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 | // *****************************************************************************
/*!
\file src/LinearSolver/ConjugateGradients.hpp
\copyright 2012-2015 J. Bakosi,
2016-2018 Los Alamos National Security, LLC.,
2019-2021 Triad National Security, LLC.,
2022-2024 J. Bakosi
All rights reserved. See the LICENSE file for details.
\brief Charm++ chare array for distributed conjugate gradients
\details Charm++ chare array for asynchronous distributed
conjugate gradients linear solver.
There are a potentially large number of ConjugateGradients Charm++ chares.
Each ConjugateGradient chare gets a chunk of the full load, due to partiting
the mesh, on which the solve is performed.
The implementation uses the Charm++ runtime system and is fully
asynchronous, overlapping computation and communication. The algorithm
utilizes the structured dagger (SDAG) Charm++ functionality.
*/
// *****************************************************************************
#pragma once
#include "Types.hpp"
#include "CSR.hpp"
#include "NoWarning/conjugategradients.decl.h"
namespace tk {
//! \brief ConjugateGradients Charm++ chare array used to perform a distributed
//! linear solve with the conjugate gradients algorithm
class ConjugateGradients : public CBase_ConjugateGradients {
public:
#if defined(__clang__)
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wunused-parameter"
#elif defined(STRICT_GNUC)
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wunused-parameter"
#endif
// Include Charm++ SDAG code. See http://charm.cs.illinois.edu/manuals/html/
// charm++/manual.html, Sec. "Structured Control Flow: Structured Dagger".
ConjugateGradients_SDAG_CODE
#if defined(__clang__)
#pragma clang diagnostic pop
#elif defined(STRICT_GNUC)
#pragma GCC diagnostic pop
#endif
//! Constructor
explicit ConjugateGradients(
const CSR& A,
const std::vector< tk::real >& x,
const std::vector< tk::real >& b,
const std::vector< std::size_t >& gid = {},
const std::unordered_map< std::size_t, std::size_t >& lid = {},
const std::unordered_map< int,
std::unordered_set< std::size_t > >& nodecommmap = {} );
#if defined(__clang__)
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wundefined-func-template"
#endif
//! Constructor taking a tuple of {A,x,b} by rvalue reference
explicit ConjugateGradients(<--- Member variable 'ConjugateGradients::m_tol' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_normr' is not initialized in the constructor.
std::tuple< tk::CSR,
std::vector< tk::real >,
std::vector< tk::real > >&& system,
const std::vector< std::size_t >& gid,
const std::unordered_map< std::size_t, std::size_t >& lid,
const std::unordered_map< int,
std::unordered_set< std::size_t > >& nodecommmap ) :
ConjugateGradients( std::move(std::get<0>(system)),
std::move(std::get<1>(system)),
std::move(std::get<2>(system)),
gid, lid, nodecommmap ) {}
#if defined(__clang__)
#pragma clang diagnostic pop
#endif
//! Migrate constructor
explicit ConjugateGradients( CkMigrateMessage* m )<--- Member variable 'ConjugateGradients::m_nr' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_na' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_nb' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_nq' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_nd' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_normb' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_it' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_maxit' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_finished' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_verbose' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_tol' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_rho' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_rho0' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_alpha' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_converged' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_nx' is not initialized in the constructor.<--- Member variable 'ConjugateGradients::m_normr' is not initialized in the constructor.
: CBase_ConjugateGradients( m ) {}
//! Solve linear system
void solve( std::size_t maxit,
tk::real tol,
int pe,
uint64_t verbose,
CkCallback c );
//! Initialize linear solve: set initial guess and boundary conditions
void init( const std::vector< tk::real >& x,
const std::vector< tk::real >& b,
const std::vector< tk::real >& neubc,
const std::unordered_map< std::size_t,
std::vector< std::pair< int, tk::real > > >& dirbc,
const std::string& pc,
CkCallback cb );
//! Setup solver
void setup( CkCallback c );
//! Compute the norm of the right hand side
void normb( tk::real n );
//! Compute rho = (r,z)
void rho( tk::real r );
//! Receive contributions to r = b - A * x on chare-boundaries
void comres( const std::vector< std::size_t >& gid,
const std::vector< std::vector< tk::real > >& rc );
//! Receive contributions to boundary conditions and rhs on chare-boundaries
void combc( const std::map< std::size_t,
std::vector< std::pair< int, tk::real > > >& dbc,
const std::vector< std::size_t >& gid,
const std::vector< std::vector< tk::real > >& qc );
//! \brief Receive contributions to rhs with Dirichlet BCs applied on
//! chare-boundaries
void comr( const std::vector< std::size_t >& gid,
const std::vector< std::vector< tk::real > >& rc );
//! Receive contributions to preconditioner chare-boundaries
void comd( const std::vector< std::size_t >& gid,
const std::vector< std::vector< tk::real > >& qc );
//! Receive contributions to q = A * p on chare-boundaries
void comq( const std::vector< std::size_t >& gid,
const std::vector< std::vector< tk::real > >& qc );
//! Receive contributions to final solution on chare-boundaries
void comx( const std::vector< std::size_t >& gid,
const std::vector< std::vector< tk::real > >& xc );
//! Compute the dot product (p,q)
void pq( tk::real d );
//! Compute the norm of the residual (r,r)
void normres( tk::real r );
//! Compute the dot product (r,z)
void rz( tk::real rz );
//! Access solution
const std::vector< tk::real >& solution() const { return m_x; }
//! Return convergence flag
bool converged() const { return m_converged; }
//! Return number of iterations taken
std::size_t it() const { return m_it; }
//! Non-const-ref access to lhs matrix
tk::CSR& lhs() { return m_A; }
/** @name Pack/unpack (Charm++ serialization) routines */
///@{
//! \brief Pack/Unpack serialize member function
//! \param[in,out] p Charm++'s PUP::er serializer object reference
void pup( PUP::er &p ) override {
p | m_A;
p | m_An;
p | m_x;
p | m_b;
p | m_pc;
p | m_gid;
p | m_lid;
p | m_nodeCommMap;
p | m_r;
p | m_z;
p | m_d;
p | m_rc;
p | m_nr;
p | m_na;
p | m_dirbc;
p | m_dirbcc;
p | m_nb;
p | m_p;
p | m_q;
p | m_qc;
p | m_nq;
p | m_nd;
p | m_initres;
p | m_solved;
p | m_normb;
p | m_it;
p | m_maxit;
p | m_finished;
p | m_verbose;
p | m_tol;
p | m_rho;
p | m_rho0;
p | m_alpha;
p | m_converged;
p | m_xc;
p | m_nx;
p | m_normr;
}
//! \brief Pack/Unpack serialize operator|
//! \param[in,out] p Charm++'s PUP::er serializer object reference
//! \param[in,out] c ConjugateGradients object reference
friend void operator|( PUP::er& p, ConjugateGradients& c ) { c.pup(p); }
///@}
private:
//! Sparse matrix
CSR m_A;
//! Sparse matrix before boundary conditions
CSR m_An;
//! Solution/unknown
std::vector< tk::real > m_x;
//! Right hand side
std::vector< tk::real > m_b;
//! Preconditioner to use
std::string m_pc;
//! Global node IDs
std::vector< std::size_t > m_gid;
//! Local node IDs associated to global ones
std::unordered_map< std::size_t, std::size_t > m_lid;
//! Global mesh node IDs shared with other chares associated to chare IDs
std::unordered_map< int, std::unordered_set< std::size_t > > m_nodeCommMap;
//! Auxiliary vector for CG solve
std::vector< tk::real > m_r;
//! Receive buffer for communication of r = b - A * x
std::unordered_map< std::size_t, std::vector< tk::real > > m_rc;
//! Auxiliary vector for preconditioned CG solve
std::vector< tk::real > m_z;
//! Jacobi preconditioner
std::vector< tk::real > m_d;
//! Counter for assembling m_r
std::size_t m_nr;
//! Counter for assembling m_r (rhs with BCs applied)
std::size_t m_na;
//! Dirichlet boundary conditions
std::map< std::size_t, std::vector< std::pair<int,tk::real> > > m_dirbc;
//! Dirichlet boundary conditions communication buffer
std::map< std::size_t, std::vector< std::pair<int,tk::real> > > m_dirbcc;
//! Counter for assembling boundary conditions
std::size_t m_nb;
//! Auxiliary vector for CG solve
std::vector< tk::real > m_p;
//! Auxiliary vector for CG solve
std::vector< tk::real > m_q;
//! Receive buffer for communication of q = A * p
std::unordered_map< std::size_t, std::vector< tk::real > > m_qc;
//! Counter for assembling m_q
std::size_t m_nq;
//! Counter for assembling the preconditioner
std::size_t m_nd;
//! Charm++ callback to continue with when the setup is complete
CkCallback m_initres;
//! Charm++ callback to continue with when the solve is complete
CkCallback m_solved;
//! L2 norm of the right hand side
tk::real m_normb;
//! Iteration count
std::size_t m_it;
//! Max iteration count
std::size_t m_maxit;
//! True if finished
bool m_finished;
//! Verbose output
uint64_t m_verbose;
//! Stop tolerance
tk::real m_tol;
//! Helper scalar for CG algorithm
tk::real m_rho;
//! Helper scalar for CG algorithm
tk::real m_rho0;
//! Helper scalar for CG algorithm
tk::real m_alpha;
//! Convergence flag: true if linear smoother converged to tolerance
bool m_converged;
//! Receive buffer for solution
std::unordered_map< std::size_t, std::vector< tk::real > > m_xc;
//! Counter for assembling the solution on chare boundaries
std::size_t m_nx;
//! Norm of the residual
tk::real m_normr;
//! Initiate computationa of dot product of two vectors
void dot( const std::vector< tk::real >& a,
const std::vector< tk::real >& b,
CkCallback c );
//! Initiate A * x for computing the residual, r = b - A * x
void residual();
//! Finish computing the initial residual, r = b - A * x
void initres();
//! Setup preconditioner
void pc();
//! Apply boundary conditions
void apply( CkCallback cb );
//! Finish computing rhs with applied BCs
void r( CkCallback cb );
//! Initiate computing q = A * p
void qAp();
//! Finish computing q = A * p
void q();
//! Start next linear solver iteration
void next();
//! Assemble solution on chare boundaries and decide what's next
void x();
};
} // tk::
|