From cb4adb40eb6a30d55070e5ccfc9bec95873c594b Mon Sep 17 00:00:00 2001
From: Yunsong Wang <yunsongw@nvidia.com>
Date: Fri, 7 Jan 2022 15:15:05 -0500
Subject: [PATCH] Add sum scan benchmark

---
 cpp/benchmarks/CMakeLists.txt                 |   1 +
 .../groupby/group_scan_benchmark.cu           | 119 ++++++++++++++++++
 2 files changed, 120 insertions(+)
 create mode 100644 cpp/benchmarks/groupby/group_scan_benchmark.cu

diff --git a/cpp/benchmarks/CMakeLists.txt b/cpp/benchmarks/CMakeLists.txt
index 63f6857ee08..370f84fc14a 100644
--- a/cpp/benchmarks/CMakeLists.txt
+++ b/cpp/benchmarks/CMakeLists.txt
@@ -177,6 +177,7 @@ ConfigureBench(
   groupby/group_shift_benchmark.cu
   groupby/group_struct_benchmark.cu
   groupby/group_no_requests_benchmark.cu
+  groupby/group_scan_benchmark.cu
 )
 
 # ##################################################################################################
diff --git a/cpp/benchmarks/groupby/group_scan_benchmark.cu b/cpp/benchmarks/groupby/group_scan_benchmark.cu
new file mode 100644
index 00000000000..b22b8ea3f14
--- /dev/null
+++ b/cpp/benchmarks/groupby/group_scan_benchmark.cu
@@ -0,0 +1,119 @@
+/*
+ * Copyright (c) 2019-2022, NVIDIA CORPORATION.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <cudf/copying.hpp>
+#include <cudf/detail/aggregation/aggregation.hpp>
+#include <cudf/groupby.hpp>
+#include <cudf/sorting.hpp>
+#include <cudf/table/table.hpp>
+#include <cudf_test/column_wrapper.hpp>
+#include <fixture/benchmark_fixture.hpp>
+#include <synchronization/synchronization.hpp>
+
+#include <random>
+
+class Groupby : public cudf::benchmark {
+};
+
+template <typename T>
+T random_int(T min, T max)
+{
+  static unsigned seed = 13377331;
+  static std::mt19937 engine{seed};
+  static std::uniform_int_distribution<T> uniform{min, max};
+
+  return uniform(engine);
+}
+
+void BM_basic_sum_scan(benchmark::State& state)
+{
+  using wrapper = cudf::test::fixed_width_column_wrapper<int64_t>;
+
+  const cudf::size_type column_size{(cudf::size_type)state.range(0)};
+
+  auto data_it = cudf::detail::make_counting_transform_iterator(
+    0, [=](cudf::size_type row) { return random_int(0, 100); });
+
+  wrapper keys(data_it, data_it + column_size);
+  wrapper vals(data_it, data_it + column_size);
+
+  cudf::groupby::groupby gb_obj(cudf::table_view({keys, keys, keys}));
+
+  std::vector<cudf::groupby::scan_request> requests;
+  requests.emplace_back(cudf::groupby::scan_request());
+  requests[0].values = vals;
+  requests[0].aggregations.push_back(cudf::make_sum_aggregation<cudf::groupby_scan_aggregation>());
+
+  for (auto _ : state) {
+    cuda_event_timer timer(state, true);
+
+    auto result = gb_obj.scan(requests);
+  }
+}
+
+BENCHMARK_DEFINE_F(Groupby, BasicSumScan)(::benchmark::State& state) { BM_basic_sum_scan(state); }
+
+BENCHMARK_REGISTER_F(Groupby, BasicSumScan)
+  ->UseManualTime()
+  ->Unit(benchmark::kMillisecond)
+  ->Arg(1000000)
+  ->Arg(10000000)
+  ->Arg(100000000);
+
+void BM_pre_sorted_sum_scan(benchmark::State& state)
+{
+  using wrapper = cudf::test::fixed_width_column_wrapper<int64_t>;
+
+  const cudf::size_type column_size{(cudf::size_type)state.range(0)};
+
+  auto data_it = cudf::detail::make_counting_transform_iterator(
+    0, [=](cudf::size_type row) { return random_int(0, 100); });
+  auto valid_it = cudf::detail::make_counting_transform_iterator(
+    0, [=](cudf::size_type row) { return random_int(0, 100) < 90; });
+
+  wrapper keys(data_it, data_it + column_size);
+  wrapper vals(data_it, data_it + column_size, valid_it);
+
+  auto keys_table  = cudf::table_view({keys});
+  auto sort_order  = cudf::sorted_order(keys_table);
+  auto sorted_keys = cudf::gather(keys_table, *sort_order);
+  // No need to sort values using sort_order because they were generated randomly
+
+  cudf::groupby::groupby gb_obj(*sorted_keys, cudf::null_policy::EXCLUDE, cudf::sorted::YES);
+
+  std::vector<cudf::groupby::scan_request> requests;
+  requests.emplace_back(cudf::groupby::scan_request());
+  requests[0].values = vals;
+  requests[0].aggregations.push_back(cudf::make_sum_aggregation<cudf::groupby_scan_aggregation>());
+
+  for (auto _ : state) {
+    cuda_event_timer timer(state, true);
+
+    auto result = gb_obj.scan(requests);
+  }
+}
+
+BENCHMARK_DEFINE_F(Groupby, PreSortedSumScan)(::benchmark::State& state)
+{
+  BM_pre_sorted_sum_scan(state);
+}
+
+BENCHMARK_REGISTER_F(Groupby, PreSortedSumScan)
+  ->UseManualTime()
+  ->Unit(benchmark::kMillisecond)
+  ->Arg(1000000)
+  ->Arg(10000000)
+  ->Arg(100000000);