tabula rasa

test/check/speed_cpp.cpp

@@ -0,0 +1,122 @@
+// Copyright 2015-2016 Hans Dembinski
+//
+// Distributed under the Boost Software License, Version 1.0.
+// (See accompanying file LICENSE_1_0.txt
+// or copy at http://www.boost.org/LICENSE_1_0.txt)
+
+#include <boost/histogram/histogram.hpp>
+#include <boost/histogram/axis.hpp>
+#include <boost/random.hpp>
+#include <boost/array.hpp>
+
+#include <algorithm>
+#include <limits>
+#include <vector>
+#include <ctime>
+#include <cstdio>
+
+using namespace std;
+using namespace boost::histogram;
+
+template <typename D>
+struct rng {
+  boost::random::mt19937 r;
+  D d;
+  rng(double a, double b) : d(a, b) {}
+  double operator()() { return d(r); }
+};
+
+vector<double> random_array(unsigned n, int type) {
+  using namespace boost::random;
+  std::vector<double> result;
+  switch (type) {
+    case 0:
+      std::generate_n(std::back_inserter(result), n, rng<uniform_real_distribution<> >(0.0, 1.0));
+      break;
+    case 1:
+      std::generate_n(std::back_inserter(result), n, rng<normal_distribution<> >(0.0, 0.3));
+      break;
+  }
+  return result;
+}
+
+void compare_1d(unsigned n, int distrib)
+{
+  vector<double> r = random_array(n, distrib);
+
+  double best_boost = std::numeric_limits<double>::max();
+  for (unsigned k = 0; k < 10; ++k) {
+    histogram h(regular_axis(100, 0, 1));
+    t = clock();
+    for (unsigned i = 0; i < n; ++i)
+      h.fill(r[i]);
+    t = clock() - t;
+    best_boost = std::min(best_boost, double(t) / CLOCKS_PER_SEC);
+    // printf("root %g this %g\n", hroot.GetSum(), h.sum());
+    assert(hroot.GetSum() == h.sum());
+  }
+
+  printf("1D\n");
+  printf("t[boost] = %.3f\n", best_boost);
+}
+
+void compare_3d(unsigned n, int distrib)
+{
+  vector<double> r = random_array(3 * n, distrib);
+
+  double best_boost = std::numeric_limits<double>::max();
+  for (unsigned k = 0; k < 10; ++k) {
+    histogram h(regular_axis(100, 0, 1),
+                 regular_axis(100, 0, 1),
+                 regular_axis(100, 0, 1));
+    t = clock();
+    for (unsigned i = 0; i < n; ++i)
+      h.fill(r[3 * i], r[3 * i + 1], r[3 * i + 2]);
+    t = clock() - t;
+    best_boost = std::min(best_boost, double(t) / CLOCKS_PER_SEC);
+    assert(hroot.GetSum() == h.sum());
+  } 
+
+  printf("3D\n");
+  printf("t[boost] = %.3f\n", best_boost);
+}
+
+void compare_6d(unsigned n, int distrib)
+{
+  vector<double> r = random_array(6 * n, distrib);
+
+  double best_boost = std::numeric_limits<double>::max();
+  for (unsigned k = 0; k < 10; ++k) {
+    double x[6];
+
+    histogram h(regular_axis(10, 0, 1),
+                regular_axis(10, 0, 1),
+                regular_axis(10, 0, 1),
+                regular_axis(10, 0, 1),
+                regular_axis(10, 0, 1),
+                regular_axis(10, 0, 1));
+
+    t = clock();
+    for (unsigned i = 0; i < n; ++i) {
+      for (unsigned k = 0; k < 6; ++k)
+        x[k] = r[6 * i + k];      
+      h.fill(x, x+6);
+    }
+    t = clock() - t;
+    best_boost = std::min(best_boost, double(t) / CLOCKS_PER_SEC);
+  } 
+
+  printf("6D\n");
+  printf("t[boost] = %.3f\n", best_boost);
+}
+
+int main() {
+  printf("uniform distribution\n");
+  compare_1d(12000000, 0);
+  compare_3d(4000000, 0);
+  compare_6d(2000000, 0);
+  printf("normal distribution\n");
+  compare_1d(12000000, 1);
+  compare_3d(4000000, 1);
+  compare_6d(2000000, 1);
+}