Actual source code: aomemscalable.c
1: /*
2: The memory scalable AO application ordering routines. These store the
3: orderings on each process for that process' range of values, this is more memory-efficient than `AOBASIC`
4: */
6: #include <../src/vec/is/ao/aoimpl.h>
8: typedef struct {
9: PetscInt *app_loc; /* app_loc[i] is the partner for the ith local PETSc slot */
10: PetscInt *petsc_loc; /* petsc_loc[j] is the partner for the jth local app slot */
11: PetscLayout map; /* determines the local sizes of ao */
12: } AO_MemoryScalable;
14: /*
15: All processors ship the data to process 0 to be printed; note that this is not scalable because
16: process 0 allocates space for all the orderings entry across all the processes
17: */
18: static PetscErrorCode AOView_MemoryScalable(AO ao, PetscViewer viewer)
19: {
20: PetscMPIInt rank, size;
21: AO_MemoryScalable *aomems = (AO_MemoryScalable *)ao->data;
22: PetscBool iascii;
23: PetscMPIInt tag_app, tag_petsc;
24: PetscLayout map = aomems->map;
25: PetscInt *app, *app_loc, *petsc, *petsc_loc, len, i, j;
26: MPI_Status status;
28: PetscFunctionBegin;
29: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &iascii));
30: PetscCheck(iascii, PetscObjectComm((PetscObject)viewer), PETSC_ERR_SUP, "Viewer type %s not supported for AO MemoryScalable", ((PetscObject)viewer)->type_name);
32: PetscCallMPI(MPI_Comm_rank(PetscObjectComm((PetscObject)ao), &rank));
33: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)ao), &size));
35: PetscCall(PetscObjectGetNewTag((PetscObject)ao, &tag_app));
36: PetscCall(PetscObjectGetNewTag((PetscObject)ao, &tag_petsc));
38: if (rank == 0) {
39: PetscCall(PetscViewerASCIIPrintf(viewer, "Number of elements in ordering %" PetscInt_FMT "\n", ao->N));
40: PetscCall(PetscViewerASCIIPrintf(viewer, "PETSc->App App->PETSc\n"));
42: PetscCall(PetscMalloc2(map->N, &app, map->N, &petsc));
43: len = map->n;
44: /* print local AO */
45: PetscCall(PetscViewerASCIIPrintf(viewer, "Process [%d]\n", rank));
46: for (i = 0; i < len; i++) PetscCall(PetscViewerASCIIPrintf(viewer, "%3" PetscInt_FMT " %3" PetscInt_FMT " %3" PetscInt_FMT " %3" PetscInt_FMT "\n", i, aomems->app_loc[i], i, aomems->petsc_loc[i]));
48: /* recv and print off-processor's AO */
49: for (i = 1; i < size; i++) {
50: len = map->range[i + 1] - map->range[i];
51: app_loc = app + map->range[i];
52: petsc_loc = petsc + map->range[i];
53: PetscCallMPI(MPI_Recv(app_loc, (PetscMPIInt)len, MPIU_INT, i, tag_app, PetscObjectComm((PetscObject)ao), &status));
54: PetscCallMPI(MPI_Recv(petsc_loc, (PetscMPIInt)len, MPIU_INT, i, tag_petsc, PetscObjectComm((PetscObject)ao), &status));
55: PetscCall(PetscViewerASCIIPrintf(viewer, "Process [%" PetscInt_FMT "]\n", i));
56: for (j = 0; j < len; j++) PetscCall(PetscViewerASCIIPrintf(viewer, "%3" PetscInt_FMT " %3" PetscInt_FMT " %3" PetscInt_FMT " %3" PetscInt_FMT "\n", map->range[i] + j, app_loc[j], map->range[i] + j, petsc_loc[j]));
57: }
58: PetscCall(PetscFree2(app, petsc));
60: } else {
61: /* send values */
62: PetscCallMPI(MPI_Send((void *)aomems->app_loc, map->n, MPIU_INT, 0, tag_app, PetscObjectComm((PetscObject)ao)));
63: PetscCallMPI(MPI_Send((void *)aomems->petsc_loc, map->n, MPIU_INT, 0, tag_petsc, PetscObjectComm((PetscObject)ao)));
64: }
65: PetscCall(PetscViewerFlush(viewer));
66: PetscFunctionReturn(PETSC_SUCCESS);
67: }
69: static PetscErrorCode AODestroy_MemoryScalable(AO ao)
70: {
71: AO_MemoryScalable *aomems = (AO_MemoryScalable *)ao->data;
73: PetscFunctionBegin;
74: PetscCall(PetscFree2(aomems->app_loc, aomems->petsc_loc));
75: PetscCall(PetscLayoutDestroy(&aomems->map));
76: PetscCall(PetscFree(aomems));
77: PetscFunctionReturn(PETSC_SUCCESS);
78: }
80: /*
81: Input Parameters:
82: + ao - the application ordering context
83: . n - the number of integers in ia[]
84: . ia - the integers; these are replaced with their mapped value
85: - maploc - app_loc or petsc_loc in struct "AO_MemoryScalable"
87: Output Parameter:
88: . ia - the mapped interges
89: */
90: static PetscErrorCode AOMap_MemoryScalable_private(AO ao, PetscInt n, PetscInt *ia, const PetscInt *maploc)
91: {
92: AO_MemoryScalable *aomems = (AO_MemoryScalable *)ao->data;
93: MPI_Comm comm;
94: PetscMPIInt rank, size, tag1, tag2;
95: PetscInt *owner, *start, *sizes, nsends, nreceives;
96: PetscInt nmax, count, *sindices, *rindices, i, j, idx, lastidx, *sindices2, *rindices2;
97: const PetscInt *owners = aomems->map->range;
98: MPI_Request *send_waits, *recv_waits, *send_waits2, *recv_waits2;
99: MPI_Status recv_status;
100: PetscMPIInt nindices, source, widx;
101: PetscInt *rbuf, *sbuf;
102: MPI_Status *send_status, *send_status2;
104: PetscFunctionBegin;
105: PetscCall(PetscObjectGetComm((PetscObject)ao, &comm));
106: PetscCallMPI(MPI_Comm_rank(comm, &rank));
107: PetscCallMPI(MPI_Comm_size(comm, &size));
109: /* first count number of contributors to each processor */
110: PetscCall(PetscMalloc1(size, &start));
111: PetscCall(PetscCalloc2(2 * size, &sizes, n, &owner));
113: j = 0;
114: lastidx = -1;
115: for (i = 0; i < n; i++) {
116: if (ia[i] < 0) owner[i] = -1; /* mark negative entries (which are not to be mapped) with a special negative value */
117: if (ia[i] >= ao->N) owner[i] = -2; /* mark out of range entries with special negative value */
118: else {
119: /* if indices are NOT locally sorted, need to start search at the beginning */
120: if (lastidx > (idx = ia[i])) j = 0;
121: lastidx = idx;
122: for (; j < size; j++) {
123: if (idx >= owners[j] && idx < owners[j + 1]) {
124: sizes[2 * j]++; /* num of indices to be sent */
125: sizes[2 * j + 1] = 1; /* send to proc[j] */
126: owner[i] = j;
127: break;
128: }
129: }
130: }
131: }
132: sizes[2 * rank] = sizes[2 * rank + 1] = 0; /* do not receive from self! */
133: nsends = 0;
134: for (i = 0; i < size; i++) nsends += sizes[2 * i + 1];
136: /* inform other processors of number of messages and max length*/
137: PetscCall(PetscMaxSum(comm, sizes, &nmax, &nreceives));
139: /* allocate arrays */
140: PetscCall(PetscObjectGetNewTag((PetscObject)ao, &tag1));
141: PetscCall(PetscObjectGetNewTag((PetscObject)ao, &tag2));
143: PetscCall(PetscMalloc2(nreceives * nmax, &rindices, nreceives, &recv_waits));
144: PetscCall(PetscMalloc2(nsends * nmax, &rindices2, nsends, &recv_waits2));
146: PetscCall(PetscMalloc3(n, &sindices, nsends, &send_waits, nsends, &send_status));
147: PetscCall(PetscMalloc3(n, &sindices2, nreceives, &send_waits2, nreceives, &send_status2));
149: /* post 1st receives: receive others requests
150: since we don't know how long each individual message is we
151: allocate the largest needed buffer for each receive. Potentially
152: this is a lot of wasted space.
153: */
154: for (i = 0, count = 0; i < nreceives; i++) PetscCallMPI(MPI_Irecv(rindices + nmax * i, nmax, MPIU_INT, MPI_ANY_SOURCE, tag1, comm, recv_waits + count++));
156: /* do 1st sends:
157: 1) starts[i] gives the starting index in svalues for stuff going to
158: the ith processor
159: */
160: start[0] = 0;
161: for (i = 1; i < size; i++) start[i] = start[i - 1] + sizes[2 * i - 2];
162: for (i = 0; i < n; i++) {
163: j = owner[i];
164: if (j == -1) continue; /* do not remap negative entries in ia[] */
165: else if (j == -2) { /* out of range entries get mapped to -1 */ ia[i] = -1;
166: continue;
167: } else if (j != rank) {
168: sindices[start[j]++] = ia[i];
169: } else { /* compute my own map */
170: ia[i] = maploc[ia[i] - owners[rank]];
171: }
172: }
174: start[0] = 0;
175: for (i = 1; i < size; i++) start[i] = start[i - 1] + sizes[2 * i - 2];
176: for (i = 0, count = 0; i < size; i++) {
177: if (sizes[2 * i + 1]) {
178: /* send my request to others */
179: PetscCallMPI(MPI_Isend(sindices + start[i], sizes[2 * i], MPIU_INT, i, tag1, comm, send_waits + count));
180: /* post receive for the answer of my request */
181: PetscCallMPI(MPI_Irecv(sindices2 + start[i], sizes[2 * i], MPIU_INT, i, tag2, comm, recv_waits2 + count));
182: count++;
183: }
184: }
185: PetscCheck(nsends == count, comm, PETSC_ERR_SUP, "nsends %" PetscInt_FMT " != count %" PetscInt_FMT, nsends, count);
187: /* wait on 1st sends */
188: if (nsends) PetscCallMPI(MPI_Waitall(nsends, send_waits, send_status));
190: /* 1st recvs: other's requests */
191: for (j = 0; j < nreceives; j++) {
192: PetscCallMPI(MPI_Waitany(nreceives, recv_waits, &widx, &recv_status)); /* idx: index of handle for operation that completed */
193: PetscCallMPI(MPI_Get_count(&recv_status, MPIU_INT, &nindices));
194: rbuf = rindices + nmax * widx; /* global index */
195: source = recv_status.MPI_SOURCE;
197: /* compute mapping */
198: sbuf = rbuf;
199: for (i = 0; i < nindices; i++) sbuf[i] = maploc[rbuf[i] - owners[rank]];
201: /* send mapping back to the sender */
202: PetscCallMPI(MPI_Isend(sbuf, nindices, MPIU_INT, source, tag2, comm, send_waits2 + widx));
203: }
205: /* wait on 2nd sends */
206: if (nreceives) PetscCallMPI(MPI_Waitall(nreceives, send_waits2, send_status2));
208: /* 2nd recvs: for the answer of my request */
209: for (j = 0; j < nsends; j++) {
210: PetscCallMPI(MPI_Waitany(nsends, recv_waits2, &widx, &recv_status));
211: PetscCallMPI(MPI_Get_count(&recv_status, MPIU_INT, &nindices));
212: source = recv_status.MPI_SOURCE;
213: /* pack output ia[] */
214: rbuf = sindices2 + start[source];
215: count = 0;
216: for (i = 0; i < n; i++) {
217: if (source == owner[i]) ia[i] = rbuf[count++];
218: }
219: }
221: /* free arrays */
222: PetscCall(PetscFree(start));
223: PetscCall(PetscFree2(sizes, owner));
224: PetscCall(PetscFree2(rindices, recv_waits));
225: PetscCall(PetscFree2(rindices2, recv_waits2));
226: PetscCall(PetscFree3(sindices, send_waits, send_status));
227: PetscCall(PetscFree3(sindices2, send_waits2, send_status2));
228: PetscFunctionReturn(PETSC_SUCCESS);
229: }
231: static PetscErrorCode AOPetscToApplication_MemoryScalable(AO ao, PetscInt n, PetscInt *ia)
232: {
233: AO_MemoryScalable *aomems = (AO_MemoryScalable *)ao->data;
234: PetscInt *app_loc = aomems->app_loc;
236: PetscFunctionBegin;
237: PetscCall(AOMap_MemoryScalable_private(ao, n, ia, app_loc));
238: PetscFunctionReturn(PETSC_SUCCESS);
239: }
241: static PetscErrorCode AOApplicationToPetsc_MemoryScalable(AO ao, PetscInt n, PetscInt *ia)
242: {
243: AO_MemoryScalable *aomems = (AO_MemoryScalable *)ao->data;
244: PetscInt *petsc_loc = aomems->petsc_loc;
246: PetscFunctionBegin;
247: PetscCall(AOMap_MemoryScalable_private(ao, n, ia, petsc_loc));
248: PetscFunctionReturn(PETSC_SUCCESS);
249: }
251: static const struct _AOOps AOOps_MemoryScalable = {
252: PetscDesignatedInitializer(view, AOView_MemoryScalable),
253: PetscDesignatedInitializer(destroy, AODestroy_MemoryScalable),
254: PetscDesignatedInitializer(petsctoapplication, AOPetscToApplication_MemoryScalable),
255: PetscDesignatedInitializer(applicationtopetsc, AOApplicationToPetsc_MemoryScalable),
256: };
258: static PetscErrorCode AOCreateMemoryScalable_private(MPI_Comm comm, PetscInt napp, const PetscInt from_array[], const PetscInt to_array[], AO ao, PetscInt *aomap_loc)
259: {
260: AO_MemoryScalable *aomems = (AO_MemoryScalable *)ao->data;
261: PetscLayout map = aomems->map;
262: PetscInt n_local = map->n, i, j;
263: PetscMPIInt rank, size, tag;
264: PetscInt *owner, *start, *sizes, nsends, nreceives;
265: PetscInt nmax, count, *sindices, *rindices, idx, lastidx;
266: PetscInt *owners = aomems->map->range;
267: MPI_Request *send_waits, *recv_waits;
268: MPI_Status recv_status;
269: PetscMPIInt nindices, widx;
270: PetscInt *rbuf;
271: PetscInt n = napp, ip, ia;
272: MPI_Status *send_status;
274: PetscFunctionBegin;
275: PetscCall(PetscArrayzero(aomap_loc, n_local));
277: PetscCallMPI(MPI_Comm_rank(comm, &rank));
278: PetscCallMPI(MPI_Comm_size(comm, &size));
280: /* first count number of contributors (of from_array[]) to each processor */
281: PetscCall(PetscCalloc1(2 * size, &sizes));
282: PetscCall(PetscMalloc1(n, &owner));
284: j = 0;
285: lastidx = -1;
286: for (i = 0; i < n; i++) {
287: /* if indices are NOT locally sorted, need to start search at the beginning */
288: if (lastidx > (idx = from_array[i])) j = 0;
289: lastidx = idx;
290: for (; j < size; j++) {
291: if (idx >= owners[j] && idx < owners[j + 1]) {
292: sizes[2 * j] += 2; /* num of indices to be sent - in pairs (ip,ia) */
293: sizes[2 * j + 1] = 1; /* send to proc[j] */
294: owner[i] = j;
295: break;
296: }
297: }
298: }
299: sizes[2 * rank] = sizes[2 * rank + 1] = 0; /* do not receive from self! */
300: nsends = 0;
301: for (i = 0; i < size; i++) nsends += sizes[2 * i + 1];
303: /* inform other processors of number of messages and max length*/
304: PetscCall(PetscMaxSum(comm, sizes, &nmax, &nreceives));
306: /* allocate arrays */
307: PetscCall(PetscObjectGetNewTag((PetscObject)ao, &tag));
308: PetscCall(PetscMalloc2(nreceives * nmax, &rindices, nreceives, &recv_waits));
309: PetscCall(PetscMalloc3(2 * n, &sindices, nsends, &send_waits, nsends, &send_status));
310: PetscCall(PetscMalloc1(size, &start));
312: /* post receives: */
313: for (i = 0; i < nreceives; i++) PetscCallMPI(MPI_Irecv(rindices + nmax * i, nmax, MPIU_INT, MPI_ANY_SOURCE, tag, comm, recv_waits + i));
315: /* do sends:
316: 1) starts[i] gives the starting index in svalues for stuff going to
317: the ith processor
318: */
319: start[0] = 0;
320: for (i = 1; i < size; i++) start[i] = start[i - 1] + sizes[2 * i - 2];
321: for (i = 0; i < n; i++) {
322: j = owner[i];
323: if (j != rank) {
324: ip = from_array[i];
325: ia = to_array[i];
326: sindices[start[j]++] = ip;
327: sindices[start[j]++] = ia;
328: } else { /* compute my own map */
329: ip = from_array[i] - owners[rank];
330: ia = to_array[i];
331: aomap_loc[ip] = ia;
332: }
333: }
335: start[0] = 0;
336: for (i = 1; i < size; i++) start[i] = start[i - 1] + sizes[2 * i - 2];
337: for (i = 0, count = 0; i < size; i++) {
338: if (sizes[2 * i + 1]) {
339: PetscCallMPI(MPI_Isend(sindices + start[i], sizes[2 * i], MPIU_INT, i, tag, comm, send_waits + count));
340: count++;
341: }
342: }
343: PetscCheck(nsends == count, comm, PETSC_ERR_SUP, "nsends %" PetscInt_FMT " != count %" PetscInt_FMT, nsends, count);
345: /* wait on sends */
346: if (nsends) PetscCallMPI(MPI_Waitall(nsends, send_waits, send_status));
348: /* recvs */
349: count = 0;
350: for (j = nreceives; j > 0; j--) {
351: PetscCallMPI(MPI_Waitany(nreceives, recv_waits, &widx, &recv_status));
352: PetscCallMPI(MPI_Get_count(&recv_status, MPIU_INT, &nindices));
353: rbuf = rindices + nmax * widx; /* global index */
355: /* compute local mapping */
356: for (i = 0; i < nindices; i += 2) { /* pack aomap_loc */
357: ip = rbuf[i] - owners[rank]; /* local index */
358: ia = rbuf[i + 1];
359: aomap_loc[ip] = ia;
360: }
361: count++;
362: }
364: PetscCall(PetscFree(start));
365: PetscCall(PetscFree3(sindices, send_waits, send_status));
366: PetscCall(PetscFree2(rindices, recv_waits));
367: PetscCall(PetscFree(owner));
368: PetscCall(PetscFree(sizes));
369: PetscFunctionReturn(PETSC_SUCCESS);
370: }
372: PETSC_INTERN PetscErrorCode AOCreate_MemoryScalable(AO ao)
373: {
374: IS isapp = ao->isapp, ispetsc = ao->ispetsc;
375: const PetscInt *mypetsc, *myapp;
376: PetscInt napp, n_local, N, i, start, *petsc, *lens, *disp;
377: MPI_Comm comm;
378: AO_MemoryScalable *aomems;
379: PetscLayout map;
380: PetscMPIInt size, rank;
382: PetscFunctionBegin;
383: PetscCheck(isapp, PetscObjectComm((PetscObject)ao), PETSC_ERR_ARG_WRONGSTATE, "AOSetIS() must be called before AOSetType()");
384: /* create special struct aomems */
385: PetscCall(PetscNew(&aomems));
386: ao->data = (void *)aomems;
387: ao->ops[0] = AOOps_MemoryScalable;
388: PetscCall(PetscObjectChangeTypeName((PetscObject)ao, AOMEMORYSCALABLE));
390: /* transmit all local lengths of isapp to all processors */
391: PetscCall(PetscObjectGetComm((PetscObject)isapp, &comm));
392: PetscCallMPI(MPI_Comm_size(comm, &size));
393: PetscCallMPI(MPI_Comm_rank(comm, &rank));
394: PetscCall(PetscMalloc2(size, &lens, size, &disp));
395: PetscCall(ISGetLocalSize(isapp, &napp));
396: PetscCallMPI(MPI_Allgather(&napp, 1, MPIU_INT, lens, 1, MPIU_INT, comm));
398: N = 0;
399: for (i = 0; i < size; i++) {
400: disp[i] = N;
401: N += lens[i];
402: }
404: /* If ispetsc is 0 then use "natural" numbering */
405: if (napp) {
406: if (!ispetsc) {
407: start = disp[rank];
408: PetscCall(PetscMalloc1(napp + 1, &petsc));
409: for (i = 0; i < napp; i++) petsc[i] = start + i;
410: } else {
411: PetscCall(ISGetIndices(ispetsc, &mypetsc));
412: petsc = (PetscInt *)mypetsc;
413: }
414: } else {
415: petsc = NULL;
416: }
418: /* create a map with global size N - used to determine the local sizes of ao - shall we use local napp instead of N? */
419: PetscCall(PetscLayoutCreate(comm, &map));
420: map->bs = 1;
421: map->N = N;
422: PetscCall(PetscLayoutSetUp(map));
424: ao->N = N;
425: ao->n = map->n;
426: aomems->map = map;
428: /* create distributed indices app_loc: petsc->app and petsc_loc: app->petsc */
429: n_local = map->n;
430: PetscCall(PetscCalloc2(n_local, &aomems->app_loc, n_local, &aomems->petsc_loc));
431: PetscCall(ISGetIndices(isapp, &myapp));
433: PetscCall(AOCreateMemoryScalable_private(comm, napp, petsc, myapp, ao, aomems->app_loc));
434: PetscCall(AOCreateMemoryScalable_private(comm, napp, myapp, petsc, ao, aomems->petsc_loc));
436: PetscCall(ISRestoreIndices(isapp, &myapp));
437: if (napp) {
438: if (ispetsc) {
439: PetscCall(ISRestoreIndices(ispetsc, &mypetsc));
440: } else {
441: PetscCall(PetscFree(petsc));
442: }
443: }
444: PetscCall(PetscFree2(lens, disp));
445: PetscFunctionReturn(PETSC_SUCCESS);
446: }
448: /*@C
449: AOCreateMemoryScalable - Creates a memory scalable application ordering using two integer arrays.
451: Collective
453: Input Parameters:
454: + comm - MPI communicator that is to share the `AO`
455: . napp - size of `myapp` and `mypetsc`
456: . myapp - integer array that defines an ordering
457: - mypetsc - integer array that defines another ordering (may be `NULL` to indicate the natural ordering, that is 0,1,2,3,...)
459: Output Parameter:
460: . aoout - the new application ordering
462: Level: beginner
464: Note:
465: The arrays `myapp` and `mypetsc` must contain the all the integers 0 to `napp`-1 with no duplicates; that is there cannot be any "holes"
466: in the indices. Use `AOCreateMapping()` or `AOCreateMappingIS()` if you wish to have "holes" in the indices.
467: Comparing with `AOCreateBasic()`, this routine trades memory with message communication.
469: .seealso: [](sec_ao), [](sec_scatter), `AO`, `AOCreateMemoryScalableIS()`, `AODestroy()`, `AOPetscToApplication()`, `AOApplicationToPetsc()`
470: @*/
471: PetscErrorCode AOCreateMemoryScalable(MPI_Comm comm, PetscInt napp, const PetscInt myapp[], const PetscInt mypetsc[], AO *aoout)
472: {
473: IS isapp, ispetsc;
474: const PetscInt *app = myapp, *petsc = mypetsc;
476: PetscFunctionBegin;
477: PetscCall(ISCreateGeneral(comm, napp, app, PETSC_USE_POINTER, &isapp));
478: if (mypetsc) {
479: PetscCall(ISCreateGeneral(comm, napp, petsc, PETSC_USE_POINTER, &ispetsc));
480: } else {
481: ispetsc = NULL;
482: }
483: PetscCall(AOCreateMemoryScalableIS(isapp, ispetsc, aoout));
484: PetscCall(ISDestroy(&isapp));
485: if (mypetsc) PetscCall(ISDestroy(&ispetsc));
486: PetscFunctionReturn(PETSC_SUCCESS);
487: }
489: /*@C
490: AOCreateMemoryScalableIS - Creates a memory scalable application ordering using two index sets.
492: Collective
494: Input Parameters:
495: + isapp - index set that defines an ordering
496: - ispetsc - index set that defines another ordering (may be `NULL` to use the natural ordering)
498: Output Parameter:
499: . aoout - the new application ordering
501: Level: beginner
503: Notes:
504: The index sets `isapp` and `ispetsc` must contain the all the integers 0 to napp-1 (where napp is the length of the index sets) with no duplicates;
505: that is there cannot be any "holes".
507: Comparing with `AOCreateBasicIS()`, this routine trades memory with message communication.
509: .seealso: [](sec_ao), [](sec_scatter), `AO`, `AOCreateBasicIS()`, `AOCreateMemoryScalable()`, `AODestroy()`
510: @*/
511: PetscErrorCode AOCreateMemoryScalableIS(IS isapp, IS ispetsc, AO *aoout)
512: {
513: MPI_Comm comm;
514: AO ao;
516: PetscFunctionBegin;
517: PetscCall(PetscObjectGetComm((PetscObject)isapp, &comm));
518: PetscCall(AOCreate(comm, &ao));
519: PetscCall(AOSetIS(ao, isapp, ispetsc));
520: PetscCall(AOSetType(ao, AOMEMORYSCALABLE));
521: PetscCall(AOViewFromOptions(ao, NULL, "-ao_view"));
522: *aoout = ao;
523: PetscFunctionReturn(PETSC_SUCCESS);
524: }