You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

dget53.f 7.2 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254
  1. *> \brief \b DGET53
  2. *
  3. * =========== DOCUMENTATION ===========
  4. *
  5. * Online html documentation available at
  6. * http://www.netlib.org/lapack/explore-html/
  7. *
  8. * Definition:
  9. * ===========
  10. *
  11. * SUBROUTINE DGET53( A, LDA, B, LDB, SCALE, WR, WI, RESULT, INFO )
  12. *
  13. * .. Scalar Arguments ..
  14. * INTEGER INFO, LDA, LDB
  15. * DOUBLE PRECISION RESULT, SCALE, WI, WR
  16. * ..
  17. * .. Array Arguments ..
  18. * DOUBLE PRECISION A( LDA, * ), B( LDB, * )
  19. * ..
  20. *
  21. *
  22. *> \par Purpose:
  23. * =============
  24. *>
  25. *> \verbatim
  26. *>
  27. *> DGET53 checks the generalized eigenvalues computed by DLAG2.
  28. *>
  29. *> The basic test for an eigenvalue is:
  30. *>
  31. *> | det( s A - w B ) |
  32. *> RESULT = ---------------------------------------------------
  33. *> ulp max( s norm(A), |w| norm(B) )*norm( s A - w B )
  34. *>
  35. *> Two "safety checks" are performed:
  36. *>
  37. *> (1) ulp*max( s*norm(A), |w|*norm(B) ) must be at least
  38. *> safe_minimum. This insures that the test performed is
  39. *> not essentially det(0*A + 0*B)=0.
  40. *>
  41. *> (2) s*norm(A) + |w|*norm(B) must be less than 1/safe_minimum.
  42. *> This insures that s*A - w*B will not overflow.
  43. *>
  44. *> If these tests are not passed, then s and w are scaled and
  45. *> tested anyway, if this is possible.
  46. *> \endverbatim
  47. *
  48. * Arguments:
  49. * ==========
  50. *
  51. *> \param[in] A
  52. *> \verbatim
  53. *> A is DOUBLE PRECISION array, dimension (LDA, 2)
  54. *> The 2x2 matrix A.
  55. *> \endverbatim
  56. *>
  57. *> \param[in] LDA
  58. *> \verbatim
  59. *> LDA is INTEGER
  60. *> The leading dimension of A. It must be at least 2.
  61. *> \endverbatim
  62. *>
  63. *> \param[in] B
  64. *> \verbatim
  65. *> B is DOUBLE PRECISION array, dimension (LDB, N)
  66. *> The 2x2 upper-triangular matrix B.
  67. *> \endverbatim
  68. *>
  69. *> \param[in] LDB
  70. *> \verbatim
  71. *> LDB is INTEGER
  72. *> The leading dimension of B. It must be at least 2.
  73. *> \endverbatim
  74. *>
  75. *> \param[in] SCALE
  76. *> \verbatim
  77. *> SCALE is DOUBLE PRECISION
  78. *> The "scale factor" s in the formula s A - w B . It is
  79. *> assumed to be non-negative.
  80. *> \endverbatim
  81. *>
  82. *> \param[in] WR
  83. *> \verbatim
  84. *> WR is DOUBLE PRECISION
  85. *> The real part of the eigenvalue w in the formula
  86. *> s A - w B .
  87. *> \endverbatim
  88. *>
  89. *> \param[in] WI
  90. *> \verbatim
  91. *> WI is DOUBLE PRECISION
  92. *> The imaginary part of the eigenvalue w in the formula
  93. *> s A - w B .
  94. *> \endverbatim
  95. *>
  96. *> \param[out] RESULT
  97. *> \verbatim
  98. *> RESULT is DOUBLE PRECISION
  99. *> If INFO is 2 or less, the value computed by the test
  100. *> described above.
  101. *> If INFO=3, this will just be 1/ulp.
  102. *> \endverbatim
  103. *>
  104. *> \param[out] INFO
  105. *> \verbatim
  106. *> INFO is INTEGER
  107. *> =0: The input data pass the "safety checks".
  108. *> =1: s*norm(A) + |w|*norm(B) > 1/safe_minimum.
  109. *> =2: ulp*max( s*norm(A), |w|*norm(B) ) < safe_minimum
  110. *> =3: same as INFO=2, but s and w could not be scaled so
  111. *> as to compute the test.
  112. *> \endverbatim
  113. *
  114. * Authors:
  115. * ========
  116. *
  117. *> \author Univ. of Tennessee
  118. *> \author Univ. of California Berkeley
  119. *> \author Univ. of Colorado Denver
  120. *> \author NAG Ltd.
  121. *
  122. *> \date November 2011
  123. *
  124. *> \ingroup double_eig
  125. *
  126. * =====================================================================
  127. SUBROUTINE DGET53( A, LDA, B, LDB, SCALE, WR, WI, RESULT, INFO )
  128. *
  129. * -- LAPACK test routine (version 3.4.0) --
  130. * -- LAPACK is a software package provided by Univ. of Tennessee, --
  131. * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
  132. * November 2011
  133. *
  134. * .. Scalar Arguments ..
  135. INTEGER INFO, LDA, LDB
  136. DOUBLE PRECISION RESULT, SCALE, WI, WR
  137. * ..
  138. * .. Array Arguments ..
  139. DOUBLE PRECISION A( LDA, * ), B( LDB, * )
  140. * ..
  141. *
  142. * =====================================================================
  143. *
  144. * .. Parameters ..
  145. DOUBLE PRECISION ZERO, ONE
  146. PARAMETER ( ZERO = 0.0D0, ONE = 1.0D0 )
  147. * ..
  148. * .. Local Scalars ..
  149. DOUBLE PRECISION ABSW, ANORM, BNORM, CI11, CI12, CI22, CNORM,
  150. $ CR11, CR12, CR21, CR22, CSCALE, DETI, DETR, S1,
  151. $ SAFMIN, SCALES, SIGMIN, TEMP, ULP, WIS, WRS
  152. * ..
  153. * .. External Functions ..
  154. DOUBLE PRECISION DLAMCH
  155. EXTERNAL DLAMCH
  156. * ..
  157. * .. Intrinsic Functions ..
  158. INTRINSIC ABS, MAX, SQRT
  159. * ..
  160. * .. Executable Statements ..
  161. *
  162. * Initialize
  163. *
  164. INFO = 0
  165. RESULT = ZERO
  166. SCALES = SCALE
  167. WRS = WR
  168. WIS = WI
  169. *
  170. * Machine constants and norms
  171. *
  172. SAFMIN = DLAMCH( 'Safe minimum' )
  173. ULP = DLAMCH( 'Epsilon' )*DLAMCH( 'Base' )
  174. ABSW = ABS( WRS ) + ABS( WIS )
  175. ANORM = MAX( ABS( A( 1, 1 ) )+ABS( A( 2, 1 ) ),
  176. $ ABS( A( 1, 2 ) )+ABS( A( 2, 2 ) ), SAFMIN )
  177. BNORM = MAX( ABS( B( 1, 1 ) ), ABS( B( 1, 2 ) )+ABS( B( 2, 2 ) ),
  178. $ SAFMIN )
  179. *
  180. * Check for possible overflow.
  181. *
  182. TEMP = ( SAFMIN*BNORM )*ABSW + ( SAFMIN*ANORM )*SCALES
  183. IF( TEMP.GE.ONE ) THEN
  184. *
  185. * Scale down to avoid overflow
  186. *
  187. INFO = 1
  188. TEMP = ONE / TEMP
  189. SCALES = SCALES*TEMP
  190. WRS = WRS*TEMP
  191. WIS = WIS*TEMP
  192. ABSW = ABS( WRS ) + ABS( WIS )
  193. END IF
  194. S1 = MAX( ULP*MAX( SCALES*ANORM, ABSW*BNORM ),
  195. $ SAFMIN*MAX( SCALES, ABSW ) )
  196. *
  197. * Check for W and SCALE essentially zero.
  198. *
  199. IF( S1.LT.SAFMIN ) THEN
  200. INFO = 2
  201. IF( SCALES.LT.SAFMIN .AND. ABSW.LT.SAFMIN ) THEN
  202. INFO = 3
  203. RESULT = ONE / ULP
  204. RETURN
  205. END IF
  206. *
  207. * Scale up to avoid underflow
  208. *
  209. TEMP = ONE / MAX( SCALES*ANORM+ABSW*BNORM, SAFMIN )
  210. SCALES = SCALES*TEMP
  211. WRS = WRS*TEMP
  212. WIS = WIS*TEMP
  213. ABSW = ABS( WRS ) + ABS( WIS )
  214. S1 = MAX( ULP*MAX( SCALES*ANORM, ABSW*BNORM ),
  215. $ SAFMIN*MAX( SCALES, ABSW ) )
  216. IF( S1.LT.SAFMIN ) THEN
  217. INFO = 3
  218. RESULT = ONE / ULP
  219. RETURN
  220. END IF
  221. END IF
  222. *
  223. * Compute C = s A - w B
  224. *
  225. CR11 = SCALES*A( 1, 1 ) - WRS*B( 1, 1 )
  226. CI11 = -WIS*B( 1, 1 )
  227. CR21 = SCALES*A( 2, 1 )
  228. CR12 = SCALES*A( 1, 2 ) - WRS*B( 1, 2 )
  229. CI12 = -WIS*B( 1, 2 )
  230. CR22 = SCALES*A( 2, 2 ) - WRS*B( 2, 2 )
  231. CI22 = -WIS*B( 2, 2 )
  232. *
  233. * Compute the smallest singular value of s A - w B:
  234. *
  235. * |det( s A - w B )|
  236. * sigma_min = ------------------
  237. * norm( s A - w B )
  238. *
  239. CNORM = MAX( ABS( CR11 )+ABS( CI11 )+ABS( CR21 ),
  240. $ ABS( CR12 )+ABS( CI12 )+ABS( CR22 )+ABS( CI22 ), SAFMIN )
  241. CSCALE = ONE / SQRT( CNORM )
  242. DETR = ( CSCALE*CR11 )*( CSCALE*CR22 ) -
  243. $ ( CSCALE*CI11 )*( CSCALE*CI22 ) -
  244. $ ( CSCALE*CR12 )*( CSCALE*CR21 )
  245. DETI = ( CSCALE*CR11 )*( CSCALE*CI22 ) +
  246. $ ( CSCALE*CI11 )*( CSCALE*CR22 ) -
  247. $ ( CSCALE*CI12 )*( CSCALE*CR21 )
  248. SIGMIN = ABS( DETR ) + ABS( DETI )
  249. RESULT = SIGMIN / S1
  250. RETURN
  251. *
  252. * End of DGET53
  253. *
  254. END