source: Sophya/trunk/SophyaProg/Tests/lpk.cc@ 3617

Last change on this file since 3617 was 3617, checked in by cmv, 16 years ago

chgt nom Gaussian -> GaussianRand etc.. pour ambiguite, cmv 02/05/2009

File size: 8.9 KB
RevLine 
[2615]1#include "sopnamsp.h"
[778]2#include "machdefs.h"
3
4#include <math.h>
[2322]5#include <iostream>
[778]6
[857]7#include "srandgen.h"
[778]8#include "tarrinit.h"
[857]9#include "array.h"
[778]10#include "timing.h"
11#include "intflapack.h"
12
13
[2647]14int lpk_linsolve_mtx(int n); // ErrorCode = 8
15int lpk_svd_mtx(int l, int c, int wsf, bool covu=true); // ErrorCode = 16
16int lpk_leastsquare_mtx(int l, int c); // ErrorCode = 32
17int lpk_inverse_mtx(int l); // ErrorCode = 64
[812]18
[1248]19static double TOLERANCE = 1.e-6;
[2647]20static int nprt = 100;
21static int prtlev = 1;
[1100]22
[778]23int main(int narg, char* arg[])
24{
25
26 SophyaInit();
27 InitTim(); // Initializing the CPU timer
28
29
[812]30
31 if (narg < 2) {
[1343]32 cout << " lpk - LinAlg/LapackServer test \n"
[2647]33 << " Usage: lpk select [sizeL,C=5,5] [prtlev=1] \n"
[1343]34 << " [nprtmax=100] [WorkSpaceSizeFactor=2] \n"
[2647]35 << " select= linsolve svd svds lss inverse / all= \n"
[1343]36 << " linsolve: lpk_linsolve_mtx() LapackServer::LinSolve with TMatrix<r_8> \n"
37 << " svd: lpk_svd_mtx() LapackServer::SVD(a,s,u,vt) with TMatrix<r_8> \n"
[1567]38 << " svds: lpk_svd_mtx() LapackServer::SVD(a,s); with TMatrix<r_8> \n"
39 << " lss: lpk_leastsquare_mtx() LapackServer::LeastSquareSolve() r_8 \n"
[2647]40 << " inverse: LapackServer::ComputeInverse() r_8 \n" << endl;
[812]41 exit(0);
42 }
[1343]43 int l,c;
44 l = c = 5;
45 int wsf = 2;
46 string opt = arg[1];
47 if (narg > 2) sscanf(arg[2], "%d,%d", &l, &c);
[812]48 if (narg > 3) prtlev = atoi(arg[3]);
49 if (narg > 4) nprt = atoi(arg[4]);
[1343]50 if (narg > 5) wsf = atoi(arg[5]);
[812]51
[2647]52 cout << " lpk - LinAlg/LapackServer_Test sizeL,C=" << l << "," << c
53 << " opt= " << opt << endl;
[1248]54 int rc = 0;
[812]55 BaseArray::SetMaxPrint(nprt, prtlev);
56 try {
[2647]57 if (opt == "all") {
58 rc += lpk_linsolve_mtx(l);
59 rc += lpk_svd_mtx(l,c,wsf,true);
60 rc += lpk_leastsquare_mtx(l,c);
61 rc += lpk_inverse_mtx(l);
62 }
63 else if (opt == "linsolve") rc = lpk_linsolve_mtx(l);
[1343]64 else if (opt == "svd") rc = lpk_svd_mtx(l,c,wsf,true);
65 else if (opt == "svds") rc = lpk_svd_mtx(l,c,wsf,false);
[1567]66 else if (opt == "lss") rc = lpk_leastsquare_mtx(l,c);
[2647]67 else if (opt == "inverse") rc = lpk_inverse_mtx(l);
[1343]68 else { cout << " Unknown option " << opt << " ! " << endl; rc = 66; }
[812]69 }
70 catch (PThrowable exc) {
71 cerr << " catched Exception (lpk.cc) " << exc.Msg() << endl;
[1248]72 rc = 77;
[812]73 }
74 catch (...) {
75 cerr << " catched unknown (...) exception (lpk.cc) " << endl;
[1248]76 rc = 78;
[812]77 }
78
79 PrtTim(" End of lpk LinAlg/Lapack test ");
[1343]80 cout << " --------------- END of Programme -------- (Rc= "
81 << rc << ") --- " << endl;
[1248]82 return(rc);
[812]83}
84
[1343]85// -----------------------------------------------------------------------
86/* Nouvelle-Fonction */
87int lpk_linsolve_mtx(int n)
[812]88{
[1343]89 int i,j;
[812]90 BaseArray::SetDefaultMemoryMapping(BaseArray::FortranMemoryMapping);
[1343]91 cout << " lpk_linsolve_mtx() - Test of LapackServer::LinSolve() " << endl;
[812]92 Matrix a(n,n);
93 for(i=0; i<n; i++)
[3617]94 for(j=0; j<n; j++) a(j,i) = GaussianRand(1.,0.);
[812]95
96 Vector x(n), b;
97 // Matrix x(n,1), b;
[2647]98 if (prtlev > 0)
99 cout << " ------------ Vector X = \n " << x << "\n" << endl;
[3617]100 for(i=0; i<n; i++) x(i) = GaussianRand(1.5,2.);
[812]101 b = a*x;
[778]102
[2647]103 if (prtlev > 0) {
104 cout << " ---- lpk_tmtx() LapackServer::LinSolve Test Using TMatrix<r_8> ----- " << endl;
105 cout << " ------------ Matrix A = \n " << a << "\n" << endl;
106 cout << " ------------ Matrix X = \n " << x << "\n" << endl;
107 cout << " ------------ Matrix B = \n " << b << "\n" << endl;
108 }
[812]109 cout << "\n Calling LapackLinSolve(a,b) .... " << endl;
[1100]110 PrtTim(" Calling LapackLinSolve(a,b) ");
[812]111 LapackLinSolve(a,b);
[1100]112 PrtTim(" End LapackLinSolve(a,b) ");
[812]113
[2647]114 if (prtlev > 0)
115 cout << " ------------ Result B(=X ?) = \n " << b << "\n" << endl;
[812]116 Vector diff = b-x;
[1343]117 PrtTim(" End of Compute(diff)");
[2647]118 if (prtlev > 0)
119 cout << " ------------ Vector diff B-X = \n " << diff << "\n" << endl;
[1248]120 double min,max;
121 diff.MinMax(min, max);
122 cout << " Min/Max difference Vector (?=0) , Min= " << min
123 << " Max= " << max << endl;
124 if ((fabs(min) > TOLERANCE) || (fabs(max) > TOLERANCE)) {
125 cout << " !!! Difference exceeding tolerance (=" << TOLERANCE << ") !!!"
126 << endl;
[2647]127 return(8);
[1248]128 }
129 return(0);
[812]130
131}
132
[1343]133// -----------------------------------------------------------------------
134/* Nouvelle-Fonction */
135int lpk_svd_mtx(int m, int n, int wsf, bool covu)
[812]136{
[1343]137 BaseArray::SetDefaultMemoryMapping(BaseArray::FortranMemoryMapping);
138 cout << " lpk_svd_mtx() - Test of LapackServer::SVD " << endl;
139
140 Matrix a(m , n), aa;
141 a = RandomSequence(RandomSequence::Gaussian, 0., 4.);
142 aa = a;
143 Vector s;
144 Matrix u, vt;
145 cout << " ---- lpk_svd_tmtx() LapackServer::SVD Test Using TMatrix<r_8> ---- " << endl;
[2647]146 if (prtlev > 0)
147 cout << " ------------ Matrix A = \n " << a << "\n" << endl;
[1343]148
149 cout << "\n Calling LapackSVD(a,s,u,vt) .... " << endl;
150 PrtTim(" Calling LapackSVD() ");
151 LapackServer<r_8> lpks;
152 lpks.SetWorkSpaceSizeFactor(wsf);
153 if (covu) lpks.SVD(aa, s, u, vt);
154 else lpks.SVD(aa, s);
155 PrtTim(" End LapackSVD() ");
156
[2647]157 if (prtlev > 0)
158 cout << " ------------ Result S = \n " << s << "\n" << endl;
[1343]159 if (!covu) return(0);
[2647]160 if (prtlev > 0) {
161 cout << " ------------ Result U = \n " << u << "\n" << endl;
162 cout << " ------------ Result VT = \n " << vt << "\n" << endl;
163 }
[1343]164 Matrix sm(m,n);
165 int minmn = (m<n) ? m : n ;
166 for(int k=0; k< minmn ; k++) sm(k,k) = s(k);
167 Matrix diff = u*(sm*vt) - a;
168 PrtTim(" End of Compute(diff)");
[2647]169 if (prtlev > 0)
170 cout << " ------------ Matrix diff U*S*Vt - A = \n " << diff << "\n" << endl;
[1343]171 double min,max;
172 diff.MinMax(min, max);
173 cout << " Min/Max difference Matrix (?=0) , Min= " << min
174 << " Max= " << max << endl;
175 if ((fabs(min) > TOLERANCE) || (fabs(max) > TOLERANCE)) {
176 cout << " !!! Difference exceeding tolerance (=" << TOLERANCE << ") !!!"
177 << endl;
[2647]178 return(16);
[1343]179 }
180 return(0);
181
182}
183
[1567]184// -----------------------------------------------------------------------
185/* Nouvelle-Fonction */
186int lpk_leastsquare_mtx(int m, int n)
187{
188 BaseArray::SetDefaultMemoryMapping(BaseArray::FortranMemoryMapping);
189 cout << " lpk_leastsquare_mtx() - Test of LapackLeastSquareSolve " << endl;
190
191 Matrix a(m , n), aa;
192 a = RandomSequence(RandomSequence::Gaussian, 0., 4.);
193 aa = a;
194 Vector x(n),noise(m),b,bx;
195 x = RandomSequence(RandomSequence::Gaussian, 2., 1.);
196 double signoise = 0.1;
197 noise = RandomSequence(RandomSequence::Gaussian, 0., 0.1);
198 b = a*x;
199 bx = b+noise;
200
201 cout << " ---- lpk_leastsquare_tmtx() LapackLeastSquareSolve<r_8> ---- " << endl;
[2647]202 if (prtlev > 0) {
203 cout << " ------------ Matrix A = \n " << a << "\n" << endl;
204 cout << " ------------ Vector B = \n " << bx << "\n" << endl;
205 cout << " ------------ Vector X = \n " << x << "\n" << endl;
206 }
[1567]207 cout << "\n Calling LapackLeastSquareSolve(a,bx) .... " << endl;
208 PrtTim(" Calling LapackLeastSquareSolve() ");
209 LapackLeastSquareSolve(aa, bx);
210 PrtTim(" End LapackLeastSquareSolve() ");
211 bx.Share(bx(Range(0,0,n)));
[2647]212 if (prtlev > 0) {
213 cout << " ------------ Result X = \n " << bx << "\n" << endl;
214 cout << " ------------ X-X_Real = \n " << bx-x << "\n" << endl;
215 }
[1567]216 Vector diff = b-a*bx;
217 PrtTim(" End of Compute(diff)");
[2647]218 if (prtlev > 0)
219 cout << " ------------ Matrix diff b-a*x = \n " << diff << "\n" << endl;
[1567]220 double mean,sigma;
221 MeanSigma(diff, mean, sigma);
222 cout << " MeanSigma(diff) , Mean= " << mean << " Sigma=" << sigma
223 << " SigNoise= " << signoise << endl;
224 if ((fabs(mean) > signoise) || (fabs(sigma-signoise) > signoise)) {
225 cout << " !!! Difference exceeding tolerance (=" << TOLERANCE << ") !!!"
226 << endl;
[2647]227 return 32;
[1567]228 }
[2647]229 return 0;
[1567]230
231}
[2647]232// -----------------------------------------------------------------------
233/* Nouvelle-Fonction */
234int lpk_inverse_mtx(int m)
[1343]235{
[2647]236 BaseArray::SetDefaultMemoryMapping(BaseArray::FortranMemoryMapping);
237 cout << " lpk_inverse_mtx() - Test of LapackComputeInverse<r_8> " << endl;
[778]238
[2647]239 Matrix a(m , m), aa, ainv, idmx(m,m), mxprod, diff;
240 a = RandomSequence(RandomSequence::Gaussian, 0., 4.);
241 aa = a; // We make a copy of a, as it is modified by LapackServer
242 PrtTim(" End of a_inint");
[778]243
[2647]244 cout << "\n Calling LapackInverse() .... " << endl;
245 ainv = LapackInverse(aa);
246 PrtTim(" End of LapackInverse(a)");
247 mxprod = a*ainv;
248 PrtTim(" End of mxprod = a*ainv");
249 idmx = IdentityMatrix();
250 PrtTim(" End of idmx = IdentityMatrix()");
251 diff = mxprod-idmx;
252 PrtTim(" End of Compute(diff)");
[778]253
[2647]254 if (prtlev > 0) {
255 cout << " ------------ Matrix A = \n " << a << "\n" << endl;
256 cout << " ------------ Matrix Inverse(A) = \n " << ainv << "\n" << endl;
257 cout << " ------------ Matrix mxprod = A*Inverse(A) = \n " << mxprod << "\n" << endl;
258 }
259 double min,max;
260 diff.MinMax(min, max);
261 PrtTim(" End of diffCheck");
262 cout << " Min/Max difference Matrix (?=0) , Min= " << min
263 << " Max= " << max << endl;
264 if ((fabs(min) > TOLERANCE) || (fabs(max) > TOLERANCE)) {
265 cout << " !!! Difference exceeding tolerance (=" << TOLERANCE << ") !!!"
266 << endl;
267 return 64;
268 }
269 return 0;
[778]270
[2647]271}
[778]272
Note: See TracBrowser for help on using the repository browser.