forked from stanfordhpccenter/HTR-solver
-
Notifications
You must be signed in to change notification settings - Fork 0
/
task_helper.h
210 lines (192 loc) · 7.26 KB
/
task_helper.h
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
// Copyright (c) "2019, by Stanford University
// Developer: Mario Di Renzo
// Affiliation: Center for Turbulence Research, Stanford University
// URL: https://ctr.stanford.edu
// Citation: Di Renzo, M., Lin, F., and Urzay, J. (2020).
// HTR solver: An open-source exascale-oriented task-based
// multi-GPU high-order code for hypersonic aerothermodynamics.
// Computer Physics Communications 255, 107262"
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are met:
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
// ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
// WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE FOR ANY
// DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
// (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
// LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
// ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#ifndef __TASK_HELPER_H__
#define __TASK_HELPER_H__
template<typename FT, int N, typename T = coord_t> using AccessorRO = FieldAccessor< READ_ONLY,FT,N,T,Realm::AffineAccessor<FT,N,T> >;
template<typename FT, int N, typename T = coord_t> using AccessorRW = FieldAccessor<READ_WRITE,FT,N,T,Realm::AffineAccessor<FT,N,T> >;
template<typename FT, int N, typename T = coord_t> using AccessorWO = FieldAccessor<WRITE_ONLY,FT,N,T,Realm::AffineAccessor<FT,N,T> >;
#ifndef __CUDA_HD__
#ifdef __CUDACC__
#define __CUDA_HD__ __host__ __device__
#else
#define __CUDA_HD__
#endif
#endif
#ifndef __CUDA_H__
#ifdef __CUDACC__
#define __CUDA_H__ __device__
#else
#define __CUDA_H__
#endif
#endif
#ifndef __UNROLL__
#ifdef __CUDACC__
#define __UNROLL__ #pragma unroll
#else
#define __UNROLL__
#endif
#endif
#ifndef nSpec
#error "nSpec is undefined"
#endif
#ifndef nEq
#error "nEq is undefined"
#endif
template<typename T, int SIZE>
struct MyArray {
public:
__CUDA_HD__
inline T& operator[](int index) {
#ifdef BOUNDS_CHECKS
assert(index >= 0);
assert(index < SIZE);
#endif
return v[index];
}
__CUDA_HD__
inline T operator[](int index) const {
#ifdef BOUNDS_CHECKS
assert(index >= 0);
assert(index < SIZE);
#endif
return v[index];
}
public:
T v[SIZE];
};
typedef MyArray<double, 3> Vec3;
typedef MyArray<double, nEq> VecNEq;
typedef MyArray<double, nSpec> VecNSp;
enum direction {
Xdir,
Ydir,
Zdir
};
enum side {
Plus,
Minus
};
// Utility that computes the size of a Rect<3>
template<direction dir>
__CUDA_HD__
inline coord_t getSize(const Rect<3> bounds);
template<>
__CUDA_HD__
inline coord_t getSize<Xdir>(const Rect<3> bounds) { return bounds.hi.x - bounds.lo.x + 1; };
template<>
__CUDA_HD__
inline coord_t getSize<Ydir>(const Rect<3> bounds) { return bounds.hi.y - bounds.lo.y + 1; };
template<>
__CUDA_HD__
inline coord_t getSize<Zdir>(const Rect<3> bounds) { return bounds.hi.z - bounds.lo.z + 1; };
// Utility that computes the stencil point warping the point aroud periodic boundaries
template<direction dir, side s>
static inline Point<3> warpPeriodic(const Rect<3> bounds, Point<3> p, const coord_t size, const int off);
template<>
__CUDA_HD__
inline Point<3> warpPeriodic<Xdir, Minus>(const Rect<3> bounds, Point<3> p, const coord_t size, const int off) {
return Point<3>(((p.x + off - bounds.lo.x) % size + size) % size + bounds.lo.x, p.y, p.z);
};
template<>
__CUDA_HD__
inline Point<3> warpPeriodic<Xdir, Plus>(const Rect<3> bounds, Point<3> p, const coord_t size, const int off) {
return Point<3>((p.x + off - bounds.lo.x) % size + bounds.lo.x, p.y, p.z);
};
template<>
__CUDA_HD__
inline Point<3> warpPeriodic<Ydir, Minus>(const Rect<3> bounds, Point<3> p, const coord_t size, const int off) {
return Point<3>(p.x, ((p.y + off - bounds.lo.y) % size + size) % size + bounds.lo.y, p.z);
};
template<>
__CUDA_HD__
inline Point<3> warpPeriodic<Ydir, Plus>(const Rect<3> bounds, Point<3> p, const coord_t size, const int off) {
return Point<3>(p.x, (p.y + off - bounds.lo.y) % size + bounds.lo.y, p.z);
};
template<>
__CUDA_HD__
inline Point<3> warpPeriodic<Zdir, Minus>(const Rect<3> bounds, Point<3> p, const coord_t size, const int off) {
return Point<3>(p.x, p.y, ((p.z + off - bounds.lo.z) % size + size) % size + bounds.lo.z);
};
template<>
__CUDA_HD__
inline Point<3> warpPeriodic<Zdir, Plus>(const Rect<3> bounds, Point<3> p, const coord_t size, const int off) {
return Point<3>(p.x, p.y, (p.z + off - bounds.lo.z) % size + bounds.lo.z);
};
// Utility that registers tasks
namespace TaskHelper {
template<class T>
void base_cpu_wrapper(const Task *task,
const std::vector<PhysicalRegion> ®ions,
Context ctx, Runtime *runtime)
{
assert(task->arglen == 0);
assert(task->local_arglen == sizeof(typename T::Args));
const typename T::Args *a = (typename T::Args*)task->local_args;
T::cpu_base_impl(*a, regions, task->futures, ctx, runtime);
}
#ifdef LEGION_USE_CUDA
template<typename T>
void base_gpu_wrapper(const Task *task,
const std::vector<PhysicalRegion> ®ions,
Context ctx, Runtime *runtime)
{
assert(task->arglen == 0);
assert(task->local_arglen == sizeof(typename T::Args));
const typename T::Args *a = (typename T::Args*)task->local_args;
T::gpu_base_impl(*a, regions, task->futures, ctx, runtime);
}
#endif
template<typename T>
void register_hybrid_variants(void)
{
{
TaskVariantRegistrar registrar(T::TASK_ID, T::TASK_NAME);
registrar.add_constraint(ProcessorConstraint(Processor::LOC_PROC));
registrar.set_leaf(T::CPU_BASE_LEAF);
Runtime::preregister_task_variant<base_cpu_wrapper<T> >(registrar, T::TASK_NAME);
}
#ifdef REALM_USE_OPENMP
{
TaskVariantRegistrar registrar(T::TASK_ID, T::TASK_NAME);
registrar.add_constraint(ProcessorConstraint(Processor::OMP_PROC));
registrar.set_leaf(T::CPU_BASE_LEAF);
Runtime::preregister_task_variant<base_cpu_wrapper<T> >(registrar, T::TASK_NAME);
}
#endif
#ifdef LEGION_USE_CUDA
{
TaskVariantRegistrar registrar(T::TASK_ID, T::TASK_NAME);
registrar.add_constraint(ProcessorConstraint(Processor::TOC_PROC));
registrar.set_leaf(T::GPU_BASE_LEAF);
Runtime::preregister_task_variant<base_gpu_wrapper<T> >(registrar, T::TASK_NAME);
}
#endif
}
};
#endif // __TASK_HELPER_H__