Actual source code: gamg.c
1: /*
2: GAMG geometric-algebric multigrid PC - Mark Adams 2011
3: */
4: #include <../src/ksp/pc/impls/gamg/gamg.h>
5: #include <../src/ksp/ksp/impls/cheby/chebyshevimpl.h>
7: #if defined(PETSC_HAVE_CUDA)
8: #include <cuda_runtime.h>
9: #endif
11: #if defined(PETSC_HAVE_HIP)
12: #include <hip/hip_runtime.h>
13: #endif
15: PetscLogEvent petsc_gamg_setup_events[GAMG_NUM_SET];
16: PetscLogEvent petsc_gamg_setup_matmat_events[PETSC_MG_MAXLEVELS][3];
18: // #define GAMG_STAGES
19: #if defined(GAMG_STAGES)
20: static PetscLogStage gamg_stages[PETSC_MG_MAXLEVELS];
21: #endif
23: static PetscFunctionList GAMGList = NULL;
24: static PetscBool PCGAMGPackageInitialized;
26: static PetscErrorCode PCReset_GAMG(PC pc)
27: {
28: PC_MG *mg = (PC_MG *)pc->data;
29: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
31: PetscFunctionBegin;
32: PetscCall(PetscFree(pc_gamg->data));
33: pc_gamg->data_sz = 0;
34: PetscCall(PetscFree(pc_gamg->orig_data));
35: for (PetscInt level = 0; level < PETSC_MG_MAXLEVELS; level++) {
36: mg->min_eigen_DinvA[level] = 0;
37: mg->max_eigen_DinvA[level] = 0;
38: }
39: pc_gamg->emin = 0;
40: pc_gamg->emax = 0;
41: PetscCall(PCReset_MG(pc));
42: PetscFunctionReturn(PETSC_SUCCESS);
43: }
45: /*
46: PCGAMGCreateLevel_GAMG: create coarse op with RAP. repartition and/or reduce number
47: of active processors.
49: Input Parameter:
50: . pc - parameters + side effect: coarse data in 'pc_gamg->data' and
51: 'pc_gamg->data_sz' are changed via repartitioning/reduction.
52: . Amat_fine - matrix on this fine (k) level
53: . cr_bs - coarse block size
54: In/Output Parameter:
55: . a_P_inout - prolongation operator to the next level (k-->k-1)
56: . a_nactive_proc - number of active procs
57: Output Parameter:
58: . a_Amat_crs - coarse matrix that is created (k-1)
59: */
60: static PetscErrorCode PCGAMGCreateLevel_GAMG(PC pc, Mat Amat_fine, PetscInt cr_bs, Mat *a_P_inout, Mat *a_Amat_crs, PetscMPIInt *a_nactive_proc, IS *Pcolumnperm, PetscBool is_last)
61: {
62: PC_MG *mg = (PC_MG *)pc->data;
63: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
64: Mat Cmat, Pold = *a_P_inout;
65: MPI_Comm comm;
66: PetscMPIInt rank, size, new_size, nactive = *a_nactive_proc;
67: PetscInt ncrs_eq, ncrs, f_bs;
69: PetscFunctionBegin;
70: PetscCall(PetscObjectGetComm((PetscObject)Amat_fine, &comm));
71: PetscCallMPI(MPI_Comm_rank(comm, &rank));
72: PetscCallMPI(MPI_Comm_size(comm, &size));
73: PetscCall(MatGetBlockSize(Amat_fine, &f_bs));
74: PetscCall(PetscLogEventBegin(petsc_gamg_setup_events[GAMG_PTAP], 0, 0, 0, 0));
75: PetscCall(PetscLogEventBegin(petsc_gamg_setup_matmat_events[pc_gamg->current_level][1], 0, 0, 0, 0));
76: PetscCall(MatPtAP(Amat_fine, Pold, MAT_INITIAL_MATRIX, 2.0, &Cmat));
77: PetscCall(PetscLogEventEnd(petsc_gamg_setup_matmat_events[pc_gamg->current_level][1], 0, 0, 0, 0));
78: PetscCall(PetscLogEventEnd(petsc_gamg_setup_events[GAMG_PTAP], 0, 0, 0, 0));
80: if (Pcolumnperm) *Pcolumnperm = NULL;
82: /* set 'ncrs' (nodes), 'ncrs_eq' (equations)*/
83: PetscCall(MatGetLocalSize(Cmat, &ncrs_eq, NULL));
84: if (pc_gamg->data_cell_rows > 0) {
85: ncrs = pc_gamg->data_sz / pc_gamg->data_cell_cols / pc_gamg->data_cell_rows;
86: } else {
87: PetscInt bs;
88: PetscCall(MatGetBlockSize(Cmat, &bs));
89: ncrs = ncrs_eq / bs;
90: }
91: /* get number of PEs to make active 'new_size', reduce, can be any integer 1-P */
92: if (pc_gamg->level_reduction_factors[pc_gamg->current_level] == 0 && PetscDefined(HAVE_CUDA) && pc_gamg->current_level == 0) { /* 0 turns reducing to 1 process/device on; do for HIP, etc. */
93: #if defined(PETSC_HAVE_CUDA)
94: PetscShmComm pshmcomm;
95: PetscMPIInt locrank;
96: MPI_Comm loccomm;
97: PetscInt s_nnodes, r_nnodes, new_new_size;
98: cudaError_t cerr;
99: int devCount;
100: PetscCall(PetscShmCommGet(comm, &pshmcomm));
101: PetscCall(PetscShmCommGetMpiShmComm(pshmcomm, &loccomm));
102: PetscCallMPI(MPI_Comm_rank(loccomm, &locrank));
103: s_nnodes = !locrank;
104: PetscCall(MPIU_Allreduce(&s_nnodes, &r_nnodes, 1, MPIU_INT, MPI_SUM, comm));
105: PetscCheck((size % r_nnodes) == 0, PETSC_COMM_SELF, PETSC_ERR_PLIB, "odd number of nodes np=%d nnodes%" PetscInt_FMT, size, r_nnodes);
106: devCount = 0;
107: cerr = cudaGetDeviceCount(&devCount);
108: cudaGetLastError(); /* Reset the last error */
109: if (cerr == cudaSuccess && devCount >= 1) { /* There are devices, else go to heuristic */
110: new_new_size = r_nnodes * devCount;
111: new_size = new_new_size;
112: PetscCall(PetscInfo(pc, "%s: Fine grid with Cuda. %" PetscInt_FMT " nodes. Change new active set size %d --> %d (devCount=%d #nodes=%" PetscInt_FMT ")\n", ((PetscObject)pc)->prefix, r_nnodes, nactive, new_size, devCount, r_nnodes));
113: } else {
114: PetscCall(PetscInfo(pc, "%s: With Cuda but no device. Use heuristics.\n", ((PetscObject)pc)->prefix));
115: goto HEURISTIC;
116: }
117: #else
118: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_PLIB, "should not be here");
119: #endif
120: } else if (pc_gamg->level_reduction_factors[pc_gamg->current_level] > 0) {
121: if (nactive < pc_gamg->level_reduction_factors[pc_gamg->current_level]) {
122: new_size = 1;
123: PetscCall(PetscInfo(pc, "%s: reduction factor too small for %d active processes: reduce to one process\n", ((PetscObject)pc)->prefix, new_size));
124: } else {
125: PetscCheck(nactive % pc_gamg->level_reduction_factors[pc_gamg->current_level] == 0, PETSC_COMM_SELF, PETSC_ERR_PLIB, "odd number of active process %d wrt reduction factor %" PetscInt_FMT, nactive, pc_gamg->level_reduction_factors[pc_gamg->current_level]);
126: new_size = nactive / pc_gamg->level_reduction_factors[pc_gamg->current_level];
127: PetscCall(PetscInfo(pc, "%s: Manually setting reduction to %d active processes (%d/%" PetscInt_FMT ")\n", ((PetscObject)pc)->prefix, new_size, nactive, pc_gamg->level_reduction_factors[pc_gamg->current_level]));
128: }
129: } else if (is_last && !pc_gamg->use_parallel_coarse_grid_solver) {
130: new_size = 1;
131: PetscCall(PetscInfo(pc, "%s: Force coarsest grid reduction to %d active processes\n", ((PetscObject)pc)->prefix, new_size));
132: } else {
133: PetscInt ncrs_eq_glob;
134: #if defined(PETSC_HAVE_CUDA)
135: HEURISTIC:
136: #endif
137: PetscCall(MatGetSize(Cmat, &ncrs_eq_glob, NULL));
138: new_size = (PetscMPIInt)((float)ncrs_eq_glob / (float)pc_gamg->min_eq_proc + 0.5); /* hardwire min. number of eq/proc */
139: if (!new_size) new_size = 1; /* not likely, possible? */
140: else if (new_size >= nactive) new_size = nactive; /* no change, rare */
141: PetscCall(PetscInfo(pc, "%s: Coarse grid reduction from %d to %d active processes\n", ((PetscObject)pc)->prefix, nactive, new_size));
142: }
143: if (new_size == nactive) {
144: *a_Amat_crs = Cmat; /* output - no repartitioning or reduction - could bail here */
145: if (new_size < size) {
146: /* odd case where multiple coarse grids are on one processor or no coarsening ... */
147: PetscCall(PetscInfo(pc, "%s: reduced grid using same number of processors (%d) as last grid (use larger coarse grid)\n", ((PetscObject)pc)->prefix, nactive));
148: if (pc_gamg->cpu_pin_coarse_grids) {
149: PetscCall(MatBindToCPU(*a_Amat_crs, PETSC_TRUE));
150: PetscCall(MatBindToCPU(*a_P_inout, PETSC_TRUE));
151: }
152: }
153: /* we know that the grid structure can be reused in MatPtAP */
154: } else { /* reduce active processors - we know that the grid structure can NOT be reused in MatPtAP */
155: PetscInt *counts, *newproc_idx, ii, jj, kk, strideNew, *tidx, ncrs_new, ncrs_eq_new, nloc_old, expand_factor = 1, rfactor = 1;
156: IS is_eq_newproc, is_eq_num, is_eq_num_prim, new_eq_indices;
157: PetscCall(PetscLogEventBegin(petsc_gamg_setup_events[GAMG_REDUCE], 0, 0, 0, 0));
158: nloc_old = ncrs_eq / cr_bs;
159: PetscCheck(ncrs_eq % cr_bs == 0, PETSC_COMM_SELF, PETSC_ERR_PLIB, "ncrs_eq %" PetscInt_FMT " not divisible by cr_bs %" PetscInt_FMT, ncrs_eq, cr_bs);
160: /* get new_size and rfactor */
161: if (pc_gamg->layout_type == PCGAMG_LAYOUT_SPREAD || !pc_gamg->repart) {
162: /* find factor */
163: if (new_size == 1) rfactor = size; /* don't modify */
164: else {
165: PetscReal best_fact = 0.;
166: jj = -1;
167: for (kk = 1; kk <= size; kk++) {
168: if (!(size % kk)) { /* a candidate */
169: PetscReal nactpe = (PetscReal)size / (PetscReal)kk, fact = nactpe / (PetscReal)new_size;
170: if (fact > 1.0) fact = 1. / fact; /* keep fact < 1 */
171: if (fact > best_fact) {
172: best_fact = fact;
173: jj = kk;
174: }
175: }
176: }
177: if (jj != -1) rfactor = jj;
178: else rfactor = 1; /* a prime */
179: if (pc_gamg->layout_type == PCGAMG_LAYOUT_COMPACT) expand_factor = 1;
180: else expand_factor = rfactor;
181: }
182: new_size = size / rfactor; /* make new size one that is factor */
183: if (new_size == nactive) { /* no repartitioning or reduction, bail out because nested here (rare) */
184: *a_Amat_crs = Cmat;
185: PetscCall(PetscInfo(pc, "%s: Finding factorable processor set stopped reduction: new_size=%d, neq(loc)=%" PetscInt_FMT "\n", ((PetscObject)pc)->prefix, new_size, ncrs_eq));
186: PetscCall(PetscLogEventEnd(petsc_gamg_setup_events[GAMG_REDUCE], 0, 0, 0, 0));
187: PetscFunctionReturn(PETSC_SUCCESS);
188: }
189: }
190: /* make 'is_eq_newproc' */
191: PetscCall(PetscMalloc1(size, &counts));
192: if (pc_gamg->repart) { /* Repartition Cmat_{k} and move columns of P^{k}_{k-1} and coordinates of primal part accordingly */
193: Mat adj;
194: PetscCall(PetscLogEventBegin(petsc_gamg_setup_events[GAMG_REPART], 0, 0, 0, 0));
195: PetscCall(PetscInfo(pc, "%s: Repartition: size (active): %d --> %d, %" PetscInt_FMT " local equations, using %s process layout\n", ((PetscObject)pc)->prefix, *a_nactive_proc, new_size, ncrs_eq, (pc_gamg->layout_type == PCGAMG_LAYOUT_COMPACT) ? "compact" : "spread"));
196: /* get 'adj' */
197: if (cr_bs == 1) {
198: PetscCall(MatConvert(Cmat, MATMPIADJ, MAT_INITIAL_MATRIX, &adj));
199: } else {
200: /* make a scalar matrix to partition (no Stokes here) */
201: Mat tMat;
202: PetscInt Istart_crs, Iend_crs, ncols, jj, Ii;
203: const PetscScalar *vals;
204: const PetscInt *idx;
205: PetscInt *d_nnz, *o_nnz, M, N;
206: static PetscInt llev = 0; /* ugly but just used for debugging */
207: MatType mtype;
209: PetscCall(PetscMalloc2(ncrs, &d_nnz, ncrs, &o_nnz));
210: PetscCall(MatGetOwnershipRange(Cmat, &Istart_crs, &Iend_crs));
211: PetscCall(MatGetSize(Cmat, &M, &N));
212: for (Ii = Istart_crs, jj = 0; Ii < Iend_crs; Ii += cr_bs, jj++) {
213: PetscCall(MatGetRow(Cmat, Ii, &ncols, NULL, NULL));
214: d_nnz[jj] = ncols / cr_bs;
215: o_nnz[jj] = ncols / cr_bs;
216: PetscCall(MatRestoreRow(Cmat, Ii, &ncols, NULL, NULL));
217: if (d_nnz[jj] > ncrs) d_nnz[jj] = ncrs;
218: if (o_nnz[jj] > (M / cr_bs - ncrs)) o_nnz[jj] = M / cr_bs - ncrs;
219: }
221: PetscCall(MatGetType(Amat_fine, &mtype));
222: PetscCall(MatCreate(comm, &tMat));
223: PetscCall(MatSetSizes(tMat, ncrs, ncrs, PETSC_DETERMINE, PETSC_DETERMINE));
224: PetscCall(MatSetType(tMat, mtype));
225: PetscCall(MatSeqAIJSetPreallocation(tMat, 0, d_nnz));
226: PetscCall(MatMPIAIJSetPreallocation(tMat, 0, d_nnz, 0, o_nnz));
227: PetscCall(PetscFree2(d_nnz, o_nnz));
229: for (ii = Istart_crs; ii < Iend_crs; ii++) {
230: PetscInt dest_row = ii / cr_bs;
231: PetscCall(MatGetRow(Cmat, ii, &ncols, &idx, &vals));
232: for (jj = 0; jj < ncols; jj++) {
233: PetscInt dest_col = idx[jj] / cr_bs;
234: PetscScalar v = 1.0;
235: PetscCall(MatSetValues(tMat, 1, &dest_row, 1, &dest_col, &v, ADD_VALUES));
236: }
237: PetscCall(MatRestoreRow(Cmat, ii, &ncols, &idx, &vals));
238: }
239: PetscCall(MatAssemblyBegin(tMat, MAT_FINAL_ASSEMBLY));
240: PetscCall(MatAssemblyEnd(tMat, MAT_FINAL_ASSEMBLY));
242: if (llev++ == -1) {
243: PetscViewer viewer;
244: char fname[32];
245: PetscCall(PetscSNPrintf(fname, sizeof(fname), "part_mat_%" PetscInt_FMT ".mat", llev));
246: PetscCall(PetscViewerBinaryOpen(comm, fname, FILE_MODE_WRITE, &viewer));
247: PetscCall(MatView(tMat, viewer));
248: PetscCall(PetscViewerDestroy(&viewer));
249: }
250: PetscCall(MatConvert(tMat, MATMPIADJ, MAT_INITIAL_MATRIX, &adj));
251: PetscCall(MatDestroy(&tMat));
252: } /* create 'adj' */
254: { /* partition: get newproc_idx */
255: char prefix[256];
256: const char *pcpre;
257: const PetscInt *is_idx;
258: MatPartitioning mpart;
259: IS proc_is;
261: PetscCall(MatPartitioningCreate(comm, &mpart));
262: PetscCall(MatPartitioningSetAdjacency(mpart, adj));
263: PetscCall(PCGetOptionsPrefix(pc, &pcpre));
264: PetscCall(PetscSNPrintf(prefix, sizeof(prefix), "%spc_gamg_", pcpre ? pcpre : ""));
265: PetscCall(PetscObjectSetOptionsPrefix((PetscObject)mpart, prefix));
266: PetscCall(MatPartitioningSetFromOptions(mpart));
267: PetscCall(MatPartitioningSetNParts(mpart, new_size));
268: PetscCall(MatPartitioningApply(mpart, &proc_is));
269: PetscCall(MatPartitioningDestroy(&mpart));
271: /* collect IS info */
272: PetscCall(PetscMalloc1(ncrs_eq, &newproc_idx));
273: PetscCall(ISGetIndices(proc_is, &is_idx));
274: for (kk = jj = 0; kk < nloc_old; kk++) {
275: for (ii = 0; ii < cr_bs; ii++, jj++) { newproc_idx[jj] = is_idx[kk] * expand_factor; /* distribution */ }
276: }
277: PetscCall(ISRestoreIndices(proc_is, &is_idx));
278: PetscCall(ISDestroy(&proc_is));
279: }
280: PetscCall(MatDestroy(&adj));
282: PetscCall(ISCreateGeneral(comm, ncrs_eq, newproc_idx, PETSC_COPY_VALUES, &is_eq_newproc));
283: PetscCall(PetscFree(newproc_idx));
284: PetscCall(PetscLogEventEnd(petsc_gamg_setup_events[GAMG_REPART], 0, 0, 0, 0));
285: } else { /* simple aggregation of parts -- 'is_eq_newproc' */
286: PetscInt targetPE;
287: PetscCheck(new_size != nactive, PETSC_COMM_SELF, PETSC_ERR_PLIB, "new_size==nactive. Should not happen");
288: PetscCall(PetscInfo(pc, "%s: Number of equations (loc) %" PetscInt_FMT " with simple aggregation\n", ((PetscObject)pc)->prefix, ncrs_eq));
289: targetPE = (rank / rfactor) * expand_factor;
290: PetscCall(ISCreateStride(comm, ncrs_eq, targetPE, 0, &is_eq_newproc));
291: } /* end simple 'is_eq_newproc' */
293: /*
294: Create an index set from the is_eq_newproc index set to indicate the mapping TO
295: */
296: PetscCall(ISPartitioningToNumbering(is_eq_newproc, &is_eq_num));
297: is_eq_num_prim = is_eq_num;
298: /*
299: Determine how many equations/vertices are assigned to each processor
300: */
301: PetscCall(ISPartitioningCount(is_eq_newproc, size, counts));
302: ncrs_eq_new = counts[rank];
303: PetscCall(ISDestroy(&is_eq_newproc));
304: ncrs_new = ncrs_eq_new / cr_bs;
306: PetscCall(PetscFree(counts));
307: /* data movement scope -- this could be moved to subclasses so that we don't try to cram all auxiliary data into some complex abstracted thing */
308: {
309: Vec src_crd, dest_crd;
310: const PetscInt *idx, ndata_rows = pc_gamg->data_cell_rows, ndata_cols = pc_gamg->data_cell_cols, node_data_sz = ndata_rows * ndata_cols;
311: VecScatter vecscat;
312: PetscScalar *array;
313: IS isscat;
314: /* move data (for primal equations only) */
315: /* Create a vector to contain the newly ordered element information */
316: PetscCall(VecCreate(comm, &dest_crd));
317: PetscCall(VecSetSizes(dest_crd, node_data_sz * ncrs_new, PETSC_DECIDE));
318: PetscCall(VecSetType(dest_crd, VECSTANDARD)); /* this is needed! */
319: /*
320: There are 'ndata_rows*ndata_cols' data items per node, (one can think of the vectors of having
321: a block size of ...). Note, ISs are expanded into equation space by 'cr_bs'.
322: */
323: PetscCall(PetscMalloc1(ncrs * node_data_sz, &tidx));
324: PetscCall(ISGetIndices(is_eq_num_prim, &idx));
325: for (ii = 0, jj = 0; ii < ncrs; ii++) {
326: PetscInt id = idx[ii * cr_bs] / cr_bs; /* get node back */
327: for (kk = 0; kk < node_data_sz; kk++, jj++) tidx[jj] = id * node_data_sz + kk;
328: }
329: PetscCall(ISRestoreIndices(is_eq_num_prim, &idx));
330: PetscCall(ISCreateGeneral(comm, node_data_sz * ncrs, tidx, PETSC_COPY_VALUES, &isscat));
331: PetscCall(PetscFree(tidx));
332: /*
333: Create a vector to contain the original vertex information for each element
334: */
335: PetscCall(VecCreateSeq(PETSC_COMM_SELF, node_data_sz * ncrs, &src_crd));
336: for (jj = 0; jj < ndata_cols; jj++) {
337: const PetscInt stride0 = ncrs * pc_gamg->data_cell_rows;
338: for (ii = 0; ii < ncrs; ii++) {
339: for (kk = 0; kk < ndata_rows; kk++) {
340: PetscInt ix = ii * ndata_rows + kk + jj * stride0, jx = ii * node_data_sz + kk * ndata_cols + jj;
341: PetscScalar tt = (PetscScalar)pc_gamg->data[ix];
342: PetscCall(VecSetValues(src_crd, 1, &jx, &tt, INSERT_VALUES));
343: }
344: }
345: }
346: PetscCall(VecAssemblyBegin(src_crd));
347: PetscCall(VecAssemblyEnd(src_crd));
348: /*
349: Scatter the element vertex information (still in the original vertex ordering)
350: to the correct processor
351: */
352: PetscCall(VecScatterCreate(src_crd, NULL, dest_crd, isscat, &vecscat));
353: PetscCall(ISDestroy(&isscat));
354: PetscCall(VecScatterBegin(vecscat, src_crd, dest_crd, INSERT_VALUES, SCATTER_FORWARD));
355: PetscCall(VecScatterEnd(vecscat, src_crd, dest_crd, INSERT_VALUES, SCATTER_FORWARD));
356: PetscCall(VecScatterDestroy(&vecscat));
357: PetscCall(VecDestroy(&src_crd));
358: /*
359: Put the element vertex data into a new allocation of the gdata->ele
360: */
361: PetscCall(PetscFree(pc_gamg->data));
362: PetscCall(PetscMalloc1(node_data_sz * ncrs_new, &pc_gamg->data));
364: pc_gamg->data_sz = node_data_sz * ncrs_new;
365: strideNew = ncrs_new * ndata_rows;
367: PetscCall(VecGetArray(dest_crd, &array));
368: for (jj = 0; jj < ndata_cols; jj++) {
369: for (ii = 0; ii < ncrs_new; ii++) {
370: for (kk = 0; kk < ndata_rows; kk++) {
371: PetscInt ix = ii * ndata_rows + kk + jj * strideNew, jx = ii * node_data_sz + kk * ndata_cols + jj;
372: pc_gamg->data[ix] = PetscRealPart(array[jx]);
373: }
374: }
375: }
376: PetscCall(VecRestoreArray(dest_crd, &array));
377: PetscCall(VecDestroy(&dest_crd));
378: }
379: /* move A and P (columns) with new layout */
380: /*
381: Invert for MatCreateSubMatrix
382: */
383: PetscCall(ISInvertPermutation(is_eq_num, ncrs_eq_new, &new_eq_indices));
384: PetscCall(ISSort(new_eq_indices)); /* is this needed? */
385: PetscCall(ISSetBlockSize(new_eq_indices, cr_bs));
386: if (is_eq_num != is_eq_num_prim) { PetscCall(ISDestroy(&is_eq_num_prim)); /* could be same as 'is_eq_num' */ }
387: if (Pcolumnperm) {
388: PetscCall(PetscObjectReference((PetscObject)new_eq_indices));
389: *Pcolumnperm = new_eq_indices;
390: }
391: PetscCall(ISDestroy(&is_eq_num));
393: /* 'a_Amat_crs' output */
394: {
395: Mat mat;
396: PetscBool isset, isspd, isher;
397: #if !defined(PETSC_USE_COMPLEX)
398: PetscBool issym;
399: #endif
401: PetscCall(MatCreateSubMatrix(Cmat, new_eq_indices, new_eq_indices, MAT_INITIAL_MATRIX, &mat));
402: PetscCall(MatIsSPDKnown(Cmat, &isset, &isspd)); // like MatPropagateSymmetryOptions, but should set MAT_STRUCTURALLY_SYMMETRIC ?
403: if (isset) PetscCall(MatSetOption(mat, MAT_SPD, isspd));
404: else {
405: PetscCall(MatIsHermitianKnown(Cmat, &isset, &isher));
406: if (isset) PetscCall(MatSetOption(mat, MAT_HERMITIAN, isher));
407: else {
408: #if !defined(PETSC_USE_COMPLEX)
409: PetscCall(MatIsSymmetricKnown(Cmat, &isset, &issym));
410: if (isset) PetscCall(MatSetOption(mat, MAT_SYMMETRIC, issym));
411: #endif
412: }
413: }
414: *a_Amat_crs = mat;
415: }
416: PetscCall(MatDestroy(&Cmat));
418: /* prolongator */
419: {
420: IS findices;
421: PetscInt Istart, Iend;
422: Mat Pnew;
424: PetscCall(MatGetOwnershipRange(Pold, &Istart, &Iend));
425: PetscCall(ISCreateStride(comm, Iend - Istart, Istart, 1, &findices));
426: PetscCall(ISSetBlockSize(findices, f_bs));
427: PetscCall(MatCreateSubMatrix(Pold, findices, new_eq_indices, MAT_INITIAL_MATRIX, &Pnew));
428: PetscCall(ISDestroy(&findices));
429: PetscCall(MatSetOption(Pnew, MAT_FORM_EXPLICIT_TRANSPOSE, PETSC_TRUE));
431: PetscCall(MatDestroy(a_P_inout));
433: /* output - repartitioned */
434: *a_P_inout = Pnew;
435: }
436: PetscCall(ISDestroy(&new_eq_indices));
438: *a_nactive_proc = new_size; /* output */
440: /* pinning on reduced grids, not a bad heuristic and optimization gets folded into process reduction optimization */
441: if (pc_gamg->cpu_pin_coarse_grids) {
442: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
443: static PetscInt llev = 2;
444: PetscCall(PetscInfo(pc, "%s: Pinning level %" PetscInt_FMT " to the CPU\n", ((PetscObject)pc)->prefix, llev++));
445: #endif
446: PetscCall(MatBindToCPU(*a_Amat_crs, PETSC_TRUE));
447: PetscCall(MatBindToCPU(*a_P_inout, PETSC_TRUE));
448: if (1) { /* HACK: move this to MatBindCPU_MPIAIJXXX; lvec is created, need to pin it, this is done in MatSetUpMultiply_MPIAIJ. Hack */
449: Mat A = *a_Amat_crs, P = *a_P_inout;
450: PetscMPIInt size;
451: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
452: if (size > 1) {
453: Mat_MPIAIJ *a = (Mat_MPIAIJ *)A->data, *p = (Mat_MPIAIJ *)P->data;
454: PetscCall(VecBindToCPU(a->lvec, PETSC_TRUE));
455: PetscCall(VecBindToCPU(p->lvec, PETSC_TRUE));
456: }
457: }
458: }
459: PetscCall(PetscLogEventEnd(petsc_gamg_setup_events[GAMG_REDUCE], 0, 0, 0, 0));
460: } // processor reduce
461: PetscFunctionReturn(PETSC_SUCCESS);
462: }
464: // used in GEO
465: PetscErrorCode PCGAMGSquareGraph_GAMG(PC a_pc, Mat Gmat1, Mat *Gmat2)
466: {
467: const char *prefix;
468: char addp[32];
469: PC_MG *mg = (PC_MG *)a_pc->data;
470: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
472: PetscFunctionBegin;
473: PetscCall(PCGetOptionsPrefix(a_pc, &prefix));
474: PetscCall(PetscInfo(a_pc, "%s: Square Graph on level %" PetscInt_FMT "\n", ((PetscObject)a_pc)->prefix, pc_gamg->current_level + 1));
475: PetscCall(MatProductCreate(Gmat1, Gmat1, NULL, Gmat2));
476: PetscCall(MatSetOptionsPrefix(*Gmat2, prefix));
477: PetscCall(PetscSNPrintf(addp, sizeof(addp), "pc_gamg_square_%" PetscInt_FMT "_", pc_gamg->current_level));
478: PetscCall(MatAppendOptionsPrefix(*Gmat2, addp));
479: if ((*Gmat2)->structurally_symmetric == PETSC_BOOL3_TRUE) {
480: PetscCall(MatProductSetType(*Gmat2, MATPRODUCT_AB));
481: } else {
482: PetscCall(MatSetOption(Gmat1, MAT_FORM_EXPLICIT_TRANSPOSE, PETSC_TRUE));
483: PetscCall(MatProductSetType(*Gmat2, MATPRODUCT_AtB));
484: }
485: PetscCall(MatProductSetFromOptions(*Gmat2));
486: PetscCall(PetscLogEventBegin(petsc_gamg_setup_matmat_events[pc_gamg->current_level][0], 0, 0, 0, 0));
487: PetscCall(MatProductSymbolic(*Gmat2));
488: PetscCall(PetscLogEventEnd(petsc_gamg_setup_matmat_events[pc_gamg->current_level][0], 0, 0, 0, 0));
489: PetscCall(MatProductClear(*Gmat2));
490: /* we only need the sparsity, cheat and tell PETSc the matrix has been assembled */
491: (*Gmat2)->assembled = PETSC_TRUE;
492: PetscFunctionReturn(PETSC_SUCCESS);
493: }
495: /*
496: PCSetUp_GAMG - Prepares for the use of the GAMG preconditioner
497: by setting data structures and options.
499: Input Parameter:
500: . pc - the preconditioner context
502: */
503: static PetscErrorCode PCSetUp_GAMG(PC pc)
504: {
505: PC_MG *mg = (PC_MG *)pc->data;
506: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
507: Mat Pmat = pc->pmat;
508: PetscInt fine_level, level, level1, bs, M, N, qq, lidx, nASMBlocksArr[PETSC_MG_MAXLEVELS], cr_bs;
509: MPI_Comm comm;
510: PetscMPIInt rank, size, nactivepe;
511: Mat Aarr[PETSC_MG_MAXLEVELS], Parr[PETSC_MG_MAXLEVELS];
512: IS *ASMLocalIDsArr[PETSC_MG_MAXLEVELS];
513: PetscBool is_last = PETSC_FALSE;
514: #if defined(PETSC_USE_INFO)
515: PetscLogDouble nnz0 = 0., nnztot = 0.;
516: MatInfo info;
517: #endif
519: PetscFunctionBegin;
520: PetscCall(PetscObjectGetComm((PetscObject)pc, &comm));
521: PetscCallMPI(MPI_Comm_rank(comm, &rank));
522: PetscCallMPI(MPI_Comm_size(comm, &size));
523: PetscCall(PetscLogEventBegin(petsc_gamg_setup_events[GAMG_SETUP], 0, 0, 0, 0));
524: if (pc->setupcalled) {
525: if (!pc_gamg->reuse_prol || pc->flag == DIFFERENT_NONZERO_PATTERN) {
526: /* reset everything */
527: PetscCall(PCReset_MG(pc));
528: pc->setupcalled = 0;
529: } else {
530: PC_MG_Levels **mglevels = mg->levels;
531: /* just do Galerkin grids */
532: Mat B, dA, dB;
533: if (pc_gamg->Nlevels > 1) {
534: PetscInt gl;
535: /* currently only handle case where mat and pmat are the same on coarser levels */
536: PetscCall(KSPGetOperators(mglevels[pc_gamg->Nlevels - 1]->smoothd, &dA, &dB));
537: /* (re)set to get dirty flag */
538: PetscCall(KSPSetOperators(mglevels[pc_gamg->Nlevels - 1]->smoothd, dA, dB));
540: for (level = pc_gamg->Nlevels - 2, gl = 0; level >= 0; level--, gl++) {
541: MatReuse reuse = MAT_INITIAL_MATRIX;
542: #if defined(GAMG_STAGES)
543: PetscCall(PetscLogStagePush(gamg_stages[gl]));
544: #endif
545: /* matrix structure can change from repartitioning or process reduction but don't know if we have process reduction here. Should fix */
546: PetscCall(KSPGetOperators(mglevels[level]->smoothd, NULL, &B));
547: if (B->product) {
548: if (B->product->A == dB && B->product->B == mglevels[level + 1]->interpolate) reuse = MAT_REUSE_MATRIX;
549: }
550: if (reuse == MAT_INITIAL_MATRIX) PetscCall(MatDestroy(&mglevels[level]->A));
551: if (reuse == MAT_REUSE_MATRIX) {
552: PetscCall(PetscInfo(pc, "%s: RAP after initial setup, reuse matrix level %" PetscInt_FMT "\n", ((PetscObject)pc)->prefix, level));
553: } else {
554: PetscCall(PetscInfo(pc, "%s: RAP after initial setup, with repartitioning (new matrix) level %" PetscInt_FMT "\n", ((PetscObject)pc)->prefix, level));
555: }
556: PetscCall(PetscLogEventBegin(petsc_gamg_setup_matmat_events[gl][1], 0, 0, 0, 0));
557: PetscCall(MatPtAP(dB, mglevels[level + 1]->interpolate, reuse, PETSC_DEFAULT, &B));
558: PetscCall(PetscLogEventEnd(petsc_gamg_setup_matmat_events[gl][1], 0, 0, 0, 0));
559: if (reuse == MAT_INITIAL_MATRIX) mglevels[level]->A = B;
560: PetscCall(KSPSetOperators(mglevels[level]->smoothd, B, B));
561: // check for redoing eigen estimates
562: if (pc_gamg->recompute_esteig) {
563: PetscBool ischeb;
564: KSP smoother;
565: PetscCall(PCMGGetSmoother(pc, level + 1, &smoother));
566: PetscCall(PetscObjectTypeCompare((PetscObject)smoother, KSPCHEBYSHEV, &ischeb));
567: if (ischeb) {
568: KSP_Chebyshev *cheb = (KSP_Chebyshev *)smoother->data;
569: cheb->emin_provided = 0;
570: cheb->emax_provided = 0;
571: }
572: /* we could call PetscCall(KSPChebyshevSetEigenvalues(smoother, 0, 0)); but the logic does not work properly */
573: }
574: // inc
575: dB = B;
576: #if defined(GAMG_STAGES)
577: PetscCall(PetscLogStagePop());
578: #endif
579: }
580: }
582: PetscCall(PCSetUp_MG(pc));
583: PetscCall(PetscLogEventEnd(petsc_gamg_setup_events[GAMG_SETUP], 0, 0, 0, 0));
584: PetscFunctionReturn(PETSC_SUCCESS);
585: }
586: }
588: if (!pc_gamg->data) {
589: if (pc_gamg->orig_data) {
590: PetscCall(MatGetBlockSize(Pmat, &bs));
591: PetscCall(MatGetLocalSize(Pmat, &qq, NULL));
593: pc_gamg->data_sz = (qq / bs) * pc_gamg->orig_data_cell_rows * pc_gamg->orig_data_cell_cols;
594: pc_gamg->data_cell_rows = pc_gamg->orig_data_cell_rows;
595: pc_gamg->data_cell_cols = pc_gamg->orig_data_cell_cols;
597: PetscCall(PetscMalloc1(pc_gamg->data_sz, &pc_gamg->data));
598: for (qq = 0; qq < pc_gamg->data_sz; qq++) pc_gamg->data[qq] = pc_gamg->orig_data[qq];
599: } else {
600: PetscCheck(pc_gamg->ops->createdefaultdata, comm, PETSC_ERR_PLIB, "'createdefaultdata' not set(?) need to support NULL data");
601: PetscCall(pc_gamg->ops->createdefaultdata(pc, Pmat));
602: }
603: }
605: /* cache original data for reuse */
606: if (!pc_gamg->orig_data && (PetscBool)(!pc_gamg->reuse_prol)) {
607: PetscCall(PetscMalloc1(pc_gamg->data_sz, &pc_gamg->orig_data));
608: for (qq = 0; qq < pc_gamg->data_sz; qq++) pc_gamg->orig_data[qq] = pc_gamg->data[qq];
609: pc_gamg->orig_data_cell_rows = pc_gamg->data_cell_rows;
610: pc_gamg->orig_data_cell_cols = pc_gamg->data_cell_cols;
611: }
613: /* get basic dims */
614: PetscCall(MatGetBlockSize(Pmat, &bs));
615: PetscCall(MatGetSize(Pmat, &M, NULL));
617: #if defined(PETSC_USE_INFO)
618: PetscCall(MatGetInfo(Pmat, MAT_GLOBAL_SUM, &info)); /* global reduction */
619: nnz0 = info.nz_used;
620: nnztot = info.nz_used;
621: #endif
622: PetscCall(PetscInfo(pc, "%s: level %d) N=%" PetscInt_FMT ", n data rows=%" PetscInt_FMT ", n data cols=%" PetscInt_FMT ", nnz/row (ave)=%" PetscInt_FMT ", block size %d, np=%d\n", ((PetscObject)pc)->prefix, 0, M, pc_gamg->data_cell_rows, pc_gamg->data_cell_cols, (PetscInt)(nnz0 / (PetscReal)M + 0.5), (int)bs, size));
624: /* Get A_i and R_i */
625: for (level = 0, Aarr[0] = Pmat, nactivepe = size; level < (pc_gamg->Nlevels - 1) && (level == 0 || M > pc_gamg->coarse_eq_limit); level++) {
626: pc_gamg->current_level = level;
627: PetscCheck(level < PETSC_MG_MAXLEVELS - 2, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Too many levels %" PetscInt_FMT, level + 1);
628: level1 = level + 1;
629: #if defined(GAMG_STAGES)
630: if (!gamg_stages[level]) {
631: char str[32];
632: PetscCall(PetscSNPrintf(str, PETSC_STATIC_ARRAY_LENGTH(str), "GAMG Level %d", (int)level));
633: PetscCall(PetscLogStageRegister(str, &gamg_stages[level]));
634: }
635: PetscCall(PetscLogStagePush(gamg_stages[level]));
636: #endif
637: /* construct prolongator - Parr[level1] */
638: if (level == 0 && pc_gamg->injection_index_size > 0) {
639: Mat Prol;
640: MatType mtype;
641: PetscInt prol_m, prol_n, Prol_N = (M / bs) * pc_gamg->injection_index_size, Istart, Iend, nn, row;
642: PetscCall(PetscInfo(pc, "Create fine grid injection space prolongation %" PetscInt_FMT " x %" PetscInt_FMT ". %s\n", M, Prol_N, pc_gamg->data ? "delete null space data" : ""));
643: PetscCall(MatGetOwnershipRange(Pmat, &Istart, &Iend));
644: PetscCall(MatGetLocalSize(Pmat, &prol_m, NULL)); // rows m x n
645: prol_n = (prol_m / bs) * pc_gamg->injection_index_size;
646: PetscCheck(pc_gamg->injection_index_size < bs, PetscObjectComm((PetscObject)pc), PETSC_ERR_ARG_INCOMP, "Injection size %d must be less that block size %d", (int)pc_gamg->injection_index_size, (int)bs);
647: PetscCall(MatGetType(Pmat, &mtype));
648: PetscCall(MatCreate(PetscObjectComm((PetscObject)pc), &Prol));
649: PetscCall(MatSetBlockSizes(Prol, bs, pc_gamg->injection_index_size));
650: PetscCall(MatSetSizes(Prol, prol_m, prol_n, M, Prol_N));
651: PetscCall(MatSetType(Prol, mtype));
652: #if PetscDefined(HAVE_DEVICE)
653: PetscBool flg;
654: PetscCall(MatBoundToCPU(Pmat, &flg));
655: PetscCall(MatBindToCPU(Prol, flg));
656: if (flg) PetscCall(MatSetBindingPropagates(Prol, PETSC_TRUE));
657: #endif
658: PetscCall(MatSeqAIJSetPreallocation(Prol, 1, NULL));
659: PetscCall(MatMPIAIJSetPreallocation(Prol, 1, NULL, 0, NULL));
660: // set I \kron [1, 1, ... ]^T
661: for (PetscInt ii = Istart, col = (Istart / bs) * pc_gamg->injection_index_size; ii < Iend; ii += bs) {
662: const PetscScalar one = 1;
663: for (PetscInt jj = 0; jj < pc_gamg->injection_index_size; jj++, col++) {
664: PetscInt row = ii + pc_gamg->injection_index[jj];
665: PetscCall(MatSetValues(Prol, 1, &row, 1, &col, &one, INSERT_VALUES));
666: }
667: }
668: PetscCall(MatAssemblyBegin(Prol, MAT_FINAL_ASSEMBLY));
669: PetscCall(MatAssemblyEnd(Prol, MAT_FINAL_ASSEMBLY));
670: PetscCall(MatViewFromOptions(Prol, NULL, "-mat_view_injection"));
671: PetscCall(MatGetBlockSizes(Prol, NULL, &cr_bs)); // column size
672: Parr[level1] = Prol;
673: // can not deal with null space -- with array of 'injection cols' we could take 'injection rows and 'injection cols' to 'data'
674: if (pc_gamg->data) {
675: pc_gamg->data_cell_cols = pc_gamg->injection_index_size;
676: pc_gamg->data_cell_rows = pc_gamg->injection_index_size;
677: pc_gamg->orig_data_cell_cols = 0;
678: pc_gamg->orig_data_cell_rows = 0;
679: PetscCall(PetscFree(pc_gamg->data));
680: pc_gamg->data_sz = pc_gamg->injection_index_size * prol_n;
681: PetscCall(PetscMalloc1(pc_gamg->data_sz, &pc_gamg->data));
682: for (row = nn = 0; row < prol_n; row += pc_gamg->injection_index_size) {
683: for (int jj = 0; jj < pc_gamg->injection_index_size; jj++) {
684: int idx = row * pc_gamg->injection_index_size + jj * pc_gamg->injection_index_size;
685: for (int kk = 0; kk < pc_gamg->injection_index_size; kk++, nn++) { pc_gamg->data[idx + kk] = (jj == kk) ? 1 : 0; }
686: }
687: }
688: PetscCheck(nn == pc_gamg->data_sz, PETSC_COMM_SELF, PETSC_ERR_PLIB, "nn != pc_gamg->data_sz %" PetscInt_FMT " %" PetscInt_FMT, pc_gamg->data_sz, nn);
689: }
690: } else {
691: Mat Gmat, mat;
692: PetscCoarsenData *agg_lists;
693: Mat Prol11;
695: PetscCall(PCGAMGCreateGraph(pc, Aarr[level], &Gmat));
696: PetscCall(pc_gamg->ops->coarsen(pc, &Gmat, &agg_lists)); // Gmat may have ghosts for QR aggregates not in matrix
697: PetscCall(PetscCDGetMat(agg_lists, &mat));
698: if (!mat) PetscCall(PetscCDSetMat(agg_lists, Gmat));
699: PetscCall(pc_gamg->ops->prolongator(pc, Aarr[level], agg_lists, &Prol11));
700: /* could have failed to create new level */
701: if (Prol11) {
702: const char *prefix;
703: char addp[32];
705: /* get new block size of coarse matrices */
706: PetscCall(MatGetBlockSizes(Prol11, NULL, &cr_bs)); // column size
708: if (pc_gamg->ops->optprolongator) {
709: /* smooth */
710: PetscCall(pc_gamg->ops->optprolongator(pc, Aarr[level], &Prol11));
711: }
713: if (pc_gamg->use_aggs_in_asm) {
714: PetscInt bs;
715: PetscCall(MatGetBlockSizes(Prol11, &bs, NULL)); // row block size
716: PetscCall(PetscCDGetASMBlocks(agg_lists, bs, &nASMBlocksArr[level], &ASMLocalIDsArr[level]));
717: PetscCall(PetscInfo(pc, "%d: %" PetscInt_FMT " ASM local domains, bs = %d\n", (int)level, nASMBlocksArr[level], (int)bs));
718: } else if (pc_gamg->asm_hem_aggs) {
719: MatCoarsen crs;
720: const char *prefix;
721: PetscInt bs;
722: PetscCall(PetscCDGetMat(agg_lists, &mat));
723: if (mat == Gmat) PetscCall(PetscCDClearMat(agg_lists)); // take the Mat away from the list (yuck)
724: PetscCall(PetscCDDestroy(agg_lists));
725: PetscCall(PetscInfo(pc, "HEM ASM passes = %d\n", (int)pc_gamg->asm_hem_aggs));
726: PetscCall(MatCoarsenCreate(PetscObjectComm((PetscObject)pc), &crs));
727: PetscCall(PetscObjectGetOptionsPrefix((PetscObject)pc, &prefix));
728: PetscCall(PetscObjectSetOptionsPrefix((PetscObject)crs, prefix));
729: PetscCall(MatCoarsenSetFromOptions(crs)); // get strength args
730: PetscCall(MatCoarsenSetType(crs, MATCOARSENHEM));
731: PetscCall(MatCoarsenSetMaximumIterations(crs, pc_gamg->asm_hem_aggs));
732: PetscCall(MatCoarsenSetAdjacency(crs, Gmat));
733: PetscCall(MatCoarsenSetStrictAggs(crs, PETSC_TRUE));
734: PetscCall(MatCoarsenApply(crs));
735: PetscCall(MatCoarsenViewFromOptions(crs, NULL, "-agg_hem_mat_coarsen_view"));
736: PetscCall(MatCoarsenGetData(crs, &agg_lists)); /* output */
737: PetscCall(MatCoarsenDestroy(&crs));
738: // create aggregates
739: PetscCall(MatGetBlockSizes(Aarr[level], &bs, NULL)); // row block size
740: PetscCall(PetscCDGetASMBlocks(agg_lists, bs, &nASMBlocksArr[level], &ASMLocalIDsArr[level]));
741: }
742: PetscCall(PCGetOptionsPrefix(pc, &prefix));
743: PetscCall(MatSetOptionsPrefix(Prol11, prefix));
744: PetscCall(PetscSNPrintf(addp, sizeof(addp), "pc_gamg_prolongator_%d_", (int)level));
745: PetscCall(MatAppendOptionsPrefix(Prol11, addp));
746: /* Always generate the transpose with CUDA
747: Such behaviour can be adapted with -pc_gamg_prolongator_ prefixed options */
748: PetscCall(MatSetOption(Prol11, MAT_FORM_EXPLICIT_TRANSPOSE, PETSC_TRUE));
749: PetscCall(MatSetFromOptions(Prol11));
750: Parr[level1] = Prol11;
751: } else Parr[level1] = NULL; /* failed to coarsen */
752: PetscCall(PetscCDGetMat(agg_lists, &mat));
753: if (mat == Gmat) PetscCall(PetscCDClearMat(agg_lists)); // take the Mat away from the list (yuck)
754: PetscCall(MatDestroy(&Gmat));
755: PetscCall(PetscCDDestroy(agg_lists));
756: } /* construct prolongator scope */
757: if (level == 0) Aarr[0] = Pmat; /* use Pmat for finest level setup */
758: if (!Parr[level1]) { /* failed to coarsen */
759: PetscCall(PetscInfo(pc, "%s: Stop gridding, level %" PetscInt_FMT "\n", ((PetscObject)pc)->prefix, level));
760: #if defined(GAMG_STAGES)
761: PetscCall(PetscLogStagePop());
762: #endif
763: break;
764: }
765: PetscCall(MatGetSize(Parr[level1], &M, &N)); /* N is next M, a loop test variables */
766: PetscCheck(!is_last, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Is last ?");
767: if (N <= pc_gamg->coarse_eq_limit) is_last = PETSC_TRUE;
768: if (level1 == pc_gamg->Nlevels - 1) is_last = PETSC_TRUE;
769: PetscCall(PetscLogEventBegin(petsc_gamg_setup_events[GAMG_LEVEL], 0, 0, 0, 0));
770: PetscCall(pc_gamg->ops->createlevel(pc, Aarr[level], cr_bs, &Parr[level1], &Aarr[level1], &nactivepe, NULL, is_last));
771: PetscCall(PetscLogEventEnd(petsc_gamg_setup_events[GAMG_LEVEL], 0, 0, 0, 0));
773: PetscCall(MatGetSize(Aarr[level1], &M, &N)); /* M is loop test variables */
774: #if defined(PETSC_USE_INFO)
775: PetscCall(MatGetInfo(Aarr[level1], MAT_GLOBAL_SUM, &info));
776: nnztot += info.nz_used;
777: #endif
778: PetscCall(PetscInfo(pc, "%s: %d) N=%" PetscInt_FMT ", n data cols=%" PetscInt_FMT ", nnz/row (ave)=%" PetscInt_FMT ", %d active pes\n", ((PetscObject)pc)->prefix, (int)level1, M, pc_gamg->data_cell_cols, (PetscInt)(info.nz_used / (PetscReal)M), nactivepe));
780: #if defined(GAMG_STAGES)
781: PetscCall(PetscLogStagePop());
782: #endif
783: /* stop if one node or one proc -- could pull back for singular problems */
784: if ((pc_gamg->data_cell_cols && M / pc_gamg->data_cell_cols < 2) || (!pc_gamg->data_cell_cols && M / bs < 2)) {
785: PetscCall(PetscInfo(pc, "%s: HARD stop of coarsening on level %" PetscInt_FMT ". Grid too small: %" PetscInt_FMT " block nodes\n", ((PetscObject)pc)->prefix, level, M / bs));
786: level++;
787: break;
788: }
789: } /* levels */
790: PetscCall(PetscFree(pc_gamg->data));
792: PetscCall(PetscInfo(pc, "%s: %" PetscInt_FMT " levels, operator complexity = %g\n", ((PetscObject)pc)->prefix, level + 1, nnztot / nnz0));
793: pc_gamg->Nlevels = level + 1;
794: fine_level = level;
795: PetscCall(PCMGSetLevels(pc, pc_gamg->Nlevels, NULL));
797: if (pc_gamg->Nlevels > 1) { /* don't setup MG if one level */
799: /* set default smoothers & set operators */
800: for (lidx = 1, level = pc_gamg->Nlevels - 2; lidx <= fine_level; lidx++, level--) {
801: KSP smoother;
802: PC subpc;
804: PetscCall(PCMGGetSmoother(pc, lidx, &smoother));
805: PetscCall(KSPGetPC(smoother, &subpc));
807: PetscCall(KSPSetNormType(smoother, KSP_NORM_NONE));
808: /* set ops */
809: PetscCall(KSPSetOperators(smoother, Aarr[level], Aarr[level]));
810: PetscCall(PCMGSetInterpolation(pc, lidx, Parr[level + 1]));
812: /* set defaults */
813: PetscCall(KSPSetType(smoother, KSPCHEBYSHEV));
815: /* set blocks for ASM smoother that uses the 'aggregates' */
816: if (pc_gamg->use_aggs_in_asm || pc_gamg->asm_hem_aggs) {
817: PetscInt sz;
818: IS *iss;
820: sz = nASMBlocksArr[level];
821: iss = ASMLocalIDsArr[level];
822: PetscCall(PCSetType(subpc, PCASM));
823: PetscCall(PCASMSetOverlap(subpc, 0));
824: PetscCall(PCASMSetType(subpc, PC_ASM_BASIC));
825: if (!sz) {
826: IS is;
827: PetscCall(ISCreateGeneral(PETSC_COMM_SELF, 0, NULL, PETSC_COPY_VALUES, &is));
828: PetscCall(PCASMSetLocalSubdomains(subpc, 1, NULL, &is));
829: PetscCall(ISDestroy(&is));
830: } else {
831: PetscInt kk;
832: PetscCall(PCASMSetLocalSubdomains(subpc, sz, iss, NULL));
833: for (kk = 0; kk < sz; kk++) PetscCall(ISDestroy(&iss[kk]));
834: PetscCall(PetscFree(iss));
835: }
836: ASMLocalIDsArr[level] = NULL;
837: nASMBlocksArr[level] = 0;
838: } else {
839: PetscCall(PCSetType(subpc, PCJACOBI));
840: }
841: }
842: {
843: /* coarse grid */
844: KSP smoother, *k2;
845: PC subpc, pc2;
846: PetscInt ii, first;
847: Mat Lmat = Aarr[(level = pc_gamg->Nlevels - 1)];
848: lidx = 0;
850: PetscCall(PCMGGetSmoother(pc, lidx, &smoother));
851: PetscCall(KSPSetOperators(smoother, Lmat, Lmat));
852: if (!pc_gamg->use_parallel_coarse_grid_solver) {
853: PetscCall(KSPSetNormType(smoother, KSP_NORM_NONE));
854: PetscCall(KSPGetPC(smoother, &subpc));
855: PetscCall(PCSetType(subpc, PCBJACOBI));
856: PetscCall(PCSetUp(subpc));
857: PetscCall(PCBJacobiGetSubKSP(subpc, &ii, &first, &k2));
858: PetscCheck(ii == 1, PETSC_COMM_SELF, PETSC_ERR_PLIB, "ii %" PetscInt_FMT " is not one", ii);
859: PetscCall(KSPGetPC(k2[0], &pc2));
860: PetscCall(PCSetType(pc2, PCLU));
861: PetscCall(PCFactorSetShiftType(pc2, MAT_SHIFT_INBLOCKS));
862: PetscCall(KSPSetTolerances(k2[0], PETSC_DEFAULT, PETSC_DEFAULT, PETSC_DEFAULT, 1));
863: PetscCall(KSPSetType(k2[0], KSPPREONLY));
864: }
865: }
867: /* should be called in PCSetFromOptions_GAMG(), but cannot be called prior to PCMGSetLevels() */
868: PetscObjectOptionsBegin((PetscObject)pc);
869: PetscCall(PCSetFromOptions_MG(pc, PetscOptionsObject));
870: PetscOptionsEnd();
871: PetscCall(PCMGSetGalerkin(pc, PC_MG_GALERKIN_EXTERNAL));
873: /* set cheby eigen estimates from SA to use in the solver */
874: if (pc_gamg->use_sa_esteig) {
875: for (lidx = 1, level = pc_gamg->Nlevels - 2; level >= 0; lidx++, level--) {
876: KSP smoother;
877: PetscBool ischeb;
879: PetscCall(PCMGGetSmoother(pc, lidx, &smoother));
880: PetscCall(PetscObjectTypeCompare((PetscObject)smoother, KSPCHEBYSHEV, &ischeb));
881: if (ischeb) {
882: KSP_Chebyshev *cheb = (KSP_Chebyshev *)smoother->data;
884: // The command line will override these settings because KSPSetFromOptions is called in PCSetUp_MG
885: if (mg->max_eigen_DinvA[level] > 0) {
886: // SA uses Jacobi for P; we use SA estimates if the smoother is also Jacobi or if the user explicitly requested it.
887: // TODO: This should test whether it's the same Jacobi variant (DIAG, ROWSUM, etc.)
888: PetscReal emax, emin;
890: emin = mg->min_eigen_DinvA[level];
891: emax = mg->max_eigen_DinvA[level];
892: PetscCall(PetscInfo(pc, "%s: PCSetUp_GAMG: call KSPChebyshevSetEigenvalues on level %" PetscInt_FMT " (N=%" PetscInt_FMT ") with emax = %g emin = %g\n", ((PetscObject)pc)->prefix, level, Aarr[level]->rmap->N, (double)emax, (double)emin));
893: cheb->emin_provided = emin;
894: cheb->emax_provided = emax;
895: }
896: }
897: }
898: }
900: PetscCall(PCSetUp_MG(pc));
902: /* clean up */
903: for (level = 1; level < pc_gamg->Nlevels; level++) {
904: PetscCall(MatDestroy(&Parr[level]));
905: PetscCall(MatDestroy(&Aarr[level]));
906: }
907: } else {
908: KSP smoother;
910: PetscCall(PetscInfo(pc, "%s: One level solver used (system is seen as DD). Using default solver.\n", ((PetscObject)pc)->prefix));
911: PetscCall(PCMGGetSmoother(pc, 0, &smoother));
912: PetscCall(KSPSetOperators(smoother, Aarr[0], Aarr[0]));
913: PetscCall(KSPSetType(smoother, KSPPREONLY));
914: PetscCall(PCSetUp_MG(pc));
915: }
916: PetscCall(PetscLogEventEnd(petsc_gamg_setup_events[GAMG_SETUP], 0, 0, 0, 0));
917: PetscFunctionReturn(PETSC_SUCCESS);
918: }
920: /*
921: PCDestroy_GAMG - Destroys the private context for the GAMG preconditioner
922: that was created with PCCreate_GAMG().
924: Input Parameter:
925: . pc - the preconditioner context
927: Application Interface Routine: PCDestroy()
928: */
929: PetscErrorCode PCDestroy_GAMG(PC pc)
930: {
931: PC_MG *mg = (PC_MG *)pc->data;
932: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
934: PetscFunctionBegin;
935: PetscCall(PCReset_GAMG(pc));
936: if (pc_gamg->ops->destroy) PetscCall((*pc_gamg->ops->destroy)(pc));
937: PetscCall(PetscFree(pc_gamg->ops));
938: PetscCall(PetscFree(pc_gamg->gamg_type_name));
939: PetscCall(PetscFree(pc_gamg));
940: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetProcEqLim_C", NULL));
941: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetCoarseEqLim_C", NULL));
942: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetRepartition_C", NULL));
943: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetEigenvalues_C", NULL));
944: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetUseSAEstEig_C", NULL));
945: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetRecomputeEstEig_C", NULL));
946: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetReuseInterpolation_C", NULL));
947: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGASMSetUseAggs_C", NULL));
948: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetParallelCoarseGridSolve_C", NULL));
949: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetCpuPinCoarseGrids_C", NULL));
950: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetCoarseGridLayoutType_C", NULL));
951: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetThreshold_C", NULL));
952: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetRankReductionFactors_C", NULL));
953: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetThresholdScale_C", NULL));
954: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetType_C", NULL));
955: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGGetType_C", NULL));
956: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetNlevels_C", NULL));
957: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGASMSetHEM_C", NULL));
958: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetInjectionIndices_C", NULL));
959: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetInjectionIndex_C", NULL));
960: PetscCall(PCDestroy_MG(pc));
961: PetscFunctionReturn(PETSC_SUCCESS);
962: }
964: /*@
965: PCGAMGSetProcEqLim - Set number of equations to aim for per process on the coarse grids via processor reduction in `PCGAMG`
967: Logically Collective
969: Input Parameters:
970: + pc - the preconditioner context
971: - n - the number of equations
973: Options Database Key:
974: . -pc_gamg_process_eq_limit <limit> - set the limit
976: Level: intermediate
978: Note:
979: `PCGAMG` will reduce the number of MPI processes used directly on the coarse grids so that there are around <limit> equations on each process
980: that has degrees of freedom
982: Developer Note:
983: Should be named `PCGAMGSetProcessEquationLimit()`.
985: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetCoarseEqLim()`, `PCGAMGSetRankReductionFactors()`, `PCGAMGSetRepartition()`
986: @*/
987: PetscErrorCode PCGAMGSetProcEqLim(PC pc, PetscInt n)
988: {
989: PetscFunctionBegin;
991: PetscTryMethod(pc, "PCGAMGSetProcEqLim_C", (PC, PetscInt), (pc, n));
992: PetscFunctionReturn(PETSC_SUCCESS);
993: }
995: static PetscErrorCode PCGAMGSetProcEqLim_GAMG(PC pc, PetscInt n)
996: {
997: PC_MG *mg = (PC_MG *)pc->data;
998: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1000: PetscFunctionBegin;
1001: if (n > 0) pc_gamg->min_eq_proc = n;
1002: PetscFunctionReturn(PETSC_SUCCESS);
1003: }
1005: /*@
1006: PCGAMGSetCoarseEqLim - Set maximum number of equations on the coarsest grid of `PCGAMG`
1008: Collective
1010: Input Parameters:
1011: + pc - the preconditioner context
1012: - n - maximum number of equations to aim for
1014: Options Database Key:
1015: . -pc_gamg_coarse_eq_limit <limit> - set the limit
1017: Level: intermediate
1019: Note:
1020: For example -pc_gamg_coarse_eq_limit 1000 will stop coarsening once the coarse grid
1021: has less than 1000 unknowns.
1023: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetProcEqLim()`, `PCGAMGSetRankReductionFactors()`, `PCGAMGSetRepartition()`,
1024: `PCGAMGSetParallelCoarseGridSolve()`
1025: @*/
1026: PetscErrorCode PCGAMGSetCoarseEqLim(PC pc, PetscInt n)
1027: {
1028: PetscFunctionBegin;
1030: PetscTryMethod(pc, "PCGAMGSetCoarseEqLim_C", (PC, PetscInt), (pc, n));
1031: PetscFunctionReturn(PETSC_SUCCESS);
1032: }
1034: static PetscErrorCode PCGAMGSetCoarseEqLim_GAMG(PC pc, PetscInt n)
1035: {
1036: PC_MG *mg = (PC_MG *)pc->data;
1037: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1039: PetscFunctionBegin;
1040: if (n > 0) pc_gamg->coarse_eq_limit = n;
1041: PetscFunctionReturn(PETSC_SUCCESS);
1042: }
1044: /*@
1045: PCGAMGSetRepartition - Repartition the degrees of freedom across the processors on the coarser grids when reducing the number of MPI ranks to use
1047: Collective
1049: Input Parameters:
1050: + pc - the preconditioner context
1051: - n - `PETSC_TRUE` or `PETSC_FALSE`
1053: Options Database Key:
1054: . -pc_gamg_repartition <true,false> - turn on the repartitioning
1056: Level: intermediate
1058: Note:
1059: This will generally improve the loading balancing of the work on each level so the solves will be faster but it adds to the
1060: preconditioner setup time.
1062: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetProcEqLim()`, `PCGAMGSetRankReductionFactors()`
1063: @*/
1064: PetscErrorCode PCGAMGSetRepartition(PC pc, PetscBool n)
1065: {
1066: PetscFunctionBegin;
1068: PetscTryMethod(pc, "PCGAMGSetRepartition_C", (PC, PetscBool), (pc, n));
1069: PetscFunctionReturn(PETSC_SUCCESS);
1070: }
1072: static PetscErrorCode PCGAMGSetRepartition_GAMG(PC pc, PetscBool n)
1073: {
1074: PC_MG *mg = (PC_MG *)pc->data;
1075: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1077: PetscFunctionBegin;
1078: pc_gamg->repart = n;
1079: PetscFunctionReturn(PETSC_SUCCESS);
1080: }
1082: /*@
1083: PCGAMGSetUseSAEstEig - Use the eigen estimate from smoothed aggregation for the Chebyshev smoother during the solution process
1085: Collective
1087: Input Parameters:
1088: + pc - the preconditioner context
1089: - b - flag
1091: Options Database Key:
1092: . -pc_gamg_use_sa_esteig <true,false> - use the eigen estimate
1094: Level: advanced
1096: Notes:
1097: Smoothed aggregation constructs the smoothed prolongator $P = (I - \omega D^{-1} A) T$ where $T$ is the tentative prolongator and $D$ is the diagonal of $A$.
1098: Eigenvalue estimates (based on a few `PCCG` or `PCGMRES` iterations) are computed to choose $\omega$ so that this is a stable smoothing operation.
1099: If `KSPCHEBYSHEV` with `PCJACOBI` (diagonal) preconditioning is used for smoothing on the finest level, then the eigenvalue estimates
1100: can be reused during the solution process.
1101: This option is only used when the smoother uses `PCJACOBI`, and should be turned off when a different `PCJacobiType` is used.
1102: It became default in PETSc 3.17.
1104: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `KSPChebyshevSetEigenvalues()`, `KSPChebyshevEstEigSet()`, `PCGAMGSetRecomputeEstEig()`
1105: @*/
1106: PetscErrorCode PCGAMGSetUseSAEstEig(PC pc, PetscBool b)
1107: {
1108: PetscFunctionBegin;
1110: PetscTryMethod(pc, "PCGAMGSetUseSAEstEig_C", (PC, PetscBool), (pc, b));
1111: PetscFunctionReturn(PETSC_SUCCESS);
1112: }
1114: static PetscErrorCode PCGAMGSetUseSAEstEig_GAMG(PC pc, PetscBool b)
1115: {
1116: PC_MG *mg = (PC_MG *)pc->data;
1117: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1119: PetscFunctionBegin;
1120: pc_gamg->use_sa_esteig = b;
1121: PetscFunctionReturn(PETSC_SUCCESS);
1122: }
1124: /*@
1125: PCGAMGSetRecomputeEstEig - Set flag for Chebyshev smoothers to recompute the eigen estimates when a new matrix is used
1127: Collective
1129: Input Parameters:
1130: + pc - the preconditioner context
1131: - b - flag, default is `PETSC_TRUE`
1133: Options Database Key:
1134: . -pc_gamg_recompute_esteig <true> - use the eigen estimate
1136: Level: advanced
1138: Note:
1139: If the matrix changes only slightly in a new solve using ``PETSC_FALSE`` will save time in the setting up of the preconditioner
1140: and may not affect the solution time much.
1142: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `KSPChebyshevSetEigenvalues()`, `KSPChebyshevEstEigSet()`
1143: @*/
1144: PetscErrorCode PCGAMGSetRecomputeEstEig(PC pc, PetscBool b)
1145: {
1146: PetscFunctionBegin;
1148: PetscTryMethod(pc, "PCGAMGSetRecomputeEstEig_C", (PC, PetscBool), (pc, b));
1149: PetscFunctionReturn(PETSC_SUCCESS);
1150: }
1152: static PetscErrorCode PCGAMGSetRecomputeEstEig_GAMG(PC pc, PetscBool b)
1153: {
1154: PC_MG *mg = (PC_MG *)pc->data;
1155: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1157: PetscFunctionBegin;
1158: pc_gamg->recompute_esteig = b;
1159: PetscFunctionReturn(PETSC_SUCCESS);
1160: }
1162: /*@
1163: PCGAMGSetEigenvalues - Set WHAT eigenvalues WHY?
1165: Collective
1167: Input Parameters:
1168: + pc - the preconditioner context
1169: . emax - max eigenvalue
1170: - emin - min eigenvalue
1172: Options Database Key:
1173: . -pc_gamg_eigenvalues <emin,emax> - estimates of the eigenvalues
1175: Level: intermediate
1177: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetUseSAEstEig()`
1178: @*/
1179: PetscErrorCode PCGAMGSetEigenvalues(PC pc, PetscReal emax, PetscReal emin)
1180: {
1181: PetscFunctionBegin;
1183: PetscTryMethod(pc, "PCGAMGSetEigenvalues_C", (PC, PetscReal, PetscReal), (pc, emax, emin));
1184: PetscFunctionReturn(PETSC_SUCCESS);
1185: }
1187: static PetscErrorCode PCGAMGSetEigenvalues_GAMG(PC pc, PetscReal emax, PetscReal emin)
1188: {
1189: PC_MG *mg = (PC_MG *)pc->data;
1190: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1192: PetscFunctionBegin;
1193: PetscCheck(emax > emin, PetscObjectComm((PetscObject)pc), PETSC_ERR_ARG_INCOMP, "Maximum eigenvalue must be larger than minimum: max %g min %g", (double)emax, (double)emin);
1194: PetscCheck(emax * emin > 0.0, PetscObjectComm((PetscObject)pc), PETSC_ERR_ARG_INCOMP, "Both eigenvalues must be of the same sign: max %g min %g", (double)emax, (double)emin);
1195: pc_gamg->emax = emax;
1196: pc_gamg->emin = emin;
1197: PetscFunctionReturn(PETSC_SUCCESS);
1198: }
1200: /*@
1201: PCGAMGSetReuseInterpolation - Reuse prolongation when rebuilding a `PCGAMG` algebraic multigrid preconditioner
1203: Collective
1205: Input Parameters:
1206: + pc - the preconditioner context
1207: - n - `PETSC_TRUE` or `PETSC_FALSE`
1209: Options Database Key:
1210: . -pc_gamg_reuse_interpolation <true,false> - reuse the previous interpolation
1212: Level: intermediate
1214: Note:
1215: May negatively affect the convergence rate of the method on new matrices if the matrix entries change a great deal, but allows
1216: rebuilding the preconditioner quicker.
1218: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`
1219: @*/
1220: PetscErrorCode PCGAMGSetReuseInterpolation(PC pc, PetscBool n)
1221: {
1222: PetscFunctionBegin;
1224: PetscTryMethod(pc, "PCGAMGSetReuseInterpolation_C", (PC, PetscBool), (pc, n));
1225: PetscFunctionReturn(PETSC_SUCCESS);
1226: }
1228: static PetscErrorCode PCGAMGSetReuseInterpolation_GAMG(PC pc, PetscBool n)
1229: {
1230: PC_MG *mg = (PC_MG *)pc->data;
1231: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1233: PetscFunctionBegin;
1234: pc_gamg->reuse_prol = n;
1235: PetscFunctionReturn(PETSC_SUCCESS);
1236: }
1238: /*@
1239: PCGAMGASMSetUseAggs - Have the `PCGAMG` smoother on each level use `PCASM` where the aggregates defined by the coarsening process are
1240: the subdomains for the additive Schwarz preconditioner used as the smoother
1242: Collective
1244: Input Parameters:
1245: + pc - the preconditioner context
1246: - flg - `PETSC_TRUE` to use aggregates, `PETSC_FALSE` to not
1248: Options Database Key:
1249: . -pc_gamg_asm_use_agg <true,false> - use aggregates to define the additive Schwarz subdomains
1251: Level: intermediate
1253: Note:
1254: This option automatically sets the preconditioner on the levels to be `PCASM`.
1256: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCASM`, `PCSetType`
1257: @*/
1258: PetscErrorCode PCGAMGASMSetUseAggs(PC pc, PetscBool flg)
1259: {
1260: PetscFunctionBegin;
1262: PetscTryMethod(pc, "PCGAMGASMSetUseAggs_C", (PC, PetscBool), (pc, flg));
1263: PetscFunctionReturn(PETSC_SUCCESS);
1264: }
1266: static PetscErrorCode PCGAMGASMSetUseAggs_GAMG(PC pc, PetscBool flg)
1267: {
1268: PC_MG *mg = (PC_MG *)pc->data;
1269: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1271: PetscFunctionBegin;
1272: pc_gamg->use_aggs_in_asm = flg;
1273: PetscFunctionReturn(PETSC_SUCCESS);
1274: }
1276: /*@
1277: PCGAMGSetParallelCoarseGridSolve - allow a parallel coarse grid solver
1279: Collective
1281: Input Parameters:
1282: + pc - the preconditioner context
1283: - flg - `PETSC_TRUE` to not force coarse grid onto one processor
1285: Options Database Key:
1286: . -pc_gamg_parallel_coarse_grid_solver - use a parallel coarse grid solver
1288: Level: intermediate
1290: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetCoarseGridLayoutType()`, `PCGAMGSetCpuPinCoarseGrids()`, `PCGAMGSetRankReductionFactors()`
1291: @*/
1292: PetscErrorCode PCGAMGSetParallelCoarseGridSolve(PC pc, PetscBool flg)
1293: {
1294: PetscFunctionBegin;
1296: PetscTryMethod(pc, "PCGAMGSetParallelCoarseGridSolve_C", (PC, PetscBool), (pc, flg));
1297: PetscFunctionReturn(PETSC_SUCCESS);
1298: }
1300: static PetscErrorCode PCGAMGSetParallelCoarseGridSolve_GAMG(PC pc, PetscBool flg)
1301: {
1302: PC_MG *mg = (PC_MG *)pc->data;
1303: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1305: PetscFunctionBegin;
1306: pc_gamg->use_parallel_coarse_grid_solver = flg;
1307: PetscFunctionReturn(PETSC_SUCCESS);
1308: }
1310: /*@
1311: PCGAMGSetCpuPinCoarseGrids - pin the coarse grids created in `PCGAMG` to run only on the CPU since the problems may be too small to run efficiently on the GPUs
1313: Collective
1315: Input Parameters:
1316: + pc - the preconditioner context
1317: - flg - `PETSC_TRUE` to pin coarse grids to the CPU
1319: Options Database Key:
1320: . -pc_gamg_cpu_pin_coarse_grids - pin the coarse grids to the CPU
1322: Level: intermediate
1324: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetCoarseGridLayoutType()`, `PCGAMGSetParallelCoarseGridSolve()`
1325: @*/
1326: PetscErrorCode PCGAMGSetCpuPinCoarseGrids(PC pc, PetscBool flg)
1327: {
1328: PetscFunctionBegin;
1330: PetscTryMethod(pc, "PCGAMGSetCpuPinCoarseGrids_C", (PC, PetscBool), (pc, flg));
1331: PetscFunctionReturn(PETSC_SUCCESS);
1332: }
1334: static PetscErrorCode PCGAMGSetCpuPinCoarseGrids_GAMG(PC pc, PetscBool flg)
1335: {
1336: PC_MG *mg = (PC_MG *)pc->data;
1337: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1339: PetscFunctionBegin;
1340: pc_gamg->cpu_pin_coarse_grids = flg;
1341: PetscFunctionReturn(PETSC_SUCCESS);
1342: }
1344: /*@
1345: PCGAMGSetCoarseGridLayoutType - place coarse grids on processors with natural order (compact type)
1347: Collective
1349: Input Parameters:
1350: + pc - the preconditioner context
1351: - flg - `PCGAMGLayoutType` type, either `PCGAMG_LAYOUT_COMPACT` or `PCGAMG_LAYOUT_SPREAD`
1353: Options Database Key:
1354: . -pc_gamg_coarse_grid_layout_type - place the coarse grids with natural ordering
1356: Level: intermediate
1358: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetParallelCoarseGridSolve()`, `PCGAMGSetCpuPinCoarseGrids()`, `PCGAMGLayoutType`, `PCGAMG_LAYOUT_COMPACT`, `PCGAMG_LAYOUT_SPREAD`
1359: @*/
1360: PetscErrorCode PCGAMGSetCoarseGridLayoutType(PC pc, PCGAMGLayoutType flg)
1361: {
1362: PetscFunctionBegin;
1364: PetscTryMethod(pc, "PCGAMGSetCoarseGridLayoutType_C", (PC, PCGAMGLayoutType), (pc, flg));
1365: PetscFunctionReturn(PETSC_SUCCESS);
1366: }
1368: static PetscErrorCode PCGAMGSetCoarseGridLayoutType_GAMG(PC pc, PCGAMGLayoutType flg)
1369: {
1370: PC_MG *mg = (PC_MG *)pc->data;
1371: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1373: PetscFunctionBegin;
1374: pc_gamg->layout_type = flg;
1375: PetscFunctionReturn(PETSC_SUCCESS);
1376: }
1378: /*@
1379: PCGAMGSetNlevels - Sets the maximum number of levels `PCGAMG` will use
1381: Collective
1383: Input Parameters:
1384: + pc - the preconditioner
1385: - n - the maximum number of levels to use
1387: Options Database Key:
1388: . -pc_mg_levels <n> - set the maximum number of levels to allow
1390: Level: intermediate
1392: Developer Notes:
1393: Should be called `PCGAMGSetMaximumNumberlevels()` and possible be shared with `PCMG`
1395: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`
1396: @*/
1397: PetscErrorCode PCGAMGSetNlevels(PC pc, PetscInt n)
1398: {
1399: PetscFunctionBegin;
1401: PetscTryMethod(pc, "PCGAMGSetNlevels_C", (PC, PetscInt), (pc, n));
1402: PetscFunctionReturn(PETSC_SUCCESS);
1403: }
1405: static PetscErrorCode PCGAMGSetNlevels_GAMG(PC pc, PetscInt n)
1406: {
1407: PC_MG *mg = (PC_MG *)pc->data;
1408: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1410: PetscFunctionBegin;
1411: pc_gamg->Nlevels = n;
1412: PetscFunctionReturn(PETSC_SUCCESS);
1413: }
1415: /*@
1416: PCGAMGASMSetHEM - Sets the number of HEM matching passed
1418: Collective
1420: Input Parameters:
1421: + pc - the preconditioner
1422: - n - number of HEM matching passed to construct ASM subdomains
1424: Options Database Key:
1425: . -pc_gamg_asm_hem <n> - set the number of HEM matching passed
1427: Level: intermediate
1429: Developer Notes:
1430: Should be called `PCGAMGSetMaximumNumberlevels()` and possible be shared with `PCMG`
1432: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`
1433: @*/
1434: PetscErrorCode PCGAMGASMSetHEM(PC pc, PetscInt n)
1435: {
1436: PetscFunctionBegin;
1438: PetscTryMethod(pc, "PCGAMGASMSetHEM_C", (PC, PetscInt), (pc, n));
1439: PetscFunctionReturn(PETSC_SUCCESS);
1440: }
1442: static PetscErrorCode PCGAMGASMSetHEM_GAMG(PC pc, PetscInt n)
1443: {
1444: PC_MG *mg = (PC_MG *)pc->data;
1445: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1447: PetscFunctionBegin;
1448: pc_gamg->asm_hem_aggs = n;
1449: PetscFunctionReturn(PETSC_SUCCESS);
1450: }
1452: /*@
1453: PCGAMGSetThreshold - Relative threshold to use for dropping edges in aggregation graph
1455: Not Collective
1457: Input Parameters:
1458: + pc - the preconditioner context
1459: . v - array of threshold values for finest n levels; 0.0 means keep all nonzero entries in the graph; negative means keep even zero entries in the graph
1460: - n - number of threshold values provided in array
1462: Options Database Key:
1463: . -pc_gamg_threshold <threshold> - the threshold to drop edges
1465: Level: intermediate
1467: Notes:
1468: Increasing the threshold decreases the rate of coarsening. Conversely reducing the threshold increases the rate of coarsening (aggressive coarsening) and thereby reduces the complexity of the coarse grids, and generally results in slower solver converge rates. Reducing coarse grid complexity reduced the complexity of Galerkin coarse grid construction considerably.
1469: Before coarsening or aggregating the graph, `PCGAMG` removes small values from the graph with this threshold, and thus reducing the coupling in the graph and a different (perhaps better) coarser set of points.
1471: If `n` is less than the total number of coarsenings (see `PCGAMGSetNlevels()`), then threshold scaling (see `PCGAMGSetThresholdScale()`) is used for each successive coarsening.
1472: In this case, `PCGAMGSetThresholdScale()` must be called before `PCGAMGSetThreshold()`.
1473: If `n` is greater than the total number of levels, the excess entries in threshold will not be used.
1475: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetAggressiveLevels()`, `PCGAMGMISkSetAggressive()`, `PCGAMGSetMinDegreeOrderingMISk()`, `PCGAMGSetThresholdScale()`
1476: @*/
1477: PetscErrorCode PCGAMGSetThreshold(PC pc, PetscReal v[], PetscInt n)
1478: {
1479: PetscFunctionBegin;
1481: if (n) PetscAssertPointer(v, 2);
1482: PetscTryMethod(pc, "PCGAMGSetThreshold_C", (PC, PetscReal[], PetscInt), (pc, v, n));
1483: PetscFunctionReturn(PETSC_SUCCESS);
1484: }
1486: static PetscErrorCode PCGAMGSetThreshold_GAMG(PC pc, PetscReal v[], PetscInt n)
1487: {
1488: PC_MG *mg = (PC_MG *)pc->data;
1489: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1490: PetscInt i;
1492: PetscFunctionBegin;
1493: for (i = 0; i < PetscMin(n, PETSC_MG_MAXLEVELS); i++) pc_gamg->threshold[i] = v[i];
1494: for (; i < PETSC_MG_MAXLEVELS; i++) pc_gamg->threshold[i] = pc_gamg->threshold[i - 1] * pc_gamg->threshold_scale;
1495: PetscFunctionReturn(PETSC_SUCCESS);
1496: }
1498: /*@
1499: PCGAMGSetRankReductionFactors - Set a manual schedule for MPI rank reduction on coarse grids
1501: Collective
1503: Input Parameters:
1504: + pc - the preconditioner context
1505: . v - array of reduction factors. 0 for first value forces a reduction to one process/device on first level in CUDA
1506: - n - number of values provided in array
1508: Options Database Key:
1509: . -pc_gamg_rank_reduction_factors <factors> - provide the schedule
1511: Level: intermediate
1513: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetProcEqLim()`, `PCGAMGSetCoarseEqLim()`, `PCGAMGSetParallelCoarseGridSolve()`
1514: @*/
1515: PetscErrorCode PCGAMGSetRankReductionFactors(PC pc, PetscInt v[], PetscInt n)
1516: {
1517: PetscFunctionBegin;
1519: if (n) PetscAssertPointer(v, 2);
1520: PetscTryMethod(pc, "PCGAMGSetRankReductionFactors_C", (PC, PetscInt[], PetscInt), (pc, v, n));
1521: PetscFunctionReturn(PETSC_SUCCESS);
1522: }
1524: static PetscErrorCode PCGAMGSetRankReductionFactors_GAMG(PC pc, PetscInt v[], PetscInt n)
1525: {
1526: PC_MG *mg = (PC_MG *)pc->data;
1527: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1528: PetscInt i;
1530: PetscFunctionBegin;
1531: for (i = 0; i < PetscMin(n, PETSC_MG_MAXLEVELS); i++) pc_gamg->level_reduction_factors[i] = v[i];
1532: for (; i < PETSC_MG_MAXLEVELS; i++) pc_gamg->level_reduction_factors[i] = -1; /* 0 stop putting one process/device on first level */
1533: PetscFunctionReturn(PETSC_SUCCESS);
1534: }
1536: /*@
1537: PCGAMGSetThresholdScale - Relative threshold reduction at each level
1539: Not Collective
1541: Input Parameters:
1542: + pc - the preconditioner context
1543: - v - the threshold value reduction, usually < 1.0
1545: Options Database Key:
1546: . -pc_gamg_threshold_scale <v> - set the relative threshold reduction on each level
1548: Level: advanced
1550: Note:
1551: The initial threshold (for an arbitrary number of levels starting from the finest) can be set with `PCGAMGSetThreshold()`.
1552: This scaling is used for each subsequent coarsening, but must be called before `PCGAMGSetThreshold()`.
1554: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetThreshold()`
1555: @*/
1556: PetscErrorCode PCGAMGSetThresholdScale(PC pc, PetscReal v)
1557: {
1558: PetscFunctionBegin;
1560: PetscTryMethod(pc, "PCGAMGSetThresholdScale_C", (PC, PetscReal), (pc, v));
1561: PetscFunctionReturn(PETSC_SUCCESS);
1562: }
1564: static PetscErrorCode PCGAMGSetThresholdScale_GAMG(PC pc, PetscReal v)
1565: {
1566: PC_MG *mg = (PC_MG *)pc->data;
1567: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1569: PetscFunctionBegin;
1570: pc_gamg->threshold_scale = v;
1571: PetscFunctionReturn(PETSC_SUCCESS);
1572: }
1574: /*@C
1575: PCGAMGSetType - Set the type of algorithm `PCGAMG` should use
1577: Collective
1579: Input Parameters:
1580: + pc - the preconditioner context
1581: - type - `PCGAMGAGG`, `PCGAMGGEO`, or `PCGAMGCLASSICAL`
1583: Options Database Key:
1584: . -pc_gamg_type <agg,geo,classical> - type of algebraic multigrid to apply - only agg is supported
1586: Level: intermediate
1588: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMGGetType()`, `PCGAMG`, `PCGAMGType`
1589: @*/
1590: PetscErrorCode PCGAMGSetType(PC pc, PCGAMGType type)
1591: {
1592: PetscFunctionBegin;
1594: PetscTryMethod(pc, "PCGAMGSetType_C", (PC, PCGAMGType), (pc, type));
1595: PetscFunctionReturn(PETSC_SUCCESS);
1596: }
1598: /*@C
1599: PCGAMGGetType - Get the type of algorithm `PCGAMG` will use
1601: Collective
1603: Input Parameter:
1604: . pc - the preconditioner context
1606: Output Parameter:
1607: . type - the type of algorithm used
1609: Level: intermediate
1611: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCGAMG`, `PCGAMGSetType()`, `PCGAMGType`
1612: @*/
1613: PetscErrorCode PCGAMGGetType(PC pc, PCGAMGType *type)
1614: {
1615: PetscFunctionBegin;
1617: PetscUseMethod(pc, "PCGAMGGetType_C", (PC, PCGAMGType *), (pc, type));
1618: PetscFunctionReturn(PETSC_SUCCESS);
1619: }
1621: static PetscErrorCode PCGAMGGetType_GAMG(PC pc, PCGAMGType *type)
1622: {
1623: PC_MG *mg = (PC_MG *)pc->data;
1624: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1626: PetscFunctionBegin;
1627: *type = pc_gamg->type;
1628: PetscFunctionReturn(PETSC_SUCCESS);
1629: }
1631: static PetscErrorCode PCGAMGSetType_GAMG(PC pc, PCGAMGType type)
1632: {
1633: PC_MG *mg = (PC_MG *)pc->data;
1634: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1635: PetscErrorCode (*r)(PC);
1637: PetscFunctionBegin;
1638: pc_gamg->type = type;
1639: PetscCall(PetscFunctionListFind(GAMGList, type, &r));
1640: PetscCheck(r, PetscObjectComm((PetscObject)pc), PETSC_ERR_ARG_UNKNOWN_TYPE, "Unknown GAMG type %s given", type);
1641: if (pc_gamg->ops->destroy) {
1642: PetscCall((*pc_gamg->ops->destroy)(pc));
1643: PetscCall(PetscMemzero(pc_gamg->ops, sizeof(struct _PCGAMGOps)));
1644: pc_gamg->ops->createlevel = PCGAMGCreateLevel_GAMG;
1645: /* cleaning up common data in pc_gamg - this should disappear someday */
1646: pc_gamg->data_cell_cols = 0;
1647: pc_gamg->data_cell_rows = 0;
1648: pc_gamg->orig_data_cell_cols = 0;
1649: pc_gamg->orig_data_cell_rows = 0;
1650: PetscCall(PetscFree(pc_gamg->data));
1651: pc_gamg->data_sz = 0;
1652: }
1653: PetscCall(PetscFree(pc_gamg->gamg_type_name));
1654: PetscCall(PetscStrallocpy(type, &pc_gamg->gamg_type_name));
1655: PetscCall((*r)(pc));
1656: PetscFunctionReturn(PETSC_SUCCESS);
1657: }
1659: static PetscErrorCode PCView_GAMG(PC pc, PetscViewer viewer)
1660: {
1661: PC_MG *mg = (PC_MG *)pc->data;
1662: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1663: PetscReal gc = 0, oc = 0;
1665: PetscFunctionBegin;
1666: PetscCall(PetscViewerASCIIPrintf(viewer, " GAMG specific options\n"));
1667: PetscCall(PetscViewerASCIIPrintf(viewer, " Threshold for dropping small values in graph on each level ="));
1668: for (PetscInt i = 0; i < mg->nlevels; i++) PetscCall(PetscViewerASCIIPrintf(viewer, " %g", (double)pc_gamg->threshold[i]));
1669: PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
1670: PetscCall(PetscViewerASCIIPrintf(viewer, " Threshold scaling factor for each level not specified = %g\n", (double)pc_gamg->threshold_scale));
1671: if (pc_gamg->use_aggs_in_asm) PetscCall(PetscViewerASCIIPrintf(viewer, " Using aggregates from coarsening process to define subdomains for PCASM\n")); // this take presedence
1672: else if (pc_gamg->asm_hem_aggs) PetscCall(PetscViewerASCIIPrintf(viewer, " Using aggregates made with %d applications of heavy edge matching (HEM) to define subdomains for PCASM\n", (int)pc_gamg->asm_hem_aggs));
1673: if (pc_gamg->use_parallel_coarse_grid_solver) PetscCall(PetscViewerASCIIPrintf(viewer, " Using parallel coarse grid solver (all coarse grid equations not put on one process)\n"));
1674: if (pc_gamg->injection_index_size) {
1675: PetscCall(PetscViewerASCIIPrintf(viewer, " Using injection restriction/prolongation on first level, dofs:"));
1676: for (int i = 0; i < pc_gamg->injection_index_size; i++) PetscCall(PetscViewerASCIIPrintf(viewer, " %d", (int)pc_gamg->injection_index[i]));
1677: PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
1678: }
1679: if (pc_gamg->ops->view) PetscCall((*pc_gamg->ops->view)(pc, viewer));
1680: PetscCall(PCMGGetGridComplexity(pc, &gc, &oc));
1681: PetscCall(PetscViewerASCIIPrintf(viewer, " Complexity: grid = %g operator = %g\n", (double)gc, (double)oc));
1682: PetscFunctionReturn(PETSC_SUCCESS);
1683: }
1685: /*@
1686: PCGAMGSetInjectionIndex - Array of subset of variables per vertex to inject into coarse grid space
1688: Logically Collective
1690: Input Parameters:
1691: + pc - the coarsen context
1692: . n - number of indices
1693: - idx - array of indices
1695: Options Database Key:
1696: . -pc_gamg_injection_index - array of subset of variables per vertex to use for injection coarse grid space
1698: Level: intermediate
1700: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), `PCGAMG`
1701: @*/
1702: PetscErrorCode PCGAMGSetInjectionIndex(PC pc, PetscInt n, PetscInt idx[])
1703: {
1704: PetscFunctionBegin;
1707: PetscTryMethod(pc, "PCGAMGSetInjectionIndex_C", (PC, PetscInt, PetscInt[]), (pc, n, idx));
1708: PetscFunctionReturn(PETSC_SUCCESS);
1709: }
1711: static PetscErrorCode PCGAMGSetInjectionIndex_GAMG(PC pc, PetscInt n, PetscInt idx[])
1712: {
1713: PC_MG *mg = (PC_MG *)pc->data;
1714: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1716: PetscFunctionBegin;
1717: pc_gamg->injection_index_size = n;
1718: PetscCheck(n < MAT_COARSEN_STRENGTH_INDEX_SIZE, PetscObjectComm((PetscObject)pc), PETSC_ERR_ARG_INCOMP, "array size %d larger than max %d", (int)n, MAT_COARSEN_STRENGTH_INDEX_SIZE);
1719: for (PetscInt i = 0; i < n; i++) pc_gamg->injection_index[i] = idx[i];
1720: PetscFunctionReturn(PETSC_SUCCESS);
1721: }
1723: static PetscErrorCode PCSetFromOptions_GAMG(PC pc, PetscOptionItems *PetscOptionsObject)
1724: {
1725: PC_MG *mg = (PC_MG *)pc->data;
1726: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
1727: PetscBool flag;
1728: MPI_Comm comm;
1729: char prefix[256], tname[32];
1730: PetscInt i, n;
1731: const char *pcpre;
1732: static const char *LayoutTypes[] = {"compact", "spread", "PCGAMGLayoutType", "PC_GAMG_LAYOUT", NULL};
1734: PetscFunctionBegin;
1735: PetscCall(PetscObjectGetComm((PetscObject)pc, &comm));
1736: PetscOptionsHeadBegin(PetscOptionsObject, "GAMG options");
1737: PetscCall(PetscOptionsInt("-pc_mg_levels", "Number of Levels", "PCMGSetLevels", -1, &n, &flag));
1738: PetscCheck(!flag, comm, PETSC_ERR_ARG_WRONG, "Invalid flag -pc_mg_levels. GAMG does not allow the number of levels to be set.");
1739: PetscCall(PetscOptionsFList("-pc_gamg_type", "Type of AMG method (only 'agg' supported and useful)", "PCGAMGSetType", GAMGList, pc_gamg->gamg_type_name, tname, sizeof(tname), &flag));
1740: if (flag) PetscCall(PCGAMGSetType(pc, tname));
1741: PetscCall(PetscOptionsBool("-pc_gamg_repartition", "Repartion coarse grids", "PCGAMGSetRepartition", pc_gamg->repart, &pc_gamg->repart, NULL));
1742: PetscCall(PetscOptionsBool("-pc_gamg_use_sa_esteig", "Use eigen estimate from smoothed aggregation for smoother", "PCGAMGSetUseSAEstEig", pc_gamg->use_sa_esteig, &pc_gamg->use_sa_esteig, NULL));
1743: PetscCall(PetscOptionsBool("-pc_gamg_recompute_esteig", "Set flag to recompute eigen estimates for Chebyshev when matrix changes", "PCGAMGSetRecomputeEstEig", pc_gamg->recompute_esteig, &pc_gamg->recompute_esteig, NULL));
1744: PetscCall(PetscOptionsBool("-pc_gamg_reuse_interpolation", "Reuse prolongation operator", "PCGAMGReuseInterpolation", pc_gamg->reuse_prol, &pc_gamg->reuse_prol, NULL));
1745: PetscCall(PetscOptionsBool("-pc_gamg_asm_use_agg", "Use aggregation aggregates for ASM smoother", "PCGAMGASMSetUseAggs", pc_gamg->use_aggs_in_asm, &pc_gamg->use_aggs_in_asm, NULL));
1746: PetscCall(PetscOptionsBool("-pc_gamg_parallel_coarse_grid_solver", "Use parallel coarse grid solver (otherwise put last grid on one process)", "PCGAMGSetParallelCoarseGridSolve", pc_gamg->use_parallel_coarse_grid_solver, &pc_gamg->use_parallel_coarse_grid_solver, NULL));
1747: PetscCall(PetscOptionsBool("-pc_gamg_cpu_pin_coarse_grids", "Pin coarse grids to the CPU", "PCGAMGSetCpuPinCoarseGrids", pc_gamg->cpu_pin_coarse_grids, &pc_gamg->cpu_pin_coarse_grids, NULL));
1748: PetscCall(PetscOptionsEnum("-pc_gamg_coarse_grid_layout_type", "compact: place reduced grids on processes in natural order; spread: distribute to whole machine for more memory bandwidth", "PCGAMGSetCoarseGridLayoutType", LayoutTypes,
1749: (PetscEnum)pc_gamg->layout_type, (PetscEnum *)&pc_gamg->layout_type, NULL));
1750: PetscCall(PetscOptionsInt("-pc_gamg_process_eq_limit", "Limit (goal) on number of equations per process on coarse grids", "PCGAMGSetProcEqLim", pc_gamg->min_eq_proc, &pc_gamg->min_eq_proc, NULL));
1751: PetscCall(PetscOptionsInt("-pc_gamg_coarse_eq_limit", "Limit on number of equations for the coarse grid", "PCGAMGSetCoarseEqLim", pc_gamg->coarse_eq_limit, &pc_gamg->coarse_eq_limit, NULL));
1752: PetscCall(PetscOptionsInt("-pc_gamg_asm_hem_aggs", "Number of HEM matching passed in aggregates for ASM smoother", "PCGAMGASMSetHEM", pc_gamg->asm_hem_aggs, &pc_gamg->asm_hem_aggs, NULL));
1753: PetscCall(PetscOptionsReal("-pc_gamg_threshold_scale", "Scaling of threshold for each level not specified", "PCGAMGSetThresholdScale", pc_gamg->threshold_scale, &pc_gamg->threshold_scale, NULL));
1754: n = PETSC_MG_MAXLEVELS;
1755: PetscCall(PetscOptionsRealArray("-pc_gamg_threshold", "Relative threshold to use for dropping edges in aggregation graph", "PCGAMGSetThreshold", pc_gamg->threshold, &n, &flag));
1756: if (!flag || n < PETSC_MG_MAXLEVELS) {
1757: if (!flag) n = 1;
1758: i = n;
1759: do {
1760: pc_gamg->threshold[i] = pc_gamg->threshold[i - 1] * pc_gamg->threshold_scale;
1761: } while (++i < PETSC_MG_MAXLEVELS);
1762: }
1763: PetscCall(PetscOptionsInt("-pc_mg_levels", "Set number of MG levels (should get from base class)", "PCGAMGSetNlevels", pc_gamg->Nlevels, &pc_gamg->Nlevels, NULL));
1764: PetscCheck(pc_gamg->Nlevels <= PETSC_MG_MAXLEVELS, PetscObjectComm((PetscObject)pc), PETSC_ERR_ARG_INCOMP, "-pc_mg_levels (%d) >= PETSC_MG_MAXLEVELS (%d)", (int)pc_gamg->Nlevels, (int)PETSC_MG_MAXLEVELS);
1765: n = PETSC_MG_MAXLEVELS;
1766: PetscCall(PetscOptionsIntArray("-pc_gamg_rank_reduction_factors", "Manual schedule of coarse grid reduction factors that overrides internal heuristics (0 for first reduction puts one process/device)", "PCGAMGSetRankReductionFactors", pc_gamg->level_reduction_factors, &n, &flag));
1767: if (!flag) i = 0;
1768: else i = n;
1769: do {
1770: pc_gamg->level_reduction_factors[i] = -1;
1771: } while (++i < PETSC_MG_MAXLEVELS);
1772: {
1773: PetscReal eminmax[2] = {0., 0.};
1774: n = 2;
1775: PetscCall(PetscOptionsRealArray("-pc_gamg_eigenvalues", "extreme eigenvalues for smoothed aggregation", "PCGAMGSetEigenvalues", eminmax, &n, &flag));
1776: if (flag) {
1777: PetscCheck(n == 2, PetscObjectComm((PetscObject)pc), PETSC_ERR_ARG_INCOMP, "-pc_gamg_eigenvalues: must specify 2 parameters, min and max eigenvalues");
1778: PetscCall(PCGAMGSetEigenvalues(pc, eminmax[1], eminmax[0]));
1779: }
1780: }
1781: pc_gamg->injection_index_size = MAT_COARSEN_STRENGTH_INDEX_SIZE;
1782: PetscCall(PetscOptionsIntArray("-pc_gamg_injection_index", "Array of indices to use to use injection coarse grid space", "PCGAMGSetInjectionIndex", pc_gamg->injection_index, &pc_gamg->injection_index_size, NULL));
1783: /* set options for subtype */
1784: PetscCall((*pc_gamg->ops->setfromoptions)(pc, PetscOptionsObject));
1786: PetscCall(PCGetOptionsPrefix(pc, &pcpre));
1787: PetscCall(PetscSNPrintf(prefix, sizeof(prefix), "%spc_gamg_", pcpre ? pcpre : ""));
1788: PetscOptionsHeadEnd();
1789: PetscFunctionReturn(PETSC_SUCCESS);
1790: }
1792: /*MC
1793: PCGAMG - Geometric algebraic multigrid (AMG) preconditioner
1795: Options Database Keys:
1796: + -pc_gamg_type <type,default=agg> - one of agg, geo, or classical (only smoothed aggregation, agg, supported)
1797: . -pc_gamg_repartition <bool,default=false> - repartition the degrees of freedom across the coarse grids as they are determined
1798: . -pc_gamg_asm_use_agg <bool,default=false> - use the aggregates from the coasening process to defined the subdomains on each level for the PCASM smoother
1799: . -pc_gamg_process_eq_limit <limit, default=50> - `PCGAMG` will reduce the number of MPI ranks used directly on the coarse grids so that there are around <limit>
1800: equations on each process that has degrees of freedom
1801: . -pc_gamg_coarse_eq_limit <limit, default=50> - Set maximum number of equations on coarsest grid to aim for.
1802: . -pc_gamg_reuse_interpolation <bool,default=true> - when rebuilding the algebraic multigrid preconditioner reuse the previously computed interpolations (should always be true)
1803: . -pc_gamg_threshold[] <thresh,default=[-1,...]> - Before aggregating the graph `PCGAMG` will remove small values from the graph on each level (< 0 does no filtering)
1804: - -pc_gamg_threshold_scale <scale,default=1> - Scaling of threshold on each coarser grid if not specified
1806: Options Database Keys for Aggregation:
1807: + -pc_gamg_agg_nsmooths <nsmooth, default=1> - number of smoothing steps to use with smooth aggregation to construct prolongation
1808: . -pc_gamg_aggressive_coarsening <n,default=1> - number of aggressive coarsening (MIS-2) levels from finest.
1809: . -pc_gamg_aggressive_square_graph <bool,default=false> - Use square graph (A'A) or MIS-k (k=2) for aggressive coarsening
1810: . -pc_gamg_mis_k_minimum_degree_ordering <bool,default=true> - Use minimum degree ordering in greedy MIS algorithm
1811: . -pc_gamg_pc_gamg_asm_hem_aggs <n,default=0> - Number of HEM aggregation steps for ASM smoother
1812: - -pc_gamg_aggressive_mis_k <n,default=2> - Number (k) distance in MIS coarsening (>2 is 'aggressive')
1814: Options Database Keys for Multigrid:
1815: + -pc_mg_cycle_type <v> - v or w, see `PCMGSetCycleType()`
1816: . -pc_mg_distinct_smoothup - configure the up and down (pre and post) smoothers separately, see PCMGSetDistinctSmoothUp()
1817: . -pc_mg_type <multiplicative> - (one of) additive multiplicative full kascade
1818: - -pc_mg_levels <levels> - Number of levels of multigrid to use. GAMG has a heuristic so pc_mg_levels is not usually used with GAMG
1820: Level: intermediate
1822: Notes:
1823: To obtain good performance for `PCGAMG` for vector valued problems you must
1824: call `MatSetBlockSize()` to indicate the number of degrees of freedom per grid point
1825: call `MatSetNearNullSpace()` (or `PCSetCoordinates()` if solving the equations of elasticity) to indicate the near null space of the operator
1827: The many options for `PCMG` also work directly for `PCGAMG` such as controlling the smoothers on each level etc.
1829: .seealso: [the Users Manual section on PCGAMG](sec_amg), [the Users Manual section on PCMG](sec_mg), [](ch_ksp), `PCCreate()`, `PCSetType()`,
1830: `MatSetBlockSize()`,
1831: `PCMGType`, `PCSetCoordinates()`, `MatSetNearNullSpace()`, `PCGAMGSetType()`, `PCGAMGAGG`, `PCGAMGGEO`, `PCGAMGCLASSICAL`, `PCGAMGSetProcEqLim()`,
1832: `PCGAMGSetCoarseEqLim()`, `PCGAMGSetRepartition()`, `PCGAMGRegister()`, `PCGAMGSetReuseInterpolation()`, `PCGAMGASMSetUseAggs()`,
1833: `PCGAMGSetParallelCoarseGridSolve()`, `PCGAMGSetNlevels()`, `PCGAMGSetThreshold()`, `PCGAMGGetType()`, `PCGAMGSetUseSAEstEig()`
1834: M*/
1835: PETSC_EXTERN PetscErrorCode PCCreate_GAMG(PC pc)
1836: {
1837: PC_GAMG *pc_gamg;
1838: PC_MG *mg;
1840: PetscFunctionBegin;
1841: /* register AMG type */
1842: PetscCall(PCGAMGInitializePackage());
1844: /* PCGAMG is an inherited class of PCMG. Initialize pc as PCMG */
1845: PetscCall(PCSetType(pc, PCMG));
1846: PetscCall(PetscObjectChangeTypeName((PetscObject)pc, PCGAMG));
1848: /* create a supporting struct and attach it to pc */
1849: PetscCall(PetscNew(&pc_gamg));
1850: PetscCall(PCMGSetGalerkin(pc, PC_MG_GALERKIN_EXTERNAL));
1851: mg = (PC_MG *)pc->data;
1852: mg->innerctx = pc_gamg;
1854: PetscCall(PetscNew(&pc_gamg->ops));
1856: /* these should be in subctx but repartitioning needs simple arrays */
1857: pc_gamg->data_sz = 0;
1858: pc_gamg->data = NULL;
1860: /* overwrite the pointers of PCMG by the functions of base class PCGAMG */
1861: pc->ops->setfromoptions = PCSetFromOptions_GAMG;
1862: pc->ops->setup = PCSetUp_GAMG;
1863: pc->ops->reset = PCReset_GAMG;
1864: pc->ops->destroy = PCDestroy_GAMG;
1865: mg->view = PCView_GAMG;
1867: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetProcEqLim_C", PCGAMGSetProcEqLim_GAMG));
1868: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetCoarseEqLim_C", PCGAMGSetCoarseEqLim_GAMG));
1869: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetRepartition_C", PCGAMGSetRepartition_GAMG));
1870: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetEigenvalues_C", PCGAMGSetEigenvalues_GAMG));
1871: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetUseSAEstEig_C", PCGAMGSetUseSAEstEig_GAMG));
1872: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetRecomputeEstEig_C", PCGAMGSetRecomputeEstEig_GAMG));
1873: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetReuseInterpolation_C", PCGAMGSetReuseInterpolation_GAMG));
1874: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGASMSetUseAggs_C", PCGAMGASMSetUseAggs_GAMG));
1875: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetParallelCoarseGridSolve_C", PCGAMGSetParallelCoarseGridSolve_GAMG));
1876: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetCpuPinCoarseGrids_C", PCGAMGSetCpuPinCoarseGrids_GAMG));
1877: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetCoarseGridLayoutType_C", PCGAMGSetCoarseGridLayoutType_GAMG));
1878: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetThreshold_C", PCGAMGSetThreshold_GAMG));
1879: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetRankReductionFactors_C", PCGAMGSetRankReductionFactors_GAMG));
1880: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetThresholdScale_C", PCGAMGSetThresholdScale_GAMG));
1881: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetType_C", PCGAMGSetType_GAMG));
1882: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGGetType_C", PCGAMGGetType_GAMG));
1883: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetNlevels_C", PCGAMGSetNlevels_GAMG));
1884: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGASMSetHEM_C", PCGAMGASMSetHEM_GAMG));
1885: PetscCall(PetscObjectComposeFunction((PetscObject)pc, "PCGAMGSetInjectionIndex_C", PCGAMGSetInjectionIndex_GAMG));
1886: pc_gamg->repart = PETSC_FALSE;
1887: pc_gamg->reuse_prol = PETSC_TRUE;
1888: pc_gamg->use_aggs_in_asm = PETSC_FALSE;
1889: pc_gamg->use_parallel_coarse_grid_solver = PETSC_FALSE;
1890: pc_gamg->cpu_pin_coarse_grids = PETSC_FALSE;
1891: pc_gamg->layout_type = PCGAMG_LAYOUT_SPREAD;
1892: pc_gamg->min_eq_proc = 50;
1893: pc_gamg->asm_hem_aggs = 0;
1894: pc_gamg->coarse_eq_limit = 50;
1895: for (int i = 0; i < PETSC_MG_MAXLEVELS; i++) pc_gamg->threshold[i] = -1;
1896: pc_gamg->threshold_scale = 1.;
1897: pc_gamg->Nlevels = PETSC_MG_MAXLEVELS;
1898: pc_gamg->current_level = 0; /* don't need to init really */
1899: pc_gamg->use_sa_esteig = PETSC_TRUE;
1900: pc_gamg->recompute_esteig = PETSC_TRUE;
1901: pc_gamg->emin = 0;
1902: pc_gamg->emax = 0;
1904: pc_gamg->ops->createlevel = PCGAMGCreateLevel_GAMG;
1906: /* PCSetUp_GAMG assumes that the type has been set, so set it to the default now */
1907: PetscCall(PCGAMGSetType(pc, PCGAMGAGG));
1908: PetscFunctionReturn(PETSC_SUCCESS);
1909: }
1911: /*@C
1912: PCGAMGInitializePackage - This function initializes everything in the `PCGAMG` package. It is called
1913: from `PCInitializePackage()`.
1915: Level: developer
1917: .seealso: [](ch_ksp), `PetscInitialize()`
1918: @*/
1919: PetscErrorCode PCGAMGInitializePackage(void)
1920: {
1921: PetscInt l;
1923: PetscFunctionBegin;
1924: if (PCGAMGPackageInitialized) PetscFunctionReturn(PETSC_SUCCESS);
1925: PCGAMGPackageInitialized = PETSC_TRUE;
1926: PetscCall(PetscFunctionListAdd(&GAMGList, PCGAMGGEO, PCCreateGAMG_GEO));
1927: PetscCall(PetscFunctionListAdd(&GAMGList, PCGAMGAGG, PCCreateGAMG_AGG));
1928: PetscCall(PetscFunctionListAdd(&GAMGList, PCGAMGCLASSICAL, PCCreateGAMG_Classical));
1929: PetscCall(PetscRegisterFinalize(PCGAMGFinalizePackage));
1931: /* general events */
1932: PetscCall(PetscLogEventRegister("PCSetUp_GAMG+", PC_CLASSID, &petsc_gamg_setup_events[GAMG_SETUP]));
1933: PetscCall(PetscLogEventRegister(" PCGAMGCreateG", PC_CLASSID, &petsc_gamg_setup_events[GAMG_GRAPH]));
1934: PetscCall(PetscLogEventRegister(" GAMG Coarsen", PC_CLASSID, &petsc_gamg_setup_events[GAMG_COARSEN]));
1935: PetscCall(PetscLogEventRegister(" GAMG MIS/Agg", PC_CLASSID, &petsc_gamg_setup_events[GAMG_MIS]));
1936: PetscCall(PetscLogEventRegister(" PCGAMGProl", PC_CLASSID, &petsc_gamg_setup_events[GAMG_PROL]));
1937: PetscCall(PetscLogEventRegister(" GAMG Prol-col", PC_CLASSID, &petsc_gamg_setup_events[GAMG_PROLA]));
1938: PetscCall(PetscLogEventRegister(" GAMG Prol-lift", PC_CLASSID, &petsc_gamg_setup_events[GAMG_PROLB]));
1939: PetscCall(PetscLogEventRegister(" PCGAMGOptProl", PC_CLASSID, &petsc_gamg_setup_events[GAMG_OPT]));
1940: PetscCall(PetscLogEventRegister(" GAMG smooth", PC_CLASSID, &petsc_gamg_setup_events[GAMG_OPTSM]));
1941: PetscCall(PetscLogEventRegister(" PCGAMGCreateL", PC_CLASSID, &petsc_gamg_setup_events[GAMG_LEVEL]));
1942: PetscCall(PetscLogEventRegister(" GAMG PtAP", PC_CLASSID, &petsc_gamg_setup_events[GAMG_PTAP]));
1943: PetscCall(PetscLogEventRegister(" GAMG Reduce", PC_CLASSID, &petsc_gamg_setup_events[GAMG_REDUCE]));
1944: PetscCall(PetscLogEventRegister(" GAMG Repart", PC_CLASSID, &petsc_gamg_setup_events[GAMG_REPART]));
1945: /* PetscCall(PetscLogEventRegister(" GAMG Inv-Srt", PC_CLASSID, &petsc_gamg_setup_events[SET13])); */
1946: /* PetscCall(PetscLogEventRegister(" GAMG Move A", PC_CLASSID, &petsc_gamg_setup_events[SET14])); */
1947: /* PetscCall(PetscLogEventRegister(" GAMG Move P", PC_CLASSID, &petsc_gamg_setup_events[SET15])); */
1948: for (l = 0; l < PETSC_MG_MAXLEVELS; l++) {
1949: char ename[32];
1951: PetscCall(PetscSNPrintf(ename, sizeof(ename), "PCGAMG Squ l%02" PetscInt_FMT, l));
1952: PetscCall(PetscLogEventRegister(ename, PC_CLASSID, &petsc_gamg_setup_matmat_events[l][0]));
1953: PetscCall(PetscSNPrintf(ename, sizeof(ename), "PCGAMG Gal l%02" PetscInt_FMT, l));
1954: PetscCall(PetscLogEventRegister(ename, PC_CLASSID, &petsc_gamg_setup_matmat_events[l][1]));
1955: PetscCall(PetscSNPrintf(ename, sizeof(ename), "PCGAMG Opt l%02" PetscInt_FMT, l));
1956: PetscCall(PetscLogEventRegister(ename, PC_CLASSID, &petsc_gamg_setup_matmat_events[l][2]));
1957: }
1958: #if defined(GAMG_STAGES)
1959: { /* create timer stages */
1960: char str[32];
1961: PetscCall(PetscSNPrintf(str, PETSC_STATIC_ARRAY_LENGTH(str), "GAMG Level %d", 0));
1962: PetscCall(PetscLogStageRegister(str, &gamg_stages[0]));
1963: }
1964: #endif
1965: PetscFunctionReturn(PETSC_SUCCESS);
1966: }
1968: /*@C
1969: PCGAMGFinalizePackage - This function frees everything from the `PCGAMG` package. It is
1970: called from `PetscFinalize()` automatically.
1972: Level: developer
1974: .seealso: [](ch_ksp), `PetscFinalize()`
1975: @*/
1976: PetscErrorCode PCGAMGFinalizePackage(void)
1977: {
1978: PetscFunctionBegin;
1979: PCGAMGPackageInitialized = PETSC_FALSE;
1980: PetscCall(PetscFunctionListDestroy(&GAMGList));
1981: PetscFunctionReturn(PETSC_SUCCESS);
1982: }
1984: /*@C
1985: PCGAMGRegister - Register a `PCGAMG` implementation.
1987: Input Parameters:
1988: + type - string that will be used as the name of the `PCGAMG` type.
1989: - create - function for creating the gamg context.
1991: Level: developer
1993: .seealso: [](ch_ksp), `PCGAMGType`, `PCGAMG`, `PCGAMGSetType()`
1994: @*/
1995: PetscErrorCode PCGAMGRegister(PCGAMGType type, PetscErrorCode (*create)(PC))
1996: {
1997: PetscFunctionBegin;
1998: PetscCall(PCGAMGInitializePackage());
1999: PetscCall(PetscFunctionListAdd(&GAMGList, type, create));
2000: PetscFunctionReturn(PETSC_SUCCESS);
2001: }
2003: /*@C
2004: PCGAMGCreateGraph - Creates a graph that is used by the ``PCGAMGType`` in the coarsening process
2006: Input Parameters:
2007: + pc - the `PCGAMG`
2008: - A - the matrix, for any level
2010: Output Parameter:
2011: . G - the graph
2013: Level: advanced
2015: .seealso: [](ch_ksp), `PCGAMGType`, `PCGAMG`, `PCGAMGSetType()`
2016: @*/
2017: PetscErrorCode PCGAMGCreateGraph(PC pc, Mat A, Mat *G)
2018: {
2019: PC_MG *mg = (PC_MG *)pc->data;
2020: PC_GAMG *pc_gamg = (PC_GAMG *)mg->innerctx;
2022: PetscFunctionBegin;
2023: PetscCall(pc_gamg->ops->creategraph(pc, A, G));
2024: PetscFunctionReturn(PETSC_SUCCESS);
2025: }