1 /* 2 Implements the Kokkos kernel 3 */ 4 #include <petscconf.h> 5 #include <petscvec_kokkos.hpp> 6 #include <petsc/private/dmpleximpl.h> /*I "petscdmplex.h" I*/ 7 #include <petsclandau.h> 8 #include <petscts.h> 9 10 #include <Kokkos_Core.hpp> 11 #include <cstdio> 12 typedef Kokkos::TeamPolicy<>::member_type team_member; 13 #include "../land_tensors.h" 14 #include <petscaijdevice.h> 15 16 namespace landau_inner_red { // namespace helps with name resolution in reduction identity 17 template< class ScalarType > 18 struct array_type { 19 ScalarType gg2[LANDAU_DIM]; 20 ScalarType gg3[LANDAU_DIM][LANDAU_DIM]; 21 22 KOKKOS_INLINE_FUNCTION // Default constructor - Initialize to 0's 23 array_type() { 24 for (int j = 0; j < LANDAU_DIM; j++) { 25 gg2[j] = 0; 26 for (int k = 0; k < LANDAU_DIM; k++) { 27 gg3[j][k] = 0; 28 } 29 } 30 } 31 KOKKOS_INLINE_FUNCTION // Copy Constructor 32 array_type(const array_type & rhs) { 33 for (int j = 0; j < LANDAU_DIM; j++) { 34 gg2[j] = rhs.gg2[j]; 35 for (int k = 0; k < LANDAU_DIM; k++) { 36 gg3[j][k] = rhs.gg3[j][k]; 37 } 38 } 39 } 40 KOKKOS_INLINE_FUNCTION // add operator 41 array_type& operator += (const array_type& src) 42 { 43 for (int j = 0; j < LANDAU_DIM; j++) { 44 gg2[j] += src.gg2[j]; 45 for (int k = 0; k < LANDAU_DIM; k++) { 46 gg3[j][k] += src.gg3[j][k]; 47 } 48 } 49 return *this; 50 } 51 KOKKOS_INLINE_FUNCTION // volatile add operator 52 void operator += (const volatile array_type& src) volatile 53 { 54 for (int j = 0; j < LANDAU_DIM; j++) { 55 gg2[j] += src.gg2[j]; 56 for (int k = 0; k < LANDAU_DIM; k++) { 57 gg3[j][k] += src.gg3[j][k]; 58 } 59 } 60 } 61 }; 62 typedef array_type<PetscReal> TensorValueType; // used to simplify code below 63 } 64 65 namespace Kokkos { //reduction identity must be defined in Kokkos namespace 66 template<> 67 struct reduction_identity< landau_inner_red::TensorValueType > { 68 KOKKOS_FORCEINLINE_FUNCTION static landau_inner_red::TensorValueType sum() { 69 return landau_inner_red::TensorValueType(); 70 } 71 }; 72 } 73 74 extern "C" { 75 PetscErrorCode LandauKokkosCreateMatMaps(P4estVertexMaps maps[], pointInterpolationP4est (*pointMaps)[LANDAU_MAX_Q_FACE], PetscInt Nf[], PetscInt Nq, PetscInt grid) 76 { 77 P4estVertexMaps h_maps; /* host container */ 78 const Kokkos::View<pointInterpolationP4est*[LANDAU_MAX_Q_FACE], Kokkos::LayoutRight, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_points ((pointInterpolationP4est*)pointMaps, maps[grid].num_reduced); 79 const Kokkos::View< LandauIdx*[LANDAU_MAX_SPECIES][LANDAU_MAX_NQ], Kokkos::LayoutRight, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_gidx ((LandauIdx*)maps[grid].gIdx, maps[grid].num_elements); 80 Kokkos::View<pointInterpolationP4est*[LANDAU_MAX_Q_FACE], Kokkos::LayoutRight> *d_points = new Kokkos::View<pointInterpolationP4est*[LANDAU_MAX_Q_FACE], Kokkos::LayoutRight>("points", maps[grid].num_reduced); 81 Kokkos::View<LandauIdx*[LANDAU_MAX_SPECIES][LANDAU_MAX_NQ], Kokkos::LayoutRight> *d_gidx = new Kokkos::View<LandauIdx*[LANDAU_MAX_SPECIES][LANDAU_MAX_NQ], Kokkos::LayoutRight>("gIdx", maps[grid].num_elements); 82 83 PetscFunctionBegin; 84 Kokkos::deep_copy (*d_gidx, h_gidx); 85 Kokkos::deep_copy (*d_points, h_points); 86 h_maps.num_elements = maps[grid].num_elements; 87 h_maps.num_face = maps[grid].num_face; 88 h_maps.num_reduced = maps[grid].num_reduced; 89 h_maps.deviceType = maps[grid].deviceType; 90 h_maps.numgrids = maps[grid].numgrids; 91 h_maps.Nf = Nf[grid]; 92 h_maps.Nq = Nq; 93 h_maps.c_maps = (pointInterpolationP4est (*)[LANDAU_MAX_Q_FACE]) d_points->data(); 94 maps[grid].vp1 = (void*)d_points; 95 h_maps.gIdx = (LandauIdx (*)[LANDAU_MAX_SPECIES][LANDAU_MAX_NQ]) d_gidx->data(); 96 maps[grid].vp2 = (void*)d_gidx; 97 { 98 Kokkos::View<P4estVertexMaps, Kokkos::HostSpace> h_maps_k(&h_maps); 99 Kokkos::View<P4estVertexMaps> *d_maps_k = new Kokkos::View<P4estVertexMaps>(Kokkos::create_mirror(Kokkos::DefaultExecutionSpace::memory_space(),h_maps_k)); 100 Kokkos::deep_copy (*d_maps_k, h_maps_k); 101 maps[grid].d_self = d_maps_k->data(); 102 maps[grid].vp3 = (void*)d_maps_k; 103 } 104 PetscFunctionReturn(0); 105 } 106 PetscErrorCode LandauKokkosDestroyMatMaps(P4estVertexMaps maps[], PetscInt num_grids) 107 { 108 PetscFunctionBegin; 109 for (PetscInt grid=0;grid<num_grids;grid++) { 110 auto a = static_cast<Kokkos::View<pointInterpolationP4est*[LANDAU_MAX_Q_FACE], Kokkos::LayoutRight>*>(maps[grid].vp1); 111 auto b = static_cast<Kokkos::View<LandauIdx*[LANDAU_MAX_SPECIES][LANDAU_MAX_NQ], Kokkos::LayoutRight>*>(maps[grid].vp2); 112 auto c = static_cast<Kokkos::View<P4estVertexMaps*>*>(maps[grid].vp3); 113 delete a; delete b; delete c; 114 } 115 PetscFunctionReturn(0); 116 } 117 118 PetscErrorCode LandauKokkosStaticDataSet(DM plex, const PetscInt Nq, const PetscInt num_grids, PetscInt a_numCells[], PetscInt a_species_offset[], PetscInt a_mat_offset[], 119 PetscReal a_nu_alpha[], PetscReal a_nu_beta[], PetscReal a_invMass[], PetscReal a_invJ[], 120 PetscReal a_x[], PetscReal a_y[], PetscReal a_z[], PetscReal a_w[], LandauStaticData *SData_d) 121 { 122 PetscReal *BB,*DD; 123 PetscErrorCode ierr; 124 PetscTabulation *Tf; 125 PetscInt dim; 126 PetscInt Nb=Nq,ip_offset[LANDAU_MAX_GRIDS+1],ipf_offset[LANDAU_MAX_GRIDS+1],elem_offset[LANDAU_MAX_GRIDS+1],nip,IPf_sz,Nftot; 127 PetscDS prob; 128 129 PetscFunctionBegin; 130 ierr = DMGetDimension(plex, &dim);CHKERRQ(ierr); 131 ierr = DMGetDS(plex, &prob);CHKERRQ(ierr); 132 if (LANDAU_DIM != dim) SETERRQ2(PETSC_COMM_WORLD, PETSC_ERR_PLIB, "dim %D != LANDAU_DIM %d",dim,LANDAU_DIM); 133 ierr = PetscDSGetTabulation(prob, &Tf);CHKERRQ(ierr); 134 BB = Tf[0]->T[0]; DD = Tf[0]->T[1]; 135 ip_offset[0] = ipf_offset[0] = elem_offset[0] = 0; 136 nip = 0; 137 IPf_sz = 0; 138 for (PetscInt grid=0 ; grid<num_grids ; grid++) { 139 PetscInt nfloc = a_species_offset[grid+1] - a_species_offset[grid]; 140 elem_offset[grid+1] = elem_offset[grid] + a_numCells[grid]; 141 nip += a_numCells[grid]*Nq; 142 ip_offset[grid+1] = nip; 143 IPf_sz += Nq*nfloc*a_numCells[grid]; 144 ipf_offset[grid+1] = IPf_sz; 145 } 146 Nftot = a_species_offset[num_grids]; 147 ierr = PetscKokkosInitializeCheck();CHKERRQ(ierr); 148 { 149 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_alpha (a_nu_alpha, Nftot); 150 auto alpha = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("alpha", Nftot); 151 SData_d->alpha = static_cast<void*>(alpha); 152 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_beta (a_nu_beta, Nftot); 153 auto beta = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("beta", Nftot); 154 SData_d->beta = static_cast<void*>(beta); 155 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_invMass (a_invMass,Nftot); 156 auto invMass = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("invMass", Nftot); 157 SData_d->invMass = static_cast<void*>(invMass); 158 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_BB (BB,Nq*Nb); 159 auto B = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("B", Nq*Nb); 160 SData_d->B = static_cast<void*>(B); 161 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_DD (DD,Nq*Nb*dim); 162 auto D = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("D", Nq*Nb*dim); 163 SData_d->D = static_cast<void*>(D); 164 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_invJ (a_invJ, nip*dim*dim); 165 auto invJ = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("invJ", nip*dim*dim); 166 SData_d->invJ = static_cast<void*>(invJ); 167 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_x (a_x, nip); 168 auto x = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("x", nip); 169 SData_d->x = static_cast<void*>(x); 170 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_y (a_y, nip); 171 auto y = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("y", nip); 172 SData_d->y = static_cast<void*>(y); 173 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_w (a_w, nip); 174 auto w = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("w", nip); 175 SData_d->w = static_cast<void*>(w); 176 177 Kokkos::deep_copy (*alpha, h_alpha); 178 Kokkos::deep_copy (*beta, h_beta); 179 Kokkos::deep_copy (*invMass, h_invMass); 180 Kokkos::deep_copy (*B, h_BB); 181 Kokkos::deep_copy (*D, h_DD); 182 Kokkos::deep_copy (*invJ, h_invJ); 183 Kokkos::deep_copy (*x, h_x); 184 Kokkos::deep_copy (*y, h_y); 185 Kokkos::deep_copy (*w, h_w); 186 187 if (dim==3) { 188 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_z (a_z , dim==3 ? nip : 0); 189 auto z = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("z", nip); 190 SData_d->z = static_cast<void*>(z); 191 Kokkos::deep_copy (*z, h_z); 192 } else SData_d->z = NULL; 193 194 // 195 const Kokkos::View<PetscInt*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_NCells (a_numCells, num_grids); 196 auto nc = new Kokkos::View<PetscInt*, Kokkos::LayoutLeft> ("NCells",num_grids); 197 SData_d->NCells = static_cast<void*>(nc); 198 Kokkos::deep_copy (*nc, h_NCells); 199 200 const Kokkos::View<PetscInt*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_species_offset (a_species_offset, num_grids+1); 201 auto soff = new Kokkos::View<PetscInt*, Kokkos::LayoutLeft> ("species_offset",num_grids+1); 202 SData_d->species_offset = static_cast<void*>(soff); 203 Kokkos::deep_copy (*soff, h_species_offset); 204 205 const Kokkos::View<PetscInt*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_mat_offset (a_mat_offset, num_grids+1); 206 auto moff = new Kokkos::View<PetscInt*, Kokkos::LayoutLeft> ("mat_offset",num_grids+1); 207 SData_d->mat_offset = static_cast<void*>(moff); 208 Kokkos::deep_copy (*moff, h_mat_offset); 209 210 const Kokkos::View<PetscInt*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_ip_offset (ip_offset, num_grids+1); 211 auto ipoff = new Kokkos::View<PetscInt*, Kokkos::LayoutLeft> ("ip_offset",num_grids+1); 212 SData_d->ip_offset = static_cast<void*>(ipoff); 213 Kokkos::deep_copy (*ipoff, h_ip_offset); 214 215 const Kokkos::View<PetscInt*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_elem_offset (elem_offset, num_grids+1); 216 auto eoff = new Kokkos::View<PetscInt*, Kokkos::LayoutLeft> ("elem_offset",num_grids+1); 217 SData_d->elem_offset = static_cast<void*>(eoff); 218 Kokkos::deep_copy (*eoff, h_elem_offset); 219 220 const Kokkos::View<PetscInt*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_ipf_offset (ipf_offset, num_grids+1); 221 auto ipfoff = new Kokkos::View<PetscInt*, Kokkos::LayoutLeft> ("ipf_offset",num_grids+1); 222 SData_d->ipf_offset = static_cast<void*>(ipfoff); 223 Kokkos::deep_copy (*ipfoff, h_ipf_offset); 224 225 #if defined(LANDAU_LAYOUT_LEFT) // preallocate dynamic data, no copy 226 auto ipfdf_data = new Kokkos::View<PetscReal**, Kokkos::LayoutLeft > ("fdf", dim+1, IPf_sz); 227 #else 228 auto ipfdf_data = new Kokkos::View<PetscReal**, Kokkos::LayoutRight > ("fdf", dim+1, IPf_sz); 229 #endif 230 SData_d->ipfdf_data = static_cast<void*>(ipfdf_data); 231 232 auto Eq_m = new Kokkos::View<PetscReal*, Kokkos::LayoutLeft> ("Eq_m",Nftot); // allocate but do not set 233 SData_d->Eq_m = static_cast<void*>(Eq_m); 234 } 235 SData_d->maps = NULL; // not used 236 PetscFunctionReturn(0); 237 } 238 239 PetscErrorCode LandauKokkosStaticDataClear(LandauStaticData *SData_d) 240 { 241 PetscFunctionBegin; 242 if (SData_d->alpha) { 243 auto alpha = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->alpha); 244 delete alpha; 245 SData_d->alpha = NULL; 246 auto beta = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->beta); 247 delete beta; 248 auto invMass = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->invMass); 249 delete invMass; 250 auto B = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->B); 251 delete B; 252 auto D = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->D); 253 delete D; 254 auto invJ = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->invJ); 255 delete invJ; 256 auto x = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->x); 257 delete x; 258 auto y = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->y); 259 delete y; 260 if (SData_d->z) { 261 auto z = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->z); 262 delete z; 263 } 264 #if defined(LANDAU_LAYOUT_LEFT) // preallocate dynamic data, no copy 265 auto z = static_cast<Kokkos::View<PetscReal**, Kokkos::LayoutLeft>*>(SData_d->ipfdf_data); 266 #else 267 auto z = static_cast<Kokkos::View<PetscReal**, Kokkos::LayoutRight>*>(SData_d->ipfdf_data); 268 #endif 269 delete z; 270 auto w = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->w); 271 delete w; 272 auto Eq_m = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->Eq_m); 273 delete Eq_m; 274 // offset 275 auto nc = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->NCells); 276 delete nc; 277 auto soff = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->species_offset); 278 delete soff; 279 auto moff = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->mat_offset); 280 delete moff; 281 auto ipoff = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->ip_offset); 282 delete ipoff; 283 auto eoff = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->elem_offset); 284 delete eoff; 285 auto ipfoff = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->ipf_offset); 286 delete ipfoff; 287 } 288 PetscFunctionReturn(0); 289 } 290 291 #define KOKKOS_SHARED_LEVEL 1 292 KOKKOS_INLINE_FUNCTION 293 PetscErrorCode landau_mat_assemble(PetscSplitCSRDataStructure d_mat, const team_member team, 294 Kokkos::View<PetscScalar**, Kokkos::LayoutRight, Kokkos::DefaultExecutionSpace::scratch_memory_space> s_fieldMats, 295 Kokkos::View<PetscInt**, Kokkos::LayoutRight, Kokkos::DefaultExecutionSpace::scratch_memory_space> s_idx, 296 Kokkos::View<PetscReal**, Kokkos::LayoutRight, Kokkos::DefaultExecutionSpace::scratch_memory_space> s_scale, 297 const PetscInt Nb, const PetscInt Nq, const PetscInt nfaces, const PetscInt moffset, const PetscInt elem, const PetscInt fieldA, const P4estVertexMaps *d_maps) 298 { 299 const LandauIdx *const Idxs = &d_maps->gIdx[elem][fieldA][0]; 300 team.team_barrier(); 301 Kokkos::parallel_for(Kokkos::TeamVectorRange(team,0,Nb), [=] (int f) { 302 PetscInt q, idx = Idxs[f]; 303 if (idx >= 0) { 304 s_idx(f,0) = idx + moffset; 305 s_scale(f,0) = 1.; 306 } else { 307 idx = -idx - 1; 308 for (q = 0; q < nfaces; q++) { 309 s_idx(f,q) = d_maps->c_maps[idx][q].gid + moffset; 310 s_scale(f,q) = d_maps->c_maps[idx][q].scale; 311 } 312 } 313 }); 314 team.team_barrier(); 315 Kokkos::parallel_for(Kokkos::TeamThreadRange(team,0,Nb), [=] (int f) { 316 PetscInt nr,idx = Idxs[f]; 317 if (idx >= 0) { 318 nr = 1; 319 } else { 320 nr = nfaces; 321 } 322 Kokkos::parallel_for(Kokkos::ThreadVectorRange(team,0,Nb), [=] (int g) { 323 PetscScalar vals[LANDAU_MAX_Q_FACE*LANDAU_MAX_Q_FACE]; 324 PetscInt q,d,nc,idx = Idxs[g]; 325 if (idx >= 0) { 326 nc = 1; 327 } else { 328 nc = nfaces; 329 } 330 for (q = 0; q < nr; q++) { 331 for (d = 0; d < nc; d++) { 332 vals[q*nc + d] = s_scale(f,q)*s_scale(g,d)*s_fieldMats(f,g); 333 } 334 } 335 MatSetValuesDevice(d_mat,nr,&s_idx(f,0),nc,&s_idx(g,0),vals,ADD_VALUES); 336 }); 337 }); 338 return 0; 339 } 340 341 PetscErrorCode LandauKokkosJacobian(DM plex[], const PetscInt Nq, const PetscInt num_grids, const PetscInt a_numCells[], PetscReal a_Eq_m[], PetscScalar a_elem_closure[], 342 const PetscInt N, const PetscScalar a_xarray[], const LandauStaticData *SData_d, const PetscInt num_sub_blocks, const PetscReal shift, 343 const PetscLogEvent events[], const PetscInt a_mat_offset[], const PetscInt a_species_offset[], Mat subJ[], Mat JacP) 344 { 345 using scr_mem_t = Kokkos::DefaultExecutionSpace::scratch_memory_space; 346 using fieldMats_scr_t = Kokkos::View<PetscScalar**, Kokkos::LayoutRight, scr_mem_t>; 347 using idx_scr_t = Kokkos::View<PetscInt**, Kokkos::LayoutRight, scr_mem_t>; 348 using scale_scr_t = Kokkos::View<PetscReal**, Kokkos::LayoutRight, scr_mem_t>; 349 using g2_scr_t = Kokkos::View<PetscReal***, Kokkos::LayoutRight, scr_mem_t>; 350 using g3_scr_t = Kokkos::View<PetscReal****, Kokkos::LayoutRight, scr_mem_t>; 351 PetscErrorCode ierr; 352 PetscInt Nb=Nq,dim,num_cells_max,num_cells_tot,Nf_max,nip_global; 353 int nfaces=0; 354 LandauCtx *ctx; 355 PetscReal *d_Eq_m=NULL; 356 PetscScalar *d_vertex_f=NULL; 357 P4estVertexMaps *maps[LANDAU_MAX_GRIDS]; // this gets captured 358 PetscSplitCSRDataStructure d_mat; 359 PetscContainer container = NULL; 360 const int conc = Kokkos::DefaultExecutionSpace().concurrency(), openmp = !!(conc < 1000), team_size = (openmp==0) ? Nq : 1; 361 auto d_alpha_k = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->alpha); //static data 362 const PetscReal *d_alpha = d_alpha_k->data(); 363 const PetscInt Nftot = d_alpha_k->size(); // total number of species 364 auto d_beta_k = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->beta); 365 const PetscReal *d_beta = d_beta_k->data(); 366 auto d_invMass_k = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->invMass); 367 const PetscReal *d_invMass = d_invMass_k->data(); 368 auto d_B_k = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->B); 369 const PetscReal *d_BB = d_B_k->data(); 370 auto d_D_k = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->D); 371 const PetscReal *d_DD = d_D_k->data(); 372 auto d_invJ_k = *static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->invJ); // use Kokkos vector in kernels 373 auto d_x_k = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->x); //static data 374 const PetscReal *d_x = d_x_k->data(); 375 auto d_y_k = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->y); //static data 376 const PetscReal *d_y = d_y_k->data(); 377 auto d_z_k = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->z); //static data 378 const PetscReal *d_z = (LANDAU_DIM==3) ? d_z_k->data() : NULL; 379 auto d_w_k = *static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->w); //static data 380 const PetscReal *d_w = d_w_k.data(); 381 auto d_Eq_m_k = static_cast<Kokkos::View<PetscReal*, Kokkos::LayoutLeft>*>(SData_d->Eq_m); // static storage, dynamci data - E(t), copy later 382 // grid offset d_numCells 383 auto d_numCells_k = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->NCells); 384 const PetscInt *d_numCells = d_numCells_k->data(); 385 auto d_species_offset_k = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->species_offset); 386 const PetscInt *d_species_offset = d_species_offset_k->data(); 387 auto d_mat_offset_k = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->mat_offset); 388 const PetscInt *d_mat_offset = d_mat_offset_k->data(); 389 auto d_ip_offset_k = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->ip_offset); 390 const PetscInt *d_ip_offset = d_ip_offset_k->data(); 391 auto d_ipf_offset_k = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->ipf_offset); 392 const PetscInt *d_ipf_offset = d_ipf_offset_k->data(); 393 auto d_elem_offset_k = static_cast<Kokkos::View<PetscInt*, Kokkos::LayoutLeft>*>(SData_d->elem_offset); 394 const PetscInt *d_elem_offset = d_elem_offset_k->data(); 395 #if defined(LANDAU_LAYOUT_LEFT) // preallocate dynamic data, no copy 396 Kokkos::View<PetscReal**, Kokkos::LayoutLeft > d_fdf_k = *static_cast<Kokkos::View<PetscReal**, Kokkos::LayoutLeft >*>(SData_d->ipfdf_data); 397 #else 398 Kokkos::View<PetscReal**, Kokkos::LayoutRight > d_fdf_k = *static_cast<Kokkos::View<PetscReal**, Kokkos::LayoutRight >*>(SData_d->ipfdf_data); 399 #endif 400 PetscFunctionBegin; 401 ierr = PetscLogEventBegin(events[3],0,0,0,0);CHKERRQ(ierr); 402 ierr = DMGetApplicationContext(plex[0], &ctx);CHKERRQ(ierr); 403 if (!ctx) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_PLIB, "no context"); 404 ierr = DMGetDimension(plex[0], &dim);CHKERRQ(ierr); 405 if (LANDAU_DIM != dim) SETERRQ2(PETSC_COMM_WORLD, PETSC_ERR_PLIB, "dim %D != LANDAU_DIM %d",dim,LANDAU_DIM); 406 if (ctx->gpu_assembly) { 407 ierr = PetscObjectQuery((PetscObject) JacP, "assembly_maps", (PetscObject *) &container);CHKERRQ(ierr); 408 if (container) { // not here first call 409 P4estVertexMaps *h_maps=NULL; 410 ierr = PetscContainerGetPointer(container, (void **) &h_maps);CHKERRQ(ierr); 411 for (PetscInt grid=0 ; grid<num_grids ; grid++) { 412 if (h_maps[grid].d_self) { 413 maps[grid] = h_maps[grid].d_self; 414 nfaces = h_maps[grid].num_face; // nface=0 for CPU assembly 415 } else { 416 SETERRQ(PETSC_COMM_SELF, PETSC_ERR_PLIB, "GPU assembly but no metadata in container"); 417 } 418 } 419 // this does the setup the first time called 420 ierr = MatKokkosGetDeviceMatWrite(JacP,&d_mat);CHKERRQ(ierr); 421 } else { // kernel output - first call assembled on device 422 for (PetscInt grid=0 ; grid<num_grids ; grid++) maps[grid] = NULL; 423 nfaces = 0; 424 } 425 } else { 426 for (PetscInt grid=0 ; grid<num_grids ; grid++) maps[grid] = NULL; 427 nfaces = 0; 428 } 429 nip_global = num_cells_tot = Nf_max = num_cells_max = 0; 430 for (PetscInt grid=0 ; grid<num_grids ; grid++) { 431 int Nfloc = a_species_offset[grid+1] - a_species_offset[grid]; 432 if (Nfloc > Nf_max) Nf_max = Nfloc; 433 if (a_numCells[grid] > num_cells_max) num_cells_max = a_numCells[grid]; 434 num_cells_tot += a_numCells[grid]; 435 nip_global += Nq*a_numCells[grid]; 436 } 437 const PetscInt totDim_max = Nf_max*Nq, elem_mat_size_max = totDim_max*totDim_max; 438 const PetscInt elem_mat_num_cells_max_grid = container ? 0 : num_cells_max; 439 Kokkos::View<PetscScalar***, Kokkos::LayoutRight> d_elem_mats("element matrices", num_grids, elem_mat_num_cells_max_grid, elem_mat_size_max); // first call have large set of global element matrices 440 const Kokkos::View<PetscReal*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_Eq_m_k (a_Eq_m, Nftot); 441 if (a_elem_closure || a_xarray) { 442 Kokkos::deep_copy (*d_Eq_m_k, h_Eq_m_k); 443 d_Eq_m = d_Eq_m_k->data(); 444 } else d_Eq_m = NULL; 445 ierr = PetscKokkosInitializeCheck();CHKERRQ(ierr); 446 ierr = PetscLogEventEnd(events[3],0,0,0,0);CHKERRQ(ierr); 447 if (a_elem_closure || a_xarray) { // Jacobian, create f & df 448 Kokkos::View<PetscScalar*, Kokkos::LayoutLeft> *d_vertex_f_k = NULL; 449 ierr = PetscLogEventBegin(events[1],0,0,0,0);CHKERRQ(ierr); 450 if (a_elem_closure) { 451 PetscInt closure_sz = 0; // argh, don't have this on the host!!! 452 for (PetscInt grid=0 ; grid<num_grids ; grid++) { 453 PetscInt nfloc = a_species_offset[grid+1] - a_species_offset[grid]; 454 closure_sz += Nq*nfloc*a_numCells[grid]; 455 } 456 d_vertex_f_k = new Kokkos::View<PetscScalar*, Kokkos::LayoutLeft> ("closure",closure_sz); 457 const Kokkos::View<PetscScalar*, Kokkos::LayoutLeft, Kokkos::HostSpace, Kokkos::MemoryTraits<Kokkos::Unmanaged> > h_closure_k (a_elem_closure, closure_sz); // Vertex data for each element 458 Kokkos::deep_copy (*d_vertex_f_k, h_closure_k); 459 d_vertex_f = d_vertex_f_k->data(); 460 } else { 461 d_vertex_f = (PetscScalar*)a_xarray; 462 } 463 ierr = PetscLogEventEnd(events[1],0,0,0,0);CHKERRQ(ierr); 464 ierr = PetscLogEventBegin(events[8],0,0,0,0);CHKERRQ(ierr); 465 ierr = PetscLogGpuTimeBegin();CHKERRQ(ierr); 466 Kokkos::parallel_for("f, df", Kokkos::TeamPolicy<>(num_cells_tot, team_size, /* Kokkos::AUTO */ 16), KOKKOS_LAMBDA (const team_member team) { 467 // find my grid 468 PetscInt grid = 0, g_cell = team.league_rank(); 469 while (g_cell >= d_elem_offset[grid+1]) grid++; // yuck search for grid 470 { 471 const PetscInt moffset = d_mat_offset[grid], nip_loc = d_numCells[grid]*Nq, Nfloc = d_species_offset[grid+1] - d_species_offset[grid], elem = g_cell - d_elem_offset[grid]; 472 const PetscInt IP_idx = d_ip_offset[grid], IPf_idx = d_ipf_offset[grid]; 473 const PetscScalar *coef; 474 PetscScalar coef_buff[LANDAU_MAX_SPECIES*LANDAU_MAX_NQ]; 475 // un pack IPData 476 if (!maps[grid]) { 477 coef = &d_vertex_f[elem*Nb*Nfloc + IPf_idx]; // closure and IP indexing are the same 478 } else { 479 coef = coef_buff; 480 for (int f = 0; f < Nfloc; ++f) { 481 LandauIdx *const Idxs = &maps[grid]->gIdx[elem][f][0]; 482 for (int b = 0; b < Nb; ++b) { 483 PetscInt idx = Idxs[b]; 484 if (idx >= 0) { 485 coef_buff[f*Nb+b] = d_vertex_f[idx+moffset]; // xarray data, not IP 486 } else { 487 idx = -idx - 1; 488 coef_buff[f*Nb+b] = 0; 489 for (int q = 0; q < maps[grid]->num_face; q++) { 490 PetscInt id = maps[grid]->c_maps[idx][q].gid; 491 PetscScalar scale = maps[grid]->c_maps[idx][q].scale; 492 coef_buff[f*Nb+b] += scale*d_vertex_f[id+moffset]; 493 } 494 } 495 } 496 } 497 } 498 Kokkos::parallel_for(Kokkos::TeamThreadRange(team,0,Nq), [=] (int myQi) { 499 const PetscInt ipidx = IP_idx + myQi + elem * Nq, ipidx_g = myQi + elem * Nq; 500 const PetscReal *const invJj = &d_invJ_k(ipidx*dim*dim); 501 const PetscReal *Bq = &d_BB[myQi*Nb], *Dq = &d_DD[myQi*Nb*dim]; 502 Kokkos::parallel_for(Kokkos::ThreadVectorRange(team,0,(int)Nfloc), [=] (int f) { 503 PetscInt b, e, d; 504 PetscReal refSpaceDer[LANDAU_DIM]; 505 const PetscInt idx = IPf_idx + f*nip_loc + ipidx_g; 506 d_fdf_k(0,idx) = 0.0; 507 for (d = 0; d < LANDAU_DIM; ++d) refSpaceDer[d] = 0.0; 508 for (b = 0; b < Nb; ++b) { 509 const PetscInt cidx = b; 510 d_fdf_k(0,idx) += Bq[cidx]*PetscRealPart(coef[f*Nb+cidx]); 511 for (d = 0; d < dim; ++d) refSpaceDer[d] += Dq[cidx*dim+d]*PetscRealPart(coef[f*Nb+cidx]); 512 } 513 for (d = 0; d < dim; ++d) { 514 for (e = 0, d_fdf_k(d+1,idx) = 0.0; e < dim; ++e) { 515 d_fdf_k(d+1,idx) += invJj[e*dim+d]*refSpaceDer[e]; 516 } 517 } 518 }); // f 519 }); // myQi 520 } 521 }); // global elems - fdf 522 // #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP) 523 // ierr = PetscLogGpuFlops(nip_loc*(PetscLogDouble)(2*Nb*(1+dim)));CHKERRQ(ierr); 524 // #else 525 // ierr = PetscLogFlops(nip_loc*(PetscLogDouble)(2*Nb*(1+dim)));CHKERRQ(ierr); 526 // #endif 527 Kokkos::fence(); 528 ierr = PetscLogGpuTimeEnd();CHKERRQ(ierr); // is this a fence? 529 ierr = PetscLogEventEnd(events[8],0,0,0,0);CHKERRQ(ierr); 530 if (d_vertex_f_k) delete d_vertex_f_k; 531 // Jacobian 532 ierr = PetscLogEventBegin(events[4],0,0,0,0);CHKERRQ(ierr); 533 ierr = PetscLogGpuTimeBegin();CHKERRQ(ierr); 534 // #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP) 535 // ierr = PetscLogGpuFlops(nip_loc*(PetscLogDouble)((nip_loc*(11*Nfloc+ 4*dim*dim) + 6*Nfloc*dim*dim*dim + 10*Nfloc*dim*dim + 4*Nfloc*dim + Nb*Nfloc*Nb*Nq*dim*dim*5)));CHKERRQ(ierr); 536 // #else 537 // ierr = PetscLogFlops(nip_loc*(PetscLogDouble)((nip_loc*(11*Nfloc+ 4*dim*dim) + 6*Nfloc*dim*dim*dim + 10*Nfloc*dim*dim + 4*Nfloc*dim + Nb*Nfloc*Nb*Nq*dim*dim*5)));CHKERRQ(ierr); 538 // #endif 539 const int scr_bytes = 2*(g2_scr_t::shmem_size(dim,Nf_max,Nq) + g3_scr_t::shmem_size(dim,dim,Nf_max,Nq))+fieldMats_scr_t::shmem_size(Nb,Nb)+idx_scr_t::shmem_size(Nb,nfaces)+scale_scr_t::shmem_size(Nb,nfaces); 540 ierr = PetscInfo6(plex[0], "Jacobian shared memory size: %d bytes in level %d num_cells_tot=%D team size=%D #face=%D Nf_max=%D\n",scr_bytes,KOKKOS_SHARED_LEVEL,num_cells_tot,team_size,nfaces,Nf_max);CHKERRQ(ierr); 541 Kokkos::parallel_for("Jacobian", Kokkos::TeamPolicy<>(num_cells_tot, team_size, /* Kokkos::AUTO */ 16).set_scratch_size(KOKKOS_SHARED_LEVEL, Kokkos::PerTeam(scr_bytes)), KOKKOS_LAMBDA (const team_member team) { 542 // find my grid 543 PetscInt grid = 0, g_cell = team.league_rank(); 544 while (g_cell >= d_elem_offset[grid+1]) grid++; // yuck search for grid 545 { 546 const PetscInt moffset = d_mat_offset[grid], Nfloc = d_species_offset[grid+1]-d_species_offset[grid], elem = g_cell-d_elem_offset[grid], totDim = Nfloc*Nq; 547 const PetscInt IP_idx = d_ip_offset[grid]; 548 const PetscInt f_off = d_species_offset[grid]; 549 g2_scr_t g2(team.team_scratch(KOKKOS_SHARED_LEVEL),dim,Nfloc,Nq); 550 g3_scr_t g3(team.team_scratch(KOKKOS_SHARED_LEVEL),dim,dim,Nfloc,Nq); 551 g2_scr_t gg2(team.team_scratch(KOKKOS_SHARED_LEVEL),dim,Nfloc,Nq); 552 g3_scr_t gg3(team.team_scratch(KOKKOS_SHARED_LEVEL),dim,dim,Nfloc,Nq); 553 // get g2[] & g3[] 554 Kokkos::parallel_for(Kokkos::TeamThreadRange(team,0,Nq), [=] (int myQi) { 555 using Kokkos::parallel_reduce; 556 const PetscInt jpidx_g = myQi + elem * Nq, jpidx = IP_idx + jpidx_g; 557 const PetscReal* const invJj = &d_invJ_k(jpidx*dim*dim); 558 const PetscReal vj[3] = {d_x[jpidx], d_y[jpidx], d_z ? d_z[jpidx] : 0}, wj = d_w[jpidx]; 559 landau_inner_red::TensorValueType gg_temp; // reduce on part of gg2 and g33 for IP jpidx_g 560 Kokkos::parallel_reduce(Kokkos::ThreadVectorRange (team, (int)nip_global), [=] (const int& ipidx, landau_inner_red::TensorValueType & ggg) { 561 const PetscReal wi = d_w[ipidx], x = d_x[ipidx], y = d_y[ipidx]; 562 PetscReal temp1[3] = {0, 0, 0}, temp2 = 0; 563 PetscInt fieldA,d2,d3,f_off_r,grid_r,ipidx_g,nip_loc_r,Nfloc_r,IPf_idx_r; 564 #if LANDAU_DIM==2 565 PetscReal Ud[2][2], Uk[2][2], mask = (PetscAbs(vj[0]-x) < 100*PETSC_SQRT_MACHINE_EPSILON && PetscAbs(vj[1]-y) < 100*PETSC_SQRT_MACHINE_EPSILON) ? 0. : 1.; 566 LandauTensor2D(vj, x, y, Ud, Uk, mask); 567 #else 568 PetscReal U[3][3], z = d_z[ipidx], mask = (PetscAbs(vj[0]-x) < 100*PETSC_SQRT_MACHINE_EPSILON && PetscAbs(vj[1]-y) < 100*PETSC_SQRT_MACHINE_EPSILON && PetscAbs(vj[2]-z) < 100*PETSC_SQRT_MACHINE_EPSILON) ? 0. : 1.; 569 LandauTensor3D(vj, x, y, z, U, mask); 570 #endif 571 grid_r = 0; 572 while (ipidx >= d_ip_offset[grid_r+1]) grid_r++; // yuck search for grid 573 f_off_r = d_species_offset[grid_r]; 574 ipidx_g = ipidx - d_ip_offset[grid_r]; 575 nip_loc_r = d_numCells[grid_r]*Nq; 576 Nfloc_r = d_species_offset[grid_r+1] - d_species_offset[grid_r]; 577 IPf_idx_r = d_ipf_offset[grid_r]; 578 for (fieldA = 0; fieldA < Nfloc_r; ++fieldA) { 579 const PetscInt idx = IPf_idx_r + fieldA*nip_loc_r + ipidx_g; 580 temp1[0] += d_fdf_k(1,idx)*d_beta[fieldA+f_off_r]*d_invMass[fieldA+f_off_r]; 581 temp1[1] += d_fdf_k(2,idx)*d_beta[fieldA+f_off_r]*d_invMass[fieldA+f_off_r]; 582 #if LANDAU_DIM==3 583 temp1[2] += d_fdf_k(3,idx)*d_beta[fieldA+f_off_r]*d_invMass[fieldA+f_off_r]; 584 #endif 585 temp2 += d_fdf_k(0,idx)*d_beta[fieldA+f_off_r]; 586 } 587 temp1[0] *= wi; 588 temp1[1] *= wi; 589 #if LANDAU_DIM==3 590 temp1[2] *= wi; 591 #endif 592 temp2 *= wi; 593 #if LANDAU_DIM==2 594 for (d2 = 0; d2 < 2; d2++) { 595 for (d3 = 0; d3 < 2; ++d3) { 596 /* K = U * grad(f): g2=e: i,A */ 597 ggg.gg2[d2] += Uk[d2][d3]*temp1[d3]; 598 /* D = -U * (I \kron (fx)): g3=f: i,j,A */ 599 ggg.gg3[d2][d3] += Ud[d2][d3]*temp2; 600 } 601 } 602 #else 603 for (d2 = 0; d2 < 3; ++d2) { 604 for (d3 = 0; d3 < 3; ++d3) { 605 /* K = U * grad(f): g2 = e: i,A */ 606 ggg.gg2[d2] += U[d2][d3]*temp1[d3]; 607 /* D = -U * (I \kron (fx)): g3 = f: i,j,A */ 608 ggg.gg3[d2][d3] += U[d2][d3]*temp2; 609 } 610 } 611 #endif 612 }, Kokkos::Sum<landau_inner_red::TensorValueType>(gg_temp)); 613 // add alpha and put in gg2/3 614 Kokkos::parallel_for(Kokkos::ThreadVectorRange (team, (int)Nfloc), [&] (const int& fieldA) { 615 PetscInt d2,d3; 616 for (d2 = 0; d2 < dim; d2++) { 617 gg2(d2,fieldA,myQi) = gg_temp.gg2[d2]*d_alpha[fieldA+f_off]; 618 for (d3 = 0; d3 < dim; d3++) { 619 gg3(d2,d3,fieldA,myQi) = -gg_temp.gg3[d2][d3]*d_alpha[fieldA+f_off]*d_invMass[fieldA+f_off]; 620 } 621 } 622 }); 623 /* add electric field term once per IP */ 624 Kokkos::parallel_for(Kokkos::ThreadVectorRange (team, (int)Nfloc), [&] (const int& fieldA) { 625 gg2(dim-1,fieldA,myQi) += d_Eq_m[fieldA+f_off]; 626 }); 627 Kokkos::parallel_for(Kokkos::ThreadVectorRange (team, (int)Nfloc), [=] (const int& fieldA) { 628 int d,d2,d3,dp; 629 /* Jacobian transform - g2, g3 - per thread (2D) */ 630 for (d = 0; d < dim; ++d) { 631 g2(d,fieldA,myQi) = 0; 632 for (d2 = 0; d2 < dim; ++d2) { 633 g2(d,fieldA,myQi) += invJj[d*dim+d2]*gg2(d2,fieldA,myQi); 634 g3(d,d2,fieldA,myQi) = 0; 635 for (d3 = 0; d3 < dim; ++d3) { 636 for (dp = 0; dp < dim; ++dp) { 637 g3(d,d2,fieldA,myQi) += invJj[d*dim + d3]*gg3(d3,dp,fieldA,myQi)*invJj[d2*dim + dp]; 638 } 639 } 640 g3(d,d2,fieldA,myQi) *= wj; 641 } 642 g2(d,fieldA,myQi) *= wj; 643 } 644 }); 645 }); // Nq 646 team.team_barrier(); 647 { /* assemble */ 648 fieldMats_scr_t s_fieldMats(team.team_scratch(KOKKOS_SHARED_LEVEL),Nb,Nb); // Only used for GPU assembly (ie, not first pass) 649 idx_scr_t s_idx(team.team_scratch(KOKKOS_SHARED_LEVEL),Nb,nfaces); 650 scale_scr_t s_scale(team.team_scratch(KOKKOS_SHARED_LEVEL),Nb,nfaces); 651 for (PetscInt fieldA = 0; fieldA < Nfloc; fieldA++) { 652 /* assemble */ 653 Kokkos::parallel_for(Kokkos::TeamThreadRange(team,0,Nb), [=] (int f) { 654 Kokkos::parallel_for(Kokkos::ThreadVectorRange(team,0,Nb), [=] (int g) { 655 PetscScalar t = 0; 656 for (int qj = 0 ; qj < Nq ; qj++) { // look at others integration points 657 const PetscReal *BJq = &d_BB[qj*Nb], *DIq = &d_DD[qj*Nb*dim]; 658 for (int d = 0; d < dim; ++d) { 659 t += DIq[f*dim+d]*g2(d,fieldA,qj)*BJq[g]; 660 for (int d2 = 0; d2 < dim; ++d2) { 661 t += DIq[f*dim + d]*g3(d,d2,fieldA,qj)*DIq[g*dim + d2]; 662 } 663 } 664 } 665 if (elem_mat_num_cells_max_grid) { // CPU assembly 666 const PetscInt fOff = (fieldA*Nb + f)*totDim + fieldA*Nb + g; 667 d_elem_mats(grid,elem,fOff) = t; 668 } else s_fieldMats(f,g) = t; 669 }); 670 }); 671 if (!elem_mat_num_cells_max_grid) { // GPU assembly 672 landau_mat_assemble (d_mat, team, s_fieldMats, s_idx, s_scale, Nb, Nq, nfaces, moffset, elem, fieldA, maps[grid]); 673 } 674 } 675 } 676 } 677 }); 678 ierr = PetscLogGpuTimeEnd();CHKERRQ(ierr); 679 ierr = PetscLogEventEnd(events[4],0,0,0,0);CHKERRQ(ierr); 680 } else { // mass 681 ierr = PetscLogEventBegin(events[4],0,0,0,0);CHKERRQ(ierr); 682 ierr = PetscLogGpuTimeBegin();CHKERRQ(ierr); 683 // #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_HIP) 684 // ierr = PetscLogGpuFlops(nip_loc*(PetscLogDouble)(Nb*Nfloc*Nb*Nq*4));CHKERRQ(ierr); 685 // if (ctx->deviceType == LANDAU_CPU) PetscInfo(plex[grid], "Warning: Landau selected CPU but no support for Kokkos using CPU\n"); 686 // #else 687 // ierr = PetscLogFlops(nip_loc*(PetscLogDouble)(Nb*Nfloc*Nb*Nq*4));CHKERRQ(ierr); 688 // #endif 689 int scr_bytes = fieldMats_scr_t::shmem_size(Nq,Nq) + idx_scr_t::shmem_size(Nb,nfaces) + scale_scr_t::shmem_size(Nb,nfaces); 690 ierr = PetscInfo6(plex[0], "Mass shared memory size: %d bytes in level %d conc=%D team size=%D #face=%D Nb=%D\n",scr_bytes,KOKKOS_SHARED_LEVEL,conc,team_size,nfaces,Nb);CHKERRQ(ierr); 691 Kokkos::parallel_for("Mass", Kokkos::TeamPolicy<>(num_cells_tot, team_size, /* Kokkos::AUTO */ 16).set_scratch_size(KOKKOS_SHARED_LEVEL, Kokkos::PerTeam(scr_bytes)), KOKKOS_LAMBDA (const team_member team) { 692 fieldMats_scr_t s_fieldMats(team.team_scratch(KOKKOS_SHARED_LEVEL),Nb,Nb); 693 idx_scr_t s_idx(team.team_scratch(KOKKOS_SHARED_LEVEL),Nb,nfaces); 694 scale_scr_t s_scale(team.team_scratch(KOKKOS_SHARED_LEVEL),Nb,nfaces); 695 // find my grid 696 PetscInt grid = 0, g_cell = team.league_rank(); 697 while (g_cell >= d_elem_offset[grid+1]) grid++; // yuck search for grid 698 { 699 const PetscInt moffset = d_mat_offset[grid], Nfloc = d_species_offset[grid+1]-d_species_offset[grid], totDim = Nfloc*Nq, elem = g_cell-d_elem_offset[grid]; 700 const PetscInt IP_idx = d_ip_offset[grid]; 701 for (int fieldA = 0; fieldA < Nfloc; fieldA++) { 702 /* assemble */ 703 Kokkos::parallel_for(Kokkos::TeamThreadRange(team,0,Nb), [=] (int f) { 704 Kokkos::parallel_for(Kokkos::ThreadVectorRange(team,0,Nb), [=] (int g) { 705 PetscScalar t = 0; 706 for (int qj = 0 ; qj < Nq ; qj++) { // look at others integration points 707 const PetscReal *BJq = &d_BB[qj*Nb]; 708 const PetscInt jpidx = IP_idx + qj + elem * Nq; 709 if (dim==2) { 710 t += BJq[f] * d_w_k(jpidx) * shift * BJq[g] * 2. * PETSC_PI; 711 } else { 712 t += BJq[f] * d_w_k(jpidx) * shift * BJq[g]; 713 } 714 } 715 if (elem_mat_num_cells_max_grid) { 716 const PetscInt fOff = (fieldA*Nb + f)*totDim + fieldA*Nb + g; 717 d_elem_mats(grid,elem,fOff) = t; 718 } else s_fieldMats(f,g) = t; 719 }); 720 }); 721 if (!elem_mat_num_cells_max_grid) { // device assembly 722 landau_mat_assemble (d_mat, team, s_fieldMats, s_idx, s_scale, Nb, Nq,nfaces, moffset, elem, fieldA, maps[grid]); 723 } // else not using GPU assembly 724 } 725 } 726 }); 727 ierr = PetscLogGpuTimeEnd();CHKERRQ(ierr); 728 ierr = PetscLogEventEnd(events[4],0,0,0,0);CHKERRQ(ierr); 729 } 730 Kokkos::fence(); 731 if (elem_mat_num_cells_max_grid) { // CPU assembly 732 Kokkos::View<PetscScalar***, Kokkos::LayoutRight>::HostMirror h_elem_mats = Kokkos::create_mirror_view(d_elem_mats); 733 Kokkos::deep_copy (h_elem_mats, d_elem_mats); 734 for (PetscInt grid=0 ; grid<num_grids ; grid++) { 735 const PetscInt Nfloc = a_species_offset[grid+1]-a_species_offset[grid], totDim = Nfloc*Nq; 736 PetscSection section, globalSection; 737 PetscInt cStart,cEnd; 738 ierr = PetscLogEventBegin(events[5],0,0,0,0);CHKERRQ(ierr); 739 ierr = DMPlexGetHeightStratum(plex[grid],0,&cStart,&cEnd);CHKERRQ(ierr); 740 ierr = DMGetLocalSection(plex[grid], §ion);CHKERRQ(ierr); 741 ierr = DMGetGlobalSection(plex[grid], &globalSection);CHKERRQ(ierr); 742 ierr = PetscLogEventEnd(events[5],0,0,0,0);CHKERRQ(ierr); 743 ierr = PetscLogEventBegin(events[6],0,0,0,0);CHKERRQ(ierr); 744 for (PetscInt ej = cStart ; ej < cEnd; ++ej) { 745 const PetscScalar *elMat = &h_elem_mats(grid,ej-cStart,0); 746 ierr = DMPlexMatSetClosure(plex[grid], section, globalSection, subJ[grid], ej, elMat, ADD_VALUES);CHKERRQ(ierr); 747 if (grid==0 && ej==-1) { 748 int d,f; 749 PetscPrintf(PETSC_COMM_SELF,"Kokkos Element matrix %d/%d\n",1,(int)a_numCells[grid]); 750 for (d = 0; d < totDim; ++d){ 751 for (f = 0; f < totDim; ++f) PetscPrintf(PETSC_COMM_SELF," %12.5e", PetscRealPart(elMat[d*totDim + f])); 752 PetscPrintf(PETSC_COMM_SELF,"\n"); 753 } 754 exit(14); 755 } 756 } 757 // transition to use of maps for a Kokkos VecGetClosure 758 if (ctx->gpu_assembly) { 759 if (!(a_elem_closure || a_xarray)) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_PLIB, "transition in Mass"); 760 } 761 if (!container) { // move nest matrix to global JacP 762 PetscInt moffset = a_mat_offset[grid], nloc, nzl, colbuf[1024], row; 763 const PetscInt *cols; 764 const PetscScalar *vals; 765 Mat B = subJ[grid]; 766 ierr = MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 767 ierr = MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 768 ierr = MatGetSize(B, &nloc, NULL);CHKERRQ(ierr); 769 if (nloc != a_mat_offset[grid+1] - moffset) SETERRQ2(PetscObjectComm((PetscObject) B), PETSC_ERR_PLIB, "nloc %D != mat_offset[grid+1] - moffset = %D",nloc, a_mat_offset[grid+1] - moffset); 770 for (int i=0 ; i<nloc ; i++) { 771 ierr = MatGetRow(B,i,&nzl,&cols,&vals);CHKERRQ(ierr); 772 if (nzl>1024) SETERRQ1(PetscObjectComm((PetscObject) B), PETSC_ERR_PLIB, "Row too big: %D",nzl); 773 for (int j=0; j<nzl; j++) colbuf[j] = cols[j] + moffset; 774 row = i + moffset; 775 ierr = MatSetValues(JacP,1,&row,nzl,colbuf,vals,ADD_VALUES);CHKERRQ(ierr); 776 ierr = MatRestoreRow(B,i,&nzl,&cols,&vals);CHKERRQ(ierr); 777 } 778 ierr = MatDestroy(&subJ[grid]);CHKERRQ(ierr); 779 } 780 ierr = PetscLogEventEnd(events[6],0,0,0,0);CHKERRQ(ierr); 781 } // grids 782 } 783 PetscFunctionReturn(0); 784 } 785 } // extern "C" 786