Actual source code: pipelcg.c
1: #include <petsc/private/kspimpl.h>
2: #include <petsc/private/vecimpl.h>
4: #define offset(j) PetscMax(((j) - (2 * l)), 0)
5: #define shift(i, j) ((i)-offset((j)))
6: #define G(i, j) (plcg->G[((j) * (2 * l + 1)) + (shift((i), (j)))])
7: #define G_noshift(i, j) (plcg->G[((j) * (2 * l + 1)) + (i)])
8: #define alpha(i) (plcg->alpha[(i)])
9: #define gamma(i) (plcg->gamma[(i)])
10: #define delta(i) (plcg->delta[(i)])
11: #define sigma(i) (plcg->sigma[(i)])
12: #define req(i) (plcg->req[(i)])
14: typedef struct KSP_CG_PIPE_L_s KSP_CG_PIPE_L;
15: struct KSP_CG_PIPE_L_s {
16: PetscInt l; /* pipeline depth */
17: Vec *Z; /* Z vectors (shifted base) */
18: Vec *U; /* U vectors (unpreconditioned shifted base) */
19: Vec *V; /* V vectors (original base) */
20: Vec *Q; /* Q vectors (auxiliary bases) */
21: Vec p; /* work vector */
22: PetscScalar *G; /* such that Z = VG (band matrix)*/
23: PetscScalar *gamma, *delta, *alpha;
24: PetscReal lmin, lmax; /* min and max eigen values estimates to compute base shifts */
25: PetscReal *sigma; /* base shifts */
26: MPI_Request *req; /* request array for asynchronous global collective */
27: PetscBool show_rstrt; /* flag to show restart information in output (default: not shown) */
28: };
30: /*
31: KSPSetUp_PIPELCG - Sets up the workspace needed by the PIPELCG method.
33: This is called once, usually automatically by KSPSolve() or KSPSetUp()
34: but can be called directly by KSPSetUp()
35: */
36: static PetscErrorCode KSPSetUp_PIPELCG(KSP ksp)
37: {
38: KSP_CG_PIPE_L *plcg = (KSP_CG_PIPE_L *)ksp->data;
39: PetscInt l = plcg->l, max_it = ksp->max_it;
40: MPI_Comm comm;
42: PetscFunctionBegin;
43: comm = PetscObjectComm((PetscObject)ksp);
44: PetscCheck(max_it >= 1, comm, PETSC_ERR_ARG_OUTOFRANGE, "%s: max_it argument must be positive.", ((PetscObject)ksp)->type_name);
45: PetscCheck(l >= 1, comm, PETSC_ERR_ARG_OUTOFRANGE, "%s: pipel argument must be positive.", ((PetscObject)ksp)->type_name);
46: PetscCheck(l <= max_it, comm, PETSC_ERR_ARG_OUTOFRANGE, "%s: pipel argument must be less than max_it.", ((PetscObject)ksp)->type_name);
48: PetscCall(KSPSetWorkVecs(ksp, 1)); /* get work vectors needed by PIPELCG */
49: plcg->p = ksp->work[0];
51: PetscCall(VecDuplicateVecs(plcg->p, PetscMax(3, l + 1), &plcg->Z));
52: PetscCall(VecDuplicateVecs(plcg->p, 3, &plcg->U));
53: PetscCall(VecDuplicateVecs(plcg->p, 3, &plcg->V));
54: PetscCall(VecDuplicateVecs(plcg->p, 3 * (l - 1) + 1, &plcg->Q));
55: PetscCall(PetscCalloc1(2, &plcg->alpha));
56: PetscCall(PetscCalloc1(l, &plcg->sigma));
58: PetscFunctionReturn(PETSC_SUCCESS);
59: }
61: static PetscErrorCode KSPReset_PIPELCG(KSP ksp)
62: {
63: KSP_CG_PIPE_L *plcg = (KSP_CG_PIPE_L *)ksp->data;
64: PetscInt l = plcg->l;
66: PetscFunctionBegin;
67: PetscCall(PetscFree(plcg->sigma));
68: PetscCall(PetscFree(plcg->alpha));
69: PetscCall(VecDestroyVecs(PetscMax(3, l + 1), &plcg->Z));
70: PetscCall(VecDestroyVecs(3, &plcg->U));
71: PetscCall(VecDestroyVecs(3, &plcg->V));
72: PetscCall(VecDestroyVecs(3 * (l - 1) + 1, &plcg->Q));
73: PetscFunctionReturn(PETSC_SUCCESS);
74: }
76: static PetscErrorCode KSPDestroy_PIPELCG(KSP ksp)
77: {
78: PetscFunctionBegin;
79: PetscCall(KSPReset_PIPELCG(ksp));
80: PetscCall(KSPDestroyDefault(ksp));
81: PetscFunctionReturn(PETSC_SUCCESS);
82: }
84: static PetscErrorCode KSPSetFromOptions_PIPELCG(KSP ksp, PetscOptionItems *PetscOptionsObject)
85: {
86: KSP_CG_PIPE_L *plcg = (KSP_CG_PIPE_L *)ksp->data;
87: PetscBool flag = PETSC_FALSE;
89: PetscFunctionBegin;
90: PetscOptionsHeadBegin(PetscOptionsObject, "KSP PIPELCG options");
91: PetscCall(PetscOptionsInt("-ksp_pipelcg_pipel", "Pipeline length", "", plcg->l, &plcg->l, &flag));
92: if (!flag) plcg->l = 1;
93: PetscCall(PetscOptionsReal("-ksp_pipelcg_lmin", "Estimate for smallest eigenvalue", "", plcg->lmin, &plcg->lmin, &flag));
94: if (!flag) plcg->lmin = 0.0;
95: PetscCall(PetscOptionsReal("-ksp_pipelcg_lmax", "Estimate for largest eigenvalue", "", plcg->lmax, &plcg->lmax, &flag));
96: if (!flag) plcg->lmax = 0.0;
97: PetscCall(PetscOptionsBool("-ksp_pipelcg_monitor", "Output information on restarts when they occur? (default: 0)", "", plcg->show_rstrt, &plcg->show_rstrt, &flag));
98: if (!flag) plcg->show_rstrt = PETSC_FALSE;
99: PetscOptionsHeadEnd();
100: PetscFunctionReturn(PETSC_SUCCESS);
101: }
103: static PetscErrorCode MPIPetsc_Iallreduce(void *sendbuf, void *recvbuf, PetscMPIInt count, MPI_Datatype datatype, MPI_Op op, MPI_Comm comm, MPI_Request *request)
104: {
105: PetscFunctionBegin;
106: #if defined(PETSC_HAVE_MPI_NONBLOCKING_COLLECTIVES)
107: PetscCallMPI(MPI_Iallreduce(sendbuf, recvbuf, count, datatype, op, comm, request));
108: #else
109: PetscCall(MPIU_Allreduce(sendbuf, recvbuf, count, datatype, op, comm));
110: *request = MPI_REQUEST_NULL;
111: #endif
112: PetscFunctionReturn(PETSC_SUCCESS);
113: }
115: static PetscErrorCode KSPView_PIPELCG(KSP ksp, PetscViewer viewer)
116: {
117: KSP_CG_PIPE_L *plcg = (KSP_CG_PIPE_L *)ksp->data;
118: PetscBool iascii = PETSC_FALSE, isstring = PETSC_FALSE;
120: PetscFunctionBegin;
121: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &iascii));
122: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSTRING, &isstring));
123: if (iascii) {
124: PetscCall(PetscViewerASCIIPrintf(viewer, " Pipeline depth: %" PetscInt_FMT "\n", plcg->l));
125: PetscCall(PetscViewerASCIIPrintf(viewer, " Minimal eigenvalue estimate %g\n", (double)plcg->lmin));
126: PetscCall(PetscViewerASCIIPrintf(viewer, " Maximal eigenvalue estimate %g\n", (double)plcg->lmax));
127: } else if (isstring) {
128: PetscCall(PetscViewerStringSPrintf(viewer, " Pipeline depth: %" PetscInt_FMT "\n", plcg->l));
129: PetscCall(PetscViewerStringSPrintf(viewer, " Minimal eigenvalue estimate %g\n", (double)plcg->lmin));
130: PetscCall(PetscViewerStringSPrintf(viewer, " Maximal eigenvalue estimate %g\n", (double)plcg->lmax));
131: }
132: PetscFunctionReturn(PETSC_SUCCESS);
133: }
135: static PetscErrorCode KSPSolve_InnerLoop_PIPELCG(KSP ksp)
136: {
137: KSP_CG_PIPE_L *plcg = (KSP_CG_PIPE_L *)ksp->data;
138: Mat A = NULL, Pmat = NULL;
139: PetscInt it = 0, max_it = ksp->max_it, l = plcg->l, i = 0, j = 0, k = 0;
140: PetscInt start = 0, middle = 0, end = 0;
141: Vec *Z = plcg->Z, *U = plcg->U, *V = plcg->V, *Q = plcg->Q;
142: Vec x = NULL, p = NULL, temp = NULL;
143: PetscScalar sum_dummy = 0.0, eta = 0.0, zeta = 0.0, lambda = 0.0;
144: PetscReal dp = 0.0, tmp = 0.0, beta = 0.0, invbeta2 = 0.0;
145: MPI_Comm comm;
147: PetscFunctionBegin;
148: x = ksp->vec_sol;
149: p = plcg->p;
151: comm = PetscObjectComm((PetscObject)ksp);
152: PetscCall(PCGetOperators(ksp->pc, &A, &Pmat));
154: for (it = 0; it < max_it + l; ++it) {
155: /* Multiplication z_{it+1} = Az_{it} */
156: /* Shift the U vector pointers */
157: temp = U[2];
158: for (i = 2; i > 0; i--) U[i] = U[i - 1];
159: U[0] = temp;
160: if (it < l) {
161: /* SpMV and Sigma-shift and Prec */
162: PetscCall(MatMult(A, Z[l - it], U[0]));
163: PetscCall(VecAXPY(U[0], -sigma(it), U[1]));
164: PetscCall(KSP_PCApply(ksp, U[0], Z[l - it - 1]));
165: if (it < l - 1) PetscCall(VecCopy(Z[l - it - 1], Q[3 * it]));
166: } else {
167: /* Shift the Z vector pointers */
168: temp = Z[PetscMax(l, 2)];
169: for (i = PetscMax(l, 2); i > 0; --i) Z[i] = Z[i - 1];
170: Z[0] = temp;
171: /* SpMV and Prec */
172: PetscCall(MatMult(A, Z[1], U[0]));
173: PetscCall(KSP_PCApply(ksp, U[0], Z[0]));
174: }
176: /* Adjust the G matrix */
177: if (it >= l) {
178: if (it == l) {
179: /* MPI_Wait for G(0,0),scale V0 and Z and U and Q vectors with 1/beta */
180: PetscCallMPI(MPI_Wait(&req(0), MPI_STATUS_IGNORE));
181: beta = PetscSqrtReal(PetscRealPart(G(0, 0)));
182: G(0, 0) = 1.0;
183: PetscCall(VecAXPY(V[0], 1.0 / beta, p)); /* this assumes V[0] to be zero initially */
184: for (j = 0; j <= PetscMax(l, 2); ++j) PetscCall(VecScale(Z[j], 1.0 / beta));
185: for (j = 0; j <= 2; ++j) PetscCall(VecScale(U[j], 1.0 / beta));
186: for (j = 0; j < l - 1; ++j) PetscCall(VecScale(Q[3 * j], 1.0 / beta));
187: }
189: /* MPI_Wait until the dot products,started l iterations ago,are completed */
190: PetscCallMPI(MPI_Wait(&req(it - l + 1), MPI_STATUS_IGNORE));
191: if (it >= 2 * l) {
192: for (j = PetscMax(0, it - 3 * l + 1); j <= it - 2 * l; j++) { G(j, it - l + 1) = G(it - 2 * l + 1, j + l); /* exploit symmetry in G matrix */ }
193: }
195: if (it <= 2 * l - 1) {
196: invbeta2 = 1.0 / (beta * beta);
197: /* Scale columns 1 up to l of G with 1/beta^2 */
198: for (j = PetscMax(it - 3 * l + 1, 0); j <= it - l + 1; ++j) G(j, it - l + 1) *= invbeta2;
199: }
201: for (j = PetscMax(it - 2 * l + 2, 0); j <= it - l; ++j) {
202: sum_dummy = 0.0;
203: for (k = PetscMax(it - 3 * l + 1, 0); k <= j - 1; ++k) sum_dummy = sum_dummy + G(k, j) * G(k, it - l + 1);
204: G(j, it - l + 1) = (G(j, it - l + 1) - sum_dummy) / G(j, j);
205: }
207: sum_dummy = 0.0;
208: for (k = PetscMax(it - 3 * l + 1, 0); k <= it - l; ++k) sum_dummy = sum_dummy + G(k, it - l + 1) * G(k, it - l + 1);
210: tmp = PetscRealPart(G(it - l + 1, it - l + 1) - sum_dummy);
211: /* Breakdown check */
212: if (tmp < 0) {
213: if (plcg->show_rstrt) PetscCall(PetscPrintf(comm, "Sqrt breakdown in iteration %" PetscInt_FMT ": sqrt argument is %e. Iteration was restarted.\n", ksp->its + 1, (double)tmp));
214: /* End hanging dot-products in the pipeline before exiting for-loop */
215: start = it - l + 2;
216: end = PetscMin(it + 1, max_it + 1); /* !warning! 'it' can actually be greater than 'max_it' */
217: for (i = start; i < end; ++i) PetscCallMPI(MPI_Wait(&req(i), MPI_STATUS_IGNORE));
218: break;
219: }
220: G(it - l + 1, it - l + 1) = PetscSqrtReal(tmp);
222: if (it < 2 * l) {
223: if (it == l) {
224: gamma(it - l) = (G(it - l, it - l + 1) + sigma(it - l) * G(it - l, it - l)) / G(it - l, it - l);
225: } else {
226: gamma(it - l) = (G(it - l, it - l + 1) + sigma(it - l) * G(it - l, it - l) - delta(it - l - 1) * G(it - l - 1, it - l)) / G(it - l, it - l);
227: }
228: delta(it - l) = G(it - l + 1, it - l + 1) / G(it - l, it - l);
229: } else {
230: gamma(it - l) = (G(it - l, it - l) * gamma(it - 2 * l) + G(it - l, it - l + 1) * delta(it - 2 * l) - G(it - l - 1, it - l) * delta(it - l - 1)) / G(it - l, it - l);
231: delta(it - l) = (G(it - l + 1, it - l + 1) * delta(it - 2 * l)) / G(it - l, it - l);
232: }
234: /* Recursively compute the next V, Q, Z and U vectors */
235: /* Shift the V vector pointers */
236: temp = V[2];
237: for (i = 2; i > 0; i--) V[i] = V[i - 1];
238: V[0] = temp;
240: /* Recurrence V vectors */
241: if (l == 1) {
242: PetscCall(VecCopy(Z[1], V[0]));
243: } else {
244: PetscCall(VecCopy(Q[0], V[0]));
245: }
246: if (it == l) {
247: PetscCall(VecAXPY(V[0], sigma(0) - gamma(it - l), V[1]));
248: } else {
249: alpha(0) = sigma(0) - gamma(it - l);
250: alpha(1) = -delta(it - l - 1);
251: PetscCall(VecMAXPY(V[0], 2, &alpha(0), &V[1]));
252: }
253: PetscCall(VecScale(V[0], 1.0 / delta(it - l)));
255: /* Recurrence Q vectors */
256: for (j = 0; j < l - 1; ++j) {
257: /* Shift the Q vector pointers */
258: temp = Q[3 * j + 2];
259: for (i = 2; i > 0; i--) Q[3 * j + i] = Q[3 * j + i - 1];
260: Q[3 * j] = temp;
262: if (j < l - 2) {
263: PetscCall(VecCopy(Q[3 * (j + 1)], Q[3 * j]));
264: } else {
265: PetscCall(VecCopy(Z[1], Q[3 * j]));
266: }
267: if (it == l) {
268: PetscCall(VecAXPY(Q[3 * j], sigma(j + 1) - gamma(it - l), Q[3 * j + 1]));
269: } else {
270: alpha(0) = sigma(j + 1) - gamma(it - l);
271: alpha(1) = -delta(it - l - 1);
272: PetscCall(VecMAXPY(Q[3 * j], 2, &alpha(0), &Q[3 * j + 1]));
273: }
274: PetscCall(VecScale(Q[3 * j], 1.0 / delta(it - l)));
275: }
277: /* Recurrence Z and U vectors */
278: if (it == l) {
279: PetscCall(VecAXPY(Z[0], -gamma(it - l), Z[1]));
280: PetscCall(VecAXPY(U[0], -gamma(it - l), U[1]));
281: } else {
282: alpha(0) = -gamma(it - l);
283: alpha(1) = -delta(it - l - 1);
284: PetscCall(VecMAXPY(Z[0], 2, &alpha(0), &Z[1]));
285: PetscCall(VecMAXPY(U[0], 2, &alpha(0), &U[1]));
286: }
287: PetscCall(VecScale(Z[0], 1.0 / delta(it - l)));
288: PetscCall(VecScale(U[0], 1.0 / delta(it - l)));
289: }
291: /* Compute and communicate the dot products */
292: if (it < l) {
293: for (j = 0; j < it + 2; ++j) { PetscCall((*U[0]->ops->dot_local)(U[0], Z[l - j], &G(j, it + 1))); /* dot-products (U[0],Z[j]) */ }
294: PetscCall(MPIPetsc_Iallreduce(MPI_IN_PLACE, &G(0, it + 1), it + 2, MPIU_SCALAR, MPIU_SUM, comm, &req(it + 1)));
295: } else if ((it >= l) && (it < max_it)) {
296: middle = it - l + 2;
297: end = it + 2;
298: PetscCall((*U[0]->ops->dot_local)(U[0], V[0], &G(it - l + 1, it + 1))); /* dot-product (U[0],V[0]) */
299: for (j = middle; j < end; ++j) { PetscCall((*U[0]->ops->dot_local)(U[0], plcg->Z[it + 1 - j], &G(j, it + 1))); /* dot-products (U[0],Z[j]) */ }
300: PetscCall(MPIPetsc_Iallreduce(MPI_IN_PLACE, &G(it - l + 1, it + 1), l + 1, MPIU_SCALAR, MPIU_SUM, comm, &req(it + 1)));
301: }
303: /* Compute solution vector and residual norm */
304: if (it >= l) {
305: if (it == l) {
306: if (ksp->its != 0) ++ksp->its;
307: eta = gamma(0);
308: zeta = beta;
309: PetscCall(VecCopy(V[1], p));
310: PetscCall(VecScale(p, 1.0 / eta));
311: PetscCall(VecAXPY(x, zeta, p));
312: dp = beta;
313: } else if (it > l) {
314: k = it - l;
315: ++ksp->its;
316: lambda = delta(k - 1) / eta;
317: eta = gamma(k) - lambda * delta(k - 1);
318: zeta = -lambda * zeta;
319: PetscCall(VecScale(p, -delta(k - 1) / eta));
320: PetscCall(VecAXPY(p, 1.0 / eta, V[1]));
321: PetscCall(VecAXPY(x, zeta, p));
322: dp = PetscAbsScalar(zeta);
323: }
324: ksp->rnorm = dp;
325: PetscCall(KSPLogResidualHistory(ksp, dp));
326: PetscCall(KSPMonitor(ksp, ksp->its, dp));
327: PetscCall((*ksp->converged)(ksp, ksp->its, dp, &ksp->reason, ksp->cnvP));
329: if (ksp->its >= max_it && !ksp->reason) ksp->reason = KSP_DIVERGED_ITS;
330: if (ksp->reason) {
331: /* End hanging dot-products in the pipeline before exiting for-loop */
332: start = it - l + 2;
333: end = PetscMin(it + 2, max_it + 1); /* !warning! 'it' can actually be greater than 'max_it' */
334: for (i = start; i < end; ++i) PetscCallMPI(MPI_Wait(&req(i), MPI_STATUS_IGNORE));
335: break;
336: }
337: }
338: } /* End inner for loop */
339: PetscFunctionReturn(PETSC_SUCCESS);
340: }
342: static PetscErrorCode KSPSolve_ReInitData_PIPELCG(KSP ksp)
343: {
344: KSP_CG_PIPE_L *plcg = (KSP_CG_PIPE_L *)ksp->data;
345: PetscInt i = 0, j = 0, l = plcg->l, max_it = ksp->max_it;
347: PetscFunctionBegin;
348: for (i = 0; i < PetscMax(3, l + 1); ++i) PetscCall(VecSet(plcg->Z[i], 0.0));
349: for (i = 1; i < 3; ++i) PetscCall(VecSet(plcg->U[i], 0.0));
350: for (i = 0; i < 3; ++i) PetscCall(VecSet(plcg->V[i], 0.0));
351: for (i = 0; i < 3 * (l - 1) + 1; ++i) PetscCall(VecSet(plcg->Q[i], 0.0));
352: for (j = 0; j < (max_it + 1); ++j) {
353: gamma(j) = 0.0;
354: delta(j) = 0.0;
355: for (i = 0; i < (2 * l + 1); ++i) G_noshift(i, j) = 0.0;
356: }
357: PetscFunctionReturn(PETSC_SUCCESS);
358: }
360: /*
361: KSPSolve_PIPELCG - This routine actually applies the pipelined(l) conjugate gradient method
362: */
363: static PetscErrorCode KSPSolve_PIPELCG(KSP ksp)
364: {
365: KSP_CG_PIPE_L *plcg = (KSP_CG_PIPE_L *)ksp->data;
366: Mat A = NULL, Pmat = NULL;
367: Vec b = NULL, x = NULL, p = NULL;
368: PetscInt max_it = ksp->max_it, l = plcg->l;
369: PetscInt i = 0, outer_it = 0, curr_guess_zero = 0;
370: PetscReal lmin = plcg->lmin, lmax = plcg->lmax;
371: PetscBool diagonalscale = PETSC_FALSE;
372: MPI_Comm comm;
374: PetscFunctionBegin;
375: comm = PetscObjectComm((PetscObject)ksp);
376: PetscCall(PCGetDiagonalScale(ksp->pc, &diagonalscale));
377: PetscCheck(!diagonalscale, comm, PETSC_ERR_SUP, "Krylov method %s does not support diagonal scaling", ((PetscObject)ksp)->type_name);
379: x = ksp->vec_sol;
380: b = ksp->vec_rhs;
381: p = plcg->p;
383: PetscCall(PetscCalloc1((max_it + 1) * (2 * l + 1), &plcg->G));
384: PetscCall(PetscCalloc1(max_it + 1, &plcg->gamma));
385: PetscCall(PetscCalloc1(max_it + 1, &plcg->delta));
386: PetscCall(PetscCalloc1(max_it + 1, &plcg->req));
388: PetscCall(PCGetOperators(ksp->pc, &A, &Pmat));
390: for (i = 0; i < l; ++i) sigma(i) = (0.5 * (lmin + lmax) + (0.5 * (lmax - lmin) * PetscCosReal(PETSC_PI * (2.0 * i + 1.0) / (2.0 * l))));
392: ksp->its = 0;
393: outer_it = 0;
394: curr_guess_zero = !!ksp->guess_zero;
396: while (ksp->its < max_it) { /* OUTER LOOP (gmres-like restart to handle breakdowns) */
397: /* RESTART LOOP */
398: if (!curr_guess_zero) {
399: PetscCall(KSP_MatMult(ksp, A, x, plcg->U[0])); /* u <- b - Ax */
400: PetscCall(VecAYPX(plcg->U[0], -1.0, b));
401: } else {
402: PetscCall(VecCopy(b, plcg->U[0])); /* u <- b (x is 0) */
403: }
404: PetscCall(KSP_PCApply(ksp, plcg->U[0], p)); /* p <- Bu */
406: if (outer_it > 0) {
407: /* Re-initialize Z,U,V,Q,gamma,delta,G after restart occurred */
408: PetscCall(KSPSolve_ReInitData_PIPELCG(ksp));
409: }
411: PetscCall((*plcg->U[0]->ops->dot_local)(plcg->U[0], p, &G(0, 0)));
412: PetscCall(MPIPetsc_Iallreduce(MPI_IN_PLACE, &G(0, 0), 1, MPIU_SCALAR, MPIU_SUM, comm, &req(0)));
413: PetscCall(VecCopy(p, plcg->Z[l]));
415: PetscCall(KSPSolve_InnerLoop_PIPELCG(ksp));
417: if (ksp->reason) break; /* convergence or divergence */
418: ++outer_it;
419: curr_guess_zero = 0;
420: }
422: if (!ksp->reason) ksp->reason = KSP_DIVERGED_ITS;
423: PetscCall(PetscFree(plcg->G));
424: PetscCall(PetscFree(plcg->gamma));
425: PetscCall(PetscFree(plcg->delta));
426: PetscCall(PetscFree(plcg->req));
427: PetscFunctionReturn(PETSC_SUCCESS);
428: }
430: /*MC
431: KSPPIPELCG - Deep pipelined (length l) Conjugate Gradient method {cite}`cornelis2018communication` and {cite}`cools2019numerically`.
432: This method has only a single non-blocking global
433: reduction per iteration, compared to 2 blocking reductions for standard `KSPCG`. The reduction is overlapped by the
434: matrix-vector product and preconditioner application of the next l iterations. The pipeline length l is a parameter
435: of the method. [](sec_pipelineksp)
437: Options Database Keys:
438: + -ksp_pipelcg_pipel - pipelined length
439: . -ksp_pipelcg_lmin - approximation to the smallest eigenvalue of the preconditioned operator (default: 0.0)
440: . -ksp_pipelcg_lmax - approximation to the largest eigenvalue of the preconditioned operator (default: 0.0)
441: - -ksp_pipelcg_monitor - output where/why the method restarts when a sqrt breakdown occurs
443: Example usage:
444: .vb
445: KSP tutorials ex2, no preconditioner, pipel = 2, lmin = 0.0, lmax = 8.0 :
446: $mpiexec -n 14 ./ex2 -m 1000 -n 1000 -ksp_type pipelcg -pc_type none -ksp_norm_type natural
447: -ksp_rtol 1e-10 -ksp_max_it 1000 -ksp_pipelcg_pipel 2 -ksp_pipelcg_lmin 0.0 -ksp_pipelcg_lmax 8.0 -log_view
448: SNES tutorials ex48, bjacobi preconditioner, pipel = 3, lmin = 0.0, lmax = 2.0, show restart information :
449: $mpiexec -n 14 ./ex48 -M 150 -P 100 -ksp_type pipelcg -pc_type bjacobi -ksp_rtol 1e-10 -ksp_pipelcg_pipel 3
450: -ksp_pipelcg_lmin 0.0 -ksp_pipelcg_lmax 2.0 -ksp_pipelcg_monitor -log_view
451: .ve
453: Level: advanced
455: Notes:
456: MPI configuration may be necessary for reductions to make asynchronous progress, which is important for
457: performance of pipelined methods. See [](doc_faq_pipelined)
459: Contributed by:
460: Siegfried Cools, University of Antwerp, Dept. Mathematics and Computer Science,
461: funded by Flemish Research Foundation (FWO) grant number 12H4617N.
463: .seealso: [](ch_ksp), [](sec_pipelineksp), [](doc_faq_pipelined), `KSPCreate()`, `KSPSetType()`, `KSPType`, `KSPCG`, `KSPPIPECG`, `KSPPIPECGRR`, `KSPPGMRES`,
464: `KSPPIPEBCGS`, `KSPSetPCSide()`, `KSPGROPPCG`
465: M*/
466: PETSC_EXTERN PetscErrorCode KSPCreate_PIPELCG(KSP ksp)
467: {
468: KSP_CG_PIPE_L *plcg = NULL;
470: PetscFunctionBegin;
471: PetscCall(PetscNew(&plcg));
472: ksp->data = (void *)plcg;
474: PetscCall(KSPSetSupportedNorm(ksp, KSP_NORM_NONE, PC_LEFT, 1));
475: PetscCall(KSPSetSupportedNorm(ksp, KSP_NORM_NATURAL, PC_LEFT, 2));
477: ksp->ops->setup = KSPSetUp_PIPELCG;
478: ksp->ops->solve = KSPSolve_PIPELCG;
479: ksp->ops->reset = KSPReset_PIPELCG;
480: ksp->ops->destroy = KSPDestroy_PIPELCG;
481: ksp->ops->view = KSPView_PIPELCG;
482: ksp->ops->setfromoptions = KSPSetFromOptions_PIPELCG;
483: ksp->ops->buildsolution = KSPBuildSolutionDefault;
484: ksp->ops->buildresidual = KSPBuildResidualDefault;
485: PetscFunctionReturn(PETSC_SUCCESS);
486: }