X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/2ab7f3664f841e597a8d90de3722ee6ee92c3036..5193d31089b0e9d2a564d425f61fdbc27e01af60:/src/smpi/smpi_global.cpp

diff --git a/src/smpi/smpi_global.cpp b/src/smpi/smpi_global.cpp
index fed7bca265..7f62db8e9e 100644
--- a/src/smpi/smpi_global.cpp
+++ b/src/smpi/smpi_global.cpp
@@ -3,10 +3,12 @@
 /* This program is free software; you can redistribute it and/or modify it
  * under the terms of the license (GNU LGPL) which comes with this package. */
 
+#include <dlfcn.h>
+#include <fcntl.h>
 #include <spawn.h>
+#include <sys/stat.h>
 #include <sys/types.h>
 #include <sys/wait.h>
-#include <dlfcn.h>
 
 #include "mc/mc.h"
 #include "private.h"
@@ -19,6 +21,8 @@
 #include "src/mc/mc_replay.h"
 #include "src/msg/msg_private.h"
 #include "src/simix/smx_private.h"
+#include "src/surf/surf_interface.hpp"
+#include "src/smpi/SmpiHost.hpp"
 #include "surf/surf.h"
 #include "xbt/replay.hpp"
 #include <xbt/config.hpp>
@@ -34,6 +38,10 @@
 #include <vector>
 #include <memory>
 
+#if HAVE_SENDFILE
+#include <sys/sendfile.h>
+#endif
+
 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 */
@@ -87,7 +95,10 @@ int smpi_process_count()
 
 simgrid::smpi::Process* smpi_process()
 {
-  simgrid::MsgActorExt* msgExt = static_cast<simgrid::MsgActorExt*>(SIMIX_process_self()->data);
+  smx_actor_t me = SIMIX_process_self();
+  if (me == nullptr) // This happens sometimes (eg, when linking against NS3 because it pulls openMPI...)
+    return nullptr;
+  simgrid::MsgActorExt* msgExt = static_cast<simgrid::MsgActorExt*>(me->data);
   return static_cast<simgrid::smpi::Process*>(msgExt->data);
 }
 
@@ -123,13 +134,14 @@ void smpi_comm_set_copy_data_callback(void (*callback) (smx_activity_t, void*, s
 }
 
 static void print(std::vector<std::pair<size_t, size_t>> vec) {
-    fprintf(stderr, "{");
-    for(auto elt: vec) {
-        fprintf(stderr, "(0x%lx, 0x%lx),", elt.first, elt.second);
+  std::fprintf(stderr, "{");
+  for (auto elt : vec) {
+    std::fprintf(stderr, "(0x%zx, 0x%zx),", elt.first, elt.second);
     }
-    fprintf(stderr, "}\n");
+    std::fprintf(stderr, "}\n");
 }
-static void memcpy_private(void *dest, const void *src, size_t n, std::vector<std::pair<size_t, size_t>> &private_blocks) {
+static void memcpy_private(void* dest, const void* src, std::vector<std::pair<size_t, size_t>>& private_blocks)
+{
   for(auto block : private_blocks) {
     memcpy((uint8_t*)dest+block.first, (uint8_t*)src+block.first, block.second-block.first);
   }
@@ -144,8 +156,10 @@ static void check_blocks(std::vector<std::pair<size_t, size_t>> &private_blocks,
 void smpi_comm_copy_buffer_callback(smx_activity_t synchro, void *buff, size_t buff_size)
 {
   simgrid::kernel::activity::Comm *comm = dynamic_cast<simgrid::kernel::activity::Comm*>(synchro);
-  int src_shared=0, dst_shared=0;
-  size_t src_offset=0, dst_offset=0;
+  int src_shared                        = 0;
+  int dst_shared                        = 0;
+  size_t src_offset                     = 0;
+  size_t dst_offset                     = 0;
   std::vector<std::pair<size_t, size_t>> src_private_blocks;
   std::vector<std::pair<size_t, size_t>> dst_private_blocks;
   XBT_DEBUG("Copy the data over");
@@ -178,7 +192,7 @@ void smpi_comm_copy_buffer_callback(smx_activity_t synchro, void *buff, size_t b
        smpi_switch_data_segment(
            (static_cast<simgrid::smpi::Process*>((static_cast<simgrid::MsgActorExt*>(comm->src_proc->data)->data))->index()));
        tmpbuff = static_cast<void*>(xbt_malloc(buff_size));
-       memcpy_private(tmpbuff, buff, buff_size, private_blocks);
+       memcpy_private(tmpbuff, buff, private_blocks);
   }
 
   if((smpi_privatize_global_variables == SMPI_PRIVATIZE_MMAP) && ((char*)comm->dst_buff >= smpi_start_data_exe)
@@ -188,7 +202,7 @@ void smpi_comm_copy_buffer_callback(smx_activity_t synchro, void *buff, size_t b
            (static_cast<simgrid::smpi::Process*>((static_cast<simgrid::MsgActorExt*>(comm->dst_proc->data)->data))->index()));
   }
   XBT_DEBUG("Copying %zu bytes from %p to %p", buff_size, tmpbuff,comm->dst_buff);
-  memcpy_private(comm->dst_buff, tmpbuff, buff_size, private_blocks);
+  memcpy_private(comm->dst_buff, tmpbuff, private_blocks);
 
   if (comm->detached) {
     // if this is a detached send, the source buffer was duplicated by SMPI
@@ -428,12 +442,14 @@ static void smpi_init_logs(){
 }
 
 static void smpi_init_options(){
-
+    //return if already called
+    if (smpi_cpu_threshold > -1)
+      return;
     simgrid::smpi::Colls::set_collectives();
     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");
-    const char* smpi_privatize_option = xbt_cfg_get_string("smpi/privatize-global-variables");
+    const char* smpi_privatize_option               = xbt_cfg_get_string("smpi/privatization");
     if (std::strcmp(smpi_privatize_option, "no") == 0)
       smpi_privatize_global_variables = SMPI_PRIVATIZE_NONE;
     else if (std::strcmp(smpi_privatize_option, "yes") == 0)
@@ -450,8 +466,7 @@ static void smpi_init_options(){
       smpi_privatize_global_variables = SMPI_PRIVATIZE_NONE;
 
     else
-      xbt_die("Invalid value for smpi/privatize-global-variables: %s",
-        smpi_privatize_option);
+      xbt_die("Invalid value for smpi/privatization: '%s'", smpi_privatize_option);
 
     if (smpi_cpu_threshold < 0)
       smpi_cpu_threshold = DBL_MAX;
@@ -469,20 +484,9 @@ static void smpi_init_options(){
     }
 }
 
-static int execute_command(const char * const argv[])
-{
-  pid_t pid;
-  int status;
-  if (posix_spawnp(&pid, argv[0], nullptr, nullptr, (char* const*) argv, environ) != 0)
-    return 127;
-  if (waitpid(pid, &status, 0) != pid)
-    return 127;
-  return status;
-}
-
 typedef std::function<int(int argc, char *argv[])> smpi_entry_point_type;
 typedef int (* smpi_c_entry_point_type)(int argc, char **argv);
-typedef void (* smpi_fortran_entry_point_type)(void);
+typedef void (*smpi_fortran_entry_point_type)();
 
 static int smpi_run_entry_point(smpi_entry_point_type entry_point, std::vector<std::string> args)
 {
@@ -503,20 +507,17 @@ static int smpi_run_entry_point(smpi_entry_point_type entry_point, std::vector<s
 // TODO, remove the number of functions involved here
 static smpi_entry_point_type smpi_resolve_function(void* handle)
 {
-  smpi_fortran_entry_point_type entry_point2 =
-    (smpi_fortran_entry_point_type) dlsym(handle, "user_main_");
-  if (entry_point2 != nullptr) {
-    // fprintf(stderr, "EP user_main_=%p\n", entry_point2);
-    return [entry_point2](int argc, char** argv) {
+  smpi_fortran_entry_point_type entry_point_fortran = (smpi_fortran_entry_point_type)dlsym(handle, "user_main_");
+  if (entry_point_fortran != nullptr) {
+    return [entry_point_fortran](int argc, char** argv) {
       smpi_process_init(&argc, &argv);
-      entry_point2();
+      entry_point_fortran();
       return 0;
     };
   }
 
-  smpi_c_entry_point_type entry_point = (smpi_c_entry_point_type) dlsym(handle, "main");
+  smpi_c_entry_point_type entry_point = (smpi_c_entry_point_type)dlsym(handle, "main");
   if (entry_point != nullptr) {
-    // fprintf(stderr, "EP main=%p\n", entry_point);
     return entry_point;
   }
 
@@ -532,18 +533,17 @@ int smpi_main(const char* executable, int argc, char *argv[])
      * configuration tools */
     return 0;
   }
-  smpi_init_logs();
 
   TRACE_global_init(&argc, argv);
-  TRACE_add_start_function(TRACE_smpi_alloc);
-  TRACE_add_end_function(TRACE_smpi_release);
 
   SIMIX_global_init(&argc, argv);
   MSG_init(&argc,argv);
 
   SMPI_switch_data_segment = &smpi_switch_data_segment;
 
-  smpi_init_options();
+  simgrid::s4u::Host::onCreation.connect([](simgrid::s4u::Host& host) {
+    host.extension_set(new simgrid::smpi::SmpiHost(&host));
+  });
 
   // parse the platform file: get the host list
   SIMIX_create_environment(argv[1]);
@@ -551,38 +551,70 @@ int smpi_main(const char* executable, int argc, char *argv[])
 
   static std::size_t rank = 0;
 
+  smpi_init_options();
+
   if (smpi_privatize_global_variables == SMPI_PRIVATIZE_DLOPEN) {
 
     std::string executable_copy = executable;
-    simix_global->default_function = [executable_copy](std::vector<std::string> args) {
-      return std::function<void()>([executable_copy, args] {
+
+    // Prepare the copy of the binary (get its size)
+    struct stat fdin_stat;
+    stat(executable_copy.c_str(), &fdin_stat);
+    off_t fdin_size = fdin_stat.st_size;
+
+    simix_global->default_function = [executable_copy, fdin_size](std::vector<std::string> args) {
+      return std::function<void()>([executable_copy, fdin_size, args] {
 
         // Copy the dynamic library:
         std::string target_executable = executable_copy
           + "_" + std::to_string(getpid())
           + "_" + std::to_string(rank++) + ".so";
-        // TODO, execute directly instead of relying on cp
-        const char* command1 [] = {
-          "cp", "--reflink=auto", "--", executable_copy.c_str(), target_executable.c_str(),
-          nullptr
-        };
-        const char* command2 [] = {
-          "cp", "--", executable_copy.c_str(), target_executable.c_str(),
-          nullptr
-        };
-        if (execute_command(command1) != 0 && execute_command(command2) != 0)
-          xbt_die("copy failed");
+
+        int fdin = open(executable_copy.c_str(), O_RDONLY);
+        xbt_assert(fdin >= 0, "Cannot read from %s", 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());
+
+#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)) {
+          if (got == -1) {
+            xbt_assert(errno == EINTR, "Cannot read from %s", executable_copy.c_str());
+          } else {
+            char* p  = buf;
+            int todo = got;
+            while (int done = write(fdout, p, todo)) {
+              if (done == -1) {
+                xbt_assert(errno == EINTR, "Cannot write into %s", target_executable.c_str());
+              } else {
+                p += done;
+                todo -= done;
+              }
+            }
+          }
+        }
+#endif
+        close(fdin);
+        close(fdout);
 
         // Load the copy and resolve the entry point:
         void* handle = dlopen(target_executable.c_str(), RTLD_LAZY | RTLD_LOCAL | RTLD_DEEPBIND);
-        unlink(target_executable.c_str());
+        if (xbt_cfg_get_boolean("smpi/keep-temps") == false)
+          unlink(target_executable.c_str());
         if (handle == nullptr)
-          xbt_die("dlopen failed");
+          xbt_die("dlopen failed: %s (errno: %d -- %s)", dlerror(), errno, strerror(errno));
         smpi_entry_point_type entry_point = smpi_resolve_function(handle);
         if (!entry_point)
           xbt_die("Could not resolve entry point");
 
-          smpi_run_entry_point(entry_point, args);
+        smpi_run_entry_point(entry_point, args);
       });
     };
 
@@ -592,7 +624,7 @@ int smpi_main(const char* executable, int argc, char *argv[])
     // Load the dynamic library and resolve the entry point:
     void* handle = dlopen(executable, RTLD_LAZY | RTLD_LOCAL | RTLD_DEEPBIND);
     if (handle == nullptr)
-      xbt_die("dlopen failed for %s", executable);
+      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 (!entry_point)
       xbt_die("main not found in %s", executable);
@@ -609,12 +641,7 @@ int smpi_main(const char* executable, int argc, char *argv[])
 
   SIMIX_launch_application(argv[2]);
 
-  smpi_global_init();
-
-  smpi_check_options();
-
-  if(smpi_privatize_global_variables == SMPI_PRIVATIZE_MMAP)
-    smpi_initialize_global_memory_segments();
+  SMPI_init();
 
   /* Clean IO before the run */
   fflush(stdout);
@@ -654,15 +681,14 @@ int smpi_main(const char* executable, int argc, char *argv[])
   return ret;
 }
 
-// This function can be called from extern file, to initialize logs, options, and processes of smpi
-// without the need of smpirun
+// Called either directly from the user code, or from the code called by smpirun
 void SMPI_init(){
   smpi_init_logs();
   smpi_init_options();
   smpi_global_init();
   smpi_check_options();
-  if (TRACE_is_enabled() && TRACE_is_configured())
-    TRACE_smpi_alloc();
+  TRACE_smpi_alloc();
+  simgrid::surf::surfExitCallbacks.connect(TRACE_smpi_release);
   if(smpi_privatize_global_variables == SMPI_PRIVATIZE_MMAP)
     smpi_initialize_global_memory_segments();
 }