xref: /petsc/src/ts/tests/ex30.c (revision a29dfd43bb0c77e2653d3bfa2c953f902720a6d2)
1 static char help[] = "Grid based Landau collision operator with PIC interface with OpenMP setup. (one species per grid)\n";
2 
3 /*
4    Support 2.5V with axisymmetric coordinates
5      - r,z coordinates
6      - Domain and species data input by Landau operator
7      - "radius" for each grid, normalized with electron thermal velocity
8      - Domain: (0,radius) x (-radius,radius), thus first coordinate x[0] is perpendicular velocity and 2pi*x[0] term is added for axisymmetric
9    Supports full 3V
10 
11  */
12 
13 #include <petscdmplex.h>
14 #include <petscds.h>
15 #include <petscdmswarm.h>
16 #include <petscksp.h>
17 #include <petsc/private/petscimpl.h>
18 #if defined(PETSC_HAVE_OPENMP) && defined(PETSC_HAVE_THREADSAFETY)
19   #include <omp.h>
20 #endif
21 #include <petsclandau.h>
22 #include <petscdmcomposite.h>
23 
24 typedef struct {
25   Mat MpTrans;
26   Mat Mp;
27   Vec ff;
28   Vec uu;
29 } MatShellCtx;
30 
31 typedef struct {
32   PetscInt   v_target;
33   PetscInt   g_target;
34   PetscInt   global_vertex_id_0;
35   DM        *globSwarmArray;
36   LandauCtx *ctx;
37   DM        *grid_dm;
38   Mat       *g_Mass;
39   Mat       *globMpArray;
40   Vec       *globXArray;
41   PetscBool  print;
42   PetscBool  print_entropy;
43 } PrintCtx;
44 
45 PetscErrorCode MatMultMtM_SeqAIJ(Mat MtM, Vec xx, Vec yy)
46 {
47   MatShellCtx *matshellctx;
48 
49   PetscFunctionBeginUser;
50   PetscCall(MatShellGetContext(MtM, &matshellctx));
51   PetscCheck(matshellctx, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "No context");
52   PetscCall(MatMult(matshellctx->Mp, xx, matshellctx->ff));
53   PetscCall(MatMult(matshellctx->MpTrans, matshellctx->ff, yy));
54   PetscFunctionReturn(PETSC_SUCCESS);
55 }
56 
57 PetscErrorCode MatMultAddMtM_SeqAIJ(Mat MtM, Vec xx, Vec yy, Vec zz)
58 {
59   MatShellCtx *matshellctx;
60 
61   PetscFunctionBeginUser;
62   PetscCall(MatShellGetContext(MtM, &matshellctx));
63   PetscCheck(matshellctx, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "No context");
64   PetscCall(MatMult(matshellctx->Mp, xx, matshellctx->ff));
65   PetscCall(MatMultAdd(matshellctx->MpTrans, matshellctx->ff, yy, zz));
66   PetscFunctionReturn(PETSC_SUCCESS);
67 }
68 
69 PetscErrorCode createSwarm(const DM dm, PetscInt dim, DM *sw)
70 {
71   PetscInt Nc = 1;
72 
73   PetscFunctionBeginUser;
74   PetscCall(DMCreate(PETSC_COMM_SELF, sw));
75   PetscCall(DMSetType(*sw, DMSWARM));
76   PetscCall(DMSetDimension(*sw, dim));
77   PetscCall(DMSwarmSetType(*sw, DMSWARM_PIC));
78   PetscCall(DMSwarmSetCellDM(*sw, dm));
79   PetscCall(DMSwarmRegisterPetscDatatypeField(*sw, "w_q", Nc, PETSC_REAL));
80   PetscCall(DMSwarmFinalizeFieldRegister(*sw));
81   PetscCall(DMSetFromOptions(*sw));
82   PetscCall(PetscObjectSetName((PetscObject)*sw, "Particle Grid"));
83   PetscFunctionReturn(PETSC_SUCCESS);
84 }
85 
86 static PetscErrorCode makeSwarm(DM sw, const PetscInt dim, const PetscInt Np, const PetscReal xx[], const PetscReal yy[], const PetscReal zz[])
87 {
88   PetscReal    *coords;
89   PetscDataType dtype;
90   PetscInt      bs, p, zero = 0;
91 
92   PetscFunctionBeginUser;
93   PetscCall(DMSwarmSetLocalSizes(sw, Np, zero));
94   PetscCall(DMSwarmGetField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
95   for (p = 0; p < Np; p++) {
96     coords[p * dim + 0] = xx[p];
97     coords[p * dim + 1] = yy[p];
98     if (dim == 3) coords[p * dim + 2] = zz[p];
99   }
100   PetscCall(DMSwarmRestoreField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
101   PetscFunctionReturn(PETSC_SUCCESS);
102 }
103 
104 static PetscErrorCode createMp(const DM dm, DM sw, Mat *Mp_out)
105 {
106   PetscBool removePoints = PETSC_TRUE;
107   Mat       M_p;
108 
109   PetscFunctionBeginUser;
110   // migrate after coords are set
111   PetscCall(DMSwarmMigrate(sw, removePoints));
112   //
113   PetscCall(PetscObjectSetName((PetscObject)sw, "Particle Grid"));
114 
115   /* PetscInt  N,*count,nmin=10000,nmax=0,ntot=0; */
116   /* // count */
117   /* PetscCall(DMSwarmCreatePointPerCellCount(sw, &N, &count)); */
118   /* for (int i=0, n; i< N ; i++) { */
119   /*   if ((n=count[i]) > nmax) nmax = n; */
120   /*   if (n < nmin) nmin = n; */
121   /*   PetscCall(PetscInfo(dm, " %d) %d particles\n", i, n)); */
122   /*   ntot += n; */
123   /* } */
124   /* PetscCall(PetscFree(count)); */
125   /* PetscCall(PetscInfo(dm, " %" PetscInt_FMT " max particle / cell, and %" PetscInt_FMT " min, ratio = %g,  %" PetscInt_FMT " total\n", nmax, nmin, (double)nmax/(double)nmin,ntot)); */
126 
127   /* This gives M f = \int_\Omega \phi f, which looks like a rhs for a PDE */
128   PetscCall(DMCreateMassMatrix(sw, dm, &M_p));
129   PetscCall(DMViewFromOptions(sw, NULL, "-ex30_sw_view"));
130   // output
131   *Mp_out = M_p;
132   PetscFunctionReturn(PETSC_SUCCESS);
133 }
134 
135 static PetscErrorCode particlesToGrid(const DM dm, DM sw, const PetscInt a_tid, const PetscInt dim, const PetscReal a_wp[], Vec rho, Mat M_p)
136 {
137   PetscReal    *wq;
138   PetscDataType dtype;
139   Vec           ff;
140   PetscInt      bs, p, Np;
141 
142   PetscFunctionBeginUser;
143   PetscCall(DMSwarmGetField(sw, "w_q", &bs, &dtype, (void **)&wq));
144   PetscCall(DMSwarmGetLocalSize(sw, &Np));
145   for (p = 0; p < Np; p++) wq[p] = a_wp[p];
146   PetscCall(DMSwarmRestoreField(sw, "w_q", &bs, &dtype, (void **)&wq));
147   PetscCall(PetscObjectSetName((PetscObject)rho, "rho"));
148   PetscCall(DMSwarmCreateGlobalVectorFromField(sw, "w_q", &ff));
149   PetscCall(PetscObjectSetName((PetscObject)ff, "weights"));
150   PetscCall(MatMultTranspose(M_p, ff, rho));
151   PetscCall(DMSwarmDestroyGlobalVectorFromField(sw, "w_q", &ff));
152   PetscFunctionReturn(PETSC_SUCCESS);
153 }
154 
155 //
156 // add grid to arg 'sw.w_q'
157 //
158 PetscErrorCode gridToParticles(const DM dm, DM sw, const Vec rhs, Vec work_ferhs, Mat M_p, Mat Mass)
159 {
160   PetscBool    is_lsqr;
161   KSP          ksp;
162   Mat          PM_p = NULL, MtM, D = NULL;
163   Vec          ff;
164   PetscInt     N, M, nzl;
165   MatShellCtx *matshellctx = NULL;
166   PC           pc;
167 
168   PetscFunctionBeginUser;
169   // (Mp' Mp)^-1 M
170   PetscCall(MatMult(Mass, rhs, work_ferhs));
171   // pseudo-inverse
172   PetscCall(KSPCreate(PETSC_COMM_SELF, &ksp));
173   PetscCall(KSPSetType(ksp, KSPCG));
174   PetscCall(KSPGetPC(ksp, &pc));
175   PetscCall(PCSetType(pc, PCJACOBI));
176   PetscCall(KSPSetOptionsPrefix(ksp, "ftop_"));
177   PetscCall(KSPSetFromOptions(ksp));
178   PetscCall(PetscObjectTypeCompare((PetscObject)ksp, KSPLSQR, &is_lsqr));
179   if (!is_lsqr) {
180     PetscCall(MatGetLocalSize(M_p, &M, &N));
181     if (N > M) {
182       PetscCall(PetscInfo(ksp, " M (%" PetscInt_FMT ") < M (%" PetscInt_FMT ") more vertices than particles: revert to lsqr\n", M, N));
183       is_lsqr = PETSC_TRUE;
184       PetscCall(KSPSetType(ksp, KSPLSQR));
185       PetscCall(PCSetType(pc, PCNONE)); // should not happen, but could solve stable (Mp^T Mp), move projection Mp before solve
186     } else {
187       PetscCall(PetscNew(&matshellctx));
188       PetscCall(MatCreateVecs(M_p, &matshellctx->uu, &matshellctx->ff));
189       if (0) {
190         PetscCall(MatTransposeMatMult(M_p, M_p, MAT_INITIAL_MATRIX, 4, &MtM));
191         PetscCall(KSPSetOperators(ksp, MtM, MtM));
192         PetscCall(PetscInfo(M_p, "createMtM KSP with explicit Mp'Mp\n"));
193         PetscCall(MatViewFromOptions(MtM, NULL, "-ftop2_MtM_mat_view"));
194       } else {
195         PetscCall(MatCreateShell(PetscObjectComm((PetscObject)dm), N, N, PETSC_DECIDE, PETSC_DECIDE, matshellctx, &MtM));
196         PetscCall(MatTranspose(M_p, MAT_INITIAL_MATRIX, &matshellctx->MpTrans));
197         matshellctx->Mp = M_p;
198         PetscCall(MatShellSetOperation(MtM, MATOP_MULT, (void (*)(void))MatMultMtM_SeqAIJ));
199         PetscCall(MatShellSetOperation(MtM, MATOP_MULT_ADD, (void (*)(void))MatMultAddMtM_SeqAIJ));
200         PetscCall(MatCreateSeqAIJ(PETSC_COMM_SELF, N, N, 1, NULL, &D));
201         PetscCall(MatViewFromOptions(matshellctx->MpTrans, NULL, "-ftop2_MpT_mat_view"));
202         for (int i = 0; i < N; i++) {
203           const PetscScalar *vals;
204           const PetscInt    *cols;
205           PetscScalar        dot = 0;
206           PetscCall(MatGetRow(matshellctx->MpTrans, i, &nzl, &cols, &vals));
207           for (int ii = 0; ii < nzl; ii++) dot += PetscSqr(vals[ii]);
208           if (dot < PETSC_MACHINE_EPSILON) {
209             PetscCall(PetscInfo(ksp, "empty row in pseudo-inverse %d\n", i));
210             is_lsqr = PETSC_TRUE; // empty rows
211             PetscCall(KSPSetType(ksp, KSPLSQR));
212             PetscCall(PCSetType(pc, PCNONE)); // should not happen, but could solve stable (Mp Mp^T), move projection Mp before solve
213             // clean up
214             PetscCall(MatDestroy(&matshellctx->MpTrans));
215             PetscCall(VecDestroy(&matshellctx->ff));
216             PetscCall(VecDestroy(&matshellctx->uu));
217             PetscCall(MatDestroy(&D));
218             PetscCall(MatDestroy(&MtM));
219             PetscCall(PetscFree(matshellctx));
220             D = NULL;
221             break;
222           }
223           PetscCall(MatSetValue(D, i, i, dot, INSERT_VALUES));
224         }
225         if (D) {
226           PetscCall(MatAssemblyBegin(D, MAT_FINAL_ASSEMBLY));
227           PetscCall(MatAssemblyEnd(D, MAT_FINAL_ASSEMBLY));
228           PetscCall(PetscInfo(M_p, "createMtMKSP Have %" PetscInt_FMT " eqs, nzl = %" PetscInt_FMT "\n", N, nzl));
229           PetscCall(KSPSetOperators(ksp, MtM, D));
230           PetscCall(MatViewFromOptions(D, NULL, "-ftop2_D_mat_view"));
231           PetscCall(MatViewFromOptions(M_p, NULL, "-ftop2_Mp_mat_view"));
232           PetscCall(MatViewFromOptions(matshellctx->MpTrans, NULL, "-ftop2_MpTranspose_mat_view"));
233           PetscCall(MatViewFromOptions(MtM, NULL, "-ftop2_MtM_mat_view"));
234         }
235       }
236     }
237   }
238   if (is_lsqr) {
239     PC        pc2;
240     PetscBool is_bjac;
241     PetscCall(KSPGetPC(ksp, &pc2));
242     PetscCall(PetscObjectTypeCompare((PetscObject)pc2, PCBJACOBI, &is_bjac));
243     if (is_bjac) {
244       PetscCall(DMSwarmCreateMassMatrixSquare(sw, dm, &PM_p));
245       PetscCall(KSPSetOperators(ksp, M_p, PM_p));
246     } else {
247       PetscCall(KSPSetOperators(ksp, M_p, M_p));
248     }
249     PetscCall(MatViewFromOptions(M_p, NULL, "-ftop2_Mp_mat_view"));
250   }
251   PetscCall(DMSwarmCreateGlobalVectorFromField(sw, "w_q", &ff)); // this grabs access
252   if (!is_lsqr) {
253     PetscErrorCode ierr;
254     ierr = KSPSolve(ksp, work_ferhs, matshellctx->uu);
255     if (!ierr) {
256       PetscCall(MatMult(M_p, matshellctx->uu, ff));
257     } else { // failed
258       PC        pc2;
259       PetscBool is_bjac;
260       PetscCall(PetscInfo(ksp, "Solver failed, probably singular, try lsqr\n"));
261       PetscCall(KSPReset(ksp));
262       PetscCall(KSPSetType(ksp, KSPLSQR));
263       PetscCall(KSPGetPC(ksp, &pc2));
264       PetscCall(PCSetType(pc2, PCNONE)); // should not happen, but could solve stable (Mp Mp^T), move projection Mp before solve
265       PetscCall(KSPSetOptionsPrefix(ksp, "ftop_"));
266       PetscCall(KSPSetFromOptions(ksp));
267       PetscCall(PetscObjectTypeCompare((PetscObject)pc2, PCBJACOBI, &is_bjac));
268       if (is_bjac) {
269         PetscCall(DMSwarmCreateMassMatrixSquare(sw, dm, &PM_p));
270         PetscCall(KSPSetOperators(ksp, M_p, PM_p));
271       } else {
272         PetscCall(KSPSetOperators(ksp, M_p, M_p));
273       }
274       ierr = KSPSolveTranspose(ksp, work_ferhs, ff);
275       if (ierr) { PetscCheck(!ierr, PETSC_COMM_WORLD, PETSC_ERR_PLIB, "backup LSQR solver failed - need to add N_v > N_p Moore-Penrose pseudo-inverse"); }
276     }
277     if (D) PetscCall(MatDestroy(&D));
278     PetscCall(MatDestroy(&MtM));
279     if (matshellctx->MpTrans) PetscCall(MatDestroy(&matshellctx->MpTrans));
280     PetscCall(VecDestroy(&matshellctx->ff));
281     PetscCall(VecDestroy(&matshellctx->uu));
282     PetscCall(PetscFree(matshellctx));
283   } else {
284     PetscErrorCode ierr;
285     ierr = KSPSolveTranspose(ksp, work_ferhs, ff);
286     if (ierr) { PetscCheck(!ierr, PETSC_COMM_WORLD, PETSC_ERR_PLIB, "backup LSQR solver failed - need to add N_v > N_p Moore-Penrose pseudo-inverse"); }
287   }
288   PetscCall(KSPDestroy(&ksp));
289   PetscCall(MatDestroy(&PM_p));
290   PetscCall(DMSwarmDestroyGlobalVectorFromField(sw, "w_q", &ff));
291   PetscFunctionReturn(PETSC_SUCCESS);
292 }
293 
294 #define EX30_MAX_NUM_THRDS 12
295 #define EX30_MAX_BATCH_SZ  1024
296 //
297 // add grid to arg 'globSwarmArray[].w_q'
298 //
299 PetscErrorCode gridToParticles_private(DM grid_dm[], DM globSwarmArray[], const PetscInt dim, const PetscInt v_target, const PetscInt numthreads, const PetscInt num_vertices, const PetscInt global_vertex_id, Mat globMpArray[], Mat g_Mass[], Vec t_fhat[][EX30_MAX_NUM_THRDS], PetscReal moments[], Vec globXArray[], LandauCtx *ctx)
300 {
301   PetscErrorCode ierr = (PetscErrorCode)0; // used for inside thread loops
302 
303   PetscFunctionBeginUser;
304   // map back to particles
305   for (PetscInt v_id_0 = 0; v_id_0 < ctx->batch_sz; v_id_0 += numthreads) {
306     PetscCall(PetscInfo(grid_dm[0], "g2p: global batch %" PetscInt_FMT " of %" PetscInt_FMT ", Landau batch %" PetscInt_FMT " of %" PetscInt_FMT ": map back to particles\n", global_vertex_id + 1, num_vertices, v_id_0 + 1, ctx->batch_sz));
307     //PetscPragmaOMP(parallel for)
308     for (int tid = 0; tid < numthreads; tid++) {
309       const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id + v_id;
310       if (glb_v_id < num_vertices) {
311         for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
312           PetscErrorCode ierr_t;
313           ierr_t = PetscInfo(grid_dm[0], "gridToParticles: global batch %" PetscInt_FMT ", local batch b=%" PetscInt_FMT ", grid g=%" PetscInt_FMT ", index(b,g) %" PetscInt_FMT "\n", global_vertex_id, v_id, grid, LAND_PACK_IDX(v_id, grid));
314           ierr_t = gridToParticles(grid_dm[grid], globSwarmArray[LAND_PACK_IDX(v_id, grid)], globXArray[LAND_PACK_IDX(v_id, grid)], t_fhat[grid][tid], globMpArray[LAND_PACK_IDX(v_id, grid)], g_Mass[grid]);
315           if (ierr_t) ierr = ierr_t;
316         }
317       }
318     }
319     PetscCheck(!ierr, PETSC_COMM_WORLD, PETSC_ERR_PLIB, "Error in OMP loop. ierr = %d", (int)ierr);
320     /* Get moments */
321     PetscCall(PetscInfo(grid_dm[0], "Cleanup batches %" PetscInt_FMT " to %" PetscInt_FMT "\n", v_id_0, v_id_0 + numthreads));
322     for (int tid = 0; tid < numthreads; tid++) {
323       const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id + v_id;
324       if (glb_v_id == v_target) {
325         for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
326           PetscDataType dtype;
327           PetscReal    *wp, *coords;
328           DM            sw = globSwarmArray[LAND_PACK_IDX(v_id, grid)];
329           PetscInt      npoints, bs = 1;
330           PetscCall(DMSwarmGetField(sw, "w_q", &bs, &dtype, (void **)&wp)); // take data out here
331           PetscCall(DMSwarmGetField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
332           PetscCall(DMSwarmGetLocalSize(sw, &npoints));
333           for (int p = 0; p < npoints; p++) {
334             PetscReal v2 = 0, fact = (dim == 2) ? 2.0 * PETSC_PI * coords[p * dim + 0] : 1, w = fact * wp[p] * ctx->n_0 * ctx->masses[ctx->species_offset[grid]];
335             for (int i = 0; i < dim; ++i) v2 += PetscSqr(coords[p * dim + i]);
336             moments[0] += w;
337             moments[1] += w * ctx->v_0 * coords[p * dim + 1]; // z-momentum
338             moments[2] += w * 0.5 * ctx->v_0 * ctx->v_0 * v2;
339           }
340           PetscCall(DMSwarmRestoreField(sw, "w_q", &bs, &dtype, (void **)&wp));
341           PetscCall(DMSwarmRestoreField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
342         }
343         const PetscReal N_inv = 1 / moments[0];
344         PetscCall(PetscInfo(grid_dm[0], "gridToParticles_private [%" PetscInt_FMT "], n = %g\n", v_id, (double)moments[0]));
345         for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
346           PetscDataType dtype;
347           PetscReal    *wp, *coords;
348           DM            sw = globSwarmArray[LAND_PACK_IDX(v_id, grid)];
349           PetscInt      npoints, bs = 1;
350           PetscCall(DMSwarmGetField(sw, "w_q", &bs, &dtype, (void **)&wp)); // take data out here
351           PetscCall(DMSwarmGetField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
352           PetscCall(DMSwarmGetLocalSize(sw, &npoints));
353           for (int p = 0; p < npoints; p++) {
354             const PetscReal fact = dim == 2 ? 2.0 * PETSC_PI * coords[p * dim + 0] : 1, w = fact * wp[p] * ctx->n_0 * ctx->masses[ctx->species_offset[grid]], ww = w * N_inv;
355             if (w > PETSC_REAL_MIN) {
356               moments[3] -= ww * PetscLogReal(ww);
357               PetscCheck(ww < 1 - PETSC_MACHINE_EPSILON, PETSC_COMM_WORLD, PETSC_ERR_PLIB, "ww (%g) > 1", (double)ww);
358             } else moments[4] -= w; // keep track of density that is lost
359           }
360           PetscCall(DMSwarmRestoreField(sw, "w_q", &bs, &dtype, (void **)&wp));
361           PetscCall(DMSwarmRestoreField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
362         }
363       }
364     } // thread batch
365   } // batch
366   PetscFunctionReturn(PETSC_SUCCESS);
367 }
368 
369 static void maxwellian(PetscInt dim, const PetscReal x[], PetscReal kt_m, PetscReal n, PetscReal shift, PetscScalar *u)
370 {
371   PetscInt  i;
372   PetscReal v2 = 0, theta = 2.0 * kt_m; /* theta = 2kT/mc^2 */
373 
374   if (shift != 0.) {
375     v2 = 0;
376     for (i = 0; i < dim - 1; ++i) v2 += x[i] * x[i];
377     v2 += (x[dim - 1] - shift) * (x[dim - 1] - shift);
378     /* evaluate the shifted Maxwellian */
379     u[0] += n * PetscPowReal(PETSC_PI * theta, -1.5) * (PetscExpReal(-v2 / theta));
380   } else {
381     /* compute the exponents, v^2 */
382     for (i = 0; i < dim; ++i) v2 += x[i] * x[i];
383     /* evaluate the Maxwellian */
384     u[0] += n * PetscPowReal(PETSC_PI * theta, -1.5) * (PetscExpReal(-v2 / theta));
385   }
386 }
387 
388 static PetscErrorCode PostStep(TS ts)
389 {
390   PetscInt   n, dim, nDMs, v_id;
391   PetscReal  t;
392   LandauCtx *ctx;
393   Vec        X;
394   PrintCtx  *printCtx;
395   DM         pack;
396   PetscReal  moments[5], e_grid[LANDAU_MAX_GRIDS];
397 
398   PetscFunctionBeginUser;
399   PetscCall(TSGetApplicationContext(ts, &printCtx));
400   if (!printCtx->print && !printCtx->print_entropy) PetscFunctionReturn(PETSC_SUCCESS);
401   ctx = printCtx->ctx;
402   if (printCtx->v_target < printCtx->global_vertex_id_0 || printCtx->v_target >= printCtx->global_vertex_id_0 + ctx->batch_sz) PetscFunctionReturn(PETSC_SUCCESS);
403   for (int i = 0; i < 5; i++) moments[i] = 0;
404   for (int i = 0; i < LANDAU_MAX_GRIDS; i++) e_grid[i] = 0;
405   v_id = printCtx->v_target % ctx->batch_sz;
406   PetscCall(TSGetDM(ts, &pack));
407   PetscCall(DMGetDimension(pack, &dim));
408   PetscCall(DMCompositeGetNumberDM(pack, &nDMs)); // number of vertices * number of grids
409   PetscCall(TSGetSolution(ts, &X));
410   PetscCall(TSGetStepNumber(ts, &n));
411   PetscCall(TSGetTime(ts, &t));
412   PetscCall(DMCompositeGetAccessArray(pack, X, nDMs, NULL, printCtx->globXArray));
413   if (printCtx->print_entropy && printCtx->v_target >= 0 && 0) {
414     for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
415       PetscDataType dtype;
416       PetscReal    *wp, *coords;
417       DM            sw = printCtx->globSwarmArray[LAND_PACK_IDX(v_id, grid)];
418       Vec           work, subX = printCtx->globXArray[LAND_PACK_IDX(v_id, grid)];
419       PetscInt      bs, NN;
420       // C-G moments
421       PetscCall(VecDuplicate(subX, &work));
422       PetscCall(gridToParticles(printCtx->grid_dm[grid], sw, subX, work, printCtx->globMpArray[LAND_PACK_IDX(v_id, grid)], printCtx->g_Mass[grid]));
423       PetscCall(VecDestroy(&work));
424       // moments
425       PetscCall(DMSwarmGetField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
426       PetscCall(DMSwarmGetLocalSize(sw, &NN));
427       PetscCall(DMSwarmGetField(sw, "w_q", &bs, &dtype, (void **)&wp));
428       for (int pp = 0; pp < NN; pp++) {
429         PetscReal v2 = 0, fact = (dim == 2) ? 2.0 * PETSC_PI * coords[pp * dim + 0] : 1, w = fact * wp[pp] * ctx->n_0 * ctx->masses[ctx->species_offset[grid]];
430         for (int i = 0; i < dim; ++i) v2 += PetscSqr(coords[pp * dim + i]);
431         moments[0] += w;
432         moments[1] += w * ctx->v_0 * coords[pp * dim + 1]; // z-momentum
433         moments[2] += w * 0.5 * ctx->v_0 * ctx->v_0 * v2;
434         e_grid[grid] += w * 0.5 * ctx->v_0 * ctx->v_0 * v2;
435       }
436       PetscCall(DMSwarmRestoreField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
437       PetscCall(DMSwarmRestoreField(sw, "w_q", &bs, &dtype, (void **)&wp));
438     }
439     // entropy
440     const PetscReal N_inv = 1 / moments[0];
441     for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
442       PetscDataType dtype;
443       PetscReal    *wp, *coords;
444       DM            sw = printCtx->globSwarmArray[LAND_PACK_IDX(v_id, grid)];
445       PetscInt      bs, NN;
446       PetscCall(DMSwarmGetField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
447       PetscCall(DMSwarmGetLocalSize(sw, &NN));
448       PetscCall(DMSwarmGetField(sw, "w_q", &bs, &dtype, (void **)&wp));
449       for (int pp = 0; pp < NN; pp++) {
450         PetscReal fact = (dim == 2) ? 2.0 * PETSC_PI * coords[pp * dim + 0] : 1, w = fact * wp[pp] * ctx->n_0 * ctx->masses[ctx->species_offset[grid]], ww = w * N_inv;
451         if (w > PETSC_REAL_MIN) {
452           moments[3] -= ww * PetscLogReal(ww);
453         } else moments[4] -= w;
454       }
455       PetscCall(DMSwarmRestoreField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
456       PetscCall(DMSwarmRestoreField(sw, "w_q", &bs, &dtype, (void **)&wp));
457     }
458     PetscCall(PetscInfo(X, "%4d) time %e, Landau particle moments: 0: %18.12e 1: %19.12e 2: %18.12e entropy: %e loss %e. energy = %e + %e + %e\n", (int)n, (double)t, (double)moments[0], (double)moments[1], (double)moments[2], (double)moments[3], (double)(moments[4] / moments[0]), (double)e_grid[0], (double)e_grid[1], (double)e_grid[2]));
459   }
460   if (printCtx->print && printCtx->g_target >= 0) {
461     PetscInt         grid   = printCtx->g_target, id;
462     static PetscReal last_t = -100000, period = .5;
463     if (last_t == -100000) last_t = -period + t;
464     if (t >= last_t + period) {
465       last_t = t;
466       PetscCall(DMGetOutputSequenceNumber(ctx->plex[grid], &id, NULL));
467       PetscCall(DMSetOutputSequenceNumber(ctx->plex[grid], id + 1, t));
468       PetscCall(VecViewFromOptions(printCtx->globXArray[LAND_PACK_IDX(v_id % ctx->batch_sz, grid)], NULL, "-ex30_vec_view"));
469       if (ctx->num_grids > grid + 1) {
470         PetscCall(DMSetOutputSequenceNumber(ctx->plex[grid + 1], id + 1, t));
471         PetscCall(VecViewFromOptions(printCtx->globXArray[LAND_PACK_IDX(v_id % ctx->batch_sz, grid + 1)], NULL, "-ex30_vec_view2"));
472       }
473       PetscCall(PetscInfo(X, "%4d) time %e View\n", (int)n, (double)t));
474     }
475   }
476   PetscCall(DMCompositeRestoreAccessArray(pack, X, nDMs, NULL, printCtx->globXArray));
477   PetscFunctionReturn(PETSC_SUCCESS);
478 }
479 
480 PetscErrorCode go(TS ts, Vec X, const PetscInt num_vertices, const PetscInt a_Np, const PetscInt dim, const PetscInt v_target, const PetscInt g_target, PetscReal shift, PetscBool use_uniform_particle_grid)
481 {
482   DM             pack, *globSwarmArray, grid_dm[LANDAU_MAX_GRIDS];
483   Mat           *globMpArray, g_Mass[LANDAU_MAX_GRIDS];
484   KSP            t_ksp[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS];
485   Vec            t_fhat[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS];
486   PetscInt       nDMs;
487   PetscErrorCode ierr = (PetscErrorCode)0; // used for inside thread loops
488 #if defined(PETSC_HAVE_OPENMP) && defined(PETSC_HAVE_THREADSAFETY)
489   PetscInt numthreads = PetscNumOMPThreads;
490 #else
491   PetscInt numthreads = 1;
492 #endif
493   LandauCtx *ctx;
494   Vec       *globXArray;
495   PetscReal  moments_0[5], moments_1a[5], moments_1b[5], dt_init;
496   PrintCtx  *printCtx;
497 
498   PetscFunctionBeginUser;
499   PetscCheck(numthreads <= EX30_MAX_NUM_THRDS, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "Too many threads %" PetscInt_FMT " > %d", numthreads, EX30_MAX_NUM_THRDS);
500   PetscCheck(numthreads > 0, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "Number threads %" PetscInt_FMT " > %d", numthreads, EX30_MAX_NUM_THRDS);
501   PetscCall(TSGetDM(ts, &pack));
502   PetscCall(DMGetApplicationContext(pack, &ctx));
503   PetscCheck(ctx->batch_sz % numthreads == 0, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "batch size (-dm_landau_batch_size) %" PetscInt_FMT "  mod #threads %" PetscInt_FMT " must equal zero", ctx->batch_sz, numthreads);
504   PetscCall(DMCompositeGetNumberDM(pack, &nDMs)); // number of vertices * number of grids
505   PetscCall(PetscInfo(pack, "Have %" PetscInt_FMT " total grids, with %" PetscInt_FMT " Landau local batched and %" PetscInt_FMT " global items (vertices) %d DMs\n", ctx->num_grids, ctx->batch_sz, num_vertices, (int)nDMs));
506   PetscCall(PetscMalloc(sizeof(*globXArray) * nDMs, &globXArray));
507   PetscCall(PetscMalloc(sizeof(*globMpArray) * nDMs, &globMpArray));
508   PetscCall(PetscMalloc(sizeof(*globSwarmArray) * nDMs, &globSwarmArray));
509   // print ctx
510   PetscCall(PetscNew(&printCtx));
511   PetscCall(TSSetApplicationContext(ts, printCtx));
512   printCtx->v_target       = v_target;
513   printCtx->g_target       = g_target;
514   printCtx->ctx            = ctx;
515   printCtx->globSwarmArray = globSwarmArray;
516   printCtx->grid_dm        = grid_dm;
517   printCtx->globMpArray    = globMpArray;
518   printCtx->g_Mass         = g_Mass;
519   printCtx->globXArray     = globXArray;
520   printCtx->print_entropy  = PETSC_FALSE;
521   PetscOptionsBegin(PETSC_COMM_SELF, "", "Print Options", "DMPLEX");
522   PetscCall(PetscOptionsBool("-print_entropy", "Print entropy and moments at each time step", "ex30.c", printCtx->print_entropy, &printCtx->print_entropy, NULL));
523   PetscOptionsEnd();
524   // view
525   PetscCall(DMViewFromOptions(ctx->plex[g_target], NULL, "-ex30_dm_view"));
526   if (ctx->num_grids > g_target + 1) { PetscCall(DMViewFromOptions(ctx->plex[g_target + 1], NULL, "-ex30_dm_view2")); }
527   // create mesh mass matrices
528   PetscCall(VecZeroEntries(X));
529   PetscCall(DMCompositeGetAccessArray(pack, X, nDMs, NULL, globXArray)); // just to duplicate
530   for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {               // add same particels for all grids
531     Vec          subX = globXArray[LAND_PACK_IDX(0, grid)];
532     DM           dm   = ctx->plex[grid];
533     PetscSection s;
534     grid_dm[grid] = dm;
535     PetscCall(DMCreateMassMatrix(dm, dm, &g_Mass[grid]));
536     //
537     PetscCall(DMGetLocalSection(dm, &s));
538     PetscCall(DMPlexCreateClosureIndex(dm, s));
539     for (int tid = 0; tid < numthreads; tid++) {
540       PC pc;
541       PetscCall(VecDuplicate(subX, &t_fhat[grid][tid]));
542       PetscCall(KSPCreate(PETSC_COMM_SELF, &t_ksp[grid][tid]));
543       PetscCall(KSPSetType(t_ksp[grid][tid], KSPCG));
544       PetscCall(KSPGetPC(t_ksp[grid][tid], &pc));
545       PetscCall(PCSetType(pc, PCJACOBI));
546       PetscCall(KSPSetOptionsPrefix(t_ksp[grid][tid], "ptof_"));
547       PetscCall(KSPSetOperators(t_ksp[grid][tid], g_Mass[grid], g_Mass[grid]));
548       PetscCall(KSPSetFromOptions(t_ksp[grid][tid]));
549     }
550   }
551   PetscCall(DMCompositeRestoreAccessArray(pack, X, nDMs, NULL, globXArray));
552   PetscCall(TSGetTimeStep(ts, &dt_init)); // we could have an adaptive time stepper
553   // loop over all vertices in chucks that are batched for TSSolve
554   for (int i = 0; i < 5; i++) moments_0[i] = moments_1a[i] = moments_1b[i] = 0;
555   for (PetscInt global_vertex_id_0 = 0; global_vertex_id_0 < num_vertices; global_vertex_id_0 += ctx->batch_sz, shift /= 2) { // outer vertex loop
556     PetscCall(TSSetTime(ts, 0));
557     PetscCall(TSSetStepNumber(ts, 0));
558     PetscCall(TSSetTimeStep(ts, dt_init));
559     PetscCall(DMCompositeGetAccessArray(pack, X, nDMs, NULL, globXArray));
560     printCtx->global_vertex_id_0 = global_vertex_id_0;
561     if (v_target >= global_vertex_id_0 && v_target < global_vertex_id_0 + ctx->batch_sz) {
562       PetscCall(PetscObjectSetName((PetscObject)globXArray[LAND_PACK_IDX(v_target % ctx->batch_sz, g_target)], "rho"));
563       printCtx->print = PETSC_TRUE;
564     } else printCtx->print = PETSC_FALSE;
565     // create fake particles in batches with threads
566     for (PetscInt v_id_0 = 0; v_id_0 < ctx->batch_sz; v_id_0 += numthreads) {
567       PetscReal *xx_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS], *yy_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS], *zz_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS], *wp_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS] /* , radiuses[80000] */;
568       PetscInt   Np_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS];
569       // make particles
570       for (int tid = 0; tid < numthreads; tid++) {
571         const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
572         if (glb_v_id < num_vertices) {                                                     // the ragged edge (in last batch)
573           PetscInt Npp0 = a_Np + (glb_v_id % (a_Np / 10 + 1)), nTargetP[LANDAU_MAX_GRIDS]; // n of particels in each dim with load imbalance
574           for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {                         // add same particels for all grids
575             // for (PetscInt sp = ctx->species_offset[grid], i0 = 0; sp < ctx->species_offset[grid + 1]; sp++, i0++) {
576             const PetscReal kT_m  = ctx->k * ctx->thermal_temps[ctx->species_offset[grid]] / ctx->masses[ctx->species_offset[grid]] / (ctx->v_0 * ctx->v_0);                      /* theta = 2kT/mc^2 per species */
577             PetscReal       lo[3] = {-ctx->radius[grid], -ctx->radius[grid], -ctx->radius[grid]}, hi[3] = {ctx->radius[grid], ctx->radius[grid], ctx->radius[grid]}, hp[3], vole; // would be nice to get box from DM
578             PetscInt        Npi = Npp0, Npj = 2 * Npp0, Npk = 1;
579             PetscRandom     rand;
580             PetscReal       sigma = ctx->thermal_speed[grid] / ctx->thermal_speed[0], p2_shift = grid == 0 ? shift : -shift; // symmetric shift of e vs ions
581             PetscCall(PetscRandomCreate(PETSC_COMM_SELF, &rand));
582             PetscCall(PetscRandomSetInterval(rand, 0., 1.));
583             PetscCall(PetscRandomSetFromOptions(rand));
584             if (dim == 2) lo[0] = 0; // Landau coordinate (r,z)
585             else Npi = Npj = Npk = Npp0;
586             // User: use glb_v_id to index into your data
587             const PetscInt NNreal = Npi * Npj * Npk, NN = NNreal + (dim == 2 ? 3 : 6); // make room for bounding box
588             Np_t[grid][tid] = NN;
589             if (glb_v_id == v_target) nTargetP[grid] = NN;
590             PetscCall(PetscMalloc4(NN, &xx_t[grid][tid], NN, &yy_t[grid][tid], NN, &wp_t[grid][tid], dim == 2 ? 1 : NN, &zz_t[grid][tid]));
591             hp[0] = (hi[0] - lo[0]) / Npi;
592             hp[1] = (hi[1] - lo[1]) / Npj;
593             hp[2] = (hi[2] - lo[2]) / Npk;
594             if (dim == 2) hp[2] = 1;
595             PetscCall(PetscInfo(pack, " lo = %14.7e, hi = %14.7e; hp = %14.7e, %14.7e; kT_m = %g; \n", (double)lo[1], (double)hi[1], (double)hp[0], (double)hp[1], (double)kT_m)); // temp
596             vole = hp[0] * hp[1] * hp[2] * ctx->n[grid];                                                                                                                           // fix for multi-species
597             PetscCall(PetscInfo(pack, "Vertex %" PetscInt_FMT ", grid %" PetscInt_FMT " with %" PetscInt_FMT " particles (diagnostic target = %" PetscInt_FMT ")\n", glb_v_id, grid, NN, v_target));
598             for (int pj = 0, pp = 0; pj < Npj; pj++) {
599               for (int pk = 0; pk < Npk; pk++) {
600                 for (int pi = 0; pi < Npi; pi++, pp++) {
601                   PetscReal p_shift   = p2_shift;
602                   wp_t[grid][tid][pp] = 0;
603                   if (use_uniform_particle_grid) {
604                     xx_t[grid][tid][pp] = lo[0] + hp[0] / 2.0 + pi * hp[0];
605                     yy_t[grid][tid][pp] = lo[1] + hp[1] / 2.0 + pj * hp[1];
606                     if (dim == 3) zz_t[grid][tid][pp] = lo[2] + hp[2] / 2.0 + pk * hp[2];
607                     PetscReal x[] = {xx_t[grid][tid][pp], yy_t[grid][tid][pp], dim == 2 ? 0 : zz_t[grid][tid][pp]};
608                     p_shift *= ctx->thermal_speed[grid] / ctx->v_0;
609                     if (ctx->sphere && PetscSqrtReal(PetscSqr(xx_t[grid][tid][pp]) + PetscSqr(yy_t[grid][tid][pp])) > 0.92 * hi[0]) {
610                       wp_t[grid][tid][pp] = 0;
611                     } else {
612                       maxwellian(dim, x, kT_m, vole, p_shift, &wp_t[grid][tid][pp]);
613                       if (ctx->num_grids == 1 && shift != 0) {                          // bi-maxwellian, electron plasma
614                         maxwellian(dim, x, kT_m, vole, -p_shift, &wp_t[grid][tid][pp]); // symmetric shift of electron plasma
615                       }
616                     }
617                   } else {
618                     PetscReal u1, u2;
619                     do {
620                       do {
621                         PetscCall(PetscRandomGetValueReal(rand, &u1));
622                       } while (u1 == 0);
623                       PetscCall(PetscRandomGetValueReal(rand, &u2));
624                       //compute z0 and z1
625                       PetscReal mag       = sigma * PetscSqrtReal(-2.0 * PetscLogReal(u1)); // is this the same scale grid Maxwellian? t_therm = sigma
626                       xx_t[grid][tid][pp] = mag * PetscCosReal(2.0 * PETSC_PI * u2);
627                       yy_t[grid][tid][pp] = mag * PetscSinReal(2.0 * PETSC_PI * u2);
628                       if (dim == 2 && xx_t[grid][tid][pp] < lo[0]) xx_t[grid][tid][pp] = -xx_t[grid][tid][pp];
629                       if (dim == 3) zz_t[grid][tid][pp] = lo[2] + hp[2] / 2.0 + pk * hp[2];
630                       if (!ctx->sphere) {
631                         if (dim == 2 && xx_t[grid][tid][pp] < 0) xx_t[grid][tid][pp] = -xx_t[grid][tid][pp]; // ???
632                         else if (dim == 3) {
633                           while (zz_t[grid][tid][pp] >= hi[2] || zz_t[grid][tid][pp] <= lo[2]) zz_t[grid][tid][pp] *= .9;
634                         }
635                         while (xx_t[grid][tid][pp] >= hi[0] || xx_t[grid][tid][pp] <= lo[0]) xx_t[grid][tid][pp] *= .9;
636                         while (yy_t[grid][tid][pp] >= hi[1] || yy_t[grid][tid][pp] <= lo[1]) yy_t[grid][tid][pp] *= .9;
637                       } else { // 2D
638                         //if (glb_v_id == v_target && pp < 80000) radiuses[pp] = PetscSqrtReal(PetscSqr(xx_t[grid][tid][pp]) + PetscSqr(yy_t[grid][tid][pp]));
639                         while (PetscSqrtReal(PetscSqr(xx_t[grid][tid][pp]) + PetscSqr(yy_t[grid][tid][pp])) > 0.92 * hi[0]) { // safety factor for facets of sphere
640                           xx_t[grid][tid][pp] *= .9;
641                           yy_t[grid][tid][pp] *= .9;
642                         }
643                       }
644                       if (ctx->num_grids == 1 && pp % 2 == 0) p_shift = 0; // one species, split bi-max
645                       p_shift *= ctx->thermal_speed[grid] / ctx->v_0;
646                       if (dim == 3) zz_t[grid][tid][pp] += p_shift;
647                       else yy_t[grid][tid][pp] += p_shift;
648                       wp_t[grid][tid][pp] += ctx->n[grid] / NNreal * PetscSqrtReal(ctx->masses[ctx->species_offset[grid]] / ctx->masses[0]);
649                       if (p_shift <= 0) break; // add bi-max for electron plasma only
650                       p_shift = -p_shift;
651                     } while (ctx->num_grids == 1); // add bi-max for electron plasma only
652                   }
653                   {
654                     if (glb_v_id == v_target) {
655                       PetscReal x[] = {xx_t[grid][tid][pp], yy_t[grid][tid][pp], dim == 2 ? 0 : zz_t[grid][tid][pp]};
656                       PetscReal v2 = 0, fact = dim == 2 ? 2.0 * PETSC_PI * x[0] : 1, w = fact * wp_t[grid][tid][pp] * ctx->n_0 * ctx->masses[ctx->species_offset[grid]];
657                       for (int i = 0; i < dim; ++i) v2 += PetscSqr(x[i]);
658                       moments_0[0] += w;                   // not thread safe
659                       moments_0[1] += w * ctx->v_0 * x[1]; // z-momentum
660                       moments_0[2] += w * 0.5 * ctx->v_0 * ctx->v_0 * v2;
661                     }
662                   }
663                 }
664               }
665             }
666             if (dim == 2) { // fix bounding box
667               int pp                = NNreal;
668               wp_t[grid][tid][pp]   = 0;
669               xx_t[grid][tid][pp]   = 1.e-7;
670               yy_t[grid][tid][pp++] = hi[1] - 5.e-7;
671               wp_t[grid][tid][pp]   = 0;
672               xx_t[grid][tid][pp]   = hi[0] - 5.e-7;
673               yy_t[grid][tid][pp++] = 0;
674               wp_t[grid][tid][pp]   = 0;
675               xx_t[grid][tid][pp]   = 1.e-7;
676               yy_t[grid][tid][pp++] = lo[1] + 5.e-7;
677             } else {
678               const int p0 = NNreal;
679               for (int pj = 0; pj < 6; pj++) { xx_t[grid][tid][p0 + pj] = yy_t[grid][tid][p0 + pj] = zz_t[grid][tid][p0 + pj] = wp_t[grid][tid][p0 + pj] = 0; }
680               xx_t[grid][tid][p0 + 0] = lo[0];
681               xx_t[grid][tid][p0 + 1] = hi[0];
682               yy_t[grid][tid][p0 + 2] = lo[1];
683               yy_t[grid][tid][p0 + 3] = hi[1];
684               zz_t[grid][tid][p0 + 4] = lo[2];
685               zz_t[grid][tid][p0 + 5] = hi[2];
686             }
687             PetscCall(PetscRandomDestroy(&rand));
688           }
689           // entropy init, need global n
690           if (glb_v_id == v_target) {
691             const PetscReal N_inv = 1 / moments_0[0];
692             PetscCall(PetscInfo(pack, "Target %" PetscInt_FMT " with %" PetscInt_FMT " particels\n", glb_v_id, nTargetP[0]));
693             for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
694               const PetscInt NN = nTargetP[grid];
695               for (int pp = 0; pp < NN; pp++) {
696                 const PetscReal fact = dim == 2 ? 2.0 * PETSC_PI * xx_t[grid][tid][pp] : 1, w = fact * ctx->n_0 * ctx->masses[ctx->species_offset[grid]] * wp_t[grid][tid][pp], ww = w * N_inv;
697                 if (w > PETSC_REAL_MIN) {
698                   moments_0[3] -= ww * PetscLogReal(ww);
699                   PetscCheck(ww < 1 - PETSC_MACHINE_EPSILON, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "ww (%g) > 1", (double)ww);
700                 } else moments_0[4] -= w;
701               }
702             } // grid
703           } // target
704         } // active
705       } // threads
706       /* Create particle swarm */
707       for (int tid = 0; tid < numthreads; tid++) {
708         const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
709         if (glb_v_id < num_vertices) {                             // the ragged edge of the last batch
710           for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
711             PetscSection section;
712             PetscInt     Nf;
713             DM           dm = grid_dm[grid];
714             PetscCall(DMGetLocalSection(dm, &section));
715             PetscCall(PetscSectionGetNumFields(section, &Nf));
716             PetscCheck(Nf == 1, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "Only on species per grid supported -- todo");
717             PetscCall(DMViewFromOptions(dm, NULL, "-dm_view"));
718             PetscCall(PetscInfo(pack, "call createSwarm [%" PetscInt_FMT ".%" PetscInt_FMT "] local block index %" PetscInt_FMT "\n", v_id, grid, LAND_PACK_IDX(v_id, grid)));
719             PetscCall(createSwarm(dm, dim, &globSwarmArray[LAND_PACK_IDX(v_id, grid)]));
720           }
721         } // active
722       } // threads
723       PetscCheck(ierr != 9999, PETSC_COMM_WORLD, PETSC_ERR_PLIB, "Only support one species per grid");
724       // make globMpArray
725       PetscPragmaOMP(parallel for)
726       for (int tid = 0; tid < numthreads; tid++) {
727         const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
728         if (glb_v_id < num_vertices) {
729           for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
730             // for (PetscInt sp = ctx->species_offset[grid], i0 = 0; sp < ctx->species_offset[grid + 1]; sp++, i0++) -- loop over species for Nf > 1 -- TODO
731             PetscErrorCode ierr_t;
732             DM             sw = globSwarmArray[LAND_PACK_IDX(v_id, grid)];
733             ierr_t            = PetscInfo(pack, "makeSwarm %" PetscInt_FMT ".%" PetscInt_FMT ") for block %" PetscInt_FMT "\n", v_id, grid, LAND_PACK_IDX(v_id, grid));
734             ierr_t            = makeSwarm(sw, dim, Np_t[grid][tid], xx_t[grid][tid], yy_t[grid][tid], zz_t[grid][tid]);
735             if (ierr_t) ierr = ierr_t;
736           }
737         }
738       }
739       for (int tid = 0; tid < numthreads; tid++) {
740         const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
741         if (glb_v_id < num_vertices) {
742           for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
743             DM dm = grid_dm[grid];
744             DM sw = globSwarmArray[LAND_PACK_IDX(v_id, grid)];
745             PetscCall(PetscInfo(pack, "createMp %" PetscInt_FMT ".%" PetscInt_FMT ") for block %" PetscInt_FMT "\n", v_id, grid, LAND_PACK_IDX(v_id, grid)));
746             PetscCall(createMp(dm, sw, &globMpArray[LAND_PACK_IDX(v_id, grid)]));
747             PetscCall(MatViewFromOptions(globMpArray[LAND_PACK_IDX(v_id, grid)], NULL, "-mp_mat_view"));
748           }
749         }
750       }
751       // p --> g: set X
752       // PetscPragmaOMP(parallel for)
753       for (int tid = 0; tid < numthreads; tid++) {
754         const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
755         if (glb_v_id < num_vertices) {
756           for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
757             PetscErrorCode ierr_t;
758             DM             dm   = grid_dm[grid];
759             DM             sw   = globSwarmArray[LAND_PACK_IDX(v_id, grid)];
760             Vec            subX = globXArray[LAND_PACK_IDX(v_id, grid)], work = t_fhat[grid][tid];
761             ierr_t = PetscInfo(pack, "particlesToGrid %" PetscInt_FMT ".%" PetscInt_FMT ") for block %" PetscInt_FMT "\n", v_id, grid, LAND_PACK_IDX(v_id, grid));
762             ierr_t = particlesToGrid(dm, sw, tid, dim, wp_t[grid][tid], subX, globMpArray[LAND_PACK_IDX(v_id, grid)]);
763             if (ierr_t) ierr = ierr_t;
764             // u = M^_1 f_w
765             ierr_t = VecCopy(subX, work);
766             ierr_t = KSPSolve(t_ksp[grid][tid], work, subX);
767             if (ierr_t) ierr = ierr_t;
768           }
769         }
770       }
771       PetscCheck(!ierr, PETSC_COMM_WORLD, PETSC_ERR_PLIB, "Error in OMP loop. ierr = %d", (int)ierr);
772       /* Cleanup */
773       for (int tid = 0; tid < numthreads; tid++) {
774         const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
775         if (glb_v_id < num_vertices) {
776           for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
777             PetscCall(PetscFree4(xx_t[grid][tid], yy_t[grid][tid], wp_t[grid][tid], zz_t[grid][tid]));
778           }
779         } // active
780       } // threads
781     } // (fake) particle loop
782     // standard view of initial conditions
783     if (v_target >= global_vertex_id_0 && v_target < global_vertex_id_0 + ctx->batch_sz) {
784       PetscCall(DMSetOutputSequenceNumber(ctx->plex[g_target], 0, 0.0));
785       PetscCall(VecViewFromOptions(globXArray[LAND_PACK_IDX(v_target % ctx->batch_sz, g_target)], NULL, "-ex30_vec_view"));
786       if (ctx->num_grids > g_target + 1) {
787         PetscCall(DMSetOutputSequenceNumber(ctx->plex[g_target + 1], 0, 0.0));
788         PetscCall(VecViewFromOptions(globXArray[LAND_PACK_IDX(v_target % ctx->batch_sz, g_target + 1)], NULL, "-ex30_vec_view2"));
789       }
790       PetscCall(MatViewFromOptions(globMpArray[LAND_PACK_IDX(v_target % ctx->batch_sz, g_target)], NULL, "-ex30_mass_mat_view"));
791       PetscCall(DMViewFromOptions(globSwarmArray[LAND_PACK_IDX(v_target % ctx->batch_sz, g_target)], NULL, "-ex30_sw_view"));
792       PetscCall(DMSwarmViewXDMF(globSwarmArray[LAND_PACK_IDX(v_target % ctx->batch_sz, g_target)], "initial_swarm.xmf")); // writes a file by default!!!
793     }
794     // coarse graining moments_1a, bring f back from grid before advance
795     if (v_target >= global_vertex_id_0 && v_target < global_vertex_id_0 + ctx->batch_sz && printCtx->print_entropy) {
796       const PetscInt v_id = v_target % ctx->batch_sz;
797       for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
798         PetscDataType dtype;
799         PetscReal    *wp, *coords;
800         DM            sw = globSwarmArray[LAND_PACK_IDX(v_id, grid)];
801         Vec           work, subX = globXArray[LAND_PACK_IDX(v_id, grid)];
802         PetscInt      bs, NN;
803         // C-G moments
804         PetscCall(VecDuplicate(subX, &work));
805         PetscCall(gridToParticles(grid_dm[grid], sw, subX, work, globMpArray[LAND_PACK_IDX(v_id, grid)], g_Mass[grid]));
806         PetscCall(VecDestroy(&work));
807         // moments
808         PetscCall(DMSwarmGetField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
809         PetscCall(DMSwarmGetLocalSize(sw, &NN));
810         PetscCall(DMSwarmGetField(sw, "w_q", &bs, &dtype, (void **)&wp));
811         for (int pp = 0; pp < NN; pp++) {
812           PetscReal v2 = 0, fact = (dim == 2) ? 2.0 * PETSC_PI * coords[pp * dim + 0] : 1, w = fact * wp[pp] * ctx->n_0 * ctx->masses[ctx->species_offset[grid]];
813           for (int i = 0; i < dim; ++i) v2 += PetscSqr(coords[pp * dim + i]);
814           moments_1a[0] += w;
815           moments_1a[1] += w * ctx->v_0 * coords[pp * dim + 1]; // z-momentum
816           moments_1a[2] += w * 0.5 * ctx->v_0 * ctx->v_0 * v2;
817         }
818         PetscCall(DMSwarmRestoreField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
819         PetscCall(DMSwarmRestoreField(sw, "w_q", &bs, &dtype, (void **)&wp));
820       }
821       // entropy
822       const PetscReal N_inv = 1 / moments_1a[0];
823       PetscCall(PetscInfo(pack, "Entropy batch %" PetscInt_FMT " of %" PetscInt_FMT ", n = %g\n", v_target, num_vertices, (double)(1 / N_inv)));
824       for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
825         PetscDataType dtype;
826         PetscReal    *wp, *coords;
827         DM            sw = globSwarmArray[LAND_PACK_IDX(v_id, grid)];
828         PetscInt      bs, NN;
829         PetscCall(DMSwarmGetLocalSize(sw, &NN));
830         PetscCall(DMSwarmGetField(sw, "w_q", &bs, &dtype, (void **)&wp));
831         PetscCall(DMSwarmGetField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
832         for (int pp = 0; pp < NN; pp++) {
833           PetscReal fact = (dim == 2) ? 2.0 * PETSC_PI * coords[pp * dim + 0] : 1, w = fact * wp[pp] * ctx->n_0 * ctx->masses[ctx->species_offset[grid]], ww = w * N_inv;
834           if (w > PETSC_REAL_MIN) {
835             moments_1a[3] -= ww * PetscLogReal(ww);
836             PetscCheck(ww < 1 - PETSC_MACHINE_EPSILON, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "ww (%g) > 1", (double)ww);
837           } else moments_1a[4] -= w;
838         }
839         PetscCall(DMSwarmRestoreField(sw, "w_q", &bs, &dtype, (void **)&wp));
840         PetscCall(DMSwarmRestoreField(sw, "DMSwarmPIC_coor", &bs, &dtype, (void **)&coords));
841       }
842     }
843     // restore vector
844     PetscCall(DMCompositeRestoreAccessArray(pack, X, nDMs, NULL, globXArray));
845     // view initial grid
846     if (v_target >= global_vertex_id_0 && v_target < global_vertex_id_0 + ctx->batch_sz) { PetscCall(DMPlexLandauPrintNorms(X, 0)); }
847     // advance
848     PetscCall(TSSetSolution(ts, X));
849     PetscCall(PetscInfo(pack, "Advance vertex %" PetscInt_FMT " to %" PetscInt_FMT "\n", global_vertex_id_0, global_vertex_id_0 + ctx->batch_sz));
850     PetscCall(TSSetPostStep(ts, PostStep));
851     PetscCall(PostStep(ts));
852     PetscCall(TSSolve(ts, X));
853     // view
854     PetscCall(DMCompositeGetAccessArray(pack, X, nDMs, NULL, globXArray));
855     if (v_target >= global_vertex_id_0 && v_target < global_vertex_id_0 + ctx->batch_sz) {
856       /* Visualize original particle field */
857       DM  sw = globSwarmArray[LAND_PACK_IDX(v_target % ctx->batch_sz, g_target)];
858       Vec f;
859       PetscCall(DMSetOutputSequenceNumber(sw, 0, 0.0));
860       PetscCall(DMViewFromOptions(grid_dm[g_target], NULL, "-weights_dm_view"));
861       PetscCall(DMViewFromOptions(sw, NULL, "-weights_sw_view"));
862       PetscCall(DMSwarmCreateGlobalVectorFromField(sw, "w_q", &f));
863       PetscCall(PetscObjectSetName((PetscObject)f, "weights"));
864       PetscCall(VecViewFromOptions(f, NULL, "-weights_vec_view"));
865       PetscCall(DMSwarmDestroyGlobalVectorFromField(sw, "w_q", &f));
866       //
867       PetscCall(DMPlexLandauPrintNorms(X, 1));
868     }
869     if (!use_uniform_particle_grid) { // resample to uniform grid
870       for (PetscInt v_id_0 = 0; v_id_0 < ctx->batch_sz; v_id_0 += numthreads) {
871         PetscReal *xx_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS], *yy_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS], *zz_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS], *wp_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS];
872         PetscInt   Np_t[LANDAU_MAX_GRIDS][EX30_MAX_NUM_THRDS];
873         for (int tid = 0; tid < numthreads; tid++) {
874           const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
875           if (glb_v_id < num_vertices) {
876             // create uniform grid w/o weights & smaller
877             PetscInt Npp0 = (a_Np + (glb_v_id % (a_Np / 10 + 1))) / 2, Nv; // 1/2 of uniform particle grid size
878             for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
879               // for (PetscInt sp = ctx->species_offset[grid], i0 = 0; sp < ctx->species_offset[grid + 1]; sp++, i0++)
880               PetscReal lo[3] = {-ctx->radius[grid], -ctx->radius[grid], -ctx->radius[grid]}, hi[3] = {ctx->radius[grid], ctx->radius[grid], ctx->radius[grid]}, hp[3];
881               PetscInt  Npi = Npp0, Npj = 2 * Npp0, Npk = 1, NN;
882               // delete old particles and particle mass matrix
883               PetscCall(DMDestroy(&globSwarmArray[LAND_PACK_IDX(v_id, grid)]));
884               PetscCall(MatDestroy(&globMpArray[LAND_PACK_IDX(v_id, grid)]));
885               // create fake particles in batches with threads
886               PetscCall(MatGetLocalSize(g_Mass[grid], &Nv, NULL));
887               if (dim == 2) lo[0] = 0;
888               else Npi = Npj = Npk = Npp0;
889               NN = Npi * Npj * Npk + (dim == 2 ? 3 : 6); // make a regular grid of particles Npp x Npp
890               while (Npi * Npj * Npk < Nv) {             // make stable - no LS
891                 Npi++;
892                 Npj++;
893                 Npk++;
894                 NN = Npi * Npj * Npk + (dim == 2 ? 3 : 6);
895               }
896               Np_t[grid][tid] = NN;
897               PetscCall(PetscMalloc4(NN, &xx_t[grid][tid], NN, &yy_t[grid][tid], NN, &wp_t[grid][tid], dim == 2 ? 1 : NN, &zz_t[grid][tid]));
898               hp[0] = (hi[0] - lo[0]) / Npi;
899               hp[1] = (hi[1] - lo[1]) / Npj;
900               hp[2] = (hi[2] - lo[2]) / Npk;
901               if (dim == 2) hp[2] = 1;
902               PetscCall(PetscInfo(pack, "Resampling %d particles, %d vertices\n", (int)NN, (int)Nv)); // temp
903               for (int pj = 0, pp = 0; pj < Npj; pj++) {
904                 for (int pk = 0; pk < Npk; pk++) {
905                   for (int pi = 0; pi < Npi; pi++, pp++) {
906                     wp_t[grid][tid][pp] = 0;
907                     xx_t[grid][tid][pp] = lo[0] + hp[0] / 2.0 + pi * hp[0];
908                     yy_t[grid][tid][pp] = lo[1] + hp[1] / 2.0 + pj * hp[1];
909                     if (dim == 3) zz_t[grid][tid][pp] = lo[2] + hp[2] / 2.0 + pk * hp[2];
910                   }
911                 }
912               }
913               if (dim == 2) { // fix bounding box
914                 int pp                = NN - 3;
915                 wp_t[grid][tid][pp]   = 0;
916                 xx_t[grid][tid][pp]   = 1.e-7;
917                 yy_t[grid][tid][pp++] = hi[1] - 5.e-7;
918                 wp_t[grid][tid][pp]   = 0;
919                 xx_t[grid][tid][pp]   = hi[0] - 5.e-7;
920                 yy_t[grid][tid][pp++] = 0;
921                 wp_t[grid][tid][pp]   = 0;
922                 xx_t[grid][tid][pp]   = 1.e-7;
923                 yy_t[grid][tid][pp++] = lo[1] + 5.e-7;
924               } else {
925                 const int p0 = NN - 6;
926                 for (int pj = 0; pj < 6; pj++) { xx_t[grid][tid][p0 + pj] = yy_t[grid][tid][p0 + pj] = zz_t[grid][tid][p0 + pj] = wp_t[grid][tid][p0 + pj] = 0; }
927                 xx_t[grid][tid][p0 + 0] = lo[0];
928                 xx_t[grid][tid][p0 + 1] = hi[0];
929                 yy_t[grid][tid][p0 + 2] = lo[1];
930                 yy_t[grid][tid][p0 + 3] = hi[1];
931                 zz_t[grid][tid][p0 + 4] = lo[2];
932                 zz_t[grid][tid][p0 + 5] = hi[2];
933               }
934             }
935           } // active
936         } // threads
937         /* Create particle swarm */
938         for (int tid = 0; tid < numthreads; tid++) {
939           const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
940           if (glb_v_id < num_vertices) {                             // the ragged edge of the last batch
941             for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
942               // for (PetscInt sp = ctx->species_offset[grid], i0 = 0; sp < ctx->species_offset[grid + 1]; sp++, i0++) -- loop over species for Nf > 1 -- TODO
943               PetscErrorCode ierr_t;
944               PetscSection   section;
945               PetscInt       Nf;
946               DM             dm = grid_dm[grid];
947               ierr_t            = DMGetLocalSection(dm, &section);
948               ierr_t            = PetscSectionGetNumFields(section, &Nf);
949               if (Nf != 1) ierr_t = (PetscErrorCode)9999;
950               else {
951                 ierr_t = DMViewFromOptions(dm, NULL, "-dm_view");
952                 ierr_t = PetscInfo(pack, "call createSwarm [%" PetscInt_FMT ".%" PetscInt_FMT "] local block index %" PetscInt_FMT "\n", v_id, grid, LAND_PACK_IDX(v_id, grid));
953                 ierr_t = createSwarm(dm, dim, &globSwarmArray[LAND_PACK_IDX(v_id, grid)]);
954               }
955               if (ierr_t) ierr = ierr_t;
956             }
957           } // active
958         } // threads
959         PetscCheck(ierr != 9999, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "Only support one species per grid");
960         PetscCheck(!ierr, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "Error in OMP loop. ierr = %d", (int)ierr);
961         // make globMpArray
962         PetscPragmaOMP(parallel for)
963         for (int tid = 0; tid < numthreads; tid++) {
964           const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
965           if (glb_v_id < num_vertices) {
966             for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
967               // for (PetscInt sp = ctx->species_offset[grid], i0 = 0; sp < ctx->species_offset[grid + 1]; sp++, i0++) -- loop over species for Nf > 1 -- TODO
968               PetscErrorCode ierr_t;
969               DM             sw = globSwarmArray[LAND_PACK_IDX(v_id, grid)];
970               ierr_t            = PetscInfo(pack, "makeSwarm %" PetscInt_FMT ".%" PetscInt_FMT ") for block %" PetscInt_FMT "\n", v_id, grid, LAND_PACK_IDX(v_id, grid));
971               ierr_t            = makeSwarm(sw, dim, Np_t[grid][tid], xx_t[grid][tid], yy_t[grid][tid], zz_t[grid][tid]);
972               if (ierr_t) ierr = ierr_t;
973             }
974           } // active
975         } // threads
976         // create particle mass matrices
977         //PetscPragmaOMP(parallel for)
978         for (int tid = 0; tid < numthreads; tid++) {
979           const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
980           if (glb_v_id < num_vertices) {
981             for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
982               PetscErrorCode ierr_t;
983               DM             dm = grid_dm[grid];
984               DM             sw = globSwarmArray[LAND_PACK_IDX(v_id, grid)];
985               ierr_t            = PetscInfo(pack, "createMp %" PetscInt_FMT ".%" PetscInt_FMT ") for block %" PetscInt_FMT "\n", v_id, grid, LAND_PACK_IDX(v_id, grid));
986               ierr_t            = createMp(dm, sw, &globMpArray[LAND_PACK_IDX(v_id, grid)]);
987               if (ierr_t) ierr = ierr_t;
988             }
989           } // active
990         } // threads
991         PetscCheck(!ierr, PETSC_COMM_WORLD, PETSC_ERR_PLIB, "Error in OMP loop. ierr = %d", (int)ierr);
992         /* Cleanup */
993         for (int tid = 0; tid < numthreads; tid++) {
994           const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
995           if (glb_v_id < num_vertices) {
996             for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
997               PetscCall(PetscFree4(xx_t[grid][tid], yy_t[grid][tid], wp_t[grid][tid], zz_t[grid][tid]));
998             }
999           } // active
1000         } // threads
1001       } // batch
1002       // view
1003       if (v_target >= global_vertex_id_0 && v_target < global_vertex_id_0 + ctx->batch_sz) {
1004         /* Visualize particle field */
1005         DM  sw = globSwarmArray[LAND_PACK_IDX(v_target % ctx->batch_sz, g_target)];
1006         Vec f;
1007         PetscCall(DMSetOutputSequenceNumber(sw, 0, 0.0));
1008         PetscCall(DMViewFromOptions(sw, NULL, "-resampled_weights_sw_view"));
1009         PetscCall(DMSwarmCreateGlobalVectorFromField(sw, "w_q", &f));
1010         PetscCall(PetscObjectSetName((PetscObject)f, "resampled_weights"));
1011         PetscCall(VecViewFromOptions(f, NULL, "-resampled_weights_vec_view"));
1012         PetscCall(DMSwarmDestroyGlobalVectorFromField(sw, "w_q", &f));
1013         PetscCall(DMSwarmViewXDMF(sw, "resampled.xmf"));
1014       }
1015     } // !uniform
1016     // particles to grid, compute moments and entropy, for target vertex only
1017     if (v_target >= global_vertex_id_0 && v_target < global_vertex_id_0 + ctx->batch_sz && printCtx->print_entropy) {
1018       PetscReal energy_error_rel;
1019       PetscCall(gridToParticles_private(grid_dm, globSwarmArray, dim, v_target, numthreads, num_vertices, global_vertex_id_0, globMpArray, g_Mass, t_fhat, moments_1b, globXArray, ctx));
1020       energy_error_rel = PetscAbsReal((moments_1b[2] - moments_0[2])) / moments_0[2];
1021       PetscCall(PetscPrintf(PETSC_COMM_WORLD, "Particle Moments:\t number density      momentum (par)     energy             entropy            negative weights  : # OMP threads %g\n", (double)numthreads));
1022       PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\tInitial:         %18.12e %19.12e %18.12e %18.12e %g %%\n", (double)moments_0[0], (double)moments_0[1], (double)moments_0[2], (double)moments_0[3], 100 * (double)(moments_0[4] / moments_0[0])));
1023       PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\tCoarse-graining: %18.12e %19.12e %18.12e %18.12e %g %%\n", (double)moments_1a[0], (double)moments_1a[1], (double)moments_1a[2], (double)moments_1a[3], 100 * (double)(moments_1a[4] / moments_0[0])));
1024       PetscCall(PetscPrintf(PETSC_COMM_WORLD, "\tLandau:          %18.12e %19.12e %18.12e %18.12e %g %%\n", (double)moments_1b[0], (double)moments_1b[1], (double)moments_1b[2], (double)moments_1b[3], 100 * (double)(moments_1b[4] / moments_0[0])));
1025       PetscCall(PetscPrintf(PETSC_COMM_WORLD, "Coarse-graining entropy generation = %e ; Landau entropy generation = %e\n", (double)(moments_1a[3] - moments_0[3]), (double)(moments_1b[3] - moments_0[3])));
1026       PetscCall(PetscPrintf(PETSC_COMM_WORLD, "(relative) energy conservation: Coarse-graining = %e, Landau = %e (%g %d)\n", (double)(moments_1a[2] - moments_0[2]) / (double)moments_0[2], (double)energy_error_rel, (double)PetscLog10Real(energy_error_rel), (int)(PetscLog10Real(energy_error_rel) + .5)));
1027     }
1028     // restore vector
1029     PetscCall(DMCompositeRestoreAccessArray(pack, X, nDMs, NULL, globXArray));
1030     // cleanup
1031     for (PetscInt v_id_0 = 0; v_id_0 < ctx->batch_sz; v_id_0 += numthreads) {
1032       for (int tid = 0; tid < numthreads; tid++) {
1033         const PetscInt v_id = v_id_0 + tid, glb_v_id = global_vertex_id_0 + v_id;
1034         if (glb_v_id < num_vertices) {
1035           for (PetscInt grid = 0; grid < ctx->num_grids; grid++) {
1036             PetscCall(DMDestroy(&globSwarmArray[LAND_PACK_IDX(v_id, grid)]));
1037             PetscCall(MatDestroy(&globMpArray[LAND_PACK_IDX(v_id, grid)]));
1038           }
1039         }
1040       }
1041     }
1042   } // user batch, not used
1043   /* Cleanup */
1044   PetscCall(PetscFree(globXArray));
1045   PetscCall(PetscFree(globSwarmArray));
1046   PetscCall(PetscFree(globMpArray));
1047   PetscCall(PetscFree(printCtx));
1048   // clean up mass matrices
1049   for (PetscInt grid = 0; grid < ctx->num_grids; grid++) { // add same particels for all grids
1050     PetscCall(MatDestroy(&g_Mass[grid]));
1051     for (int tid = 0; tid < numthreads; tid++) {
1052       PetscCall(VecDestroy(&t_fhat[grid][tid]));
1053       PetscCall(KSPDestroy(&t_ksp[grid][tid]));
1054     }
1055   }
1056   PetscFunctionReturn(PETSC_SUCCESS);
1057 }
1058 
1059 int main(int argc, char **argv)
1060 {
1061   DM         pack;
1062   Vec        X;
1063   PetscInt   dim = 2, num_vertices = 1, Np = 10, v_target = 0, g_target = 0;
1064   TS         ts;
1065   Mat        J;
1066   LandauCtx *ctx;
1067   PetscReal  shift                     = 0;
1068   PetscBool  use_uniform_particle_grid = PETSC_TRUE;
1069 
1070   PetscFunctionBeginUser;
1071   PetscCall(PetscInitialize(&argc, &argv, NULL, help));
1072   // process args
1073   PetscOptionsBegin(PETSC_COMM_SELF, "", "Collision Options", "DMPLEX");
1074   PetscCall(PetscOptionsInt("-dim", "Velocity space dimension", "ex30.c", dim, &dim, NULL));
1075   PetscCall(PetscOptionsInt("-number_spatial_vertices", "Number of user spatial vertices to be batched for Landau", "ex30.c", num_vertices, &num_vertices, NULL));
1076   PetscCall(PetscOptionsInt("-number_particles_per_dimension", "Number of particles per grid, with slight modification per spatial vertex, in each dimension of base Cartesian grid", "ex30.c", Np, &Np, NULL));
1077   PetscCall(PetscOptionsBool("-use_uniform_particle_grid", "Use uniform particle grid", "ex30.c", use_uniform_particle_grid, &use_uniform_particle_grid, NULL));
1078   PetscCall(PetscOptionsInt("-vertex_view_target", "Global vertex for diagnostics", "ex30.c", v_target, &v_target, NULL));
1079   PetscCall(PetscOptionsReal("-e_shift", "Bi-Maxwellian shift", "ex30.c", shift, &shift, NULL));
1080   PetscCheck(v_target < num_vertices, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "Batch to view %" PetscInt_FMT " should be < number of vertices %" PetscInt_FMT, v_target, num_vertices);
1081   PetscCall(PetscOptionsInt("-grid_view_target", "Grid to view with diagnostics", "ex30.c", g_target, &g_target, NULL));
1082   PetscOptionsEnd();
1083   /* Create a mesh */
1084   PetscCall(DMPlexLandauCreateVelocitySpace(PETSC_COMM_SELF, dim, "", &X, &J, &pack));
1085   PetscCall(DMGetApplicationContext(pack, &ctx));
1086   PetscCall(DMSetUp(pack));
1087   PetscCall(DMSetOutputSequenceNumber(pack, 0, 0.0));
1088   PetscCheck(g_target < ctx->num_grids, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "Grid to view %" PetscInt_FMT " should be < number of grids %" PetscInt_FMT, g_target, ctx->num_grids);
1089   PetscCheck(ctx->batch_view_idx == v_target % ctx->batch_sz, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "Global view index %" PetscInt_FMT " mode batch size %" PetscInt_FMT " != ctx->batch_view_idx %" PetscInt_FMT, v_target, ctx->batch_sz, ctx->batch_view_idx);
1090   // PetscCheck(!use_uniform_particle_grid || !ctx->sphere, PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "Can not use -use_uniform_particle_grid and -dm_landau_sphere");
1091   /* Create timestepping solver context */
1092   PetscCall(TSCreate(PETSC_COMM_SELF, &ts));
1093   PetscCall(TSSetDM(ts, pack));
1094   PetscCall(TSSetIFunction(ts, NULL, DMPlexLandauIFunction, NULL));
1095   PetscCall(TSSetIJacobian(ts, J, J, DMPlexLandauIJacobian, NULL));
1096   PetscCall(TSSetExactFinalTime(ts, TS_EXACTFINALTIME_STEPOVER));
1097   PetscCall(TSSetFromOptions(ts));
1098   PetscCall(PetscObjectSetName((PetscObject)X, "X"));
1099   // do particle advance
1100   PetscCall(go(ts, X, num_vertices, Np, dim, v_target, g_target, shift, use_uniform_particle_grid));
1101   PetscCall(MatZeroEntries(J)); // need to zero out so as to not reuse it in Landau's logic
1102   /* clean up */
1103   PetscCall(DMPlexLandauDestroyVelocitySpace(&pack));
1104   PetscCall(TSDestroy(&ts));
1105   PetscCall(VecDestroy(&X));
1106   PetscCall(PetscFinalize());
1107   return 0;
1108 }
1109 
1110 /*TEST
1111 
1112   build:
1113     requires: !complex
1114 
1115   testset:
1116     requires: double defined(PETSC_USE_DMLANDAU_2D)
1117     output_file: output/ex30_0.out
1118     args: -dim 2 -petscspace_degree 3 -dm_landau_num_species_grid 1,1,1 -dm_refine 1 -number_particles_per_dimension 20 \
1119           -dm_landau_batch_size 4 -number_spatial_vertices 6 -vertex_view_target 5 -grid_view_target 1 -dm_landau_batch_view_idx 1 \
1120           -dm_landau_n 1.000018,1,1e-6 -dm_landau_thermal_temps 2,1,1 -dm_landau_ion_masses 2,180 -dm_landau_ion_charges 1,18 \
1121           -ftop_ksp_rtol 1e-10 -ftop_ksp_type lsqr -ftop_pc_type bjacobi -ftop_sub_pc_factor_shift_type nonzero -ftop_sub_pc_type lu -ftop_ksp_error_if_not_converged \
1122           -ksp_type gmres -ksp_error_if_not_converged -dm_landau_verbose 4 -print_entropy \
1123           -ptof_ksp_type cg -ptof_pc_type jacobi -ptof_ksp_rtol 1e-12 -ptof_ksp_error_if_not_converged\
1124           -snes_converged_reason -snes_monitor -snes_rtol 1e-12 -snes_stol 1e-12 \
1125           -ts_dt 0.01 -ts_rtol 1e-1 -ts_exact_final_time stepover -ts_max_snes_failures -1 -ts_max_steps 1 -ts_monitor -ts_type beuler
1126     test:
1127       suffix: cpu
1128       args: -dm_landau_device_type cpu -pc_type jacobi
1129     test:
1130       suffix: kokkos
1131       # failed on Sunspot@ALCF with sycl
1132       requires: kokkos_kernels !openmp !sycl
1133       args: -dm_landau_device_type kokkos -dm_mat_type aijkokkos -dm_vec_type kokkos -pc_type bjkokkos -pc_bjkokkos_ksp_type tfqmr -pc_bjkokkos_pc_type jacobi
1134 
1135   testset:
1136     requires: double !defined(PETSC_USE_DMLANDAU_2D)
1137     output_file: output/ex30_3d.out
1138     args: -dim 3 -petscspace_degree 2 -dm_landau_num_species_grid 1,1 -dm_refine 0 -number_particles_per_dimension 10 -dm_plex_hash_location \
1139           -dm_landau_batch_size 1 -number_spatial_vertices 1 -vertex_view_target 0 -grid_view_target 0 -dm_landau_batch_view_idx 0 \
1140           -dm_landau_n 1.000018,1 -dm_landau_thermal_temps 2,1 -dm_landau_ion_masses 2 -dm_landau_ion_charges 1 \
1141           -ftop_ksp_type cg -ftop_pc_type jacobi -ftop_ksp_rtol 1e-12 -ftop_ksp_error_if_not_converged -ksp_type preonly -pc_type lu -ksp_error_if_not_converged \
1142           -ptof_ksp_type cg -ptof_pc_type jacobi -ptof_ksp_rtol 1e-12 -ptof_ksp_error_if_not_converged \
1143           -snes_converged_reason -snes_monitor -snes_rtol 1e-12 -snes_stol 1e-12 \
1144           -ts_dt 0.1 -ts_exact_final_time stepover -ts_max_snes_failures -1 -ts_max_steps 1 -ts_monitor -ts_type beuler -print_entropy
1145     test:
1146       suffix: cpu_3d
1147       args: -dm_landau_device_type cpu
1148     test:
1149       suffix: kokkos_3d
1150       requires: kokkos_kernels !openmp
1151       args: -dm_landau_device_type kokkos -dm_mat_type aijkokkos -dm_vec_type kokkos -pc_type bjkokkos -pc_bjkokkos_ksp_type tfqmr -pc_bjkokkos_pc_type jacobi
1152 
1153   test:
1154     suffix: conserve
1155     requires: !complex double defined(PETSC_USE_DMLANDAU_2D) !cuda
1156     args: -dm_landau_batch_size 4 -dm_refine 0 -dm_landau_num_species_grid 1 -dm_landau_thermal_temps 1 -petscspace_degree 3 -snes_converged_reason -ts_type beuler -ts_dt .1 \
1157           -ts_max_steps 1 -ksp_type preonly -ksp_error_if_not_converged -snes_rtol 1e-14 -snes_stol 1e-14 -dm_landau_device_type cpu -number_particles_per_dimension 20 \
1158           -ptof_ksp_type cg -ptof_pc_type jacobi -ptof_ksp_rtol 1e-14 -ptof_ksp_error_if_not_converged -pc_type lu -dm_landau_simplex 1 -use_uniform_particle_grid false -dm_landau_sphere -print_entropy -number_particles_per_dimension 50 -ftop_ksp_type cg -ftop_pc_type jacobi -ftop_ksp_rtol 1e-14
1159 
1160 TEST*/
1161