include/pops/coupling/schur/amr/amr_condensed_schur_source_stepper.hpp Source File

adc_cpp 0.3.0 Model-free C++23 core for coupled hyperbolic-elliptic systems on adaptive (AMR) meshes, with MPI and GPU (Kokkos) backends

• include
• pops
• coupling
• schur
• amr

amr_condensed_schur_source_stepper.hpp

Go to the documentation of this file.

1#pragma once

2

3#include <pops/coupling/schur/source/condensed_schur_source_stepper.hpp> // CondensedSchurSourceStepper (#126) + detail kernels

4#include <pops/amr/hierarchy/refinement_ratio.hpp>

5#include <pops/coupling/schur/core/schur_condensation.hpp> // ElectrostaticLorentzCondensation (assemble per level)

6#include <pops/numerics/elliptic/mg/composite_fac_poisson.hpp> // CompositeFacPoisson (composite FAC elliptic solve)

7#include <pops/numerics/time/amr/reflux/amr_reflux_mf.hpp> // mf_average_down_mb (fine -> coarse cascade)

8#include <pops/numerics/time/amr/levels/amr_subcycling.hpp> // AmrLevelMP (multi-patch hierarchy)

9

10#include <memory>

11#include <stdexcept>

12#include <string>

13#include <vector>

14

49

50namespace pops {

51

54class AmrCondensedSchurSourceStepper {

55 public:

63 AmrCondensedSchurSourceStepper(const VariableSet& vars, const Geometry& coarse_geom,

64 const BoxArray& coarse_ba, const BCRec& bcPhi, Real alpha,

65 int n_precond_vcycles = 1)

66 : AmrCondensedSchurSourceStepper(

67 vars, vars.index_of(VariableRole::Density), vars.index_of(VariableRole::MomentumX),

68 vars.index_of(VariableRole::MomentumY), vars.index_of(VariableRole::Energy),

69 coarse_geom, coarse_ba, bcPhi, alpha, n_precond_vcycles) {}

70

73 AmrCondensedSchurSourceStepper(const VariableSet& vars, int c_rho, int c_mx, int c_my, int c_E,

74 const Geometry& coarse_geom, const BoxArray& coarse_ba,

75 const BCRec& bcPhi, Real alpha, int n_precond_vcycles = 1)

76 : vars_(vars),

77 coarse_geom_(coarse_geom),

78 coarse_ba_(coarse_ba),

79 bcPhi_(bcPhi),

80 alpha_(alpha),

81 c_rho_(c_rho),

82 c_mx_(c_mx),

83 c_my_(c_my),

84 c_E_(c_E),

85 coarse_(vars, c_rho, c_mx, c_my, c_E, coarse_geom, coarse_ba, bcPhi, alpha,

86 n_precond_vcycles) {}

87

91 void set_krylov(Real tol, int max_iters) { coarse_.set_krylov(tol, max_iters); }

92

94 bool has_energy() const { return coarse_.energy_comp() >= 0; }

95

105 void step(std::vector<AmrLevelMP>& levels, MultiFab& coarse_phi, const MultiFab& coarse_bz,

106 int c_bz, Real theta, Real dt) {

107 if (levels.empty())

108 return;

109 // A fine level EFFECTIVELY POPULATED (>= one patch) signals a multi-level hierarchy. NB: the

110 // compiled path (build_amr_compiled) ALWAYS allocates a seed fine level, EMPTY after regrid when

111 // no refinement is requested (refine_threshold disabled) -> levels.size() is 2 but the

112 // hierarchy is EFFECTIVELY mono-level. So we gate on the NUMBER OF fine PATCHES, not on

113 // levels.size(), to avoid refusing the mono-level case with an allocated but empty fine level.

114 int n_fine_patches = 0;

115 for (std::size_t k = 1; k < levels.size(); ++k)

116 n_fine_patches += static_cast<int>(levels[k].U.box_array().size());

117 if (n_fine_patches == 0) {

118 // MONO-LEVEL (no fine patch): COMPLETE uniform stage on the coarse level (assemble + solve +

119 // reconstruction + extrapolation + energy + ghosts), bit-for-bit identical to #126.

120 coarse_.step(levels[0].U, coarse_phi, coarse_bz, c_bz, theta, dt);

121 return;

122 }

123 // MULTI-LEVEL (Phase 4a): COMPOSITE condensed source stage -- the fine patches REALLY refine

124 // the elliptic (tensor Schur operator solved by FAC on coarse + fine), then velocity

125 // reconstruction PER LEVEL and average_down cascade. Phase 4a frame: 2 levels, 1..N fine patches

126 // disjoint NON ADJACENT (separated by at least one coarse cell -- guard imposed by the FAC),

127 // coarse replicated mono-block, MONO-RANK. Beyond that -> clear error (> 2 levels / MPI / multi-block

128 // = Phase 4b). The fine-fine join between adjacent patches is rejected at the FAC ctor (Phase 4b).

129 if (levels.size() != 2 || n_ranks() != 1)

130 throw std::runtime_error(

131 "AmrCondensedSchurSourceStepper: COMPOSITE condensed source stage wired for 2 levels + "

132 "NON ADJACENT multi-box fine patches, mono-rank ; > 2 levels / MPI / multi-block = Phase "

133 "4b.");

134 step_multilevel(levels, coarse_phi, coarse_bz, c_bz, theta, dt);

135 }

136

138 const KrylovResult& last_solve() const { return coarse_.last_solve(); }

139

140 int density_comp() const { return coarse_.density_comp(); }

141 int momentum_x_comp() const { return coarse_.momentum_x_comp(); }

142 int momentum_y_comp() const { return coarse_.momentum_y_comp(); }

143 int energy_comp() const { return coarse_.energy_comp(); }

144

145 private:

151 void step_multilevel(std::vector<AmrLevelMP>& levels, MultiFab& coarse_phi,

152 const MultiFab& coarse_bz, int c_bz, Real theta, Real dt) {

153 // COMPLETE fine BoxArray (1..N patches): the FAC is built on this tiling; the patches being separated

154 // by at least one coarse cell (FAC ctor guard), each edge is a true C-F join.

155 const BoxArray& fine_ba = levels[1].U.box_array();

156 ensure_fac(fine_ba);

157 const Geometry geom_c = coarse_geom_;

158 const Geometry geom_f = coarse_geom_.refine(kAmrRefRatio);

159 ElectrostaticLorentzCondensation builder(vars_, alpha_, theta, dt);

160

161 MultiFab& Uc = levels[0].U;

162 MultiFab& Uf = levels[1].U;

163 const BoxArray bac = Uc.box_array();

164 const DistributionMapping dmc = Uc.dmap();

165 const BoxArray baf = Uf.box_array();

166 const DistributionMapping dmf = Uf.dmap();

167

168 // --- B_z 1-component per level (coarse: extracted from coarse_bz; fine: bilerp of the coarse) ---

169 MultiFab bz_c(bac, dmc, 1, 1), bz_f(baf, dmf, 1, 1);

170 copy_comp(bz_c, coarse_bz, c_bz);

171 device_fence();

172 fill_ghosts(bz_c, geom_c.domain, coeff_bc(bcPhi_));

173 bilerp_coarse_to_fine(bz_f, bz_c); // fine B_z from the coarse (B0 uniform -> exact)

174

175 // --- phi^n per level (coarse = coarse_phi; fine = injected aux, levels[1].aux comp 0) ---

176 MultiFab phi_n_c(bac, dmc, 1, 1), phi_n_f(baf, dmf, 1, 1);

177 copy0(phi_n_c, coarse_phi);

178 copy0(phi_n_f, *levels[1].aux);

179

180 // --- v^n per level (before the solve: the reconstruction overwrites mom) ---

181 MultiFab vx_n_c(bac, dmc, 1, 0), vy_n_c(bac, dmc, 1, 0);

182 MultiFab vx_n_f(baf, dmf, 1, 0), vy_n_f(baf, dmf, 1, 0);

183 extract_v(Uc, vx_n_c, vy_n_c);

184 extract_v(Uf, vx_n_f, vy_n_f);

185

186 // --- operator + condensed RHS assembly PER LEVEL, into the composite solver fields ---

187 // eps_x == eps_y for the Schur (A_xx = A_yy = 1 + c rho/det): we write eps_x into the single eps of the

188 // composite and eps_y into a discarded scratch. f_composite = -rhs_schur (sign convention #126).

189 MultiFab eps_y_c(bac, dmc, 1, 1), eps_y_f(baf, dmf, 1, 1);

190 MultiFab rhs_c(bac, dmc, 1, 0), rhs_f(baf, dmf, 1, 0);

191 builder.assemble_operator(Uc, bz_c, geom_c, bcPhi_, fac_->eps_coarse(), eps_y_c,

192 fac_->a_xy_coarse(), fac_->a_yx_coarse());

193 builder.assemble_operator(Uf, bz_f, geom_f, bcPhi_, fac_->eps_fine(), eps_y_f,

194 fac_->a_xy_fine(), fac_->a_yx_fine());

195 {

196 MultiFab pn(bac, dmc, 1, 1);

197 copy0(pn, phi_n_c);

198 builder.assemble_rhs(pn, Uc, bz_c, geom_c, bcPhi_, rhs_c);

199 negate_into(fac_->rhs_coarse(), rhs_c);

200 }

201 {

202 MultiFab pn(baf, dmf, 1, 1);

203 copy0(pn, phi_n_f);

204 builder.assemble_rhs(pn, Uf, bz_f, geom_f, bcPhi_, rhs_f);

205 negate_into(fac_->rhs_fine(), rhs_f);

206 }

207

208 // --- COMPOSITE SOLVE: phi^{n+theta} per level (the fine patch refines the elliptic) ---

209 fac_->use_variable_coefficient(true);

210 fac_->use_cross_terms(true);

211 fac_->solve();

212

213 // --- velocity reconstruction + phi/v extrapolation + energy, PER LEVEL ---

214 reconstruct_level(Uc, fac_->phi_coarse(), phi_n_c, bz_c, vx_n_c, vy_n_c, geom_c, theta, dt,

215 /*fill_phi_ghosts=*/true);

216 reconstruct_level(Uf, fac_->phi_fine(), phi_n_f, bz_f, vx_n_f, vy_n_f, geom_f, theta, dt,

217 /*fill_phi_ghosts=*/false); // C-F ghosts already set by the composite solve

218

219 // coarse phi^{n+1} (extrapolated in place into fac_->phi_coarse()) -> published into coarse_phi.

220 copy0(coarse_phi, fac_->phi_coarse());

221

222 // --- fine -> coarse cascade: the COVERED coarse cells = 2x2 average of the fine cells (#169) ---

223 device_fence();

224 mf_average_down_mb(Uf, Uc);

225 device_fence();

226 fill_ghosts(coarse_phi, geom_c.domain, bcPhi_);

227 }

228

233 void reconstruct_level(MultiFab& state, MultiFab& phi_nt, const MultiFab& phi_n,

234 const MultiFab& bz, const MultiFab& vx_n, const MultiFab& vy_n,

235 const Geometry& geom, Real theta, Real dt, bool fill_phi_ghosts) {

236 const Real th_dt = theta * dt, inv_theta = Real(1) / theta;

237 const Real half_idx = Real(1) / (Real(2) * geom.dx());

238 const Real half_idy = Real(1) / (Real(2) * geom.dy());

239 device_fence();

240 if (fill_phi_ghosts)

241 fill_ghosts(phi_nt, geom.domain, bcPhi_);

242 device_fence();

243 MultiFab vx_t(state.box_array(), state.dmap(), 1, 0),

244 vy_t(state.box_array(), state.dmap(), 1, 0);

245 for (int li = 0; li < state.local_size(); ++li)

246 for_each_cell(

247 state.box(li),

248 detail::SchurReconstructKernel{

249 phi_nt.fab(li).const_array(), vx_n.fab(li).const_array(), vy_n.fab(li).const_array(),

250 bz.fab(li).const_array(), state.fab(li).array(), vx_t.fab(li).array(),

251 vy_t.fab(li).array(), th_dt, half_idx, half_idy, c_rho_, c_mx_, c_my_});

252 for (int li = 0; li < phi_nt.local_size(); ++li)

253 for_each_cell(phi_nt.box(li),

254 detail::SchurExtrapolateScalarKernel{phi_n.fab(li).const_array(),

255 phi_nt.fab(li).array(), inv_theta});

256 for (int li = 0; li < state.local_size(); ++li)

257 for_each_cell(state.box(li), detail::SchurExtrapolateVelocityKernel{

258 vx_n.fab(li).const_array(), vy_n.fab(li).const_array(),

259 vx_t.fab(li).array(), vy_t.fab(li).array(),

260 state.fab(li).array(), inv_theta, c_rho_, c_mx_, c_my_});

261 if (c_E_ >= 0)

262 for (int li = 0; li < state.local_size(); ++li)

263 for_each_cell(state.box(li), detail::SchurEnergyKernel{

264 vx_n.fab(li).const_array(), vy_n.fab(li).const_array(),

265 vx_t.fab(li).const_array(), vy_t.fab(li).const_array(),

266 state.fab(li).array(), c_rho_, c_E_});

267 device_fence();

268 fill_ghosts(state, geom.domain, coeff_bc(bcPhi_));

269 }

270

274 void ensure_fac(const BoxArray& fine_ba) {

275 if (fac_ && fac_fine_boxes_ == fine_ba.boxes())

276 return;

277 fac_ = std::make_unique<CompositeFacPoisson>(coarse_geom_, coarse_ba_, bcPhi_, fine_ba,

278 kAmrRefRatio);

279 fac_fine_boxes_ = fine_ba.boxes();

280 }

281

283 static BCRec coeff_bc(const BCRec& b) {

284 auto fo = [](BCType t) { return t == BCType::Periodic ? t : BCType::Foextrap; };

285 BCRec c;

286 c.xlo = fo(b.xlo);

287 c.xhi = fo(b.xhi);

288 c.ylo = fo(b.ylo);

289 c.yhi = fo(b.yhi);

290 return c;

291 }

292

293 void copy0(MultiFab& dst, const MultiFab& src) {

294 device_fence();

295 for (int li = 0; li < dst.local_size(); ++li)

296 for_each_cell(dst.box(li),

297 detail::CopyComp0Kernel{dst.fab(li).array(), src.fab(li).const_array()});

298 }

299 void copy_comp(MultiFab& dst, const MultiFab& src, int c) { // dst comp0 <- src comp c

300 device_fence();

301 for (int li = 0; li < dst.local_size(); ++li)

302 for_each_cell(dst.box(li),

303 detail::CopyBzKernel{src.fab(li).const_array(), dst.fab(li).array(), c});

304 }

305 void negate_into(MultiFab& dst, const MultiFab& src) {

306 device_fence();

307 for (int li = 0; li < dst.local_size(); ++li)

308 for_each_cell(dst.box(li),

309 detail::NegateKernel{src.fab(li).const_array(), dst.fab(li).array()});

310 }

311 void extract_v(const MultiFab& state, MultiFab& vx, MultiFab& vy) {

312 device_fence();

313 for (int li = 0; li < state.local_size(); ++li)

314 for_each_cell(state.box(li),

315 detail::ExtractVelocityKernel{state.fab(li).const_array(), vx.fab(li).array(),

316 vy.fab(li).array(), c_rho_, c_mx_, c_my_});

317 }

320 void bilerp_coarse_to_fine(MultiFab& fine, const MultiFab& coarse) {

321 device_fence();

322 const ConstArray4 C = coarse.fab(0).const_array();

323 const int ng = fine.n_grow();

324 for (int li = 0; li < fine.local_size(); ++li) {

325 Array4 F = fine.fab(li).array();

326 const Box2D vb = fine.box(li);

327 for (int j = vb.lo[1] - ng; j <= vb.hi[1] + ng; ++j)

328 for (int i = vb.lo[0] - ng; i <= vb.hi[0] + ng; ++i)

329 F(i, j, 0) = detail::fac_bilerp_coarse(C, i, j, kAmrRefRatio);

330 }

331 }

332

333 VariableSet vars_;

334 Geometry coarse_geom_;

335 BoxArray coarse_ba_;

336 BCRec bcPhi_;

337 Real alpha_;

338 int c_rho_, c_mx_, c_my_, c_E_;

340 CondensedSchurSourceStepper coarse_;

342 std::unique_ptr<CompositeFacPoisson> fac_;

344 std::vector<Box2D> fac_fine_boxes_;

345};

346

347} // namespace pops

amr_reflux_mf.hpp

Umbrella for the AMR MultiFab stack: includes the numerics/time sub-headers in dependency order (flux...

amr_subcycling.hpp

AMR multi-patch subcycling engine (several fine boxes per level): 2-level step (amr_step_2level_multi...

pops::AmrCondensedSchurSourceStepper

Schur-condensed SOURCE stage over an AMR hierarchy.

Definition amr_condensed_schur_source_stepper.hpp:54

pops::AmrCondensedSchurSourceStepper::step

void step(std::vector< AmrLevelMP > &levels, MultiFab &coarse_phi, const MultiFab &coarse_bz, int c_bz, Real theta, Real dt)

Condensed SOURCE stage, IN-PLACE on the hierarchy levels and the coarse potential coarse_phi.

Definition amr_condensed_schur_source_stepper.hpp:105

pops::AmrCondensedSchurSourceStepper::momentum_y_comp

int momentum_y_comp() const

Definition amr_condensed_schur_source_stepper.hpp:142

pops::AmrCondensedSchurSourceStepper::momentum_x_comp

int momentum_x_comp() const

Definition amr_condensed_schur_source_stepper.hpp:141

pops::AmrCondensedSchurSourceStepper::AmrCondensedSchurSourceStepper

AmrCondensedSchurSourceStepper(const VariableSet &vars, int c_rho, int c_mx, int c_my, int c_E, const Geometry &coarse_geom, const BoxArray &coarse_ba, const BCRec &bcPhi, Real alpha, int n_precond_vcycles=1)

EXPLICIT-COMPONENT variant (audit wave 3, parity with the System steppers): roles carried by the ABI ...

Definition amr_condensed_schur_source_stepper.hpp:73

pops::AmrCondensedSchurSourceStepper::density_comp

int density_comp() const

Definition amr_condensed_schur_source_stepper.hpp:140

pops::AmrCondensedSchurSourceStepper::last_solve

const KrylovResult & last_solve() const

Diagnostic of the last coarse stage solve (BiCGStab iterations, relative residual,...

Definition amr_condensed_schur_source_stepper.hpp:138

pops::AmrCondensedSchurSourceStepper::AmrCondensedSchurSourceStepper

AmrCondensedSchurSourceStepper(const VariableSet &vars, const Geometry &coarse_geom, const BoxArray &coarse_ba, const BCRec &bcPhi, Real alpha, int n_precond_vcycles=1)

vars: descriptor of the fluid block (MUST expose Density / MomentumX / MomentumY; Energy optional).

Definition amr_condensed_schur_source_stepper.hpp:63

pops::AmrCondensedSchurSourceStepper::set_krylov

void set_krylov(Real tol, int max_iters)

Tolerance / budget of the COARSE stage Krylov solve (delegated to the uniform stage #126; historical ...

Definition amr_condensed_schur_source_stepper.hpp:91

pops::AmrCondensedSchurSourceStepper::energy_comp

int energy_comp() const

Definition amr_condensed_schur_source_stepper.hpp:143

pops::AmrCondensedSchurSourceStepper::has_energy

bool has_energy() const

true if the model carries an Energy role (energy update active in the coarse stage).

Definition amr_condensed_schur_source_stepper.hpp:94

pops::BoxArray

Ordered list of boxes tiling a level.

Definition box_array.hpp:22

pops::CondensedSchurSourceStepper::set_krylov

void set_krylov(Real tol, int max_iters)

Tolerance / iteration budget of the stage Krylov solve (BiCGStab).

Definition condensed_schur_source_stepper.hpp:277

pops::CondensedSchurSourceStepper::density_comp

int density_comp() const

Definition condensed_schur_source_stepper.hpp:283

pops::CondensedSchurSourceStepper::momentum_x_comp

int momentum_x_comp() const

Definition condensed_schur_source_stepper.hpp:284

pops::CondensedSchurSourceStepper::energy_comp

int energy_comp() const

Definition condensed_schur_source_stepper.hpp:286

pops::CondensedSchurSourceStepper::last_solve

const KrylovResult & last_solve() const

Diagnostic of the last solve (BiCGStab iterations, relative residual, convergence).

Definition condensed_schur_source_stepper.hpp:272

pops::CondensedSchurSourceStepper::step

void step(MultiFab &state, MultiFab &phi, const MultiFab &bz_field, int c_bz, Real theta, Real dt)

Condensed SOURCE STAGE, IN-PLACE on state and phi.

Definition condensed_schur_source_stepper.hpp:190

pops::CondensedSchurSourceStepper::momentum_y_comp

int momentum_y_comp() const

Definition condensed_schur_source_stepper.hpp:285

pops::DistributionMapping

Owning MPI rank of each box, indexed by GLOBAL box index (parallel to a BoxArray).

Definition distribution_mapping.hpp:19

pops::ElectrostaticLorentzCondensation

GENERIC builder of the condensed source stage for the electrostatic + Lorentz source (kind="electrost...

Definition schur_condensation.hpp:153

pops::MultiFab

Field distributed over a level: decomposition (BoxArray) + distribution (DistributionMapping) + ncomp...

Definition multifab.hpp:33

pops::MultiFab::dmap

const DistributionMapping & dmap() const

GLOBAL distribution (owner rank per box).

Definition multifab.hpp:58

pops::MultiFab::box_array

const BoxArray & box_array() const

GLOBAL decomposition of the level (all boxes, all ranks).

Definition multifab.hpp:56

composite_fac_poisson.hpp

CompositeFacPoisson: 2-level AMR COMPOSITE elliptic solver (Fast Adaptive Composite,...

condensed_schur_source_stepper.hpp

CondensedSchurSourceStepper: Schur-condensed SOURCE STAGE (level 4 of docs/SCHUR_CONDENSATION_DESIGN....

pops::detail::fac_bilerp_coarse

POPS_HD Real fac_bilerp_coarse(const ConstArray4 &C, int i, int j, int r)

BILINEAR interpolation of the coarse potential (cell-centered, C with ghosts) at the CENTER of the fi...

Definition composite_fac_poisson.hpp:70

pops

Definition amr_hierarchy.hpp:29

pops::for_each_cell

void for_each_cell(const Box2D &b, F f)

Applies f to EACH cell (i, j) of box b (bounds inclusive), via Kokkos::parallel_for (Serial / OpenMP ...

Definition for_each.hpp:138

pops::Real

double Real

Definition types.hpp:30

pops::mf_average_down_mb

void mf_average_down_mb(const MultiFab &Uf, MultiFab &Uc)

Definition amr_subcycling.hpp:305

pops::n_ranks

int n_ranks()

Definition comm.hpp:139

pops::device_fence

void device_fence()

Device barrier: waits for in-flight kernels to finish before a HOST access to unified memory.

Definition kokkos_env.hpp:43

pops::kAmrRefRatio

constexpr int kAmrRefRatio

The native AMR refinement ratio between two consecutive levels.

Definition refinement_ratio.hpp:30

pops::VariableRole

VariableRole

PHYSICAL role of a component.

Definition variables.hpp:27

pops::VariableRole::Energy

@ Energy

pops::VariableRole::Density

@ Density

pops::VariableRole::MomentumX

@ MomentumX

pops::VariableRole::MomentumY

@ MomentumY

pops::BCType

BCType

Boundary condition type for a face: Periodic (handled by fill_boundary), Foextrap (zero gradient,...

Definition physical_bc.hpp:25

pops::BCType::Foextrap

@ Foextrap

pops::BCType::Periodic

@ Periodic

pops::fill_ghosts

void fill_ghosts(MultiFab &mf, const Box2D &domain, const BCRec &bc)

COMPLETE ghost filling: fill_boundary (interior + periodic, periodicity deduced from bc) THEN fill_ph...

Definition physical_bc.hpp:227

refinement_ratio.hpp

Single source of truth for the native AMR refinement ratio.

schur_condensation.hpp

BUILDER (NOT solver) of the Schur-condensed source stage of the implicit source coupling potential / ...

pops::BCRec

Boundary conditions for the FOUR faces of the domain (type + associated Dirichlet value).

Definition physical_bc.hpp:29

pops::Geometry

Cartesian geometry of a level: index domain + physical bounds [xlo, xhi] x [ylo, yhi].

Definition geometry.hpp:20

pops::Geometry::refine

Geometry refine(int r) const

Geometry refined by ratio r: SAME physical extent, refined index domain (dx -> dx/r).

Definition geometry.hpp:40

pops::Geometry::domain

Box2D domain

Definition geometry.hpp:21

pops::KrylovResult

Outcome of a Krylov solve: iterations performed, final relative residual, convergence flag.

Definition krylov_result.hpp:16

pops::VariableSet

A model's variable set: kind (cons/prim), names, size, canonical roles (optional, parallel to names; ...

Definition variables.hpp:58

---

Generated by  1.9.8