未验证 提交 3ca7328f 编写于 作者: P PuQing 提交者: GitHub

[PHI decoupling] move "thread pool" from fluid to phi (#48075)

* move threadpool

fix cmake

* fix make
上级 468f8815
......@@ -225,10 +225,6 @@ cc_test(
SRCS reader_test.cc
DEPS reader)
cc_library(
threadpool
SRCS threadpool.cc
DEPS enforce)
cc_test(
threadpool_test
SRCS threadpool_test.cc
......
......@@ -155,7 +155,7 @@ void Scope::DeleteScope(Scope* scope) const {
if (FLAGS_benchmark || FLAGS_eager_delete_scope) {
delete scope;
} else {
Async([scope] { delete scope; });
phi::Async([scope] { delete scope; });
}
}
}
......
......@@ -14,138 +14,16 @@ limitations under the License. */
#pragma once
#include <condition_variable> // NOLINT
#include <functional>
#include <future> // NOLINT
#include <memory>
#include <mutex> // NOLINT
#include <queue>
#include <thread> // NOLINT
#include <utility>
#include <vector>
#include "glog/logging.h"
#include "paddle/fluid/platform/enforce.h"
#include "paddle/fluid/platform/macros.h" // for DISABLE_COPY_AND_ASSIGN
#include "paddle/phi/core/threadpool.h"
namespace paddle {
namespace framework {
struct ExceptionHandler {
mutable std::future<std::unique_ptr<platform::EnforceNotMet>> future_;
explicit ExceptionHandler(
std::future<std::unique_ptr<platform::EnforceNotMet>>&& f)
: future_(std::move(f)) {}
void operator()() const {
auto ex = this->future_.get();
if (ex != nullptr) {
PADDLE_THROW(platform::errors::Fatal(
"The exception is thrown inside the thread pool. You "
"should use RunAndGetException to handle the exception."
"The exception is:\n %s.",
ex->what()));
}
}
};
// ThreadPool maintains a queue of tasks, and runs them using a fixed
// number of threads.
class ThreadPool {
public:
explicit ThreadPool(int num_threads);
using Task = std::packaged_task<std::unique_ptr<platform::EnforceNotMet>()>;
// Returns the singleton of ThreadPool.
static ThreadPool* GetInstance();
~ThreadPool();
// Run pushes a function to the task queue and returns a std::future
// object. To wait for the completion of the task, call
// std::future::wait().
template <typename Callback>
std::future<void> Run(Callback fn) {
auto f = this->RunAndGetException(fn);
return std::async(std::launch::deferred, ExceptionHandler(std::move(f)));
}
template <typename Callback>
std::future<std::unique_ptr<platform::EnforceNotMet>> RunAndGetException(
Callback fn) {
Task task([fn]() -> std::unique_ptr<platform::EnforceNotMet> {
try {
fn();
} catch (platform::EnforceNotMet& ex) {
return std::unique_ptr<platform::EnforceNotMet>(
new platform::EnforceNotMet(ex));
} catch (const std::exception& e) {
PADDLE_THROW(platform::errors::Fatal(
"Unexpected exception is catched in thread pool. All "
"throwable exception in Paddle should be an EnforceNotMet."
"The exception is:\n %s.",
e.what()));
}
return nullptr;
});
std::future<std::unique_ptr<platform::EnforceNotMet>> f = task.get_future();
{
std::unique_lock<std::mutex> lock(mutex_);
if (!running_) {
PADDLE_THROW(platform::errors::Unavailable(
"Task is enqueued into stopped ThreadPool."));
}
tasks_.push(std::move(task));
}
scheduled_.notify_one();
return f;
}
private:
DISABLE_COPY_AND_ASSIGN(ThreadPool);
// The constructor starts threads to run TaskLoop, which retrieves
// and runs tasks from the queue.
void TaskLoop();
// Init is called by GetInstance.
static void Init();
private:
static std::unique_ptr<ThreadPool> threadpool_;
static std::once_flag init_flag_;
std::vector<std::unique_ptr<std::thread>> threads_;
std::queue<Task> tasks_;
std::mutex mutex_;
bool running_;
std::condition_variable scheduled_;
};
class ThreadPoolIO : ThreadPool {
public:
static ThreadPool* GetInstanceIO();
static void InitIO();
private:
// NOTE: threadpool in base will be inhereted here.
static std::unique_ptr<ThreadPool> io_threadpool_;
static std::once_flag io_init_flag_;
};
using ExceptionHandler = phi::ExceptionHandler;
// Run a function asynchronously.
// NOTE: The function must return void. If the function need to return a value,
// you can use lambda to capture a value pointer.
template <typename Callback>
std::future<void> Async(Callback callback) {
return ThreadPool::GetInstance()->Run(callback);
}
using ThreadPool = phi::ThreadPool;
template <typename Callback>
std::future<void> AsyncIO(Callback callback) {
return ThreadPoolIO::GetInstanceIO()->Run(callback);
}
using ThreadPoolIO = phi::ThreadPoolIO;
} // namespace framework
} // namespace paddle
......@@ -26,7 +26,7 @@ void do_sum(std::vector<std::future<void>>* fs,
int cnt) {
for (int i = 0; i < cnt; ++i) {
std::lock_guard<std::mutex> l(*mu);
fs->push_back(framework::Async([sum]() { sum->fetch_add(1); }));
fs->push_back(phi::Async([sum]() { sum->fetch_add(1); }));
}
}
......
......@@ -43,6 +43,10 @@ cc_library(
lod_utils
SRCS lod_utils.cc
DEPS phi_enforce)
cc_library(
threadpool
SRCS threadpool.cc
DEPS phi_enforce)
cc_library(
dense_tensor
......
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "paddle/fluid/framework/threadpool.h"
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "paddle/phi/core/threadpool.h"
#include <thread>
#include "gflags/gflags.h"
#include "glog/logging.h"
#include "paddle/fluid/platform/enforce.h"
#include "paddle/phi/core/enforce.h"
DECLARE_int32(dist_threadpool_size);
DEFINE_int32(io_threadpool_size,
100,
"number of threads used for doing IO, default 100");
DECLARE_int32(dist_threadpool_size);
namespace phi {
namespace paddle {
namespace framework {
std::unique_ptr<ThreadPool> ThreadPool::threadpool_(nullptr);
std::once_flag ThreadPool::init_flag_;
......@@ -47,7 +46,7 @@ void ThreadPool::Init() {
PADDLE_ENFORCE_GT(
num_threads,
0,
platform::errors::InvalidArgument("The number of threads is 0."));
phi::errors::InvalidArgument("The number of threads is 0."));
threadpool_.reset(new ThreadPool(num_threads));
}
}
......@@ -88,8 +87,8 @@ void ThreadPool::TaskLoop() {
}
if (tasks_.empty()) {
PADDLE_THROW(platform::errors::Unavailable(
"Current thread has no task to Run."));
PADDLE_THROW(
phi::errors::Unavailable("Current thread has no task to Run."));
}
// pop a task from the task queue
......@@ -115,6 +114,4 @@ void ThreadPoolIO::InitIO() {
io_threadpool_.reset(new ThreadPool(FLAGS_io_threadpool_size));
}
}
} // namespace framework
} // namespace paddle
} // namespace phi
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include <condition_variable> // NOLINT
#include <functional>
#include <future> // NOLINT
#include <memory>
#include <mutex> // NOLINT
#include <queue>
#include <thread> // NOLINT
#include <utility>
#include <vector>
#include "paddle/phi/core/enforce.h"
#include "paddle/phi/core/macros.h" // for DISABLE_COPY_AND_ASSIGN
namespace phi {
struct ExceptionHandler {
mutable std::future<std::unique_ptr<phi::enforce::EnforceNotMet>> future_;
explicit ExceptionHandler(
std::future<std::unique_ptr<phi::enforce::EnforceNotMet>>&& f)
: future_(std::move(f)) {}
void operator()() const {
auto ex = this->future_.get();
if (ex != nullptr) {
PADDLE_THROW(phi::errors::Fatal(
"The exception is thrown inside the thread pool. You "
"should use RunAndGetException to handle the exception."
"The exception is:\n %s.",
ex->what()));
}
}
};
// ThreadPool maintains a queue of tasks, and runs them using a fixed
// number of threads.
class ThreadPool {
public:
explicit ThreadPool(int num_threads);
using Task =
std::packaged_task<std::unique_ptr<phi::enforce::EnforceNotMet>()>;
// Returns the singleton of ThreadPool.
static ThreadPool* GetInstance();
~ThreadPool();
// Run pushes a function to the task queue and returns a std::future
// object. To wait for the completion of the task, call
// std::future::wait().
template <typename Callback>
std::future<void> Run(Callback fn) {
auto f = this->RunAndGetException(fn);
return std::async(std::launch::deferred, ExceptionHandler(std::move(f)));
}
template <typename Callback>
std::future<std::unique_ptr<phi::enforce::EnforceNotMet>> RunAndGetException(
Callback fn) {
Task task([fn]() -> std::unique_ptr<phi::enforce::EnforceNotMet> {
try {
fn();
} catch (phi::enforce::EnforceNotMet& ex) {
return std::unique_ptr<phi::enforce::EnforceNotMet>(
new phi::enforce::EnforceNotMet(ex));
} catch (const std::exception& e) {
PADDLE_THROW(phi::errors::Fatal(
"Unexpected exception is catched in thread pool. All "
"throwable exception in Paddle should be an EnforceNotMet."
"The exception is:\n %s.",
e.what()));
}
return nullptr;
});
std::future<std::unique_ptr<phi::enforce::EnforceNotMet>> f =
task.get_future();
{
std::unique_lock<std::mutex> lock(mutex_);
if (!running_) {
PADDLE_THROW(phi::errors::Unavailable(
"Task is enqueued into stopped ThreadPool."));
}
tasks_.push(std::move(task));
}
scheduled_.notify_one();
return f;
}
private:
DISABLE_COPY_AND_ASSIGN(ThreadPool);
// The constructor starts threads to run TaskLoop, which retrieves
// and runs tasks from the queue.
void TaskLoop();
// Init is called by GetInstance.
static void Init();
private:
static std::unique_ptr<ThreadPool> threadpool_;
static std::once_flag init_flag_;
std::vector<std::unique_ptr<std::thread>> threads_;
std::queue<Task> tasks_;
std::mutex mutex_;
bool running_;
std::condition_variable scheduled_;
};
class ThreadPoolIO : ThreadPool {
public:
static ThreadPool* GetInstanceIO();
static void InitIO();
private:
// NOTE: threadpool in base will be inhereted here.
static std::unique_ptr<ThreadPool> io_threadpool_;
static std::once_flag io_init_flag_;
};
// Run a function asynchronously.
// NOTE: The function must return void. If the function need to return a value,
// you can use lambda to capture a value pointer.
template <typename Callback>
std::future<void> Async(Callback callback) {
return ThreadPool::GetInstance()->Run(callback);
}
template <typename Callback>
std::future<void> AsyncIO(Callback callback) {
return ThreadPoolIO::GetInstanceIO()->Run(callback);
}
} // namespace phi
......@@ -15,10 +15,10 @@
#include "paddle/phi/kernels/selected_rows/adam_kernel.h"
#include "paddle/fluid/framework/tensor_util.h"
#include "paddle/fluid/framework/threadpool.h"
#include "paddle/phi/backends/cpu/cpu_context.h"
#include "paddle/phi/core/kernel_registry.h"
#include "paddle/phi/core/tensor_utils.h"
#include "paddle/phi/core/threadpool.h"
#include "paddle/phi/kernels/funcs/adam_functors.h"
#include "paddle/phi/kernels/funcs/selected_rows_functor.h"
......@@ -201,12 +201,12 @@ void AdamDenseParamSparseGradKernel(
if (end > static_cast<int64_t>(param_row_count)) {
end = static_cast<int64_t>(param_row_count);
}
fs.push_back(paddle::framework::Async([&functor,
&row_id_to_grad_row_offset,
&grad_data,
row_numel,
start,
end]() {
fs.push_back(phi::Async([&functor,
&row_id_to_grad_row_offset,
&grad_data,
row_numel,
start,
end]() {
for (int64_t row_id = start; row_id < end; ++row_id) {
auto iter = row_id_to_grad_row_offset.find(row_id);
if (iter != row_id_to_grad_row_offset.end()) {
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册