/* Copyright 2005-2014 Intel Corporation. All Rights Reserved. This file is part of Threading Building Blocks. Threading Building Blocks is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License version 2 as published by the Free Software Foundation. Threading Building Blocks 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 General Public License for more details. You should have received a copy of the GNU General Public License along with Threading Building Blocks; if not, write to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA As a special exception, you may use this file as part of a free software library without restriction. Specifically, if other files instantiate templates or use macros or inline functions from this file, or you compile this file and link it with other files to produce an executable, this file does not by itself cause the resulting executable to be covered by the GNU General Public License. This exception does not however invalidate any other reasons why the executable file might be covered by the GNU General Public License. */ #ifndef __TBB__aggregator_H #define __TBB__aggregator_H #if !TBB_PREVIEW_AGGREGATOR #error Set TBB_PREVIEW_AGGREGATOR before including aggregator.h #endif #include "atomic.h" #include "tbb_profiling.h" namespace tbb { namespace interface6 { using namespace tbb::internal; class aggregator_operation { template friend class aggregator_ext; uintptr_t status; aggregator_operation* my_next; public: enum aggregator_operation_status { agg_waiting=0, agg_finished }; aggregator_operation() : status(agg_waiting), my_next(NULL) {} /// Call start before handling this operation void start() { call_itt_notify(acquired, &status); } /// Call finish when done handling this operation /** The operation will be released to its originating thread, and possibly deleted. */ void finish() { itt_store_word_with_release(status, uintptr_t(agg_finished)); } aggregator_operation* next() { return itt_hide_load_word(my_next);} void set_next(aggregator_operation* n) { itt_hide_store_word(my_next, n); } }; namespace internal { class basic_operation_base : public aggregator_operation { friend class basic_handler; virtual void apply_body() = 0; public: basic_operation_base() : aggregator_operation() {} virtual ~basic_operation_base() {} }; template class basic_operation : public basic_operation_base, no_assign { const Body& my_body; /*override*/ void apply_body() { my_body(); } public: basic_operation(const Body& b) : basic_operation_base(), my_body(b) {} }; class basic_handler { public: basic_handler() {} void operator()(aggregator_operation* op_list) const { while (op_list) { // ITT note: &(op_list->status) tag is used to cover accesses to the operation data. // The executing thread "acquires" the tag (see start()) and then performs // the associated operation w/o triggering a race condition diagnostics. // A thread that created the operation is waiting for its status (see execute_impl()), // so when this thread is done with the operation, it will "release" the tag // and update the status (see finish()) to give control back to the waiting thread. basic_operation_base& request = static_cast(*op_list); // IMPORTANT: need to advance op_list to op_list->next() before calling request.finish() op_list = op_list->next(); request.start(); request.apply_body(); request.finish(); } } }; } // namespace internal //! Aggregator base class and expert interface /** An aggregator for collecting operations coming from multiple sources and executing them serially on a single thread. */ template class aggregator_ext : tbb::internal::no_copy { public: aggregator_ext(const handler_type& h) : handler_busy(0), handle_operations(h) { mailbox = NULL; } //! EXPERT INTERFACE: Enter a user-made operation into the aggregator's mailbox. /** Details of user-made operations must be handled by user-provided handler */ void process(aggregator_operation *op) { execute_impl(*op); } protected: /** Place operation in mailbox, then either handle mailbox or wait for the operation to be completed by a different thread. */ void execute_impl(aggregator_operation& op) { aggregator_operation* res; // ITT note: &(op.status) tag is used to cover accesses to this operation. This // thread has created the operation, and now releases it so that the handler // thread may handle the associated operation w/o triggering a race condition; // thus this tag will be acquired just before the operation is handled in the // handle_operations functor. call_itt_notify(releasing, &(op.status)); // insert the operation in the queue do { // ITT may flag the following line as a race; it is a false positive: // This is an atomic read; we don't provide itt_hide_load_word for atomics op.my_next = res = mailbox; // NOT A RACE } while (mailbox.compare_and_swap(&op, res) != res); if (!res) { // first in the list; handle the operations // ITT note: &mailbox tag covers access to the handler_busy flag, which this // waiting handler thread will try to set before entering handle_operations. call_itt_notify(acquired, &mailbox); start_handle_operations(); __TBB_ASSERT(op.status, NULL); } else { // not first; wait for op to be ready call_itt_notify(prepare, &(op.status)); spin_wait_while_eq(op.status, uintptr_t(aggregator_operation::agg_waiting)); itt_load_word_with_acquire(op.status); } } private: //! An atomically updated list (aka mailbox) of aggregator_operations atomic mailbox; //! Controls thread access to handle_operations /** Behaves as boolean flag where 0=false, 1=true */ uintptr_t handler_busy; handler_type handle_operations; //! Trigger the handling of operations when the handler is free void start_handle_operations() { aggregator_operation *pending_operations; // ITT note: &handler_busy tag covers access to mailbox as it is passed // between active and waiting handlers. Below, the waiting handler waits until // the active handler releases, and the waiting handler acquires &handler_busy as // it becomes the active_handler. The release point is at the end of this // function, when all operations in mailbox have been handled by the // owner of this aggregator. call_itt_notify(prepare, &handler_busy); // get handler_busy: only one thread can possibly spin here at a time spin_wait_until_eq(handler_busy, uintptr_t(0)); call_itt_notify(acquired, &handler_busy); // acquire fence not necessary here due to causality rule and surrounding atomics __TBB_store_with_release(handler_busy, uintptr_t(1)); // ITT note: &mailbox tag covers access to the handler_busy flag itself. // Capturing the state of the mailbox signifies that handler_busy has been // set and a new active handler will now process that list's operations. call_itt_notify(releasing, &mailbox); // grab pending_operations pending_operations = mailbox.fetch_and_store(NULL); // handle all the operations handle_operations(pending_operations); // release the handler itt_store_word_with_release(handler_busy, uintptr_t(0)); } }; //! Basic aggregator interface class aggregator : private aggregator_ext { public: aggregator() : aggregator_ext(internal::basic_handler()) {} //! BASIC INTERFACE: Enter a function for exclusvie execution by the aggregator. /** The calling thread stores the function object in a basic_operation and places the operation in the aggregator's mailbox */ template void execute(const Body& b) { internal::basic_operation op(b); this->execute_impl(op); } }; } // namespace interface6 using interface6::aggregator; using interface6::aggregator_ext; using interface6::aggregator_operation; } // namespace tbb #endif // __TBB__aggregator_H