* under the terms of the license (GNU LGPL) which comes with this package. */
#include "private.h"
+#include "private.hpp"
#include "smpi_mpi_dt_private.h"
#include "mc/mc.h"
#include "src/mc/mc_record.h"
#include "simgrid/sg_config.h"
#include "src/mc/mc_replay.h"
#include "src/msg/msg_private.h"
+#include "src/simix/SynchroComm.hpp"
+
#include <float.h> /* DBL_MAX */
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
+#include <fstream>
XBT_LOG_NEW_DEFAULT_SUBCATEGORY(smpi_kernel, smpi, "Logging specific to SMPI (kernel)");
+#include <boost/tokenizer.hpp>
+#include <boost/algorithm/string.hpp> /* trim_right / trim_left */
+
+std::map<std::string, double> location2speedup;
typedef struct s_smpi_process_data {
double simulated;
char* instance_id;
int replaying; /* is the process replaying a trace */
xbt_bar_t finalization_barrier;
+ int return_value;
+ smpi_trace_call_location_t* trace_call_loc;
} s_smpi_process_data_t;
static smpi_process_data_t *process_data = NULL;
extern double smpi_total_benched_time;
xbt_os_timer_t global_timer;
MPI_Comm MPI_COMM_WORLD = MPI_COMM_UNINITIALIZED;
-
MPI_Errhandler *MPI_ERRORS_RETURN = NULL;
MPI_Errhandler *MPI_ERRORS_ARE_FATAL = NULL;
MPI_Errhandler *MPI_ERRHANDLER_NULL = NULL;
if (argc && argv) {
proc = SIMIX_process_self();
//FIXME: dirty cleanup method to avoid using msg cleanup functions on these processes when using MSG+SMPI
- SIMIX_process_set_cleanup_function(proc, SIMIX_process_cleanup);
+ SIMIX_process_set_cleanup_function(proc, MSG_process_cleanup_from_SIMIX);
char* instance_id = (*argv)[1];
int rank = xbt_str_parse_int((*argv)[2], "Invalid rank: %s");
index = smpi_process_index_of_smx_process(proc);
return process_count;
}
+/**
+ * \brief Returns a structure that stores the location (filename + linenumber)
+ * of the last calls to MPI_* functions.
+ *
+ * \see smpi_trace_set_call_location
+ */
+smpi_trace_call_location_t* smpi_process_get_call_location(void)
+{
+ smpi_process_data_t process_data = smpi_process_data();
+ return process_data->trace_call_loc;
+}
+
int smpi_process_index(void)
{
smpi_process_data_t data = smpi_process_data();
message, request, request->buf, request->size, request->src, request->dst, request->tag, request->flags);
}
-void smpi_comm_copy_buffer_callback(smx_synchro_t comm, void *buff, size_t buff_size)
+void smpi_comm_copy_buffer_callback(smx_synchro_t synchro, void *buff, size_t buff_size)
{
XBT_DEBUG("Copy the data over");
void* tmpbuff=buff;
+ simgrid::simix::Comm *comm = dynamic_cast<simgrid::simix::Comm*>(synchro);
if((smpi_privatize_global_variables) && ((char*)buff >= smpi_start_data_exe)
&& ((char*)buff < smpi_start_data_exe + smpi_size_data_exe )
){
XBT_DEBUG("Privatization : We are copying from a zone inside global memory... Saving data to temp buffer !");
- smpi_switch_data_segment(((smpi_process_data_t)(((simdata_process_t)SIMIX_process_get_data(comm->comm.src_proc))->data))->index);
+
+
+ smpi_switch_data_segment(((smpi_process_data_t)(((simdata_process_t)SIMIX_process_get_data(comm->src_proc))->data))->index);
tmpbuff = (void*)xbt_malloc(buff_size);
memcpy(tmpbuff, buff, buff_size);
}
- if((smpi_privatize_global_variables) && ((char*)comm->comm.dst_buff >= smpi_start_data_exe)
- && ((char*)comm->comm.dst_buff < smpi_start_data_exe + smpi_size_data_exe )){
+ if((smpi_privatize_global_variables) && ((char*)comm->dst_buff >= smpi_start_data_exe)
+ && ((char*)comm->dst_buff < smpi_start_data_exe + smpi_size_data_exe )){
XBT_DEBUG("Privatization : We are copying to a zone inside global memory - Switch data segment");
- smpi_switch_data_segment(((smpi_process_data_t)(((simdata_process_t)SIMIX_process_get_data(comm->comm.dst_proc))->data))->index);
+ smpi_switch_data_segment(((smpi_process_data_t)(((simdata_process_t)SIMIX_process_get_data(comm->dst_proc))->data))->index);
}
- memcpy(comm->comm.dst_buff, tmpbuff, buff_size);
- if (comm->comm.detached) {
+ memcpy(comm->dst_buff, tmpbuff, buff_size);
+ if (comm->detached) {
// if this is a detached send, the source buffer was duplicated by SMPI
// sender to make the original buffer available to the application ASAP
xbt_free(buff);
//It seems that the request is used after the call there this should be free somewhere else but where???
//xbt_free(comm->comm.src_data);// inside SMPI the request is kept inside the user data and should be free
- comm->comm.src_buff = NULL;
+ comm->src_buff = NULL;
}
if(tmpbuff!=buff)xbt_free(tmpbuff);
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(xbt_cfg_get_int("smpi/async-small-thresh") <= xbt_cfg_get_int("smpi/send-is-detached-thresh"));
- if (xbt_cfg_is_default_value("smpi/running_power")) {
+ if (xbt_cfg_is_default_value("smpi/running-power")) {
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/running_power:<flops>\" to set its value."
+ "Use the option \"--cfg=smpi/running-power:<flops>\" to set its value."
"Check http://simgrid.org/simgrid/latest/doc/options.html#options_smpi_bench for more information.");
}
}
global_timer = xbt_os_timer_new();
xbt_os_walltimer_start(global_timer);
}
+
+ if (xbt_cfg_get_string("smpi/comp-adjustment-file")[0] != '\0') {
+ std::string filename {xbt_cfg_get_string("smpi/comp-adjustment-file")};
+ std::ifstream fstream(filename);
+ if (!fstream.is_open()) {
+ xbt_die("Could not open file %s. Does it exist?", filename.c_str());
+ }
+
+ std::string line;
+ typedef boost::tokenizer< boost::escaped_list_separator<char>> Tokenizer;
+ std::getline(fstream, line); // Skip the header line
+ while (std::getline(fstream, line)) {
+ Tokenizer tok(line);
+ Tokenizer::iterator it = tok.begin();
+ Tokenizer::iterator end = std::next(tok.begin());
+
+ std::string location = *it;
+ boost::trim(location);
+ location2speedup.insert(std::pair<std::string, double>(location, std::stod(*end)));
+ }
+ }
+
if (process_count == 0){
process_count = SIMIX_process_count();
smpirun=1;
process_data[i]->state = SMPI_UNINITIALIZED;
process_data[i]->sampling = 0;
process_data[i]->finalization_barrier = NULL;
+ process_data[i]->return_value = 0;
+ process_data[i]->trace_call_loc = xbt_new(smpi_trace_call_location_t, 1);
}
//if the process was launched through smpirun script we generate a global mpi_comm_world
//if not, we let MPI_COMM_NULL, and the comm world will be private to each mpi instance
smpi_bench_destroy();
if (MPI_COMM_WORLD != MPI_COMM_UNINITIALIZED){
while (smpi_group_unuse(smpi_comm_group(MPI_COMM_WORLD)) > 0);
- xbt_free(MPI_COMM_WORLD);
xbt_barrier_destroy(process_data[0]->finalization_barrier);
}else{
smpi_deployment_cleanup_instances();
}
- MPI_COMM_WORLD = MPI_COMM_NULL;
for (i = 0; i < count; i++) {
if(process_data[i]->comm_self!=MPI_COMM_NULL){
- smpi_group_unuse(smpi_comm_group(process_data[i]->comm_self));
smpi_comm_destroy(process_data[i]->comm_self);
}
if(process_data[i]->comm_intra!=MPI_COMM_NULL){
- smpi_group_unuse(smpi_comm_group(process_data[i]->comm_intra));
smpi_comm_destroy(process_data[i]->comm_intra);
}
xbt_os_timer_free(process_data[i]->timer);
- simcall_mbox_destroy(process_data[i]->mailbox);
- simcall_mbox_destroy(process_data[i]->mailbox_small);
xbt_mutex_destroy(process_data[i]->mailboxes_mutex);
xbt_free(process_data[i]);
}
xbt_free(process_data);
process_data = NULL;
+ if (MPI_COMM_WORLD != MPI_COMM_UNINITIALIZED){
+ smpi_comm_cleanup_smp(MPI_COMM_WORLD);
+ smpi_comm_cleanup_attributes(MPI_COMM_WORLD);
+ if(smpi_coll_cleanup_callback!=NULL)
+ smpi_coll_cleanup_callback();
+ xbt_free(MPI_COMM_WORLD);
+ }
+
+ MPI_COMM_WORLD = MPI_COMM_NULL;
+
+ if (!MC_is_active()) {
+ xbt_os_timer_free(global_timer);
+ }
+
xbt_free(index_to_process_data);
if(smpi_privatize_global_variables)
smpi_destroy_global_memory_segments();
}
#ifndef WIN32
+
void __attribute__ ((weak)) user_main_()
{
xbt_die("Should not be in this smpi_simulated_main");
return 0;
}
+inline static int smpi_main_wrapper(int argc, char **argv){
+ int ret = smpi_simulated_main_(argc,argv);
+ if(ret !=0){
+ XBT_WARN("SMPI process did not return 0. Return value : %d", ret);
+ smpi_process_data()->return_value=ret;
+ }
+ return 0;
+}
+
int __attribute__ ((weak)) main(int argc, char **argv)
{
- return smpi_main(smpi_simulated_main_, argc, argv);
+ return smpi_main(smpi_main_wrapper, argc, argv);
}
#endif
int reduce_scatter_id =
find_coll_description(mpi_coll_reduce_scatter_description,
- xbt_cfg_get_string("smpi/reduce_scatter"),"reduce_scatter");
+ xbt_cfg_get_string("smpi/reduce-scatter"),"reduce_scatter");
mpi_coll_reduce_scatter_fun = (int (*)(void *sbuf, void *rbuf, int *rcounts,MPI_Datatype dtype, MPI_Op op,
MPI_Comm comm)) mpi_coll_reduce_scatter_description[reduce_scatter_id].coll;
int barrier_id = find_coll_description(mpi_coll_barrier_description, xbt_cfg_get_string("smpi/barrier"),"barrier");
mpi_coll_barrier_fun = (int (*)(MPI_Comm comm)) mpi_coll_barrier_description[barrier_id].coll;
- smpi_cpu_threshold = xbt_cfg_get_double("smpi/cpu_threshold");
- smpi_running_power = xbt_cfg_get_double("smpi/running_power");
- smpi_privatize_global_variables = xbt_cfg_get_boolean("smpi/privatize_global_variables");
+ smpi_coll_cleanup_callback=NULL;
+ smpi_cpu_threshold = xbt_cfg_get_double("smpi/cpu-threshold");
+ smpi_running_power = xbt_cfg_get_double("smpi/running-power");
+ smpi_privatize_global_variables = xbt_cfg_get_boolean("smpi/privatize-global-variables");
if (smpi_cpu_threshold < 0)
smpi_cpu_threshold = DBL_MAX;
}
SIMIX_run();
xbt_os_walltimer_stop(global_timer);
- if (xbt_cfg_get_boolean("smpi/display_timing")){
+ if (xbt_cfg_get_boolean("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"
"You may want to use sampling functions or trace replay to reduce this.");
}
}
+ int count = smpi_process_count();
+ int i, ret=0;
+ for (i = 0; i < count; i++) {
+ if(process_data[i]->return_value!=0){
+ ret=process_data[i]->return_value;//return first non 0 value
+ break;
+ }
+ }
smpi_global_destroy();
TRACE_end();
- return 0;
+ return ret;
}
// This function can be called from extern file, to initialize logs, options, and processes of smpi