X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/d53a2108a1ea38b2f952abb48c39730ccc91c2a8..ab4d43a39df33bb5f24c3b2cc6d576130b16a649:/src/smpi/internals/smpi_global.cpp diff --git a/src/smpi/internals/smpi_global.cpp b/src/smpi/internals/smpi_global.cpp index 5aade555a7..f2a5b48c51 100644 --- a/src/smpi/internals/smpi_global.cpp +++ b/src/smpi/internals/smpi_global.cpp @@ -5,25 +5,17 @@ #include "SmpiHost.hpp" #include "mc/mc.h" -#include "private.hpp" -#include "simgrid/s4u/Host.hpp" -#include "simgrid/s4u/Mailbox.hpp" -#include "simgrid/s4u/forward.hpp" +#include "simgrid/s4u/Engine.hpp" #include "smpi_coll.hpp" -#include "smpi_comm.hpp" -#include "smpi_group.hpp" -#include "smpi_info.hpp" #include "smpi_process.hpp" #include "src/msg/msg_private.hpp" #include "src/simix/smx_private.hpp" -#include "src/surf/surf_interface.hpp" #include "xbt/config.hpp" #include /* DBL_MAX */ #include #include #include -#include #if HAVE_SENDFILE #include @@ -49,14 +41,15 @@ struct papi_process_data { papi_counter_t counter_data; int event_set; }; - #endif + using simgrid::s4u::Actor; using simgrid::s4u::ActorPtr; std::unordered_map location2speedup; static std::map process_data; int process_count = 0; +static int smpi_exit_status = 0; int smpi_universe_size = 0; extern double smpi_total_benched_time; xbt_os_timer_t global_timer; @@ -76,7 +69,7 @@ MPI_Errhandler *MPI_ERRORS_ARE_FATAL = nullptr; MPI_Errhandler *MPI_ERRHANDLER_NULL = nullptr; // No instance gets manually created; check also the smpirun.in script as // this default name is used there as well (when the tag is generated). -static const char* smpi_default_instance_name = "smpirun"; +static const std::string smpi_default_instance_name("smpirun"); static simgrid::config::Flag smpi_wtime_sleep( "smpi/wtime", "Minimum time to inject inside a call to MPI_Wtime", 0.0); static simgrid::config::Flag smpi_init_sleep( @@ -84,11 +77,6 @@ static simgrid::config::Flag smpi_init_sleep( void (*smpi_comm_copy_data_callback) (smx_activity_t, void*, size_t) = &smpi_comm_copy_buffer_callback; -void smpi_add_process(ActorPtr actor) -{ - process_data.insert({actor, new simgrid::smpi::Process(actor, nullptr)}); -} - int smpi_process_count() { return process_count; @@ -99,7 +87,7 @@ simgrid::smpi::Process* smpi_process() ActorPtr me = Actor::self(); if (me == nullptr) // This happens sometimes (eg, when linking against NS3 because it pulls openMPI...) return nullptr; - simgrid::msg::ActorExt* msgExt = static_cast(me->getImpl()->userdata); + simgrid::msg::ActorExt* msgExt = static_cast(me->get_impl()->userdata); return static_cast(msgExt->data); } @@ -117,7 +105,7 @@ void smpi_process_init(int *argc, char ***argv){ } int smpi_process_index(){ - return simgrid::s4u::Actor::self()->getPid(); + return simgrid::s4u::this_actor::get_pid(); } void * smpi_process_get_user_data(){ @@ -192,19 +180,19 @@ void smpi_comm_copy_buffer_callback(smx_activity_t synchro, void *buff, size_t b auto private_blocks = merge_private_blocks(src_private_blocks, dst_private_blocks); check_blocks(private_blocks, buff_size); void* tmpbuff=buff; - if ((smpi_privatize_global_variables == SMPI_PRIVATIZE_MMAP) && (static_cast(buff) >= smpi_data_exe_start) && + if ((smpi_privatize_global_variables == SmpiPrivStrategies::Mmap) && + (static_cast(buff) >= smpi_data_exe_start) && (static_cast(buff) < smpi_data_exe_start + smpi_data_exe_size)) { XBT_DEBUG("Privatization : We are copying from a zone inside global memory... Saving data to temp buffer !"); - - smpi_switch_data_segment(Actor::self()->getPid()); + smpi_switch_data_segment(comm->src_proc->iface()); tmpbuff = static_cast(xbt_malloc(buff_size)); memcpy_private(tmpbuff, buff, private_blocks); } - if ((smpi_privatize_global_variables == SMPI_PRIVATIZE_MMAP) && ((char*)comm->dst_buff >= smpi_data_exe_start) && + if ((smpi_privatize_global_variables == SmpiPrivStrategies::Mmap) && ((char*)comm->dst_buff >= smpi_data_exe_start) && ((char*)comm->dst_buff < smpi_data_exe_start + smpi_data_exe_size)) { XBT_DEBUG("Privatization : We are copying to a zone inside global memory - Switch data segment"); - smpi_switch_data_segment(Actor::self()->getPid()); + smpi_switch_data_segment(comm->dst_proc->iface()); } XBT_DEBUG("Copying %zu bytes from %p to %p", buff_size, tmpbuff,comm->dst_buff); memcpy_private(comm->dst_buff, tmpbuff, private_blocks); @@ -226,21 +214,23 @@ void smpi_comm_null_copy_buffer_callback(smx_activity_t comm, void *buff, size_t /* nothing done in this version */ } -static void smpi_check_options(){ +static void smpi_check_options() +{ //check correctness of MPI parameters - xbt_assert(xbt_cfg_get_int("smpi/async-small-thresh") <= xbt_cfg_get_int("smpi/send-is-detached-thresh")); + xbt_assert(simgrid::config::get_config("smpi/async-small-thresh") <= + simgrid::config::get_config("smpi/send-is-detached-thresh")); - if (xbt_cfg_is_default_value("smpi/host-speed")) { - XBT_INFO("You did not set the power of the host running the simulation. " - "The timings will certainly not be accurate. " - "Use the option \"--cfg=smpi/host-speed:\" to set its value." - "Check http://simgrid.org/simgrid/latest/doc/options.html#options_smpi_bench for more information."); - } + if (xbt_cfg_is_default_value("smpi/host-speed")) { + XBT_INFO("You did not set the power of the host running the simulation. " + "The timings will certainly not be accurate. " + "Use the option \"--cfg=smpi/host-speed:\" to set its value." + "Check http://simgrid.org/simgrid/latest/doc/options.html#options_smpi_bench for more information."); + } - xbt_assert(xbt_cfg_get_double("smpi/cpu-threshold") >=0, - "The 'smpi/cpu-threshold' option cannot have negative values [anymore]. If you want to discard " - "the simulation of any computation, please use 'smpi/simulate-computation:no' instead."); + xbt_assert(xbt_cfg_get_double("smpi/cpu-threshold") >= 0, + "The 'smpi/cpu-threshold' option cannot have negative values [anymore]. If you want to discard " + "the simulation of any computation, please use 'smpi/simulate-computation:no' instead."); } int smpi_enabled() { @@ -349,18 +339,6 @@ void smpi_global_destroy() smpi_bench_destroy(); smpi_shared_destroy(); smpi_deployment_cleanup_instances(); - for (auto& pair : process_data) { - auto& process = pair.second; - if (process->comm_self() != MPI_COMM_NULL) { - simgrid::smpi::Comm::destroy(process->comm_self()); - } - if (process->comm_intra() != MPI_COMM_NULL) { - simgrid::smpi::Comm::destroy(process->comm_intra()); - } - xbt_os_timer_free(process->timer()); - xbt_mutex_destroy(process->mailboxes_mutex()); - } - process_data.clear(); if (simgrid::smpi::Colls::smpi_coll_cleanup_callback != nullptr) simgrid::smpi::Colls::smpi_coll_cleanup_callback(); @@ -371,7 +349,7 @@ void smpi_global_destroy() xbt_os_timer_free(global_timer); } - if(smpi_privatize_global_variables == SMPI_PRIVATIZE_MMAP) + if (smpi_privatize_global_variables == SmpiPrivStrategies::Mmap) smpi_destroy_global_memory_segments(); smpi_free_static(); } @@ -384,23 +362,28 @@ static void smpi_init_options(){ simgrid::smpi::Colls::smpi_coll_cleanup_callback = nullptr; smpi_cpu_threshold = xbt_cfg_get_double("smpi/cpu-threshold"); smpi_host_speed = xbt_cfg_get_double("smpi/host-speed"); + xbt_assert(smpi_host_speed >= 0, "You're trying to set the host_speed to a negative value (%f)", smpi_host_speed); std::string smpi_privatize_option = xbt_cfg_get_string("smpi/privatization"); if (smpi_privatize_option == "no" || smpi_privatize_option == "0") - smpi_privatize_global_variables = SMPI_PRIVATIZE_NONE; + smpi_privatize_global_variables = SmpiPrivStrategies::None; else if (smpi_privatize_option == "yes" || smpi_privatize_option == "1") - smpi_privatize_global_variables = SMPI_PRIVATIZE_DEFAULT; + smpi_privatize_global_variables = SmpiPrivStrategies::Default; else if (smpi_privatize_option == "mmap") - smpi_privatize_global_variables = SMPI_PRIVATIZE_MMAP; + smpi_privatize_global_variables = SmpiPrivStrategies::Mmap; else if (smpi_privatize_option == "dlopen") - smpi_privatize_global_variables = SMPI_PRIVATIZE_DLOPEN; + smpi_privatize_global_variables = SmpiPrivStrategies::Dlopen; else xbt_die("Invalid value for smpi/privatization: '%s'", smpi_privatize_option.c_str()); + if (not SMPI_switch_data_segment) { + XBT_DEBUG("Running without smpi_main(); disable smpi/privatization."); + smpi_privatize_global_variables = SmpiPrivStrategies::None; + } #if defined(__FreeBSD__) - if (smpi_privatize_global_variables == SMPI_PRIVATIZE_MMAP) { - XBT_INFO("Mixing mmap privatization is broken on FreeBSD, switching to dlopen privatization instead."); - smpi_privatize_global_variables = SMPI_PRIVATIZE_DLOPEN; - } + if (smpi_privatize_global_variables == SmpiPrivStrategies::Mmap) { + XBT_INFO("mmap privatization is broken on FreeBSD, switching to dlopen privatization instead."); + smpi_privatize_global_variables = SmpiPrivStrategies::Dlopen; + } #endif if (smpi_cpu_threshold < 0) @@ -435,7 +418,8 @@ static int smpi_run_entry_point(smpi_entry_point_type entry_point, std::vectorset_return_value(res); + if (smpi_exit_status == 0) + smpi_exit_status = res; } return 0; } @@ -489,7 +473,7 @@ int smpi_main(const char* executable, int argc, char *argv[]) SIMIX_comm_set_copy_data_callback(smpi_comm_copy_buffer_callback); smpi_init_options(); - if (smpi_privatize_global_variables == SMPI_PRIVATIZE_DLOPEN) { + if (smpi_privatize_global_variables == SmpiPrivStrategies::Dlopen) { std::string executable_copy = executable; @@ -508,17 +492,18 @@ int smpi_main(const char* executable, int argc, char *argv[]) + "_" + std::to_string(rank++) + ".so"; int fdin = open(executable_copy.c_str(), O_RDONLY); - xbt_assert(fdin >= 0, "Cannot read from %s", executable_copy.c_str()); + xbt_assert(fdin >= 0, "Cannot read from %s. Please make sure that the file exists and is executable.", + executable_copy.c_str()); int fdout = open(target_executable.c_str(), O_CREAT | O_RDWR, S_IRWXU); xbt_assert(fdout >= 0, "Cannot write into %s", target_executable.c_str()); + XBT_DEBUG("Copy %ld bytes into %s", static_cast(fdin_size), target_executable.c_str()); #if HAVE_SENDFILE ssize_t sent_size = sendfile(fdout, fdin, NULL, fdin_size); xbt_assert(sent_size == fdin_size, "Error while copying %s: only %zd bytes copied instead of %ld (errno: %d -- %s)", target_executable.c_str(), sent_size, fdin_size, errno, strerror(errno)); #else - XBT_VERB("Copy %d bytes into %s", static_cast(fdin_size), target_executable.c_str()); const int bufsize = 1024 * 1024 * 4; char buf[bufsize]; while (int got = read(fdin, buf, bufsize)) { @@ -544,7 +529,7 @@ int smpi_main(const char* executable, int argc, char *argv[]) // Load the copy and resolve the entry point: void* handle = dlopen(target_executable.c_str(), RTLD_LAZY | RTLD_LOCAL | RTLD_DEEPBIND); int saved_errno = errno; - if (xbt_cfg_get_boolean("smpi/keep-temps") == false) + if (simgrid::config::get_config("smpi/keep-temps") == false) unlink(target_executable.c_str()); if (handle == nullptr) xbt_die("dlopen failed: %s (errno: %d -- %s)", dlerror(), saved_errno, strerror(saved_errno)); @@ -555,18 +540,19 @@ int smpi_main(const char* executable, int argc, char *argv[]) smpi_run_entry_point(entry_point, args); }); }; - } else { - + if (smpi_privatize_global_variables == SmpiPrivStrategies::Mmap) + smpi_prepare_global_memory_segment(); // Load the dynamic library and resolve the entry point: - void* handle = dlopen(executable, RTLD_LAZY | RTLD_LOCAL | RTLD_DEEPBIND); + void* handle = dlopen(executable, RTLD_LAZY | RTLD_LOCAL); if (handle == nullptr) xbt_die("dlopen failed for %s: %s (errno: %d -- %s)", executable, dlerror(), errno, strerror(errno)); smpi_entry_point_type entry_point = smpi_resolve_function(handle); if (not entry_point) xbt_die("main not found in %s", executable); - // TODO, register the executable for SMPI privatization + if (smpi_privatize_global_variables == SmpiPrivStrategies::Mmap) + smpi_backup_global_memory_segment(); // Execute the same entry point for each simulated process: simix_global->default_function = [entry_point](std::vector args) { @@ -574,13 +560,12 @@ int smpi_main(const char* executable, int argc, char *argv[]) smpi_run_entry_point(entry_point, args); }); }; - } SMPI_init(); SIMIX_launch_application(argv[2]); - SMPI_app_instance_register(smpi_default_instance_name, nullptr, - SIMIX_process_count()); // This call has a side effect on process_count... + SMPI_app_instance_register(smpi_default_instance_name.c_str(), nullptr, + process_data.size()); // This call has a side effect on process_count... MPI_COMM_WORLD = *smpi_deployment_comm_world(smpi_default_instance_name); smpi_universe_size = process_count; @@ -596,7 +581,7 @@ int smpi_main(const char* executable, int argc, char *argv[]) SIMIX_run(); xbt_os_walltimer_stop(global_timer); - if (xbt_cfg_get_boolean("smpi/display-timing")){ + if (simgrid::config::get_config("smpi/display-timing")) { double global_time = xbt_os_timer_elapsed(global_timer); XBT_INFO("Simulated time: %g seconds. \n\n" "The simulation took %g seconds (after parsing and platform setup)\n" @@ -608,38 +593,32 @@ int smpi_main(const char* executable, int argc, char *argv[]) "You may want to use sampling functions or trace replay to reduce this."); } } - int ret = 0; - for (auto& pair : process_data) { - auto& smpi_process = pair.second; - if (smpi_process->return_value() != 0) { - ret = smpi_process->return_value(); // return first non 0 value - break; - } - } smpi_global_destroy(); TRACE_end(); - return ret; + return smpi_exit_status; } // Called either directly from the user code, or from the code called by smpirun void SMPI_init(){ - simgrid::s4u::Actor::onCreation.connect([](simgrid::s4u::ActorPtr actor) { - smpi_add_process(actor); + simgrid::s4u::Actor::on_creation.connect([](simgrid::s4u::ActorPtr actor) { + if (not actor->is_daemon()) { + process_data.insert({actor, new simgrid::smpi::Process(actor, nullptr)}); + } }); - simgrid::s4u::Actor::onDestruction.connect([](simgrid::s4u::ActorPtr actor) { - delete process_data.at(actor); - process_data.erase(actor); + simgrid::s4u::Actor::on_destruction.connect([](simgrid::s4u::ActorPtr actor) { + auto it = process_data.find(actor); + if (it != process_data.end()) { + delete it->second; + process_data.erase(it); + } }); smpi_init_options(); smpi_global_init(); smpi_check_options(); - TRACE_smpi_alloc(); - simgrid::surf::surfExitCallbacks.connect(TRACE_smpi_release); - if(smpi_privatize_global_variables == SMPI_PRIVATIZE_MMAP) - smpi_backup_global_memory_segment(); + simgrid::s4u::onSimulationEnd.connect(TRACE_smpi_release); } void SMPI_finalize(){