Merge pull request #24778 from vespa-engine/havardpe/generator-micro-bench

added simple micro-bench to show generator overhead

3 files changed, 66 insertions, 0 deletions

diff --git a/vespalib/src/tests/coro/generator/.gitignore b/vespalib/src/tests/coro/generator/.gitignore
new file mode 100644
index 00000000000..748003a81fe
--- /dev/null
+++ b/vespalib/src/tests/coro/generator/.gitignore
@@ -0,0 +1 @@
+/vespalib_generator_bench_app
diff --git a/vespalib/src/tests/coro/generator/CMakeLists.txt b/vespalib/src/tests/coro/generator/CMakeLists.txt
index b4f59c69451..e2534274f7c 100644
--- a/vespalib/src/tests/coro/generator/CMakeLists.txt
+++ b/vespalib/src/tests/coro/generator/CMakeLists.txt
@@ -6,4 +6,11 @@ vespa_add_executable(vespalib_generator_test_app TEST
     vespalib
     GTest::GTest
 )
+vespa_add_executable(vespalib_generator_bench_app TEST
+    SOURCES
+    generator_bench.cpp
+    DEPENDS
+    vespalib
+    GTest::GTest
+)
 vespa_add_test(NAME vespalib_generator_test_app COMMAND vespalib_generator_test_app)
diff --git a/vespalib/src/tests/coro/generator/generator_bench.cpp b/vespalib/src/tests/coro/generator/generator_bench.cpp
new file mode 100644
index 00000000000..664132b7ba4
--- /dev/null
+++ b/vespalib/src/tests/coro/generator/generator_bench.cpp
@@ -0,0 +1,58 @@
+// Copyright Yahoo. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+
+#include <vespa/vespalib/coro/generator.h>
+#include <vespa/vespalib/util/benchmark_timer.h>
+#include <vespa/vespalib/gtest/gtest.h>
+#include <ranges>
+#include <vector>
+
+using vespalib::coro::Generator;
+using vespalib::BenchmarkTimer;
+
+std::vector<size_t> make_data() __attribute__((noinline));
+std::vector<size_t> make_data(size_t size) {
+    std::vector<size_t> data;
+    for (size_t i = 0; i < size; ++i) {
+        data.push_back(i);
+    }
+    return data;
+}
+
+template <std::ranges::input_range T>
+size_t calc_sum(T&& values) {
+    size_t sum = 0;
+    for (auto&& value: values) {
+        sum += value;
+    }
+    return sum;
+}
+
+size_t calc_sum_direct(const std::vector<size_t> &values) {
+    return calc_sum(values);
+}
+
+size_t calc_sum_wrapped(const std::vector<size_t> &values) {
+    return calc_sum([](const std::vector<size_t> &inner_values)->Generator<size_t>
+                    {
+                        for (auto&& value: inner_values) {
+                            co_yield value;
+                        }
+                    }(values));
+}
+
+TEST(GeneratorBench, direct_vs_wrapped_vector_for_loop) {
+    std::vector<size_t> data = make_data(100000);
+    double direct_ms = BenchmarkTimer::benchmark([&data](){
+                                                     size_t sink = calc_sum_direct(data);
+                                                     (void) sink;
+                                                 }, 5.0) * 1000.0;
+    fprintf(stderr, "direct: %g ms\n", direct_ms);
+    double wrapped_ms = BenchmarkTimer::benchmark([&data](){
+                                                      size_t sink = calc_sum_wrapped(data);
+                                                      (void) sink;
+                                                  }, 5.0) * 1000.0;
+    fprintf(stderr, "wrapped: %g ms\n", wrapped_ms);
+    fprintf(stderr, "ratio: %g\n", (wrapped_ms/direct_ms));
+}
+
+GTEST_MAIN_RUN_ALL_TESTS()