Bug fix: we really need to consume all messages in proc::receive!

[loba.git] / main.cpp

diff --git a/main.cpp b/main.cpp
index 0ec5b6da05da76d9d8d80571af3497cc4118df94..907869eaad4aa4d2d517160d8648200a84e7198e 100644 (file)
--- a/main.cpp
+++ b/main.cpp
@@ -1,15 +1,13 @@
+#include <algorithm>
+#include <cmath>

 #include <cstring>
 #include <cstring>

+#include <tr1/functional>

 #include <iostream>
 #include <iostream>

+#include <numeric>

 #include <stdexcept>
 #include <stdexcept>

+#include <vector>

 #include <msg/msg.h>
 #include <xbt/log.h>
 #include <msg/msg.h>
 #include <xbt/log.h>

-#include "deployment.h"
-#include "hostdata.h"
-#include "misc.h"
-#include "options.h"
-#include "process.h"
-#include "timer.h"
-#include "version.h"

 
 // Creates log categories
 XBT_LOG_NEW_CATEGORY(simu, "Root of simulation messages");
 
 // Creates log categories
 XBT_LOG_NEW_CATEGORY(simu, "Root of simulation messages");


@@ -21,22 +19,37 @@ XBT_LOG_NEW_SUBCATEGORY(loba, simu, "Messages from load-balancer");
 
 XBT_LOG_EXTERNAL_DEFAULT_CATEGORY(main);
 
 
 XBT_LOG_EXTERNAL_DEFAULT_CATEGORY(main);
 

-// Failure exit status
-enum {
-    EXIT_NO_FAILURE    = 0x00,  // no error
-    EXIT_FAILURE_ARGS  = 0x01,  // bad arguments
-    EXIT_FAILURE_INIT  = 0x02,  // failed to initialize simulator
-    EXIT_FAILURE_SIMU  = 0x04,  // simulation failed
-    EXIT_FAILURE_CLEAN = 0x08,  // error at cleanup
-};
+#include "deployment.h"
+#include "hostdata.h"
+#include "misc.h"
+#include "options.h"
+#include "process.h"
+#include "timer.h"
+#include "version.h"
+
+namespace {
+    // Failure exit status
+    enum {
+        EXIT_NO_FAILURE    = 0x00,  // no error
+        EXIT_FAILURE_ARGS  = 0x01,  // bad arguments
+        EXIT_FAILURE_INIT  = 0x02,  // failed to initialize simulator
+        EXIT_FAILURE_SIMU  = 0x04,  // simulation failed
+        EXIT_FAILURE_CLEAN = 0x08,  // error at cleanup
+    };
+
+    std::vector<double> loads;
+    double load_stddev;
+    double load_avg;
+}

 
 

-int simulation_main(int argc, char* argv[])
+static int simulation_main(int argc, char* argv[])

 {
     int result;
     process* proc;
     try {
         proc = opt::loba_algorithms.new_instance(opt::loba_algo, argc, argv);
         result = proc->run();
 {
     int result;
     process* proc;
     try {
         proc = opt::loba_algorithms.new_instance(opt::loba_algo, argc, argv);
         result = proc->run();

+        loads.push_back(proc->get_load());

         delete proc;
     }
     catch (std::invalid_argument& e) {
         delete proc;
     }
     catch (std::invalid_argument& e) {


@@ -45,6 +58,53 @@ int simulation_main(int argc, char* argv[])
     return result;
 }
 
     return result;
 }
 

+static void check_for_lost_load()
+{
+    double total_init = process::get_total_load_init();
+
+    double total_exit = process::get_total_load_exit();
+    double lost = total_init - total_exit;
+    double lost_ratio = 100.0 * lost / total_init;
+    if (lost_ratio < -opt::load_ratio_threshold)
+        CRITICAL2("Gained load at exit! %g (%g%%) <============",
+                  lost, lost_ratio);
+    else if (lost_ratio > opt::load_ratio_threshold)
+        CRITICAL2("Lost load at exit! %g (%g%%) <============",
+                  lost, lost_ratio);
+    else
+        DEBUG2("Total load at exit looks good: %g (%g%%)", lost, lost_ratio);
+
+    double total_running = process::get_total_load_running();
+    double running_ratio = 100.0 * total_running / total_init;
+    if (running_ratio < -opt::load_ratio_threshold)
+        CRITICAL2("Negative running load at exit! %g (%g%%) <============",
+                  total_running, running_ratio);
+    else if (running_ratio > opt::load_ratio_threshold)
+        CRITICAL2("Remaining running load at exit! %g (%g%%) <============",
+                  total_running, running_ratio);
+    else
+        DEBUG2("Running load at exit looks good: %g (%g%%)",
+               total_running, running_ratio);
+}
+
+static void compute_load_imbalance()
+{
+    using std::tr1::bind;
+    using std::tr1::placeholders::_1;
+
+    unsigned n = loads.size();
+    load_avg = std::accumulate(loads.begin(), loads.end(), 0.0) / n;
+
+    std::vector<double> diff(loads);
+    std::transform(diff.begin(), diff.end(), diff.begin(),
+                   bind(std::minus<double>(), _1, load_avg));
+    double epsilon = std::accumulate(diff.begin(), diff.end(), 0.0);
+    double square_sum = std::inner_product(diff.begin(), diff.end(),
+                                           diff.begin(), 0.0);
+    double variance = (square_sum - (epsilon * epsilon) / n) / n;
+    load_stddev = sqrt(variance);
+}
+

 int main(int argc, char* argv[])
 {
     // Note: variables used after THROW must be declared as volatile.
 int main(int argc, char* argv[])
 {
     // Note: variables used after THROW must be declared as volatile.


@@ -57,9 +117,15 @@ int main(int argc, char* argv[])
     simulation_time.start();
 
     // Set default logging parameters
     simulation_time.start();
 
     // Set default logging parameters

-    // xbt_log_control_set("simu.thres:verbose");
-    xbt_log_control_set("simu.fmt:'[%h %r] [%c/%p] %m%n'");
-    xbt_log_control_set("main.fmt:'[%c/%p] %m%n'");
+    bool do_log_control_set = true;
+    for (int i = 1 ; do_log_control_set && i < argc ; i++)
+        do_log_control_set = !(argv[i][0] == '-' && argv[i][1] != '-' &&
+                               strchr(argv[i] + 1, 'v'));
+    if (do_log_control_set) {
+        // xbt_log_control_set("simu.thres:verbose");
+        xbt_log_control_set("simu.fmt:'[%h %r] [%c/%p] %m%n'");
+        xbt_log_control_set("main.fmt:'[%c/%p] %m%n'");
+    }

 
     // Initialize some MSG internal data.
     // Note: MSG_global_init() may throw an exception, but it seems
 
     // Initialize some MSG internal data.
     // Note: MSG_global_init() may throw an exception, but it seems


@@ -67,7 +133,7 @@ int main(int argc, char* argv[])
     MSG_global_init(&argc, argv);
 
     // Parse global parameters
     MSG_global_init(&argc, argv);
 
     // Parse global parameters

-    int parse_res = opt::parse_args(&argc, argv);
+    bool parse_res = opt::parse_args(&argc, argv);

     if (!parse_res
         || opt::version_requested || opt::help_requested) {
         if (opt::version_requested)
     if (!parse_res
         || opt::version_requested || opt::help_requested) {
         if (opt::version_requested)


@@ -80,6 +146,8 @@ int main(int argc, char* argv[])
         MSG_clean();
         exit(parse_res ? EXIT_NO_FAILURE : EXIT_FAILURE_ARGS);
     }
         MSG_clean();
         exit(parse_res ? EXIT_NO_FAILURE : EXIT_FAILURE_ARGS);
     }

+    INFO3("%s v%s (%s)", opt::program_name.c_str(), version::num.c_str(),
+          version::date.c_str());

     opt::print();
 
     TRY {
     opt::print();
 
     TRY {


@@ -95,8 +163,13 @@ int main(int argc, char* argv[])
         if (opt::auto_depl::enabled) {
             if (!opt::auto_depl::nhosts)
                 opt::auto_depl::nhosts = hostdata::size();
         if (opt::auto_depl::enabled) {
             if (!opt::auto_depl::nhosts)
                 opt::auto_depl::nhosts = hostdata::size();

+            if (opt::auto_depl::nhosts > hostdata::size()) {
+                WARN2("%u hosts is too much: limiting to %u",
+                      opt::auto_depl::nhosts, (unsigned )hostdata::size());
+                opt::auto_depl::nhosts = hostdata::size();
+            }

             if (!opt::auto_depl::load)
             if (!opt::auto_depl::load)

-                opt::auto_depl::load = hostdata::size();
+                opt::auto_depl::load = opt::auto_depl::nhosts;

             MY_launch_application(); // it is already opt::* aware...
         } else {
             MSG_launch_application(opt::deployment_file.c_str());
             MY_launch_application(); // it is already opt::* aware...
         } else {
             MSG_launch_application(opt::deployment_file.c_str());


@@ -107,8 +180,10 @@ int main(int argc, char* argv[])
         // Launch the MSG simulation.
         INFO1("Starting simulation at %f...", MSG_get_clock());
         res = MSG_main();
         // Launch the MSG simulation.
         INFO1("Starting simulation at %f...", MSG_get_clock());
         res = MSG_main();

-        INFO1("Simulation ended at %f.", MSG_get_clock());

         simulated_time = MSG_get_clock();
         simulated_time = MSG_get_clock();

+        INFO1("Simulation ended at %f.", simulated_time);
+        check_for_lost_load();
+        compute_load_imbalance();

         if (res != MSG_OK)
             THROW1(0, 0, "MSG_main() failed with status %#x", res);
 
         if (res != MSG_OK)
             THROW1(0, 0, "MSG_main() failed with status %#x", res);
 


@@ -135,8 +210,9 @@ int main(int argc, char* argv[])
     if (simulated_time >= 0.0) {
         simulation_time.stop();
         INFO0(",----[ Results ]");
     if (simulated_time >= 0.0) {
         simulation_time.stop();
         INFO0(",----[ Results ]");

-        INFO1("| Total simulated time...: %g", simulated_time);
-        INFO1("| Total simulation time..: %g", simulation_time.duration());
+        INFO2("| Load avg./stddev. at exit.: %g / %g", load_avg, load_stddev);
+        INFO1("| Total simulated time......: %g", simulated_time);
+        INFO1("| Total simulation time.....: %g", simulation_time.duration());

         INFO0("`----");
     }
     if (exit_status)
         INFO0("`----");
     }
     if (exit_status)