Actual source code: test11.c
slepc-3.21.1 2024-04-26
1: /*
2: - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
3: SLEPc - Scalable Library for Eigenvalue Problem Computations
4: Copyright (c) 2002-, Universitat Politecnica de Valencia, Spain
6: This file is part of SLEPc.
7: SLEPc is distributed under a 2-clause BSD license (see LICENSE).
8: - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
9: */
10: /*
11: Define the function
13: f(x) = (exp(x)-1)/x (the phi_1 function)
15: with the following tree:
17: f(x) f(x) (combined by division)
18: / \ p(x) = x (polynomial)
19: a(x) p(x) a(x) (combined by addition)
20: / \ e(x) = exp(x) (exponential)
21: e(x) c(x) c(x) = -1 (constant)
22: */
24: static char help[] = "Another test of a combined function.\n\n";
26: #include <slepcfn.h>
28: /*
29: Compute matrix function B = A\(exp(A)-I)
30: */
31: PetscErrorCode TestMatCombine(FN fn,Mat A,PetscViewer viewer,PetscBool verbose,PetscBool inplace)
32: {
33: PetscBool set,flg;
34: PetscInt n;
35: Mat F,Acopy;
36: Vec v,f0;
37: PetscReal nrm;
39: PetscFunctionBeginUser;
40: PetscCall(MatGetSize(A,&n,NULL));
41: PetscCall(MatDuplicate(A,MAT_DO_NOT_COPY_VALUES,&F));
42: PetscCall(PetscObjectSetName((PetscObject)F,"F"));
43: /* compute matrix function */
44: if (inplace) {
45: PetscCall(MatCopy(A,F,SAME_NONZERO_PATTERN));
46: PetscCall(MatIsHermitianKnown(A,&set,&flg));
47: if (set && flg) PetscCall(MatSetOption(F,MAT_HERMITIAN,PETSC_TRUE));
48: PetscCall(FNEvaluateFunctionMat(fn,F,NULL));
49: } else {
50: PetscCall(MatDuplicate(A,MAT_COPY_VALUES,&Acopy));
51: PetscCall(FNEvaluateFunctionMat(fn,A,F));
52: /* check that A has not been modified */
53: PetscCall(MatAXPY(Acopy,-1.0,A,SAME_NONZERO_PATTERN));
54: PetscCall(MatNorm(Acopy,NORM_1,&nrm));
55: if (nrm>100*PETSC_MACHINE_EPSILON) PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Warning: the input matrix has changed by %g\n",(double)nrm));
56: PetscCall(MatDestroy(&Acopy));
57: }
58: if (verbose) {
59: PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Matrix A - - - - - - - -\n"));
60: PetscCall(MatView(A,viewer));
61: PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Computed f(A) - - - - - - -\n"));
62: PetscCall(MatView(F,viewer));
63: }
64: /* print matrix norm for checking */
65: PetscCall(MatNorm(F,NORM_1,&nrm));
66: PetscCall(PetscPrintf(PETSC_COMM_WORLD,"The 1-norm of f(A) is %6.3f\n",(double)nrm));
67: /* check FNEvaluateFunctionMatVec() */
68: PetscCall(MatCreateVecs(A,&v,&f0));
69: PetscCall(MatGetColumnVector(F,f0,0));
70: PetscCall(FNEvaluateFunctionMatVec(fn,A,v));
71: PetscCall(VecAXPY(v,-1.0,f0));
72: PetscCall(VecNorm(v,NORM_2,&nrm));
73: if (nrm>100*PETSC_MACHINE_EPSILON) PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Warning: the norm of f(A)*e_1-v is %g\n",(double)nrm));
74: PetscCall(MatDestroy(&F));
75: PetscCall(VecDestroy(&v));
76: PetscCall(VecDestroy(&f0));
77: PetscFunctionReturn(PETSC_SUCCESS);
78: }
80: int main(int argc,char **argv)
81: {
82: FN f,p,a,e,c,f1,f2;
83: FNCombineType ctype;
84: Mat A=NULL;
85: PetscInt i,j,n=10,np;
86: PetscScalar x,y,yp,*As,coeffs[10];
87: char strx[50],str[50];
88: PetscViewer viewer;
89: PetscBool verbose,inplace,matcuda;
91: PetscFunctionBeginUser;
92: PetscCall(SlepcInitialize(&argc,&argv,(char*)0,help));
93: PetscCall(PetscOptionsGetInt(NULL,NULL,"-n",&n,NULL));
94: PetscCall(PetscOptionsHasName(NULL,NULL,"-verbose",&verbose));
95: PetscCall(PetscOptionsHasName(NULL,NULL,"-inplace",&inplace));
96: PetscCall(PetscOptionsHasName(NULL,NULL,"-matcuda",&matcuda));
97: PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Phi1 via a combined function, n=%" PetscInt_FMT ".\n",n));
99: /* Create function */
101: /* e(x) = exp(x) */
102: PetscCall(FNCreate(PETSC_COMM_WORLD,&e));
103: PetscCall(PetscObjectSetName((PetscObject)e,"e"));
104: PetscCall(FNSetType(e,FNEXP));
105: PetscCall(FNSetFromOptions(e));
106: /* c(x) = -1 */
107: PetscCall(FNCreate(PETSC_COMM_WORLD,&c));
108: PetscCall(PetscObjectSetName((PetscObject)c,"c"));
109: PetscCall(FNSetType(c,FNRATIONAL));
110: PetscCall(FNSetFromOptions(c));
111: np = 1;
112: coeffs[0] = -1.0;
113: PetscCall(FNRationalSetNumerator(c,np,coeffs));
114: /* a(x) */
115: PetscCall(FNCreate(PETSC_COMM_WORLD,&a));
116: PetscCall(PetscObjectSetName((PetscObject)a,"a"));
117: PetscCall(FNSetType(a,FNCOMBINE));
118: PetscCall(FNSetFromOptions(a));
119: PetscCall(FNCombineSetChildren(a,FN_COMBINE_ADD,e,c));
120: /* p(x) = x */
121: PetscCall(FNCreate(PETSC_COMM_WORLD,&p));
122: PetscCall(PetscObjectSetName((PetscObject)p,"p"));
123: PetscCall(FNSetType(p,FNRATIONAL));
124: PetscCall(FNSetFromOptions(p));
125: np = 2;
126: coeffs[0] = 1.0; coeffs[1] = 0.0;
127: PetscCall(FNRationalSetNumerator(p,np,coeffs));
128: /* f(x) */
129: PetscCall(FNCreate(PETSC_COMM_WORLD,&f));
130: PetscCall(PetscObjectSetName((PetscObject)f,"f"));
131: PetscCall(FNSetType(f,FNCOMBINE));
132: PetscCall(FNSetFromOptions(f));
133: PetscCall(FNCombineSetChildren(f,FN_COMBINE_DIVIDE,a,p));
135: /* Set up viewer */
136: PetscCall(PetscViewerASCIIGetStdout(PETSC_COMM_WORLD,&viewer));
137: PetscCall(FNCombineGetChildren(f,&ctype,&f1,&f2));
138: PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Two functions combined with division:\n"));
139: PetscCall(FNView(f1,viewer));
140: PetscCall(FNView(f2,viewer));
141: if (verbose) PetscCall(PetscViewerPushFormat(viewer,PETSC_VIEWER_ASCII_MATLAB));
143: /* Scalar evaluation */
144: x = 2.2;
145: PetscCall(SlepcSNPrintfScalar(strx,sizeof(strx),x,PETSC_FALSE));
146: PetscCall(FNEvaluateFunction(f,x,&y));
147: PetscCall(FNEvaluateDerivative(f,x,&yp));
148: PetscCall(SlepcSNPrintfScalar(str,sizeof(str),y,PETSC_FALSE));
149: PetscCall(PetscPrintf(PETSC_COMM_WORLD," f(%s)=%s\n",strx,str));
150: PetscCall(SlepcSNPrintfScalar(str,sizeof(str),yp,PETSC_FALSE));
151: PetscCall(PetscPrintf(PETSC_COMM_WORLD," f'(%s)=%s\n",strx,str));
153: /* Create matrices */
154: if (matcuda) {
155: #if defined(PETSC_HAVE_CUDA)
156: PetscCall(MatCreateSeqDenseCUDA(PETSC_COMM_SELF,n,n,NULL,&A));
157: #endif
158: } else PetscCall(MatCreateSeqDense(PETSC_COMM_SELF,n,n,NULL,&A));
159: PetscCall(PetscObjectSetName((PetscObject)A,"A"));
161: /* Fill A with 1-D Laplacian matrix */
162: PetscCall(MatDenseGetArray(A,&As));
163: for (i=0;i<n;i++) As[i+i*n]=2.0;
164: j=1;
165: for (i=0;i<n-j;i++) { As[i+(i+j)*n]=-1.0; As[(i+j)+i*n]=-1.0; }
166: PetscCall(MatDenseRestoreArray(A,&As));
167: PetscCall(MatSetOption(A,MAT_HERMITIAN,PETSC_TRUE));
168: PetscCall(TestMatCombine(f,A,viewer,verbose,inplace));
170: /* Repeat with same matrix as non-symmetric */
171: PetscCall(MatSetOption(A,MAT_HERMITIAN,PETSC_FALSE));
172: PetscCall(TestMatCombine(f,A,viewer,verbose,inplace));
174: PetscCall(MatDestroy(&A));
175: PetscCall(FNDestroy(&f));
176: PetscCall(FNDestroy(&p));
177: PetscCall(FNDestroy(&a));
178: PetscCall(FNDestroy(&e));
179: PetscCall(FNDestroy(&c));
180: PetscCall(SlepcFinalize());
181: return 0;
182: }
184: /*TEST
186: testset:
187: output_file: output/test11_1.out
188: test:
189: suffix: 1
190: test:
191: suffix: 1_cuda
192: args: -matcuda
193: requires: cuda
194: test:
195: suffix: 2
196: args: -inplace
197: test:
198: suffix: 2_cuda
199: args: -inplace -matcuda
200: requires: cuda
202: TEST*/