Actual source code: rk.c
petsc-3.13.6 2020-09-29
1: /*
2: Code for time stepping with the Runge-Kutta method
4: Notes:
5: The general system is written as
7: Udot = F(t,U)
9: */
11: #include <petsc/private/tsimpl.h>
12: #include <petscdm.h>
13: #include <../src/ts/impls/explicit/rk/rk.h>
14: #include <../src/ts/impls/explicit/rk/mrk.h>
16: static TSRKType TSRKDefault = TSRK3BS;
17: static PetscBool TSRKRegisterAllCalled;
18: static PetscBool TSRKPackageInitialized;
20: static RKTableauLink RKTableauList;
22: /*MC
23: TSRK1FE - First order forward Euler scheme.
25: This method has one stage.
27: Options database:
28: . -ts_rk_type 1fe
30: Level: advanced
32: .seealso: TSRK, TSRKType, TSRKSetType()
33: M*/
34: /*MC
35: TSRK2A - Second order RK scheme.
37: This method has two stages.
39: Options database:
40: . -ts_rk_type 2a
42: Level: advanced
44: .seealso: TSRK, TSRKType, TSRKSetType()
45: M*/
46: /*MC
47: TSRK3 - Third order RK scheme.
49: This method has three stages.
51: Options database:
52: . -ts_rk_type 3
54: Level: advanced
56: .seealso: TSRK, TSRKType, TSRKSetType()
57: M*/
58: /*MC
59: TSRK3BS - Third order RK scheme of Bogacki-Shampine with 2nd order embedded method.
61: This method has four stages with the First Same As Last (FSAL) property.
63: Options database:
64: . -ts_rk_type 3bs
66: Level: advanced
68: References: https://doi.org/10.1016/0893-9659(89)90079-7
70: .seealso: TSRK, TSRKType, TSRKSetType()
71: M*/
72: /*MC
73: TSRK4 - Fourth order RK scheme.
75: This is the classical Runge-Kutta method with four stages.
77: Options database:
78: . -ts_rk_type 4
80: Level: advanced
82: .seealso: TSRK, TSRKType, TSRKSetType()
83: M*/
84: /*MC
85: TSRK5F - Fifth order Fehlberg RK scheme with a 4th order embedded method.
87: This method has six stages.
89: Options database:
90: . -ts_rk_type 5f
92: Level: advanced
94: .seealso: TSRK, TSRKType, TSRKSetType()
95: M*/
96: /*MC
97: TSRK5DP - Fifth order Dormand-Prince RK scheme with the 4th order embedded method.
99: This method has seven stages with the First Same As Last (FSAL) property.
101: Options database:
102: . -ts_rk_type 5dp
104: Level: advanced
106: References: https://doi.org/10.1016/0771-050X(80)90013-3
108: .seealso: TSRK, TSRKType, TSRKSetType()
109: M*/
110: /*MC
111: TSRK5BS - Fifth order Bogacki-Shampine RK scheme with 4th order embedded method.
113: This method has eight stages with the First Same As Last (FSAL) property.
115: Options database:
116: . -ts_rk_type 5bs
118: Level: advanced
120: References: https://doi.org/10.1016/0898-1221(96)00141-1
122: .seealso: TSRK, TSRKType, TSRKSetType()
123: M*/
124: /*MC
125: TSRK6VR - Sixth order robust Verner RK scheme with fifth order embedded method.
127: This method has nine stages with the First Same As Last (FSAL) property.
129: Options database:
130: . -ts_rk_type 6vr
132: Level: advanced
134: References: http://people.math.sfu.ca/~jverner/RKV65.IIIXb.Robust.00010102836.081204.CoeffsOnlyRAT
136: .seealso: TSRK, TSRKType, TSRKSetType()
137: M*/
138: /*MC
139: TSRK7VR - Seventh order robust Verner RK scheme with sixth order embedded method.
141: This method has ten stages with the First Same As Last (FSAL) property.
143: Options database:
144: . -ts_rk_type 7vr
146: Level: advanced
148: References: http://people.math.sfu.ca/~jverner/RKV76.IIa.Robust.000027015646.081206.CoeffsOnlyRAT
150: .seealso: TSRK, TSRKType, TSRKSetType()
151: M*/
152: /*MC
153: TSRK8VR - Eigth order robust Verner RK scheme with seventh order embedded method.
155: This method has thirteen stages with the First Same As Last (FSAL) property.
157: Options database:
158: . -ts_rk_type 8vr
160: Level: advanced
162: References: http://people.math.sfu.ca/~jverner/RKV87.IIa.Robust.00000754677.081208.CoeffsOnlyRATandFLOAT
164: .seealso: TSRK, TSRKType, TSRKSetType()
165: M*/
167: /*@C
168: TSRKRegisterAll - Registers all of the Runge-Kutta explicit methods in TSRK
170: Not Collective, but should be called by all processes which will need the schemes to be registered
172: Level: advanced
174: .seealso: TSRKRegisterDestroy()
175: @*/
176: PetscErrorCode TSRKRegisterAll(void)
177: {
181: if (TSRKRegisterAllCalled) return(0);
182: TSRKRegisterAllCalled = PETSC_TRUE;
184: #define RC PetscRealConstant
185: {
186: const PetscReal
187: A[1][1] = {{0}},
188: b[1] = {RC(1.0)};
189: TSRKRegister(TSRK1FE,1,1,&A[0][0],b,NULL,NULL,0,NULL);
190: }
191: {
192: const PetscReal
193: A[2][2] = {{0,0},
194: {RC(1.0),0}},
195: b[2] = {RC(0.5),RC(0.5)},
196: bembed[2] = {RC(1.0),0};
197: TSRKRegister(TSRK2A,2,2,&A[0][0],b,NULL,bembed,0,NULL);
198: }
199: {
200: const PetscReal
201: A[3][3] = {{0,0,0},
202: {RC(2.0)/RC(3.0),0,0},
203: {RC(-1.0)/RC(3.0),RC(1.0),0}},
204: b[3] = {RC(0.25),RC(0.5),RC(0.25)};
205: TSRKRegister(TSRK3,3,3,&A[0][0],b,NULL,NULL,0,NULL);
206: }
207: {
208: const PetscReal
209: A[4][4] = {{0,0,0,0},
210: {RC(1.0)/RC(2.0),0,0,0},
211: {0,RC(3.0)/RC(4.0),0,0},
212: {RC(2.0)/RC(9.0),RC(1.0)/RC(3.0),RC(4.0)/RC(9.0),0}},
213: b[4] = {RC(2.0)/RC(9.0),RC(1.0)/RC(3.0),RC(4.0)/RC(9.0),0},
214: bembed[4] = {RC(7.0)/RC(24.0),RC(1.0)/RC(4.0),RC(1.0)/RC(3.0),RC(1.0)/RC(8.0)};
215: TSRKRegister(TSRK3BS,3,4,&A[0][0],b,NULL,bembed,0,NULL);
216: }
217: {
218: const PetscReal
219: A[4][4] = {{0,0,0,0},
220: {RC(0.5),0,0,0},
221: {0,RC(0.5),0,0},
222: {0,0,RC(1.0),0}},
223: b[4] = {RC(1.0)/RC(6.0),RC(1.0)/RC(3.0),RC(1.0)/RC(3.0),RC(1.0)/RC(6.0)};
224: TSRKRegister(TSRK4,4,4,&A[0][0],b,NULL,NULL,0,NULL);
225: }
226: {
227: const PetscReal
228: A[6][6] = {{0,0,0,0,0,0},
229: {RC(0.25),0,0,0,0,0},
230: {RC(3.0)/RC(32.0),RC(9.0)/RC(32.0),0,0,0,0},
231: {RC(1932.0)/RC(2197.0),RC(-7200.0)/RC(2197.0),RC(7296.0)/RC(2197.0),0,0,0},
232: {RC(439.0)/RC(216.0),RC(-8.0),RC(3680.0)/RC(513.0),RC(-845.0)/RC(4104.0),0,0},
233: {RC(-8.0)/RC(27.0),RC(2.0),RC(-3544.0)/RC(2565.0),RC(1859.0)/RC(4104.0),RC(-11.0)/RC(40.0),0}},
234: b[6] = {RC(16.0)/RC(135.0),0,RC(6656.0)/RC(12825.0),RC(28561.0)/RC(56430.0),RC(-9.0)/RC(50.0),RC(2.0)/RC(55.0)},
235: bembed[6] = {RC(25.0)/RC(216.0),0,RC(1408.0)/RC(2565.0),RC(2197.0)/RC(4104.0),RC(-1.0)/RC(5.0),0};
236: TSRKRegister(TSRK5F,5,6,&A[0][0],b,NULL,bembed,0,NULL);
237: }
238: {
239: const PetscReal
240: A[7][7] = {{0,0,0,0,0,0,0},
241: {RC(1.0)/RC(5.0),0,0,0,0,0,0},
242: {RC(3.0)/RC(40.0),RC(9.0)/RC(40.0),0,0,0,0,0},
243: {RC(44.0)/RC(45.0),RC(-56.0)/RC(15.0),RC(32.0)/RC(9.0),0,0,0,0},
244: {RC(19372.0)/RC(6561.0),RC(-25360.0)/RC(2187.0),RC(64448.0)/RC(6561.0),RC(-212.0)/RC(729.0),0,0,0},
245: {RC(9017.0)/RC(3168.0),RC(-355.0)/RC(33.0),RC(46732.0)/RC(5247.0),RC(49.0)/RC(176.0),RC(-5103.0)/RC(18656.0),0,0},
246: {RC(35.0)/RC(384.0),0,RC(500.0)/RC(1113.0),RC(125.0)/RC(192.0),RC(-2187.0)/RC(6784.0),RC(11.0)/RC(84.0),0}},
247: b[7] = {RC(35.0)/RC(384.0),0,RC(500.0)/RC(1113.0),RC(125.0)/RC(192.0),RC(-2187.0)/RC(6784.0),RC(11.0)/RC(84.0),0},
248: bembed[7] = {RC(5179.0)/RC(57600.0),0,RC(7571.0)/RC(16695.0),RC(393.0)/RC(640.0),RC(-92097.0)/RC(339200.0),RC(187.0)/RC(2100.0),RC(1.0)/RC(40.0)},
249: binterp[7][5] = {{RC(1.0),RC(-4034104133.0)/RC(1410260304.0),RC(105330401.0)/RC(33982176.0),RC(-13107642775.0)/RC(11282082432.0),RC(6542295.0)/RC(470086768.0)},
250: {0,0,0,0,0},
251: {0,RC(132343189600.0)/RC(32700410799.0),RC(-833316000.0)/RC(131326951.0),RC(91412856700.0)/RC(32700410799.0),RC(-523383600.0)/RC(10900136933.0)},
252: {0,RC(-115792950.0)/RC(29380423.0),RC(185270875.0)/RC(16991088.0),RC(-12653452475.0)/RC(1880347072.0),RC(98134425.0)/RC(235043384.0)},
253: {0,RC(70805911779.0)/RC(24914598704.0),RC(-4531260609.0)/RC(600351776.0),RC(988140236175.0)/RC(199316789632.0),RC(-14307999165.0)/RC(24914598704.0)},
254: {0,RC(-331320693.0)/RC(205662961.0),RC(31361737.0)/RC(7433601.0),RC(-2426908385.0)/RC(822651844.0),RC(97305120.0)/RC(205662961.0)},
255: {0,RC(44764047.0)/RC(29380423.0),RC(-1532549.0)/RC(353981.0),RC(90730570.0)/RC(29380423.0),RC(-8293050.0)/RC(29380423.0)}};
256: TSRKRegister(TSRK5DP,5,7,&A[0][0],b,NULL,bembed,5,binterp[0]);
257: }
258: {
259: const PetscReal
260: A[8][8] = {{0,0,0,0,0,0,0,0},
261: {RC(1.0)/RC(6.0),0,0,0,0,0,0,0},
262: {RC(2.0)/RC(27.0),RC(4.0)/RC(27.0),0,0,0,0,0,0},
263: {RC(183.0)/RC(1372.0),RC(-162.0)/RC(343.0),RC(1053.0)/RC(1372.0),0,0,0,0,0},
264: {RC(68.0)/RC(297.0),RC(-4.0)/RC(11.0),RC(42.0)/RC(143.0),RC(1960.0)/RC(3861.0),0,0,0,0},
265: {RC(597.0)/RC(22528.0),RC(81.0)/RC(352.0),RC(63099.0)/RC(585728.0),RC(58653.0)/RC(366080.0),RC(4617.0)/RC(20480.0),0,0,0},
266: {RC(174197.0)/RC(959244.0),RC(-30942.0)/RC(79937.0),RC(8152137.0)/RC(19744439.0),RC(666106.0)/RC(1039181.0),RC(-29421.0)/RC(29068.0),RC(482048.0)/RC(414219.0),0,0},
267: {RC(587.0)/RC(8064.0),0,RC(4440339.0)/RC(15491840.0),RC(24353.0)/RC(124800.0),RC(387.0)/RC(44800.0),RC(2152.0)/RC(5985.0),RC(7267.0)/RC(94080.0),0}},
268: b[8] = {RC(587.0)/RC(8064.0),0,RC(4440339.0)/RC(15491840.0),RC(24353.0)/RC(124800.0),RC(387.0)/RC(44800.0),RC(2152.0)/RC(5985.0),RC(7267.0)/RC(94080.0),0},
269: bembed[8] = {RC(2479.0)/RC(34992.0),0,RC(123.0)/RC(416.0),RC(612941.0)/RC(3411720.0),RC(43.0)/RC(1440.0),RC(2272.0)/RC(6561.0),RC(79937.0)/RC(1113912.0),RC(3293.0)/RC(556956.0)};
270: TSRKRegister(TSRK5BS,5,8,&A[0][0],b,NULL,bembed,0,NULL);
271: }
272: {
273: const PetscReal
274: A[9][9] = {{0,0,0,0,0,0,0,0,0},
275: {RC(1.8000000000000000000000000000000000000000e-01),0,0,0,0,0,0,0,0},
276: {RC(8.9506172839506172839506172839506172839506e-02),RC(7.7160493827160493827160493827160493827160e-02),0,0,0,0,0,0,0},
277: {RC(6.2500000000000000000000000000000000000000e-02),0,RC(1.8750000000000000000000000000000000000000e-01),0,0,0,0,0,0},
278: {RC(3.1651600000000000000000000000000000000000e-01),0,RC(-1.0449480000000000000000000000000000000000e+00),RC(1.2584320000000000000000000000000000000000e+00),0,0,0,0,0},
279: {RC(2.7232612736485626257225065566674305502508e-01),0,RC(-8.2513360323886639676113360323886639676113e-01),RC(1.0480917678812415654520917678812415654521e+00),RC(1.0471570799276856873679117969088177628396e-01),0,0,0,0},
280: {RC(-1.6699418599716514314329607278961797333198e-01),0,RC(6.3170850202429149797570850202429149797571e-01),RC(1.7461044552773876082146758838488161796432e-01),RC(-1.0665356459086066122525194734018680677781e+00),RC(1.2272108843537414965986394557823129251701e+00),0,0,0},
281: {RC(3.6423751686909581646423751686909581646424e-01),0,RC(-2.0404858299595141700404858299595141700405e-01),RC(-3.4883737816068643136312309244640071707741e-01),RC(3.2619323032856867443333608747142581729048e+00),RC(-2.7551020408163265306122448979591836734694e+00),RC(6.8181818181818181818181818181818181818182e-01),0,0},
282: {RC(7.6388888888888888888888888888888888888889e-02),0,0,RC(3.6940836940836940836940836940836940836941e-01),0,RC(2.4801587301587301587301587301587301587302e-01),RC(2.3674242424242424242424242424242424242424e-01),RC(6.9444444444444444444444444444444444444444e-02),0}},
283: b[9] = {RC(7.6388888888888888888888888888888888888889e-02),0,0,RC(3.6940836940836940836940836940836940836941e-01),0,RC(2.4801587301587301587301587301587301587302e-01),RC(2.3674242424242424242424242424242424242424e-01),RC(6.9444444444444444444444444444444444444444e-02),0},
284: bembed[9] = {RC(5.8700209643605870020964360587002096436059e-02),0,0,RC(4.8072562358276643990929705215419501133787e-01),RC(-8.5341242076919085578832094861228313083563e-01),RC(1.2046485260770975056689342403628117913832e+00),0,RC(-5.9242373072160306202859394348756050883710e-02),RC(1.6858043453788134639198468985703028256220e-01)};
285: TSRKRegister(TSRK6VR,6,9,&A[0][0],b,NULL,bembed,0,NULL);
286: }
287: {
288: const PetscReal
289: A[10][10] = {{0,0,0,0,0,0,0,0,0,0},
290: {RC(5.0000000000000000000000000000000000000000e-03),0,0,0,0,0,0,0,0,0},
291: {RC(-1.0767901234567901234567901234567901234568e+00),RC(1.1856790123456790123456790123456790123457e+00),0,0,0,0,0,0,0,0},
292: {RC(4.0833333333333333333333333333333333333333e-02),0,RC(1.2250000000000000000000000000000000000000e-01),0,0,0,0,0,0,0},
293: {RC(6.3607142857142857142857142857142857142857e-01),0,RC(-2.4444642857142857142857142857142857142857e+00),RC(2.2633928571428571428571428571428571428571e+00),0,0,0,0,0,0},
294: {RC(-2.5351211079349245229256383554660215487207e+00),0,RC(1.0299374654449267920438514460756024913612e+01),RC(-7.9513032885990579949493217458266876536482e+00),RC(7.9301148923100592201226014271115261823800e-01),0,0,0,0,0},
295: {RC(1.0018765812524632961969196583094999808207e+00),0,RC(-4.1665712824423798331313938005470971453189e+00),RC(3.8343432929128642412552665218251378665197e+00),RC(-5.0233333560710847547464330228611765612403e-01),RC(6.6768474388416077115385092269857695410259e-01),0,0,0,0},
296: {RC(2.7255018354630767130333963819175005717348e+01),0,RC(-4.2004617278410638355318645443909295369611e+01),RC(-1.0535713126619489917921081600546526103722e+01),RC(8.0495536711411937147983652158926826634202e+01),RC(-6.7343882271790513468549075963212975640927e+01),RC(1.3048657610777937463471187029566964762710e+01),0,0,0},
297: {RC(-3.0397378057114965146943658658755763226883e+00),0,RC(1.0138161410329801111857946190709700150441e+01),RC(-6.4293056748647215721462825629555298064437e+00),RC(-1.5864371483408276587115312853798610579467e+00),RC(1.8921781841968424410864308909131353365021e+00),RC(1.9699335407608869061292360163336442838006e-02),RC(5.4416989827933235465102724247952572977903e-03),0,0},
298: {RC(-1.4449518916777735137351003179355712360517e+00),0,RC(8.0318913859955919224117033223019560435041e+00),RC(-7.5831741663401346820798883023671588604984e+00),RC(3.5816169353190074211247685442452878696855e+00),RC(-2.4369722632199529411183809065693752383733e+00),RC(8.5158999992326179339689766032486142173390e-01),0,0,0}},
299: b[10] = {RC(4.7425837833706756083569172717574534698932e-02),0,0,RC(2.5622361659370562659961727458274623448160e-01),RC(2.6951376833074206619473817258075952886764e-01),RC(1.2686622409092782845989138364739173247882e-01),RC(2.4887225942060071622046449427647492767292e-01),RC(3.0744837408200631335304388479099184768645e-03),RC(4.8023809989496943308189063347143123323209e-02),0},
300: bembed[10] = {RC(4.7485247699299631037531273805727961552268e-02),0,0,RC(2.5599412588690633297154918245905393870497e-01),RC(2.7058478081067688722530891099268135732387e-01),RC(1.2505618684425992913638822323746917920448e-01),RC(2.5204468723743860507184043820197442562182e-01),0,0,RC(4.8834971521418614557381971303093137592592e-02)};
301: TSRKRegister(TSRK7VR,7,10,&A[0][0],b,NULL,bembed,0,NULL);
302: }
303: {
304: const PetscReal
305: A[13][13] = {{0,0,0,0,0,0,0,0,0,0,0,0,0},
306: {RC(2.5000000000000000000000000000000000000000e-01),0,0,0,0,0,0,0,0,0,0,0,0},
307: {RC(8.7400846504915232052686327594877411977046e-02),RC(2.5487604938654321753087950620345685135815e-02),0,0,0,0,0,0,0,0,0,0,0},
308: {RC(4.2333169291338582677165354330708661417323e-02),0,RC(1.2699950787401574803149606299212598425197e-01),0,0,0,0,0,0,0,0,0,0},
309: {RC(4.2609505888742261494881445237572274090942e-01),0,RC(-1.5987952846591523265427733230657181117089e+00),RC(1.5967002257717297115939588706899953707994e+00),0,0,0,0,0,0,0,0,0},
310: {RC(5.0719337296713929515090618138513639239329e-02),0,0,RC(2.5433377264600407582754714408877778031369e-01),RC(2.0394689005728199465736223777270858044698e-01),0,0,0,0,0,0,0,0},
311: {RC(-2.9000374717523110970388379285425896124091e-01),0,0,RC(1.3441873910260789889438681109414337003184e+00),RC(-2.8647779433614427309611103827036562829470e+00),RC(2.6775942995105948517211260646164815438695e+00),0,0,0,0,0,0,0},
312: {RC(9.8535011337993546469740402980727014284756e-02),0,0,0,RC(2.2192680630751384842024036498197387903583e-01),RC(-1.8140622911806994312690338288073952457474e-01),RC(1.0944411472562548236922614918038631254153e-02),0,0,0,0,0,0},
313: {RC(3.8711052545731144679444618165166373405645e-01),0,0,RC(-1.4424454974855277571256745553077927767173e+00),RC(2.9053981890699509317691346449233848441744e+00),RC(-1.8537710696301059290843332675811978025183e+00),RC(1.4003648098728154269497325109771241479223e-01),RC(5.7273940811495816575746774624447706488753e-01),0,0,0,0,0},
314: {RC(-1.6124403444439308100630016197913480595436e-01),0,0,RC(-1.7339602957358984083578404473962567894901e-01),RC(-1.3012892814065147406016812745172492529744e+00),RC(1.1379503751738617308558792131431003472124e+00),RC(-3.1747649663966880106923521138043024698980e-02),RC(9.3351293824933666439811064486056884856590e-01),RC(-8.3786318334733852703300855629616433201504e-02),0,0,0,0},
315: {RC(-1.9199444881589533281510804651483576073142e-02),0,0,RC(2.7330857265264284907942326254016124275617e-01),RC(-6.7534973206944372919691611210942380856240e-01),RC(3.4151849813846016071738489974728382711981e-01),RC(-6.7950064803375772478920516198524629391910e-02),RC(9.6591752247623878884265586491216376509746e-02),RC(1.3253082511182101180721038466545389951226e-01),RC(3.6854959360386113446906329951531666812946e-01),0,0,0},
316: {RC(6.0918774036452898676888412111588817784584e-01),0,0,RC(-2.2725690858980016768999800931413088399719e+00),RC(4.7578983426940290068155255881914785497547e+00),RC(-5.5161067066927584824294689667844248244842e+00),RC(2.9005963696801192709095818565946174378180e-01),RC(5.6914239633590368229109858454801849145630e-01),RC(7.9267957603321670271339916205893327579951e-01),RC(1.5473720453288822894126190771849898232047e-01),RC(1.6149708956621816247083215106334544434974e+00),0,0},
317: {RC(8.8735762208534719663211694051981022704884e-01),0,0,RC(-2.9754597821085367558513632804709301581977e+00),RC(5.6007170094881630597990392548350098923829e+00),RC(-5.9156074505366744680014930189941657351840e+00),RC(2.2029689156134927016879142540807638331238e-01),RC(1.0155097824462216666143271340902996997549e-01),RC(1.1514345647386055909780397752125850553556e+00),RC(1.9297101665271239396134361900805843653065e+00),0,0,0}},
318: b[13] = {RC(4.4729564666695714203015840429049382466467e-02),0,0,0,0,RC(1.5691033527708199813368698010726645409175e-01),RC(1.8460973408151637740702451873526277892035e-01),RC(2.2516380602086991042479419400350721970920e-01),RC(1.4794615651970234687005179885449141753736e-01),RC(7.6055542444955825269798361910336491012732e-02),RC(1.2277290235018619610824346315921437388535e-01),RC(4.1811958638991631583384842800871882376786e-02),0},
319: bembed[13] = {RC(4.5847111400495925878664730122010282095875e-02),0,0,0,0,RC(2.6231891404152387437443356584845803392392e-01),RC(1.9169372337852611904485738635688429008025e-01),RC(2.1709172327902618330978407422906448568196e-01),RC(1.2738189624833706796803169450656737867900e-01),RC(1.1510530385365326258240515750043192148894e-01),0,0,RC(4.0561327798437566841823391436583608050053e-02)};
320: TSRKRegister(TSRK8VR,8,13,&A[0][0],b,NULL,bembed,0,NULL);
321: }
322: #undef RC
323: return(0);
324: }
326: /*@C
327: TSRKRegisterDestroy - Frees the list of schemes that were registered by TSRKRegister().
329: Not Collective
331: Level: advanced
333: .seealso: TSRKRegister(), TSRKRegisterAll()
334: @*/
335: PetscErrorCode TSRKRegisterDestroy(void)
336: {
338: RKTableauLink link;
341: while ((link = RKTableauList)) {
342: RKTableau t = &link->tab;
343: RKTableauList = link->next;
344: PetscFree3(t->A,t->b,t->c);
345: PetscFree (t->bembed);
346: PetscFree (t->binterp);
347: PetscFree (t->name);
348: PetscFree (link);
349: }
350: TSRKRegisterAllCalled = PETSC_FALSE;
351: return(0);
352: }
354: /*@C
355: TSRKInitializePackage - This function initializes everything in the TSRK package. It is called
356: from TSInitializePackage().
358: Level: developer
360: .seealso: PetscInitialize()
361: @*/
362: PetscErrorCode TSRKInitializePackage(void)
363: {
367: if (TSRKPackageInitialized) return(0);
368: TSRKPackageInitialized = PETSC_TRUE;
369: TSRKRegisterAll();
370: PetscRegisterFinalize(TSRKFinalizePackage);
371: return(0);
372: }
374: /*@C
375: TSRKFinalizePackage - This function destroys everything in the TSRK package. It is
376: called from PetscFinalize().
378: Level: developer
380: .seealso: PetscFinalize()
381: @*/
382: PetscErrorCode TSRKFinalizePackage(void)
383: {
387: TSRKPackageInitialized = PETSC_FALSE;
388: TSRKRegisterDestroy();
389: return(0);
390: }
392: /*@C
393: TSRKRegister - register an RK scheme by providing the entries in the Butcher tableau and optionally embedded approximations and interpolation
395: Not Collective, but the same schemes should be registered on all processes on which they will be used
397: Input Parameters:
398: + name - identifier for method
399: . order - approximation order of method
400: . s - number of stages, this is the dimension of the matrices below
401: . A - stage coefficients (dimension s*s, row-major)
402: . b - step completion table (dimension s; NULL to use last row of A)
403: . c - abscissa (dimension s; NULL to use row sums of A)
404: . bembed - completion table for embedded method (dimension s; NULL if not available)
405: . p - Order of the interpolation scheme, equal to the number of columns of binterp
406: - binterp - Coefficients of the interpolation formula (dimension s*p; NULL to reuse b with p=1)
408: Notes:
409: Several RK methods are provided, this function is only needed to create new methods.
411: Level: advanced
413: .seealso: TSRK
414: @*/
415: PetscErrorCode TSRKRegister(TSRKType name,PetscInt order,PetscInt s,
416: const PetscReal A[],const PetscReal b[],const PetscReal c[],
417: const PetscReal bembed[],PetscInt p,const PetscReal binterp[])
418: {
419: PetscErrorCode ierr;
420: RKTableauLink link;
421: RKTableau t;
422: PetscInt i,j;
432: TSRKInitializePackage();
433: PetscNew(&link);
434: t = &link->tab;
436: PetscStrallocpy(name,&t->name);
437: t->order = order;
438: t->s = s;
439: PetscMalloc3(s*s,&t->A,s,&t->b,s,&t->c);
440: PetscArraycpy(t->A,A,s*s);
441: if (b) { PetscArraycpy(t->b,b,s); }
442: else for (i=0; i<s; i++) t->b[i] = A[(s-1)*s+i];
443: if (c) { PetscArraycpy(t->c,c,s); }
444: else for (i=0; i<s; i++) for (j=0,t->c[i]=0; j<s; j++) t->c[i] += A[i*s+j];
445: t->FSAL = PETSC_TRUE;
446: for (i=0; i<s; i++) if (t->A[(s-1)*s+i] != t->b[i]) t->FSAL = PETSC_FALSE;
448: if (bembed) {
449: PetscMalloc1(s,&t->bembed);
450: PetscArraycpy(t->bembed,bembed,s);
451: }
453: if (!binterp) { p = 1; binterp = t->b; }
454: t->p = p;
455: PetscMalloc1(s*p,&t->binterp);
456: PetscArraycpy(t->binterp,binterp,s*p);
458: link->next = RKTableauList;
459: RKTableauList = link;
460: return(0);
461: }
463: PetscErrorCode TSRKGetTableau_RK(TS ts, PetscInt *s, const PetscReal **A, const PetscReal **b, const PetscReal **c, const PetscReal **bembed,
464: PetscInt *p, const PetscReal **binterp, PetscBool *FSAL)
465: {
466: TS_RK *rk = (TS_RK*)ts->data;
467: RKTableau tab = rk->tableau;
470: if (s) *s = tab->s;
471: if (A) *A = tab->A;
472: if (b) *b = tab->b;
473: if (c) *c = tab->c;
474: if (bembed) *bembed = tab->bembed;
475: if (p) *p = tab->p;
476: if (binterp) *binterp = tab->binterp;
477: if (FSAL) *FSAL = tab->FSAL;
478: return(0);
479: }
481: /*@C
482: TSRKGetTableau - Get info on the RK tableau
484: Not Collective
486: Input Parameters:
487: . ts - timestepping context
489: Output Parameters:
490: + s - number of stages, this is the dimension of the matrices below
491: . A - stage coefficients (dimension s*s, row-major)
492: . b - step completion table (dimension s)
493: . c - abscissa (dimension s)
494: . bembed - completion table for embedded method (dimension s; NULL if not available)
495: . p - Order of the interpolation scheme, equal to the number of columns of binterp
496: . binterp - Coefficients of the interpolation formula (dimension s*p)
497: - FSAL - wheather or not the scheme has the First Same As Last property
499: Level: developer
501: .seealso: TSRK
502: @*/
503: PetscErrorCode TSRKGetTableau(TS ts, PetscInt *s, const PetscReal **A, const PetscReal **b, const PetscReal **c, const PetscReal **bembed,
504: PetscInt *p, const PetscReal **binterp, PetscBool *FSAL)
505: {
510: PetscUseMethod(ts,"TSRKGetTableau_C",(TS,PetscInt*,const PetscReal**,const PetscReal**,const PetscReal**,const PetscReal**,
511: PetscInt*,const PetscReal**,PetscBool*),(ts,s,A,b,c,bembed,p,binterp,FSAL));
512: return(0);
513: }
515: /*
516: This is for single-step RK method
517: The step completion formula is
519: x1 = x0 + h b^T YdotRHS
521: This function can be called before or after ts->vec_sol has been updated.
522: Suppose we have a completion formula (b) and an embedded formula (be) of different order.
523: We can write
525: x1e = x0 + h be^T YdotRHS
526: = x1 - h b^T YdotRHS + h be^T YdotRHS
527: = x1 + h (be - b)^T YdotRHS
529: so we can evaluate the method with different order even after the step has been optimistically completed.
530: */
531: static PetscErrorCode TSEvaluateStep_RK(TS ts,PetscInt order,Vec X,PetscBool *done)
532: {
533: TS_RK *rk = (TS_RK*)ts->data;
534: RKTableau tab = rk->tableau;
535: PetscScalar *w = rk->work;
536: PetscReal h;
537: PetscInt s = tab->s,j;
541: switch (rk->status) {
542: case TS_STEP_INCOMPLETE:
543: case TS_STEP_PENDING:
544: h = ts->time_step; break;
545: case TS_STEP_COMPLETE:
546: h = ts->ptime - ts->ptime_prev; break;
547: default: SETERRQ(PetscObjectComm((PetscObject)ts),PETSC_ERR_PLIB,"Invalid TSStepStatus");
548: }
549: if (order == tab->order) {
550: if (rk->status == TS_STEP_INCOMPLETE) {
551: VecCopy(ts->vec_sol,X);
552: for (j=0; j<s; j++) w[j] = h*tab->b[j]/rk->dtratio;
553: VecMAXPY(X,s,w,rk->YdotRHS);
554: } else {VecCopy(ts->vec_sol,X);}
555: return(0);
556: } else if (order == tab->order-1) {
557: if (!tab->bembed) goto unavailable;
558: if (rk->status == TS_STEP_INCOMPLETE) { /*Complete with the embedded method (be)*/
559: VecCopy(ts->vec_sol,X);
560: for (j=0; j<s; j++) w[j] = h*tab->bembed[j];
561: VecMAXPY(X,s,w,rk->YdotRHS);
562: } else { /*Rollback and re-complete using (be-b) */
563: VecCopy(ts->vec_sol,X);
564: for (j=0; j<s; j++) w[j] = h*(tab->bembed[j] - tab->b[j]);
565: VecMAXPY(X,s,w,rk->YdotRHS);
566: }
567: if (done) *done = PETSC_TRUE;
568: return(0);
569: }
570: unavailable:
571: if (done) *done = PETSC_FALSE;
572: else SETERRQ3(PetscObjectComm((PetscObject)ts),PETSC_ERR_SUP,"RK '%s' of order %D cannot evaluate step at order %D. Consider using -ts_adapt_type none or a different method that has an embedded estimate.",tab->name,tab->order,order);
573: return(0);
574: }
576: static PetscErrorCode TSForwardCostIntegral_RK(TS ts)
577: {
578: TS_RK *rk = (TS_RK*)ts->data;
579: TS quadts = ts->quadraturets;
580: RKTableau tab = rk->tableau;
581: const PetscInt s = tab->s;
582: const PetscReal *b = tab->b,*c = tab->c;
583: Vec *Y = rk->Y;
584: PetscInt i;
585: PetscErrorCode ierr;
588: /* No need to backup quadts->vec_sol since it can be reverted in TSRollBack_RK */
589: for (i=s-1; i>=0; i--) {
590: /* Evolve quadrature TS solution to compute integrals */
591: TSComputeRHSFunction(quadts,rk->ptime+rk->time_step*c[i],Y[i],ts->vec_costintegrand);
592: VecAXPY(quadts->vec_sol,rk->time_step*b[i],ts->vec_costintegrand);
593: }
594: return(0);
595: }
597: static PetscErrorCode TSAdjointCostIntegral_RK(TS ts)
598: {
599: TS_RK *rk = (TS_RK*)ts->data;
600: RKTableau tab = rk->tableau;
601: TS quadts = ts->quadraturets;
602: const PetscInt s = tab->s;
603: const PetscReal *b = tab->b,*c = tab->c;
604: Vec *Y = rk->Y;
605: PetscInt i;
606: PetscErrorCode ierr;
609: for (i=s-1; i>=0; i--) {
610: /* Evolve quadrature TS solution to compute integrals */
611: TSComputeRHSFunction(quadts,ts->ptime+ts->time_step*(1.0-c[i]),Y[i],ts->vec_costintegrand);
612: VecAXPY(quadts->vec_sol,-ts->time_step*b[i],ts->vec_costintegrand);
613: }
614: return(0);
615: }
617: static PetscErrorCode TSRollBack_RK(TS ts)
618: {
619: TS_RK *rk = (TS_RK*)ts->data;
620: TS quadts = ts->quadraturets;
621: RKTableau tab = rk->tableau;
622: const PetscInt s = tab->s;
623: const PetscReal *b = tab->b,*c = tab->c;
624: PetscScalar *w = rk->work;
625: Vec *Y = rk->Y,*YdotRHS = rk->YdotRHS;
626: PetscInt j;
627: PetscReal h;
628: PetscErrorCode ierr;
631: switch (rk->status) {
632: case TS_STEP_INCOMPLETE:
633: case TS_STEP_PENDING:
634: h = ts->time_step; break;
635: case TS_STEP_COMPLETE:
636: h = ts->ptime - ts->ptime_prev; break;
637: default: SETERRQ(PetscObjectComm((PetscObject)ts),PETSC_ERR_PLIB,"Invalid TSStepStatus");
638: }
639: for (j=0; j<s; j++) w[j] = -h*b[j];
640: VecMAXPY(ts->vec_sol,s,w,YdotRHS);
641: if (quadts && ts->costintegralfwd) {
642: for (j=0; j<s; j++) {
643: /* Revert the quadrature TS solution */
644: TSComputeRHSFunction(quadts,rk->ptime+h*c[j],Y[j],ts->vec_costintegrand);
645: VecAXPY(quadts->vec_sol,-h*b[j],ts->vec_costintegrand);
646: }
647: }
648: return(0);
649: }
651: static PetscErrorCode TSForwardStep_RK(TS ts)
652: {
653: TS_RK *rk = (TS_RK*)ts->data;
654: RKTableau tab = rk->tableau;
655: Mat J,*MatsFwdSensipTemp = rk->MatsFwdSensipTemp;
656: const PetscInt s = tab->s;
657: const PetscReal *A = tab->A,*c = tab->c,*b = tab->b;
658: Vec *Y = rk->Y;
659: PetscInt i,j;
660: PetscReal stage_time,h = ts->time_step;
661: PetscBool zero;
662: PetscErrorCode ierr;
665: MatCopy(ts->mat_sensip,rk->MatFwdSensip0,SAME_NONZERO_PATTERN);
666: TSGetRHSJacobian(ts,&J,NULL,NULL,NULL);
668: for (i=0; i<s; i++) {
669: stage_time = ts->ptime + h*c[i];
670: zero = PETSC_FALSE;
671: if (b[i] == 0 && i == s-1) zero = PETSC_TRUE;
672: /* TLM Stage values */
673: if(!i) {
674: MatCopy(ts->mat_sensip,rk->MatsFwdStageSensip[i],SAME_NONZERO_PATTERN);
675: } else if (!zero) {
676: MatZeroEntries(rk->MatsFwdStageSensip[i]);
677: for (j=0; j<i; j++) {
678: MatAXPY(rk->MatsFwdStageSensip[i],h*A[i*s+j],MatsFwdSensipTemp[j],SAME_NONZERO_PATTERN);
679: }
680: MatAXPY(rk->MatsFwdStageSensip[i],1.,ts->mat_sensip,SAME_NONZERO_PATTERN);
681: } else {
682: MatZeroEntries(rk->MatsFwdStageSensip[i]);
683: }
685: TSComputeRHSJacobian(ts,stage_time,Y[i],J,J);
686: MatMatMult(J,rk->MatsFwdStageSensip[i],MAT_REUSE_MATRIX,PETSC_DEFAULT,&MatsFwdSensipTemp[i]);
687: if (ts->Jacprhs) {
688: TSComputeRHSJacobianP(ts,stage_time,Y[i],ts->Jacprhs); /* get f_p */
689: if (ts->vecs_sensi2p) { /* TLM used for 2nd-order adjoint */
690: PetscScalar *xarr;
691: MatDenseGetColumn(MatsFwdSensipTemp[i],0,&xarr);
692: VecPlaceArray(rk->VecDeltaFwdSensipCol,xarr);
693: MatMultAdd(ts->Jacprhs,ts->vec_dir,rk->VecDeltaFwdSensipCol,rk->VecDeltaFwdSensipCol);
694: VecResetArray(rk->VecDeltaFwdSensipCol);
695: MatDenseRestoreColumn(MatsFwdSensipTemp[i],&xarr);
696: } else {
697: MatAXPY(MatsFwdSensipTemp[i],1.,ts->Jacprhs,SUBSET_NONZERO_PATTERN);
698: }
699: }
700: }
702: for (i=0; i<s; i++) {
703: MatAXPY(ts->mat_sensip,h*b[i],rk->MatsFwdSensipTemp[i],SAME_NONZERO_PATTERN);
704: }
705: rk->status = TS_STEP_COMPLETE;
706: return(0);
707: }
709: static PetscErrorCode TSForwardGetStages_RK(TS ts,PetscInt *ns,Mat **stagesensip)
710: {
711: TS_RK *rk = (TS_RK*)ts->data;
712: RKTableau tab = rk->tableau;
715: if (ns) *ns = tab->s;
716: if (stagesensip) *stagesensip = rk->MatsFwdStageSensip;
717: return(0);
718: }
720: static PetscErrorCode TSForwardSetUp_RK(TS ts)
721: {
722: TS_RK *rk = (TS_RK*)ts->data;
723: RKTableau tab = rk->tableau;
724: PetscInt i;
728: /* backup sensitivity results for roll-backs */
729: MatDuplicate(ts->mat_sensip,MAT_DO_NOT_COPY_VALUES,&rk->MatFwdSensip0);
731: PetscMalloc1(tab->s,&rk->MatsFwdStageSensip);
732: PetscMalloc1(tab->s,&rk->MatsFwdSensipTemp);
733: for(i=0; i<tab->s; i++) {
734: MatDuplicate(ts->mat_sensip,MAT_DO_NOT_COPY_VALUES,&rk->MatsFwdStageSensip[i]);
735: MatDuplicate(ts->mat_sensip,MAT_DO_NOT_COPY_VALUES,&rk->MatsFwdSensipTemp[i]);
736: }
737: VecDuplicate(ts->vec_sol,&rk->VecDeltaFwdSensipCol);
738: return(0);
739: }
741: static PetscErrorCode TSForwardReset_RK(TS ts)
742: {
743: TS_RK *rk = (TS_RK*)ts->data;
744: RKTableau tab = rk->tableau;
745: PetscInt i;
749: MatDestroy(&rk->MatFwdSensip0);
750: if (rk->MatsFwdStageSensip) {
751: for (i=0; i<tab->s; i++) {
752: MatDestroy(&rk->MatsFwdStageSensip[i]);
753: }
754: PetscFree(rk->MatsFwdStageSensip);
755: }
756: if (rk->MatsFwdSensipTemp) {
757: for (i=0; i<tab->s; i++) {
758: MatDestroy(&rk->MatsFwdSensipTemp[i]);
759: }
760: PetscFree(rk->MatsFwdSensipTemp);
761: }
762: VecDestroy(&rk->VecDeltaFwdSensipCol);
763: return(0);
764: }
766: static PetscErrorCode TSStep_RK(TS ts)
767: {
768: TS_RK *rk = (TS_RK*)ts->data;
769: RKTableau tab = rk->tableau;
770: const PetscInt s = tab->s;
771: const PetscReal *A = tab->A,*c = tab->c;
772: PetscScalar *w = rk->work;
773: Vec *Y = rk->Y,*YdotRHS = rk->YdotRHS;
774: PetscBool FSAL = tab->FSAL;
775: TSAdapt adapt;
776: PetscInt i,j;
777: PetscInt rejections = 0;
778: PetscBool stageok,accept = PETSC_TRUE;
779: PetscReal next_time_step = ts->time_step;
780: PetscErrorCode ierr;
783: if (ts->steprollback || ts->steprestart) FSAL = PETSC_FALSE;
784: if (FSAL) { VecCopy(YdotRHS[s-1],YdotRHS[0]); }
786: rk->status = TS_STEP_INCOMPLETE;
787: while (!ts->reason && rk->status != TS_STEP_COMPLETE) {
788: PetscReal t = ts->ptime;
789: PetscReal h = ts->time_step;
790: for (i=0; i<s; i++) {
791: rk->stage_time = t + h*c[i];
792: TSPreStage(ts,rk->stage_time);
793: VecCopy(ts->vec_sol,Y[i]);
794: for (j=0; j<i; j++) w[j] = h*A[i*s+j];
795: VecMAXPY(Y[i],i,w,YdotRHS);
796: TSPostStage(ts,rk->stage_time,i,Y);
797: TSGetAdapt(ts,&adapt);
798: TSAdaptCheckStage(adapt,ts,rk->stage_time,Y[i],&stageok);
799: if (!stageok) goto reject_step;
800: if (FSAL && !i) continue;
801: TSComputeRHSFunction(ts,t+h*c[i],Y[i],YdotRHS[i]);
802: }
804: rk->status = TS_STEP_INCOMPLETE;
805: TSEvaluateStep(ts,tab->order,ts->vec_sol,NULL);
806: rk->status = TS_STEP_PENDING;
807: TSGetAdapt(ts,&adapt);
808: TSAdaptCandidatesClear(adapt);
809: TSAdaptCandidateAdd(adapt,tab->name,tab->order,1,tab->ccfl,(PetscReal)tab->s,PETSC_TRUE);
810: TSAdaptChoose(adapt,ts,ts->time_step,NULL,&next_time_step,&accept);
811: rk->status = accept ? TS_STEP_COMPLETE : TS_STEP_INCOMPLETE;
812: if (!accept) { /* Roll back the current step */
813: TSRollBack_RK(ts);
814: ts->time_step = next_time_step;
815: goto reject_step;
816: }
818: if (ts->costintegralfwd) { /* Save the info for the later use in cost integral evaluation */
819: rk->ptime = ts->ptime;
820: rk->time_step = ts->time_step;
821: }
823: ts->ptime += ts->time_step;
824: ts->time_step = next_time_step;
825: break;
827: reject_step:
828: ts->reject++; accept = PETSC_FALSE;
829: if (!ts->reason && ++rejections > ts->max_reject && ts->max_reject >= 0) {
830: ts->reason = TS_DIVERGED_STEP_REJECTED;
831: PetscInfo2(ts,"Step=%D, step rejections %D greater than current TS allowed, stopping solve\n",ts->steps,rejections);
832: }
833: }
834: return(0);
835: }
837: static PetscErrorCode TSAdjointSetUp_RK(TS ts)
838: {
839: TS_RK *rk = (TS_RK*)ts->data;
840: RKTableau tab = rk->tableau;
841: PetscInt s = tab->s;
845: if (ts->adjointsetupcalled++) return(0);
846: VecDuplicateVecs(ts->vecs_sensi[0],s*ts->numcost,&rk->VecsDeltaLam);
847: VecDuplicateVecs(ts->vecs_sensi[0],ts->numcost,&rk->VecsSensiTemp);
848: if(ts->vecs_sensip) {
849: VecDuplicate(ts->vecs_sensip[0],&rk->VecDeltaMu);
850: }
851: if (ts->vecs_sensi2) {
852: VecDuplicateVecs(ts->vecs_sensi[0],s*ts->numcost,&rk->VecsDeltaLam2);
853: VecDuplicateVecs(ts->vecs_sensi2[0],ts->numcost,&rk->VecsSensi2Temp);
854: }
855: if (ts->vecs_sensi2p) {
856: VecDuplicate(ts->vecs_sensi2p[0],&rk->VecDeltaMu2);
857: }
858: return(0);
859: }
861: /*
862: Assumptions:
863: - TSStep_RK() always evaluates the step with b, not bembed.
864: */
865: static PetscErrorCode TSAdjointStep_RK(TS ts)
866: {
867: TS_RK *rk = (TS_RK*)ts->data;
868: TS quadts = ts->quadraturets;
869: RKTableau tab = rk->tableau;
870: Mat J,Jpre,Jquad;
871: const PetscInt s = tab->s;
872: const PetscReal *A = tab->A,*b = tab->b,*c = tab->c;
873: PetscScalar *w = rk->work,*xarr;
874: Vec *Y = rk->Y,*VecsDeltaLam = rk->VecsDeltaLam,VecDeltaMu = rk->VecDeltaMu,*VecsSensiTemp = rk->VecsSensiTemp;
875: Vec *VecsDeltaLam2 = rk->VecsDeltaLam2,VecDeltaMu2 = rk->VecDeltaMu2,*VecsSensi2Temp = rk->VecsSensi2Temp;
876: Vec VecDRDUTransCol = ts->vec_drdu_col,VecDRDPTransCol = ts->vec_drdp_col;
877: PetscInt i,j,nadj;
878: PetscReal t = ts->ptime;
879: PetscReal h = ts->time_step;
880: PetscErrorCode ierr;
883: rk->status = TS_STEP_INCOMPLETE;
885: TSGetRHSJacobian(ts,&J,&Jpre,NULL,NULL);
886: if (quadts) {
887: TSGetRHSJacobian(quadts,&Jquad,NULL,NULL,NULL);
888: }
889: for (i=s-1; i>=0; i--) {
890: if (tab->FSAL && i == s-1) {
891: /* VecsDeltaLam[nadj*s+s-1] are initialized with zeros and the values never change.*/
892: continue;
893: }
894: rk->stage_time = t + h*(1.0-c[i]);
895: TSComputeSNESJacobian(ts,Y[i],J,Jpre);
896: if (quadts) {
897: TSComputeRHSJacobian(quadts,rk->stage_time,Y[i],Jquad,Jquad); /* get r_u^T */
898: }
899: if (ts->vecs_sensip) {
900: TSComputeRHSJacobianP(ts,rk->stage_time,Y[i],ts->Jacprhs); /* get f_p */
901: if (quadts) {
902: TSComputeRHSJacobianP(quadts,rk->stage_time,Y[i],quadts->Jacprhs); /* get f_p for the quadrature */
903: }
904: }
906: if (b[i]) {
907: for (j=i+1; j<s; j++) w[j-i-1] = A[j*s+i]/b[i]; /* coefficients for computing VecsSensiTemp */
908: } else {
909: for (j=i+1; j<s; j++) w[j-i-1] = A[j*s+i]; /* coefficients for computing VecsSensiTemp */
910: }
912: for (nadj=0; nadj<ts->numcost; nadj++) {
913: /* Stage values of lambda */
914: if (b[i]) {
915: /* lambda_{n+1} + \sum_{j=i+1}^s a_{ji}/b[i]*lambda_{s,j} */
916: VecCopy(ts->vecs_sensi[nadj],VecsSensiTemp[nadj]); /* VecDeltaLam is an vec array of size s by numcost */
917: VecMAXPY(VecsSensiTemp[nadj],s-i-1,w,&VecsDeltaLam[nadj*s+i+1]);
918: MatMultTranspose(J,VecsSensiTemp[nadj],VecsDeltaLam[nadj*s+i]); /* VecsSensiTemp will be reused by 2nd-order adjoint */
919: VecScale(VecsDeltaLam[nadj*s+i],-h*b[i]);
920: if (quadts) {
921: MatDenseGetColumn(Jquad,nadj,&xarr);
922: VecPlaceArray(VecDRDUTransCol,xarr);
923: VecAXPY(VecsDeltaLam[nadj*s+i],-h*b[i],VecDRDUTransCol);
924: VecResetArray(VecDRDUTransCol);
925: MatDenseRestoreColumn(Jquad,&xarr);
926: }
927: } else {
928: /* \sum_{j=i+1}^s a_{ji}*lambda_{s,j} */
929: VecSet(VecsSensiTemp[nadj],0);
930: VecMAXPY(VecsSensiTemp[nadj],s-i-1,w,&VecsDeltaLam[nadj*s+i+1]);
931: MatMultTranspose(J,VecsSensiTemp[nadj],VecsDeltaLam[nadj*s+i]);
932: VecScale(VecsDeltaLam[nadj*s+i],-h);
933: }
935: /* Stage values of mu */
936: if (ts->vecs_sensip) {
937: MatMultTranspose(ts->Jacprhs,VecsSensiTemp[nadj],VecDeltaMu);
938: if (b[i]) {
939: VecScale(VecDeltaMu,-h*b[i]);
940: if (quadts) {
941: MatDenseGetColumn(quadts->Jacprhs,nadj,&xarr);
942: VecPlaceArray(VecDRDPTransCol,xarr);
943: VecAXPY(VecDeltaMu,-h*b[i],VecDRDPTransCol);
944: VecResetArray(VecDRDPTransCol);
945: MatDenseRestoreColumn(quadts->Jacprhs,&xarr);
946: }
947: } else {
948: VecScale(VecDeltaMu,-h);
949: }
950: VecAXPY(ts->vecs_sensip[nadj],1.,VecDeltaMu); /* update sensip for each stage */
951: }
952: }
954: if (ts->vecs_sensi2 && ts->forward_solve) { /* 2nd-order adjoint, TLM mode has to be turned on */
955: /* Get w1 at t_{n+1} from TLM matrix */
956: MatDenseGetColumn(rk->MatsFwdStageSensip[i],0,&xarr);
957: VecPlaceArray(ts->vec_sensip_col,xarr);
958: /* lambda_s^T F_UU w_1 */
959: TSComputeRHSHessianProductFunctionUU(ts,rk->stage_time,Y[i],VecsSensiTemp,ts->vec_sensip_col,ts->vecs_guu);
960: if (quadts) {
961: /* R_UU w_1 */
962: TSComputeRHSHessianProductFunctionUU(quadts,rk->stage_time,Y[i],NULL,ts->vec_sensip_col,ts->vecs_guu);
963: }
964: if (ts->vecs_sensip) {
965: /* lambda_s^T F_UP w_2 */
966: TSComputeRHSHessianProductFunctionUP(ts,rk->stage_time,Y[i],VecsSensiTemp,ts->vec_dir,ts->vecs_gup);
967: if (quadts) {
968: /* R_UP w_2 */
969: TSComputeRHSHessianProductFunctionUP(quadts,rk->stage_time,Y[i],NULL,ts->vec_sensip_col,ts->vecs_gup);
970: }
971: }
972: if (ts->vecs_sensi2p) {
973: /* lambda_s^T F_PU w_1 */
974: TSComputeRHSHessianProductFunctionPU(ts,rk->stage_time,Y[i],VecsSensiTemp,ts->vec_sensip_col,ts->vecs_gpu);
975: /* lambda_s^T F_PP w_2 */
976: TSComputeRHSHessianProductFunctionPP(ts,rk->stage_time,Y[i],VecsSensiTemp,ts->vec_dir,ts->vecs_gpp);
977: if (b[i] && quadts) {
978: /* R_PU w_1 */
979: TSComputeRHSHessianProductFunctionPU(quadts,rk->stage_time,Y[i],NULL,ts->vec_sensip_col,ts->vecs_gpu);
980: /* R_PP w_2 */
981: TSComputeRHSHessianProductFunctionPP(quadts,rk->stage_time,Y[i],NULL,ts->vec_dir,ts->vecs_gpp);
982: }
983: }
984: VecResetArray(ts->vec_sensip_col);
985: MatDenseRestoreColumn(rk->MatsFwdStageSensip[i],&xarr);
987: for (nadj=0; nadj<ts->numcost; nadj++) {
988: /* Stage values of lambda */
989: if (b[i]) {
990: /* J_i^T*(Lambda_{n+1}+\sum_{j=i+1}^s a_{ji}/b_i*Lambda_{s,j} */
991: VecCopy(ts->vecs_sensi2[nadj],VecsSensi2Temp[nadj]);
992: VecMAXPY(VecsSensi2Temp[nadj],s-i-1,w,&VecsDeltaLam2[nadj*s+i+1]);
993: MatMultTranspose(J,VecsSensi2Temp[nadj],VecsDeltaLam2[nadj*s+i]);
994: VecScale(VecsDeltaLam2[nadj*s+i],-h*b[i]);
995: VecAXPY(VecsDeltaLam2[nadj*s+i],-h*b[i],ts->vecs_guu[nadj]);
996: if (ts->vecs_sensip) {
997: VecAXPY(VecsDeltaLam2[nadj*s+i],-h*b[i],ts->vecs_gup[nadj]);
998: }
999: } else {
1000: /* \sum_{j=i+1}^s a_{ji}*Lambda_{s,j} */
1001: VecSet(VecsDeltaLam2[nadj*s+i],0);
1002: VecMAXPY(VecsSensi2Temp[nadj],s-i-1,w,&VecsDeltaLam2[nadj*s+i+1]);
1003: MatMultTranspose(J,VecsSensi2Temp[nadj],VecsDeltaLam2[nadj*s+i]);
1004: VecScale(VecsDeltaLam2[nadj*s+i],-h);
1005: VecAXPY(VecsDeltaLam2[nadj*s+i],-h,ts->vecs_guu[nadj]);
1006: if (ts->vecs_sensip) {
1007: VecAXPY(VecsDeltaLam2[nadj*s+i],-h,ts->vecs_gup[nadj]);
1008: }
1009: }
1010: if (ts->vecs_sensi2p) { /* 2nd-order adjoint for parameters */
1011: MatMultTranspose(ts->Jacprhs,VecsSensi2Temp[nadj],VecDeltaMu2);
1012: if (b[i]) {
1013: VecScale(VecDeltaMu2,-h*b[i]);
1014: VecAXPY(VecDeltaMu2,-h*b[i],ts->vecs_gpu[nadj]);
1015: VecAXPY(VecDeltaMu2,-h*b[i],ts->vecs_gpp[nadj]);
1016: } else {
1017: VecScale(VecDeltaMu2,-h);
1018: VecAXPY(VecDeltaMu2,-h,ts->vecs_gpu[nadj]);
1019: VecAXPY(VecDeltaMu2,-h,ts->vecs_gpp[nadj]);
1020: }
1021: VecAXPY(ts->vecs_sensi2p[nadj],1,VecDeltaMu2); /* update sensi2p for each stage */
1022: }
1023: }
1024: }
1025: }
1027: for (j=0; j<s; j++) w[j] = 1.0;
1028: for (nadj=0; nadj<ts->numcost; nadj++) { /* no need to do this for mu's */
1029: VecMAXPY(ts->vecs_sensi[nadj],s,w,&VecsDeltaLam[nadj*s]);
1030: if (ts->vecs_sensi2) {
1031: VecMAXPY(ts->vecs_sensi2[nadj],s,w,&VecsDeltaLam2[nadj*s]);
1032: }
1033: }
1034: rk->status = TS_STEP_COMPLETE;
1035: return(0);
1036: }
1038: static PetscErrorCode TSAdjointReset_RK(TS ts)
1039: {
1040: TS_RK *rk = (TS_RK*)ts->data;
1041: RKTableau tab = rk->tableau;
1045: VecDestroyVecs(tab->s*ts->numcost,&rk->VecsDeltaLam);
1046: VecDestroyVecs(ts->numcost,&rk->VecsSensiTemp);
1047: VecDestroy(&rk->VecDeltaMu);
1048: VecDestroyVecs(tab->s*ts->numcost,&rk->VecsDeltaLam2);
1049: VecDestroy(&rk->VecDeltaMu2);
1050: VecDestroyVecs(ts->numcost,&rk->VecsSensi2Temp);
1051: return(0);
1052: }
1054: static PetscErrorCode TSInterpolate_RK(TS ts,PetscReal itime,Vec X)
1055: {
1056: TS_RK *rk = (TS_RK*)ts->data;
1057: PetscInt s = rk->tableau->s,p = rk->tableau->p,i,j;
1058: PetscReal h;
1059: PetscReal tt,t;
1060: PetscScalar *b;
1061: const PetscReal *B = rk->tableau->binterp;
1062: PetscErrorCode ierr;
1065: if (!B) SETERRQ1(PetscObjectComm((PetscObject)ts),PETSC_ERR_SUP,"TSRK %s does not have an interpolation formula",rk->tableau->name);
1067: switch (rk->status) {
1068: case TS_STEP_INCOMPLETE:
1069: case TS_STEP_PENDING:
1070: h = ts->time_step;
1071: t = (itime - ts->ptime)/h;
1072: break;
1073: case TS_STEP_COMPLETE:
1074: h = ts->ptime - ts->ptime_prev;
1075: t = (itime - ts->ptime)/h + 1; /* In the interval [0,1] */
1076: break;
1077: default: SETERRQ(PetscObjectComm((PetscObject)ts),PETSC_ERR_PLIB,"Invalid TSStepStatus");
1078: }
1079: PetscMalloc1(s,&b);
1080: for (i=0; i<s; i++) b[i] = 0;
1081: for (j=0,tt=t; j<p; j++,tt*=t) {
1082: for (i=0; i<s; i++) {
1083: b[i] += h * B[i*p+j] * tt;
1084: }
1085: }
1086: VecCopy(rk->Y[0],X);
1087: VecMAXPY(X,s,b,rk->YdotRHS);
1088: PetscFree(b);
1089: return(0);
1090: }
1092: /*------------------------------------------------------------*/
1094: static PetscErrorCode TSRKTableauReset(TS ts)
1095: {
1096: TS_RK *rk = (TS_RK*)ts->data;
1097: RKTableau tab = rk->tableau;
1101: if (!tab) return(0);
1102: PetscFree(rk->work);
1103: VecDestroyVecs(tab->s,&rk->Y);
1104: VecDestroyVecs(tab->s,&rk->YdotRHS);
1105: VecDestroyVecs(tab->s*ts->numcost,&rk->VecsDeltaLam);
1106: VecDestroyVecs(ts->numcost,&rk->VecsSensiTemp);
1107: VecDestroy(&rk->VecDeltaMu);
1108: return(0);
1109: }
1111: static PetscErrorCode TSReset_RK(TS ts)
1112: {
1116: TSRKTableauReset(ts);
1117: if (ts->use_splitrhsfunction) {
1118: PetscTryMethod(ts,"TSReset_RK_MultirateSplit_C",(TS),(ts));
1119: } else {
1120: PetscTryMethod(ts,"TSReset_RK_MultirateNonsplit_C",(TS),(ts));
1121: }
1122: return(0);
1123: }
1125: static PetscErrorCode DMCoarsenHook_TSRK(DM fine,DM coarse,void *ctx)
1126: {
1128: return(0);
1129: }
1131: static PetscErrorCode DMRestrictHook_TSRK(DM fine,Mat restrct,Vec rscale,Mat inject,DM coarse,void *ctx)
1132: {
1134: return(0);
1135: }
1138: static PetscErrorCode DMSubDomainHook_TSRK(DM dm,DM subdm,void *ctx)
1139: {
1141: return(0);
1142: }
1144: static PetscErrorCode DMSubDomainRestrictHook_TSRK(DM dm,VecScatter gscat,VecScatter lscat,DM subdm,void *ctx)
1145: {
1148: return(0);
1149: }
1151: static PetscErrorCode TSRKTableauSetUp(TS ts)
1152: {
1153: TS_RK *rk = (TS_RK*)ts->data;
1154: RKTableau tab = rk->tableau;
1158: PetscMalloc1(tab->s,&rk->work);
1159: VecDuplicateVecs(ts->vec_sol,tab->s,&rk->Y);
1160: VecDuplicateVecs(ts->vec_sol,tab->s,&rk->YdotRHS);
1161: return(0);
1162: }
1164: static PetscErrorCode TSSetUp_RK(TS ts)
1165: {
1166: TS quadts = ts->quadraturets;
1168: DM dm;
1171: TSCheckImplicitTerm(ts);
1172: TSRKTableauSetUp(ts);
1173: if (quadts && ts->costintegralfwd) {
1174: Mat Jquad;
1175: TSGetRHSJacobian(quadts,&Jquad,NULL,NULL,NULL);
1176: }
1177: TSGetDM(ts,&dm);
1178: DMCoarsenHookAdd(dm,DMCoarsenHook_TSRK,DMRestrictHook_TSRK,ts);
1179: DMSubDomainHookAdd(dm,DMSubDomainHook_TSRK,DMSubDomainRestrictHook_TSRK,ts);
1180: if (ts->use_splitrhsfunction) {
1181: PetscTryMethod(ts,"TSSetUp_RK_MultirateSplit_C",(TS),(ts));
1182: } else {
1183: PetscTryMethod(ts,"TSSetUp_RK_MultirateNonsplit_C",(TS),(ts));
1184: }
1185: return(0);
1186: }
1188: static PetscErrorCode TSSetFromOptions_RK(PetscOptionItems *PetscOptionsObject,TS ts)
1189: {
1190: TS_RK *rk = (TS_RK*)ts->data;
1194: PetscOptionsHead(PetscOptionsObject,"RK ODE solver options");
1195: {
1196: RKTableauLink link;
1197: PetscInt count,choice;
1198: PetscBool flg,use_multirate = PETSC_FALSE;
1199: const char **namelist;
1201: for (link=RKTableauList,count=0; link; link=link->next,count++) ;
1202: PetscMalloc1(count,(char***)&namelist);
1203: for (link=RKTableauList,count=0; link; link=link->next,count++) namelist[count] = link->tab.name;
1204: PetscOptionsBool("-ts_rk_multirate","Use interpolation-based multirate RK method","TSRKSetMultirate",rk->use_multirate,&use_multirate,&flg);
1205: if (flg) {
1206: TSRKSetMultirate(ts,use_multirate);
1207: }
1208: PetscOptionsEList("-ts_rk_type","Family of RK method","TSRKSetType",(const char*const*)namelist,count,rk->tableau->name,&choice,&flg);
1209: if (flg) {TSRKSetType(ts,namelist[choice]);}
1210: PetscFree(namelist);
1211: }
1212: PetscOptionsTail();
1213: PetscOptionsBegin(PETSC_COMM_WORLD,NULL,"Multirate methods options","");
1214: PetscOptionsInt("-ts_rk_dtratio","time step ratio between slow and fast","",rk->dtratio,&rk->dtratio,NULL);
1215: PetscOptionsEnd();
1216: return(0);
1217: }
1219: static PetscErrorCode TSView_RK(TS ts,PetscViewer viewer)
1220: {
1221: TS_RK *rk = (TS_RK*)ts->data;
1222: PetscBool iascii;
1226: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
1227: if (iascii) {
1228: RKTableau tab = rk->tableau;
1229: TSRKType rktype;
1230: const PetscReal *c;
1231: PetscInt s;
1232: char buf[512];
1233: PetscBool FSAL;
1235: TSRKGetType(ts,&rktype);
1236: TSRKGetTableau(ts,&s,NULL,NULL,&c,NULL,NULL,NULL,&FSAL);
1237: PetscViewerASCIIPrintf(viewer," RK type %s\n",rktype);
1238: PetscViewerASCIIPrintf(viewer," Order: %D\n",tab->order);
1239: PetscViewerASCIIPrintf(viewer," FSAL property: %s\n",FSAL ? "yes" : "no");
1240: PetscFormatRealArray(buf,sizeof(buf),"% 8.6f",s,c);
1241: PetscViewerASCIIPrintf(viewer," Abscissa c = %s\n",buf);
1242: }
1243: return(0);
1244: }
1246: static PetscErrorCode TSLoad_RK(TS ts,PetscViewer viewer)
1247: {
1249: TSAdapt adapt;
1252: TSGetAdapt(ts,&adapt);
1253: TSAdaptLoad(adapt,viewer);
1254: return(0);
1255: }
1257: /*@
1258: TSRKGetOrder - Get the order of RK scheme
1260: Not collective
1262: Input Parameter:
1263: . ts - timestepping context
1265: Output Parameter:
1266: . order - order of RK-scheme
1268: Level: intermediate
1270: .seealso: TSRKGetType()
1271: @*/
1272: PetscErrorCode TSRKGetOrder(TS ts,PetscInt *order)
1273: {
1279: PetscUseMethod(ts,"TSRKGetOrder_C",(TS,PetscInt*),(ts,order));
1280: return(0);
1281: }
1283: /*@C
1284: TSRKSetType - Set the type of RK scheme
1286: Logically collective
1288: Input Parameter:
1289: + ts - timestepping context
1290: - rktype - type of RK-scheme
1292: Options Database:
1293: . -ts_rk_type - <1fe,2a,3,3bs,4,5f,5dp,5bs>
1295: Level: intermediate
1297: .seealso: TSRKGetType(), TSRK, TSRKType, TSRK1FE, TSRK2A, TSRK3, TSRK3BS, TSRK4, TSRK5F, TSRK5DP, TSRK5BS, TSRK6VR, TSRK7VR, TSRK8VR
1298: @*/
1299: PetscErrorCode TSRKSetType(TS ts,TSRKType rktype)
1300: {
1306: PetscTryMethod(ts,"TSRKSetType_C",(TS,TSRKType),(ts,rktype));
1307: return(0);
1308: }
1310: /*@C
1311: TSRKGetType - Get the type of RK scheme
1313: Not collective
1315: Input Parameter:
1316: . ts - timestepping context
1318: Output Parameter:
1319: . rktype - type of RK-scheme
1321: Level: intermediate
1323: .seealso: TSRKSetType()
1324: @*/
1325: PetscErrorCode TSRKGetType(TS ts,TSRKType *rktype)
1326: {
1331: PetscUseMethod(ts,"TSRKGetType_C",(TS,TSRKType*),(ts,rktype));
1332: return(0);
1333: }
1335: static PetscErrorCode TSRKGetOrder_RK(TS ts,PetscInt *order)
1336: {
1337: TS_RK *rk = (TS_RK*)ts->data;
1340: *order = rk->tableau->order;
1341: return(0);
1342: }
1344: static PetscErrorCode TSRKGetType_RK(TS ts,TSRKType *rktype)
1345: {
1346: TS_RK *rk = (TS_RK*)ts->data;
1349: *rktype = rk->tableau->name;
1350: return(0);
1351: }
1353: static PetscErrorCode TSRKSetType_RK(TS ts,TSRKType rktype)
1354: {
1355: TS_RK *rk = (TS_RK*)ts->data;
1357: PetscBool match;
1358: RKTableauLink link;
1361: if (rk->tableau) {
1362: PetscStrcmp(rk->tableau->name,rktype,&match);
1363: if (match) return(0);
1364: }
1365: for (link = RKTableauList; link; link=link->next) {
1366: PetscStrcmp(link->tab.name,rktype,&match);
1367: if (match) {
1368: if (ts->setupcalled) {TSRKTableauReset(ts);}
1369: rk->tableau = &link->tab;
1370: if (ts->setupcalled) {TSRKTableauSetUp(ts);}
1371: ts->default_adapt_type = rk->tableau->bembed ? TSADAPTBASIC : TSADAPTNONE;
1372: return(0);
1373: }
1374: }
1375: SETERRQ1(PetscObjectComm((PetscObject)ts),PETSC_ERR_ARG_UNKNOWN_TYPE,"Could not find '%s'",rktype);
1376: return(0);
1377: }
1379: static PetscErrorCode TSGetStages_RK(TS ts,PetscInt *ns,Vec **Y)
1380: {
1381: TS_RK *rk = (TS_RK*)ts->data;
1384: if (ns) *ns = rk->tableau->s;
1385: if (Y) *Y = rk->Y;
1386: return(0);
1387: }
1389: static PetscErrorCode TSDestroy_RK(TS ts)
1390: {
1394: TSReset_RK(ts);
1395: if (ts->dm) {
1396: DMCoarsenHookRemove(ts->dm,DMCoarsenHook_TSRK,DMRestrictHook_TSRK,ts);
1397: DMSubDomainHookRemove(ts->dm,DMSubDomainHook_TSRK,DMSubDomainRestrictHook_TSRK,ts);
1398: }
1399: PetscFree(ts->data);
1400: PetscObjectComposeFunction((PetscObject)ts,"TSRKGetOrder_C",NULL);
1401: PetscObjectComposeFunction((PetscObject)ts,"TSRKGetType_C",NULL);
1402: PetscObjectComposeFunction((PetscObject)ts,"TSRKSetType_C",NULL);
1403: PetscObjectComposeFunction((PetscObject)ts,"TSRKGetTableau_C",NULL);
1404: PetscObjectComposeFunction((PetscObject)ts,"TSRKSetMultirate_C",NULL);
1405: PetscObjectComposeFunction((PetscObject)ts,"TSRKGetMultirate_C",NULL);
1406: return(0);
1407: }
1409: /*
1410: This defines the nonlinear equation that is to be solved with SNES
1411: We do not need to solve the equation; we just use SNES to approximate the Jacobian
1412: */
1413: static PetscErrorCode SNESTSFormFunction_RK(SNES snes,Vec x,Vec y,TS ts)
1414: {
1415: TS_RK *rk = (TS_RK*)ts->data;
1416: DM dm,dmsave;
1420: SNESGetDM(snes,&dm);
1421: /* DM monkey-business allows user code to call TSGetDM() inside of functions evaluated on levels of FAS */
1422: dmsave = ts->dm;
1423: ts->dm = dm;
1424: TSComputeRHSFunction(ts,rk->stage_time,x,y);
1425: ts->dm = dmsave;
1426: return(0);
1427: }
1429: static PetscErrorCode SNESTSFormJacobian_RK(SNES snes,Vec x,Mat A,Mat B,TS ts)
1430: {
1431: TS_RK *rk = (TS_RK*)ts->data;
1432: DM dm,dmsave;
1436: SNESGetDM(snes,&dm);
1437: dmsave = ts->dm;
1438: ts->dm = dm;
1439: TSComputeRHSJacobian(ts,rk->stage_time,x,A,B);
1440: ts->dm = dmsave;
1441: return(0);
1442: }
1444: /*@C
1445: TSRKSetMultirate - Use the interpolation-based multirate RK method
1447: Logically collective
1449: Input Parameter:
1450: + ts - timestepping context
1451: - use_multirate - PETSC_TRUE enables the multirate RK method, sets the basic method to be RK2A and sets the ratio between slow stepsize and fast stepsize to be 2
1453: Options Database:
1454: . -ts_rk_multirate - <true,false>
1456: Notes:
1457: The multirate method requires interpolation. The default interpolation works for 1st- and 2nd- order RK, but not for high-order RKs except TSRK5DP which comes with the interpolation coeffcients (binterp).
1459: Level: intermediate
1461: .seealso: TSRKGetMultirate()
1462: @*/
1463: PetscErrorCode TSRKSetMultirate(TS ts,PetscBool use_multirate)
1464: {
1468: PetscTryMethod(ts,"TSRKSetMultirate_C",(TS,PetscBool),(ts,use_multirate));
1469: return(0);
1470: }
1472: /*@C
1473: TSRKGetMultirate - Gets whether to Use the interpolation-based multirate RK method
1475: Not collective
1477: Input Parameter:
1478: . ts - timestepping context
1480: Output Parameter:
1481: . use_multirate - PETSC_TRUE if the multirate RK method is enabled, PETSC_FALSE otherwise
1483: Level: intermediate
1485: .seealso: TSRKSetMultirate()
1486: @*/
1487: PetscErrorCode TSRKGetMultirate(TS ts,PetscBool *use_multirate)
1488: {
1492: PetscUseMethod(ts,"TSRKGetMultirate_C",(TS,PetscBool*),(ts,use_multirate));
1493: return(0);
1494: }
1496: /*MC
1497: TSRK - ODE and DAE solver using Runge-Kutta schemes
1499: The user should provide the right hand side of the equation
1500: using TSSetRHSFunction().
1502: Notes:
1503: The default is TSRK3BS, it can be changed with TSRKSetType() or -ts_rk_type
1505: Level: beginner
1507: .seealso: TSCreate(), TS, TSSetType(), TSRKSetType(), TSRKGetType(), TSRK2D, TTSRK2E, TSRK3,
1508: TSRK4, TSRK5, TSRKPRSSP2, TSRKBPR3, TSRKType, TSRKRegister(), TSRKSetMultirate(), TSRKGetMultirate()
1510: M*/
1511: PETSC_EXTERN PetscErrorCode TSCreate_RK(TS ts)
1512: {
1513: TS_RK *rk;
1517: TSRKInitializePackage();
1519: ts->ops->reset = TSReset_RK;
1520: ts->ops->destroy = TSDestroy_RK;
1521: ts->ops->view = TSView_RK;
1522: ts->ops->load = TSLoad_RK;
1523: ts->ops->setup = TSSetUp_RK;
1524: ts->ops->interpolate = TSInterpolate_RK;
1525: ts->ops->step = TSStep_RK;
1526: ts->ops->evaluatestep = TSEvaluateStep_RK;
1527: ts->ops->rollback = TSRollBack_RK;
1528: ts->ops->setfromoptions = TSSetFromOptions_RK;
1529: ts->ops->getstages = TSGetStages_RK;
1531: ts->ops->snesfunction = SNESTSFormFunction_RK;
1532: ts->ops->snesjacobian = SNESTSFormJacobian_RK;
1533: ts->ops->adjointintegral = TSAdjointCostIntegral_RK;
1534: ts->ops->adjointsetup = TSAdjointSetUp_RK;
1535: ts->ops->adjointstep = TSAdjointStep_RK;
1536: ts->ops->adjointreset = TSAdjointReset_RK;
1538: ts->ops->forwardintegral = TSForwardCostIntegral_RK;
1539: ts->ops->forwardsetup = TSForwardSetUp_RK;
1540: ts->ops->forwardreset = TSForwardReset_RK;
1541: ts->ops->forwardstep = TSForwardStep_RK;
1542: ts->ops->forwardgetstages= TSForwardGetStages_RK;
1544: PetscNewLog(ts,&rk);
1545: ts->data = (void*)rk;
1547: PetscObjectComposeFunction((PetscObject)ts,"TSRKGetOrder_C",TSRKGetOrder_RK);
1548: PetscObjectComposeFunction((PetscObject)ts,"TSRKGetType_C",TSRKGetType_RK);
1549: PetscObjectComposeFunction((PetscObject)ts,"TSRKSetType_C",TSRKSetType_RK);
1550: PetscObjectComposeFunction((PetscObject)ts,"TSRKGetTableau_C",TSRKGetTableau_RK);
1551: PetscObjectComposeFunction((PetscObject)ts,"TSRKSetMultirate_C",TSRKSetMultirate_RK);
1552: PetscObjectComposeFunction((PetscObject)ts,"TSRKGetMultirate_C",TSRKGetMultirate_RK);
1554: TSRKSetType(ts,TSRKDefault);
1555: rk->dtratio = 1;
1556: return(0);
1557: }