From: Matthias Maier Date: Thu, 2 Apr 2020 22:14:30 +0000 (-0500) Subject: Refactor output thread; only perform IO asynchronously X-Git-Tag: v9.2.0-rc1~305^2~2 X-Git-Url: https://gitweb.dealii.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=6bd5b1014eaea189e551b4fafec430e7ae8f27a4;p=dealii.git Refactor output thread; only perform IO asynchronously --- diff --git a/examples/step-69/step-69.cc b/examples/step-69/step-69.cc index b75a712b09..8beef68072 100644 --- a/examples/step-69/step-69.cc +++ b/examples/step-69/step-69.cc @@ -82,9 +82,10 @@ #include #include -// For std::isnan, std::isinf, and std::ifstream +// For std::isnan, std::isinf, std::ifstream, std::async, and std::future #include #include +#include // @sect3{Class template declarations} // @@ -505,9 +506,11 @@ namespace Step69 InitialValues initial_values; TimeStepping time_stepping; SchlierenPostprocessor schlieren_postprocessor; + DataOut data_out; - std::thread output_thread; vector_type output_vector; + + std::future background_thread_state; }; // @sect3{Implementation} @@ -2571,28 +2574,43 @@ namespace Step69 ParameterAcceptor::initialize("step-69.prm"); pcout << "done" << std::endl; - // Next we create the triangulation: + // Next we create the triangulation, assemble all matrices, set up + // scratch space, and initialize the DataOut object: - print_head(pcout, "create triangulation"); - discretization.setup(); + { + print_head(pcout, "create triangulation"); + discretization.setup(); - pcout << "Number of active cells: " - << discretization.triangulation.n_global_active_cells() << std::endl; + pcout << "Number of active cells: " + << discretization.triangulation.n_global_active_cells() + << std::endl; - // Assemble all matrices: + print_head(pcout, "compute offline data"); + offline_data.setup(); + offline_data.assemble(); - print_head(pcout, "compute offline data"); - offline_data.setup(); - offline_data.assemble(); + pcout << "Number of degrees of freedom: " + << offline_data.dof_handler.n_dofs() << std::endl; - pcout << "Number of degrees of freedom: " - << offline_data.dof_handler.n_dofs() << std::endl; + print_head(pcout, "set up time step"); + time_stepping.prepare(); + schlieren_postprocessor.prepare(); - // And set up scratch space: + data_out.attach_dof_handler(offline_data.dof_handler); - print_head(pcout, "set up time step"); - time_stepping.prepare(); - schlieren_postprocessor.prepare(); + constexpr auto problem_dimension = + ProblemDescription::problem_dimension; + const auto &component_names = ProblemDescription::component_names; + + for (unsigned int i = 0; i < problem_dimension; ++i) + { + output_vector[i].reinit(offline_data.partitioner); + data_out.add_data_vector(output_vector[i], component_names[i]); + } + + data_out.add_data_vector(schlieren_postprocessor.schlieren, + "schlieren_plot"); + } // We will store the current time and state in the variable // t and vector U: @@ -2686,9 +2704,10 @@ namespace Step69 // We wait for any remaining background output thread to finish before // printing a summary and exiting. - - if (output_thread.joinable()) - output_thread.join(); + if (background_thread_state.valid()) + { + background_thread_state.wait(); + } computing_timer.print_summary(); pcout << timer_output.str() << std::endl; @@ -2762,23 +2781,34 @@ namespace Step69 pcout << "MainLoop::output(t = " << t << ", checkpoint = " << checkpoint << ")" << std::endl; - // We check whether the output thread is still running. If so, we have - // to wait to for it to finish because we would otherwise overwrite - // output_vector and rerun the - // schlieren_postprocessor before the output of the - // previous output cycle has been fully written back to disk. - - if (output_thread.joinable()) + // If the asynchronous writeback option is set we launch a background + // thread performing all the slow IO to disc. In that case we have to + // make sure that the background thread actually finished running. If + // not, we have to wait to for it to finish because we would otherwise + // overwrite output_vector, rerun the + // schlieren_postprocessor and + // DataOut::build_patches() prematurely before the output of the + // previous output cycle has been fully written out to disk. + // + // We launch said background thread with std::async + // that returns a std::future + // object. This std::future object contains the return + // value of the function, which is in our case simply + // void. + + if (background_thread_state.valid()) { TimerOutput::Scope timer(computing_timer, "main_loop - stalled output"); - output_thread.join(); + background_thread_state.wait(); } constexpr auto problem_dimension = ProblemDescription::problem_dimension; - // At this point we make a copy of the state vector and run the - // schlieren postprocessor. + // At this point we make a copy of the state vector, run the schlieren + // postprocessor, and run DataOut::build_patches() for (unsigned int i = 0; i < problem_dimension; ++i) { @@ -2788,6 +2818,9 @@ namespace Step69 schlieren_postprocessor.compute_schlieren(output_vector); + data_out.build_patches(discretization.mapping, + discretization.finite_element.degree - 1); + // Next we create a lambda function for the background thread. We capture // the this pointer as well as most of the arguments of @@ -2815,26 +2848,6 @@ namespace Step69 oa << it2; } - // The actual output code is standard. We create a (local) DataOut - // instance, attach all data vectors we want to output and finally - // call to DataOut::write_vtu_with_pvtu_record - - DataOut data_out; - data_out.attach_dof_handler(offline_data.dof_handler); - - constexpr auto problem_dimension = - ProblemDescription::problem_dimension; - const auto &component_names = ProblemDescription::component_names; - - for (unsigned int i = 0; i < problem_dimension; ++i) - data_out.add_data_vector(output_vector[i], component_names[i]); - - data_out.add_data_vector(schlieren_postprocessor.schlieren, - "schlieren_plot"); - - data_out.build_patches(discretization.mapping, - discretization.finite_element.degree - 1); - DataOutBase::VtkFlags flags(t, cycle, true, @@ -2844,18 +2857,20 @@ namespace Step69 data_out.write_vtu_with_pvtu_record("", name, cycle, mpi_communicator, 6); }; - // We launch the thread by creating a + // If the asynchronous writeback option is set we launch a new + // background thread with the help of // std::thread - // object from the lambda function and moving it into the - // output_thread thread object. At this point we can - // return from the output() function and resume with the - // time stepping in the main loop - the thread will run in the - // background. + // href="https://en.cppreference.com/w/cpp/thread/async">std::async + // function. The function returns a std::future + // object that we can use to query the status of the background thread. + // At this point we can return from the output() function + // and resume with the time stepping in the main loop - the thread will + // run in the background. if (!asynchronous_writeback) { - output_thread = std::move(std::thread(output_worker)); + background_thread_state = std::async(std::launch::async, output_worker); } else {