sgbfa.f 5.8 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187
  1. *DECK SGBFA
  2. SUBROUTINE SGBFA (ABD, LDA, N, ML, MU, IPVT, INFO)
  3. C***BEGIN PROLOGUE SGBFA
  4. C***PURPOSE Factor a band matrix using Gaussian elimination.
  5. C***LIBRARY SLATEC (LINPACK)
  6. C***CATEGORY D2A2
  7. C***TYPE SINGLE PRECISION (SGBFA-S, DGBFA-D, CGBFA-C)
  8. C***KEYWORDS BANDED, LINEAR ALGEBRA, LINPACK, MATRIX FACTORIZATION
  9. C***AUTHOR Moler, C. B., (U. of New Mexico)
  10. C***DESCRIPTION
  11. C
  12. C SGBFA factors a real band matrix by elimination.
  13. C
  14. C SGBFA is usually called by SBGCO, but it can be called
  15. C directly with a saving in time if RCOND is not needed.
  16. C
  17. C On Entry
  18. C
  19. C ABD REAL(LDA, N)
  20. C contains the matrix in band storage. The columns
  21. C of the matrix are stored in the columns of ABD and
  22. C the diagonals of the matrix are stored in rows
  23. C ML+1 through 2*ML+MU+1 of ABD .
  24. C See the comments below for details.
  25. C
  26. C LDA INTEGER
  27. C the leading dimension of the array ABD .
  28. C LDA must be .GE. 2*ML + MU + 1 .
  29. C
  30. C N INTEGER
  31. C the order of the original matrix.
  32. C
  33. C ML INTEGER
  34. C number of diagonals below the main diagonal.
  35. C 0 .LE. ML .LT. N .
  36. C
  37. C MU INTEGER
  38. C number of diagonals above the main diagonal.
  39. C 0 .LE. MU .LT. N .
  40. C More efficient if ML .LE. MU .
  41. C On Return
  42. C
  43. C ABD an upper triangular matrix in band storage and
  44. C the multipliers which were used to obtain it.
  45. C The factorization can be written A = L*U , where
  46. C L is a product of permutation and unit lower
  47. C triangular matrices and U is upper triangular.
  48. C
  49. C IPVT INTEGER(N)
  50. C an integer vector of pivot indices.
  51. C
  52. C INFO INTEGER
  53. C = 0 normal value.
  54. C = K if U(K,K) .EQ. 0.0 . This is not an error
  55. C condition for this subroutine, but it does
  56. C indicate that SGBSL will divide by zero if
  57. C called. Use RCOND in SBGCO for a reliable
  58. C indication of singularity.
  59. C
  60. C Band Storage
  61. C
  62. C If A is a band matrix, the following program segment
  63. C will set up the input.
  64. C
  65. C ML = (band width below the diagonal)
  66. C MU = (band width above the diagonal)
  67. C M = ML + MU + 1
  68. C DO 20 J = 1, N
  69. C I1 = MAX(1, J-MU)
  70. C I2 = MIN(N, J+ML)
  71. C DO 10 I = I1, I2
  72. C K = I - J + M
  73. C ABD(K,J) = A(I,J)
  74. C 10 CONTINUE
  75. C 20 CONTINUE
  76. C
  77. C This uses rows ML+1 through 2*ML+MU+1 of ABD .
  78. C In addition, the first ML rows in ABD are used for
  79. C elements generated during the triangularization.
  80. C The total number of rows needed in ABD is 2*ML+MU+1 .
  81. C The ML+MU by ML+MU upper left triangle and the
  82. C ML by ML lower right triangle are not referenced.
  83. C
  84. C***REFERENCES J. J. Dongarra, J. R. Bunch, C. B. Moler, and G. W.
  85. C Stewart, LINPACK Users' Guide, SIAM, 1979.
  86. C***ROUTINES CALLED ISAMAX, SAXPY, SSCAL
  87. C***REVISION HISTORY (YYMMDD)
  88. C 780814 DATE WRITTEN
  89. C 890531 Changed all specific intrinsics to generic. (WRB)
  90. C 890831 Modified array declarations. (WRB)
  91. C 890831 REVISION DATE from Version 3.2
  92. C 891214 Prologue converted to Version 4.0 format. (BAB)
  93. C 900326 Removed duplicate information from DESCRIPTION section.
  94. C (WRB)
  95. C 920501 Reformatted the REFERENCES section. (WRB)
  96. C***END PROLOGUE SGBFA
  97. INTEGER LDA,N,ML,MU,IPVT(*),INFO
  98. REAL ABD(LDA,*)
  99. C
  100. REAL T
  101. INTEGER I,ISAMAX,I0,J,JU,JZ,J0,J1,K,KP1,L,LM,M,MM,NM1
  102. C
  103. C***FIRST EXECUTABLE STATEMENT SGBFA
  104. M = ML + MU + 1
  105. INFO = 0
  106. C
  107. C ZERO INITIAL FILL-IN COLUMNS
  108. C
  109. J0 = MU + 2
  110. J1 = MIN(N,M) - 1
  111. IF (J1 .LT. J0) GO TO 30
  112. DO 20 JZ = J0, J1
  113. I0 = M + 1 - JZ
  114. DO 10 I = I0, ML
  115. ABD(I,JZ) = 0.0E0
  116. 10 CONTINUE
  117. 20 CONTINUE
  118. 30 CONTINUE
  119. JZ = J1
  120. JU = 0
  121. C
  122. C GAUSSIAN ELIMINATION WITH PARTIAL PIVOTING
  123. C
  124. NM1 = N - 1
  125. IF (NM1 .LT. 1) GO TO 130
  126. DO 120 K = 1, NM1
  127. KP1 = K + 1
  128. C
  129. C ZERO NEXT FILL-IN COLUMN
  130. C
  131. JZ = JZ + 1
  132. IF (JZ .GT. N) GO TO 50
  133. IF (ML .LT. 1) GO TO 50
  134. DO 40 I = 1, ML
  135. ABD(I,JZ) = 0.0E0
  136. 40 CONTINUE
  137. 50 CONTINUE
  138. C
  139. C FIND L = PIVOT INDEX
  140. C
  141. LM = MIN(ML,N-K)
  142. L = ISAMAX(LM+1,ABD(M,K),1) + M - 1
  143. IPVT(K) = L + K - M
  144. C
  145. C ZERO PIVOT IMPLIES THIS COLUMN ALREADY TRIANGULARIZED
  146. C
  147. IF (ABD(L,K) .EQ. 0.0E0) GO TO 100
  148. C
  149. C INTERCHANGE IF NECESSARY
  150. C
  151. IF (L .EQ. M) GO TO 60
  152. T = ABD(L,K)
  153. ABD(L,K) = ABD(M,K)
  154. ABD(M,K) = T
  155. 60 CONTINUE
  156. C
  157. C COMPUTE MULTIPLIERS
  158. C
  159. T = -1.0E0/ABD(M,K)
  160. CALL SSCAL(LM,T,ABD(M+1,K),1)
  161. C
  162. C ROW ELIMINATION WITH COLUMN INDEXING
  163. C
  164. JU = MIN(MAX(JU,MU+IPVT(K)),N)
  165. MM = M
  166. IF (JU .LT. KP1) GO TO 90
  167. DO 80 J = KP1, JU
  168. L = L - 1
  169. MM = MM - 1
  170. T = ABD(L,J)
  171. IF (L .EQ. MM) GO TO 70
  172. ABD(L,J) = ABD(MM,J)
  173. ABD(MM,J) = T
  174. 70 CONTINUE
  175. CALL SAXPY(LM,T,ABD(M+1,K),1,ABD(MM+1,J),1)
  176. 80 CONTINUE
  177. 90 CONTINUE
  178. GO TO 110
  179. 100 CONTINUE
  180. INFO = K
  181. 110 CONTINUE
  182. 120 CONTINUE
  183. 130 CONTINUE
  184. IPVT(N) = N
  185. IF (ABD(M,N) .EQ. 0.0E0) INFO = N
  186. RETURN
  187. END