Add statistics on total computation power.

diff --git a/TODO b/TODO
index 6fedf482ad7723201ad91a62d37e13d65beec43d..30586a5efde0489a5a7d404a82cf4746d1a59a71 100644 (file)
--- a/TODO
+++ b/TODO
@@ -1,5 +1,11 @@
+* segfault with ./loba cluster1000.xml -N64 -a fairstrategy
+
 * verify bookkeeping version.
 
+* add several metrics
+  - flops (total, avg, stddev)
+  - message exchanges (number, volume)
+
 * add options -j/-J : minimum number of iterations ?
 
 * add a variant to (not) change neighbor load information at send.

diff --git a/main.cpp b/main.cpp
index 907869eaad4aa4d2d517160d8648200a84e7198e..92e48827635c67a5f3c2b9ceef8e0502e635adab 100644 (file)
--- a/main.cpp
+++ b/main.cpp
@@ -37,9 +37,14 @@ namespace {
         EXIT_FAILURE_CLEAN = 0x08,  // error at cleanup
     };
 
+    std::vector<double> comps;
+    double comp_total;
+    double comp_avg;
+    double comp_stddev;
+
     std::vector<double> loads;
-    double load_stddev;
     double load_avg;
+    double load_stddev;
 }
 
 static int simulation_main(int argc, char* argv[])
@@ -49,6 +54,7 @@ static int simulation_main(int argc, char* argv[])
     try {
         proc = opt::loba_algorithms.new_instance(opt::loba_algo, argc, argv);
         result = proc->run();
+        comps.push_back(proc->get_comp());
         loads.push_back(proc->get_load());
         delete proc;
     }
@@ -87,22 +93,37 @@ static void check_for_lost_load()
                total_running, running_ratio);
 }
 
-static void compute_load_imbalance()
+static void statistics(const std::vector<double>& vec,
+                       double* sum, double* avg, double* stddev)
 {
     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);
+    unsigned n = vec.size();
+    double vec_sum = std::accumulate(vec.begin(), vec.end(), 0.0);
+    double vec_avg = vec_sum / n;
+
+    if (sum)
+        *sum = vec_sum;
+    if (avg)
+        *avg = vec_avg;
+
+    if (stddev) {
+        std::vector<double> diff(vec);
+        std::transform(diff.begin(), diff.end(), diff.begin(),
+                       bind(std::minus<double>(), _1, vec_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;
+        *stddev = sqrt(variance);
+    }
+}
+
+static void compute_metrics()
+{
+    statistics(comps, &comp_total, &comp_avg, &comp_stddev);
+    statistics(loads, NULL, &load_avg, &load_stddev);
 }
 
 int main(int argc, char* argv[])
@@ -182,8 +203,6 @@ int main(int argc, char* argv[])
         res = MSG_main();
         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);
 
@@ -209,10 +228,16 @@ int main(int argc, char* argv[])
     // Report final simulation status.
     if (simulated_time >= 0.0) {
         simulation_time.stop();
+        check_for_lost_load();
+        compute_metrics();
         INFO0(",----[ Results ]");
-        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());
+        INFO2("| Load avg./stddev. at exit..............: %g / %g",
+              load_avg, load_stddev);
+        INFO3("| Computation total/avg./stddev. at exit.: %g / %g / %g",
+              comp_total, comp_avg, comp_stddev);
+        INFO1("| Total simulated time...................: %g", simulated_time);
+        INFO1("| Total simulation time..................: %g",
+              simulation_time.duration());
         INFO0("`----");
     }
     if (exit_status)

diff --git a/process.cpp b/process.cpp
index 9f46f8f8d0c1f067cf34b725747771c063b8740c..d3d92268c524f87ef67d6db3b4c19d3dfc1073c0 100644 (file)
--- a/process.cpp
+++ b/process.cpp
@@ -33,6 +33,8 @@ process::process(int argc, char* argv[])
         rev_neigh.insert(std::make_pair(host, ptr));
     }
 
+    comp = 0.0;
+
     prev_load_broadcast = -1;   // force sending of load on first send()
     expected_load = load;
     total_load_running += load;
@@ -189,10 +191,11 @@ double process::load_balance(double /*my_load*/)
 void process::compute()
 {
     if (load > 0.0) {
-        double duration = opt::comp_cost(load);
-        m_task_t task = MSG_task_create("computation", duration, 0.0, NULL);
-        DEBUG2("compute %g flop%s", duration, ESSE(duration));
+        double flops = opt::comp_cost(load);
+        m_task_t task = MSG_task_create("computation", flops, 0.0, NULL);
+        DEBUG2("compute %g flop%s", flops, ESSE(flops));
         MSG_task_execute(task);
+        comp += flops;
         MSG_task_destroy(task);
     } else {
         DEBUG0("nothing to compute !");

diff --git a/process.h b/process.h
index 1b4f9caa8ad72e1e0ed209598981dee60ae93deb..b124593cae368cc943779ce85bcbc6b861e61088 100644 (file)
--- a/process.h
+++ b/process.h
@@ -27,6 +27,7 @@ public:
     process(int argc, char* argv[]);
     virtual ~process();
 
+    double get_comp() const                { return comp; }
     double get_load() const                { return load; }
 
     int run();
@@ -53,7 +54,7 @@ protected:
 
 private:
     static double total_load_init; // sum of process loads at init
-    static double total_load_running; // summ of loads while running
+    static double total_load_running; // sum of loads while running
     static double total_load_exit; // sum of process loads at exit
 
     typedef MAP_TEMPLATE<m_host_t, neighbor*> rev_neigh_type;
@@ -72,6 +73,8 @@ private:
     unsigned lb_iter;           // counter of load-balancing iterations
     unsigned comp_iter;         // counter of computation iterations
 
+    double comp;                // total computing done so far (flops)
+
     double prev_load_broadcast; // used to ensure that we do not send
                                 // a same information messages
     double load;                // current load