cross-platform coroutine library in c++
License under the MIT license
Generate document with doxygen.
Doxygen file located at docs/libcopp.doxyfile .
libcopp use cmake to generate makefile and switch build tools.
- [required] GCC or Clang or VC support ISO C++ 03 and upper
- [required] cmake 3.7.0 and upper
- [optional] gtest 1.6.0 and upper (better test supported)
- [optional] Boost.Test (Boost.Test supported)
- [required] masm (in vc)
- [optional] if using gtest, pthread is required.
1. make a build directory
mkdir build
2. run cmake command
cmake <libcopp dir> [options]
options can be cmake options. such as set compile toolchains, source directory or options of libcopp that control build actions. libcopp options are listed below:
-DBUILD_SHARED_LIBS=YES|NO [default=NO] enable build dynamic library.
-DLIBCOPP_ENABLE_SEGMENTED_STACKS=YES|NO [default=NO] enable split stack supported context.(it's only availabe in linux and gcc 4.7.0 or upper)
-DLIBCOPP_ENABLE_VALGRIND=YES|NO [default=YES] enable valgrind supported context.
-DGTEST_ROOT=[path] set gtest library install prefix path
3. make libcopp
make [options]
4. run test [optional]
make run_test
5. run benchmark [optional]
make benchmark
6. install [optional]
make install
Or you can just copy include directory and libcopp.a in lib or lib64 into your project to use it.
Just include headers and linking library file of your platform to use libcopp
There is a simple example of using coroutine context below:
#include <cstdio>
#include <cstring>
#include <iostream>
#include <inttypes.h>
#include <stdint.h>
// include context header file
#include <libcopp/coroutine/coroutine_context_container.h>
// define a coroutine runner
class my_runner : public copp::detail::coroutine_runnable_base
{
public:
int operator()() {
// ... your code here ...printf("cortoutine %" PRIxPTR " exit and return %d.\n", (intptr_t)&co_obj, co_obj.get_ret_code());
copp::coroutine_context_default* addr = get_coroutine_context<copp::coroutine_context_default>();
std::cout<< "cortoutine "<< addr<< " is running."<< std::endl;
addr->yield();
std::cout<< "cortoutine "<< addr<< " is resumed."<< std::endl;
return 1;
}
};
int main() {
// create a coroutine
copp::coroutine_context_default co_obj;
std::cout<< "cortoutine "<< &co_obj<< " is created."<< std::endl;
// create a runner
my_runner runner;
// bind runner to coroutine object
co_obj.create(&runner);
// start a coroutine
co_obj.start();
// yield from runner
std::cout<< "cortoutine "<< &co_obj<< " is yield."<< std::endl;
co_obj.resume();
std::cout<< "cortoutine "<< &co_obj<< " exit and return "<< co_obj.get_ret_code()<< "."<< std::endl;
return 0;
}
And then, you can custom many function such as set your stack allocator, coroutine type and etc. by set your template parameters of coroutine context. Notice: One coroutine runner can only below to one coroutine context
There is a simple example of using coroutine task below:
#include <iostream>
// include task header file
#include <libcotask/task.h>
typedef cotask::task<> my_task_t;
int main(int argc, char* argv[]) {
// create a task using factory function [with lambda expression]
my_task_t::prt_t task = my_task_t::create([](){
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " started"<< std::endl;
cotask::this_task::get_task()->yield();
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " resumed"<< std::endl;
return 0;
});
std::cout<< "task "<< task->get_id()<< " created"<< std::endl;
// start a task
task->start();
std::cout<< "task "<< task->get_id()<< " yield"<< std::endl;
task->resume();
std::cout<< "task "<< task->get_id()<< " stoped, ready to be destroyed."<< std::endl;
return 0;
}
And then, you can custom many functions by set your macro type of coroutine and task to do some other function.
There is a simple example of using coroutine context below:
#include <cstdio>
#include <cstring>
#include <iostream>
#include <inttypes.h>
#include <stdint.h>
#include <ctime>
// include context header file
#include <libcotask/task.h>
#include <libcotask/task_manager.h>
// create a task manager
typedef cotask::task_manager<cotask::task<>::id_t> mgr_t;
mgr_t::ptr_t task_mgr = mgr_t::create();
// If you task manager to manage timeout, it's important to call tick interval
void tick() {
// the first parameter is second, and the second is nanosecond
task_mgr->tick(time(NULL), 0);
}
int main() {
// type define
typedef std::shared_ptr<cotask::task<> > task_ptr_type;
// create two coroutine task
task_ptr_type co_task = cotask::task<>::create([](){
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " started"<< std::endl;
cotask::this_task::get_task()->yield();
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " resumed"<< std::endl;
return 0;
});
task_ptr_type co_another_task = cotask::task<>::create([](){
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " started"<< std::endl;
cotask::this_task::get_task()->yield();
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " resumed"<< std::endl;
return 0;
});
int res = task_mgr->add_task(co_task, 5, 0); // add task and setup 5s for timeout
if (res < 0) {
std::cerr<< "some error: "<< res<< std::endl;
return res;
}
res = task_mgr->add_task(co_another_task); // add task without timeout
if (res < 0) {
std::cerr<< "some error: "<< res<< std::endl;
return res;
}
res = task_mgr->start(co_task->get_id());
if (res < 0) {
std::cerr<< "start task "<< co_task->get_id()<< " failed, error code: "<< res<< std::endl;
}
res = task_mgr->start(co_another_task->get_id());
if (res < 0) {
std::cerr<< "start task "<< co_another_task->get_id()<< " failed, error code: "<< res<< std::endl;
}
res = task_mgr->resume(co_task->get_id());
if (res < 0) {
std::cerr<< "resume task "<< co_task->get_id()<< " failed, error code: "<< res<< std::endl;
}
res = task_mgr->kill(co_another_task->get_id());
if (res < 0) {
std::cerr<< "kill task "<< co_another_task->get_id()<< " failed, error code: "<< res<< std::endl;
} else {
std::cout<< "kill task "<< co_another_task->get_id()<< " finished."<< std::endl;
}
return 0;
}
There is a simple example of using coroutine context below:
#include <cstdio>
#include <cstring>
#include <iostream>
#include <inttypes.h>
#include <stdint.h>
#include <ctime>
// include context header file
#include <libcopp/stack/stack_pool.h>
#include <libcotask/task.h>
// define the stack pool type
typedef copp::stack_pool<copp::allocator::default_statck_allocator> stack_pool_t;
// define how to create coroutine context
struct sample_macro_coroutine {
typedef copp::detail::coroutine_context_base coroutine_t;
typedef copp::allocator::stack_allocator_pool<stack_pool_t> stack_allocator_t;
typedef copp::detail::coroutine_context_container<coroutine_t, stack_allocator_t> coroutine_container_t;
};
// create a stack pool
static stack_pool_t::ptr_t global_stack_pool = stack_pool_t::create();
// define how to task allocator
struct sample_task_allocator : public cotask::macro_task::task_allocator_t {
public:
#if defined(COPP_MACRO_ENABLE_VARIADIC_TEMPLATE) && COPP_MACRO_ENABLE_VARIADIC_TEMPLATE
template <typename Ty, typename... TARGS>
static std::shared_ptr<Ty> allocate(Ty *t, TARGS COPP_MACRO_RV_REF... args) {
std::shared_ptr<Ty> ret = cotask::macro_task::task_allocator_t::allocate(t, COPP_MACRO_STD_FORWARD(TARGS, args)...);
ret->get_coroutine_context().get_allocator().attach(global_stack_pool);
return COPP_MACRO_STD_MOVE(ret);
}
#else
template <typename Ty>
static std::shared_ptr<Ty> allocate(Ty *t) {
std::shared_ptr<Ty> ret = cotask::macro_task::task_allocator_t::allocate(t);
ret->get_coroutine_context().get_allocator().attach(global_stack_pool);
return COPP_MACRO_STD_MOVE(ret);
}
#endif
};
// define how to create coroutine task
struct sample_macro_task {
typedef cotask::macro_task::id_t id_t;
typedef cotask::macro_task::id_allocator_t id_allocator_t;
typedef cotask::macro_task::action_allocator_t action_allocator_t;
typedef sample_task_allocator task_allocator_t;
};
typedef cotask::task<sample_macro_coroutine, sample_macro_task> sample_task_t;
int main() {
// create two coroutine task
sample_task_t::ptr_t co_task = cotask::task<>::create([](){
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " started"<< std::endl;
cotask::this_task::get_task()->yield();
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " resumed"<< std::endl;
return 0;
});
sample_task_t::ptr_t co_another_task = cotask::task<>::create([](){
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " started"<< std::endl;
cotask::this_task::get_task()->yield();
std::cout<< "task "<< cotask::this_task::get_task()->get_id()<< " resumed"<< std::endl;
return 0;
});
if (!co_task) {
std::cerr<< "create coroutine task with stack pool failed"<< std::endl;
return 0;
}
if (!co_another_task) {
std::cerr<< "create coroutine task with stack pool failed"<< std::endl;
return 0;
}
// ..., then do anything you want to do with these tasks
return 0;
}
split stack support: if in Linux and user gcc 4.7.0 or upper, add -DLIBCOPP_ENABLE_SEGMENTED_STACKS=YES to use split stack supported context.
Please see CI output for latest benchmark report. the benchmark on Linux and macOS can be see here and the benchmark on Windows can be see here.
basic coroutine object summary
- [BOOST] merge boost.context 1.64.0
- [OPTIMIZE] add stack pool manager and unit test
- [OPTIMIZE] reduce memory fragment when allocate coroutine task and task action
- [CI] benchmark and sample will always be run in Travis CI and Appveyor CI
- [BOOST] merge boost.context 1.61.0 and use the new jump progress(see https://owent.net/2016/1270.html for detail)
- [BOOST] enable valgrind support if valgrind/valgrind.h exists
- [CXX] use cmake to detect the function of compiler
- [OPTIMIZE] using pthread key when c++11 TLS not available
- [OPTIMIZE] remove coroutine_context_safe_base.coroutine_context_base is also thread safe now
- [OPTIMIZE] remove all global variables of cotask
- [OPTIMIZE] remove std/thread.h, add noexpect if available
- [CI] CI use build matrix to test more compiler
- [BUILD] use RelWithDebInfo for default
- v0.2.0, this version is used in our server for about one year.
- add support for valgrind
- add ci configure
- merge boost.context 1.60.0
- add -fPIC, fix spin lock
- some environment do not support TLS, make these environment can compile success
v0.1.0