Actual source code: plexland.c
1: #include <../src/mat/impls/aij/seq/aij.h>
2: #include <petsc/private/dmpleximpl.h>
3: #include <petsclandau.h>
4: #include <petscts.h>
5: #include <petscdmforest.h>
6: #include <petscdmcomposite.h>
8: /* Landau collision operator */
10: /* relativistic terms */
11: #if defined(PETSC_USE_REAL_SINGLE)
12: #define SPEED_OF_LIGHT 2.99792458e8F
13: #define C_0(v0) (SPEED_OF_LIGHT / v0) /* needed for relativistic tensor on all architectures */
14: #else
15: #define SPEED_OF_LIGHT 2.99792458e8
16: #define C_0(v0) (SPEED_OF_LIGHT / v0) /* needed for relativistic tensor on all architectures */
17: #endif
19: #define PETSC_THREAD_SYNC
20: #include "land_tensors.h"
22: #if defined(PETSC_HAVE_OPENMP)
23: #include <omp.h>
24: #endif
26: static PetscErrorCode LandauGPUMapsDestroy(void *ptr)
27: {
28: P4estVertexMaps *maps = (P4estVertexMaps *)ptr;
29: PetscFunctionBegin;
30: // free device data
31: if (maps[0].deviceType != LANDAU_CPU) {
32: #if defined(PETSC_HAVE_KOKKOS_KERNELS)
33: if (maps[0].deviceType == LANDAU_KOKKOS) {
34: PetscCall(LandauKokkosDestroyMatMaps(maps, maps[0].numgrids)); // implies Kokkos does
35: } // else could be CUDA
36: #elif defined(PETSC_HAVE_CUDA)
37: if (maps[0].deviceType == LANDAU_CUDA) {
38: PetscCall(LandauCUDADestroyMatMaps(maps, maps[0].numgrids));
39: } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_PLIB, "maps->deviceType %d ?????", maps->deviceType);
40: #endif
41: }
42: // free host data
43: for (PetscInt grid = 0; grid < maps[0].numgrids; grid++) {
44: PetscCall(PetscFree(maps[grid].c_maps));
45: PetscCall(PetscFree(maps[grid].gIdx));
46: }
47: PetscCall(PetscFree(maps));
49: PetscFunctionReturn(PETSC_SUCCESS);
50: }
51: static PetscErrorCode energy_f(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nf_dummy, PetscScalar *u, void *actx)
52: {
53: PetscReal v2 = 0;
54: PetscFunctionBegin;
55: /* compute v^2 / 2 */
56: for (int i = 0; i < dim; ++i) v2 += x[i] * x[i];
57: /* evaluate the Maxwellian */
58: u[0] = v2 / 2;
59: PetscFunctionReturn(PETSC_SUCCESS);
60: }
62: /* needs double */
63: static PetscErrorCode gamma_m1_f(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nf_dummy, PetscScalar *u, void *actx)
64: {
65: PetscReal *c2_0_arr = ((PetscReal *)actx);
66: double u2 = 0, c02 = (double)*c2_0_arr, xx;
68: PetscFunctionBegin;
69: /* compute u^2 / 2 */
70: for (int i = 0; i < dim; ++i) u2 += x[i] * x[i];
71: /* gamma - 1 = g_eps, for conditioning and we only take derivatives */
72: xx = u2 / c02;
73: #if defined(PETSC_USE_DEBUG)
74: u[0] = PetscSqrtReal(1. + xx);
75: #else
76: u[0] = xx / (PetscSqrtReal(1. + xx) + 1.) - 1.; // better conditioned. -1 might help condition and only used for derivative
77: #endif
78: PetscFunctionReturn(PETSC_SUCCESS);
79: }
81: /*
82: LandauFormJacobian_Internal - Evaluates Jacobian matrix.
84: Input Parameters:
85: . globX - input vector
86: . actx - optional user-defined context
87: . dim - dimension
89: Output Parameter:
90: . J0acP - Jacobian matrix filled, not created
91: */
92: static PetscErrorCode LandauFormJacobian_Internal(Vec a_X, Mat JacP, const PetscInt dim, PetscReal shift, void *a_ctx)
93: {
94: LandauCtx *ctx = (LandauCtx *)a_ctx;
95: PetscInt numCells[LANDAU_MAX_GRIDS], Nq, Nb;
96: PetscQuadrature quad;
97: PetscReal Eq_m[LANDAU_MAX_SPECIES]; // could be static data w/o quench (ex2)
98: PetscScalar *cellClosure = NULL;
99: const PetscScalar *xdata = NULL;
100: PetscDS prob;
101: PetscContainer container;
102: P4estVertexMaps *maps;
103: Mat subJ[LANDAU_MAX_GRIDS * LANDAU_MAX_BATCH_SZ];
105: PetscFunctionBegin;
109: /* check for matrix container for GPU assembly. Support CPU assembly for debugging */
110: PetscCheck(ctx->plex[0] != NULL, ctx->comm, PETSC_ERR_ARG_WRONG, "Plex not created");
111: PetscCall(PetscLogEventBegin(ctx->events[10], 0, 0, 0, 0));
112: PetscCall(DMGetDS(ctx->plex[0], &prob)); // same DS for all grids
113: PetscCall(PetscObjectQuery((PetscObject)JacP, "assembly_maps", (PetscObject *)&container));
114: if (container) {
115: PetscCheck(ctx->gpu_assembly, ctx->comm, PETSC_ERR_ARG_WRONG, "maps but no GPU assembly");
116: PetscCall(PetscContainerGetPointer(container, (void **)&maps));
117: PetscCheck(maps, ctx->comm, PETSC_ERR_ARG_WRONG, "empty GPU matrix container");
118: for (PetscInt i = 0; i < ctx->num_grids * ctx->batch_sz; i++) subJ[i] = NULL;
119: } else {
120: PetscCheck(!ctx->gpu_assembly, ctx->comm, PETSC_ERR_ARG_WRONG, "No maps but GPU assembly");
121: for (PetscInt tid = 0; tid < ctx->batch_sz; tid++) {
122: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) PetscCall(DMCreateMatrix(ctx->plex[grid], &subJ[LAND_PACK_IDX(tid, grid)]));
123: }
124: maps = NULL;
125: }
126: // get dynamic data (Eq is odd, for quench and Spitzer test) for CPU assembly and raw data for Jacobian GPU assembly. Get host numCells[], Nq (yuck)
127: PetscCall(PetscFEGetQuadrature(ctx->fe[0], &quad));
128: PetscCall(PetscQuadratureGetData(quad, NULL, NULL, &Nq, NULL, NULL));
129: Nb = Nq;
130: PetscCheck(Nq <= LANDAU_MAX_NQ, ctx->comm, PETSC_ERR_ARG_WRONG, "Order too high. Nq = %" PetscInt_FMT " > LANDAU_MAX_NQ (%d)", Nq, LANDAU_MAX_NQ);
131: // get metadata for collecting dynamic data
132: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
133: PetscInt cStart, cEnd;
134: PetscCheck(ctx->plex[grid] != NULL, ctx->comm, PETSC_ERR_ARG_WRONG, "Plex not created");
135: PetscCall(DMPlexGetHeightStratum(ctx->plex[grid], 0, &cStart, &cEnd));
136: numCells[grid] = cEnd - cStart; // grids can have different topology
137: }
138: PetscCall(PetscLogEventEnd(ctx->events[10], 0, 0, 0, 0));
139: if (shift == 0) { /* create dynamic point data: f_alpha for closure of each cell (cellClosure[nbatch,ngrids,ncells[g],f[Nb,ns[g]]]) or xdata */
140: DM pack;
141: PetscCall(VecGetDM(a_X, &pack));
142: PetscCheck(pack, PETSC_COMM_SELF, PETSC_ERR_PLIB, "pack has no DM");
143: PetscCall(PetscLogEventBegin(ctx->events[1], 0, 0, 0, 0));
144: for (PetscInt fieldA = 0; fieldA < ctx->num_species; fieldA++) {
145: Eq_m[fieldA] = ctx->Ez * ctx->t_0 * ctx->charges[fieldA] / (ctx->v_0 * ctx->masses[fieldA]); /* normalize dimensionless */
146: if (dim == 2) Eq_m[fieldA] *= 2 * PETSC_PI; /* add the 2pi term that is not in Landau */
147: }
148: if (!ctx->gpu_assembly) {
149: Vec *locXArray, *globXArray;
150: PetscScalar *cellClosure_it;
151: PetscInt cellClosure_sz = 0, nDMs, Nf[LANDAU_MAX_GRIDS];
152: PetscSection section[LANDAU_MAX_GRIDS], globsection[LANDAU_MAX_GRIDS];
153: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
154: PetscCall(DMGetLocalSection(ctx->plex[grid], §ion[grid]));
155: PetscCall(DMGetGlobalSection(ctx->plex[grid], &globsection[grid]));
156: PetscCall(PetscSectionGetNumFields(section[grid], &Nf[grid]));
157: }
158: /* count cellClosure size */
159: PetscCall(DMCompositeGetNumberDM(pack, &nDMs));
160: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) cellClosure_sz += Nb * Nf[grid] * numCells[grid];
161: PetscCall(PetscMalloc1(cellClosure_sz * ctx->batch_sz, &cellClosure));
162: cellClosure_it = cellClosure;
163: PetscCall(PetscMalloc(sizeof(*locXArray) * nDMs, &locXArray));
164: PetscCall(PetscMalloc(sizeof(*globXArray) * nDMs, &globXArray));
165: PetscCall(DMCompositeGetLocalAccessArray(pack, a_X, nDMs, NULL, locXArray));
166: PetscCall(DMCompositeGetAccessArray(pack, a_X, nDMs, NULL, globXArray));
167: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) { // OpenMP (once)
168: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
169: Vec locX = locXArray[LAND_PACK_IDX(b_id, grid)], globX = globXArray[LAND_PACK_IDX(b_id, grid)], locX2;
170: PetscInt cStart, cEnd, ei;
171: PetscCall(VecDuplicate(locX, &locX2));
172: PetscCall(DMGlobalToLocalBegin(ctx->plex[grid], globX, INSERT_VALUES, locX2));
173: PetscCall(DMGlobalToLocalEnd(ctx->plex[grid], globX, INSERT_VALUES, locX2));
174: PetscCall(DMPlexGetHeightStratum(ctx->plex[grid], 0, &cStart, &cEnd));
175: for (ei = cStart; ei < cEnd; ++ei) {
176: PetscScalar *coef = NULL;
177: PetscCall(DMPlexVecGetClosure(ctx->plex[grid], section[grid], locX2, ei, NULL, &coef));
178: PetscCall(PetscMemcpy(cellClosure_it, coef, Nb * Nf[grid] * sizeof(*cellClosure_it))); /* change if LandauIPReal != PetscScalar */
179: PetscCall(DMPlexVecRestoreClosure(ctx->plex[grid], section[grid], locX2, ei, NULL, &coef));
180: cellClosure_it += Nb * Nf[grid];
181: }
182: PetscCall(VecDestroy(&locX2));
183: }
184: }
185: PetscCheck(cellClosure_it - cellClosure == cellClosure_sz * ctx->batch_sz, PETSC_COMM_SELF, PETSC_ERR_PLIB, "iteration wrong %" PetscCount_FMT " != cellClosure_sz = %" PetscInt_FMT, (PetscCount)(cellClosure_it - cellClosure),
186: cellClosure_sz * ctx->batch_sz);
187: PetscCall(DMCompositeRestoreLocalAccessArray(pack, a_X, nDMs, NULL, locXArray));
188: PetscCall(DMCompositeRestoreAccessArray(pack, a_X, nDMs, NULL, globXArray));
189: PetscCall(PetscFree(locXArray));
190: PetscCall(PetscFree(globXArray));
191: xdata = NULL;
192: } else {
193: PetscMemType mtype;
194: if (ctx->jacobian_field_major_order) { // get data in batch ordering
195: PetscCall(VecScatterBegin(ctx->plex_batch, a_X, ctx->work_vec, INSERT_VALUES, SCATTER_FORWARD));
196: PetscCall(VecScatterEnd(ctx->plex_batch, a_X, ctx->work_vec, INSERT_VALUES, SCATTER_FORWARD));
197: PetscCall(VecGetArrayReadAndMemType(ctx->work_vec, &xdata, &mtype));
198: } else {
199: PetscCall(VecGetArrayReadAndMemType(a_X, &xdata, &mtype));
200: }
201: PetscCheck(mtype == PETSC_MEMTYPE_HOST || ctx->deviceType != LANDAU_CPU, ctx->comm, PETSC_ERR_ARG_WRONG, "CPU run with device data: use -mat_type aij");
202: cellClosure = NULL;
203: }
204: PetscCall(PetscLogEventEnd(ctx->events[1], 0, 0, 0, 0));
205: } else xdata = cellClosure = NULL;
207: /* do it */
208: if (ctx->deviceType == LANDAU_CUDA || ctx->deviceType == LANDAU_KOKKOS) {
209: if (ctx->deviceType == LANDAU_CUDA) {
210: #if defined(PETSC_HAVE_CUDA)
211: PetscCall(LandauCUDAJacobian(ctx->plex, Nq, ctx->batch_sz, ctx->num_grids, numCells, Eq_m, cellClosure, xdata, &ctx->SData_d, shift, ctx->events, ctx->mat_offset, ctx->species_offset, subJ, JacP));
212: #else
213: SETERRQ(ctx->comm, PETSC_ERR_ARG_WRONG, "-landau_device_type %s not built", "cuda");
214: #endif
215: } else if (ctx->deviceType == LANDAU_KOKKOS) {
216: #if defined(PETSC_HAVE_KOKKOS_KERNELS)
217: PetscCall(LandauKokkosJacobian(ctx->plex, Nq, ctx->batch_sz, ctx->num_grids, numCells, Eq_m, cellClosure, xdata, &ctx->SData_d, shift, ctx->events, ctx->mat_offset, ctx->species_offset, subJ, JacP));
218: #else
219: SETERRQ(ctx->comm, PETSC_ERR_ARG_WRONG, "-landau_device_type %s not built", "kokkos");
220: #endif
221: }
222: } else { /* CPU version */
223: PetscTabulation *Tf; // used for CPU and print info. Same on all grids and all species
224: PetscInt ip_offset[LANDAU_MAX_GRIDS + 1], ipf_offset[LANDAU_MAX_GRIDS + 1], elem_offset[LANDAU_MAX_GRIDS + 1], IPf_sz_glb, IPf_sz_tot, num_grids = ctx->num_grids, Nf[LANDAU_MAX_GRIDS];
225: PetscReal *ff, *dudx, *dudy, *dudz, *invJ_a = (PetscReal *)ctx->SData_d.invJ, *xx = (PetscReal *)ctx->SData_d.x, *yy = (PetscReal *)ctx->SData_d.y, *zz = (PetscReal *)ctx->SData_d.z, *ww = (PetscReal *)ctx->SData_d.w;
226: PetscReal *nu_alpha = (PetscReal *)ctx->SData_d.alpha, *nu_beta = (PetscReal *)ctx->SData_d.beta, *invMass = (PetscReal *)ctx->SData_d.invMass;
227: PetscReal(*lambdas)[LANDAU_MAX_GRIDS][LANDAU_MAX_GRIDS] = (PetscReal(*)[LANDAU_MAX_GRIDS][LANDAU_MAX_GRIDS])ctx->SData_d.lambdas;
228: PetscSection section[LANDAU_MAX_GRIDS], globsection[LANDAU_MAX_GRIDS];
229: PetscScalar *coo_vals = NULL;
230: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
231: PetscCall(DMGetLocalSection(ctx->plex[grid], §ion[grid]));
232: PetscCall(DMGetGlobalSection(ctx->plex[grid], &globsection[grid]));
233: PetscCall(PetscSectionGetNumFields(section[grid], &Nf[grid]));
234: }
235: /* count IPf size, etc */
236: PetscCall(PetscDSGetTabulation(prob, &Tf)); // Bf, &Df same for all grids
237: const PetscReal *const BB = Tf[0]->T[0], *const DD = Tf[0]->T[1];
238: ip_offset[0] = ipf_offset[0] = elem_offset[0] = 0;
239: for (PetscInt grid = 0; grid < num_grids; grid++) {
240: PetscInt nfloc = ctx->species_offset[grid + 1] - ctx->species_offset[grid];
241: elem_offset[grid + 1] = elem_offset[grid] + numCells[grid];
242: ip_offset[grid + 1] = ip_offset[grid] + numCells[grid] * Nq;
243: ipf_offset[grid + 1] = ipf_offset[grid] + Nq * nfloc * numCells[grid];
244: }
245: IPf_sz_glb = ipf_offset[num_grids];
246: IPf_sz_tot = IPf_sz_glb * ctx->batch_sz;
247: // prep COO
248: if (ctx->coo_assembly) {
249: PetscCall(PetscMalloc1(ctx->SData_d.coo_size, &coo_vals)); // allocate every time?
250: PetscCall(PetscInfo(ctx->plex[0], "COO Allocate %" PetscInt_FMT " values\n", (PetscInt)ctx->SData_d.coo_size));
251: }
252: if (shift == 0.0) { /* compute dynamic data f and df and init data for Jacobian */
253: #if defined(PETSC_HAVE_THREADSAFETY)
254: double starttime, endtime;
255: starttime = MPI_Wtime();
256: #endif
257: PetscCall(PetscLogEventBegin(ctx->events[8], 0, 0, 0, 0));
258: PetscCall(PetscMalloc4(IPf_sz_tot, &ff, IPf_sz_tot, &dudx, IPf_sz_tot, &dudy, dim == 3 ? IPf_sz_tot : 0, &dudz));
259: // F df/dx
260: for (PetscInt tid = 0; tid < ctx->batch_sz * elem_offset[num_grids]; tid++) { // for each element
261: const PetscInt b_Nelem = elem_offset[num_grids], b_elem_idx = tid % b_Nelem, b_id = tid / b_Nelem; // b_id == OMP thd_id in batch
262: // find my grid:
263: PetscInt grid = 0;
264: while (b_elem_idx >= elem_offset[grid + 1]) grid++; // yuck search for grid
265: {
266: const PetscInt loc_nip = numCells[grid] * Nq, loc_Nf = ctx->species_offset[grid + 1] - ctx->species_offset[grid], loc_elem = b_elem_idx - elem_offset[grid];
267: const PetscInt moffset = LAND_MOFFSET(b_id, grid, ctx->batch_sz, ctx->num_grids, ctx->mat_offset); //b_id*b_N + ctx->mat_offset[grid];
268: PetscScalar *coef, coef_buff[LANDAU_MAX_SPECIES * LANDAU_MAX_NQ];
269: PetscReal *invJe = &invJ_a[(ip_offset[grid] + loc_elem * Nq) * dim * dim]; // ingJ is static data on batch 0
270: PetscInt b, f, q;
271: if (cellClosure) {
272: coef = &cellClosure[b_id * IPf_sz_glb + ipf_offset[grid] + loc_elem * Nb * loc_Nf]; // this is const
273: } else {
274: coef = coef_buff;
275: for (f = 0; f < loc_Nf; ++f) {
276: LandauIdx *const Idxs = &maps[grid].gIdx[loc_elem][f][0];
277: for (b = 0; b < Nb; ++b) {
278: PetscInt idx = Idxs[b];
279: if (idx >= 0) {
280: coef[f * Nb + b] = xdata[idx + moffset];
281: } else {
282: idx = -idx - 1;
283: coef[f * Nb + b] = 0;
284: for (q = 0; q < maps[grid].num_face; q++) {
285: PetscInt id = maps[grid].c_maps[idx][q].gid;
286: PetscScalar scale = maps[grid].c_maps[idx][q].scale;
287: coef[f * Nb + b] += scale * xdata[id + moffset];
288: }
289: }
290: }
291: }
292: }
293: /* get f and df */
294: for (PetscInt qi = 0; qi < Nq; qi++) {
295: const PetscReal *invJ = &invJe[qi * dim * dim];
296: const PetscReal *Bq = &BB[qi * Nb];
297: const PetscReal *Dq = &DD[qi * Nb * dim];
298: PetscReal u_x[LANDAU_DIM];
299: /* get f & df */
300: for (f = 0; f < loc_Nf; ++f) {
301: const PetscInt idx = b_id * IPf_sz_glb + ipf_offset[grid] + f * loc_nip + loc_elem * Nq + qi;
302: PetscInt b, e;
303: PetscReal refSpaceDer[LANDAU_DIM];
304: ff[idx] = 0.0;
305: for (int d = 0; d < LANDAU_DIM; ++d) refSpaceDer[d] = 0.0;
306: for (b = 0; b < Nb; ++b) {
307: const PetscInt cidx = b;
308: ff[idx] += Bq[cidx] * PetscRealPart(coef[f * Nb + cidx]);
309: for (int d = 0; d < dim; ++d) refSpaceDer[d] += Dq[cidx * dim + d] * PetscRealPart(coef[f * Nb + cidx]);
310: }
311: for (int d = 0; d < LANDAU_DIM; ++d) {
312: for (e = 0, u_x[d] = 0.0; e < LANDAU_DIM; ++e) u_x[d] += invJ[e * dim + d] * refSpaceDer[e];
313: }
314: dudx[idx] = u_x[0];
315: dudy[idx] = u_x[1];
316: #if LANDAU_DIM == 3
317: dudz[idx] = u_x[2];
318: #endif
319: }
320: } // q
321: } // grid
322: } // grid*batch
323: PetscCall(PetscLogEventEnd(ctx->events[8], 0, 0, 0, 0));
324: #if defined(PETSC_HAVE_THREADSAFETY)
325: endtime = MPI_Wtime();
326: if (ctx->stage) ctx->times[LANDAU_F_DF] += (endtime - starttime);
327: #endif
328: } // Jacobian setup
329: // assemble Jacobian (or mass)
330: for (PetscInt tid = 0; tid < ctx->batch_sz * elem_offset[num_grids]; tid++) { // for each element
331: const PetscInt b_Nelem = elem_offset[num_grids];
332: const PetscInt glb_elem_idx = tid % b_Nelem, b_id = tid / b_Nelem;
333: PetscInt grid = 0;
334: #if defined(PETSC_HAVE_THREADSAFETY)
335: double starttime, endtime;
336: starttime = MPI_Wtime();
337: #endif
338: while (glb_elem_idx >= elem_offset[grid + 1]) grid++;
339: {
340: const PetscInt loc_Nf = ctx->species_offset[grid + 1] - ctx->species_offset[grid], loc_elem = glb_elem_idx - elem_offset[grid];
341: const PetscInt moffset = LAND_MOFFSET(b_id, grid, ctx->batch_sz, ctx->num_grids, ctx->mat_offset), totDim = loc_Nf * Nq, elemMatSize = totDim * totDim;
342: PetscScalar *elemMat;
343: const PetscReal *invJe = &invJ_a[(ip_offset[grid] + loc_elem * Nq) * dim * dim];
344: PetscCall(PetscMalloc1(elemMatSize, &elemMat));
345: PetscCall(PetscMemzero(elemMat, elemMatSize * sizeof(*elemMat)));
346: if (shift == 0.0) { // Jacobian
347: PetscCall(PetscLogEventBegin(ctx->events[4], 0, 0, 0, 0));
348: } else { // mass
349: PetscCall(PetscLogEventBegin(ctx->events[16], 0, 0, 0, 0));
350: }
351: for (PetscInt qj = 0; qj < Nq; ++qj) {
352: const PetscInt jpidx_glb = ip_offset[grid] + qj + loc_elem * Nq;
353: PetscReal g0[LANDAU_MAX_SPECIES], g2[LANDAU_MAX_SPECIES][LANDAU_DIM], g3[LANDAU_MAX_SPECIES][LANDAU_DIM][LANDAU_DIM]; // could make a LANDAU_MAX_SPECIES_GRID ~ number of ions - 1
354: PetscInt d, d2, dp, d3, IPf_idx;
355: if (shift == 0.0) { // Jacobian
356: const PetscReal *const invJj = &invJe[qj * dim * dim];
357: PetscReal gg2[LANDAU_MAX_SPECIES][LANDAU_DIM], gg3[LANDAU_MAX_SPECIES][LANDAU_DIM][LANDAU_DIM], gg2_temp[LANDAU_DIM], gg3_temp[LANDAU_DIM][LANDAU_DIM];
358: const PetscReal vj[3] = {xx[jpidx_glb], yy[jpidx_glb], zz ? zz[jpidx_glb] : 0}, wj = ww[jpidx_glb];
359: // create g2 & g3
360: for (d = 0; d < LANDAU_DIM; d++) { // clear accumulation data D & K
361: gg2_temp[d] = 0;
362: for (d2 = 0; d2 < LANDAU_DIM; d2++) gg3_temp[d][d2] = 0;
363: }
364: /* inner beta reduction */
365: IPf_idx = 0;
366: for (PetscInt grid_r = 0, f_off = 0, ipidx = 0; grid_r < ctx->num_grids; grid_r++, f_off = ctx->species_offset[grid_r]) { // IPf_idx += nip_loc_r*Nfloc_r
367: PetscInt nip_loc_r = numCells[grid_r] * Nq, Nfloc_r = Nf[grid_r];
368: for (PetscInt ei_r = 0, loc_fdf_idx = 0; ei_r < numCells[grid_r]; ++ei_r) {
369: for (PetscInt qi = 0; qi < Nq; qi++, ipidx++, loc_fdf_idx++) {
370: const PetscReal wi = ww[ipidx], x = xx[ipidx], y = yy[ipidx];
371: PetscReal temp1[3] = {0, 0, 0}, temp2 = 0;
372: #if LANDAU_DIM == 2
373: 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.;
374: LandauTensor2D(vj, x, y, Ud, Uk, mask);
375: #else
376: PetscReal U[3][3], z = zz[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.;
377: if (ctx->use_relativistic_corrections) {
378: LandauTensor3DRelativistic(vj, x, y, z, U, mask, C_0(ctx->v_0));
379: } else {
380: LandauTensor3D(vj, x, y, z, U, mask);
381: }
382: #endif
383: for (int f = 0; f < Nfloc_r; ++f) {
384: const PetscInt idx = b_id * IPf_sz_glb + ipf_offset[grid_r] + f * nip_loc_r + ei_r * Nq + qi; // IPf_idx + f*nip_loc_r + loc_fdf_idx;
385: temp1[0] += dudx[idx] * nu_beta[f + f_off] * invMass[f + f_off] * (*lambdas)[grid][grid_r];
386: temp1[1] += dudy[idx] * nu_beta[f + f_off] * invMass[f + f_off] * (*lambdas)[grid][grid_r];
387: #if LANDAU_DIM == 3
388: temp1[2] += dudz[idx] * nu_beta[f + f_off] * invMass[f + f_off] * (*lambdas)[grid][grid_r];
389: #endif
390: temp2 += ff[idx] * nu_beta[f + f_off] * (*lambdas)[grid][grid_r];
391: }
392: temp1[0] *= wi;
393: temp1[1] *= wi;
394: #if LANDAU_DIM == 3
395: temp1[2] *= wi;
396: #endif
397: temp2 *= wi;
398: #if LANDAU_DIM == 2
399: for (d2 = 0; d2 < 2; d2++) {
400: for (d3 = 0; d3 < 2; ++d3) {
401: /* K = U * grad(f): g2=e: i,A */
402: gg2_temp[d2] += Uk[d2][d3] * temp1[d3];
403: /* D = -U * (I \kron (fx)): g3=f: i,j,A */
404: gg3_temp[d2][d3] += Ud[d2][d3] * temp2;
405: }
406: }
407: #else
408: for (d2 = 0; d2 < 3; ++d2) {
409: for (d3 = 0; d3 < 3; ++d3) {
410: /* K = U * grad(f): g2 = e: i,A */
411: gg2_temp[d2] += U[d2][d3] * temp1[d3];
412: /* D = -U * (I \kron (fx)): g3 = f: i,j,A */
413: gg3_temp[d2][d3] += U[d2][d3] * temp2;
414: }
415: }
416: #endif
417: } // qi
418: } // ei_r
419: IPf_idx += nip_loc_r * Nfloc_r;
420: } /* grid_r - IPs */
421: PetscCheck(IPf_idx == IPf_sz_glb, PETSC_COMM_SELF, PETSC_ERR_PLIB, "IPf_idx != IPf_sz %" PetscInt_FMT " %" PetscInt_FMT, IPf_idx, IPf_sz_glb);
422: // add alpha and put in gg2/3
423: for (PetscInt fieldA = 0, f_off = ctx->species_offset[grid]; fieldA < loc_Nf; ++fieldA) {
424: for (d2 = 0; d2 < LANDAU_DIM; d2++) {
425: gg2[fieldA][d2] = gg2_temp[d2] * nu_alpha[fieldA + f_off];
426: for (d3 = 0; d3 < LANDAU_DIM; d3++) gg3[fieldA][d2][d3] = -gg3_temp[d2][d3] * nu_alpha[fieldA + f_off] * invMass[fieldA + f_off];
427: }
428: }
429: /* add electric field term once per IP */
430: for (PetscInt fieldA = 0, f_off = ctx->species_offset[grid]; fieldA < loc_Nf; ++fieldA) gg2[fieldA][LANDAU_DIM - 1] += Eq_m[fieldA + f_off];
431: /* Jacobian transform - g2, g3 */
432: for (PetscInt fieldA = 0; fieldA < loc_Nf; ++fieldA) {
433: for (d = 0; d < dim; ++d) {
434: g2[fieldA][d] = 0.0;
435: for (d2 = 0; d2 < dim; ++d2) {
436: g2[fieldA][d] += invJj[d * dim + d2] * gg2[fieldA][d2];
437: g3[fieldA][d][d2] = 0.0;
438: for (d3 = 0; d3 < dim; ++d3) {
439: for (dp = 0; dp < dim; ++dp) g3[fieldA][d][d2] += invJj[d * dim + d3] * gg3[fieldA][d3][dp] * invJj[d2 * dim + dp];
440: }
441: g3[fieldA][d][d2] *= wj;
442: }
443: g2[fieldA][d] *= wj;
444: }
445: }
446: } else { // mass
447: PetscReal wj = ww[jpidx_glb];
448: /* Jacobian transform - g0 */
449: for (PetscInt fieldA = 0; fieldA < loc_Nf; ++fieldA) {
450: if (dim == 2) {
451: g0[fieldA] = wj * shift * 2. * PETSC_PI; // move this to below and remove g0
452: } else {
453: g0[fieldA] = wj * shift; // move this to below and remove g0
454: }
455: }
456: }
457: /* FE matrix construction */
458: {
459: PetscInt fieldA, d, f, d2, g;
460: const PetscReal *BJq = &BB[qj * Nb], *DIq = &DD[qj * Nb * dim];
461: /* assemble - on the diagonal (I,I) */
462: for (fieldA = 0; fieldA < loc_Nf; fieldA++) {
463: for (f = 0; f < Nb; f++) {
464: const PetscInt i = fieldA * Nb + f; /* Element matrix row */
465: for (g = 0; g < Nb; ++g) {
466: const PetscInt j = fieldA * Nb + g; /* Element matrix column */
467: const PetscInt fOff = i * totDim + j;
468: if (shift == 0.0) {
469: for (d = 0; d < dim; ++d) {
470: elemMat[fOff] += DIq[f * dim + d] * g2[fieldA][d] * BJq[g];
471: for (d2 = 0; d2 < dim; ++d2) elemMat[fOff] += DIq[f * dim + d] * g3[fieldA][d][d2] * DIq[g * dim + d2];
472: }
473: } else { // mass
474: elemMat[fOff] += BJq[f] * g0[fieldA] * BJq[g];
475: }
476: }
477: }
478: }
479: }
480: } /* qj loop */
481: if (shift == 0.0) { // Jacobian
482: PetscCall(PetscLogEventEnd(ctx->events[4], 0, 0, 0, 0));
483: } else {
484: PetscCall(PetscLogEventEnd(ctx->events[16], 0, 0, 0, 0));
485: }
486: #if defined(PETSC_HAVE_THREADSAFETY)
487: endtime = MPI_Wtime();
488: if (ctx->stage) ctx->times[LANDAU_KERNEL] += (endtime - starttime);
489: #endif
490: /* assemble matrix */
491: if (!container) {
492: PetscInt cStart;
493: PetscCall(PetscLogEventBegin(ctx->events[6], 0, 0, 0, 0));
494: PetscCall(DMPlexGetHeightStratum(ctx->plex[grid], 0, &cStart, NULL));
495: PetscCall(DMPlexMatSetClosure(ctx->plex[grid], section[grid], globsection[grid], subJ[LAND_PACK_IDX(b_id, grid)], loc_elem + cStart, elemMat, ADD_VALUES));
496: PetscCall(PetscLogEventEnd(ctx->events[6], 0, 0, 0, 0));
497: } else { // GPU like assembly for debugging
498: PetscInt fieldA, q, f, g, d, nr, nc, rows0[LANDAU_MAX_Q_FACE] = {0}, cols0[LANDAU_MAX_Q_FACE] = {0}, rows[LANDAU_MAX_Q_FACE], cols[LANDAU_MAX_Q_FACE];
499: PetscScalar vals[LANDAU_MAX_Q_FACE * LANDAU_MAX_Q_FACE] = {0}, row_scale[LANDAU_MAX_Q_FACE] = {0}, col_scale[LANDAU_MAX_Q_FACE] = {0};
500: LandauIdx *coo_elem_offsets = (LandauIdx *)ctx->SData_d.coo_elem_offsets, *coo_elem_fullNb = (LandauIdx *)ctx->SData_d.coo_elem_fullNb, (*coo_elem_point_offsets)[LANDAU_MAX_NQ + 1] = (LandauIdx(*)[LANDAU_MAX_NQ + 1]) ctx->SData_d.coo_elem_point_offsets;
501: /* assemble - from the diagonal (I,I) in this format for DMPlexMatSetClosure */
502: for (fieldA = 0; fieldA < loc_Nf; fieldA++) {
503: LandauIdx *const Idxs = &maps[grid].gIdx[loc_elem][fieldA][0];
504: for (f = 0; f < Nb; f++) {
505: PetscInt idx = Idxs[f];
506: if (idx >= 0) {
507: nr = 1;
508: rows0[0] = idx;
509: row_scale[0] = 1.;
510: } else {
511: idx = -idx - 1;
512: for (q = 0, nr = 0; q < maps[grid].num_face; q++, nr++) {
513: if (maps[grid].c_maps[idx][q].gid < 0) break;
514: rows0[q] = maps[grid].c_maps[idx][q].gid;
515: row_scale[q] = maps[grid].c_maps[idx][q].scale;
516: }
517: }
518: for (g = 0; g < Nb; ++g) {
519: idx = Idxs[g];
520: if (idx >= 0) {
521: nc = 1;
522: cols0[0] = idx;
523: col_scale[0] = 1.;
524: } else {
525: idx = -idx - 1;
526: nc = maps[grid].num_face;
527: for (q = 0, nc = 0; q < maps[grid].num_face; q++, nc++) {
528: if (maps[grid].c_maps[idx][q].gid < 0) break;
529: cols0[q] = maps[grid].c_maps[idx][q].gid;
530: col_scale[q] = maps[grid].c_maps[idx][q].scale;
531: }
532: }
533: const PetscInt i = fieldA * Nb + f; /* Element matrix row */
534: const PetscInt j = fieldA * Nb + g; /* Element matrix column */
535: const PetscScalar Aij = elemMat[i * totDim + j];
536: if (coo_vals) { // mirror (i,j) in CreateStaticGPUData
537: const int fullNb = coo_elem_fullNb[glb_elem_idx], fullNb2 = fullNb * fullNb;
538: const int idx0 = b_id * coo_elem_offsets[elem_offset[num_grids]] + coo_elem_offsets[glb_elem_idx] + fieldA * fullNb2 + fullNb * coo_elem_point_offsets[glb_elem_idx][f] + nr * coo_elem_point_offsets[glb_elem_idx][g];
539: for (int q = 0, idx2 = idx0; q < nr; q++) {
540: for (int d = 0; d < nc; d++, idx2++) coo_vals[idx2] = row_scale[q] * col_scale[d] * Aij;
541: }
542: } else {
543: for (q = 0; q < nr; q++) rows[q] = rows0[q] + moffset;
544: for (d = 0; d < nc; d++) cols[d] = cols0[d] + moffset;
545: for (q = 0; q < nr; q++) {
546: for (d = 0; d < nc; d++) vals[q * nc + d] = row_scale[q] * col_scale[d] * Aij;
547: }
548: PetscCall(MatSetValues(JacP, nr, rows, nc, cols, vals, ADD_VALUES));
549: }
550: }
551: }
552: }
553: }
554: if (loc_elem == -1) {
555: PetscCall(PetscPrintf(ctx->comm, "CPU Element matrix\n"));
556: for (int d = 0; d < totDim; ++d) {
557: for (int f = 0; f < totDim; ++f) PetscCall(PetscPrintf(ctx->comm, " %12.5e", (double)PetscRealPart(elemMat[d * totDim + f])));
558: PetscCall(PetscPrintf(ctx->comm, "\n"));
559: }
560: exit(12);
561: }
562: PetscCall(PetscFree(elemMat));
563: } /* grid */
564: } /* outer element & batch loop */
565: if (shift == 0.0) { // mass
566: PetscCall(PetscFree4(ff, dudx, dudy, dudz));
567: }
568: if (!container) { // 'CPU' assembly move nest matrix to global JacP
569: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) { // OpenMP
570: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
571: const PetscInt moffset = LAND_MOFFSET(b_id, grid, ctx->batch_sz, ctx->num_grids, ctx->mat_offset); // b_id*b_N + ctx->mat_offset[grid];
572: PetscInt nloc, nzl, colbuf[1024], row;
573: const PetscInt *cols;
574: const PetscScalar *vals;
575: Mat B = subJ[LAND_PACK_IDX(b_id, grid)];
576: PetscCall(MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY));
577: PetscCall(MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY));
578: PetscCall(MatGetSize(B, &nloc, NULL));
579: for (int i = 0; i < nloc; i++) {
580: PetscCall(MatGetRow(B, i, &nzl, &cols, &vals));
581: PetscCheck(nzl <= 1024, PetscObjectComm((PetscObject)B), PETSC_ERR_PLIB, "Row too big: %" PetscInt_FMT, nzl);
582: for (int j = 0; j < nzl; j++) colbuf[j] = moffset + cols[j];
583: row = moffset + i;
584: PetscCall(MatSetValues(JacP, 1, &row, nzl, colbuf, vals, ADD_VALUES));
585: PetscCall(MatRestoreRow(B, i, &nzl, &cols, &vals));
586: }
587: PetscCall(MatDestroy(&B));
588: }
589: }
590: }
591: if (coo_vals) {
592: PetscCall(MatSetValuesCOO(JacP, coo_vals, ADD_VALUES));
593: PetscCall(PetscFree(coo_vals));
594: }
595: } /* CPU version */
596: PetscCall(MatAssemblyBegin(JacP, MAT_FINAL_ASSEMBLY));
597: PetscCall(MatAssemblyEnd(JacP, MAT_FINAL_ASSEMBLY));
598: /* clean up */
599: if (cellClosure) PetscCall(PetscFree(cellClosure));
600: if (xdata) PetscCall(VecRestoreArrayReadAndMemType(a_X, &xdata));
601: PetscFunctionReturn(PETSC_SUCCESS);
602: }
604: static PetscErrorCode GeometryDMLandau(DM base, PetscInt point, PetscInt dim, const PetscReal abc[], PetscReal xyz[], void *a_ctx)
605: {
606: PetscReal r = abc[0], z = abc[1];
608: PetscFunctionBegin;
609: xyz[0] = r;
610: xyz[1] = z;
611: if (dim == 3) xyz[2] = abc[2];
613: PetscFunctionReturn(PETSC_SUCCESS);
614: }
616: /* create DMComposite of meshes for each species group */
617: static PetscErrorCode LandauDMCreateVMeshes(MPI_Comm comm_self, const PetscInt dim, const char prefix[], LandauCtx *ctx, DM pack)
618: {
619: PetscFunctionBegin;
620: { /* p4est, quads */
621: /* Create plex mesh of Landau domain */
622: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
623: PetscReal par_radius = ctx->radius_par[grid], perp_radius = ctx->radius_perp[grid];
624: if (!ctx->sphere) {
625: PetscReal lo[] = {-perp_radius, -par_radius, -par_radius}, hi[] = {perp_radius, par_radius, par_radius};
626: DMBoundaryType periodicity[3] = {DM_BOUNDARY_NONE, dim == 2 ? DM_BOUNDARY_NONE : DM_BOUNDARY_NONE, DM_BOUNDARY_NONE};
627: if (dim == 2) lo[0] = 0;
628: else {
629: lo[1] = -perp_radius;
630: hi[1] = perp_radius; // 3D y is a perp
631: }
632: PetscCall(DMPlexCreateBoxMesh(comm_self, dim, PETSC_FALSE, ctx->cells0, lo, hi, periodicity, PETSC_TRUE, &ctx->plex[grid])); // todo: make composite and create dm[grid] here
633: PetscCall(DMLocalizeCoordinates(ctx->plex[grid])); /* needed for periodic */
634: if (dim == 3) PetscCall(PetscObjectSetName((PetscObject)ctx->plex[grid], "cube"));
635: else PetscCall(PetscObjectSetName((PetscObject)ctx->plex[grid], "half-plane"));
636: } else if (dim == 2) {
637: PetscInt numCells = 6, cells[11][4], i, j;
638: PetscInt numVerts = 11;
639: PetscReal *flatCoords = NULL;
640: PetscInt *flatCells = NULL, *pcell;
641: int cells2[][4] = {
642: {0, 1, 6, 5 },
643: {1, 2, 7, 6 },
644: {2, 3, 8, 7 },
645: {3, 4, 9, 8 },
646: {5, 6, 7, 10},
647: {10, 7, 8, 9 },
648: };
649: for (i = 0; i < numCells; i++)
650: for (j = 0; j < 4; j++) cells[i][j] = cells2[i][j];
651: PetscCall(PetscMalloc2(numVerts * 2, &flatCoords, numCells * 4, &flatCells));
652: {
653: PetscReal(*coords)[2] = (PetscReal(*)[2])flatCoords;
654: PetscReal rad = ctx->radius[grid];
655: for (j = 0; j < 5; j++) {
656: PetscReal z, r, theta = -PETSC_PI / 2 + (j % 5) * PETSC_PI / 4;
657: r = rad * PetscCosReal(theta);
658: coords[j][0] = r;
659: z = rad * PetscSinReal(theta);
660: coords[j][1] = z;
661: }
662: coords[j][0] = 0;
663: coords[j++][1] = -rad * ctx->sphere_inner_radius_90degree;
664: coords[j][0] = rad * ctx->sphere_inner_radius_45degree;
665: coords[j++][1] = -rad * ctx->sphere_inner_radius_45degree;
666: coords[j][0] = rad * ctx->sphere_inner_radius_90degree;
667: coords[j++][1] = 0;
668: coords[j][0] = rad * ctx->sphere_inner_radius_45degree;
669: coords[j++][1] = rad * ctx->sphere_inner_radius_45degree;
670: coords[j][0] = 0;
671: coords[j++][1] = rad * ctx->sphere_inner_radius_90degree;
672: coords[j][0] = 0;
673: coords[j++][1] = 0;
674: }
675: for (j = 0, pcell = flatCells; j < numCells; j++, pcell += 4) {
676: for (int jj = 0; jj < 4; jj++) pcell[jj] = cells[j][jj];
677: }
678: PetscCall(DMPlexCreateFromCellListPetsc(comm_self, 2, numCells, numVerts, 4, ctx->interpolate, flatCells, 2, flatCoords, &ctx->plex[grid]));
679: PetscCall(PetscFree2(flatCoords, flatCells));
680: PetscCall(PetscObjectSetName((PetscObject)ctx->plex[grid], "semi-circle"));
681: } else SETERRQ(ctx->comm, PETSC_ERR_PLIB, "Velocity space meshes does not support 3V cubed sphere");
682: PetscCall(DMSetFromOptions(ctx->plex[grid]));
683: } // grid loop
684: PetscCall(PetscObjectSetOptionsPrefix((PetscObject)pack, prefix));
686: { /* convert to p4est (or whatever), wait for discretization to create pack */
687: char convType[256];
688: PetscBool flg;
690: PetscOptionsBegin(ctx->comm, prefix, "Mesh conversion options", "DMPLEX");
691: PetscCall(PetscOptionsFList("-dm_landau_type", "Convert DMPlex to another format (p4est)", "plexland.c", DMList, DMPLEX, convType, 256, &flg));
692: PetscOptionsEnd();
693: if (flg) {
694: ctx->use_p4est = PETSC_TRUE; /* flag for Forest */
695: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
696: DM dmforest;
697: PetscCall(DMConvert(ctx->plex[grid], convType, &dmforest));
698: if (dmforest) {
699: PetscBool isForest;
700: PetscCall(PetscObjectSetOptionsPrefix((PetscObject)dmforest, prefix));
701: PetscCall(DMIsForest(dmforest, &isForest));
702: if (isForest) {
703: if (ctx->sphere) PetscCall(DMForestSetBaseCoordinateMapping(dmforest, GeometryDMLandau, ctx));
704: PetscCall(DMDestroy(&ctx->plex[grid]));
705: ctx->plex[grid] = dmforest; // Forest for adaptivity
706: } else SETERRQ(ctx->comm, PETSC_ERR_PLIB, "Converted to non Forest?");
707: } else SETERRQ(ctx->comm, PETSC_ERR_PLIB, "Convert failed?");
708: }
709: } else ctx->use_p4est = PETSC_FALSE; /* flag for Forest */
710: }
711: } /* non-file */
712: PetscCall(DMSetDimension(pack, dim));
713: PetscCall(PetscObjectSetName((PetscObject)pack, "Mesh"));
714: PetscCall(DMSetApplicationContext(pack, ctx));
716: PetscFunctionReturn(PETSC_SUCCESS);
717: }
719: static PetscErrorCode SetupDS(DM pack, PetscInt dim, PetscInt grid, LandauCtx *ctx)
720: {
721: PetscInt ii, i0;
722: char buf[256];
723: PetscSection section;
725: PetscFunctionBegin;
726: for (ii = ctx->species_offset[grid], i0 = 0; ii < ctx->species_offset[grid + 1]; ii++, i0++) {
727: if (ii == 0) PetscCall(PetscSNPrintf(buf, sizeof(buf), "e"));
728: else PetscCall(PetscSNPrintf(buf, sizeof(buf), "i%" PetscInt_FMT, ii));
729: /* Setup Discretization - FEM */
730: PetscCall(PetscFECreateDefault(PETSC_COMM_SELF, dim, 1, PETSC_FALSE, NULL, PETSC_DECIDE, &ctx->fe[ii]));
731: PetscCall(PetscObjectSetName((PetscObject)ctx->fe[ii], buf));
732: PetscCall(DMSetField(ctx->plex[grid], i0, NULL, (PetscObject)ctx->fe[ii]));
733: }
734: PetscCall(DMCreateDS(ctx->plex[grid]));
735: PetscCall(DMGetSection(ctx->plex[grid], §ion));
736: for (PetscInt ii = ctx->species_offset[grid], i0 = 0; ii < ctx->species_offset[grid + 1]; ii++, i0++) {
737: if (ii == 0) PetscCall(PetscSNPrintf(buf, sizeof(buf), "se"));
738: else PetscCall(PetscSNPrintf(buf, sizeof(buf), "si%" PetscInt_FMT, ii));
739: PetscCall(PetscSectionSetComponentName(section, i0, 0, buf));
740: }
741: PetscFunctionReturn(PETSC_SUCCESS);
742: }
744: /* Define a Maxwellian function for testing out the operator. */
746: /* Using cartesian velocity space coordinates, the particle */
747: /* density, [1/m^3], is defined according to */
749: /* $$ n=\int_{R^3} dv^3 \left(\frac{m}{2\pi T}\right)^{3/2}\exp [- mv^2/(2T)] $$ */
751: /* Using some constant, c, we normalize the velocity vector into a */
752: /* dimensionless variable according to v=c*x. Thus the density, $n$, becomes */
754: /* $$ n=\int_{R^3} dx^3 \left(\frac{mc^2}{2\pi T}\right)^{3/2}\exp [- mc^2/(2T)*x^2] $$ */
756: /* Defining $\theta=2T/mc^2$, we thus find that the probability density */
757: /* for finding the particle within the interval in a box dx^3 around x is */
759: /* f(x;\theta)=\left(\frac{1}{\pi\theta}\right)^{3/2} \exp [ -x^2/\theta ] */
761: typedef struct {
762: PetscReal v_0;
763: PetscReal kT_m;
764: PetscReal n;
765: PetscReal shift;
766: } MaxwellianCtx;
768: static PetscErrorCode maxwellian(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nf_dummy, PetscScalar *u, void *actx)
769: {
770: MaxwellianCtx *mctx = (MaxwellianCtx *)actx;
771: PetscInt i;
772: PetscReal v2 = 0, theta = 2 * mctx->kT_m / (mctx->v_0 * mctx->v_0), shift; /* theta = 2kT/mc^2 */
773: PetscFunctionBegin;
774: /* compute the exponents, v^2 */
775: for (i = 0; i < dim; ++i) v2 += x[i] * x[i];
776: /* evaluate the Maxwellian */
777: if (mctx->shift < 0) shift = -mctx->shift;
778: else {
779: u[0] = mctx->n * PetscPowReal(PETSC_PI * theta, -1.5) * (PetscExpReal(-v2 / theta));
780: shift = mctx->shift;
781: }
782: if (shift != 0.) {
783: v2 = 0;
784: for (i = 0; i < dim - 1; ++i) v2 += x[i] * x[i];
785: v2 += (x[dim - 1] - shift) * (x[dim - 1] - shift);
786: /* evaluate the shifted Maxwellian */
787: u[0] += mctx->n * PetscPowReal(PETSC_PI * theta, -1.5) * (PetscExpReal(-v2 / theta));
788: }
789: PetscFunctionReturn(PETSC_SUCCESS);
790: }
792: /*@
793: DMPlexLandauAddMaxwellians - Add a Maxwellian distribution to a state
795: Collective
797: Input Parameters:
798: . dm - The mesh (local)
799: + time - Current time
800: - temps - Temperatures of each species (global)
801: . ns - Number density of each species (global)
802: - grid - index into current grid - just used for offset into temp and ns
803: . b_id - batch index
804: - n_batch - number of batches
805: + actx - Landau context
807: Output Parameter:
808: . X - The state (local to this grid)
810: Level: beginner
812: .keywords: mesh
813: .seealso: `DMPlexLandauCreateVelocitySpace()`
814: @*/
815: PetscErrorCode DMPlexLandauAddMaxwellians(DM dm, Vec X, PetscReal time, PetscReal temps[], PetscReal ns[], PetscInt grid, PetscInt b_id, PetscInt n_batch, void *actx)
816: {
817: LandauCtx *ctx = (LandauCtx *)actx;
818: PetscErrorCode (*initu[LANDAU_MAX_SPECIES])(PetscInt, PetscReal, const PetscReal[], PetscInt, PetscScalar[], void *);
819: PetscInt dim;
820: MaxwellianCtx *mctxs[LANDAU_MAX_SPECIES], data[LANDAU_MAX_SPECIES];
822: PetscFunctionBegin;
823: PetscCall(DMGetDimension(dm, &dim));
824: if (!ctx) PetscCall(DMGetApplicationContext(dm, &ctx));
825: for (PetscInt ii = ctx->species_offset[grid], i0 = 0; ii < ctx->species_offset[grid + 1]; ii++, i0++) {
826: mctxs[i0] = &data[i0];
827: data[i0].v_0 = ctx->v_0; // v_0 same for all grids
828: data[i0].kT_m = ctx->k * temps[ii] / ctx->masses[ii]; /* kT/m */
829: data[i0].n = ns[ii];
830: initu[i0] = maxwellian;
831: data[i0].shift = 0;
832: }
833: data[0].shift = ctx->electronShift;
834: /* need to make ADD_ALL_VALUES work - TODO */
835: PetscCall(DMProjectFunction(dm, time, initu, (void **)mctxs, INSERT_ALL_VALUES, X));
836: PetscFunctionReturn(PETSC_SUCCESS);
837: }
839: /*
840: LandauSetInitialCondition - Adds Maxwellians with context
842: Collective
844: Input Parameters:
845: . dm - The mesh
846: - grid - index into current grid - just used for offset into temp and ns
847: . b_id - batch index
848: - n_batch - number of batches
849: + actx - Landau context with T and n
851: Output Parameter:
852: . X - The state
854: Level: beginner
856: .keywords: mesh
857: .seealso: `DMPlexLandauCreateVelocitySpace()`, `DMPlexLandauAddMaxwellians()`
858: */
859: static PetscErrorCode LandauSetInitialCondition(DM dm, Vec X, PetscInt grid, PetscInt b_id, PetscInt n_batch, void *actx)
860: {
861: LandauCtx *ctx = (LandauCtx *)actx;
862: PetscFunctionBegin;
863: if (!ctx) PetscCall(DMGetApplicationContext(dm, &ctx));
864: PetscCall(VecZeroEntries(X));
865: PetscCall(DMPlexLandauAddMaxwellians(dm, X, 0.0, ctx->thermal_temps, ctx->n, grid, b_id, n_batch, ctx));
866: PetscFunctionReturn(PETSC_SUCCESS);
867: }
869: // adapt a level once. Forest in/out
870: #if defined(PETSC_USE_INFO)
871: static const char *s_refine_names[] = {"RE", "Z1", "Origin", "Z2", "Uniform"};
872: #endif
873: static PetscErrorCode adaptToleranceFEM(PetscFE fem, Vec sol, PetscInt type, PetscInt grid, LandauCtx *ctx, DM *newForest)
874: {
875: DM forest, plex, adaptedDM = NULL;
876: PetscDS prob;
877: PetscBool isForest;
878: PetscQuadrature quad;
879: PetscInt Nq, *Nb, cStart, cEnd, c, dim, qj, k;
880: DMLabel adaptLabel = NULL;
882: PetscFunctionBegin;
883: forest = ctx->plex[grid];
884: PetscCall(DMCreateDS(forest));
885: PetscCall(DMGetDS(forest, &prob));
886: PetscCall(DMGetDimension(forest, &dim));
887: PetscCall(DMIsForest(forest, &isForest));
888: PetscCheck(isForest, ctx->comm, PETSC_ERR_ARG_WRONG, "! Forest");
889: PetscCall(DMConvert(forest, DMPLEX, &plex));
890: PetscCall(DMPlexGetHeightStratum(plex, 0, &cStart, &cEnd));
891: PetscCall(DMLabelCreate(PETSC_COMM_SELF, "adapt", &adaptLabel));
892: PetscCall(PetscFEGetQuadrature(fem, &quad));
893: PetscCall(PetscQuadratureGetData(quad, NULL, NULL, &Nq, NULL, NULL));
894: PetscCheck(Nq <= LANDAU_MAX_NQ, ctx->comm, PETSC_ERR_ARG_WRONG, "Order too high. Nq = %" PetscInt_FMT " > LANDAU_MAX_NQ (%d)", Nq, LANDAU_MAX_NQ);
895: PetscCall(PetscDSGetDimensions(prob, &Nb));
896: PetscCall(PetscInfo(sol, "%" PetscInt_FMT ") Refine phase: %s\n", grid, s_refine_names[type]));
897: if (type == 4) {
898: for (c = cStart; c < cEnd; c++) PetscCall(DMLabelSetValue(adaptLabel, c, DM_ADAPT_REFINE));
899: } else if (type == 2) {
900: PetscInt rCellIdx[8], nr = 0, nrmax = (dim == 3) ? 8 : 2;
901: PetscReal minRad = PETSC_INFINITY, r;
902: for (c = cStart; c < cEnd; c++) {
903: PetscReal tt, v0[LANDAU_MAX_NQ * 3], detJ[LANDAU_MAX_NQ];
904: PetscCall(DMPlexComputeCellGeometryFEM(plex, c, quad, v0, NULL, NULL, detJ));
905: for (qj = 0; qj < Nq; ++qj) {
906: tt = PetscSqr(v0[dim * qj + 0]) + PetscSqr(v0[dim * qj + 1]) + PetscSqr(((dim == 3) ? v0[dim * qj + 2] : 0));
907: r = PetscSqrtReal(tt);
908: if (r < minRad - PETSC_SQRT_MACHINE_EPSILON * 10.) {
909: minRad = r;
910: nr = 0;
911: rCellIdx[nr++] = c;
912: PetscCall(PetscInfo(sol, "\t\t%" PetscInt_FMT ") Found first inner r=%e, cell %" PetscInt_FMT ", qp %" PetscInt_FMT "/%" PetscInt_FMT "\n", grid, (double)r, c, qj + 1, Nq));
913: } else if ((r - minRad) < PETSC_SQRT_MACHINE_EPSILON * 100. && nr < nrmax) {
914: for (k = 0; k < nr; k++)
915: if (c == rCellIdx[k]) break;
916: if (k == nr) {
917: rCellIdx[nr++] = c;
918: PetscCall(PetscInfo(sol, "\t\t\t%" PetscInt_FMT ") Found another inner r=%e, cell %" PetscInt_FMT ", qp %" PetscInt_FMT "/%" PetscInt_FMT ", d=%e\n", grid, (double)r, c, qj + 1, Nq, (double)(r - minRad)));
919: }
920: }
921: }
922: }
923: for (k = 0; k < nr; k++) PetscCall(DMLabelSetValue(adaptLabel, rCellIdx[k], DM_ADAPT_REFINE));
924: PetscCall(PetscInfo(sol, "\t\t\t%" PetscInt_FMT ") Refined %" PetscInt_FMT " origin cells %" PetscInt_FMT ",%" PetscInt_FMT " r=%g\n", grid, nr, rCellIdx[0], rCellIdx[1], (double)minRad));
925: } else if (type == 0 || type == 1 || type == 3) { /* refine along r=0 axis */
926: PetscScalar *coef = NULL;
927: Vec coords;
928: PetscInt csize, Nv, d, nz, nrefined = 0;
929: DM cdm;
930: PetscSection cs;
931: PetscCall(DMGetCoordinatesLocal(forest, &coords));
932: PetscCall(DMGetCoordinateDM(forest, &cdm));
933: PetscCall(DMGetLocalSection(cdm, &cs));
934: for (c = cStart; c < cEnd; c++) {
935: PetscInt doit = 0, outside = 0;
936: PetscCall(DMPlexVecGetClosure(cdm, cs, coords, c, &csize, &coef));
937: Nv = csize / dim;
938: for (nz = d = 0; d < Nv; d++) {
939: PetscReal z = PetscRealPart(coef[d * dim + (dim - 1)]), x = PetscSqr(PetscRealPart(coef[d * dim + 0])) + ((dim == 3) ? PetscSqr(PetscRealPart(coef[d * dim + 1])) : 0);
940: x = PetscSqrtReal(x);
941: if (type == 0) {
942: if (ctx->re_radius > PETSC_SQRT_MACHINE_EPSILON && (z < -PETSC_MACHINE_EPSILON * 10. || z > ctx->re_radius + PETSC_MACHINE_EPSILON * 10.)) outside++; /* first pass don't refine bottom */
943: } else if (type == 1 && (z > ctx->vperp0_radius1 || z < -ctx->vperp0_radius1)) {
944: outside++; /* don't refine outside electron refine radius */
945: PetscCall(PetscInfo(sol, "\t%" PetscInt_FMT ") (debug) found %s cells\n", grid, s_refine_names[type]));
946: } else if (type == 3 && (z > ctx->vperp0_radius2 || z < -ctx->vperp0_radius2)) {
947: outside++; /* refine r=0 cells on refinement front */
948: PetscCall(PetscInfo(sol, "\t%" PetscInt_FMT ") (debug) found %s cells\n", grid, s_refine_names[type]));
949: }
950: if (x < PETSC_MACHINE_EPSILON * 10. && (type != 0 || ctx->re_radius > PETSC_SQRT_MACHINE_EPSILON)) nz++;
951: }
952: PetscCall(DMPlexVecRestoreClosure(cdm, cs, coords, c, &csize, &coef));
953: if (doit || (outside < Nv && nz)) {
954: PetscCall(DMLabelSetValue(adaptLabel, c, DM_ADAPT_REFINE));
955: nrefined++;
956: }
957: }
958: PetscCall(PetscInfo(sol, "\t%" PetscInt_FMT ") Refined %" PetscInt_FMT " cells\n", grid, nrefined));
959: }
960: PetscCall(DMDestroy(&plex));
961: PetscCall(DMAdaptLabel(forest, adaptLabel, &adaptedDM));
962: PetscCall(DMLabelDestroy(&adaptLabel));
963: *newForest = adaptedDM;
964: if (adaptedDM) {
965: if (isForest) {
966: PetscCall(DMForestSetAdaptivityForest(adaptedDM, NULL)); // ????
967: }
968: PetscCall(DMConvert(adaptedDM, DMPLEX, &plex));
969: PetscCall(DMPlexGetHeightStratum(plex, 0, &cStart, &cEnd));
970: PetscCall(PetscInfo(sol, "\t\t\t\t%" PetscInt_FMT ") %" PetscInt_FMT " cells, %" PetscInt_FMT " total quadrature points\n", grid, cEnd - cStart, Nq * (cEnd - cStart)));
971: PetscCall(DMDestroy(&plex));
972: } else *newForest = NULL;
973: PetscFunctionReturn(PETSC_SUCCESS);
974: }
976: // forest goes in (ctx->plex[grid]), plex comes out
977: static PetscErrorCode adapt(PetscInt grid, LandauCtx *ctx, Vec *uu)
978: {
979: PetscInt adaptIter;
981: PetscFunctionBegin;
982: PetscInt type, limits[5] = {(grid == 0) ? ctx->numRERefine : 0, (grid == 0) ? ctx->nZRefine1 : 0, ctx->numAMRRefine[grid], (grid == 0) ? ctx->nZRefine2 : 0, ctx->postAMRRefine[grid]};
983: for (type = 0; type < 5; type++) {
984: for (adaptIter = 0; adaptIter < limits[type]; adaptIter++) {
985: DM newForest = NULL;
986: PetscCall(adaptToleranceFEM(ctx->fe[0], *uu, type, grid, ctx, &newForest));
987: if (newForest) {
988: PetscCall(DMDestroy(&ctx->plex[grid]));
989: PetscCall(VecDestroy(uu));
990: PetscCall(DMCreateGlobalVector(newForest, uu));
991: PetscCall(PetscObjectSetName((PetscObject)*uu, "uAMR"));
992: PetscCall(LandauSetInitialCondition(newForest, *uu, grid, 0, 1, ctx));
993: ctx->plex[grid] = newForest;
994: } else {
995: PetscCall(PetscInfo(*uu, "No refinement\n"));
996: }
997: }
998: }
999: PetscFunctionReturn(PETSC_SUCCESS);
1000: }
1002: // make log(Lambdas) from NRL Plasma formulary
1003: static PetscErrorCode makeLambdas(LandauCtx *ctx)
1004: {
1005: PetscFunctionBegin;
1006: for (PetscInt gridi = 0; gridi < ctx->num_grids; gridi++) {
1007: int iii = ctx->species_offset[gridi];
1008: PetscReal Ti_ev = (ctx->thermal_temps[iii] / 1.1604525e7) * 1000; // convert (back) to eV
1009: PetscReal ni = ctx->n[iii] * ctx->n_0;
1010: for (PetscInt gridj = gridi; gridj < ctx->num_grids; gridj++) {
1011: PetscInt jjj = ctx->species_offset[gridj];
1012: PetscReal Zj = ctx->charges[jjj] / 1.6022e-19;
1013: if (gridi == 0) {
1014: if (gridj == 0) { // lam_ee
1015: ctx->lambdas[gridi][gridj] = 23.5 - PetscLogReal(PetscSqrtReal(ni) * PetscPowReal(Ti_ev, -1.25)) - PetscSqrtReal(1e-5 + PetscSqr(PetscLogReal(Ti_ev) - 2) / 16);
1016: } else { // lam_ei == lam_ie
1017: if (10 * Zj * Zj > Ti_ev) {
1018: ctx->lambdas[gridi][gridj] = ctx->lambdas[gridj][gridi] = 23 - PetscLogReal(PetscSqrtReal(ni) * Zj * PetscPowReal(Ti_ev, -1.5));
1019: } else {
1020: ctx->lambdas[gridi][gridj] = ctx->lambdas[gridj][gridi] = 24 - PetscLogReal(PetscSqrtReal(ni) / Ti_ev);
1021: }
1022: }
1023: } else { // lam_ii'
1024: PetscReal mui = ctx->masses[iii] / 1.6720e-27, Zi = ctx->charges[iii] / 1.6022e-19;
1025: PetscReal Tj_ev = (ctx->thermal_temps[jjj] / 1.1604525e7) * 1000; // convert (back) to eV
1026: PetscReal muj = ctx->masses[jjj] / 1.6720e-27;
1027: PetscReal nj = ctx->n[jjj] * ctx->n_0;
1028: ctx->lambdas[gridi][gridj] = ctx->lambdas[gridj][gridi] = 23 - PetscLogReal(Zi * Zj * (mui + muj) / (mui * Tj_ev + muj * Ti_ev) * PetscSqrtReal(ni * Zi * Zi / Ti_ev + nj * Zj * Zj / Tj_ev));
1029: }
1030: }
1031: }
1032: //PetscReal v0 = PetscSqrtReal(ctx->k * ctx->thermal_temps[iii] / ctx->masses[iii]); /* arbitrary units for non-dimensionalization: plasma formulary def */
1033: PetscFunctionReturn(PETSC_SUCCESS);
1034: }
1036: static PetscErrorCode ProcessOptions(LandauCtx *ctx, const char prefix[])
1037: {
1038: PetscBool flg;
1039: PetscInt ii, nt, nm, nc, num_species_grid[LANDAU_MAX_GRIDS], non_dim_grid;
1040: PetscReal v0_grid[LANDAU_MAX_GRIDS], lnLam = 10;
1041: DM dummy;
1043: PetscFunctionBegin;
1044: PetscCall(DMCreate(ctx->comm, &dummy));
1045: /* get options - initialize context */
1046: ctx->verbose = 1; // should be 0 for silent compliance
1047: #if defined(PETSC_HAVE_THREADSAFETY) && defined(PETSC_HAVE_OPENMP)
1048: ctx->batch_sz = PetscNumOMPThreads;
1049: #else
1050: ctx->batch_sz = 1;
1051: #endif
1052: ctx->batch_view_idx = 0;
1053: ctx->interpolate = PETSC_TRUE;
1054: ctx->gpu_assembly = PETSC_TRUE;
1055: ctx->norm_state = 0;
1056: ctx->electronShift = 0;
1057: ctx->M = NULL;
1058: ctx->J = NULL;
1059: /* geometry and grids */
1060: ctx->sphere = PETSC_FALSE;
1061: ctx->use_p4est = PETSC_FALSE;
1062: for (PetscInt grid = 0; grid < LANDAU_MAX_GRIDS; grid++) {
1063: ctx->radius[grid] = 5.; /* thermal radius (velocity) */
1064: ctx->radius_perp[grid] = 5.; /* thermal radius (velocity) */
1065: ctx->radius_par[grid] = 5.; /* thermal radius (velocity) */
1066: ctx->numAMRRefine[grid] = 0;
1067: ctx->postAMRRefine[grid] = 0;
1068: ctx->species_offset[grid + 1] = 1; // one species default
1069: num_species_grid[grid] = 0;
1070: ctx->plex[grid] = NULL; /* cache as expensive to Convert */
1071: }
1072: ctx->species_offset[0] = 0;
1073: ctx->re_radius = 0.;
1074: ctx->vperp0_radius1 = 0;
1075: ctx->vperp0_radius2 = 0;
1076: ctx->nZRefine1 = 0;
1077: ctx->nZRefine2 = 0;
1078: ctx->numRERefine = 0;
1079: num_species_grid[0] = 1; // one species default
1080: /* species - [0] electrons, [1] one ion species eg, duetarium, [2] heavy impurity ion, ... */
1081: ctx->charges[0] = -1; /* electron charge (MKS) */
1082: ctx->masses[0] = 1 / 1835.469965278441013; /* temporary value in proton mass */
1083: ctx->n[0] = 1;
1084: ctx->v_0 = 1; /* thermal velocity, we could start with a scale != 1 */
1085: ctx->thermal_temps[0] = 1;
1086: /* constants, etc. */
1087: ctx->epsilon0 = 8.8542e-12; /* permittivity of free space (MKS) F/m */
1088: ctx->k = 1.38064852e-23; /* Boltzmann constant (MKS) J/K */
1089: ctx->n_0 = 1.e20; /* typical plasma n, but could set it to 1 */
1090: ctx->Ez = 0;
1091: for (PetscInt grid = 0; grid < LANDAU_NUM_TIMERS; grid++) ctx->times[grid] = 0;
1092: for (PetscInt ii = 0; ii < LANDAU_DIM; ii++) ctx->cells0[ii] = 2;
1093: if (LANDAU_DIM == 2) ctx->cells0[0] = 1;
1094: ctx->use_matrix_mass = PETSC_FALSE;
1095: ctx->use_relativistic_corrections = PETSC_FALSE;
1096: ctx->use_energy_tensor_trick = PETSC_FALSE; /* Use Eero's trick for energy conservation v --> grad(v^2/2) */
1097: ctx->SData_d.w = NULL;
1098: ctx->SData_d.x = NULL;
1099: ctx->SData_d.y = NULL;
1100: ctx->SData_d.z = NULL;
1101: ctx->SData_d.invJ = NULL;
1102: ctx->jacobian_field_major_order = PETSC_FALSE;
1103: ctx->SData_d.coo_elem_offsets = NULL;
1104: ctx->SData_d.coo_elem_point_offsets = NULL;
1105: ctx->coo_assembly = PETSC_FALSE;
1106: ctx->SData_d.coo_elem_fullNb = NULL;
1107: ctx->SData_d.coo_size = 0;
1108: PetscOptionsBegin(ctx->comm, prefix, "Options for Fokker-Plank-Landau collision operator", "none");
1109: {
1110: char opstring[256];
1111: #if defined(PETSC_HAVE_KOKKOS_KERNELS)
1112: ctx->deviceType = LANDAU_KOKKOS;
1113: PetscCall(PetscStrncpy(opstring, "kokkos", sizeof(opstring)));
1114: #elif defined(PETSC_HAVE_CUDA)
1115: ctx->deviceType = LANDAU_CUDA;
1116: PetscCall(PetscStrncpy(opstring, "cuda", sizeof(opstring)));
1117: #else
1118: ctx->deviceType = LANDAU_CPU;
1119: PetscCall(PetscStrncpy(opstring, "cpu", sizeof(opstring)));
1120: #endif
1121: PetscCall(PetscOptionsString("-dm_landau_device_type", "Use kernels on 'cpu', 'cuda', or 'kokkos'", "plexland.c", opstring, opstring, sizeof(opstring), NULL));
1122: PetscCall(PetscStrcmp("cpu", opstring, &flg));
1123: if (flg) {
1124: ctx->deviceType = LANDAU_CPU;
1125: } else {
1126: PetscCall(PetscStrcmp("cuda", opstring, &flg));
1127: if (flg) {
1128: ctx->deviceType = LANDAU_CUDA;
1129: } else {
1130: PetscCall(PetscStrcmp("kokkos", opstring, &flg));
1131: if (flg) ctx->deviceType = LANDAU_KOKKOS;
1132: else SETERRQ(ctx->comm, PETSC_ERR_ARG_WRONG, "-dm_landau_device_type %s", opstring);
1133: }
1134: }
1135: }
1136: PetscCall(PetscOptionsReal("-dm_landau_electron_shift", "Shift in thermal velocity of electrons", "none", ctx->electronShift, &ctx->electronShift, NULL));
1137: PetscCall(PetscOptionsInt("-dm_landau_verbose", "Level of verbosity output", "plexland.c", ctx->verbose, &ctx->verbose, NULL));
1138: PetscCall(PetscOptionsInt("-dm_landau_batch_size", "Number of 'vertices' to batch", "ex2.c", ctx->batch_sz, &ctx->batch_sz, NULL));
1139: PetscCheck(LANDAU_MAX_BATCH_SZ >= ctx->batch_sz, ctx->comm, PETSC_ERR_ARG_WRONG, "LANDAU_MAX_BATCH_SZ %" PetscInt_FMT " < ctx->batch_sz %" PetscInt_FMT, (PetscInt)LANDAU_MAX_BATCH_SZ, ctx->batch_sz);
1140: PetscCall(PetscOptionsInt("-dm_landau_batch_view_idx", "Index of batch for diagnostics like plotting", "ex2.c", ctx->batch_view_idx, &ctx->batch_view_idx, NULL));
1141: PetscCheck(ctx->batch_view_idx < ctx->batch_sz, ctx->comm, PETSC_ERR_ARG_WRONG, "-ctx->batch_view_idx %" PetscInt_FMT " > ctx->batch_sz %" PetscInt_FMT, ctx->batch_view_idx, ctx->batch_sz);
1142: PetscCall(PetscOptionsReal("-dm_landau_Ez", "Initial parallel electric field in unites of Conner-Hastie critical field", "plexland.c", ctx->Ez, &ctx->Ez, NULL));
1143: PetscCall(PetscOptionsReal("-dm_landau_n_0", "Normalization constant for number density", "plexland.c", ctx->n_0, &ctx->n_0, NULL));
1144: PetscCall(PetscOptionsBool("-dm_landau_use_mataxpy_mass", "Use fast but slightly fragile MATAXPY to add mass term", "plexland.c", ctx->use_matrix_mass, &ctx->use_matrix_mass, NULL));
1145: PetscCall(PetscOptionsBool("-dm_landau_use_relativistic_corrections", "Use relativistic corrections", "plexland.c", ctx->use_relativistic_corrections, &ctx->use_relativistic_corrections, NULL));
1146: if (LANDAU_DIM == 2 && ctx->use_relativistic_corrections) ctx->use_relativistic_corrections = PETSC_FALSE; // should warn
1147: PetscCall(PetscOptionsBool("-dm_landau_use_energy_tensor_trick", "Use Eero's trick of using grad(v^2/2) instead of v as args to Landau tensor to conserve energy with relativistic corrections and Q1 elements", "plexland.c", ctx->use_energy_tensor_trick,
1148: &ctx->use_energy_tensor_trick, NULL));
1150: /* get num species with temperature, set defaults */
1151: for (ii = 1; ii < LANDAU_MAX_SPECIES; ii++) {
1152: ctx->thermal_temps[ii] = 1;
1153: ctx->charges[ii] = 1;
1154: ctx->masses[ii] = 1;
1155: ctx->n[ii] = 1;
1156: }
1157: nt = LANDAU_MAX_SPECIES;
1158: PetscCall(PetscOptionsRealArray("-dm_landau_thermal_temps", "Temperature of each species [e,i_0,i_1,...] in keV (must be set to set number of species)", "plexland.c", ctx->thermal_temps, &nt, &flg));
1159: if (flg) {
1160: PetscCall(PetscInfo(dummy, "num_species set to number of thermal temps provided (%" PetscInt_FMT ")\n", nt));
1161: ctx->num_species = nt;
1162: } else SETERRQ(ctx->comm, PETSC_ERR_ARG_WRONG, "-dm_landau_thermal_temps ,t1,t2,.. must be provided to set the number of species");
1163: for (ii = 0; ii < ctx->num_species; ii++) ctx->thermal_temps[ii] *= 1.1604525e7; /* convert to Kelvin */
1164: nm = LANDAU_MAX_SPECIES - 1;
1165: PetscCall(PetscOptionsRealArray("-dm_landau_ion_masses", "Mass of each species in units of proton mass [i_0=2,i_1=40...]", "plexland.c", &ctx->masses[1], &nm, &flg));
1166: PetscCheck(!flg || nm == ctx->num_species - 1, ctx->comm, PETSC_ERR_ARG_WRONG, "num ion masses %" PetscInt_FMT " != num species %" PetscInt_FMT, nm, ctx->num_species - 1);
1167: nm = LANDAU_MAX_SPECIES;
1168: PetscCall(PetscOptionsRealArray("-dm_landau_n", "Number density of each species = n_s * n_0", "plexland.c", ctx->n, &nm, &flg));
1169: PetscCheck(!flg || nm == ctx->num_species, ctx->comm, PETSC_ERR_ARG_WRONG, "wrong num n: %" PetscInt_FMT " != num species %" PetscInt_FMT, nm, ctx->num_species);
1170: for (ii = 0; ii < LANDAU_MAX_SPECIES; ii++) ctx->masses[ii] *= 1.6720e-27; /* scale by proton mass kg */
1171: ctx->masses[0] = 9.10938356e-31; /* electron mass kg (should be about right already) */
1172: nc = LANDAU_MAX_SPECIES - 1;
1173: PetscCall(PetscOptionsRealArray("-dm_landau_ion_charges", "Charge of each species in units of proton charge [i_0=2,i_1=18,...]", "plexland.c", &ctx->charges[1], &nc, &flg));
1174: if (flg) PetscCheck(nc == ctx->num_species - 1, ctx->comm, PETSC_ERR_ARG_WRONG, "num charges %" PetscInt_FMT " != num species %" PetscInt_FMT, nc, ctx->num_species - 1);
1175: for (ii = 0; ii < LANDAU_MAX_SPECIES; ii++) ctx->charges[ii] *= 1.6022e-19; /* electron/proton charge (MKS) */
1176: /* geometry and grids */
1177: nt = LANDAU_MAX_GRIDS;
1178: PetscCall(PetscOptionsIntArray("-dm_landau_num_species_grid", "Number of species on each grid: [ 1, ....] or [S, 0 ....] for single grid", "plexland.c", num_species_grid, &nt, &flg));
1179: if (flg) {
1180: ctx->num_grids = nt;
1181: for (ii = nt = 0; ii < ctx->num_grids; ii++) nt += num_species_grid[ii];
1182: PetscCheck(ctx->num_species == nt, ctx->comm, PETSC_ERR_ARG_WRONG, "-dm_landau_num_species_grid: sum %" PetscInt_FMT " != num_species = %" PetscInt_FMT ". %" PetscInt_FMT " grids (check that number of grids <= LANDAU_MAX_GRIDS = %d)", nt, ctx->num_species,
1183: ctx->num_grids, LANDAU_MAX_GRIDS);
1184: } else {
1185: ctx->num_grids = 1; // go back to a single grid run
1186: num_species_grid[0] = ctx->num_species;
1187: }
1188: for (ctx->species_offset[0] = ii = 0; ii < ctx->num_grids; ii++) ctx->species_offset[ii + 1] = ctx->species_offset[ii] + num_species_grid[ii];
1189: PetscCheck(ctx->species_offset[ctx->num_grids] == ctx->num_species, ctx->comm, PETSC_ERR_ARG_WRONG, "ctx->species_offset[ctx->num_grids] %" PetscInt_FMT " != ctx->num_species = %" PetscInt_FMT " ???????????", ctx->species_offset[ctx->num_grids],
1190: ctx->num_species);
1191: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1192: int iii = ctx->species_offset[grid]; // normalize with first (arbitrary) species on grid
1193: v0_grid[grid] = PetscSqrtReal(ctx->k * ctx->thermal_temps[iii] / ctx->masses[iii]); /* arbitrary units for non-dimensionalization: plasma formulary def */
1194: }
1195: // get lambdas here because we need them for t_0 etc
1196: PetscCall(PetscOptionsReal("-dm_landau_ln_lambda", "Universal cross section parameter. Default uses NRL formulas", "plexland.c", lnLam, &lnLam, &flg));
1197: if (flg) {
1198: for (PetscInt grid = 0; grid < LANDAU_MAX_GRIDS; grid++) {
1199: for (PetscInt gridj = 0; gridj < LANDAU_MAX_GRIDS; gridj++) ctx->lambdas[gridj][grid] = lnLam; /* cross section ratio large - small angle collisions */
1200: }
1201: } else {
1202: PetscCall(makeLambdas(ctx));
1203: }
1204: non_dim_grid = 0;
1205: PetscCall(PetscOptionsInt("-dm_landau_normalization_grid", "Index of grid to use for setting v_0, m_0, t_0. (Not recommended)", "plexland.c", non_dim_grid, &non_dim_grid, &flg));
1206: if (non_dim_grid != 0) PetscCall(PetscInfo(dummy, "Normalization grid set to %" PetscInt_FMT ", but non-default not well verified\n", non_dim_grid));
1207: PetscCheck(non_dim_grid >= 0 && non_dim_grid < ctx->num_species, ctx->comm, PETSC_ERR_ARG_WRONG, "Normalization grid wrong: %" PetscInt_FMT, non_dim_grid);
1208: ctx->v_0 = v0_grid[non_dim_grid]; /* arbitrary units for non dimensionalization: global mean velocity in 1D of electrons */
1209: ctx->m_0 = ctx->masses[non_dim_grid]; /* arbitrary reference mass, electrons */
1210: ctx->t_0 = 8 * PETSC_PI * PetscSqr(ctx->epsilon0 * ctx->m_0 / PetscSqr(ctx->charges[non_dim_grid])) / ctx->lambdas[non_dim_grid][non_dim_grid] / ctx->n_0 * PetscPowReal(ctx->v_0, 3); /* note, this t_0 makes nu[non_dim_grid,non_dim_grid]=1 */
1211: /* domain */
1212: nt = LANDAU_MAX_GRIDS;
1213: PetscCall(PetscOptionsRealArray("-dm_landau_domain_radius", "Phase space size in units of thermal velocity of grid", "plexland.c", ctx->radius, &nt, &flg));
1214: if (flg) {
1215: PetscCheck(nt >= ctx->num_grids, ctx->comm, PETSC_ERR_ARG_WRONG, "-dm_landau_domain_radius: given %" PetscInt_FMT " radius != number grids %" PetscInt_FMT, nt, ctx->num_grids);
1216: while (nt--) ctx->radius_par[nt] = ctx->radius_perp[nt] = ctx->radius[nt];
1217: } else {
1218: nt = LANDAU_MAX_GRIDS;
1219: PetscCall(PetscOptionsRealArray("-dm_landau_domain_max_par", "Parallel velocity domain size in units of thermal velocity of grid", "plexland.c", ctx->radius_par, &nt, &flg));
1220: if (flg) PetscCheck(nt >= ctx->num_grids, ctx->comm, PETSC_ERR_ARG_WRONG, "-dm_landau_domain_max_par: given %" PetscInt_FMT " radius != number grids %" PetscInt_FMT, nt, ctx->num_grids);
1221: PetscCall(PetscOptionsRealArray("-dm_landau_domain_max_perp", "Perpendicular velocity domain size in units of thermal velocity of grid", "plexland.c", ctx->radius_perp, &nt, &flg));
1222: if (flg) PetscCheck(nt >= ctx->num_grids, ctx->comm, PETSC_ERR_ARG_WRONG, "-dm_landau_domain_max_perp: given %" PetscInt_FMT " radius != number grids %" PetscInt_FMT, nt, ctx->num_grids);
1223: }
1224: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1225: if (flg && ctx->radius[grid] <= 0) { /* negative is ratio of c - need to set par and perp with this -- todo */
1226: if (ctx->radius[grid] == 0) ctx->radius[grid] = 0.75;
1227: else ctx->radius[grid] = -ctx->radius[grid];
1228: ctx->radius[grid] = ctx->radius[grid] * SPEED_OF_LIGHT / ctx->v_0; // use any species on grid to normalize (v_0 same for all on grid)
1229: PetscCall(PetscInfo(dummy, "Change domain radius to %g for grid %" PetscInt_FMT "\n", (double)ctx->radius[grid], grid));
1230: }
1231: ctx->radius[grid] *= v0_grid[grid] / ctx->v_0; // scale domain by thermal radius relative to v_0
1232: ctx->radius_perp[grid] *= v0_grid[grid] / ctx->v_0; // scale domain by thermal radius relative to v_0
1233: ctx->radius_par[grid] *= v0_grid[grid] / ctx->v_0; // scale domain by thermal radius relative to v_0
1234: }
1235: /* amr parameters */
1236: nt = LANDAU_MAX_GRIDS;
1237: PetscCall(PetscOptionsIntArray("-dm_landau_amr_levels_max", "Number of AMR levels of refinement around origin, after (RE) refinements along z", "plexland.c", ctx->numAMRRefine, &nt, &flg));
1238: PetscCheck(!flg || nt >= ctx->num_grids, ctx->comm, PETSC_ERR_ARG_WRONG, "-dm_landau_amr_levels_max: given %" PetscInt_FMT " != number grids %" PetscInt_FMT, nt, ctx->num_grids);
1239: nt = LANDAU_MAX_GRIDS;
1240: PetscCall(PetscOptionsIntArray("-dm_landau_amr_post_refine", "Number of levels to uniformly refine after AMR", "plexland.c", ctx->postAMRRefine, &nt, &flg));
1241: for (ii = 1; ii < ctx->num_grids; ii++) ctx->postAMRRefine[ii] = ctx->postAMRRefine[0]; // all grids the same now
1242: PetscCall(PetscOptionsInt("-dm_landau_amr_re_levels", "Number of levels to refine along v_perp=0, z>0", "plexland.c", ctx->numRERefine, &ctx->numRERefine, &flg));
1243: PetscCall(PetscOptionsInt("-dm_landau_amr_z_refine_pre", "Number of levels to refine along v_perp=0 before origin refine", "plexland.c", ctx->nZRefine1, &ctx->nZRefine1, &flg));
1244: PetscCall(PetscOptionsInt("-dm_landau_amr_z_refine_post", "Number of levels to refine along v_perp=0 after origin refine", "plexland.c", ctx->nZRefine2, &ctx->nZRefine2, &flg));
1245: PetscCall(PetscOptionsReal("-dm_landau_re_radius", "velocity range to refine on positive (z>0) r=0 axis for runaways", "plexland.c", ctx->re_radius, &ctx->re_radius, &flg));
1246: PetscCall(PetscOptionsReal("-dm_landau_z_radius_pre", "velocity range to refine r=0 axis (for electrons)", "plexland.c", ctx->vperp0_radius1, &ctx->vperp0_radius1, &flg));
1247: PetscCall(PetscOptionsReal("-dm_landau_z_radius_post", "velocity range to refine r=0 axis (for electrons) after origin AMR", "plexland.c", ctx->vperp0_radius2, &ctx->vperp0_radius2, &flg));
1248: /* spherical domain (not used) */
1249: PetscCall(PetscOptionsBool("-dm_landau_sphere", "use sphere/semi-circle domain instead of rectangle", "plexland.c", ctx->sphere, &ctx->sphere, NULL));
1250: if (ctx->sphere) {
1251: ctx->sphere_inner_radius_90degree = 0.40;
1252: ctx->sphere_inner_radius_45degree = 0.35;
1253: PetscCall(PetscOptionsReal("-dm_landau_sphere_inner_radius_90degree_scale", "Scaling of radius for inner circle on 90 degree grid", "plexland.c", ctx->sphere_inner_radius_90degree, &ctx->sphere_inner_radius_90degree, NULL));
1254: PetscCall(PetscOptionsReal("-dm_landau_sphere_inner_radius_45degree_scale", "Scaling of radius for inner circle on 45 degree grid", "plexland.c", ctx->sphere_inner_radius_45degree, &ctx->sphere_inner_radius_45degree, NULL));
1255: } else {
1256: nt = LANDAU_DIM;
1257: PetscCall(PetscOptionsIntArray("-dm_landau_num_cells", "Number of cells in each dimension of base grid", "plexland.c", ctx->cells0, &nt, &flg));
1258: }
1259: /* processing options */
1260: PetscCall(PetscOptionsBool("-dm_landau_gpu_assembly", "Assemble Jacobian on GPU", "plexland.c", ctx->gpu_assembly, &ctx->gpu_assembly, NULL));
1261: if (ctx->deviceType == LANDAU_CPU || ctx->deviceType == LANDAU_KOKKOS) { // make Kokkos
1262: PetscCall(PetscOptionsBool("-dm_landau_coo_assembly", "Assemble Jacobian with Kokkos on 'device'", "plexland.c", ctx->coo_assembly, &ctx->coo_assembly, NULL));
1263: if (ctx->coo_assembly) PetscCheck(ctx->gpu_assembly, ctx->comm, PETSC_ERR_ARG_WRONG, "COO assembly requires 'gpu assembly' even if Kokkos 'CPU' back-end %d", ctx->coo_assembly);
1264: }
1265: PetscCall(PetscOptionsBool("-dm_landau_jacobian_field_major_order", "Reorder Jacobian for GPU assembly with field major, or block diagonal, ordering (DEPRECATED)", "plexland.c", ctx->jacobian_field_major_order, &ctx->jacobian_field_major_order, NULL));
1266: if (ctx->jacobian_field_major_order) PetscCheck(ctx->gpu_assembly, ctx->comm, PETSC_ERR_ARG_WRONG, "-dm_landau_jacobian_field_major_order requires -dm_landau_gpu_assembly");
1267: PetscCheck(!ctx->jacobian_field_major_order, ctx->comm, PETSC_ERR_ARG_WRONG, "-dm_landau_jacobian_field_major_order DEPRECATED");
1268: PetscOptionsEnd();
1270: for (ii = ctx->num_species; ii < LANDAU_MAX_SPECIES; ii++) ctx->masses[ii] = ctx->thermal_temps[ii] = ctx->charges[ii] = 0;
1271: if (ctx->verbose != 0) {
1272: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "masses: e=%10.3e; ions in proton mass units: %10.3e %10.3e ...\n", (double)ctx->masses[0], (double)(ctx->masses[1] / 1.6720e-27), (double)(ctx->num_species > 2 ? ctx->masses[2] / 1.6720e-27 : 0)));
1273: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "charges: e=%10.3e; charges in elementary units: %10.3e %10.3e\n", (double)ctx->charges[0], (double)(-ctx->charges[1] / ctx->charges[0]), (double)(ctx->num_species > 2 ? -ctx->charges[2] / ctx->charges[0] : 0)));
1274: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "n: e: %10.3e i: %10.3e %10.3e\n", (double)ctx->n[0], (double)ctx->n[1], (double)(ctx->num_species > 2 ? ctx->n[2] : 0)));
1275: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "thermal T (K): e=%10.3e i=%10.3e %10.3e. Normalization grid %d: v_0=%10.3e (%10.3ec) n_0=%10.3e t_0=%10.3e %" PetscInt_FMT " batched, view batch %" PetscInt_FMT "\n", (double)ctx->thermal_temps[0],
1276: (double)ctx->thermal_temps[1], (double)((ctx->num_species > 2) ? ctx->thermal_temps[2] : 0), (int)non_dim_grid, (double)ctx->v_0, (double)(ctx->v_0 / SPEED_OF_LIGHT), (double)ctx->n_0, (double)ctx->t_0, ctx->batch_sz, ctx->batch_view_idx));
1277: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "Domain radius (AMR levels) grid %d: par=%10.3e perp=%10.3e (%" PetscInt_FMT ") ", 0, (double)ctx->radius_par[0], (double)ctx->radius_perp[0], ctx->numAMRRefine[0]));
1278: for (ii = 1; ii < ctx->num_grids; ii++) PetscCall(PetscPrintf(PETSC_COMM_WORLD, ", %" PetscInt_FMT ": par=%10.3e perp=%10.3e (%" PetscInt_FMT ") ", ii, (double)ctx->radius_par[ii], (double)ctx->radius_perp[ii], ctx->numAMRRefine[ii]));
1279: if (ctx->use_relativistic_corrections) PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\nUse relativistic corrections\n"));
1280: else PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\n"));
1281: }
1282: PetscCall(DMDestroy(&dummy));
1283: {
1284: PetscMPIInt rank;
1285: PetscCallMPI(MPI_Comm_rank(PETSC_COMM_WORLD, &rank));
1286: ctx->stage = 0;
1287: PetscCall(PetscLogEventRegister("Landau Create", DM_CLASSID, &ctx->events[13])); /* 13 */
1288: PetscCall(PetscLogEventRegister(" GPU ass. setup", DM_CLASSID, &ctx->events[2])); /* 2 */
1289: PetscCall(PetscLogEventRegister(" Build matrix", DM_CLASSID, &ctx->events[12])); /* 12 */
1290: PetscCall(PetscLogEventRegister(" Assembly maps", DM_CLASSID, &ctx->events[15])); /* 15 */
1291: PetscCall(PetscLogEventRegister("Landau Mass mat", DM_CLASSID, &ctx->events[14])); /* 14 */
1292: PetscCall(PetscLogEventRegister("Landau Operator", DM_CLASSID, &ctx->events[11])); /* 11 */
1293: PetscCall(PetscLogEventRegister("Landau Jacobian", DM_CLASSID, &ctx->events[0])); /* 0 */
1294: PetscCall(PetscLogEventRegister("Landau Mass", DM_CLASSID, &ctx->events[9])); /* 9 */
1295: PetscCall(PetscLogEventRegister(" Preamble", DM_CLASSID, &ctx->events[10])); /* 10 */
1296: PetscCall(PetscLogEventRegister(" static IP Data", DM_CLASSID, &ctx->events[7])); /* 7 */
1297: PetscCall(PetscLogEventRegister(" dynamic IP-Jac", DM_CLASSID, &ctx->events[1])); /* 1 */
1298: PetscCall(PetscLogEventRegister(" Kernel-init", DM_CLASSID, &ctx->events[3])); /* 3 */
1299: PetscCall(PetscLogEventRegister(" Jac-f-df (GPU)", DM_CLASSID, &ctx->events[8])); /* 8 */
1300: PetscCall(PetscLogEventRegister(" J Kernel (GPU)", DM_CLASSID, &ctx->events[4])); /* 4 */
1301: PetscCall(PetscLogEventRegister(" M Kernel (GPU)", DM_CLASSID, &ctx->events[16])); /* 16 */
1302: PetscCall(PetscLogEventRegister(" Copy to CPU", DM_CLASSID, &ctx->events[5])); /* 5 */
1303: PetscCall(PetscLogEventRegister(" CPU assemble", DM_CLASSID, &ctx->events[6])); /* 6 */
1305: if (rank) { /* turn off output stuff for duplicate runs - do we need to add the prefix to all this? */
1306: PetscCall(PetscOptionsClearValue(NULL, "-snes_converged_reason"));
1307: PetscCall(PetscOptionsClearValue(NULL, "-ksp_converged_reason"));
1308: PetscCall(PetscOptionsClearValue(NULL, "-snes_monitor"));
1309: PetscCall(PetscOptionsClearValue(NULL, "-ksp_monitor"));
1310: PetscCall(PetscOptionsClearValue(NULL, "-ts_monitor"));
1311: PetscCall(PetscOptionsClearValue(NULL, "-ts_view"));
1312: PetscCall(PetscOptionsClearValue(NULL, "-ts_adapt_monitor"));
1313: PetscCall(PetscOptionsClearValue(NULL, "-dm_landau_amr_dm_view"));
1314: PetscCall(PetscOptionsClearValue(NULL, "-dm_landau_amr_vec_view"));
1315: PetscCall(PetscOptionsClearValue(NULL, "-dm_landau_mass_dm_view"));
1316: PetscCall(PetscOptionsClearValue(NULL, "-dm_landau_mass_view"));
1317: PetscCall(PetscOptionsClearValue(NULL, "-dm_landau_jacobian_view"));
1318: PetscCall(PetscOptionsClearValue(NULL, "-dm_landau_mat_view"));
1319: PetscCall(PetscOptionsClearValue(NULL, "-pc_bjkokkos_ksp_converged_reason"));
1320: PetscCall(PetscOptionsClearValue(NULL, "-pc_bjkokkos_ksp_monitor"));
1321: PetscCall(PetscOptionsClearValue(NULL, "-"));
1322: PetscCall(PetscOptionsClearValue(NULL, "-info"));
1323: }
1324: }
1325: PetscFunctionReturn(PETSC_SUCCESS);
1326: }
1328: static PetscErrorCode CreateStaticData(PetscInt dim, IS grid_batch_is_inv[], LandauCtx *ctx)
1329: {
1330: PetscSection section[LANDAU_MAX_GRIDS], globsection[LANDAU_MAX_GRIDS];
1331: PetscQuadrature quad;
1332: const PetscReal *quadWeights;
1333: PetscReal invMass[LANDAU_MAX_SPECIES], nu_alpha[LANDAU_MAX_SPECIES], nu_beta[LANDAU_MAX_SPECIES];
1334: PetscInt numCells[LANDAU_MAX_GRIDS], Nq, Nf[LANDAU_MAX_GRIDS], ncellsTot = 0, MAP_BF_SIZE = 64 * LANDAU_DIM * LANDAU_DIM * LANDAU_MAX_Q_FACE * LANDAU_MAX_SPECIES;
1335: PetscTabulation *Tf;
1336: PetscDS prob;
1338: PetscFunctionBegin;
1339: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1340: for (PetscInt ii = ctx->species_offset[grid]; ii < ctx->species_offset[grid + 1]; ii++) {
1341: invMass[ii] = ctx->m_0 / ctx->masses[ii];
1342: nu_alpha[ii] = PetscSqr(ctx->charges[ii] / ctx->m_0) * ctx->m_0 / ctx->masses[ii];
1343: nu_beta[ii] = PetscSqr(ctx->charges[ii] / ctx->epsilon0) / (8 * PETSC_PI) * ctx->t_0 * ctx->n_0 / PetscPowReal(ctx->v_0, 3);
1344: }
1345: }
1346: if (ctx->verbose == 4) {
1347: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "nu_alpha: "));
1348: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1349: int iii = ctx->species_offset[grid];
1350: for (PetscInt ii = iii; ii < ctx->species_offset[grid + 1]; ii++) PetscCall(PetscPrintf(PETSC_COMM_WORLD, " %e", (double)nu_alpha[ii]));
1351: }
1352: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\nnu_beta: "));
1353: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1354: int iii = ctx->species_offset[grid];
1355: for (PetscInt ii = iii; ii < ctx->species_offset[grid + 1]; ii++) PetscCall(PetscPrintf(PETSC_COMM_WORLD, " %e", (double)nu_beta[ii]));
1356: }
1357: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\nnu_alpha[i]*nu_beta[j]*lambda[i][j]:\n"));
1358: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1359: int iii = ctx->species_offset[grid];
1360: for (PetscInt ii = iii; ii < ctx->species_offset[grid + 1]; ii++) {
1361: for (PetscInt gridj = 0; gridj < ctx->num_grids; gridj++) {
1362: int jjj = ctx->species_offset[gridj];
1363: for (PetscInt jj = jjj; jj < ctx->species_offset[gridj + 1]; jj++) PetscCall(PetscPrintf(PETSC_COMM_WORLD, " %14.9e", (double)(nu_alpha[ii] * nu_beta[jj] * ctx->lambdas[grid][gridj])));
1364: }
1365: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\n"));
1366: }
1367: }
1368: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "lambda[i][j]:\n"));
1369: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1370: int iii = ctx->species_offset[grid];
1371: for (PetscInt ii = iii; ii < ctx->species_offset[grid + 1]; ii++) {
1372: for (PetscInt gridj = 0; gridj < ctx->num_grids; gridj++) {
1373: int jjj = ctx->species_offset[gridj];
1374: for (PetscInt jj = jjj; jj < ctx->species_offset[gridj + 1]; jj++) PetscCall(PetscPrintf(PETSC_COMM_WORLD, " %14.9e", (double)ctx->lambdas[grid][gridj]));
1375: }
1376: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\n"));
1377: }
1378: }
1379: }
1380: PetscCall(DMGetDS(ctx->plex[0], &prob)); // same DS for all grids
1381: PetscCall(PetscDSGetTabulation(prob, &Tf)); // Bf, &Df same for all grids
1382: /* DS, Tab and quad is same on all grids */
1383: PetscCheck(ctx->plex[0], ctx->comm, PETSC_ERR_ARG_WRONG, "Plex not created");
1384: PetscCall(PetscFEGetQuadrature(ctx->fe[0], &quad));
1385: PetscCall(PetscQuadratureGetData(quad, NULL, NULL, &Nq, NULL, &quadWeights));
1386: PetscCheck(Nq <= LANDAU_MAX_NQ, ctx->comm, PETSC_ERR_ARG_WRONG, "Order too high. Nq = %" PetscInt_FMT " > LANDAU_MAX_NQ (%d)", Nq, LANDAU_MAX_NQ);
1387: /* setup each grid */
1388: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1389: PetscInt cStart, cEnd;
1390: PetscCheck(ctx->plex[grid] != NULL, ctx->comm, PETSC_ERR_ARG_WRONG, "Plex not created");
1391: PetscCall(DMPlexGetHeightStratum(ctx->plex[grid], 0, &cStart, &cEnd));
1392: numCells[grid] = cEnd - cStart; // grids can have different topology
1393: PetscCall(DMGetLocalSection(ctx->plex[grid], §ion[grid]));
1394: PetscCall(DMGetGlobalSection(ctx->plex[grid], &globsection[grid]));
1395: PetscCall(PetscSectionGetNumFields(section[grid], &Nf[grid]));
1396: ncellsTot += numCells[grid];
1397: }
1398: /* create GPU assembly data */
1399: if (ctx->gpu_assembly) { /* we need GPU object with GPU assembly */
1400: PetscContainer container;
1401: PetscScalar *elemMatrix, *elMat;
1402: pointInterpolationP4est(*pointMaps)[LANDAU_MAX_Q_FACE];
1403: P4estVertexMaps *maps;
1404: const PetscInt *plex_batch = NULL, Nb = Nq, elMatSz = Nq * Nq * ctx->num_species * ctx->num_species; // tensor elements;
1405: LandauIdx *coo_elem_offsets = NULL, *coo_elem_fullNb = NULL, (*coo_elem_point_offsets)[LANDAU_MAX_NQ + 1] = NULL;
1406: /* create GPU assembly data */
1407: PetscCall(PetscInfo(ctx->plex[0], "Make GPU maps %d\n", 1));
1408: PetscCall(PetscLogEventBegin(ctx->events[2], 0, 0, 0, 0));
1409: PetscCall(PetscMalloc(sizeof(*maps) * ctx->num_grids, &maps));
1410: PetscCall(PetscMalloc(sizeof(*pointMaps) * MAP_BF_SIZE, &pointMaps));
1411: PetscCall(PetscMalloc(sizeof(*elemMatrix) * elMatSz, &elemMatrix));
1413: if (ctx->coo_assembly) { // setup COO assembly -- put COO metadata directly in ctx->SData_d
1414: PetscCall(PetscMalloc3(ncellsTot + 1, &coo_elem_offsets, ncellsTot, &coo_elem_fullNb, ncellsTot, &coo_elem_point_offsets)); // array of integer pointers
1415: coo_elem_offsets[0] = 0; // finish later
1416: PetscCall(PetscInfo(ctx->plex[0], "COO initialization, %" PetscInt_FMT " cells\n", ncellsTot));
1417: ctx->SData_d.coo_n_cellsTot = ncellsTot;
1418: ctx->SData_d.coo_elem_offsets = (void *)coo_elem_offsets;
1419: ctx->SData_d.coo_elem_fullNb = (void *)coo_elem_fullNb;
1420: ctx->SData_d.coo_elem_point_offsets = (void *)coo_elem_point_offsets;
1421: } else {
1422: ctx->SData_d.coo_elem_offsets = ctx->SData_d.coo_elem_fullNb = NULL;
1423: ctx->SData_d.coo_elem_point_offsets = NULL;
1424: ctx->SData_d.coo_n_cellsTot = 0;
1425: }
1427: ctx->SData_d.coo_max_fullnb = 0;
1428: for (PetscInt grid = 0, glb_elem_idx = 0; grid < ctx->num_grids; grid++) {
1429: PetscInt cStart, cEnd, Nfloc = Nf[grid], totDim = Nfloc * Nq;
1430: if (grid_batch_is_inv[grid]) PetscCall(ISGetIndices(grid_batch_is_inv[grid], &plex_batch));
1431: PetscCall(DMPlexGetHeightStratum(ctx->plex[grid], 0, &cStart, &cEnd));
1432: // make maps
1433: maps[grid].d_self = NULL;
1434: maps[grid].num_elements = numCells[grid];
1435: maps[grid].num_face = (PetscInt)(pow(Nq, 1. / ((double)dim)) + .001); // Q
1436: maps[grid].num_face = (PetscInt)(pow(maps[grid].num_face, (double)(dim - 1)) + .001); // Q^2
1437: maps[grid].num_reduced = 0;
1438: maps[grid].deviceType = ctx->deviceType;
1439: maps[grid].numgrids = ctx->num_grids;
1440: // count reduced and get
1441: PetscCall(PetscMalloc(maps[grid].num_elements * sizeof(*maps[grid].gIdx), &maps[grid].gIdx));
1442: for (int ej = cStart, eidx = 0; ej < cEnd; ++ej, ++eidx, glb_elem_idx++) {
1443: if (coo_elem_offsets) coo_elem_offsets[glb_elem_idx + 1] = coo_elem_offsets[glb_elem_idx]; // start with last one, then add
1444: for (int fieldA = 0; fieldA < Nf[grid]; fieldA++) {
1445: int fullNb = 0;
1446: for (int q = 0; q < Nb; ++q) {
1447: PetscInt numindices, *indices;
1448: PetscScalar *valuesOrig = elMat = elemMatrix;
1449: PetscCall(PetscArrayzero(elMat, totDim * totDim));
1450: elMat[(fieldA * Nb + q) * totDim + fieldA * Nb + q] = 1;
1451: PetscCall(DMPlexGetClosureIndices(ctx->plex[grid], section[grid], globsection[grid], ej, PETSC_TRUE, &numindices, &indices, NULL, (PetscScalar **)&elMat));
1452: for (PetscInt f = 0; f < numindices; ++f) { // look for a non-zero on the diagonal
1453: if (PetscAbs(PetscRealPart(elMat[f * numindices + f])) > PETSC_MACHINE_EPSILON) {
1454: // found it
1455: if (PetscAbs(PetscRealPart(elMat[f * numindices + f] - 1.)) < PETSC_MACHINE_EPSILON) { // normal vertex 1.0
1456: if (plex_batch) {
1457: maps[grid].gIdx[eidx][fieldA][q] = (LandauIdx)plex_batch[indices[f]];
1458: } else {
1459: maps[grid].gIdx[eidx][fieldA][q] = (LandauIdx)indices[f];
1460: }
1461: fullNb++;
1462: } else { //found a constraint
1463: int jj = 0;
1464: PetscReal sum = 0;
1465: const PetscInt ff = f;
1466: maps[grid].gIdx[eidx][fieldA][q] = -maps[grid].num_reduced - 1; // store (-)index: id = -(idx+1): idx = -id - 1
1468: do { // constraints are continuous in Plex - exploit that here
1469: int ii; // get 'scale'
1470: for (ii = 0, pointMaps[maps[grid].num_reduced][jj].scale = 0; ii < maps[grid].num_face; ii++) { // sum row of outer product to recover vector value
1471: if (ff + ii < numindices) { // 3D has Q and Q^2 interps so might run off end. We could test that elMat[f*numindices + ff + ii] > 0, and break if not
1472: pointMaps[maps[grid].num_reduced][jj].scale += PetscRealPart(elMat[f * numindices + ff + ii]);
1473: }
1474: }
1475: sum += pointMaps[maps[grid].num_reduced][jj].scale; // diagnostic
1476: // get 'gid'
1477: if (pointMaps[maps[grid].num_reduced][jj].scale == 0) pointMaps[maps[grid].num_reduced][jj].gid = -1; // 3D has Q and Q^2 interps
1478: else {
1479: if (plex_batch) {
1480: pointMaps[maps[grid].num_reduced][jj].gid = plex_batch[indices[f]];
1481: } else {
1482: pointMaps[maps[grid].num_reduced][jj].gid = indices[f];
1483: }
1484: fullNb++;
1485: }
1486: } while (++jj < maps[grid].num_face && ++f < numindices); // jj is incremented if we hit the end
1487: while (jj < maps[grid].num_face) {
1488: pointMaps[maps[grid].num_reduced][jj].scale = 0;
1489: pointMaps[maps[grid].num_reduced][jj].gid = -1;
1490: jj++;
1491: }
1492: if (PetscAbs(sum - 1.0) > 10 * PETSC_MACHINE_EPSILON) { // debug
1493: int d, f;
1494: PetscReal tmp = 0;
1495: PetscCall(PetscPrintf(PETSC_COMM_SELF, "\t\t%d.%d.%d) ERROR total I = %22.16e (LANDAU_MAX_Q_FACE=%d, #face=%d)\n", eidx, q, fieldA, (double)sum, LANDAU_MAX_Q_FACE, maps[grid].num_face));
1496: for (d = 0, tmp = 0; d < numindices; ++d) {
1497: if (tmp != 0 && PetscAbs(tmp - 1.0) > 10 * PETSC_MACHINE_EPSILON) PetscCall(PetscPrintf(PETSC_COMM_WORLD, "%3d) %3" PetscInt_FMT ": ", d, indices[d]));
1498: for (f = 0; f < numindices; ++f) tmp += PetscRealPart(elMat[d * numindices + f]);
1499: if (tmp != 0) PetscCall(PetscPrintf(ctx->comm, " | %22.16e\n", (double)tmp));
1500: }
1501: }
1502: maps[grid].num_reduced++;
1503: PetscCheck(maps[grid].num_reduced < MAP_BF_SIZE, PETSC_COMM_SELF, PETSC_ERR_PLIB, "maps[grid].num_reduced %d > %" PetscInt_FMT, maps[grid].num_reduced, MAP_BF_SIZE);
1504: }
1505: break;
1506: }
1507: }
1508: // cleanup
1509: PetscCall(DMPlexRestoreClosureIndices(ctx->plex[grid], section[grid], globsection[grid], ej, PETSC_TRUE, &numindices, &indices, NULL, (PetscScalar **)&elMat));
1510: if (elMat != valuesOrig) PetscCall(DMRestoreWorkArray(ctx->plex[grid], numindices * numindices, MPIU_SCALAR, &elMat));
1511: }
1512: if (ctx->coo_assembly) { // setup COO assembly
1513: coo_elem_offsets[glb_elem_idx + 1] += fullNb * fullNb; // one species block, adds a block for each species, on this element in this grid
1514: if (fieldA == 0) { // cache full Nb for this element, on this grid per species
1515: coo_elem_fullNb[glb_elem_idx] = fullNb;
1516: if (fullNb > ctx->SData_d.coo_max_fullnb) ctx->SData_d.coo_max_fullnb = fullNb;
1517: } else PetscCheck(coo_elem_fullNb[glb_elem_idx] == fullNb, PETSC_COMM_SELF, PETSC_ERR_PLIB, "full element size change with species %d %d", coo_elem_fullNb[glb_elem_idx], fullNb);
1518: }
1519: } // field
1520: } // cell
1521: // allocate and copy point data maps[grid].gIdx[eidx][field][q]
1522: PetscCall(PetscMalloc(maps[grid].num_reduced * sizeof(*maps[grid].c_maps), &maps[grid].c_maps));
1523: for (int ej = 0; ej < maps[grid].num_reduced; ++ej) {
1524: for (int q = 0; q < maps[grid].num_face; ++q) {
1525: maps[grid].c_maps[ej][q].scale = pointMaps[ej][q].scale;
1526: maps[grid].c_maps[ej][q].gid = pointMaps[ej][q].gid;
1527: }
1528: }
1529: #if defined(PETSC_HAVE_KOKKOS_KERNELS)
1530: if (ctx->deviceType == LANDAU_KOKKOS) {
1531: PetscCall(LandauKokkosCreateMatMaps(maps, pointMaps, Nf, Nq, grid)); // implies Kokkos does
1532: } // else could be CUDA
1533: #endif
1534: #if defined(PETSC_HAVE_CUDA)
1535: if (ctx->deviceType == LANDAU_CUDA) PetscCall(LandauCUDACreateMatMaps(maps, pointMaps, Nf, Nq, grid));
1536: #endif
1537: if (plex_batch) {
1538: PetscCall(ISRestoreIndices(grid_batch_is_inv[grid], &plex_batch));
1539: PetscCall(ISDestroy(&grid_batch_is_inv[grid])); // we are done with this
1540: }
1541: } /* grids */
1542: // finish COO
1543: if (ctx->coo_assembly) { // setup COO assembly
1544: PetscInt *oor, *ooc;
1545: ctx->SData_d.coo_size = coo_elem_offsets[ncellsTot] * ctx->batch_sz;
1546: PetscCall(PetscMalloc2(ctx->SData_d.coo_size, &oor, ctx->SData_d.coo_size, &ooc));
1547: for (int i = 0; i < ctx->SData_d.coo_size; i++) oor[i] = ooc[i] = -1;
1548: // get
1549: for (int grid = 0, glb_elem_idx = 0; grid < ctx->num_grids; grid++) {
1550: for (int ej = 0; ej < numCells[grid]; ++ej, glb_elem_idx++) {
1551: const int fullNb = coo_elem_fullNb[glb_elem_idx];
1552: const LandauIdx *const Idxs = &maps[grid].gIdx[ej][0][0]; // just use field-0 maps, They should be the same but this is just for COO storage
1553: coo_elem_point_offsets[glb_elem_idx][0] = 0;
1554: for (int f = 0, cnt2 = 0; f < Nb; f++) {
1555: int idx = Idxs[f];
1556: coo_elem_point_offsets[glb_elem_idx][f + 1] = coo_elem_point_offsets[glb_elem_idx][f]; // start at last
1557: if (idx >= 0) {
1558: cnt2++;
1559: coo_elem_point_offsets[glb_elem_idx][f + 1]++; // inc
1560: } else {
1561: idx = -idx - 1;
1562: for (int q = 0; q < maps[grid].num_face; q++) {
1563: if (maps[grid].c_maps[idx][q].gid < 0) break;
1564: cnt2++;
1565: coo_elem_point_offsets[glb_elem_idx][f + 1]++; // inc
1566: }
1567: }
1568: PetscCheck(cnt2 <= fullNb, PETSC_COMM_SELF, PETSC_ERR_PLIB, "wrong count %d < %d", fullNb, cnt2);
1569: }
1570: PetscCheck(coo_elem_point_offsets[glb_elem_idx][Nb] == fullNb, PETSC_COMM_SELF, PETSC_ERR_PLIB, "coo_elem_point_offsets size %d != fullNb=%d", coo_elem_point_offsets[glb_elem_idx][Nb], fullNb);
1571: }
1572: }
1573: // set
1574: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) {
1575: for (int grid = 0, glb_elem_idx = 0; grid < ctx->num_grids; grid++) {
1576: const PetscInt moffset = LAND_MOFFSET(b_id, grid, ctx->batch_sz, ctx->num_grids, ctx->mat_offset);
1577: for (int ej = 0; ej < numCells[grid]; ++ej, glb_elem_idx++) {
1578: const int fullNb = coo_elem_fullNb[glb_elem_idx], fullNb2 = fullNb * fullNb;
1579: // set (i,j)
1580: for (int fieldA = 0; fieldA < Nf[grid]; fieldA++) {
1581: const LandauIdx *const Idxs = &maps[grid].gIdx[ej][fieldA][0];
1582: int rows[LANDAU_MAX_Q_FACE], cols[LANDAU_MAX_Q_FACE];
1583: for (int f = 0; f < Nb; ++f) {
1584: const int nr = coo_elem_point_offsets[glb_elem_idx][f + 1] - coo_elem_point_offsets[glb_elem_idx][f];
1585: if (nr == 1) rows[0] = Idxs[f];
1586: else {
1587: const int idx = -Idxs[f] - 1;
1588: for (int q = 0; q < nr; q++) rows[q] = maps[grid].c_maps[idx][q].gid;
1589: }
1590: for (int g = 0; g < Nb; ++g) {
1591: const int nc = coo_elem_point_offsets[glb_elem_idx][g + 1] - coo_elem_point_offsets[glb_elem_idx][g];
1592: if (nc == 1) cols[0] = Idxs[g];
1593: else {
1594: const int idx = -Idxs[g] - 1;
1595: for (int q = 0; q < nc; q++) cols[q] = maps[grid].c_maps[idx][q].gid;
1596: }
1597: const int idx0 = b_id * coo_elem_offsets[ncellsTot] + coo_elem_offsets[glb_elem_idx] + fieldA * fullNb2 + fullNb * coo_elem_point_offsets[glb_elem_idx][f] + nr * coo_elem_point_offsets[glb_elem_idx][g];
1598: for (int q = 0, idx = idx0; q < nr; q++) {
1599: for (int d = 0; d < nc; d++, idx++) {
1600: oor[idx] = rows[q] + moffset;
1601: ooc[idx] = cols[d] + moffset;
1602: }
1603: }
1604: }
1605: }
1606: }
1607: } // cell
1608: } // grid
1609: } // batch
1610: PetscCall(MatSetPreallocationCOO(ctx->J, ctx->SData_d.coo_size, oor, ooc));
1611: PetscCall(PetscFree2(oor, ooc));
1612: }
1613: PetscCall(PetscFree(pointMaps));
1614: PetscCall(PetscFree(elemMatrix));
1615: PetscCall(PetscContainerCreate(PETSC_COMM_SELF, &container));
1616: PetscCall(PetscContainerSetPointer(container, (void *)maps));
1617: PetscCall(PetscContainerSetUserDestroy(container, LandauGPUMapsDestroy));
1618: PetscCall(PetscObjectCompose((PetscObject)ctx->J, "assembly_maps", (PetscObject)container));
1619: PetscCall(PetscContainerDestroy(&container));
1620: PetscCall(PetscLogEventEnd(ctx->events[2], 0, 0, 0, 0));
1621: } // end GPU assembly
1622: { /* create static point data, Jacobian called first, only one vertex copy */
1623: PetscReal *invJe, *ww, *xx, *yy, *zz = NULL, *invJ_a;
1624: PetscInt outer_ipidx, outer_ej, grid, nip_glb = 0;
1625: PetscFE fe;
1626: const PetscInt Nb = Nq;
1627: PetscCall(PetscLogEventBegin(ctx->events[7], 0, 0, 0, 0));
1628: PetscCall(PetscInfo(ctx->plex[0], "Initialize static data\n"));
1629: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) nip_glb += Nq * numCells[grid];
1630: /* collect f data, first time is for Jacobian, but make mass now */
1631: if (ctx->verbose != 0) {
1632: PetscInt ncells = 0, N;
1633: PetscCall(MatGetSize(ctx->J, &N, NULL));
1634: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) ncells += numCells[grid];
1635: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "%d) %s %" PetscInt_FMT " IPs, %" PetscInt_FMT " cells total, Nb=%" PetscInt_FMT ", Nq=%" PetscInt_FMT ", dim=%" PetscInt_FMT ", Tab: Nb=%" PetscInt_FMT " Nf=%" PetscInt_FMT " Np=%" PetscInt_FMT " cdim=%" PetscInt_FMT " N=%" PetscInt_FMT "\n", 0, "FormLandau", nip_glb, ncells, Nb, Nq, dim, Nb,
1636: ctx->num_species, Nb, dim, N));
1637: }
1638: PetscCall(PetscMalloc4(nip_glb, &ww, nip_glb, &xx, nip_glb, &yy, nip_glb * dim * dim, &invJ_a));
1639: if (dim == 3) PetscCall(PetscMalloc1(nip_glb, &zz));
1640: if (ctx->use_energy_tensor_trick) {
1641: PetscCall(PetscFECreateDefault(PETSC_COMM_SELF, dim, 1, PETSC_FALSE, NULL, PETSC_DECIDE, &fe));
1642: PetscCall(PetscObjectSetName((PetscObject)fe, "energy"));
1643: }
1644: /* init each grids static data - no batch */
1645: for (grid = 0, outer_ipidx = 0, outer_ej = 0; grid < ctx->num_grids; grid++) { // OpenMP (once)
1646: Vec v2_2 = NULL; // projected function: v^2/2 for non-relativistic, gamma... for relativistic
1647: PetscSection e_section;
1648: DM dmEnergy;
1649: PetscInt cStart, cEnd, ej;
1651: PetscCall(DMPlexGetHeightStratum(ctx->plex[grid], 0, &cStart, &cEnd));
1652: // prep energy trick, get v^2 / 2 vector
1653: if (ctx->use_energy_tensor_trick) {
1654: PetscErrorCode (*energyf[1])(PetscInt, PetscReal, const PetscReal[], PetscInt, PetscScalar[], void *) = {ctx->use_relativistic_corrections ? gamma_m1_f : energy_f};
1655: Vec glob_v2;
1656: PetscReal *c2_0[1], data[1] = {PetscSqr(C_0(ctx->v_0))};
1658: PetscCall(DMClone(ctx->plex[grid], &dmEnergy));
1659: PetscCall(PetscObjectSetName((PetscObject)dmEnergy, "energy"));
1660: PetscCall(DMSetField(dmEnergy, 0, NULL, (PetscObject)fe));
1661: PetscCall(DMCreateDS(dmEnergy));
1662: PetscCall(DMGetSection(dmEnergy, &e_section));
1663: PetscCall(DMGetGlobalVector(dmEnergy, &glob_v2));
1664: PetscCall(PetscObjectSetName((PetscObject)glob_v2, "trick"));
1665: c2_0[0] = &data[0];
1666: PetscCall(DMProjectFunction(dmEnergy, 0., energyf, (void **)c2_0, INSERT_ALL_VALUES, glob_v2));
1667: PetscCall(DMGetLocalVector(dmEnergy, &v2_2));
1668: PetscCall(VecZeroEntries(v2_2)); /* zero BCs so don't set */
1669: PetscCall(DMGlobalToLocalBegin(dmEnergy, glob_v2, INSERT_VALUES, v2_2));
1670: PetscCall(DMGlobalToLocalEnd(dmEnergy, glob_v2, INSERT_VALUES, v2_2));
1671: PetscCall(DMViewFromOptions(dmEnergy, NULL, "-energy_dm_view"));
1672: PetscCall(VecViewFromOptions(glob_v2, NULL, "-energy_vec_view"));
1673: PetscCall(DMRestoreGlobalVector(dmEnergy, &glob_v2));
1674: }
1675: /* append part of the IP data for each grid */
1676: for (ej = 0; ej < numCells[grid]; ++ej, ++outer_ej) {
1677: PetscScalar *coefs = NULL;
1678: PetscReal vj[LANDAU_MAX_NQ * LANDAU_DIM], detJj[LANDAU_MAX_NQ], Jdummy[LANDAU_MAX_NQ * LANDAU_DIM * LANDAU_DIM], c0 = C_0(ctx->v_0), c02 = PetscSqr(c0);
1679: invJe = invJ_a + outer_ej * Nq * dim * dim;
1680: PetscCall(DMPlexComputeCellGeometryFEM(ctx->plex[grid], ej + cStart, quad, vj, Jdummy, invJe, detJj));
1681: if (ctx->use_energy_tensor_trick) PetscCall(DMPlexVecGetClosure(dmEnergy, e_section, v2_2, ej + cStart, NULL, &coefs));
1682: /* create static point data */
1683: for (PetscInt qj = 0; qj < Nq; qj++, outer_ipidx++) {
1684: const PetscInt gidx = outer_ipidx;
1685: const PetscReal *invJ = &invJe[qj * dim * dim];
1686: ww[gidx] = detJj[qj] * quadWeights[qj];
1687: if (dim == 2) ww[gidx] *= vj[qj * dim + 0]; /* cylindrical coordinate, w/o 2pi */
1688: // get xx, yy, zz
1689: if (ctx->use_energy_tensor_trick) {
1690: double refSpaceDer[3], eGradPhi[3];
1691: const PetscReal *const DD = Tf[0]->T[1];
1692: const PetscReal *Dq = &DD[qj * Nb * dim];
1693: for (int d = 0; d < 3; ++d) refSpaceDer[d] = eGradPhi[d] = 0.0;
1694: for (int b = 0; b < Nb; ++b) {
1695: for (int d = 0; d < dim; ++d) refSpaceDer[d] += Dq[b * dim + d] * PetscRealPart(coefs[b]);
1696: }
1697: xx[gidx] = 1e10;
1698: if (ctx->use_relativistic_corrections) {
1699: double dg2_c2 = 0;
1700: //for (int d = 0; d < dim; ++d) refSpaceDer[d] *= c02;
1701: for (int d = 0; d < dim; ++d) dg2_c2 += PetscSqr(refSpaceDer[d]);
1702: dg2_c2 *= (double)c02;
1703: if (dg2_c2 >= .999) {
1704: xx[gidx] = vj[qj * dim + 0]; /* coordinate */
1705: yy[gidx] = vj[qj * dim + 1];
1706: if (dim == 3) zz[gidx] = vj[qj * dim + 2];
1707: PetscCall(PetscPrintf(ctx->comm, "Error: %12.5e %" PetscInt_FMT ".%" PetscInt_FMT ") dg2/c02 = %12.5e x= %12.5e %12.5e %12.5e\n", (double)PetscSqrtReal(xx[gidx] * xx[gidx] + yy[gidx] * yy[gidx] + zz[gidx] * zz[gidx]), ej, qj, dg2_c2, (double)xx[gidx], (double)yy[gidx], (double)zz[gidx]));
1708: } else {
1709: PetscReal fact = c02 / PetscSqrtReal(1. - dg2_c2);
1710: for (int d = 0; d < dim; ++d) refSpaceDer[d] *= fact;
1711: // could test with other point u' that (grad - grad') * U (refSpaceDer, refSpaceDer') == 0
1712: }
1713: }
1714: if (xx[gidx] == 1e10) {
1715: for (int d = 0; d < dim; ++d) {
1716: for (int e = 0; e < dim; ++e) eGradPhi[d] += invJ[e * dim + d] * refSpaceDer[e];
1717: }
1718: xx[gidx] = eGradPhi[0];
1719: yy[gidx] = eGradPhi[1];
1720: if (dim == 3) zz[gidx] = eGradPhi[2];
1721: }
1722: } else {
1723: xx[gidx] = vj[qj * dim + 0]; /* coordinate */
1724: yy[gidx] = vj[qj * dim + 1];
1725: if (dim == 3) zz[gidx] = vj[qj * dim + 2];
1726: }
1727: } /* q */
1728: if (ctx->use_energy_tensor_trick) PetscCall(DMPlexVecRestoreClosure(dmEnergy, e_section, v2_2, ej + cStart, NULL, &coefs));
1729: } /* ej */
1730: if (ctx->use_energy_tensor_trick) {
1731: PetscCall(DMRestoreLocalVector(dmEnergy, &v2_2));
1732: PetscCall(DMDestroy(&dmEnergy));
1733: }
1734: } /* grid */
1735: if (ctx->use_energy_tensor_trick) PetscCall(PetscFEDestroy(&fe));
1736: /* cache static data */
1737: if (ctx->deviceType == LANDAU_CUDA || ctx->deviceType == LANDAU_KOKKOS) {
1738: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_KOKKOS_KERNELS)
1739: if (ctx->deviceType == LANDAU_CUDA) {
1740: #if defined(PETSC_HAVE_CUDA)
1741: PetscCall(LandauCUDAStaticDataSet(ctx->plex[0], Nq, ctx->batch_sz, ctx->num_grids, numCells, ctx->species_offset, ctx->mat_offset, nu_alpha, nu_beta, invMass, (PetscReal *)ctx->lambdas, invJ_a, xx, yy, zz, ww, &ctx->SData_d));
1742: #else
1743: SETERRQ(ctx->comm, PETSC_ERR_ARG_WRONG, "-landau_device_type cuda not built");
1744: #endif
1745: } else if (ctx->deviceType == LANDAU_KOKKOS) {
1746: #if defined(PETSC_HAVE_KOKKOS_KERNELS)
1747: PetscCall(LandauKokkosStaticDataSet(ctx->plex[0], Nq, ctx->batch_sz, ctx->num_grids, numCells, ctx->species_offset, ctx->mat_offset, nu_alpha, nu_beta, invMass, (PetscReal *)ctx->lambdas, invJ_a, xx, yy, zz, ww, &ctx->SData_d));
1748: #else
1749: SETERRQ(ctx->comm, PETSC_ERR_ARG_WRONG, "-landau_device_type kokkos not built");
1750: #endif
1751: }
1752: #endif
1753: /* free */
1754: PetscCall(PetscFree4(ww, xx, yy, invJ_a));
1755: if (dim == 3) PetscCall(PetscFree(zz));
1756: } else { /* CPU version, just copy in, only use part */
1757: PetscReal *nu_alpha_p = (PetscReal *)ctx->SData_d.alpha, *nu_beta_p = (PetscReal *)ctx->SData_d.beta, *invMass_p = (PetscReal *)ctx->SData_d.invMass, *lambdas_p = NULL; // why set these ?
1758: ctx->SData_d.w = (void *)ww;
1759: ctx->SData_d.x = (void *)xx;
1760: ctx->SData_d.y = (void *)yy;
1761: ctx->SData_d.z = (void *)zz;
1762: ctx->SData_d.invJ = (void *)invJ_a;
1763: PetscCall(PetscMalloc4(ctx->num_species, &nu_alpha_p, ctx->num_species, &nu_beta_p, ctx->num_species, &invMass_p, LANDAU_MAX_GRIDS * LANDAU_MAX_GRIDS, &lambdas_p));
1764: for (PetscInt ii = 0; ii < ctx->num_species; ii++) {
1765: nu_alpha_p[ii] = nu_alpha[ii];
1766: nu_beta_p[ii] = nu_beta[ii];
1767: invMass_p[ii] = invMass[ii];
1768: }
1769: ctx->SData_d.alpha = (void *)nu_alpha_p;
1770: ctx->SData_d.beta = (void *)nu_beta_p;
1771: ctx->SData_d.invMass = (void *)invMass_p;
1772: ctx->SData_d.lambdas = (void *)lambdas_p;
1773: for (PetscInt grid = 0; grid < LANDAU_MAX_GRIDS; grid++) {
1774: PetscReal(*lambdas)[LANDAU_MAX_GRIDS][LANDAU_MAX_GRIDS] = (PetscReal(*)[LANDAU_MAX_GRIDS][LANDAU_MAX_GRIDS])ctx->SData_d.lambdas;
1775: for (PetscInt gridj = 0; gridj < LANDAU_MAX_GRIDS; gridj++) { (*lambdas)[grid][gridj] = ctx->lambdas[grid][gridj]; }
1776: }
1777: }
1778: PetscCall(PetscLogEventEnd(ctx->events[7], 0, 0, 0, 0));
1779: } // initialize
1780: PetscFunctionReturn(PETSC_SUCCESS);
1781: }
1783: /* < v, u > */
1784: static void g0_1(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, PetscReal u_tShift, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar g0[])
1785: {
1786: g0[0] = 1.;
1787: }
1789: /* < v, u > */
1790: static void g0_fake(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, PetscReal u_tShift, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar g0[])
1791: {
1792: static double ttt = 1e-12;
1793: g0[0] = ttt++;
1794: }
1796: /* < v, u > */
1797: static void g0_r(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, PetscReal u_tShift, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar g0[])
1798: {
1799: g0[0] = 2. * PETSC_PI * x[0];
1800: }
1802: static PetscErrorCode MatrixNfDestroy(void *ptr)
1803: {
1804: PetscInt *nf = (PetscInt *)ptr;
1805: PetscFunctionBegin;
1806: PetscCall(PetscFree(nf));
1807: PetscFunctionReturn(PETSC_SUCCESS);
1808: }
1810: /*
1811: LandauCreateJacobianMatrix - creates ctx->J with without real data. Hard to keep sparse.
1812: - Like DMPlexLandauCreateMassMatrix. Should remove one and combine
1813: - has old support for field major ordering
1814: */
1815: static PetscErrorCode LandauCreateJacobianMatrix(MPI_Comm comm, Vec X, IS grid_batch_is_inv[LANDAU_MAX_GRIDS], LandauCtx *ctx)
1816: {
1817: PetscInt *idxs = NULL;
1818: Mat subM[LANDAU_MAX_GRIDS];
1820: PetscFunctionBegin;
1821: if (!ctx->gpu_assembly) { /* we need GPU object with GPU assembly */
1822: PetscFunctionReturn(PETSC_SUCCESS);
1823: }
1824: // get the RCM for this grid to separate out species into blocks -- create 'idxs' & 'ctx->batch_is' -- not used
1825: if (ctx->gpu_assembly && ctx->jacobian_field_major_order) PetscCall(PetscMalloc1(ctx->mat_offset[ctx->num_grids] * ctx->batch_sz, &idxs));
1826: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1827: const PetscInt *values, n = ctx->mat_offset[grid + 1] - ctx->mat_offset[grid];
1828: Mat gMat;
1829: DM massDM;
1830: PetscDS prob;
1831: Vec tvec;
1832: // get "mass" matrix for reordering
1833: PetscCall(DMClone(ctx->plex[grid], &massDM));
1834: PetscCall(DMCopyFields(ctx->plex[grid], massDM));
1835: PetscCall(DMCreateDS(massDM));
1836: PetscCall(DMGetDS(massDM, &prob));
1837: for (int ix = 0, ii = ctx->species_offset[grid]; ii < ctx->species_offset[grid + 1]; ii++, ix++) PetscCall(PetscDSSetJacobian(prob, ix, ix, g0_fake, NULL, NULL, NULL));
1838: PetscCall(PetscOptionsInsertString(NULL, "-dm_preallocate_only")); // this trick is need to both sparsify the matrix and avoid runtime error
1839: PetscCall(DMCreateMatrix(massDM, &gMat));
1840: PetscCall(PetscOptionsInsertString(NULL, "-dm_preallocate_only false"));
1841: PetscCall(MatSetOption(gMat, MAT_STRUCTURALLY_SYMMETRIC, PETSC_TRUE));
1842: PetscCall(MatSetOption(gMat, MAT_IGNORE_ZERO_ENTRIES, PETSC_TRUE));
1843: PetscCall(DMCreateLocalVector(ctx->plex[grid], &tvec));
1844: PetscCall(DMPlexSNESComputeJacobianFEM(massDM, tvec, gMat, gMat, ctx));
1845: PetscCall(MatViewFromOptions(gMat, NULL, "-dm_landau_reorder_mat_view"));
1846: PetscCall(DMDestroy(&massDM));
1847: PetscCall(VecDestroy(&tvec));
1848: subM[grid] = gMat;
1849: if (ctx->gpu_assembly && ctx->jacobian_field_major_order) {
1850: MatOrderingType rtype = MATORDERINGRCM;
1851: IS isrow, isicol;
1852: PetscCall(MatGetOrdering(gMat, rtype, &isrow, &isicol));
1853: PetscCall(ISInvertPermutation(isrow, PETSC_DECIDE, &grid_batch_is_inv[grid]));
1854: PetscCall(ISGetIndices(isrow, &values));
1855: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) { // add batch size DMs for this species grid
1856: #if !defined(LANDAU_SPECIES_MAJOR)
1857: PetscInt N = ctx->mat_offset[ctx->num_grids], n0 = ctx->mat_offset[grid] + b_id * N;
1858: for (int ii = 0; ii < n; ++ii) idxs[n0 + ii] = values[ii] + n0;
1859: #else
1860: PetscInt n0 = ctx->mat_offset[grid] * ctx->batch_sz + b_id * n;
1861: for (int ii = 0; ii < n; ++ii) idxs[n0 + ii] = values[ii] + n0;
1862: #endif
1863: }
1864: PetscCall(ISRestoreIndices(isrow, &values));
1865: PetscCall(ISDestroy(&isrow));
1866: PetscCall(ISDestroy(&isicol));
1867: }
1868: }
1869: if (ctx->gpu_assembly && ctx->jacobian_field_major_order) PetscCall(ISCreateGeneral(comm, ctx->mat_offset[ctx->num_grids] * ctx->batch_sz, idxs, PETSC_OWN_POINTER, &ctx->batch_is));
1870: // get a block matrix
1871: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1872: Mat B = subM[grid];
1873: PetscInt nloc, nzl, *colbuf, row, COL_BF_SIZE = 1024;
1874: PetscCall(PetscMalloc(sizeof(*colbuf) * COL_BF_SIZE, &colbuf));
1875: PetscCall(MatGetSize(B, &nloc, NULL));
1876: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) {
1877: const PetscInt moffset = LAND_MOFFSET(b_id, grid, ctx->batch_sz, ctx->num_grids, ctx->mat_offset);
1878: const PetscInt *cols;
1879: const PetscScalar *vals;
1880: for (int i = 0; i < nloc; i++) {
1881: PetscCall(MatGetRow(B, i, &nzl, NULL, NULL));
1882: if (nzl > COL_BF_SIZE) {
1883: PetscCall(PetscFree(colbuf));
1884: PetscCall(PetscInfo(ctx->plex[grid], "Realloc buffer %" PetscInt_FMT " to %" PetscInt_FMT " (row size %" PetscInt_FMT ") \n", COL_BF_SIZE, 2 * COL_BF_SIZE, nzl));
1885: COL_BF_SIZE = nzl;
1886: PetscCall(PetscMalloc(sizeof(*colbuf) * COL_BF_SIZE, &colbuf));
1887: }
1888: PetscCall(MatGetRow(B, i, &nzl, &cols, &vals));
1889: for (int j = 0; j < nzl; j++) colbuf[j] = cols[j] + moffset;
1890: row = i + moffset;
1891: PetscCall(MatSetValues(ctx->J, 1, &row, nzl, colbuf, vals, INSERT_VALUES));
1892: PetscCall(MatRestoreRow(B, i, &nzl, &cols, &vals));
1893: }
1894: }
1895: PetscCall(PetscFree(colbuf));
1896: }
1897: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) PetscCall(MatDestroy(&subM[grid]));
1898: PetscCall(MatAssemblyBegin(ctx->J, MAT_FINAL_ASSEMBLY));
1899: PetscCall(MatAssemblyEnd(ctx->J, MAT_FINAL_ASSEMBLY));
1901: // debug
1902: PetscCall(MatViewFromOptions(ctx->J, NULL, "-dm_landau_mat_view"));
1903: if (ctx->gpu_assembly && ctx->jacobian_field_major_order) {
1904: Mat mat_block_order;
1905: PetscCall(MatCreateSubMatrix(ctx->J, ctx->batch_is, ctx->batch_is, MAT_INITIAL_MATRIX, &mat_block_order)); // use MatPermute
1906: PetscCall(MatViewFromOptions(mat_block_order, NULL, "-dm_landau_mat_view"));
1907: PetscCall(MatDestroy(&mat_block_order));
1908: PetscCall(VecScatterCreate(X, ctx->batch_is, X, NULL, &ctx->plex_batch));
1909: PetscCall(VecDuplicate(X, &ctx->work_vec));
1910: }
1912: PetscFunctionReturn(PETSC_SUCCESS);
1913: }
1915: PetscErrorCode DMPlexLandauCreateMassMatrix(DM pack, Mat *Amat);
1916: /*@C
1917: DMPlexLandauCreateVelocitySpace - Create a DMPlex velocity space mesh
1919: Collective
1921: Input Parameters:
1922: + comm - The MPI communicator
1923: . dim - velocity space dimension (2 for axisymmetric, 3 for full 3X + 3V solver)
1924: - prefix - prefix for options (not tested)
1926: Output Parameter:
1927: . pack - The DM object representing the mesh
1928: + X - A vector (user destroys)
1929: - J - Optional matrix (object destroys)
1931: Level: beginner
1933: .keywords: mesh
1934: .seealso: `DMPlexCreate()`, `DMPlexLandauDestroyVelocitySpace()`
1935: @*/
1936: PetscErrorCode DMPlexLandauCreateVelocitySpace(MPI_Comm comm, PetscInt dim, const char prefix[], Vec *X, Mat *J, DM *pack)
1937: {
1938: LandauCtx *ctx;
1939: Vec Xsub[LANDAU_MAX_GRIDS];
1940: IS grid_batch_is_inv[LANDAU_MAX_GRIDS];
1942: PetscFunctionBegin;
1943: PetscCheck(dim == 2 || dim == 3, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Only 2D and 3D supported");
1944: PetscCheck(LANDAU_DIM == dim, PETSC_COMM_SELF, PETSC_ERR_PLIB, "dim %" PetscInt_FMT " != LANDAU_DIM %d", dim, LANDAU_DIM);
1945: PetscCall(PetscNew(&ctx));
1946: ctx->comm = comm; /* used for diagnostics and global errors */
1947: /* process options */
1948: PetscCall(ProcessOptions(ctx, prefix));
1949: if (dim == 2) ctx->use_relativistic_corrections = PETSC_FALSE;
1950: /* Create Mesh */
1951: PetscCall(DMCompositeCreate(PETSC_COMM_SELF, pack));
1952: PetscCall(PetscLogEventBegin(ctx->events[13], 0, 0, 0, 0));
1953: PetscCall(PetscLogEventBegin(ctx->events[15], 0, 0, 0, 0));
1954: PetscCall(LandauDMCreateVMeshes(PETSC_COMM_SELF, dim, prefix, ctx, *pack)); // creates grids (Forest of AMR)
1955: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1956: /* create FEM */
1957: PetscCall(SetupDS(ctx->plex[grid], dim, grid, ctx));
1958: /* set initial state */
1959: PetscCall(DMCreateGlobalVector(ctx->plex[grid], &Xsub[grid]));
1960: PetscCall(PetscObjectSetName((PetscObject)Xsub[grid], "u_orig"));
1961: /* initial static refinement, no solve */
1962: PetscCall(LandauSetInitialCondition(ctx->plex[grid], Xsub[grid], grid, 0, 1, ctx));
1963: /* forest refinement - forest goes in (if forest), plex comes out */
1964: if (ctx->use_p4est) {
1965: DM plex;
1966: PetscCall(adapt(grid, ctx, &Xsub[grid])); // forest goes in, plex comes out
1967: PetscCall(DMViewFromOptions(ctx->plex[grid], NULL, "-dm_landau_amr_dm_view")); // need to differentiate - todo
1968: PetscCall(VecViewFromOptions(Xsub[grid], NULL, "-dm_landau_amr_vec_view"));
1969: // convert to plex, all done with this level
1970: PetscCall(DMConvert(ctx->plex[grid], DMPLEX, &plex));
1971: PetscCall(DMDestroy(&ctx->plex[grid]));
1972: ctx->plex[grid] = plex;
1973: }
1974: #if !defined(LANDAU_SPECIES_MAJOR)
1975: PetscCall(DMCompositeAddDM(*pack, ctx->plex[grid]));
1976: #else
1977: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) { // add batch size DMs for this species grid
1978: PetscCall(DMCompositeAddDM(*pack, ctx->plex[grid]));
1979: }
1980: #endif
1981: PetscCall(DMSetApplicationContext(ctx->plex[grid], ctx));
1982: }
1983: #if !defined(LANDAU_SPECIES_MAJOR)
1984: // stack the batched DMs, could do it all here!!! b_id=0
1985: for (PetscInt b_id = 1; b_id < ctx->batch_sz; b_id++) {
1986: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) PetscCall(DMCompositeAddDM(*pack, ctx->plex[grid]));
1987: }
1988: #endif
1989: // create ctx->mat_offset
1990: ctx->mat_offset[0] = 0;
1991: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1992: PetscInt n;
1993: PetscCall(VecGetLocalSize(Xsub[grid], &n));
1994: ctx->mat_offset[grid + 1] = ctx->mat_offset[grid] + n;
1995: }
1996: // creat DM & Jac
1997: PetscCall(DMSetApplicationContext(*pack, ctx));
1998: PetscCall(PetscOptionsInsertString(NULL, "-dm_preallocate_only"));
1999: PetscCall(DMCreateMatrix(*pack, &ctx->J));
2000: PetscCall(PetscOptionsInsertString(NULL, "-dm_preallocate_only false"));
2001: PetscCall(MatSetOption(ctx->J, MAT_STRUCTURALLY_SYMMETRIC, PETSC_TRUE));
2002: PetscCall(MatSetOption(ctx->J, MAT_IGNORE_ZERO_ENTRIES, PETSC_TRUE));
2003: PetscCall(PetscObjectSetName((PetscObject)ctx->J, "Jac"));
2004: // construct initial conditions in X
2005: PetscCall(DMCreateGlobalVector(*pack, X));
2006: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
2007: PetscInt n;
2008: PetscCall(VecGetLocalSize(Xsub[grid], &n));
2009: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) {
2010: PetscScalar const *values;
2011: const PetscInt moffset = LAND_MOFFSET(b_id, grid, ctx->batch_sz, ctx->num_grids, ctx->mat_offset);
2012: PetscCall(LandauSetInitialCondition(ctx->plex[grid], Xsub[grid], grid, b_id, ctx->batch_sz, ctx));
2013: PetscCall(VecGetArrayRead(Xsub[grid], &values)); // Drop whole grid in Plex ordering
2014: for (int i = 0, idx = moffset; i < n; i++, idx++) PetscCall(VecSetValue(*X, idx, values[i], INSERT_VALUES));
2015: PetscCall(VecRestoreArrayRead(Xsub[grid], &values));
2016: }
2017: }
2018: // cleanup
2019: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) PetscCall(VecDestroy(&Xsub[grid]));
2020: /* check for correct matrix type */
2021: if (ctx->gpu_assembly) { /* we need GPU object with GPU assembly */
2022: PetscBool flg;
2023: if (ctx->deviceType == LANDAU_CUDA) {
2024: PetscCall(PetscObjectTypeCompareAny((PetscObject)ctx->J, &flg, MATSEQAIJCUSPARSE, MATMPIAIJCUSPARSE, MATAIJCUSPARSE, ""));
2025: PetscCheck(flg, ctx->comm, PETSC_ERR_ARG_WRONG, "must use '-dm_mat_type aijcusparse -dm_vec_type cuda' for GPU assembly and Cuda or use '-dm_landau_device_type cpu'");
2026: } else if (ctx->deviceType == LANDAU_KOKKOS) {
2027: PetscCall(PetscObjectTypeCompareAny((PetscObject)ctx->J, &flg, MATSEQAIJKOKKOS, MATMPIAIJKOKKOS, MATAIJKOKKOS, ""));
2028: #if defined(PETSC_HAVE_KOKKOS_KERNELS)
2029: PetscCheck(flg, ctx->comm, PETSC_ERR_ARG_WRONG, "must use '-dm_mat_type aijkokkos -dm_vec_type kokkos' for GPU assembly and Kokkos or use '-dm_landau_device_type cpu'");
2030: #else
2031: PetscCheck(flg, ctx->comm, PETSC_ERR_ARG_WRONG, "must configure with '--download-kokkos-kernels' for GPU assembly and Kokkos or use '-dm_landau_device_type cpu'");
2032: #endif
2033: }
2034: }
2035: PetscCall(PetscLogEventEnd(ctx->events[15], 0, 0, 0, 0));
2037: // create field major ordering
2038: ctx->work_vec = NULL;
2039: ctx->plex_batch = NULL;
2040: ctx->batch_is = NULL;
2041: for (int i = 0; i < LANDAU_MAX_GRIDS; i++) grid_batch_is_inv[i] = NULL;
2042: PetscCall(PetscLogEventBegin(ctx->events[12], 0, 0, 0, 0));
2043: PetscCall(LandauCreateJacobianMatrix(comm, *X, grid_batch_is_inv, ctx));
2044: PetscCall(PetscLogEventEnd(ctx->events[12], 0, 0, 0, 0));
2046: // create AMR GPU assembly maps and static GPU data
2047: PetscCall(CreateStaticData(dim, grid_batch_is_inv, ctx));
2049: PetscCall(PetscLogEventEnd(ctx->events[13], 0, 0, 0, 0));
2051: // create mass matrix
2052: PetscCall(DMPlexLandauCreateMassMatrix(*pack, NULL));
2054: if (J) *J = ctx->J;
2056: if (ctx->gpu_assembly && ctx->jacobian_field_major_order) {
2057: PetscContainer container;
2058: // cache ctx for KSP with batch/field major Jacobian ordering -ksp_type gmres/etc -dm_landau_jacobian_field_major_order
2059: PetscCall(PetscContainerCreate(PETSC_COMM_SELF, &container));
2060: PetscCall(PetscContainerSetPointer(container, (void *)ctx));
2061: PetscCall(PetscObjectCompose((PetscObject)ctx->J, "LandauCtx", (PetscObject)container));
2062: PetscCall(PetscContainerDestroy(&container));
2063: // batch solvers need to map -- can batch solvers work
2064: PetscCall(PetscContainerCreate(PETSC_COMM_SELF, &container));
2065: PetscCall(PetscContainerSetPointer(container, (void *)ctx->plex_batch));
2066: PetscCall(PetscObjectCompose((PetscObject)ctx->J, "plex_batch_is", (PetscObject)container));
2067: PetscCall(PetscContainerDestroy(&container));
2068: }
2069: // for batch solvers
2070: {
2071: PetscContainer container;
2072: PetscInt *pNf;
2073: PetscCall(PetscContainerCreate(PETSC_COMM_SELF, &container));
2074: PetscCall(PetscMalloc1(sizeof(*pNf), &pNf));
2075: *pNf = ctx->batch_sz;
2076: PetscCall(PetscContainerSetPointer(container, (void *)pNf));
2077: PetscCall(PetscContainerSetUserDestroy(container, MatrixNfDestroy));
2078: PetscCall(PetscObjectCompose((PetscObject)ctx->J, "batch size", (PetscObject)container));
2079: PetscCall(PetscContainerDestroy(&container));
2080: }
2082: PetscFunctionReturn(PETSC_SUCCESS);
2083: }
2085: /*@
2086: DMPlexLandauAccess - Access to the distribution function with user callback
2088: Collective
2090: Input Parameters:
2091: . pack - the DMComposite
2092: + func - call back function
2093: . user_ctx - user context
2095: Input/Output Parameter:
2096: . X - Vector to data to
2098: Level: advanced
2100: .keywords: mesh
2101: .seealso: `DMPlexLandauCreateVelocitySpace()`
2102: @*/
2103: PetscErrorCode DMPlexLandauAccess(DM pack, Vec X, PetscErrorCode (*func)(DM, Vec, PetscInt, PetscInt, PetscInt, void *), void *user_ctx)
2104: {
2105: LandauCtx *ctx;
2106: PetscFunctionBegin;
2107: PetscCall(DMGetApplicationContext(pack, &ctx)); // uses ctx->num_grids; ctx->plex[grid]; ctx->batch_sz; ctx->mat_offset
2108: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
2109: PetscInt dim, n;
2110: PetscCall(DMGetDimension(pack, &dim));
2111: for (PetscInt sp = ctx->species_offset[grid], i0 = 0; sp < ctx->species_offset[grid + 1]; sp++, i0++) {
2112: Vec vec;
2113: PetscInt vf[1] = {i0};
2114: IS vis;
2115: DM vdm;
2116: PetscCall(DMCreateSubDM(ctx->plex[grid], 1, vf, &vis, &vdm));
2117: PetscCall(DMSetApplicationContext(vdm, ctx)); // the user might want this
2118: PetscCall(DMCreateGlobalVector(vdm, &vec));
2119: PetscCall(VecGetSize(vec, &n));
2120: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) {
2121: const PetscInt moffset = LAND_MOFFSET(b_id, grid, ctx->batch_sz, ctx->num_grids, ctx->mat_offset);
2122: PetscCall(VecZeroEntries(vec));
2123: /* Add your data with 'dm' for species 'sp' to 'vec' */
2124: PetscCall(func(vdm, vec, i0, grid, b_id, user_ctx));
2125: /* add to global */
2126: PetscScalar const *values;
2127: const PetscInt *offsets;
2128: PetscCall(VecGetArrayRead(vec, &values));
2129: PetscCall(ISGetIndices(vis, &offsets));
2130: for (int i = 0; i < n; i++) PetscCall(VecSetValue(X, moffset + offsets[i], values[i], ADD_VALUES));
2131: PetscCall(VecRestoreArrayRead(vec, &values));
2132: PetscCall(ISRestoreIndices(vis, &offsets));
2133: } // batch
2134: PetscCall(VecDestroy(&vec));
2135: PetscCall(ISDestroy(&vis));
2136: PetscCall(DMDestroy(&vdm));
2137: }
2138: } // grid
2139: PetscFunctionReturn(PETSC_SUCCESS);
2140: }
2142: /*@
2143: DMPlexLandauDestroyVelocitySpace - Destroy a DMPlex velocity space mesh
2145: Collective
2147: Input/Output Parameters:
2148: . dm - the dm to destroy
2150: Level: beginner
2152: .keywords: mesh
2153: .seealso: `DMPlexLandauCreateVelocitySpace()`
2154: @*/
2155: PetscErrorCode DMPlexLandauDestroyVelocitySpace(DM *dm)
2156: {
2157: LandauCtx *ctx;
2158: PetscFunctionBegin;
2159: PetscCall(DMGetApplicationContext(*dm, &ctx));
2160: PetscCall(MatDestroy(&ctx->M));
2161: PetscCall(MatDestroy(&ctx->J));
2162: for (PetscInt ii = 0; ii < ctx->num_species; ii++) PetscCall(PetscFEDestroy(&ctx->fe[ii]));
2163: PetscCall(ISDestroy(&ctx->batch_is));
2164: PetscCall(VecDestroy(&ctx->work_vec));
2165: PetscCall(VecScatterDestroy(&ctx->plex_batch));
2166: if (ctx->deviceType == LANDAU_CUDA) {
2167: #if defined(PETSC_HAVE_CUDA)
2168: PetscCall(LandauCUDAStaticDataClear(&ctx->SData_d));
2169: #else
2170: SETERRQ(ctx->comm, PETSC_ERR_ARG_WRONG, "-landau_device_type %s not built", "cuda");
2171: #endif
2172: } else if (ctx->deviceType == LANDAU_KOKKOS) {
2173: #if defined(PETSC_HAVE_KOKKOS_KERNELS)
2174: PetscCall(LandauKokkosStaticDataClear(&ctx->SData_d));
2175: #else
2176: SETERRQ(ctx->comm, PETSC_ERR_ARG_WRONG, "-landau_device_type %s not built", "kokkos");
2177: #endif
2178: } else {
2179: if (ctx->SData_d.x) { /* in a CPU run */
2180: PetscReal *invJ = (PetscReal *)ctx->SData_d.invJ, *xx = (PetscReal *)ctx->SData_d.x, *yy = (PetscReal *)ctx->SData_d.y, *zz = (PetscReal *)ctx->SData_d.z, *ww = (PetscReal *)ctx->SData_d.w;
2181: LandauIdx *coo_elem_offsets = (LandauIdx *)ctx->SData_d.coo_elem_offsets, *coo_elem_fullNb = (LandauIdx *)ctx->SData_d.coo_elem_fullNb, (*coo_elem_point_offsets)[LANDAU_MAX_NQ + 1] = (LandauIdx(*)[LANDAU_MAX_NQ + 1]) ctx->SData_d.coo_elem_point_offsets;
2182: PetscCall(PetscFree4(ww, xx, yy, invJ));
2183: if (zz) PetscCall(PetscFree(zz));
2184: if (coo_elem_offsets) {
2185: PetscCall(PetscFree3(coo_elem_offsets, coo_elem_fullNb, coo_elem_point_offsets)); // could be NULL
2186: }
2187: PetscCall(PetscFree4(ctx->SData_d.alpha, ctx->SData_d.beta, ctx->SData_d.invMass, ctx->SData_d.lambdas));
2188: }
2189: }
2191: if (ctx->times[LANDAU_MATRIX_TOTAL] > 0) { // OMP timings
2192: PetscCall(PetscPrintf(ctx->comm, "TSStep N 1.0 %10.3e\n", ctx->times[LANDAU_EX2_TSSOLVE]));
2193: PetscCall(PetscPrintf(ctx->comm, "2: Solve: %10.3e with %" PetscInt_FMT " threads\n", ctx->times[LANDAU_EX2_TSSOLVE] - ctx->times[LANDAU_MATRIX_TOTAL], ctx->batch_sz));
2194: PetscCall(PetscPrintf(ctx->comm, "3: Landau: %10.3e\n", ctx->times[LANDAU_MATRIX_TOTAL]));
2195: PetscCall(PetscPrintf(ctx->comm, "Landau Jacobian %" PetscInt_FMT " 1.0 %10.3e\n", (PetscInt)ctx->times[LANDAU_JACOBIAN_COUNT], ctx->times[LANDAU_JACOBIAN]));
2196: PetscCall(PetscPrintf(ctx->comm, "Landau Operator N 1.0 %10.3e\n", ctx->times[LANDAU_OPERATOR]));
2197: PetscCall(PetscPrintf(ctx->comm, "Landau Mass N 1.0 %10.3e\n", ctx->times[LANDAU_MASS]));
2198: PetscCall(PetscPrintf(ctx->comm, " Jac-f-df (GPU) N 1.0 %10.3e\n", ctx->times[LANDAU_F_DF]));
2199: PetscCall(PetscPrintf(ctx->comm, " Kernel (GPU) N 1.0 %10.3e\n", ctx->times[LANDAU_KERNEL]));
2200: PetscCall(PetscPrintf(ctx->comm, "MatLUFactorNum X 1.0 %10.3e\n", ctx->times[KSP_FACTOR]));
2201: PetscCall(PetscPrintf(ctx->comm, "MatSolve X 1.0 %10.3e\n", ctx->times[KSP_SOLVE]));
2202: }
2203: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) PetscCall(DMDestroy(&ctx->plex[grid]));
2204: PetscCall(PetscFree(ctx));
2205: PetscCall(DMDestroy(dm));
2206: PetscFunctionReturn(PETSC_SUCCESS);
2207: }
2209: /* < v, ru > */
2210: static void f0_s_den(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar *f0)
2211: {
2212: PetscInt ii = (PetscInt)PetscRealPart(constants[0]);
2213: f0[0] = u[ii];
2214: }
2216: /* < v, ru > */
2217: static void f0_s_mom(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar *f0)
2218: {
2219: PetscInt ii = (PetscInt)PetscRealPart(constants[0]), jj = (PetscInt)PetscRealPart(constants[1]);
2220: f0[0] = x[jj] * u[ii]; /* x momentum */
2221: }
2223: static void f0_s_v2(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar *f0)
2224: {
2225: PetscInt i, ii = (PetscInt)PetscRealPart(constants[0]);
2226: double tmp1 = 0.;
2227: for (i = 0; i < dim; ++i) tmp1 += x[i] * x[i];
2228: f0[0] = tmp1 * u[ii];
2229: }
2231: static PetscErrorCode gamma_n_f(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nf, PetscScalar *u, void *actx)
2232: {
2233: const PetscReal *c2_0_arr = ((PetscReal *)actx);
2234: const PetscReal c02 = c2_0_arr[0];
2236: PetscFunctionBegin;
2237: for (int s = 0; s < Nf; s++) {
2238: PetscReal tmp1 = 0.;
2239: for (int i = 0; i < dim; ++i) tmp1 += x[i] * x[i];
2240: #if defined(PETSC_USE_DEBUG)
2241: u[s] = PetscSqrtReal(1. + tmp1 / c02); // u[0] = PetscSqrtReal(1. + xx);
2242: #else
2243: {
2244: PetscReal xx = tmp1 / c02;
2245: u[s] = xx / (PetscSqrtReal(1. + xx) + 1.); // better conditioned = xx/(PetscSqrtReal(1. + xx) + 1.)
2246: }
2247: #endif
2248: }
2249: PetscFunctionReturn(PETSC_SUCCESS);
2250: }
2252: /* < v, ru > */
2253: static void f0_s_rden(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar *f0)
2254: {
2255: PetscInt ii = (PetscInt)PetscRealPart(constants[0]);
2256: f0[0] = 2. * PETSC_PI * x[0] * u[ii];
2257: }
2259: /* < v, ru > */
2260: static void f0_s_rmom(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar *f0)
2261: {
2262: PetscInt ii = (PetscInt)PetscRealPart(constants[0]);
2263: f0[0] = 2. * PETSC_PI * x[0] * x[1] * u[ii];
2264: }
2266: static void f0_s_rv2(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar *f0)
2267: {
2268: PetscInt ii = (PetscInt)PetscRealPart(constants[0]);
2269: f0[0] = 2. * PETSC_PI * x[0] * (x[0] * x[0] + x[1] * x[1]) * u[ii];
2270: }
2272: /*@
2273: DMPlexLandauPrintNorms - collects moments and prints them
2275: Collective
2277: Input Parameters:
2278: + X - the state
2279: - stepi - current step to print
2281: Level: beginner
2283: .keywords: mesh
2284: .seealso: `DMPlexLandauCreateVelocitySpace()`
2285: @*/
2286: PetscErrorCode DMPlexLandauPrintNorms(Vec X, PetscInt stepi)
2287: {
2288: LandauCtx *ctx;
2289: PetscDS prob;
2290: DM pack;
2291: PetscInt cStart, cEnd, dim, ii, i0, nDMs;
2292: PetscScalar xmomentumtot = 0, ymomentumtot = 0, zmomentumtot = 0, energytot = 0, densitytot = 0, tt[LANDAU_MAX_SPECIES];
2293: PetscScalar xmomentum[LANDAU_MAX_SPECIES], ymomentum[LANDAU_MAX_SPECIES], zmomentum[LANDAU_MAX_SPECIES], energy[LANDAU_MAX_SPECIES], density[LANDAU_MAX_SPECIES];
2294: Vec *globXArray;
2296: PetscFunctionBegin;
2297: PetscCall(VecGetDM(X, &pack));
2298: PetscCheck(pack, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Vector has no DM");
2299: PetscCall(DMGetDimension(pack, &dim));
2300: PetscCheck(dim == 2 || dim == 3, PETSC_COMM_SELF, PETSC_ERR_PLIB, "dim %" PetscInt_FMT " not in [2,3]", dim);
2301: PetscCall(DMGetApplicationContext(pack, &ctx));
2302: PetscCheck(ctx, PETSC_COMM_SELF, PETSC_ERR_PLIB, "no context");
2303: /* print momentum and energy */
2304: PetscCall(DMCompositeGetNumberDM(pack, &nDMs));
2305: PetscCheck(nDMs == ctx->num_grids * ctx->batch_sz, PETSC_COMM_WORLD, PETSC_ERR_PLIB, "#DM wrong %" PetscInt_FMT " %" PetscInt_FMT, nDMs, ctx->num_grids * ctx->batch_sz);
2306: PetscCall(PetscMalloc(sizeof(*globXArray) * nDMs, &globXArray));
2307: PetscCall(DMCompositeGetAccessArray(pack, X, nDMs, NULL, globXArray));
2308: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
2309: Vec Xloc = globXArray[LAND_PACK_IDX(ctx->batch_view_idx, grid)];
2310: PetscCall(DMGetDS(ctx->plex[grid], &prob));
2311: for (ii = ctx->species_offset[grid], i0 = 0; ii < ctx->species_offset[grid + 1]; ii++, i0++) {
2312: PetscScalar user[2] = {(PetscScalar)i0, (PetscScalar)ctx->charges[ii]};
2313: PetscCall(PetscDSSetConstants(prob, 2, user));
2314: if (dim == 2) { /* 2/3X + 3V (cylindrical coordinates) */
2315: PetscCall(PetscDSSetObjective(prob, 0, &f0_s_rden));
2316: PetscCall(DMPlexComputeIntegralFEM(ctx->plex[grid], Xloc, tt, ctx));
2317: density[ii] = tt[0] * ctx->n_0 * ctx->charges[ii];
2318: PetscCall(PetscDSSetObjective(prob, 0, &f0_s_rmom));
2319: PetscCall(DMPlexComputeIntegralFEM(ctx->plex[grid], Xloc, tt, ctx));
2320: zmomentum[ii] = tt[0] * ctx->n_0 * ctx->v_0 * ctx->masses[ii];
2321: PetscCall(PetscDSSetObjective(prob, 0, &f0_s_rv2));
2322: PetscCall(DMPlexComputeIntegralFEM(ctx->plex[grid], Xloc, tt, ctx));
2323: energy[ii] = tt[0] * 0.5 * ctx->n_0 * ctx->v_0 * ctx->v_0 * ctx->masses[ii];
2324: zmomentumtot += zmomentum[ii];
2325: energytot += energy[ii];
2326: densitytot += density[ii];
2327: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "%3" PetscInt_FMT ") species-%" PetscInt_FMT ": charge density= %20.13e z-momentum= %20.13e energy= %20.13e", stepi, ii, (double)PetscRealPart(density[ii]), (double)PetscRealPart(zmomentum[ii]), (double)PetscRealPart(energy[ii])));
2328: } else { /* 2/3Xloc + 3V */
2329: PetscCall(PetscDSSetObjective(prob, 0, &f0_s_den));
2330: PetscCall(DMPlexComputeIntegralFEM(ctx->plex[grid], Xloc, tt, ctx));
2331: density[ii] = tt[0] * ctx->n_0 * ctx->charges[ii];
2332: PetscCall(PetscDSSetObjective(prob, 0, &f0_s_mom));
2333: user[1] = 0;
2334: PetscCall(PetscDSSetConstants(prob, 2, user));
2335: PetscCall(DMPlexComputeIntegralFEM(ctx->plex[grid], Xloc, tt, ctx));
2336: xmomentum[ii] = tt[0] * ctx->n_0 * ctx->v_0 * ctx->masses[ii];
2337: user[1] = 1;
2338: PetscCall(PetscDSSetConstants(prob, 2, user));
2339: PetscCall(DMPlexComputeIntegralFEM(ctx->plex[grid], Xloc, tt, ctx));
2340: ymomentum[ii] = tt[0] * ctx->n_0 * ctx->v_0 * ctx->masses[ii];
2341: user[1] = 2;
2342: PetscCall(PetscDSSetConstants(prob, 2, user));
2343: PetscCall(DMPlexComputeIntegralFEM(ctx->plex[grid], Xloc, tt, ctx));
2344: zmomentum[ii] = tt[0] * ctx->n_0 * ctx->v_0 * ctx->masses[ii];
2345: if (ctx->use_relativistic_corrections) {
2346: /* gamma * M * f */
2347: if (ii == 0 && grid == 0) { // do all at once
2348: Vec Mf, globGamma, *globMfArray, *globGammaArray;
2349: PetscErrorCode (*gammaf[1])(PetscInt, PetscReal, const PetscReal[], PetscInt, PetscScalar[], void *) = {gamma_n_f};
2350: PetscReal *c2_0[1], data[1];
2352: PetscCall(VecDuplicate(X, &globGamma));
2353: PetscCall(VecDuplicate(X, &Mf));
2354: PetscCall(PetscMalloc(sizeof(*globMfArray) * nDMs, &globMfArray));
2355: PetscCall(PetscMalloc(sizeof(*globMfArray) * nDMs, &globGammaArray));
2356: /* M * f */
2357: PetscCall(MatMult(ctx->M, X, Mf));
2358: /* gamma */
2359: PetscCall(DMCompositeGetAccessArray(pack, globGamma, nDMs, NULL, globGammaArray));
2360: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // yes a grid loop in a grid loop to print nice, need to fix for batching
2361: Vec v1 = globGammaArray[LAND_PACK_IDX(ctx->batch_view_idx, grid)];
2362: data[0] = PetscSqr(C_0(ctx->v_0));
2363: c2_0[0] = &data[0];
2364: PetscCall(DMProjectFunction(ctx->plex[grid], 0., gammaf, (void **)c2_0, INSERT_ALL_VALUES, v1));
2365: }
2366: PetscCall(DMCompositeRestoreAccessArray(pack, globGamma, nDMs, NULL, globGammaArray));
2367: /* gamma * Mf */
2368: PetscCall(DMCompositeGetAccessArray(pack, globGamma, nDMs, NULL, globGammaArray));
2369: PetscCall(DMCompositeGetAccessArray(pack, Mf, nDMs, NULL, globMfArray));
2370: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // yes a grid loop in a grid loop to print nice
2371: PetscInt Nf = ctx->species_offset[grid + 1] - ctx->species_offset[grid], N, bs;
2372: Vec Mfsub = globMfArray[LAND_PACK_IDX(ctx->batch_view_idx, grid)], Gsub = globGammaArray[LAND_PACK_IDX(ctx->batch_view_idx, grid)], v1, v2;
2373: // get each component
2374: PetscCall(VecGetSize(Mfsub, &N));
2375: PetscCall(VecCreate(ctx->comm, &v1));
2376: PetscCall(VecSetSizes(v1, PETSC_DECIDE, N / Nf));
2377: PetscCall(VecCreate(ctx->comm, &v2));
2378: PetscCall(VecSetSizes(v2, PETSC_DECIDE, N / Nf));
2379: PetscCall(VecSetFromOptions(v1)); // ???
2380: PetscCall(VecSetFromOptions(v2));
2381: // get each component
2382: PetscCall(VecGetBlockSize(Gsub, &bs));
2383: PetscCheck(bs == Nf, PETSC_COMM_SELF, PETSC_ERR_PLIB, "bs %" PetscInt_FMT " != num_species %" PetscInt_FMT " in Gsub", bs, Nf);
2384: PetscCall(VecGetBlockSize(Mfsub, &bs));
2385: PetscCheck(bs == Nf, PETSC_COMM_SELF, PETSC_ERR_PLIB, "bs %" PetscInt_FMT " != num_species %" PetscInt_FMT, bs, Nf);
2386: for (int i = 0, ix = ctx->species_offset[grid]; i < Nf; i++, ix++) {
2387: PetscScalar val;
2388: PetscCall(VecStrideGather(Gsub, i, v1, INSERT_VALUES)); // this is not right -- TODO
2389: PetscCall(VecStrideGather(Mfsub, i, v2, INSERT_VALUES));
2390: PetscCall(VecDot(v1, v2, &val));
2391: energy[ix] = PetscRealPart(val) * ctx->n_0 * ctx->v_0 * ctx->v_0 * ctx->masses[ix];
2392: }
2393: PetscCall(VecDestroy(&v1));
2394: PetscCall(VecDestroy(&v2));
2395: } /* grids */
2396: PetscCall(DMCompositeRestoreAccessArray(pack, globGamma, nDMs, NULL, globGammaArray));
2397: PetscCall(DMCompositeRestoreAccessArray(pack, Mf, nDMs, NULL, globMfArray));
2398: PetscCall(PetscFree(globGammaArray));
2399: PetscCall(PetscFree(globMfArray));
2400: PetscCall(VecDestroy(&globGamma));
2401: PetscCall(VecDestroy(&Mf));
2402: }
2403: } else {
2404: PetscCall(PetscDSSetObjective(prob, 0, &f0_s_v2));
2405: PetscCall(DMPlexComputeIntegralFEM(ctx->plex[grid], Xloc, tt, ctx));
2406: energy[ii] = 0.5 * tt[0] * ctx->n_0 * ctx->v_0 * ctx->v_0 * ctx->masses[ii];
2407: }
2408: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "%3" PetscInt_FMT ") species %" PetscInt_FMT ": density=%20.13e, x-momentum=%20.13e, y-momentum=%20.13e, z-momentum=%20.13e, energy=%21.13e", stepi, ii, (double)PetscRealPart(density[ii]), (double)PetscRealPart(xmomentum[ii]), (double)PetscRealPart(ymomentum[ii]), (double)PetscRealPart(zmomentum[ii]), (double)PetscRealPart(energy[ii])));
2409: xmomentumtot += xmomentum[ii];
2410: ymomentumtot += ymomentum[ii];
2411: zmomentumtot += zmomentum[ii];
2412: energytot += energy[ii];
2413: densitytot += density[ii];
2414: }
2415: if (ctx->num_species > 1) PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\n"));
2416: }
2417: }
2418: PetscCall(DMCompositeRestoreAccessArray(pack, X, nDMs, NULL, globXArray));
2419: PetscCall(PetscFree(globXArray));
2420: /* totals */
2421: PetscCall(DMPlexGetHeightStratum(ctx->plex[0], 0, &cStart, &cEnd));
2422: if (ctx->num_species > 1) {
2423: if (dim == 2) {
2424: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\t%3" PetscInt_FMT ") Total: charge density=%21.13e, momentum=%21.13e, energy=%21.13e (m_i[0]/m_e = %g, %" PetscInt_FMT " cells on electron grid)", stepi, (double)PetscRealPart(densitytot), (double)PetscRealPart(zmomentumtot), (double)PetscRealPart(energytot),
2425: (double)(ctx->masses[1] / ctx->masses[0]), cEnd - cStart));
2426: } else {
2427: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\t%3" PetscInt_FMT ") Total: charge density=%21.13e, x-momentum=%21.13e, y-momentum=%21.13e, z-momentum=%21.13e, energy=%21.13e (m_i[0]/m_e = %g, %" PetscInt_FMT " cells)", stepi, (double)PetscRealPart(densitytot), (double)PetscRealPart(xmomentumtot), (double)PetscRealPart(ymomentumtot), (double)PetscRealPart(zmomentumtot), (double)PetscRealPart(energytot),
2428: (double)(ctx->masses[1] / ctx->masses[0]), cEnd - cStart));
2429: }
2430: } else PetscCall(PetscPrintf(PETSC_COMM_WORLD, " -- %" PetscInt_FMT " cells", cEnd - cStart));
2431: PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\n"));
2432: PetscFunctionReturn(PETSC_SUCCESS);
2433: }
2435: /*@
2436: DMPlexLandauCreateMassMatrix - Create mass matrix for Landau in Plex space (not field major order of Jacobian)
2437: - puts mass matrix into ctx->M
2439: Collective
2441: Input Parameter:
2442: . pack - the DM object. Puts matrix in Landau context M field
2444: Output Parameter:
2445: . Amat - The mass matrix (optional), mass matrix is added to the DM context
2447: Level: beginner
2449: .keywords: mesh
2450: .seealso: `DMPlexLandauCreateVelocitySpace()`
2451: @*/
2452: PetscErrorCode DMPlexLandauCreateMassMatrix(DM pack, Mat *Amat)
2453: {
2454: DM mass_pack, massDM[LANDAU_MAX_GRIDS];
2455: PetscDS prob;
2456: PetscInt ii, dim, N1 = 1, N2;
2457: LandauCtx *ctx;
2458: Mat packM, subM[LANDAU_MAX_GRIDS];
2460: PetscFunctionBegin;
2463: PetscCall(DMGetApplicationContext(pack, &ctx));
2464: PetscCheck(ctx, PETSC_COMM_SELF, PETSC_ERR_PLIB, "no context");
2465: PetscCall(PetscLogEventBegin(ctx->events[14], 0, 0, 0, 0));
2466: PetscCall(DMGetDimension(pack, &dim));
2467: PetscCall(DMCompositeCreate(PetscObjectComm((PetscObject)pack), &mass_pack));
2468: /* create pack mass matrix */
2469: for (PetscInt grid = 0, ix = 0; grid < ctx->num_grids; grid++) {
2470: PetscCall(DMClone(ctx->plex[grid], &massDM[grid]));
2471: PetscCall(DMCopyFields(ctx->plex[grid], massDM[grid]));
2472: PetscCall(DMCreateDS(massDM[grid]));
2473: PetscCall(DMGetDS(massDM[grid], &prob));
2474: for (ix = 0, ii = ctx->species_offset[grid]; ii < ctx->species_offset[grid + 1]; ii++, ix++) {
2475: if (dim == 3) PetscCall(PetscDSSetJacobian(prob, ix, ix, g0_1, NULL, NULL, NULL));
2476: else PetscCall(PetscDSSetJacobian(prob, ix, ix, g0_r, NULL, NULL, NULL));
2477: }
2478: #if !defined(LANDAU_SPECIES_MAJOR)
2479: PetscCall(DMCompositeAddDM(mass_pack, massDM[grid]));
2480: #else
2481: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) { // add batch size DMs for this species grid
2482: PetscCall(DMCompositeAddDM(mass_pack, massDM[grid]));
2483: }
2484: #endif
2485: PetscCall(DMCreateMatrix(massDM[grid], &subM[grid]));
2486: }
2487: #if !defined(LANDAU_SPECIES_MAJOR)
2488: // stack the batched DMs
2489: for (PetscInt b_id = 1; b_id < ctx->batch_sz; b_id++) {
2490: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) PetscCall(DMCompositeAddDM(mass_pack, massDM[grid]));
2491: }
2492: #endif
2493: PetscCall(PetscOptionsInsertString(NULL, "-dm_preallocate_only"));
2494: PetscCall(DMCreateMatrix(mass_pack, &packM));
2495: PetscCall(PetscOptionsInsertString(NULL, "-dm_preallocate_only false"));
2496: PetscCall(MatSetOption(packM, MAT_STRUCTURALLY_SYMMETRIC, PETSC_TRUE));
2497: PetscCall(MatSetOption(packM, MAT_IGNORE_ZERO_ENTRIES, PETSC_TRUE));
2498: PetscCall(DMDestroy(&mass_pack));
2499: /* make mass matrix for each block */
2500: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
2501: Vec locX;
2502: DM plex = massDM[grid];
2503: PetscCall(DMGetLocalVector(plex, &locX));
2504: /* Mass matrix is independent of the input, so no need to fill locX */
2505: PetscCall(DMPlexSNESComputeJacobianFEM(plex, locX, subM[grid], subM[grid], ctx));
2506: PetscCall(DMRestoreLocalVector(plex, &locX));
2507: PetscCall(DMDestroy(&massDM[grid]));
2508: }
2509: PetscCall(MatGetSize(ctx->J, &N1, NULL));
2510: PetscCall(MatGetSize(packM, &N2, NULL));
2511: PetscCheck(N1 == N2, PetscObjectComm((PetscObject)pack), PETSC_ERR_PLIB, "Incorrect matrix sizes: |Jacobian| = %" PetscInt_FMT ", |Mass|=%" PetscInt_FMT, N1, N2);
2512: /* assemble block diagonals */
2513: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
2514: Mat B = subM[grid];
2515: PetscInt nloc, nzl, *colbuf, COL_BF_SIZE = 1024, row;
2516: PetscCall(PetscMalloc(sizeof(*colbuf) * COL_BF_SIZE, &colbuf));
2517: PetscCall(MatGetSize(B, &nloc, NULL));
2518: for (PetscInt b_id = 0; b_id < ctx->batch_sz; b_id++) {
2519: const PetscInt moffset = LAND_MOFFSET(b_id, grid, ctx->batch_sz, ctx->num_grids, ctx->mat_offset);
2520: const PetscInt *cols;
2521: const PetscScalar *vals;
2522: for (int i = 0; i < nloc; i++) {
2523: PetscCall(MatGetRow(B, i, &nzl, NULL, NULL));
2524: if (nzl > COL_BF_SIZE) {
2525: PetscCall(PetscFree(colbuf));
2526: PetscCall(PetscInfo(pack, "Realloc buffer %" PetscInt_FMT " to %" PetscInt_FMT " (row size %" PetscInt_FMT ") \n", COL_BF_SIZE, 2 * COL_BF_SIZE, nzl));
2527: COL_BF_SIZE = nzl;
2528: PetscCall(PetscMalloc(sizeof(*colbuf) * COL_BF_SIZE, &colbuf));
2529: }
2530: PetscCall(MatGetRow(B, i, &nzl, &cols, &vals));
2531: for (int j = 0; j < nzl; j++) colbuf[j] = cols[j] + moffset;
2532: row = i + moffset;
2533: PetscCall(MatSetValues(packM, 1, &row, nzl, colbuf, vals, INSERT_VALUES));
2534: PetscCall(MatRestoreRow(B, i, &nzl, &cols, &vals));
2535: }
2536: }
2537: PetscCall(PetscFree(colbuf));
2538: }
2539: // cleanup
2540: for (PetscInt grid = 0; grid < ctx->num_grids; grid++) PetscCall(MatDestroy(&subM[grid]));
2541: PetscCall(MatAssemblyBegin(packM, MAT_FINAL_ASSEMBLY));
2542: PetscCall(MatAssemblyEnd(packM, MAT_FINAL_ASSEMBLY));
2543: PetscCall(PetscObjectSetName((PetscObject)packM, "mass"));
2544: PetscCall(MatViewFromOptions(packM, NULL, "-dm_landau_mass_view"));
2545: ctx->M = packM;
2546: if (Amat) *Amat = packM;
2547: PetscCall(PetscLogEventEnd(ctx->events[14], 0, 0, 0, 0));
2548: PetscFunctionReturn(PETSC_SUCCESS);
2549: }
2551: /*@
2552: DMPlexLandauIFunction - TS residual calculation, confusingly this computes the Jacobian w/o mass
2554: Collective
2556: Input Parameters:
2557: + TS - The time stepping context
2558: . time_dummy - current time (not used)
2559: . X - Current state
2560: . X_t - Time derivative of current state
2561: - actx - Landau context
2563: Output Parameter:
2564: . F - The residual
2566: Level: beginner
2568: .keywords: mesh
2569: .seealso: `DMPlexLandauCreateVelocitySpace()`, `DMPlexLandauIJacobian()`
2570: @*/
2571: PetscErrorCode DMPlexLandauIFunction(TS ts, PetscReal time_dummy, Vec X, Vec X_t, Vec F, void *actx)
2572: {
2573: LandauCtx *ctx = (LandauCtx *)actx;
2574: PetscInt dim;
2575: DM pack;
2576: #if defined(PETSC_HAVE_THREADSAFETY)
2577: double starttime, endtime;
2578: #endif
2579: PetscObjectState state;
2581: PetscFunctionBegin;
2582: PetscCall(TSGetDM(ts, &pack));
2583: PetscCall(DMGetApplicationContext(pack, &ctx));
2584: PetscCheck(ctx, PETSC_COMM_SELF, PETSC_ERR_PLIB, "no context");
2585: if (ctx->stage) PetscCall(PetscLogStagePush(ctx->stage));
2586: PetscCall(PetscLogEventBegin(ctx->events[11], 0, 0, 0, 0));
2587: PetscCall(PetscLogEventBegin(ctx->events[0], 0, 0, 0, 0));
2588: #if defined(PETSC_HAVE_THREADSAFETY)
2589: starttime = MPI_Wtime();
2590: #endif
2591: PetscCall(DMGetDimension(pack, &dim));
2592: PetscCall(PetscObjectStateGet((PetscObject)ctx->J, &state));
2593: if (state != ctx->norm_state) {
2594: PetscCall(PetscInfo(ts, "Create Landau Jacobian t=%g J.state %" PetscInt64_FMT " --> %" PetscInt64_FMT "\n", (double)time_dummy, ctx->norm_state, state));
2595: PetscCall(MatZeroEntries(ctx->J));
2596: PetscCall(LandauFormJacobian_Internal(X, ctx->J, dim, 0.0, (void *)ctx));
2597: PetscCall(MatViewFromOptions(ctx->J, NULL, "-dm_landau_jacobian_view"));
2598: PetscCall(PetscObjectStateGet((PetscObject)ctx->J, &state));
2599: ctx->norm_state = state;
2600: } else {
2601: PetscCall(PetscInfo(ts, "WARNING Skip forming Jacobian, has not changed %" PetscInt64_FMT "\n", state));
2602: }
2603: /* mat vec for op */
2604: PetscCall(MatMult(ctx->J, X, F)); /* C*f */
2605: /* add time term */
2606: if (X_t) PetscCall(MatMultAdd(ctx->M, X_t, F, F));
2607: #if defined(PETSC_HAVE_THREADSAFETY)
2608: if (ctx->stage) {
2609: endtime = MPI_Wtime();
2610: ctx->times[LANDAU_OPERATOR] += (endtime - starttime);
2611: ctx->times[LANDAU_JACOBIAN] += (endtime - starttime);
2612: ctx->times[LANDAU_MATRIX_TOTAL] += (endtime - starttime);
2613: ctx->times[LANDAU_JACOBIAN_COUNT] += 1;
2614: }
2615: #endif
2616: PetscCall(PetscLogEventEnd(ctx->events[0], 0, 0, 0, 0));
2617: PetscCall(PetscLogEventEnd(ctx->events[11], 0, 0, 0, 0));
2618: if (ctx->stage) PetscCall(PetscLogStagePop());
2619: PetscFunctionReturn(PETSC_SUCCESS);
2620: }
2622: /*@
2623: DMPlexLandauIJacobian - TS Jacobian construction, confusingly this adds mass
2625: Collective
2627: Input Parameters:
2628: + TS - The time stepping context
2629: . time_dummy - current time (not used)
2630: . X - Current state
2631: . U_tdummy - Time derivative of current state (not used)
2632: . shift - shift for du/dt term
2633: - actx - Landau context
2635: Output Parameters:
2636: + Amat - Jacobian
2637: - Pmat - same as Amat
2639: Level: beginner
2641: .keywords: mesh
2642: .seealso: `DMPlexLandauCreateVelocitySpace()`, `DMPlexLandauIFunction()`
2643: @*/
2644: PetscErrorCode DMPlexLandauIJacobian(TS ts, PetscReal time_dummy, Vec X, Vec U_tdummy, PetscReal shift, Mat Amat, Mat Pmat, void *actx)
2645: {
2646: LandauCtx *ctx = NULL;
2647: PetscInt dim;
2648: DM pack;
2649: #if defined(PETSC_HAVE_THREADSAFETY)
2650: double starttime, endtime;
2651: #endif
2652: PetscObjectState state;
2654: PetscFunctionBegin;
2655: PetscCall(TSGetDM(ts, &pack));
2656: PetscCall(DMGetApplicationContext(pack, &ctx));
2657: PetscCheck(ctx, PETSC_COMM_SELF, PETSC_ERR_PLIB, "no context");
2658: PetscCheck(Amat == Pmat && Amat == ctx->J, ctx->comm, PETSC_ERR_PLIB, "Amat!=Pmat || Amat!=ctx->J");
2659: PetscCall(DMGetDimension(pack, &dim));
2660: /* get collision Jacobian into A */
2661: if (ctx->stage) PetscCall(PetscLogStagePush(ctx->stage));
2662: PetscCall(PetscLogEventBegin(ctx->events[11], 0, 0, 0, 0));
2663: PetscCall(PetscLogEventBegin(ctx->events[9], 0, 0, 0, 0));
2664: #if defined(PETSC_HAVE_THREADSAFETY)
2665: starttime = MPI_Wtime();
2666: #endif
2667: PetscCall(PetscInfo(ts, "Adding mass to Jacobian t=%g, shift=%g\n", (double)time_dummy, (double)shift));
2668: PetscCheck(shift != 0.0, ctx->comm, PETSC_ERR_PLIB, "zero shift");
2669: PetscCall(PetscObjectStateGet((PetscObject)ctx->J, &state));
2670: PetscCheck(state == ctx->norm_state, ctx->comm, PETSC_ERR_PLIB, "wrong state, %" PetscInt64_FMT " %" PetscInt64_FMT "", ctx->norm_state, state);
2671: if (!ctx->use_matrix_mass) {
2672: PetscCall(LandauFormJacobian_Internal(X, ctx->J, dim, shift, (void *)ctx));
2673: } else { /* add mass */
2674: PetscCall(MatAXPY(Pmat, shift, ctx->M, SAME_NONZERO_PATTERN));
2675: }
2676: #if defined(PETSC_HAVE_THREADSAFETY)
2677: if (ctx->stage) {
2678: endtime = MPI_Wtime();
2679: ctx->times[LANDAU_OPERATOR] += (endtime - starttime);
2680: ctx->times[LANDAU_MASS] += (endtime - starttime);
2681: ctx->times[LANDAU_MATRIX_TOTAL] += (endtime - starttime);
2682: }
2683: #endif
2684: PetscCall(PetscLogEventEnd(ctx->events[9], 0, 0, 0, 0));
2685: PetscCall(PetscLogEventEnd(ctx->events[11], 0, 0, 0, 0));
2686: if (ctx->stage) PetscCall(PetscLogStagePop());
2687: PetscFunctionReturn(PETSC_SUCCESS);
2688: }