axis benchmark and simpler circular index calculation

CMakeLists.txt

@@ -151,6 +151,7 @@ if (BUILD_BENCHMARKS)
   endfunction()
 
   benchmark(test/iteration_bench.cpp)
+  benchmark(test/axis_bench.cpp)
 
   add_executable(speed_cpp test/speed_cpp.cpp)
   target_include_directories(speed_cpp PRIVATE include ${Boost_INCLUDE_DIR})

examples/guide_access_bin_counts.cpp

@@ -51,7 +51,7 @@ int main() {
     const auto b1 = x.bin(1); // current bin interval along second axis
     const auto v = *x; // "dereference" to get bin value
     os << boost::format("%i %i [%2i, %i) [%2i, %i): %i\n") % i % j % b0.lower() %
-              b0.upper() % b1.lower() % b1.upper() % *x;
+              b0.upper() % b1.lower() % b1.upper() % v;
   }
 
   std::cout << os.str() << std::flush;

include/boost/histogram/axis/circular.hpp

@@ -71,10 +71,8 @@ public:
   /// Returns the bin index for the passed argument.
   int operator()(value_type x) const noexcept {
     const auto z = std::floor((x - phase_) / delta_);
-    if (std::isfinite(z)) {
-      const auto i = static_cast<int>(z) % base_type::size();
-      return i + (i < 0) * base_type::size();
-    }
+    if (std::isfinite(z))
+      return static_cast<int>(z - std::floor(z / base_type::size()) * base_type::size());
     return base_type::size();
   }
 

include/boost/histogram/histogram_fwd.hpp

@@ -23,9 +23,10 @@ struct null_type {};
 
 enum class option_type {
   none = 0,
-  underflow = 1,
-  overflow = 2,
-  uoflow = 3,
+  underflow = 1 << 0,
+  overflow = 1 << 1,
+  uoflow = 1 << 2,
+  circular = 1 << 3,
 };
 
 constexpr inline option_type operator|(option_type a, option_type b) {

test/axis_bench.cpp

@@ -0,0 +1,95 @@
+// Copyright 2018 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/axis.hpp>
+#include <benchmark/benchmark.h>
+
+using namespace boost::histogram;
+
+template <bool include_extra_bins>
+static void null(benchmark::State& state) {
+  for (auto _ : state) {
+    for (volatile int i = 0 - include_extra_bins; i < 10 + include_extra_bins; ++i);
+  }
+}
+
+template <bool include_extra_bins>
+static void regular(benchmark::State& state) {
+  auto a = axis::regular<>(10, 0, 10);
+  for (auto _ : state) {
+    for (int i = 0 - include_extra_bins; i < a.size() + include_extra_bins; ++i)
+      benchmark::DoNotOptimize(a(i));
+  }
+}
+
+template <bool include_extra_bins>
+static void circular(benchmark::State& state) {
+  auto a = axis::circular<>(10, 0, 10);
+  for (auto _ : state) {
+    for (int i = 0 - include_extra_bins; i < a.size() + include_extra_bins; ++i)
+      benchmark::DoNotOptimize(a(i));
+  }
+}
+
+template <bool include_extra_bins>
+static void integer_int(benchmark::State& state) {
+  auto a = axis::integer<int>(0, 10);
+  for (auto _ : state) {
+    for (int i = 0 - include_extra_bins; i < a.size() + include_extra_bins; ++i)
+      benchmark::DoNotOptimize(a(i));
+  }
+}
+
+template <bool include_extra_bins>
+static void integer_double(benchmark::State& state) {
+  auto a = axis::integer<double>(0, 10);
+  for (auto _ : state) {
+    for (int i = 0 - include_extra_bins; i < a.size() + include_extra_bins; ++i)
+      benchmark::DoNotOptimize(a(i));
+  }
+}
+
+template <bool include_extra_bins>
+static void variable(benchmark::State& state) {
+  auto a = axis::variable<>({0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10});
+  for (auto _ : state) {
+    for (int i = 0 - include_extra_bins; i < a.size() + include_extra_bins; ++i)
+      benchmark::DoNotOptimize(a(i));
+  }
+}
+
+template <bool include_extra_bins>
+static void category(benchmark::State& state) {
+  auto a = axis::category<int>({0, 1, 2, 3, 4, 5, 6, 7, 8, 9});
+  for (auto _ : state) {
+    for (int i = 0 - include_extra_bins; i < a.size() + include_extra_bins; ++i)
+      benchmark::DoNotOptimize(a(i));
+  }
+}
+
+template <bool include_extra_bins>
+static void variant(benchmark::State& state) {
+  auto a = axis::variant<axis::regular<>>(axis::regular<>(10, 0, 10));
+  for (auto _ : state) {
+    for (int i = 0 - include_extra_bins; i < a.size() + include_extra_bins; ++i)
+      benchmark::DoNotOptimize(a(i));
+  }
+}
+
+BENCHMARK_TEMPLATE(null, false);
+BENCHMARK_TEMPLATE(null, true);
+BENCHMARK_TEMPLATE(regular, false);
+BENCHMARK_TEMPLATE(regular, true);
+BENCHMARK_TEMPLATE(circular, false);
+BENCHMARK_TEMPLATE(circular, true);
+BENCHMARK_TEMPLATE(integer_int, false);
+BENCHMARK_TEMPLATE(integer_int, true);
+BENCHMARK_TEMPLATE(integer_double, false);
+BENCHMARK_TEMPLATE(integer_double, true);
+BENCHMARK_TEMPLATE(variable, false);
+BENCHMARK_TEMPLATE(variable, true);
+BENCHMARK_TEMPLATE(category, false);
+BENCHMARK_TEMPLATE(category, true);