#include "SmpiHost.hpp"
#include "mc/mc.h"
#include "private.hpp"
+#include "simgrid/s4u/Engine.hpp"
#include "simgrid/s4u/Host.hpp"
#include "simgrid/s4u/Mailbox.hpp"
#include "simgrid/s4u/forward.hpp"
static std::map</*process_id*/ ActorPtr, simgrid::smpi::Process*> 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;
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 <actor> tag is generated).
-static const char* smpi_default_instance_name = "smpirun";
+static const std::string smpi_default_instance_name("smpirun");
static simgrid::config::Flag<double> smpi_wtime_sleep(
"smpi/wtime", "Minimum time to inject inside a call to MPI_Wtime", 0.0);
static simgrid::config::Flag<double> 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;
}
int smpi_process_index(){
- return simgrid::s4u::Actor::self()->getPid();
+ return simgrid::s4u::this_actor::getPid();
}
void * smpi_process_get_user_data(){
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<char*>(buff) >= smpi_data_exe_start) &&
+ if ((smpi_privatize_global_variables == SmpiPrivStrategies::Mmap) &&
+ (static_cast<char*>(buff) >= smpi_data_exe_start) &&
(static_cast<char*>(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<void*>(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);
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();
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();
}
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)
int res = entry_point(argc, argv.get());
if (res != 0){
XBT_WARN("SMPI process did not return 0. Return value : %d", res);
- smpi_process()->set_return_value(res);
+ if (smpi_exit_status == 0)
+ smpi_exit_status = res;
}
return 0;
}
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;
+ "_" + 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<long>(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<int>(fdin_size), target_executable.c_str());
const int bufsize = 1024 * 1024 * 4;
char buf[bufsize];
while (int got = read(fdin, buf, bufsize)) {
smpi_run_entry_point(entry_point, args);
});
};
-
}
else {
-
// 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);
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;
"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::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)
+ simgrid::s4u::onSimulationEnd.connect(TRACE_smpi_release);
+ if (smpi_privatize_global_variables == SmpiPrivStrategies::Mmap)
smpi_backup_global_memory_segment();
}
