File: fn_cartesian_vertical_advection.cpp

package info (click to toggle)
gridtools 2.3.9-1
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid
  • size: 29,480 kB
  • sloc: cpp: 228,792; python: 17,561; javascript: 9,164; ansic: 4,101; sh: 850; makefile: 231; f90: 201
file content (188 lines) | stat: -rw-r--r-- 8,543 bytes parent folder | download | duplicates (2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
/*
 * GridTools
 *
 * Copyright (c) 2014-2023, ETH Zurich
 * All rights reserved.
 *
 * Please, refer to the LICENSE file in the root directory.
 * SPDX-License-Identifier: BSD-3-Clause
 */

#include <gtest/gtest.h>

#include <gridtools/fn/cartesian.hpp>
#include <gridtools/stencil/global_parameter.hpp>

#include <fn_select.hpp>
#include <nvcc_workarounds.hpp>
#include <test_environment.hpp>

#include "../vertical_advection_repository.hpp"

namespace {
    using namespace gridtools;
    using namespace fn;
    using namespace cartesian;
    using namespace literals;
    using stencil::global_parameter;

    struct u_forward_scan : fwd {
        static GT_FUNCTION constexpr auto prologue() {
            return nvcc_workarounds::make_1_tuple(scan_pass(
                [](auto /*acc*/,
                    auto const &utens_stage,
                    auto const &utens,
                    auto const &u_stage,
                    auto const &u_pos,
                    auto const &wcon,
                    auto const &dtr_stage) {
                    constexpr auto i = dim::i();
                    constexpr auto k = dim::k();
                    using float_t = std::decay_t<decltype(deref(wcon))>;
                    auto gcv = float_t(0.25) * (deref(shift(wcon, i, 1, k, 1)) + deref(shift(wcon, k, 1)));
                    auto cs = gcv * float_t(BET_M);
                    auto c = gcv * float_t(BET_P);
                    auto b = deref(dtr_stage) - c;
                    auto correction = -cs * (deref(shift(u_stage, k, 1)) - deref(u_stage));
                    auto d = deref(dtr_stage) * deref(u_pos) + deref(utens) + deref(utens_stage) + correction;
                    auto divided = float_t(1) / b;
                    return make_tuple(c * divided, d * divided);
                },
                host_device::identity()));
        }

        static GT_FUNCTION constexpr auto body() {
            return scan_pass(
                [](auto acc,
                    auto const &utens_stage,
                    auto const &utens,
                    auto const &u_stage,
                    auto const &u_pos,
                    auto const &wcon,
                    auto const &dtr_stage) {
                    constexpr auto i = dim::i();
                    constexpr auto k = dim::k();
                    using float_t = std::decay_t<decltype(deref(wcon))>;
                    auto gav = -float_t(0.25) * (deref(shift(wcon, i, 1)) + deref(wcon));
                    auto gcv = float_t(0.25) * (deref(shift(wcon, i, 1, k, 1)) + deref(shift(wcon, k, 1)));
                    auto as = gav * float_t(BET_M);
                    auto cs = gcv * float_t(BET_M);
                    auto a = gav * float_t(BET_P);
                    auto c = gcv * float_t(BET_P);
                    auto b = deref(dtr_stage) - a - c;
                    auto correction = -as * (deref(shift(u_stage, k, -1)) - deref(u_stage)) -
                                      cs * (deref(shift(u_stage, k, -1)) - deref(u_stage));
                    auto d = deref(dtr_stage) * deref(u_pos) + deref(utens) + deref(utens_stage) + correction;
                    auto [cp, dp] = acc;
                    auto divided = float_t(1) / (b - cp * a);
                    return make_tuple(c * divided, (d - dp * a) * divided);
                },
                host_device::identity());
        }

        static GT_FUNCTION constexpr auto epilogue() {
            return nvcc_workarounds::make_1_tuple(scan_pass(
                [](auto acc,
                    auto const &utens_stage,
                    auto const &utens,
                    auto const &u_stage,
                    auto const &u_pos,
                    auto const &wcon,
                    auto const &dtr_stage) {
                    constexpr auto i = dim::i();
                    constexpr auto k = dim::k();
                    using float_t = std::decay_t<decltype(deref(wcon))>;
                    auto gav = -float_t(0.25) * (deref(shift(wcon, i, 1)) + deref(wcon));
                    auto as = gav * float_t(BET_M);
                    auto a = gav * float_t(BET_P);
                    auto b = deref(dtr_stage) - a;
                    auto correction = -as * (deref(shift(u_stage, k, -1)) - deref(u_stage));
                    auto d = deref(dtr_stage) * deref(u_pos) + deref(utens) + deref(utens_stage) + correction;
                    auto [cp, dp] = acc;
                    auto divided = float_t(1) / (b - cp * a);
                    return make_tuple(float_t(0), (d - dp * a) * divided);
                },
                host_device::identity()));
        }
    };

    struct u_backward_scan : bwd {
        static GT_FUNCTION constexpr auto prologue() {
            return nvcc_workarounds::make_1_tuple(scan_pass(
                [](auto /*acc*/, auto const &cd, auto const &u_pos, auto const &dtr_stage) {
                    auto d = tuple_get(1_c, deref(cd));
                    return make_tuple(deref(dtr_stage) * (d - deref(u_pos)), d);
                },
                [](auto const &acc) { return tuple_get(0_c, acc); }));
        }

        static GT_FUNCTION constexpr auto body() {
            return scan_pass(
                [](auto acc, auto const &cd, auto const &u_pos, auto const &dtr_stage) {
                    auto [c, d] = deref(cd);
                    auto data = d - c * tuple_get(1_c, acc);
                    return tuple(deref(dtr_stage) * (data - deref(u_pos)), data);
                },
                [](auto const &acc) { return tuple_get(0_c, acc); });
        }
    };

    constexpr inline auto vadv_solver = [](auto &&executor,
                                            auto &cd,
                                            auto &utens_stage,
                                            auto const &utens,
                                            auto const &u_stage,
                                            auto const &u_pos,
                                            auto const &wcon,
                                            auto const &dtr_stage) {
        using float_t = sid::element_type<decltype(utens_stage)>;
        executor()
            .arg(cd)
            .arg(utens_stage)
            .arg(utens)
            .arg(u_stage)
            .arg(u_pos)
            .arg(wcon)
            .arg(dtr_stage)
            .assign(0_c, u_forward_scan(), tuple<float_t, float_t>(0, 0), 1_c, 2_c, 3_c, 4_c, 5_c, 6_c)
            .assign(1_c, u_backward_scan(), tuple<float_t, float_t>(0, 0), 0_c, 4_c, 6_c)
            .execute();
    };

    GT_REGRESSION_TEST(fn_cartesian_vertical_advection, vertical_test_environment<3>, fn_backend_t) {
        using float_t = typename TypeParam::float_t;
        vertical_advection_repository repo{TypeParam::d(0), TypeParam::d(1), TypeParam::d(2)};

        auto fencil = [](int i,
                          int j,
                          int k,
                          auto &utens_stage,
                          auto const &utens,
                          auto const &u_stage,
                          auto const &u_pos,
                          auto const &wcon,
                          auto const &dtr_stage) {
            using sizes_t = hymap::keys<dim::i, dim::j, dim::k>::values<int, int, int>;
            auto be = fn_backend_t();
            auto domain = cartesian_domain(sizes_t(i - 6, j - 6, k), sizes_t(3, 3, 0));
            auto backend = make_backend(be, domain);
            auto alloc = tmp_allocator(be);
            auto cd = allocate_global_tmp<tuple<float_t, float_t>>(alloc, sizes_t(i, j, k));
            vadv_solver(backend.vertical_executor(), cd, utens_stage, utens, u_stage, u_pos, wcon, dtr_stage);
        };

        auto utens_stage = TypeParam::make_storage(repo.utens_stage_in);
        auto comp = [&,
                        utens = TypeParam::make_storage(repo.utens),
                        u_stage = TypeParam::make_storage(repo.u_stage),
                        u_pos = TypeParam::make_storage(repo.u_pos),
                        wcon = TypeParam::make_storage(repo.wcon),
                        dtr_stage = stencil::global_parameter(float_t(repo.dtr_stage))] {
            fencil(
                TypeParam::d(0), TypeParam::d(1), TypeParam::d(2), utens_stage, utens, u_stage, u_pos, wcon, dtr_stage);
        };
        comp();
        TypeParam::verify(repo.utens_stage_out, utens_stage);
        TypeParam::benchmark("fn_cartesian_vertical_advection", comp);
    }
} // namespace