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
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
|
/*
* 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/sid_neighbor_table.hpp>
#include <gridtools/fn/unstructured.hpp>
#include <gridtools/sid/dimension_to_tuple_like.hpp>
#include <fn_select.hpp>
#include <test_environment.hpp>
namespace {
using namespace gridtools;
using namespace fn;
using namespace literals;
struct zavg_stencil {
constexpr auto operator()() const {
return [](auto const &pp, auto const &s) {
std::decay_t<decltype(deref(pp))> tmp = 0;
tuple_util::host_device::for_each(
[&](auto i) {
auto shifted_pp = shift(pp, e2v(), i);
tmp += deref(shifted_pp);
},
meta::rename<tuple, meta::make_indices_c<2>>());
tmp /= 2;
auto ss = deref(s);
return make_tuple(tmp * tuple_get(0_c, ss), tmp * tuple_get(1_c, ss));
};
}
};
struct nabla_stencil {
constexpr auto operator()() const {
return [](auto const &zavg, auto const &sign, auto const &vol) {
using float_t = std::decay_t<decltype(deref(vol))>;
auto signs = deref(sign);
tuple<float_t, float_t> tmp(0, 0);
tuple_util::host_device::for_each(
[&](auto i) {
auto const shifted_zavg = shift(zavg, v2e(), i);
bool const edge_exists = can_deref(shifted_zavg);
auto const value = edge_exists ? deref(shifted_zavg) : tuple<float_t, float_t>(0, 0);
tuple_get(0_c, tmp) += tuple_get(0_c, value) * get<i.value>(signs);
tuple_get(1_c, tmp) += tuple_get(1_c, value) * get<i.value>(signs);
},
meta::rename<tuple, meta::make_indices_c<6>>());
auto v = deref(vol);
return make_tuple(tuple_get(0_c, tmp) / v, tuple_get(1_c, tmp) / v);
};
}
};
struct nabla_stencil_fused {
constexpr auto operator()() const {
return [](auto const &sign, auto const &vol, auto const &pp, auto const &s) {
using float_t = std::decay_t<decltype(deref(vol))>;
auto signs = deref(sign);
tuple<float_t, float_t> tmp(0, 0);
tuple_util::host_device::for_each(
[&](auto i) {
auto const shifted_s = shift(s, v2e(), i);
bool const edge_exists = can_deref(shifted_s);
float_t tmp2 = 0;
tuple_util::host_device::for_each(
[&](auto const &j) {
auto shifted_pp = shift(pp, v2e(), i, e2v(), j);
tmp2 += edge_exists ? deref(shifted_pp) : 0;
},
meta::rename<tuple, meta::make_indices_c<2>>());
tmp2 /= 2;
auto ss = edge_exists ? deref(shifted_s) : tuple<float_t, float_t>(0, 0);
auto zavg = tuple(tmp2 * tuple_get(0_c, ss), tmp2 * tuple_get(1_c, ss));
tuple_get(0_c, tmp) += tuple_get(0_c, zavg) * get<i.value>(signs);
tuple_get(1_c, tmp) += tuple_get(1_c, zavg) * get<i.value>(signs);
},
meta::rename<tuple, meta::make_indices_c<6>>());
auto v = deref(vol);
return make_tuple(tuple_get(0_c, tmp) / v, tuple_get(1_c, tmp) / v);
};
}
};
constexpr inline auto pp = [](int vertex, int k) { return (vertex + k) % 19; };
constexpr inline auto sign = [](int vertex) { return array<int, 6>{0, 1, vertex % 2, 1, (vertex + 1) % 2, 0}; };
constexpr inline auto vol = [](int vertex) { return vertex % 13 + 1; };
constexpr inline auto s = [](int edge, int k) { return tuple((edge + k) % 17, (edge + k) % 7); };
constexpr inline auto zavg = [](auto const &e2v) {
return [&e2v](int edge, int k) {
double tmp = 0.0;
for (int neighbor = 0; neighbor < 2; ++neighbor)
tmp += pp(e2v(edge, neighbor), k);
tmp /= 2.0;
return tuple{tmp * get<0>(s(edge, k)), tmp * get<1>(s(edge, k))};
};
};
constexpr inline auto expected = [](auto const &v2e, auto const &e2v) {
return [&v2e, zavg = zavg(e2v)](int vertex, int k) {
auto res = tuple(0.0, 0.0);
for (int neighbor = 0; neighbor < 6; ++neighbor) {
int edge = v2e(vertex, neighbor);
if (edge != -1) {
get<0>(res) += get<0>(zavg(edge, k)) * sign(vertex)[neighbor];
get<1>(res) += get<1>(zavg(edge, k)) * sign(vertex)[neighbor];
}
}
get<0>(res) /= vol(vertex);
get<1>(res) /= vol(vertex);
return res;
};
};
constexpr inline auto apply_zavg = [](auto &&executor, auto &zavg, auto const &pp, auto const &s) {
executor().arg(zavg).arg(pp).arg(s).assign(0_c, zavg_stencil(), 1_c, 2_c).execute();
};
constexpr inline auto apply_nabla =
[](auto executor, auto &nabla, auto const &zavg, auto const &sign, auto const &vol) {
executor().arg(nabla).arg(zavg).arg(sign).arg(vol).assign(0_c, nabla_stencil(), 1_c, 2_c, 3_c).execute();
};
constexpr inline auto apply_nabla_fused =
[](auto executor, auto &nabla, auto const &sign, auto const &vol, auto const &pp, auto const &s) {
executor()
.arg(nabla)
.arg(sign)
.arg(vol)
.arg(pp)
.arg(s)
.assign(0_c, nabla_stencil_fused(), 1_c, 2_c, 3_c, 4_c)
.execute();
};
constexpr inline auto fencil = [](auto backend,
int nvertices,
int nedges,
int nlevels,
auto const &v2e_table,
auto const &e2v_table,
auto &nabla,
auto const &pp,
auto const &s,
auto const &sign,
auto const &vol) {
using float_t = std::remove_const_t<sid::element_type<decltype(pp)>>;
auto v2e_conn = connectivity<v2e>(v2e_table);
auto e2v_conn = connectivity<e2v>(e2v_table);
auto edge_domain = unstructured_domain({nedges, nlevels}, {}, e2v_conn);
auto vertex_domain = unstructured_domain({nvertices, nlevels}, {}, v2e_conn);
auto edge_backend = make_backend(backend, edge_domain);
auto vertex_backend = make_backend(backend, vertex_domain);
auto alloc = tmp_allocator(backend);
auto zavg = allocate_global_tmp<tuple<float_t, float_t>>(alloc, edge_domain.sizes());
apply_zavg(edge_backend.stencil_executor(), zavg, pp, s);
apply_nabla(vertex_backend.stencil_executor(), nabla, zavg, sign, vol);
};
constexpr inline auto fencil_fused = [](auto backend,
int nvertices,
int nlevels,
auto const &v2e_table,
auto const &e2v_table,
auto &nabla,
auto const &pp,
auto const &s,
auto const &sign,
auto const &vol) {
auto v2e_conn = connectivity<v2e>(v2e_table);
auto e2v_conn = connectivity<e2v>(e2v_table);
auto vertex_domain = unstructured_domain({nvertices, nlevels}, {}, v2e_conn, e2v_conn);
auto vertex_backend = make_backend(backend, vertex_domain);
apply_nabla_fused(vertex_backend.stencil_executor(), nabla, sign, vol, pp, s);
};
static constexpr int vertex_field_id = 0;
static constexpr int edge_field_id = 1;
using k_blocked_backend_t = meta::if_<meta::is_instantiation_of<backend::gpu, fn_backend_t>,
backend::gpu<meta::list<meta::list<integral_constant<int, 0>, integral_constant<int, 32>>,
meta::list<integral_constant<int, 1>, integral_constant<int, 8>>>,
meta::list<meta::list<integral_constant<int, 1>, integral_constant<int, 5>>>>,
fn_backend_t>;
constexpr inline auto make_comp = [](auto backend, auto const &mesh, auto &nabla) {
using mesh_t = std::remove_reference_t<decltype(mesh)>;
using float_t = typename mesh_t::float_t;
return
[backend,
&nabla,
nvertices = mesh.nvertices(),
nedges = mesh.nedges(),
nlevels = mesh.nlevels(),
v2e_table = mesh.v2e_table(),
e2v_table = mesh.e2v_table(),
pp = mesh.template make_const_storage<float_t, vertex_field_id>(pp, mesh.nvertices(), mesh.nlevels()),
sign = mesh.template make_const_storage<array<float_t, 6>>(sign, mesh.nvertices()),
vol = mesh.make_const_storage(vol, mesh.nvertices()),
s = mesh.template make_const_storage<tuple<float_t, float_t>>(s, mesh.nedges(), mesh.nlevels())] {
auto v2e_ptr = sid_neighbor_table::as_neighbor_table<integral_constant<int, 0>,
integral_constant<int, 1>,
mesh_t::max_v2e_neighbors_t::value>(v2e_table);
auto e2v_ptr = sid_neighbor_table::as_neighbor_table<integral_constant<int, 0>,
integral_constant<int, 1>,
mesh_t::max_e2v_neighbors_t::value>(e2v_table);
fencil(backend, nvertices, nedges, nlevels, v2e_ptr, e2v_ptr, nabla, pp, s, sign, vol);
};
};
constexpr inline auto make_comp_fused = [](auto backend, auto const &mesh, auto &nabla) {
using mesh_t = std::remove_reference_t<decltype(mesh)>;
using float_t = typename mesh_t::float_t;
return
[backend,
&nabla,
nvertices = mesh.nvertices(),
nlevels = mesh.nlevels(),
v2e_table = mesh.v2e_table(),
e2v_table = mesh.e2v_table(),
pp = mesh.template make_const_storage<float_t, vertex_field_id>(pp, mesh.nvertices(), mesh.nlevels()),
sign = mesh.template make_const_storage<array<float_t, 6>>(sign, mesh.nvertices()),
vol = mesh.make_const_storage(vol, mesh.nvertices()),
s = mesh.template make_const_storage<tuple<float_t, float_t>>(s, mesh.nedges(), mesh.nlevels())] {
auto v2e_ptr = sid_neighbor_table::as_neighbor_table<integral_constant<int, 0>,
integral_constant<int, 1>,
mesh_t::max_v2e_neighbors_t::value>(v2e_table);
auto e2v_ptr = sid_neighbor_table::as_neighbor_table<integral_constant<int, 0>,
integral_constant<int, 1>,
mesh_t::max_e2v_neighbors_t::value>(e2v_table);
fencil_fused(backend, nvertices, nlevels, v2e_ptr, e2v_ptr, nabla, pp, s, sign, vol);
};
};
constexpr inline auto make_expected = [](auto const &mesh) {
return [v2e_table = mesh.v2e_table(), e2v_table = mesh.e2v_table()](int vertex, int k) {
auto v2e = v2e_table->const_host_view();
auto e2v = e2v_table->const_host_view();
return expected(v2e, e2v)(vertex, k);
};
};
GT_REGRESSION_TEST(fn_unstructured_nabla_field_of_tuples, test_environment<>, fn_backend_t) {
using float_t = typename TypeParam::float_t;
auto mesh = TypeParam::fn_unstructured_mesh();
auto nabla = mesh.template make_storage<tuple<float_t, float_t>>(mesh.nvertices(), mesh.nlevels());
auto comp = make_comp(fn_backend_t(), mesh, nabla);
comp();
auto expected = make_expected(mesh);
TypeParam::verify(expected, nabla);
TypeParam::benchmark("fn_unstructured_nabla_field_of_tuples", comp);
}
GT_REGRESSION_TEST(fn_unstructured_nabla_fused_field_of_tuples, test_environment<>, k_blocked_backend_t) {
using float_t = typename TypeParam::float_t;
auto mesh = TypeParam::fn_unstructured_mesh();
auto nabla = mesh.template make_storage<tuple<float_t, float_t>>(mesh.nvertices(), mesh.nlevels());
auto comp = make_comp_fused(k_blocked_backend_t(), mesh, nabla);
comp();
auto expected = make_expected(mesh);
TypeParam::verify(expected, nabla);
TypeParam::benchmark("fn_unstructured_nabla_fused_field_of_tuples", comp);
}
GT_REGRESSION_TEST(fn_unstructured_nabla_tuple_of_fields, test_environment<>, fn_backend_t) {
using float_t = typename TypeParam::float_t;
auto mesh = TypeParam::fn_unstructured_mesh();
auto nabla0 = mesh.template make_storage<float_t, vertex_field_id>(mesh.nvertices(), mesh.nlevels());
auto nabla1 = mesh.template make_storage<float_t, vertex_field_id>(mesh.nvertices(), mesh.nlevels());
auto nabla =
sid::composite::keys<integral_constant<int, 0>, integral_constant<int, 1>>::make_values(nabla0, nabla1);
auto comp = make_comp(fn_backend_t(), mesh, nabla);
comp();
auto expected = make_expected(mesh);
TypeParam::verify([&](int vertex, int k) { return get<0>(expected(vertex, k)); }, nabla0);
TypeParam::verify([&](int vertex, int k) { return get<1>(expected(vertex, k)); }, nabla1);
TypeParam::benchmark("fn_unstructured_nabla_tuple_of_fields", comp);
}
GT_REGRESSION_TEST(fn_unstructured_nabla_fused_tuple_of_fields, test_environment<>, k_blocked_backend_t) {
using float_t = typename TypeParam::float_t;
auto mesh = TypeParam::fn_unstructured_mesh();
auto nabla0 = mesh.template make_storage<float_t, vertex_field_id>(mesh.nvertices(), mesh.nlevels());
auto nabla1 = mesh.template make_storage<float_t, vertex_field_id>(mesh.nvertices(), mesh.nlevels());
auto nabla =
sid::composite::keys<integral_constant<int, 0>, integral_constant<int, 1>>::make_values(nabla0, nabla1);
auto comp = make_comp_fused(k_blocked_backend_t(), mesh, nabla);
comp();
auto expected = make_expected(mesh);
TypeParam::verify([&](int vertex, int k) { return get<0>(expected(vertex, k)); }, nabla0);
TypeParam::verify([&](int vertex, int k) { return get<1>(expected(vertex, k)); }, nabla1);
TypeParam::benchmark("fn_unstructured_nabla_fused_tuple_of_fields", comp);
}
GT_REGRESSION_TEST(fn_unstructured_nabla_field_of_dimension_to_tuple_like, test_environment<>, fn_backend_t) {
using float_t = typename TypeParam::float_t;
auto mesh = TypeParam::fn_unstructured_mesh();
auto nabla_tmp =
mesh.template make_storage<float_t>(mesh.nvertices(), mesh.nlevels(), integral_constant<int, 2>{});
auto nabla = sid::dimension_to_tuple_like<integral_constant<int, 2>, 2>(nabla_tmp);
auto comp = make_comp(fn_backend_t(), mesh, nabla);
comp();
auto expected = make_expected(mesh);
TypeParam::verify(
[&](int vertex, int k, int t) {
return t == 0 ? get<0>(expected(vertex, k)) : get<1>(expected(vertex, k));
},
nabla_tmp);
TypeParam::benchmark("fn_unstructured_nabla_dimension_to_tuple_like", comp);
}
} // namespace
|
