/** * @file llcoros.h * @author Nat Goodspeed * @date 2009-06-02 * @brief Manage running boost::coroutine instances * * $LicenseInfo:firstyear=2009&license=viewerlgpl$ * Second Life Viewer Source Code * Copyright (C) 2010, Linden Research, Inc. * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; * version 2.1 of the License only. * * This library is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA * * Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA * $/LicenseInfo$ */ #if ! defined(LL_LLCOROS_H) #define LL_LLCOROS_H #include #include "llsingleton.h" #include #include #include #include #include // forward-declare helper class namespace llcoro { class Suspending; } /** * Registry of named Boost.Coroutine instances * * The Boost.Coroutine library supports the general case of a coroutine * accepting arbitrary parameters and yielding multiple (sets of) results. For * such use cases, it's natural for the invoking code to retain the coroutine * instance: the consumer repeatedly calls into the coroutine, perhaps passing * new parameter values, prompting it to yield its next result. * * Our typical coroutine usage is different, though. For us, coroutines * provide an alternative to the @c Responder pattern. Our typical coroutine * has @c void return, invoked in fire-and-forget mode: the handler for some * user gesture launches the coroutine and promptly returns to the main loop. * The coroutine initiates some action that will take multiple frames (e.g. a * capability request), waits for its result, processes it and silently steals * away. * * This usage poses two (related) problems: * * # Who should own the coroutine instance? If it's simply local to the * handler code that launches it, return from the handler will destroy the * coroutine object, terminating the coroutine. * # Once the coroutine terminates, in whatever way, who's responsible for * cleaning up the coroutine object? * * LLCoros is a Singleton collection of currently-active coroutine instances. * Each has a name. You ask LLCoros to launch a new coroutine with a suggested * name prefix; from your prefix it generates a distinct name, registers the * new coroutine and returns the actual name. * * The name can be used to kill off the coroutine prematurely, if needed. It * can also provide diagnostic info: we can look up the name of the * currently-running coroutine. * * Finally, the next frame ("mainloop" event) after the coroutine terminates, * LLCoros will notice its demise and destroy it. */ class LL_COMMON_API LLCoros: public LLSingleton { public: /// Canonical boost::dcoroutines::coroutine signature we use typedef boost::dcoroutines::coroutine coro; /// Canonical callable type typedef boost::function callable_t; /** * Create and start running a new coroutine with specified name. The name * string you pass is a suggestion; it will be tweaked for uniqueness. The * actual name is returned to you. * * Usage looks like this, for (e.g.) two coroutine parameters: * @code * class MyClass * { * public: * ... * // Do NOT NOT NOT accept reference params! * // Pass by value only! * void myCoroutineMethod(std::string, LLSD); * ... * }; * ... * std::string name = LLCoros::instance().launch( * "mycoro", boost::bind(&MyClass::myCoroutineMethod, this, * "somestring", LLSD(17)); * @endcode * * Your function/method can accept any parameters you want -- but ONLY BY * VALUE! Reference parameters are a BAD IDEA! You Have Been Warned. See * DEV-32777 comments for an explanation. * * Pass a nullary callable. It works to directly pass a nullary free * function (or static method); for all other cases use boost::bind(). Of * course, for a non-static class method, the first parameter must be the * class instance. Any other parameters should be passed via the bind() * expression. * * launch() tweaks the suggested name so it won't collide with any * existing coroutine instance, creates the coroutine instance, registers * it with the tweaked name and runs it until its first wait. At that * point it returns the tweaked name. */ std::string launch(const std::string& prefix, const callable_t& callable); /** * Abort a running coroutine by name. Normally, when a coroutine either * runs to completion or terminates with an exception, LLCoros quietly * cleans it up. This is for use only when you must explicitly interrupt * one prematurely. Returns @c true if the specified name was found and * still running at the time. */ bool kill(const std::string& name); /** * From within a coroutine, look up the (tweaked) name string by which * this coroutine is registered. Returns the empty string if not found * (e.g. if the coroutine was launched by hand rather than using * LLCoros::launch()). */ std::string getName() const; /// for delayed initialization void setStackSize(S32 stacksize); /// get the current coro::self& for those who really really care static coro::self& get_self(); private: LLCoros(); friend class LLSingleton; friend class llcoro::Suspending; std::string generateDistinctName(const std::string& prefix) const; bool cleanup(const LLSD&); struct CoroData; static void no_cleanup(CoroData*); static void toplevel(coro::self& self, CoroData* data, const callable_t& callable); S32 mStackSize; // coroutine-local storage, as it were: one per coro we track struct CoroData { CoroData(CoroData* prev, const std::string& name, const callable_t& callable, S32 stacksize); // The boost::dcoroutines library supports asymmetric coroutines. Every // time we context switch out of a coroutine, we pass control to the // previously-active one (or to the non-coroutine stack owned by the // thread). So our management of the "current" coroutine must be able to // restore the previous value when we're about to switch away. CoroData* mPrev; // tweaked name of the current coroutine const std::string mName; // the actual coroutine instance LLCoros::coro mCoro; // When the dcoroutine library calls a top-level callable, it implicitly // passes coro::self& as the first parameter. All our consumer code used // to explicitly pass coro::self& down through all levels of call stack, // because at the leaf level we need it for context-switching. But since // coroutines are based on cooperative switching, we can cause the // top-level entry point to stash a pointer to the currently-running // coroutine, and manage it appropriately as we switch out and back in. // That eliminates the need to pass it as an explicit parameter down // through every level, which is unfortunately viral in nature. Finding it // implicitly rather than explicitly allows minor maintenance in which a // leaf-level function adds a new async I/O call that suspends the calling // coroutine, WITHOUT having to propagate coro::self& through every // function signature down to that point -- and of course through every // other caller of every such function. LLCoros::coro::self* mSelf; }; typedef boost::ptr_map CoroMap; CoroMap mCoros; // identify the current coroutine's CoroData static boost::thread_specific_ptr sCurrentCoro; }; namespace llcoro { /// Instantiate one of these in a block surrounding any leaf point when /// control literally switches away from this coroutine. class Suspending { public: Suspending(); ~Suspending(); private: LLCoros::CoroData* mSuspended; }; } // namespace llcoro #endif /* ! defined(LL_LLCOROS_H) */