item() == WorkItem::Result::kRunOptimization) {
{
common::MutexLocker locker(&mutex_);
common::MutexLocker locker(&work_queue_mutex_);
work_queue_ = common::make_unique<WorkQueue>();
}
constraint_builder_.WhenDone(
@ -489,7 +489,7 @@ void PoseGraph3D::HandleWorkQueue(
while (process_work_queue) {
std::function<WorkItem::Result()> work_item;
{
common::MutexLocker locker(&mutex_);
common::MutexLocker locker(&work_queue_mutex_);
if (work_queue_->empty()) {
work_queue_.reset();
return;
@ -509,40 +509,60 @@ void PoseGraph3D::HandleWorkQueue(
}
void PoseGraph3D::WaitForAllComputations() {
bool notification = false;
common::MutexLocker locker(&mutex_);
int num_trajectory_nodes;
{
common::MutexLocker locker(&mutex_);
num_trajectory_nodes = data_.num_trajectory_nodes;
}
const int num_finished_nodes_at_start =
constraint_builder_.GetNumFinishedNodes();
while (!locker.AwaitWithTimeout(
[this]() REQUIRES(mutex_) {
return ((constraint_builder_.GetNumFinishedNodes() ==
data_.num_trajectory_nodes) &&
!work_queue_);
},
common::FromSeconds(1.))) {
auto report_progress = [this, num_trajectory_nodes,
num_finished_nodes_at_start]() {
// Log progress on nodes only when we are actually processing nodes.
if (data_.num_trajectory_nodes != num_finished_nodes_at_start) {
if (num_trajectory_nodes != num_finished_nodes_at_start) {
std::ostringstream progress_info;
progress_info << "Optimizing: " << std::fixed << std::setprecision(1)
<< 100. *
(constraint_builder_.GetNumFinishedNodes() -
num_finished_nodes_at_start) /
(data_.num_trajectory_nodes -
num_finished_nodes_at_start)
(num_trajectory_nodes - num_finished_nodes_at_start)
<< "%...";
std::cout << "\r\x1b[K" << progress_info.str() << std::flush;
}
};
// First wait for the work queue to drain so that it's safe to schedule
// a WhenDone() callback.
{
common::MutexLocker locker(&work_queue_mutex_);
while (!locker.AwaitWithTimeout(
[this]() REQUIRES(work_queue_mutex_) { return work_queue_ == nullptr; },
common::FromSeconds(1.))) {
report_progress();
}
}
std::cout << "\r\x1b[KOptimizing: Done. " << std::endl;
// Now wait for any pending constraint computations to finish.
common::MutexLocker locker(&mutex_);
bool notification GUARDED_BY(mutex_) = false;
constraint_builder_.WhenDone(
[this,
&notification](const constraints::ConstraintBuilder3D::Result& result) {
common::MutexLocker locker(&mutex_);
data_.constraints.insert(data_.constraints.end(), result.begin(),
result.end());
notification = true;
});
locker.Await([&notification]() { return notification; });
&notification](const constraints::ConstraintBuilder3D::Result& result)
EXCLUDES(mutex_) {
common::MutexLocker locker(&mutex_);
data_.constraints.insert(data_.constraints.end(), result.begin(),
result.end());
notification = true;
});
while (!locker.AwaitWithTimeout([&notification]()
REQUIRES(mutex_) { return notification; },
common::FromSeconds(1.))) {
report_progress();
}
CHECK_EQ(constraint_builder_.GetNumFinishedNodes(), num_trajectory_nodes);
std::cout << "\r\x1b[KOptimizing: Done. " << std::endl;
}
void PoseGraph3D::DeleteTrajectory(const int trajectory_id) {

7
cartographer/mapping/internal/3d/pose_graph_3d.h
View File

@ -162,7 +162,7 @@ class PoseGraph3D : public PoseGraph {
// Handles a new work item.
void AddWorkItem(const std::function<WorkItem::Result()>& work_item)
EXCLUDES(mutex_);
EXCLUDES(mutex_) EXCLUDES(work_queue_mutex_);
// Adds connectivity and sampler for a trajectory if it does not exist.
void AddTrajectoryIfNeeded(int trajectory_id) REQUIRES(mutex_);
@ -197,7 +197,7 @@ class PoseGraph3D : public PoseGraph {
// Runs the optimization, executes the trimmers and processes the work queue.
void HandleWorkQueue(const constraints::ConstraintBuilder3D::Result& result)
EXCLUDES(mutex_);
EXCLUDES(mutex_) EXCLUDES(work_queue_mutex_);
// Runs the optimization. Callers have to make sure, that there is only one
// optimization being run at a time.
@ -229,10 +229,11 @@ class PoseGraph3D : public PoseGraph {
const proto::PoseGraphOptions options_;
GlobalSlamOptimizationCallback global_slam_optimization_callback_;
mutable common::Mutex mutex_;
common::Mutex work_queue_mutex_;
// If it exists, further work items must be added to this queue, and will be
// considered later.
std::unique_ptr<WorkQueue> work_queue_ GUARDED_BY(mutex_);
std::unique_ptr<WorkQueue> work_queue_ GUARDED_BY(work_queue_mutex_);
// We globally localize a fraction of the nodes from each trajectory.
std::unordered_map<int, std::unique_ptr<common::FixedRatioSampler>>