//@HEADER // ************************************************************************ // // Kokkos v. 4.0 // Copyright (2022) National Technology & Engineering // Solutions of Sandia, LLC (NTESS). // // Under the terms of Contract DE-NA0003525 with NTESS, // the U.S. Government retains certain rights in this software. // // Part of Kokkos, under the Apache License v2.0 with LLVM Exceptions. // See https://kokkos.org/LICENSE for license information. // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //@HEADER #include #include namespace Test { namespace stdalgos { namespace TeamPartitionPoint { namespace KE = Kokkos::Experimental; template struct UnifDist; template <> struct UnifDist { using dist_type = std::uniform_int_distribution; std::mt19937 m_gen; dist_type m_dist; UnifDist(int a, int b, std::size_t seedIn) : m_dist(a, b) { m_gen.seed(seedIn); } int operator()() { return m_dist(m_gen); } }; template struct GreaterThanValueFunctor { ValueType m_val; KOKKOS_INLINE_FUNCTION GreaterThanValueFunctor(ValueType val) : m_val(val) {} KOKKOS_INLINE_FUNCTION bool operator()(ValueType val) const { return (val > m_val); } }; template struct TestFunctorA { ViewType m_view; DistancesViewType m_distancesView; IntraTeamSentinelView m_intraTeamSentinelView; ValueType m_threshold; int m_apiPick; TestFunctorA(const ViewType view, const DistancesViewType distancesView, const IntraTeamSentinelView intraTeamSentinelView, ValueType threshold, int apiPick) : m_view(view), m_distancesView(distancesView), m_intraTeamSentinelView(intraTeamSentinelView), m_threshold(threshold), m_apiPick(apiPick) {} template KOKKOS_INLINE_FUNCTION void operator()(const MemberType& member) const { const auto myRowIndex = member.league_rank(); auto myRowView = Kokkos::subview(m_view, myRowIndex, Kokkos::ALL()); ptrdiff_t resultDist = 0; GreaterThanValueFunctor predicate(m_threshold); if (m_apiPick == 0) { const auto it = KE::partition_point(member, KE::cbegin(myRowView), KE::cend(myRowView), predicate); resultDist = KE::distance(KE::cbegin(myRowView), it); Kokkos::single(Kokkos::PerTeam(member), [=, *this]() { m_distancesView(myRowIndex) = resultDist; }); } else if (m_apiPick == 1) { const auto it = KE::partition_point(member, myRowView, predicate); resultDist = KE::distance(KE::begin(myRowView), it); Kokkos::single(Kokkos::PerTeam(member), [=, *this]() { m_distancesView(myRowIndex) = resultDist; }); } // store result of checking if all members have their local // values matching the one stored in m_distancesView member.team_barrier(); const bool intraTeamCheck = team_members_have_matching_result( member, resultDist, m_distancesView(myRowIndex)); Kokkos::single(Kokkos::PerTeam(member), [=, *this]() { m_intraTeamSentinelView(myRowIndex) = intraTeamCheck; }); } }; template void test_A(std::size_t numTeams, std::size_t numCols, int apiId, const std::string& sIn) { /* description: use a rank-2 view randomly filled with values in a range (a,b) and run a team-level (one team per row) partition_point with predicate = IsGreaterThanValue where threshold is set to a number larger than b above */ const auto threshold = static_cast(1103); const auto valueForSureGreater = static_cast(2103); const auto valueForSureSmaller = static_cast(111); // ----------------------------------------------- // prepare data // ----------------------------------------------- // construct in memory space associated with default exespace auto dataView = create_view(LayoutTag{}, numTeams, numCols, "dataView"); // dataView might not deep copyable (e.g. strided layout) so to // randomize it, we make a new view that is for sure deep copyable, // modify it on the host, deep copy to device and then launch // a kernel to copy to dataView auto dataView_dc = create_deep_copyable_compatible_view_with_same_extent(dataView); auto dataView_dc_h = create_mirror_view(Kokkos::HostSpace(), dataView_dc); if (sIn == "trivialEmpty") { // do nothing } else if (sIn == "allTrue") { // randomly fill with values greater than threshold // so that all elements in each row satisfy the predicate // so this counts as being partitioned Kokkos::Random_XorShift64_Pool pool( 452377); Kokkos::fill_random(dataView_dc_h, pool, ValueType(2001), ValueType(2501)); } else if (sIn == "allFalse") { // randomly fill the view with values smaller than threshold // and even in this case each row counts as partitioned Kokkos::Random_XorShift64_Pool pool( 452377); Kokkos::fill_random(dataView_dc_h, pool, ValueType(0), ValueType(101)); } else if (sIn == "random") { // randomly select a location and make all values before that // larger than threshol and all values after to be smaller than threshold // so that this picked location does partition the range UnifDist indexProducer(0, numCols - 1, 3432779); for (std::size_t i = 0; i < dataView_dc_h.extent(0); ++i) { const std::size_t a = indexProducer(); for (std::size_t j = 0; j < a; ++j) { dataView_dc_h(i, j) = valueForSureGreater; } for (std::size_t j = a; j < numCols; ++j) { dataView_dc_h(i, j) = valueForSureSmaller; } } } // copy to dataView_dc and then to dataView Kokkos::deep_copy(dataView_dc, dataView_dc_h); // use CTAD CopyFunctorRank2 F1(dataView_dc, dataView); Kokkos::parallel_for("copy", dataView.extent(0) * dataView.extent(1), F1); // ----------------------------------------------- // launch kokkos kernel // ----------------------------------------------- using space_t = Kokkos::DefaultExecutionSpace; Kokkos::TeamPolicy policy(numTeams, Kokkos::AUTO()); // to verify that things work, each team stores the result // and then we check that these match what we expect Kokkos::View distancesView("distances", numTeams); // sentinel to check if all members of the team compute the same result Kokkos::View intraTeamSentinelView("intraTeamSameResult", numTeams); // use CTAD for functor TestFunctorA fnc(dataView, distancesView, intraTeamSentinelView, threshold, apiId); Kokkos::parallel_for(policy, fnc); // ----------------------------------------------- // check // ----------------------------------------------- auto distancesView_h = create_host_space_copy(distancesView); auto dataViewAfterOp_h = create_host_space_copy(dataView); auto intraTeamSentinelView_h = create_host_space_copy(intraTeamSentinelView); GreaterThanValueFunctor predicate(threshold); for (std::size_t i = 0; i < dataView_dc_h.extent(0); ++i) { auto myRow = Kokkos::subview(dataView_dc_h, i, Kokkos::ALL()); const auto stdResult = std::partition_point(KE::cbegin(myRow), KE::cend(myRow), predicate); // our result must match std const std::size_t stdDistance = KE::distance(KE::cbegin(myRow), stdResult); ASSERT_EQ(stdDistance, distancesView_h(i)); ASSERT_TRUE(intraTeamSentinelView_h(i)); } expect_equal_host_views(dataView_dc_h, dataViewAfterOp_h); } template void run_all_scenarios(const std::string& name, const std::vector& cols) { for (int numTeams : teamSizesToTest) { for (const auto& numCols : cols) { for (int apiId : {0, 1}) { test_A(numTeams, numCols, apiId, name); } } } } TEST(std_algorithms_partition_point_team_test, empty) { const std::string name = "trivialEmpty"; const std::vector cols = {0}; run_all_scenarios(name, cols); run_all_scenarios(name, cols); run_all_scenarios(name, cols); } TEST(std_algorithms_partition_point_team_test, all_true) { const std::string name = "allTrue"; const std::vector cols = {13, 101, 1444, 5153}; run_all_scenarios(name, cols); run_all_scenarios(name, cols); run_all_scenarios(name, cols); } TEST(std_algorithms_partition_point_team_test, all_false) { const std::string name = "allFalse"; const std::vector cols = {13, 101, 1444, 5153}; run_all_scenarios(name, cols); run_all_scenarios(name, cols); run_all_scenarios(name, cols); } TEST(std_algorithms_partition_point_team_test, random) { const std::string name = "random"; const std::vector cols = {13, 101, 1444, 5153}; run_all_scenarios(name, cols); run_all_scenarios(name, cols); run_all_scenarios(name, cols); } } // namespace TeamPartitionPoint } // namespace stdalgos } // namespace Test