|
|
| |
| # lines: |
2132 | | # code: |
2132 | | # comment: | 0 | |
# blank: | 0 |
| # Variables: | 77 |
| # Callers: | 0 |
| # Callings: | 2 |
| # Words: | 2241 |
| # Keywords: | 917 |
|
|
|
|
|
..
.. Local Scalars ..
..
.. Local Arrays ..
..
.. External Functions ..
..
.. External Subroutines ..
..
.. Intrinsic Functions ..
..
.. Executable Statements ..
ITERMAX = 0
NODE (IAFIRST,JAFIRST) OWNS A(1,1)
Determine the number of columns we have so we can check workspace
Set work array indices
Find a value for ROTN
Set machine-dependent constants for the stopping criterion.
If NORM(H) <= SQRT(OVFL), overflow should not occur.
I1 and I2 are the indices of the first row and last column of H
to which transformations must be applied. If eigenvalues only are
being computed, I1 and I2 are set inside the main loop.
ITN is the total number of QR iterations allowed.
The main loop begins here. I is the loop index and decreases from
IHI to ILO in steps of our schur block size (<=2*IBLK). Each
iteration of the loop works with the active submatrix in rows
and columns L to I. Eigenvalues I+1 to IHI have already
converged. Either L = ILO or the global A(L,L-1) is negligible
so that the matrix splits.
Perform QR iterations on rows and columns ILO to I until a
submatrix of order 1 or 2 splits off at the bottom because a
subdiagonal element has become negligible.
Look for a single small subdiagonal element.
H(L,L-1) is negligible
Exit from loop if a submatrix of order 1 or 2 has split off.
IF ( L .GE. I - (2*IBLK-1) )
IF ( L .GE. I - MAX(2*IBLK-1,HBL) )
Now the active submatrix is in rows and columns L to I. If
eigenvalues only are being computed, only the active submatrix
need be transformed.
Copy submatrix of size 2*JBLK and prepare to do generalized
Wilkinson shift or an exceptional shift
Make sure it's divisible by LCM (we want even workloads!)
Exceptional shift.
Prepare to use Wilkinson's double shift
Real roots: Use Wilkinson's shift twice
Look for two consecutive small subdiagonal elements:
PSLACONSB is the routine that does this.
Skip small submatrices
IF ( M .GE. I - 5 )
$ GO TO 80
Double-shift QR step
NBULGE is the number of bulges that will be attempted
Do not exceed maximum determined.
Make sure it's divisible by LCM (we want even workloads!)
sort the eigenpairs so that they are in twos for double
shifts. only call if several need sorting
IBULGE is the number of bulges going so far
"A" row defs : main row transforms from LOCALK to LOCALI2
"A" col defs : main col transforms from LOCALI1 to LOCALM
Which row & column will start the bulges
Set all values for bulges. All bulges are stored in
intermediate steps as loops over KI. Their current "task"
over the global M to I-1 values is always K1(KI) to K2(KI).
However, because there are many bulges, K1(KI) & K2(KI) might
go past that range while later bulges (KI+1,KI+2,etc..) are
finishing up.
Rules:
If MOD(K1(KI)-1,HBL) < HBL-2 then MOD(K2(KI)-1,HBL)<HBL-2
If MOD(K1(KI)-1,HBL) = HBL-2 then MOD(K2(KI)-1,HBL)=HBL-2
If MOD(K1(KI)-1,HBL) = HBL-1 then MOD(K2(KI)-1,HBL)=HBL-1
K2(KI)-K1(KI) <= ROTN
We first hit a border when MOD(K1(KI)-1,HBL)=HBL-2 and we hit
it again when MOD(K1(KI)-1,HBL)=HBL-1.
Get first transform on node who owns M+2,M+2
When we hit a border, there are row and column transforms that
overlap over several processors and the code gets very
"congested." As a remedy, when we first hit a border, a 6x6
*local* matrix is generated on one node (called SMALLA) and
work is done on that. At the end of the border, the data is
passed back and everything stays a lot simpler.
Copy 6 elements from global A(K-1:K+4,K-1:K+4)
Copy 6 elements from global A(K-2:K+3,K-2:K+3)
SLAHQR used to have a single row application and a single
column application to H. Here we do something a little
more clever. We break each transformation down into 3
parts:
1.) The minimum amount of work it takes to determine
a group of ROTN transformations (this is on
the critical path.) (Loops 130-180)
2.) The small work it takes so that each of the rows
and columns is at the same place. For example,
all ROTN row transforms are all complete
through some column TMP. (Loops within 190)
3.) The majority of the row and column transforms
are then applied in a block fashion.
(Loops 290 on.)
Each of these three parts are further subdivided into 3
parts:
A.) Work at the start of a border when
MOD(ISTART-1,HBL) = HBL-2
B.) Work at the end of a border when
MOD(ISTART-1,HBL) = HBL-1
C.) Work in the middle of the block when
MOD(ISTART-1,HBL) < HBL-2
Set a subdiagonal to zero now if it's possible
H11 = SMALLA(1,1,KI)
H10 = SMALLA(2,1,KI)
H22 = SMALLA(2,2,KI)
IF ( ABS(H10) .LE. MAX(ULP*(ABS(H11)+ABS(H22)),
$ SMLNUM) ) THEN
SMALLA(2,1,KI) = ZERO
WORK(ISUB+K-2) = ZERO
END IF
(IROW1,ICOL1) is (I,J)-coordinates of H(ISTART,ISTART)
(IROW1,ICOL1) is (I,J)-coordinates of H(ISTART,ISTART)
Create and do these transforms
Set a subdiagonal to zero now if it's possible
IF ( (IROW1.GT.2) .AND. (ICOL1.GT.2) .AND.
$ (MOD(K-1,HBL) .GT. 1) ) THEN
H11 = A((ICOL1-3)*LDA+IROW1-2)
H10 = A((ICOL1-3)*LDA+IROW1-1)
H22 = A((ICOL1-2)*LDA+IROW1-1)
IF ( ABS(H10).LE.MAX(ULP*(ABS(H11)+ABS(H22)),
$ SMLNUM) ) THEN
A((ICOL1-3)*LDA+IROW1-1) = ZERO
END IF
END IF
|
|
|
|
0001 SUBROUTINE PSLAHQR( WANTT , WANTZ , N , ILO , IHI , A , DESCA , WR , WI ,
0002 $ILOZ , IHIZ , Z , DESCZ , WORK , LWORK , IWORK ,
0003 $ILWORK , INFO )
0004
0005 * -- ScaLAPACK routine(version 1.7.3) --
0006 * University of Tennessee , Knoxville , Oak Ridge National Laboratory ,
0007 * and University of California , Berkeley.
0008 * 1.7.3 : March 22 , 2006
0009 * modification suggested by Mark Fahey and Greg Henry
0010 * 1.7.1 : January 30 , 2006
0011 * 1.7.0 : December 31 , 1998
0012
0013 * .. Scalar Arguments ..
0014 LOGICAL WANTT , WANTZ
0015 INTEGER IHI , IHIZ , ILO , ILOZ , ILWORK , INFO , LWORK , N
0016 * ..
0017 * .. Array Arguments ..
0018 INTEGER DESCA( * ) , DESCZ( * ) , IWORK( * )
0019 REAL A( * ) , WI( * ) , WORK( * ) , WR( * ) , Z( * )
0020 * ..
0021
0022 * Purpose
0023 * === ====
0024
0025 * PSLAHQR is an auxiliary routine used to find the Schur decomposition
0026 * and or eigenvalues of a matrix already in Hessenberg form from
0027 * cols ILO to IHI.
0028
0029 * Notes
0030 * === ==
0031
0032 * Each global data object is described by an associated description
0033 * vector. This vector stores the information required to establish
0034 * the mapping between an object element and its corresponding process
0035 * and memory location.
0036
0037 * Let A be a generic term for any 2D block cyclicly distributed array.
0038 * Such a global array has an associated description vector DESCA.
0039 * In the following comments , the character _ should be read as
0040 * "of the global array".
0041
0042 * NOTATION STORED IN EXPLANATION
0043 * --- ------------ -------------- --------------------------------------
0044 * DTYPE_A(global) DESCA( DTYPE_ )The descriptor type. In this case ,
0045 * DTYPE_A = 1.
0046 * CTXT_A(global) DESCA( CTXT_ ) The BLACS context handle , indicating
0047 * the BLACS process grid A is distribu -
0048 * ted over. The context itself is glo -
0049 * bal , but the handle(the integer
0050 * value) may vary.
0051 * M_A(global) DESCA( M_ ) The number of rows in the global
0052 * array A.
0053 * N_A(global) DESCA( N_ ) The number of columns in the global
0054 * array A.
0055 * MB_A(global) DESCA( MB_ ) The blocking factor used to distribute
0056 * the rows of the array.
0057 * NB_A(global) DESCA( NB_ ) The blocking factor used to distribute
0058 * the columns of the array.
0059 * RSRC_A(global) DESCA( RSRC_ ) The process row over which the first
0060 * row of the array A is distributed.
0061 * CSRC_A(global) DESCA( CSRC_ ) The process column over which the
0062 * first column of the array A is
0063 * distributed.
0064 * LLD_A(local) DESCA( LLD_ ) The leading dimension of the local
0065 * array. LLD_A >= MAX(1 , LOCr(M_A)).
0066
0067 * Let K be the number of rows or columns of a distributed matrix ,
0068 * and assume that its process grid has dimension p x q.
0069 * LOCr( K ) denotes the number of elements of K that a process
0070 * would receive if K were distributed over the p processes of its
0071 * process column.
0072 * Similarly , LOCc( K ) denotes the number of elements of K that a
0073 * process would receive if K were distributed over the q processes of
0074 * its process row.
0075 * The values of LOCr() and LOCc() may be determined via a call to the
0076 * ScaLAPACK tool function , NUMROC :
0077 * LOCr( M ) = NUMROC( M , MB_A , MYROW , RSRC_A , NPROW ) ,
0078 * LOCc( N ) = NUMROC( N , NB_A , MYCOL , CSRC_A , NPCOL ).
0079 * An upper bound for these quantities may be computed by :
0080 * LOCr( M ) <= ceil( ceil(M / MB_A) / NPROW )*MB_A
0081 * LOCc( N ) <= ceil( ceil(N / NB_A) / NPCOL )*NB_A
0082
0083 * Arguments
0084 * === ======
0085
0086 * WANTT(global input) LOGICAL
0087 * = .TRUE. : the full Schur form T is required ;
0088 * = .FALSE. : only eigenvalues are required.
0089
0090 * WANTZ(global input) LOGICAL
0091 * = .TRUE. : the matrix of Schur vectors Z is required ;
0092 * = .FALSE. : Schur vectors are not required.
0093
0094 * N(global input) INTEGER
0095 * The order of the Hessenberg matrix A(and Z if WANTZ).
0096 * N >= 0.
0097
0098 * ILO(global input) INTEGER
0099 * IHI(global input) INTEGER
0100 * It is assumed that A is already upper quasi - triangular in
0101 * rows and columns IHI + 1 : N , and that A(ILO , ILO - 1) = 0(unless
0102 * ILO = 1). PSLAHQR works primarily with the Hessenberg
0103 * submatrix in rows and columns ILO to IHI , but applies
0104 * transformations to all of H if WANTT is .TRUE..
0105 * 1 <= ILO <= max(1 , IHI) ; IHI <= N.
0106
0107 * A(global input / output) REAL array , dimension
0108 * (DESCA(LLD_) ,*)
0109 * On entry , the upper Hessenberg matrix A.
0110 * On exit , if WANTT is .TRUE. , A is upper quasi - triangular in
0111 * rows and columns ILO : IHI , with any 2 - by - 2 or larger diagonal
0112 * blocks not yet in standard form. If WANTT is .FALSE. , the
0113 * contents of A are unspecified on exit.
0114
0115 * DESCA(global and local input) INTEGER array of dimension DLEN_.
0116 * The array descriptor for the distributed matrix A.
0117
0118 * WR(global replicated output) REAL array ,
0119 * dimension(N)
0120 * WI(global replicated output) REAL array ,
0121 * dimension(N)
0122 * The real and imaginary parts , respectively , of the computed
0123 * eigenvalues ILO to IHI are stored in the corresponding
0124 * elements of WR and WI. If two eigenvalues are computed as a
0125 * complex conjugate pair , they are stored in consecutive
0126 * elements of WR and WI , say the i - th and(i + 1)th , with
0127 * WI(i) > 0 and WI(i + 1) < 0. If WANTT is .TRUE. , the
0128 * eigenvalues are stored in the same order as on the diagonal
0129 * of the Schur form returned in A. A may be returned with
0130 * larger diagonal blocks until the next release.
0131
0132 * ILOZ(global input) INTEGER
0133 * IHIZ(global input) INTEGER
0134 * Specify the rows of Z to which transformations must be
0135 * applied if WANTZ is .TRUE..
0136 * 1 <= ILOZ <= ILO ; IHI <= IHIZ <= N.
0137
0138 * Z(global input / output) REAL array.
0139 * If WANTZ is .TRUE. , on entry Z must contain the current
0140 * matrix Z of transformations accumulated by PDHSEQR , and on
0141 * exit Z has been updated ; transformations are applied only to
0142 * the submatrix Z(ILOZ : IHIZ , ILO : IHI).
0143 * If WANTZ is .FALSE. , Z is not referenced.
0144
0145 * DESCZ(global and local input) INTEGER array of dimension DLEN_.
0146 * The array descriptor for the distributed matrix Z.
0147
0148 * WORK(local output) REAL array of size LWORK
0149
0150 * LWORK(local input) INTEGER
0151 * WORK(LWORK) is a local array and LWORK is assumed big enough
0152 * so that LWORK >= 3*N +
0153 * MAX( 2*MAX(DESCZ(LLD_) , DESCA(LLD_)) + 2*LOCc(N) ,
0154 * 7*Ceil(N / HBL) / LCM(NPROW , NPCOL)) )
0155
0156 * IWORK(global and local input) INTEGER array of size ILWORK
0157
0158 * ILWORK(local input) INTEGER
0159 * This holds the some of the IBLK integer arrays. This is held
0160 * as a place holder for the next release.
0161
0162 * INFO(global output) INTEGER
0163 * < 0 : parameter number - INFO incorrect or inconsistent
0164 * = 0 : successful exit
0165 * > 0 : PSLAHQR failed to compute all the eigenvalues ILO to IHI
0166 * in a total of 30*(IHI - ILO + 1) iterations ; if INFO = i ,
0167 * elements i + 1 : ihi of WR and WI contain those eigenvalues
0168 * which have been successfully computed.
0169
0170 * Logic :
0171 * This algorithm is very similar to _LAHQR. Unlike _LAHQR ,
0172 * instead of sending one double shift through the largest
0173 * unreduced submatrix , this algorithm sends multiple double shifts
0174 * and spaces them apart so that there can be parallelism across
0175 * several processor row / columns. Another critical difference is
0176 * that this algorithm aggregrates multiple transforms together in
0177 * order to apply them in a block fashion.
0178
0179 * Important Local Variables :
0180 * IBLK = The maximum number of bulges that can be computed.
0181 * Currently fixed. Future releases this won't be fixed.
0182 * HBL = The square block size(HBL = DESCA(MB_) = DESCA(NB_))
0183 * ROTN = The number of transforms to block together
0184 * NBULGE = The number of bulges that will be attempted on the
0185 * current submatrix.
0186 * IBULGE = The current number of bulges started.
0187 * K1(*) , K2(*) = The current bulge loops from K1(*) to K2(*).
0188
0189 * Subroutines :
0190 * This routine calls :
0191 * PSLACONSB -> To determine where to start each iteration
0192 * PSLAWIL -> Given the shift , get the transformation
0193 * SLASORTE -> Pair up eigenvalues so that reals are paired.
0194 * PSLACP3 -> Parallel array to local replicated array copy &
0195 * back.
0196 * SLAREF -> Row / column reflector applier. Core routine
0197 * here.
0198 * PSLASMSUB -> Finds negligible subdiagonal elements.
0199
0200 * Current Notes and / or Restrictions :
0201 * 1.) This code requires the distributed block size to be square
0202 * and at least six(6) ; unlike simpler codes like LU , this
0203 * algorithm is extremely sensitive to block size. Unwise
0204 * choices of too small a block size can lead to bad
0205 * performance.
0206 * 2.) This code requires A and Z to be distributed identically
0207 * and have identical contxts.
0208 * 3.) This release currently does not have a routine for
0209 * resolving the Schur blocks into regular 2x2 form after
0210 * this code is completed. Because of this , a significant
0211 * performance impact is required while the deflation is done
0212 * by sometimes a single column of processors.
0213 * 4.) This code does not currently block the initial transforms
0214 * so that none of the rows or columns for any bulge are
0215 * completed until all are started. To offset pipeline
0216 * start - up it is recommended that at least 2*LCM(NPROW , NPCOL)
0217 * bulges are used(if possible)
0218 * 5.) The maximum number of bulges currently supported is fixed at
0219 * 32. In future versions this will be limited only by the
0220 * incoming WORK array.
0221 * 6.) The matrix A must be in upper Hessenberg form. If elements
0222 * below the subdiagonal are nonzero , the resulting transforms
0223 * may be nonsimilar. This is also true with the LAPACK
0224 * routine.
0225 * 7.) For this release , it is assumed RSRC_ = CSRC_ = 0
0226 * 8.) Currently , all the eigenvalues are distributed to all the
0227 * nodes. Future releases will probably distribute the
0228 * eigenvalues by the column partitioning.
0229 * 9.) The internals of this routine are subject to change.
0230
0231 * Implemented by : G. Henry , November 17 , 1996
0232
0233 * === ==================================================================
0234
0235 * .. Parameters ..
0236 INTEGER BLOCK_CYCLIC_2D , CSRC_ , CTXT_ , DLEN_ , DTYPE_ ,
0237 $LLD_ , MB_ , M_ , NB_ , N_ , RSRC_
0238 PARAMETER( BLOCK_CYCLIC_2D = 1 , DLEN_ = 9 , DTYPE_ = 1 ,
0239 $CTXT_ = 2 , M_ = 3 , N_ = 4 , MB_ = 5 , NB_ = 6 ,
0240 $RSRC_ = 7 , CSRC_ = 8 , LLD_ = 9 )
0241 REAL ZERO , ONE , HALF
0242 PARAMETER( ZERO = 0.0E + 0 , ONE = 1.0E + 0 , HALF = 0.5E + 0 )
0243 REAL CONST
0244 PARAMETER( CONST = 1.50E + 0 )
0245 INTEGER IBLK
0246 PARAMETER( IBLK = 32 )
0247 END IF
0248 ELSE IF( M.GT.L ) THEN
0248  
0249 IF( MOD( K - 1 , HBL ).GT.0 ) THEN
0249
0250 A(( ICOL1 - 2 )*LDA + IROW1 ) = - A(( ICOL1 - 2 )*
0251 $ LDA + IROW1 )
0252 END IF
0253 END IF
0254 V2 = VCOPY( 2 )
0255 T2 = T1COPY*V2
0256 WORK( VECSIDX + ( K - 1 )*3 + 1 ) = VCOPY( 2 )
0257 WORK( VECSIDX + ( K - 1 )*3 + 2 ) = VCOPY( 3 )
0258 WORK( VECSIDX + ( K - 1 )*3 + 3 ) = T1COPY
0259 T1 = T1COPY
0260 IF( K.LT.ISTOP ) THEN
0261
0262 * Do some work so next step is ready...
0263
0263
0264 V3 = VCOPY( 3 )
0265 T3 = T1*V3
0266 DO 50 J = ICOL1 , MIN( K2( KI ) + 1 , I - 1 ) +
0266
0267 $ ICOL1 - K
0268 SUM = A(( J - 1 )*LDA + IROW1 ) +
0269 $ V2*A(( J - 1 )*LDA + IROW1 + 1 ) +
0270 $ V3*A(( J - 1 )*LDA + IROW1 + 2 )
0271 A(( J - 1 )*LDA + IROW1 ) = A(( J - 1 )*LDA +
0272 $ IROW1 ) - SUM*T1
0273 A(( J - 1 )*LDA + IROW1 + 1 ) = A(( J - 1 )*LDA +
0274 $ IROW1 + 1 ) - SUM*T2
0275 A(( J - 1 )*LDA + IROW1 + 2 ) = A(( J - 1 )*LDA +
0276 $ IROW1 + 2 ) - SUM*T3
0277 50 CONTINUE
0277
0278 ITMP1 = LOCALK2( KI )
0279 DO 60 J = IROW1 + 1 , IROW1 + 3
0279
0280 SUM = A(( ICOL1 - 1 )*LDA + J ) +
0281 $ V2*A( ICOL1*LDA + J ) +
0282 $ V3*A(( ICOL1 + 1 )*LDA + J )
0283 A(( ICOL1 - 1 )*LDA + J ) = A(( ICOL1 - 1 )*LDA +
0284 $ J ) - SUM*T1
0285 A( ICOL1*LDA + J ) = A( ICOL1*LDA + J ) - SUM*T2
0286 A(( ICOL1 + 1 )*LDA + J ) = A(( ICOL1 + 1 )*LDA +
0287 $ J ) - SUM*T3
0288 60 CONTINUE
0289 END IF
0290 IROW1 = IROW1 + 1
0291 ICOL1 = ICOL1 + 1
0292 70 CONTINUE
0293 END IF
0294
0295 IF( MODKM1.EQ.HBL - 2 ) THEN
0295
0296 IF(( DOWN.EQ.ICURROW( KI ) ) .AND.
0297 $( RIGHT.EQ.ICURCOL( KI ) ) .AND.( NUM.GT.1 ) )
0297
0298 $ THEN
0299 CALL SGERV2D( CONTXT , 3 , 1 ,
0300 $ WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) , 3 ,
0301 $ DOWN , RIGHT )
0302 END IF
0303 IF(( MYROW.EQ.ICURROW( KI ) ) .AND.
0304 $( MYCOL.EQ.ICURCOL( KI ) ) .AND.( NUM.GT.1 ) )
0305 $THEN
0306 CALL SGESD2D( CONTXT , 3 , 1 ,
0307 $WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) , 3 ,
0308 $UP , LEFT )
0309 END IF
0310 IF(( DOWN.EQ.ICURROW( KI ) ) .AND.
0311 $( NPCOL.GT.1 ) .AND.( ISTART.LE.ISTOP ) ) THEN
0312 JJ = MOD( ICURCOL( KI ) + NPCOL - 1 , NPCOL )
0313 IF( MYCOL.NE.JJ ) THEN
0313
0314 CALL SGEBR2D( CONTXT , 'ROW' , ' ' ,
0315 $ 3*( ISTOP - ISTART + 1 ) , 1 ,
0316 $ WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) ,
0317 $ 3*( ISTOP - ISTART + 1 ) , MYROW , JJ )
0318 ELSE
0318
0319 CALL SGEBS2D( CONTXT , 'ROW' , ' ' ,
0320 $ 3*( ISTOP - ISTART + 1 ) , 1 ,
0321 $ WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) ,
0322 $ 3*( ISTOP - ISTART + 1 ) )
0323 END IF
0324 END IF
0325 END IF
0326
0327 * Broadcast Householder information from the block
0328
0329 IF(( MYROW.EQ.ICURROW( KI ) ) .AND.( NPCOL.GT.1 ) .AND.
0330 $( ISTART.LE.ISTOP ) ) THEN
0331 IF( MYCOL.NE.ICURCOL( KI ) ) THEN
0331
0332 CALL SGEBR2D( CONTXT , 'ROW' , ' ' ,
0333 $ 3*( ISTOP - ISTART + 1 ) , 1 ,
0334 $ WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) ,
0335 $ 3*( ISTOP - ISTART + 1 ) , MYROW ,
0336 $ ICURCOL( KI ) )
0337 ELSE
0337
0338 CALL SGEBS2D( CONTXT , 'ROW' , ' ' ,
0339 $ 3*( ISTOP - ISTART + 1 ) , 1 ,
0340 $ WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) ,
0341 $ 3*( ISTOP - ISTART + 1 ) )
0342 END IF
0343 END IF
0344 80 CONTINUE
0345
0346 * Now do column transforms and finish work
0347
0348 DO 90 KI = 1 , IBULGE
0349
0349
0350 ISTART = MAX( K1( KI ) , M )
0351 ISTOP = MIN( K2( KI ) , I - 1 )
0352
0353 IF( MOD( ISTART - 1 , HBL ).EQ.HBL - 2 ) THEN
0353
0354 IF(( RIGHT.EQ.ICURCOL( KI ) ) .AND.
0355 $( NPROW.GT.1 ) .AND.( ISTART.LE.ISTOP ) ) THEN
0355
0356 JJ = MOD( ICURROW( KI ) + NPROW - 1 , NPROW )
0357 IF( MYROW.NE.JJ ) THEN
0357
0358 CALL SGEBR2D( CONTXT , 'COL' , ' ' ,
0359 $ 3*( ISTOP - ISTART + 1 ) , 1 ,
0360 $ WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) ,
0361 $ 3*( ISTOP - ISTART + 1 ) , JJ , MYCOL )
0362 ELSE
0362
0363 CALL SGEBS2D( CONTXT , 'COL' , ' ' ,
0364 $ 3*( ISTOP - ISTART + 1 ) , 1 ,
0365 $ WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) ,
0366 $ 3*( ISTOP - ISTART + 1 ) )
0367 END IF
0368 END IF
0369 END IF
0370
0371 IF(( MYCOL.EQ.ICURCOL( KI ) ) .AND.( NPROW.GT.1 ) .AND.
0372 $( ISTART.LE.ISTOP ) ) THEN
0373 IF( MYROW.NE.ICURROW( KI ) ) THEN
0373
0374 CALL SGEBR2D( CONTXT , 'COL' , ' ' ,
0375 $ 3*( ISTOP - ISTART + 1 ) , 1 ,
0376 $ WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) ,
0377 $ 3*( ISTOP - ISTART + 1 ) , ICURROW( KI ) ,
0378 $ MYCOL )
0379 ELSE
0379
0380 CALL SGEBS2D( CONTXT , 'COL' , ' ' ,
0381 $ 3*( ISTOP - ISTART + 1 ) , 1 ,
0382 $ WORK( VECSIDX + ( ISTART - 1 )*3 + 1 ) ,
0383 $ 3*( ISTOP - ISTART + 1 ) )
0384 END IF
0385 END IF
0386 90 CONTINUE
0387
0388 * Now do make up work to have things in block fashion
0389
0390 DO 150 KI = 1 , IBULGE
0390
0391 ISTART = MAX( K1( KI ) , M )
0392 ISTOP = MIN( K2( KI ) , I - 1 )
0393
0394 MODKM1 = MOD( ISTART - 1 , HBL )
0395 IF(( MYROW.EQ.ICURROW( KI ) ) .AND.
0396 $( MYCOL.EQ.ICURCOL( KI ) ) .AND.
0397 $( MODKM1.EQ.HBL - 2 ) .AND.( ISTART.LT.I - 1 ) ) THEN
0397
0398 K = ISTART
0399
0400 * Catch up on column & border work
0401
0402 NR = MIN( 3 , I - K + 1 )
0403 V2 = WORK( VECSIDX + ( K - 1 )*3 + 1 )
0404 V3 = WORK( VECSIDX + ( K - 1 )*3 + 2 )
0405 T1 = WORK( VECSIDX + ( K - 1 )*3 + 3 )
0406 IF( NR.EQ.3 ) THEN
0407
0408 * Do some work so next step is ready...
0409
0410 * V3 = VCOPY( 3 )
0410
0411 T2 = T1*V2
0412 T3 = T1*V3
0413 ITMP1 = MIN( 6 , I2 + 2 - K )
0414 ITMP2 = MAX( I1 - K + 2 , 1 )
0415 DO 100 J = 2 , ITMP1
0415
0416 SUM = SMALLA( 2 , J , KI ) +
0417 $ V2*SMALLA( 3 , J , KI ) +
0418 $ V3*SMALLA( 4 , J , KI )
0419 SMALLA( 2 , J , KI ) = SMALLA( 2 , J , KI ) - SUM*T1
0420 SMALLA( 3 , J , KI ) = SMALLA( 3 , J , KI ) - SUM*T2
0421 SMALLA( 4 , J , KI ) = SMALLA( 4 , J , KI ) - SUM*T3
0422 100 CONTINUE
0422
0423 DO 110 J = ITMP2 , 5
0423
0424 SUM = SMALLA( J , 2 , KI ) +
0425 $ V2*SMALLA( J , 3 , KI ) +
0426 $ V3*SMALLA( J , 4 , KI )
0427 SMALLA( J , 2 , KI ) = SMALLA( J , 2 , KI ) - SUM*T1
0428 SMALLA( J , 3 , KI ) = SMALLA( J , 3 , KI ) - SUM*T2
0429 SMALLA( J , 4 , KI ) = SMALLA( J , 4 , KI ) - SUM*T3
0430 110 CONTINUE
0430
0431 END IF
0432 END IF
0433
0434 IF(( MOD( ISTART - 1 , HBL ).EQ.HBL - 1 ) .AND.
0435 $( ISTART.LE.ISTOP ) .AND.
0436 $( MYROW.EQ.ICURROW( KI ) ) .AND.
0437 $( MYCOL.EQ.ICURCOL( KI ) ) ) THEN
0438 K = ISTOP
0439
0440 * Catch up on column & border work
0441
0442 NR = MIN( 3 , I - K + 1 )
0443 V2 = WORK( VECSIDX + ( K - 1 )*3 + 1 )
0444 V3 = WORK( VECSIDX + ( K - 1 )*3 + 2 )
0445 T1 = WORK( VECSIDX + ( K - 1 )*3 + 3 )
0446 IF( NR.EQ.3 ) THEN
0447
0448 * Do some work so next step is ready...
0449
0450 * V3 = VCOPY( 3 )
0450
0451 T2 = T1*V2
0452 T3 = T1*V3
0453 ITMP1 = MIN( 6 , I2 - K + 3 )
0454 ITMP2 = MAX( I1 - K + 3 , 1 )
0455 DO 120 J = 3 , ITMP1
0455
0456 SUM = SMALLA( 3 , J , KI ) +
0457 $ V2*SMALLA( 4 , J , KI ) +
0458 $ V3*SMALLA( 5 , J , KI )
0459 SMALLA( 3 , J , KI ) = SMALLA( 3 , J , KI ) - SUM*T1
0460 SMALLA( 4 , J , KI ) = SMALLA( 4 , J , KI ) - SUM*T2
0461 SMALLA( 5 , J , KI ) = SMALLA( 5 , J , KI ) - SUM*T3
0462 120 CONTINUE
0462
0463 DO 130 J = ITMP2 , 6
0463
0464 SUM = SMALLA( J , 3 , KI ) +
0465 $ V2*SMALLA( J , 4 , KI ) +
0466 $ V3*SMALLA( J , 5 , KI )
0467 SMALLA( J , 3 , KI ) = SMALLA( J , 3 , KI ) - SUM*T1
0468 SMALLA( J , 4 , KI ) = SMALLA( J , 4 , KI ) - SUM*T2
0469 SMALLA( J , 5 , KI ) = SMALLA( J , 5 , KI ) - SUM*T3
0470 130 CONTINUE
0471 END IF
0472 END IF
0473
0474 MODKM1 = MOD( ISTART - 1 , HBL )
0475 IF(( MYROW.EQ.ICURROW( KI ) ) .AND.
0476 $( MYCOL.EQ.ICURCOL( KI ) ) .AND.
0477 $((( MODKM1.EQ.HBL - 2 ) .AND.( ISTART.EQ.I -
0478 $1 ) ) .OR.(( MODKM1.LT.HBL - 2 ) .AND.( ISTART.LE.I -
0479 $1 ) ) ) ) THEN
0480
0481 * (IROW1 , ICOL1) is(I , J) - coordinates of H(ISTART , ISTART)
0482
0483 IROW1 = KROW( KI )
0484 ICOL1 = LOCALK2( KI )
0485 DO 140 K = ISTART , ISTOP
0486
0487 * Catch up on column & border work
0488
0488
0489 NR = MIN( 3 , I - K + 1 )
0490 V2 = WORK( VECSIDX + ( K - 1 )*3 + 1 )
0491 V3 = WORK( VECSIDX + ( K - 1 )*3 + 2 )
0492 T1 = WORK( VECSIDX + ( K - 1 )*3 + 3 )
0493 IF( K.LT.ISTOP ) THEN
0494
0495 * Do some work so next step is ready...
0496
0496
0497 T2 = T1*V2
0498 T3 = T1*V3
0499 CALL SLAREF ( 'Col' , A , LDA , .FALSE. , Z , LDZ ,
0500 $ .FALSE. , ICOL1 , ICOL1 , ISTART ,
0501 $ ISTOP , MIN( ISTART + 1 , I ) - K + IROW1 ,
0502 $ IROW1 , LILOZ , LIHIZ ,
0503 $ WORK( VECSIDX + 1 ) , V2 , V3 , T1 , T2 ,
0504 $ T3 )
0505 IROW1 = IROW1 + 1
0506 ICOL1 = ICOL1 + 1
0507 ELSE
0507
0508 IF(( NR.EQ.3 ) .AND.( MOD( K - 1 ,
0509 $ HBL ).LT.HBL - 2 ) ) THEN
0510 T2 = T1*V2
0511 T3 = T1*V3
0512 CALL SLAREF ( 'Row' , A , LDA , .FALSE. , Z , LDZ ,
0513 $ .FALSE. , IROW1 , IROW1 , ISTART ,
0514 $ ISTOP , ICOL1 , MIN( MIN( K2( KI )
0515 $ + 1 , I - 1 ) , I2 ) - K + ICOL1 , LILOZ ,
0516 $ LIHIZ , WORK( VECSIDX + 1 ) , V2 ,
0517 $ V3 , T1 , T2 , T3 )
0518 END IF
0519 END IF
0520 140 CONTINUE
0521 END IF
0522
0523 * Send SMALLA back again.
0524
0525 K = ISTART
0526 MODKM1 = MOD( K - 1 , HBL )
0527 IF(( MODKM1.GE.HBL - 2 ) .AND.( K.LE.I - 1 ) ) THEN
0527
0528 IF(( MODKM1.EQ.HBL - 2 ) .AND.( K.LT.I - 1 ) ) THEN
0529
0530 * Copy 6 elements from global A(K - 1 : K + 4 , K - 1 : K + 4)
0531
0531
0532 CALL INFOG2L( K + 2 , K + 2 , DESCA , NPROW , NPCOL , MYROW ,
0533 $ MYCOL , IROW1 , ICOL1 , ITMP1 , ITMP2 )
0534 CALL PSLACP3 ( MIN( 6 , N - K + 2 ) , K - 1 , A , DESCA ,
0535 $ SMALLA( 1 , 1 , KI ) , 6 , ITMP1 , ITMP2 ,
0536 $ 1 )
0537
0538 END IF
0539 IF( MODKM1.EQ.HBL - 1 ) THEN
0540
0541 * Copy 6 elements from global A(K - 2 : K + 3 , K - 2 : K + 3)
0542
0542
0543 CALL INFOG2L( K + 1 , K + 1 , DESCA , NPROW , NPCOL , MYROW ,
0544 $ MYCOL , IROW1 , ICOL1 , ITMP1 , ITMP2 )
0545 CALL PSLACP3 ( MIN( 6 , N - K + 3 ) , K - 2 , A , DESCA ,
0546 $ SMALLA( 1 , 1 , KI ) , 6 , ITMP1 , ITMP2 ,
0547 $ 1 )
0548 END IF
0549 END IF
0550
0551 150 CONTINUE
0552
0553 * Now start major set of block ROW reflections
0554
0555 DO 160 KI = 1 , IBULGE
0555
0556 IF(( MYROW.NE.ICURROW( KI ) ) .AND.
0557 $( DOWN.NE.ICURROW( KI ) ) )GO TO 160
0557
0558 ISTART = MAX( K1( KI ) , M )
0559 ISTOP = MIN( K2( KI ) , I - 1 )
0560
0561 IF(( ISTOP.GT.ISTART ) .AND.
0562 $( MOD( ISTART - 1 , HBL ).LT.HBL - 2 ) .AND.
0563 $( ICURROW( KI ).EQ.MYROW ) ) THEN
0563
0564 IROW1 = MIN( K2( KI ) + 1 , I - 1 ) + 1
0565 CALL INFOG1L( IROW1 , HBL , NPCOL , MYCOL , 0 , ITMP1 ,
0566 $ ITMP2 )
0567 ITMP2 = NUMROC( I2 , HBL , MYCOL , 0 , NPCOL )
0568 II = KROW( KI )
0569 CALL SLAREF ( 'Row' , A , LDA , WANTZ , Z , LDZ , .TRUE. , II ,
0570 $ II , ISTART , ISTOP , ITMP1 , ITMP2 , LILOZ ,
0571 $ LIHIZ , WORK( VECSIDX + 1 ) , V2 , V3 , T1 , T2 ,
0572 $ T3 )
0573 END IF
0574 160 CONTINUE
0575
0576 DO 180 KI = 1 , IBULGE
0576
0577 IF( KROW( KI ).GT.KP2ROW( KI ) )
0577
0578 $ GO TO 180
0579 IF(( MYROW.NE.ICURROW( KI ) ) .AND.
0580 $( DOWN.NE.ICURROW( KI ) ) )GO TO 180
0580
0581 ISTART = MAX( K1( KI ) , M )
0582 ISTOP = MIN( K2( KI ) , I - 1 )
0583 IF(( ISTART.EQ.ISTOP ) .OR.
0584 $( MOD( ISTART - 1 , HBL ).GE.HBL - 2 ) .OR.
0585 $( ICURROW( KI ).NE.MYROW ) ) THEN
0585
0586 DO 170 K = ISTART , ISTOP
0586
0587 V2 = WORK( VECSIDX + ( K - 1 )*3 + 1 )
0588 V3 = WORK( VECSIDX + ( K - 1 )*3 + 2 )
0589 T1 = WORK( VECSIDX + ( K - 1 )*3 + 3 )
0590 NR = MIN( 3 , I - K + 1 )
0591 IF(( NR.EQ.3 ) .AND.( KROW( KI ).LE.
0592 $ KP2ROW( KI ) ) ) THEN
0593 IF(( K.LT.ISTOP ) .AND.
0594 $( MOD( K - 1 , HBL ).LT.HBL - 2 ) ) THEN
0594
0595 ITMP1 = MIN( K2( KI ) + 1 , I - 1 ) + 1
0596 ELSE
0596
0597 IF( MOD( K - 1 , HBL ).LT.HBL - 2 ) THEN
0597
0598 ITMP1 = MIN( K2( KI ) + 1 , I - 1 ) + 1
0599 END IF
0600 IF( MOD( K - 1 , HBL ).EQ.HBL - 2 ) THEN
0600
0601 ITMP1 = MIN( K + 4 , I2 ) + 1
0602 END IF
0603 IF( MOD( K - 1 , HBL ).EQ.HBL - 1 ) THEN
0603
0604 ITMP1 = MIN( K + 3 , I2 ) + 1
0605 END IF
0606 END IF
0607
0608 * Find local coor of rows K through K + 2
0609
0610 IROW1 = KROW( KI )
0611 IROW2 = KP2ROW( KI )
0612 CALL INFOG1L( ITMP1 , HBL , NPCOL , MYCOL , 0 ,
0613 $ ICOL1 , ICOL2 )
0614 ICOL2 = NUMROC( I2 , HBL , MYCOL , 0 , NPCOL )
0615 IF(( MOD( K - 1 , HBL ).LT.HBL - 2 ) .OR.
0616 $( NPROW.EQ.1 ) ) THEN
0616
0617 T2 = T1*V2
0618 T3 = T1*V3
0619 CALL SLAREF ( 'Row' , A , LDA , WANTZ , Z , LDZ ,
0620 $ .FALSE. , IROW1 , IROW1 , ISTART ,
0621 $ ISTOP , ICOL1 , ICOL2 , LILOZ ,
0622 $ LIHIZ , WORK( VECSIDX + 1 ) , V2 ,
0623 $ V3 , T1 , T2 , T3 )
0624 END IF
0625 IF(( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND.
0626 $( NPROW.GT.1 ) ) THEN
0626
0627 IF( IROW1.EQ.IROW2 ) THEN
0627
0628 CALL SGESD2D( CONTXT , 1 , ICOL2 - ICOL1 + 1 ,
0629 $ A(( ICOL1 - 1 )*LDA + IROW2 ) ,
0630 $ LDA , UP , MYCOL )
0631 END IF
0632 END IF
0633 IF(( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND.
0634 $( NPROW.GT.1 ) ) THEN
0635 IF( IROW1.EQ.IROW2 ) THEN
0635
0636 CALL SGESD2D( CONTXT , 1 , ICOL2 - ICOL1 + 1 ,
0637 $ A(( ICOL1 - 1 )*LDA + IROW1 ) ,
0638 $ LDA , DOWN , MYCOL )
0639 END IF
0640 END IF
0641 END IF
0642 170 CONTINUE
0643 END IF
0644 180 CONTINUE
0645
0646 DO 220 KI = 1 , IBULGE
0646
0647 IF( KROW( KI ).GT.KP2ROW( KI ) )
0647
0648 $ GO TO 220
0649 IF(( MYROW.NE.ICURROW( KI ) ) .AND.
0650 $( DOWN.NE.ICURROW( KI ) ) )GO TO 220
0650
0651 ISTART = MAX( K1( KI ) , M )
0652 ISTOP = MIN( K2( KI ) , I - 1 )
0653 IF(( ISTART.EQ.ISTOP ) .OR.
0654 $( MOD( ISTART - 1 , HBL ).GE.HBL - 2 ) .OR.
0655 $( ICURROW( KI ).NE.MYROW ) ) THEN
0655
0656 DO 210 K = ISTART , ISTOP
0656
0657 V2 = WORK( VECSIDX + ( K - 1 )*3 + 1 )
0658 V3 = WORK( VECSIDX + ( K - 1 )*3 + 2 )
0659 T1 = WORK( VECSIDX + ( K - 1 )*3 + 3 )
0660 NR = MIN( 3 , I - K + 1 )
0661 IF(( NR.EQ.3 ) .AND.( KROW( KI ).LE.
0662 $ KP2ROW( KI ) ) ) THEN
0663 IF(( K.LT.ISTOP ) .AND.
0664 $( MOD( K - 1 , HBL ).LT.HBL - 2 ) ) THEN
0664
0665 ITMP1 = MIN( K2( KI ) + 1 , I - 1 ) + 1
0666 ELSE
0666
0667 IF( MOD( K - 1 , HBL ).LT.HBL - 2 ) THEN
0667
0668 ITMP1 = MIN( K2( KI ) + 1 , I - 1 ) + 1
0669 END IF
0670 IF( MOD( K - 1 , HBL ).EQ.HBL - 2 ) THEN
0670
0671 ITMP1 = MIN( K + 4 , I2 ) + 1
0672 END IF
0673 IF( MOD( K - 1 , HBL ).EQ.HBL - 1 ) THEN
0673
0674 ITMP1 = MIN( K + 3 , I2 ) + 1
0675 END IF
0676 END IF
0677
0678 IROW1 = KROW( KI ) + K - ISTART
0679 IROW2 = KP2ROW( KI ) + K - ISTART
0680 CALL INFOG1L( ITMP1 , HBL , NPCOL , MYCOL , 0 ,
0681 $ ICOL1 , ICOL2 )
0682 ICOL2 = NUMROC( I2 , HBL , MYCOL , 0 , NPCOL )
0683 IF(( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND.
0684 $( NPROW.GT.1 ) ) THEN
0684
0685 IF( IROW1.NE.IROW2 ) THEN
0685
0686 CALL SGERV2D( CONTXT , 1 , ICOL2 - ICOL1 + 1 ,
0687 $ WORK( IRBUF + 1 ) , 1 , DOWN ,
0688 $ MYCOL )
0689 T2 = T1*V2
0690 T3 = T1*V3
0691 DO 190 J = ICOL1 , ICOL2
0691
0692 SUM = A(( J - 1 )*LDA + IROW1 ) +
0693 $ V2*A(( J - 1 )*LDA + IROW1 + 1 ) +
0694 $ V3*WORK( IRBUF + J - ICOL1 + 1 )
0695 A(( J - 1 )*LDA + IROW1 ) = A(( J - 1 )*
0696 $ LDA + IROW1 ) - SUM*T1
0697 A(( J - 1 )*LDA + IROW1 + 1 ) = A(( J - 1 )*
0698 $ LDA + IROW1 + 1 ) - SUM*T2
0699 WORK( IRBUF + J - ICOL1 + 1 ) = WORK( IRBUF +
0700 $ J - ICOL1 + 1 ) - SUM*T3
0701 190 CONTINUE
0701
0702 CALL SGESD2D( CONTXT , 1 , ICOL2 - ICOL1 + 1 ,
0703 $ WORK( IRBUF + 1 ) , 1 , DOWN ,
0704 $ MYCOL )
0705 END IF
0706 END IF
0707 IF(( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND.
0708 $( NPROW.GT.1 ) ) THEN
0708
0709 IF( IROW1.NE.IROW2 ) THEN
0709
0710 CALL SGERV2D( CONTXT , 1 , ICOL2 - ICOL1 + 1 ,
0711 $ WORK( IRBUF + 1 ) , 1 , UP ,
0712 $ MYCOL )
0713 T2 = T1*V2
0714 T3 = T1*V3
0715 DO 200 J = ICOL1 , ICOL2
0715
0716 SUM = WORK( IRBUF + J - ICOL1 + 1 ) +
0717 $ V2*A(( J - 1 )*LDA + IROW1 ) +
0718 $ V3*A(( J - 1 )*LDA + IROW1 + 1 )
0719 WORK( IRBUF + J - ICOL1 + 1 ) = WORK( IRBUF +
0720 $ J - ICOL1 + 1 ) - SUM*T1
0721 A(( J - 1 )*LDA + IROW1 ) = A(( J - 1 )*
0722 $ LDA + IROW1 ) - SUM*T2
0723 A(( J - 1 )*LDA + IROW1 + 1 ) = A(( J - 1 )*
0724 $ LDA + IROW1 + 1 ) - SUM*T3
0725 200 CONTINUE
0725
0726 CALL SGESD2D( CONTXT , 1 , ICOL2 - ICOL1 + 1 ,
0727 $ WORK( IRBUF + 1 ) , 1 , UP ,
0728 $ MYCOL )
0729 END IF
0730 END IF
0731 END IF
0732 210 CONTINUE
0733 END IF
0734 220 CONTINUE
0735
0736 DO 240 KI = 1 , IBULGE
0736
0737 IF( KROW( KI ).GT.KP2ROW( KI ) )
0737
0738 $ GO TO 240
0739 IF(( MYROW.NE.ICURROW( KI ) ) .AND.
0740 $( DOWN.NE.ICURROW( KI ) ) )GO TO 240
0740
0741 ISTART = MAX( K1( KI ) , M )
0742 ISTOP = MIN( K2( KI ) , I - 1 )
0743 IF(( ISTART.EQ.ISTOP ) .OR.
0744 $( MOD( ISTART - 1 , HBL ).GE.HBL - 2 ) .OR.
0745 $( ICURROW( KI ).NE.MYROW ) ) THEN
0745
0746 DO 230 K = ISTART , ISTOP
0746
0747 V2 = WORK( VECSIDX + ( K - 1 )*3 + 1 )
0748 V3 = WORK( VECSIDX + ( K - 1 )*3 + 2 )
0749 T1 = WORK( VECSIDX + ( K - 1 )*3 + 3 )
0750 NR = MIN( 3 , I - K + 1 )
0751 IF(( NR.EQ.3 ) .AND.( KROW( KI ).LE.
0752 $ KP2ROW( KI ) ) ) THEN
0753 IF(( K.LT.ISTOP ) .AND.
0754 $( MOD( K - 1 , HBL ).LT.HBL - 2 ) ) THEN
0754
0755 ITMP1 = MIN( K2( KI ) + 1 , I - 1 ) + 1
0756 ELSE
0756
0757 IF( MOD( K - 1 , HBL ).LT.HBL - 2 ) THEN
0757
0758 ITMP1 = MIN( K2( KI ) + 1 , I - 1 ) + 1
0759 END IF
0760 IF( MOD( K - 1 , HBL ).EQ.HBL - 2 ) THEN
0760
0761 ITMP1 = MIN( K + 4 , I2 ) + 1
0762 END IF
0763 IF( MOD( K - 1 , HBL ).EQ.HBL - 1 ) THEN
0763
0764 ITMP1 = MIN( K + 3 , I2 ) + 1
0765 END IF
0766 END IF
0767
0768 IROW1 = KROW( KI ) + K - ISTART
0769 IROW2 = KP2ROW( KI ) + K - ISTART
0770 CALL INFOG1L( ITMP1 , HBL , NPCOL , MYCOL , 0 ,
0771 $ ICOL1 , ICOL2 )
0772 ICOL2 = NUMROC( I2 , HBL , MYCOL , 0 , NPCOL )
0773 IF(( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND.
0774 $( NPROW.GT.1 ) ) THEN
0774
0775 IF( IROW1.EQ.IROW2 ) THEN
0775
0776 CALL SGERV2D( CONTXT , 1 , ICOL2 - ICOL1 + 1 ,
0777 $ A(( ICOL1 - 1 )*LDA + IROW2 ) ,
0778 $ LDA , UP , MYCOL )
0779 END IF
0780 END IF
0781 IF(( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND.
0782 $( NPROW.GT.1 ) ) THEN
0782
0783 IF( IROW1.EQ.IROW2 ) THEN
0783
0784 CALL SGERV2D( CONTXT , 1 , ICOL2 - ICOL1 + 1 ,
0785 $ A(( ICOL1 - 1 )*LDA + IROW1 ) ,
0786 $ LDA , DOWN , MYCOL )
0787 END IF
0788 END IF
0789 END IF
0790 230 CONTINUE
0791 END IF
0792 240 CONTINUE
0793 250 CONTINUE
0794
0795 * Now start major set of block COL reflections
0796
0797 DO 260 KI = 1 , IBULGE
0797
0798 IF(( MYCOL.NE.ICURCOL( KI ) ) .AND.
0799 $( RIGHT.NE.ICURCOL( KI ) ) )GO TO 260
0799
0800 ISTART = MAX( K1( KI ) , M )
0801 ISTOP = MIN( K2( KI ) , I - 1 )
0802
0803 IF((( MOD( ISTART - 1 , HBL ).LT.HBL - 2 ) .OR.( NPCOL.EQ.
0804 $ 1 ) ) .AND.( ICURCOL( KI ).EQ.MYCOL ) .AND.
0805 $( I - ISTOP + 1.GE.3 ) ) THEN
0805
0806 K = ISTART
0807 IF(( K.LT.ISTOP ) .AND.( MOD( K - 1 ,
0808 $ HBL ).LT.HBL - 2 ) ) THEN
0809 ITMP1 = MIN( ISTART + 1 , I ) - 1
0810 ELSE
0810
0811 IF( MOD( K - 1 , HBL ).LT.HBL - 2 ) THEN
0811
0812 ITMP1 = MIN( K + 3 , I )
0813 END IF
0814 IF( MOD( K - 1 , HBL ).EQ.HBL - 2 ) THEN
0814
0815 ITMP1 = MAX( I1 , K - 1 ) - 1
0816 END IF
0817 IF( MOD( K - 1 , HBL ).EQ.HBL - 1 ) THEN
0817
0818 ITMP1 = MAX( I1 , K - 2 ) - 1
0819 END IF
0820 END IF
0821
0822 ICOL1 = KCOL( KI )
0823 CALL INFOG1L( I1 , HBL , NPROW , MYROW , 0 , IROW1 , IROW2 )
0824 IROW2 = NUMROC( ITMP1 , HBL , MYROW , 0 , NPROW )
0825 IF( IROW1.LE.IROW2 ) THEN
0825
0826 ITMP2 = IROW2
0827 ELSE
0827
0828 ITMP2 = - 1
0829 END IF
0830 CALL SLAREF ( 'Col' , A , LDA , WANTZ , Z , LDZ , .TRUE. ,
0831 $ICOL1 , ICOL1 , ISTART , ISTOP , IROW1 ,
0832 $IROW2 , LILOZ , LIHIZ , WORK( VECSIDX + 1 ) ,
0833 $V2 , V3 , T1 , T2 , T3 )
0834 K = ISTOP
0835 IF( MOD( K - 1 , HBL ).LT.HBL - 2 ) THEN
0836
0837 * Do from ITMP1 + 1 to MIN(K + 3 , I)
0838
0838
0839 IF( MOD( K - 1 , HBL ).LT.HBL - 3 ) THEN
0839
0840 IROW1 = ITMP2 + 1
0841 IF( MOD(( ITMP1 / HBL ) , NPROW ).EQ.MYROW )
0841
0842 $ THEN
0843 IF( ITMP2.GT.0 ) THEN
0843
0844 IROW2 = ITMP2 + MIN( K + 3 , I ) - ITMP1
0845 ELSE
0845
0846 IROW2 = IROW1 - 1
0847 END IF
0848 ELSE
0848
0849 IROW2 = IROW1 - 1
0850 END IF
0851 ELSE
0851
0852 CALL INFOG1L( ITMP1 + 1 , HBL , NPROW , MYROW , 0 ,
0853 $ IROW1 , IROW2 )
0854 IROW2 = NUMROC( MIN( K + 3 , I ) , HBL , MYROW , 0 ,
0855 $ NPROW )
0856 END IF
0857 V2 = WORK( VECSIDX + ( K - 1 )*3 + 1 )
0858 V3 = WORK( VECSIDX + ( K - 1 )*3 + 2 )
0859 T1 = WORK( VECSIDX + ( K - 1 )*3 + 3 )
0860 T2 = T1*V2
0861 T3 = T1*V3
0862 ICOL1 = KCOL( KI ) + ISTOP - ISTART
0863 CALL SLAREF ( 'Col' , A , LDA , .FALSE. , Z , LDZ ,
0864 $ .FALSE. , ICOL1 , ICOL1 , ISTART , ISTOP ,
0865 $ IROW1 , IROW2 , LILOZ , LIHIZ ,
0866 $ WORK( VECSIDX + 1 ) , V2 , V3 , T1 , T2 ,
0867 $ T3 )
0868 END IF
0869 END IF
0870 260 CONTINUE
0871
0872 DO 320 KI = 1 , IBULGE
0872
0873 IF( KCOL( KI ).GT.KP2COL( KI ) )
0873
0874 $ GO TO 320
0875 IF(( MYCOL.NE.ICURCOL( KI ) ) .AND.
0876 $( RIGHT.NE.ICURCOL( KI ) ) )GO TO 320
0876
0877 ISTART = MAX( K1( KI ) , M )
0878 ISTOP = MIN( K2( KI ) , I - 1 )
0879 IF( MOD( ISTART - 1 , HBL ).GE.HBL - 2 ) THEN
0880
0881 * INFO is found in a buffer
0882
0882
0883 ISPEC = 1
0884 ELSE
0885
0886 * All INFO is local
0887
0887
0888 ISPEC = 0
0889 END IF
0890
0891 DO 310 K = ISTART , ISTOP
0892
0892
0893 V2 = WORK( VECSIDX + ( K - 1 )*3 + 1 )
0894 V3 = WORK( VECSIDX + ( K - 1 )*3 + 2 )
0895 T1 = WORK( VECSIDX + ( K - 1 )*3 + 3 )
0896 NR = MIN( 3 , I - K + 1 )
0897 IF(( NR.EQ.3 ) .AND.( KCOL( KI ).LE.KP2COL( KI ) ) )
0897
0898 $ THEN
0899
0900 IF(( K.LT.ISTOP ) .AND.
0901 $( MOD( K - 1 , HBL ).LT.HBL - 2 ) ) THEN
0901
0902 ITMP1 = MIN( ISTART + 1 , I ) - 1
0903 ELSE
0903
0904 IF( MOD( K - 1 , HBL ).LT.HBL - 2 ) THEN
0904
0905 ITMP1 = MIN( K + 3 , I )
0906 END IF
0907 IF( MOD( K - 1 , HBL ).EQ.HBL - 2 ) THEN
0907
0908 ITMP1 = MAX( I1 , K - 1 ) - 1
0909 END IF
0910 IF( MOD( K - 1 , HBL ).EQ.HBL - 1 ) THEN
0910
0911 ITMP1 = MAX( I1 , K - 2 ) - 1
0912 END IF
0913 END IF
0914 ICOL1 = KCOL( KI ) + K - ISTART
0915 ICOL2 = KP2COL( KI ) + K - ISTART
0916 CALL INFOG1L( I1 , HBL , NPROW , MYROW , 0 , IROW1 ,
0917 $ IROW2 )
0918 IROW2 = NUMROC( ITMP1 , HBL , MYROW , 0 , NPROW )
0919 IF(( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND.
0920 $( NPCOL.GT.1 ) ) THEN
0920
0921 IF( ICOL1.EQ.ICOL2 ) THEN
0921
0922 CALL SGESD2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0923 $ A(( ICOL1 - 1 )*LDA + IROW1 ) ,
0924 $ LDA , MYROW , LEFT )
0925 CALL SGERV2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0926 $ A(( ICOL1 - 1 )*LDA + IROW1 ) ,
0927 $ LDA , MYROW , LEFT )
0928 ELSE
0928
0929 CALL SGERV2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0930 $ WORK( ICBUF + 1 ) , IROW2 - IROW1 + 1 ,
0931 $ MYROW , RIGHT )
0932 T2 = T1*V2
0933 T3 = T1*V3
0934 DO 270 J = IROW1 , IROW2
0934
0935 SUM = A(( ICOL1 - 1 )*LDA + J ) +
0936 $ V2*A( ICOL1*LDA + J ) +
0937 $ V3*WORK( ICBUF + J - IROW1 + 1 )
0938 A(( ICOL1 - 1 )*LDA + J ) = A(( ICOL1 - 1 )*
0939 $ LDA + J ) - SUM*T1
0940 A( ICOL1*LDA + J ) = A( ICOL1*LDA + J ) -
0941 $ SUM*T2
0942 WORK( ICBUF + J - IROW1 + 1 ) = WORK( ICBUF + J -
0943 $ IROW1 + 1 ) - SUM*T3
0944 270 CONTINUE
0944
0945 CALL SGESD2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0946 $ WORK( ICBUF + 1 ) , IROW2 - IROW1 + 1 ,
0947 $ MYROW , RIGHT )
0948 END IF
0949 END IF
0950 IF(( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND.
0951 $( NPCOL.GT.1 ) ) THEN
0951
0952 IF( ICOL1.EQ.ICOL2 ) THEN
0952
0953 CALL SGESD2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0954 $ A(( ICOL1 - 1 )*LDA + IROW1 ) ,
0955 $ LDA , MYROW , RIGHT )
0956 CALL SGERV2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0957 $ A(( ICOL1 - 1 )*LDA + IROW1 ) ,
0958 $ LDA , MYROW , RIGHT )
0959 ELSE
0959
0960 CALL SGERV2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0961 $ WORK( ICBUF + 1 ) , IROW2 - IROW1 + 1 ,
0962 $ MYROW , LEFT )
0963 T2 = T1*V2
0964 T3 = T1*V3
0965 DO 280 J = IROW1 , IROW2
0965
0966 SUM = WORK( ICBUF + J - IROW1 + 1 ) +
0967 $ V2*A(( ICOL1 - 1 )*LDA + J ) +
0968 $ V3*A( ICOL1*LDA + J )
0969 WORK( ICBUF + J - IROW1 + 1 ) = WORK( ICBUF + J -
0970 $ IROW1 + 1 ) - SUM*T1
0971 A(( ICOL1 - 1 )*LDA + J ) = A(( ICOL1 - 1 )*
0972 $ LDA + J ) - SUM*T2
0973 A( ICOL1*LDA + J ) = A( ICOL1*LDA + J ) -
0974 $ SUM*T3
0975 280 CONTINUE
0975
0976 CALL SGESD2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0977 $ WORK( ICBUF + 1 ) , IROW2 - IROW1 + 1 ,
0978 $ MYROW , LEFT )
0979 END IF
0980 END IF
0981
0982 * If we want Z and we haven't already done any Z
0983 IF(( WANTZ ) .AND.( MOD( K - 1 ,
0984 $ HBL ).GE.HBL - 2 ) .AND.( NPCOL.GT.1 ) ) THEN
0985
0986 * Accumulate transformations in the matrix Z
0987
0988 IROW1 = LILOZ
0989 IROW2 = LIHIZ
0990 IF( MOD( K - 1 , HBL ).EQ.HBL - 2 ) THEN
0990
0991 IF( ICOL1.EQ.ICOL2 ) THEN
0991
0992 CALL SGESD2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0993 $ Z(( ICOL1 - 1 )*LDZ + IROW1 ) ,
0994 $ LDZ , MYROW , LEFT )
0995 CALL SGERV2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
0996 $ Z(( ICOL1 - 1 )*LDZ + IROW1 ) ,
0997 $ LDZ , MYROW , LEFT )
0998 ELSE
0998
0999 CALL SGERV2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
1000 $ WORK( ICBUF + 1 ) ,
1001 $ IROW2 - IROW1 + 1 , MYROW ,
1002 $ RIGHT )
1003 T2 = T1*V2
1004 T3 = T1*V3
1005 ICOL1 =( ICOL1 - 1 )*LDZ
1006 DO 290 J = IROW1 , IROW2
1006
1007 SUM = Z( ICOL1 + J ) +
1008 $ V2*Z( ICOL1 + J + LDZ ) +
1009 $ V3*WORK( ICBUF + J - IROW1 + 1 )
1010 Z( J + ICOL1 ) = Z( J + ICOL1 ) - SUM*T1
1011 Z( J + ICOL1 + LDZ ) = Z( J + ICOL1 + LDZ ) -
1012 $ SUM*T2
1013 WORK( ICBUF + J - IROW1 + 1 ) = WORK( ICBUF +
1014 $ J - IROW1 + 1 ) - SUM*T3
1015 290 CONTINUE
1015
1016 CALL SGESD2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
1017 $ WORK( ICBUF + 1 ) ,
1018 $ IROW2 - IROW1 + 1 , MYROW ,
1019 $ RIGHT )
1020 END IF
1021 END IF
1022 IF( MOD( K - 1 , HBL ).EQ.HBL - 1 ) THEN
1022
1023 IF( ICOL1.EQ.ICOL2 ) THEN
1023
1024 CALL SGESD2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
1025 $ Z(( ICOL1 - 1 )*LDZ + IROW1 ) ,
1026 $ LDZ , MYROW , RIGHT )
1027 CALL SGERV2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
1028 $ Z(( ICOL1 - 1 )*LDZ + IROW1 ) ,
1029 $ LDZ , MYROW , RIGHT )
1030 ELSE
1030
1031 CALL SGERV2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
1032 $ WORK( ICBUF + 1 ) ,
1033 $ IROW2 - IROW1 + 1 , MYROW , LEFT )
1034 T2 = T1*V2
1035 T3 = T1*V3
1036 ICOL1 =( ICOL1 - 1 )*LDZ
1037 DO 300 J = IROW1 , IROW2
1037
1038 SUM = WORK( ICBUF + J - IROW1 + 1 ) +
1039 $ V2*Z( J + ICOL1 ) +
1040 $ V3*Z( J + ICOL1 + LDZ )
1041 WORK( ICBUF + J - IROW1 + 1 ) = WORK( ICBUF +
1042 $ J - IROW1 + 1 ) - SUM*T1
1043 Z( J + ICOL1 ) = Z( J + ICOL1 ) - SUM*T2
1044 Z( J + ICOL1 + LDZ ) = Z( J + ICOL1 + LDZ ) -
1045 $ SUM*T3
1046 300 CONTINUE
1046
1047 CALL SGESD2D( CONTXT , IROW2 - IROW1 + 1 , 1 ,
1048 $ WORK( ICBUF + 1 ) ,
1049 $ IROW2 - IROW1 + 1 , MYROW , LEFT )
1050 END IF
1051 END IF
1052 END IF
1053 IF( ICURCOL( KI ).EQ.MYCOL ) THEN
1053
1054 IF(( ISPEC.EQ.0 ) .OR.( NPCOL.EQ.1 ) ) THEN
1054
1055 LOCALK2( KI ) = LOCALK2( KI ) + 1
1056 END IF
1057 ELSE
1057
1058 IF(( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND.
1059 $( ICURCOL( KI ).EQ.RIGHT ) ) THEN
1059
1060 IF( K.GT.M ) THEN
1060
1061 LOCALK2( KI ) = LOCALK2( KI ) + 2
1062 ELSE
1062
1063 LOCALK2( KI ) = LOCALK2( KI ) + 1
1064 END IF
1065 END IF
1066 IF(( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND.
1067 $( I - K.EQ.2 ) .AND.( ICURCOL( KI ).EQ.
1068 $RIGHT ) ) THEN
1069 LOCALK2( KI ) = LOCALK2( KI ) + 2
1070 END IF
1071 END IF
1072 END IF
1073 310 CONTINUE
1074 320 CONTINUE
1075
1076 * Column work done
1077
1078 330 CONTINUE
1079
1080 * Now do NR = 2 work
1081
1082 DO 410 KI = 1 , IBULGE
1082
1083 ISTART = MAX( K1( KI ) , M )
1084 ISTOP = MIN( K2( KI ) , I - 1 )
1085 IF( MOD( ISTART - 1 , HBL ).GE.HBL - 2 ) THEN
1086
1087 * INFO is found in a buffer
1088
1088
1089 ISPEC = 1
1090 ELSE
1091
1092 * All INFO is local
1093
1093
1094 ISPEC = 0
1095 END IF
1096
1097 DO 400 K = ISTART , ISTOP
1098
1098
1099 V2 = WORK( VECSIDX + ( K - 1 )*3 + 1 )
1100 V3 = WORK( VECSIDX + ( K - 1 )*3 + 2 )
1101 T1 = WORK( VECSIDX + ( K - 1 )*3 + 3 )
1102 NR = MIN( 3 , I - K + 1 )
1103 IF( NR.EQ.2 ) THEN
1103
1104 IF( ICURROW( KI ).EQ.MYROW ) THEN
1104
1105 T2 = T1*V2
1106 END IF
1107 IF( ICURCOL( KI ).EQ.MYCOL ) THEN
1107
1108 T2 = T1*V2
1109 END IF
1110
1111 * Apply G from the left to transform the rows of the matrix
1112 * in columns K to I2.
1113
1114 CALL INFOG1L( K , HBL , NPCOL , MYCOL , 0 , LILOH ,
1115 $ LIHIH )
1116 LIHIH = NUMROC( I2 , HBL , MYCOL , 0 , NPCOL )
1117 CALL INFOG1L( 1 , HBL , NPROW , MYROW , 0 , ITMP2 ,
1118 $ ITMP1 )
1119 ITMP1 = NUMROC( K + 1 , HBL , MYROW , 0 , NPROW )
1120 IF( ICURROW( KI ).EQ.MYROW ) THEN
1120
1121 IF(( ISPEC.EQ.0 ) .OR.( NPROW.EQ.1 ) .OR.
1122 $( MOD( K - 1 , HBL ).EQ.HBL - 2 ) ) THEN
1122
1123 ITMP1 = ITMP1 - 1
1124 DO 340 J =( LILOH - 1 )*LDA ,
1125 $( LIHIH - 1 )*LDA , LDA
1125
1126 SUM = A( ITMP1 + J ) + V2*A( ITMP1 + 1 + J )
1127 A( ITMP1 + J ) = A( ITMP1 + J ) - SUM*T1
1128 A( ITMP1 + 1 + J ) = A( ITMP1 + 1 + J ) - SUM*T2
1129 340 CONTINUE
1129
1130 ELSE
1130
1131 IF( MOD( K - 1 , HBL ).EQ.HBL - 1 ) THEN
1131
1132 CALL SGERV2D( CONTXT , 1 , LIHIH - LILOH + 1 ,
1133 $ WORK( IRBUF + 1 ) , 1 , UP ,
1134 $ MYCOL )
1135 DO 350 J = LILOH , LIHIH
1135
1136 SUM = WORK( IRBUF + J - LILOH + 1 ) +
1137 $ V2*A(( J - 1 )*LDA + ITMP1 )
1138 WORK( IRBUF + J - LILOH + 1 ) = WORK( IRBUF +
1139 $ J - LILOH + 1 ) - SUM*T1
1140 A(( J - 1 )*LDA + ITMP1 ) = A(( J - 1 )*
1141 $ LDA + ITMP1 ) - SUM*T2
1142 350 CONTINUE
1142
1143 CALL SGESD2D( CONTXT , 1 , LIHIH - LILOH + 1 ,
1144 $ WORK( IRBUF + 1 ) , 1 , UP ,
1145 $ MYCOL )
1146 END IF
1147 END IF
1148 ELSE
1148
1149 IF(( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND.
1150 $( ICURROW( KI ).EQ.DOWN ) ) THEN
1150
1151 CALL SGESD2D( CONTXT , 1 , LIHIH - LILOH + 1 ,
1152 $ A(( LILOH - 1 )*LDA + ITMP1 ) ,
1153 $ LDA , DOWN , MYCOL )
1154 CALL SGERV2D( CONTXT , 1 , LIHIH - LILOH + 1 ,
1155 $ A(( LILOH - 1 )*LDA + ITMP1 ) ,
1156 $ LDA , DOWN , MYCOL )
1157 END IF
1158 END IF
1159
1160 * Apply G from the right to transform the columns of the
1161 * matrix in rows I1 to MIN(K + 3 , I).
1162
1163 CALL INFOG1L( I1 , HBL , NPROW , MYROW , 0 , LILOH ,
1164 $ LIHIH )
1165 LIHIH = NUMROC( I , HBL , MYROW , 0 , NPROW )
1166
1167 IF( ICURCOL( KI ).EQ.MYCOL ) THEN
1168 * LOCAL A(LILOZ : LIHIZ , LOCALK2 : LOCALK2 + 2)
1168
1169 IF(( ISPEC.EQ.0 ) .OR.( NPCOL.EQ.1 ) .OR.
1170 $( MOD( K - 1 , HBL ).EQ.HBL - 2 ) ) THEN
1170
1171 CALL INFOG1L( K , HBL , NPCOL , MYCOL , 0 , ITMP1 ,
1172 $ ITMP2 )
1173 ITMP2 = NUMROC( K + 1 , HBL , MYCOL , 0 , NPCOL )
1174 DO 360 J = LILOH , LIHIH
1174
1175 SUM = A(( ITMP1 - 1 )*LDA + J ) +
1176 $ V2*A( ITMP1*LDA + J )
1177 A(( ITMP1 - 1 )*LDA + J ) = A(( ITMP1 - 1 )*
1178 $ LDA + J ) - SUM*T1
1179 A( ITMP1*LDA + J ) = A( ITMP1*LDA + J ) -
1180 $ SUM*T2
1181 360 CONTINUE
1181
1182 ELSE
1182
1183 ITMP1 = LOCALK2( KI )
1184 IF( MOD( K - 1 , HBL ).EQ.HBL - 1 ) THEN
1184
1185 CALL SGERV2D( CONTXT , LIHIH - LILOH + 1 , 1 ,
1186 $ WORK( ICBUF + 1 ) ,
1187 $ LIHIH - LILOH + 1 , MYROW , LEFT )
1188 DO 370 J = LILOH , LIHIH
1188
1189 SUM = WORK( ICBUF + J ) +
1190 $ V2*A(( ITMP1 - 1 )*LDA + J )
1191 WORK( ICBUF + J ) = WORK( ICBUF + J ) -
1192 $ SUM*T1
1193 A(( ITMP1 - 1 )*LDA + J )
1194 $ = A(( ITMP1 - 1 )*LDA + J ) - SUM*T2
1195 370 CONTINUE
1195
1196 CALL SGESD2D( CONTXT , LIHIH - LILOH + 1 , 1 ,
1197 $ WORK( ICBUF + 1 ) ,
1198 $ LIHIH - LILOH + 1 , MYROW , LEFT )
1199 END IF
1200 END IF
1201 ELSE
1201
1202 IF(( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND.
1203 $( ICURCOL( KI ).EQ.RIGHT ) ) THEN
1203
1204 ITMP1 = KCOL( KI )
1205 CALL SGESD2D( CONTXT , LIHIH - LILOH + 1 , 1 ,
1206 $ A(( ITMP1 - 1 )*LDA + LILOH ) ,
1207 $ LDA , MYROW , RIGHT )
1208 CALL INFOG1L( K , HBL , NPCOL , MYCOL , 0 , ITMP1 ,
1209 $ ITMP2 )
1210 ITMP2 = NUMROC( K + 1 , HBL , MYCOL , 0 , NPCOL )
1211 CALL SGERV2D( CONTXT , LIHIH - LILOH + 1 , 1 ,
1212 $ A(( ITMP1 - 1 )*LDA + LILOH ) ,
1213 $ LDA , MYROW , RIGHT )
1214 END IF
1215 END IF
1216
1217 IF( WANTZ ) THEN
1218
1219 * Accumulate transformations in the matrix Z
1220
1220
1221 IF( ICURCOL( KI ).EQ.MYCOL ) THEN
1222 * LOCAL Z(LILOZ : LIHIZ , LOCALK2 : LOCALK2 + 2)
1222
1223 IF(( ISPEC.EQ.0 ) .OR.( NPCOL.EQ.1 ) .OR.
1224 $( MOD( K - 1 , HBL ).EQ.HBL - 2 ) ) THEN
1224
1225 ITMP1 = KCOL( KI ) + K - ISTART
1226 ITMP1 =( ITMP1 - 1 )*LDZ
1227 DO 380 J = LILOZ , LIHIZ
1227
1228 SUM = Z( J + ITMP1 ) +
1229 $ V2*Z( J + ITMP1 + LDZ )
1230 Z( J + ITMP1 ) = Z( J + ITMP1 ) - SUM*T1
1231 Z( J + ITMP1 + LDZ ) = Z( J + ITMP1 + LDZ ) -
1232 $ SUM*T2
1233 380 CONTINUE
1233
1234 LOCALK2( KI ) = LOCALK2( KI ) + 1
1235 ELSE
1235
1236 ITMP1 = LOCALK2( KI )
1237 * IF WE ACTUALLY OWN COLUMN K
1238 IF( MOD( K - 1 , HBL ).EQ.HBL - 1 ) THEN
1238
1239 CALL SGERV2D( CONTXT , LIHIZ - LILOZ + 1 , 1 ,
1240 $ WORK( ICBUF + 1 ) , LDZ ,
1241 $ MYROW , LEFT )
1242 ITMP1 =( ITMP1 - 1 )*LDZ
1243 DO 390 J = LILOZ , LIHIZ
1243
1244 SUM = WORK( ICBUF + J ) +
1245 $ V2*Z( J + ITMP1 )
1246 WORK( ICBUF + J ) = WORK( ICBUF + J ) -
1247 $ SUM*T1
1248 Z( J + ITMP1 ) = Z( J + ITMP1 ) - SUM*T2
1249 390 CONTINUE
1249
1250 CALL SGESD2D( CONTXT , LIHIZ - LILOZ + 1 , 1 ,
1251 $ WORK( ICBUF + 1 ) , LDZ ,
1252 $ MYROW , LEFT )
1253 LOCALK2( KI ) = LOCALK2( KI ) + 1
1254 END IF
1255 END IF
1256 ELSE
1257
1258 * NO WORK BUT NEED TO UPDATE ANYWAY????
1259
1259
1260 IF(( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND.
1261 $( ICURCOL( KI ).EQ.RIGHT ) ) THEN
1261
1262 ITMP1 = KCOL( KI )
1263 ITMP1 =( ITMP1 - 1 )*LDZ
1264 CALL SGESD2D( CONTXT , LIHIZ - LILOZ + 1 , 1 ,
1265 $ Z( LILOZ + ITMP1 ) , LDZ ,
1266 $ MYROW , RIGHT )
1267 CALL SGERV2D( CONTXT , LIHIZ - LILOZ + 1 , 1 ,
1268 $ Z( LILOZ + ITMP1 ) , LDZ ,
1269 $ MYROW , RIGHT )
1270 LOCALK2( KI ) = LOCALK2( KI ) + 1
1271 END IF
1272 END IF
1273 END IF
1274 END IF
1275 400 CONTINUE
1276
1277 * Adjust local information for this bulge
1278
1279 IF( NPROW.EQ.1 ) THEN
1279
1280 KROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1
1281 KP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1
1282 END IF
1283 IF(( MOD( K1( KI ) - 1 , HBL ).LT.HBL - 2 ) .AND.
1284 $( ICURROW( KI ).EQ.MYROW ) .AND.( NPROW.GT.1 ) )
1285 $THEN
1286 KROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1
1287 END IF
1288 IF(( MOD( K2( KI ) , HBL ).LT.HBL - 2 ) .AND.
1289 $( ICURROW( KI ).EQ.MYROW ) .AND.( NPROW.GT.1 ) )
1290 $THEN
1291 KP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1
1292 END IF
1293 IF(( MOD( K1( KI ) - 1 , HBL ).GE.HBL - 2 ) .AND.
1294 $(( MYROW.EQ.ICURROW( KI ) ) .OR.( DOWN.EQ.
1295 $ICURROW( KI ) ) ) .AND.( NPROW.GT.1 ) ) THEN
1296 CALL INFOG1L( K2( KI ) + 1 , HBL , NPROW , MYROW , 0 ,
1297 $KROW( KI ) , ITMP2 )
1298 ITMP2 = NUMROC( N , HBL , MYROW , 0 , NPROW )
1299 END IF
1300 IF(( MOD( K2( KI ) , HBL ).GE.HBL - 2 ) .AND.
1301 $(( MYROW.EQ.ICURROW( KI ) ) .OR.( UP.EQ.
1302 $ICURROW( KI ) ) ) .AND.( NPROW.GT.1 ) ) THEN
1303 CALL INFOG1L( 1 , HBL , NPROW , MYROW , 0 , ITMP2 ,
1304 $KP2ROW( KI ) )
1305 KP2ROW( KI ) = NUMROC( K2( KI ) + 3 , HBL , MYROW , 0 ,
1306 $NPROW )
1307 END IF
1308 IF( NPCOL.EQ.1 ) THEN
1308
1309 KCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1
1310 KP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1
1311 END IF
1312 IF(( MOD( K1( KI ) - 1 , HBL ).LT.HBL - 2 ) .AND.
1313 $( ICURCOL( KI ).EQ.MYCOL ) .AND.( NPCOL.GT.1 ) )
1314 $THEN
1315 KCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1
1316 END IF
1317 IF(( MOD( K2( KI ) , HBL ).LT.HBL - 2 ) .AND.
1318 $( ICURCOL( KI ).EQ.MYCOL ) .AND.( NPCOL.GT.1 ) )
1319 $THEN
1320 KP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1
1321 END IF
1322 IF(( MOD( K1( KI ) - 1 , HBL ).GE.HBL - 2 ) .AND.
1323 $(( MYCOL.EQ.ICURCOL( KI ) ) .OR.( RIGHT.EQ.
1324 $ICURCOL( KI ) ) ) .AND.( NPCOL.GT.1 ) ) THEN
1325 CALL INFOG1L( K2( KI ) + 1 , HBL , NPCOL , MYCOL , 0 ,
1326 $KCOL( KI ) , ITMP2 )
1327 ITMP2 = NUMROC( N , HBL , MYCOL , 0 , NPCOL )
1328 END IF
1329 IF(( MOD( K2( KI ) , HBL ).GE.HBL - 2 ) .AND.
1330 $(( MYCOL.EQ.ICURCOL( KI ) ) .OR.( LEFT.EQ.
1331 $ICURCOL( KI ) ) ) .AND.( NPCOL.GT.1 ) ) THEN
1332 CALL INFOG1L( 1 , HBL , NPCOL , MYCOL , 0 , ITMP2 ,
1333 $KP2COL( KI ) )
1334 KP2COL( KI ) = NUMROC( K2( KI ) + 3 , HBL , MYCOL , 0 ,
1335 $NPCOL )
1336 END IF
1337 K1( KI ) = K2( KI ) + 1
1338 ISTOP = MIN( K1( KI ) + ROTN - MOD( K1( KI ) , ROTN ) , I - 2 )
1339 ISTOP = MIN( ISTOP , K1( KI ) + HBL - 3 -
1340 $MOD( K1( KI ) - 1 , HBL ) )
1341 ISTOP = MIN( ISTOP , I2 - 2 )
1342 ISTOP = MAX( ISTOP , K1( KI ) )
1343 * ISTOP = MIN( ISTOP , I - 1 )
1344 K2( KI ) = ISTOP
1345 IF( K1( KI ).EQ.ISTOP ) THEN
1345
1346 IF(( MOD( ISTOP - 1 , HBL ).EQ.HBL - 2 ) .AND.
1347 $( I - ISTOP.GT.1 ) ) THEN
1348
1349 * Next step switches rows & cols
1350
1350
1351 ICURROW( KI ) = MOD( ICURROW( KI ) + 1 , NPROW )
1352 ICURCOL( KI ) = MOD( ICURCOL( KI ) + 1 , NPCOL )
1353 END IF
1354 END IF
1355 410 CONTINUE
1356 IF( K2( IBULGE ).LE.I - 1 )
1356
1357 $ GO TO 40
1358 END IF
1359
1360 420 CONTINUE
1361
1362 * Failure to converge in remaining number of iterations
1363
1364 INFO = I
1365 RETURN
1366
1367 430 CONTINUE
1368
1369 IF( L.EQ.I ) THEN
1370
1371 * H(I , I - 1) is negligible : one eigenvalue has converged.
1372
1372
1373 CALL INFOG2L( I , I , DESCA , NPROW , NPCOL , MYROW , MYCOL , IROW ,
1374 $ ICOL , ITMP1 , ITMP2 )
1375 IF(( MYROW.EQ.ITMP1 ) .AND.( MYCOL.EQ.ITMP2 ) ) THEN
1375
1376 WR( I ) = A(( ICOL - 1 )*LDA + IROW )
1377 ELSE
1377
1378 WR( I ) = ZERO
1379 END IF
1380 WI( I ) = ZERO
1381 ELSE IF( L.EQ.I - 1 ) THEN
1382
1383 * H(I - 1 , I - 2) is negligible : a pair of eigenvalues have converged.
1384
1384
1385 WR( I - 1 ) = ZERO
1386 WR( I ) = ZERO
1387 WI( I - 1 ) = ZERO
1388 WI( I ) = ZERO
1389 MODKM1 = MOD( I - 1 + HBL , HBL )
1390 CALL INFOG2L( I - 1 , I - 1 , DESCA , NPROW , NPCOL , MYROW , MYCOL ,
1391 $ IROW1 , ICOL1 , II , JJ )
1392 IF(( MYROW.EQ.II ) .AND.( MYCOL.EQ.JJ ) ) THEN
1392
1393 H11 = A(( ICOL1 - 1 )*LDA + IROW1 )
1394 IF( MODKM1.NE.0 ) THEN
1394
1395 H21 = A(( ICOL1 - 1 )*LDA + IROW1 + 1 )
1396 H12 = A( ICOL1*LDA + IROW1 )
1397 H22 = A( ICOL1*LDA + IROW1 + 1 )
1398 ELSE
1398
1399 IF( NPROW.GT.1 ) THEN
1399
1400 CALL SGERV2D( CONTXT , 1 , 1 , H21 , 1 , DOWN , MYCOL )
1401 ELSE
1401
1402 H21 = A(( ICOL1 - 1 )*LDA + IROW1 + 1 )
1403 END IF
1404 IF( NPCOL.GT.1 ) THEN
1404
1405 CALL SGERV2D( CONTXT , 1 , 1 , H12 , 1 , MYROW , RIGHT )
1406 ELSE
1406
1407 H12 = A( ICOL1*LDA + IROW1 )
1408 END IF
1409 IF( NUM.GT.1 ) THEN
1409
1410 CALL SGERV2D( CONTXT , 1 , 1 , H22 , 1 , DOWN , RIGHT )
1411 ELSE
1411
1412 H22 = A( ICOL1*LDA + IROW1 + 1 )
1413 END IF
1414 END IF
1415 H00 = HALF*( H11 + H22 )
1416 H10 = H11*H22 - H12*H21
1417 ELSE
1417
1418 IF( MODKM1.EQ.0 ) THEN
1418
1419 IF(( NPROW.GT.1 ) .AND.( MYCOL.EQ.JJ ) .AND.
1420 $( UP.EQ.II ) ) THEN
1420
1421 CALL INFOG2L( I , I - 1 , DESCA , NPROW , NPCOL , MYROW ,
1422 $ MYCOL , IROW1 , ICOL1 , ITMP1 , ITMP2 )
1423 CALL SGESD2D( CONTXT , 1 , 1 ,
1424 $ A(( ICOL1 - 1 )*LDA + IROW1 ) , 1 , II , JJ )
1425 END IF
1426 IF(( NPCOL.GT.1 ) .AND.( LEFT.EQ.JJ ) .AND.
1427 $( MYROW.EQ.II ) ) THEN
1427
1428 CALL INFOG2L( I - 1 , I , DESCA , NPROW , NPCOL , MYROW ,
1429 $ MYCOL , IROW1 , ICOL1 , ITMP1 , ITMP2 )
1430 CALL SGESD2D( CONTXT , 1 , 1 ,
1431 $ A(( ICOL1 - 1 )*LDA + IROW1 ) , 1 , II , JJ )
1432 END IF
1433 IF(( NUM.GT.1 ) .AND.( LEFT.EQ.JJ ) .AND.
1434 $( UP.EQ.II ) ) THEN
1434
1435 CALL INFOG2L( I , I , DESCA , NPROW , NPCOL , MYROW , MYCOL ,
1436 $ IROW1 , ICOL1 , ITMP1 , ITMP2 )
1437 CALL SGESD2D( CONTXT , 1 , 1 ,
1438 $ A(( ICOL1 - 1 )*LDA + IROW1 ) , 1 , II , JJ )
1439 END IF
1440 END IF
1441 H00 = ZERO
1442 H10 = ZERO
1443 END IF
1444 H21 = H00*H00 - H10
1445 IF( H21.GE.ZERO ) THEN
1445
1446 H21 = SQRT( H21 )
1447 WR( I - 1 ) = H00 + H21
1448 WI( I - 1 ) = ZERO
1449 WR( I ) = H00 - H21
1450 WI( I ) = ZERO
1451 ELSE
1451
1452 H21 = SQRT( ABS( H21 ) )
1453 WR( I - 1 ) = H00
1454 WI( I - 1 ) = H21
1455 WR( I ) = H00
1456 WI( I ) = - H21
1457 END IF
1458 ELSE
1459
1460 * Find the eigenvalues in H(L : I , L : I) , L < I - 1
1461
1461
1462 JBLK = I - L + 1
1463 IF( JBLK.LE.2*IBLK ) THEN
1463
1464 CALL PSLACP3 ( I - L + 1 , L , A , DESCA , S1 , 2*IBLK , 0 , 0 , 0 )
1465 CALL SLAHQR( .FALSE. , .FALSE. , JBLK , 1 , JBLK , S1 , 2*IBLK ,
1466 $ WR( L ) , WI( L ) , 1 , JBLK , Z , LDZ , IERR )
1467 IF( NODE.NE.0 ) THEN
1468
1469 * Erase the eigenvalues
1470
1470
1471 DO 440 K = L , I
1471
1472 WR( K ) = ZERO
1473 WI( K ) = ZERO
1474 440 CONTINUE
1474
1475 END IF
1476 END IF
1477 END IF
1478
1479 * Decrement number of remaining iterations , and return to start of
1480 * the main loop with new value of I.
1481
1482 ITN = ITN - ITS
1483 IF( M.EQ.L - 10 ) THEN
1483
1484 I = L - 1
1485 ELSE
1485
1486 I = M
1487 END IF
1488 * I = L - 1
1489 GO TO 10
1490
1491 450 CONTINUE
1492 CALL SGSUM2D( CONTXT , 'All' , ' ' , N , 1 , WR , N , - 1 , - 1 )
1493 CALL SGSUM2D( CONTXT , 'All' , ' ' , N , 1 , WI , N , - 1 , - 1 )
1494 RETURN
1495
1496 * END OF PSLAHQR
1497
1498 END138
209
|
|
Variables in Routine PSLAHQR()
| Summary Report |
| Data Type | Quantity | Size(byte) |
| INTEGER | 52 | 208 |
| LOGICAL | 2 | 2 |
| REAL | 23 | 92 |
| TOTAL | 77 | 302 |
List of Variables
INTEGER
| BLOCK_CYCLIC_2D | CSRC_ | CTXT_ | DESCA( * ) | DESCZ( * ) |
| DLEN_ | DTYPE_ | I | IBLK | ICOL1 |
| ICOL2 | ICURCOL | ICURROW | IHI | IHIZ |
| II | ILO | ILOZ | ILWORK | INFO |
| IROW1 | IROW2 | ISPEC | ISTART | ISTOP |
| ITMP1 | ITMP2 | ITN | IWORK( * ) | J |
| JBLK | JJ | K | K1 | K2 |
| KCOL | KI | KP2COL | KP2ROW | KROW |
| LIHIH | LLD_ | LOCALK2 | LWORK | M_ |
| MB_ | MODKM1 | N | N_ | NB_ |
| NR | RSRC_ | | | |
LOGICAL
REAL
| $ | A( * ) | CONST | H00 | H10 |
| H11 | H12 | H21 | H22 | HALF |
| ONE | SMALLA | SUM | T1 | T2 |
| T3 | V2 | V3 | WI( * ) | WORK( * ) |
| WR( * ) | Z( * ) | ZERO | | |
Variables Dependence Graph
Put the mouse over a right hand side variable to display the corresponding line of the dependence | | - | | - | - | | A | <--- | AA( ITMP1+J ) = A( ITMP1+J ) - SUM*T1{2A( ITMP1+1+J ) = A( ITMP1+1+J ) - SUM*T2, 3A( ( J-1 )*LDA+ITMP1 ) = A( ( J-1 )*, 4A( ( ITMP1-1 )*LDA+J ) = A( ( ITMP1-1 )*, 5A( ITMP1*LDA+J ) = A( ITMP1*LDA+J ) -, 6A( ( ITMP1-1 )*LDA+J ), 7A( ( ICOL1-2 )*LDA+IROW1 ) = -A( ( ICOL1-2 )*, 8A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*LDA+, 9A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*LDA+, 10A( ( J-1 )*LDA+IROW1+2 ) = A( ( J-1 )*LDA+, 11A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*LDA+, 12A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) - SUM*T2, 13A( ( ICOL1+1 )*LDA+J ) = A( ( ICOL1+1 )*LDA+, 14A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 15A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*, 16A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 17A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*, 18A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*, 19A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -, 20A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*, 21A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -}, ICOL1A( ( ICOL1-2 )*LDA+IROW1 ) = -A( ( ICOL1-2 )*{2A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*LDA+, 3A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) - SUM*T2, 4A( ( ICOL1+1 )*LDA+J ) = A( ( ICOL1+1 )*LDA+, 5A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*, 6A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -, 7A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*, 8A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -}, IROW1A( ( ICOL1-2 )*LDA+IROW1 ) = -A( ( ICOL1-2 )*{2A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*LDA+, 3A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*LDA+, 4A( ( J-1 )*LDA+IROW1+2 ) = A( ( J-1 )*LDA+, 5A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 6A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*, 7A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 8A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*}, ITMP1A( ITMP1+J ) = A( ITMP1+J ) - SUM*T1{2A( ITMP1+1+J ) = A( ITMP1+1+J ) - SUM*T2, 3A( ( J-1 )*LDA+ITMP1 ) = A( ( J-1 )*, 4A( ( ITMP1-1 )*LDA+J ) = A( ( ITMP1-1 )*, 5A( ITMP1*LDA+J ) = A( ITMP1*LDA+J ) -, 6A( ( ITMP1-1 )*LDA+J )}, JA( ITMP1+J ) = A( ITMP1+J ) - SUM*T1{2A( ITMP1+1+J ) = A( ITMP1+1+J ) - SUM*T2, 3A( ( J-1 )*LDA+ITMP1 ) = A( ( J-1 )*, 4A( ( ITMP1-1 )*LDA+J ) = A( ( ITMP1-1 )*, 5A( ITMP1*LDA+J ) = A( ITMP1*LDA+J ) -, 6A( ( ITMP1-1 )*LDA+J ), 7A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*LDA+, 8A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*LDA+, 9A( ( J-1 )*LDA+IROW1+2 ) = A( ( J-1 )*LDA+, 10A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*LDA+, 11A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) - SUM*T2, 12A( ( ICOL1+1 )*LDA+J ) = A( ( ICOL1+1 )*LDA+, 13A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 14A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*, 15A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 16A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*, 17A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*, 18A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -, 19A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*, 20A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -}, SUMA( ITMP1+J ) = A( ITMP1+J ) - SUM*T1{2A( ITMP1+1+J ) = A( ITMP1+1+J ) - SUM*T2, 3A( ( J-1 )*LDA+ITMP1 ) = A( ( J-1 )*, 4A( ( ITMP1-1 )*LDA+J ) = A( ( ITMP1-1 )*, 5A( ITMP1*LDA+J ) = A( ITMP1*LDA+J ) -, 6A( ( ITMP1-1 )*LDA+J ), 7A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*LDA+, 8A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*LDA+, 9A( ( J-1 )*LDA+IROW1+2 ) = A( ( J-1 )*LDA+, 10A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*LDA+, 11A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) - SUM*T2, 12A( ( ICOL1+1 )*LDA+J ) = A( ( ICOL1+1 )*LDA+, 13A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 14A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*, 15A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 16A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*, 17A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*, 18A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -, 19A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*, 20A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -}, T1A( ITMP1+J ) = A( ITMP1+J ) - SUM*T1{2A( ( ITMP1-1 )*LDA+J ) = A( ( ITMP1-1 )*, 3A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*LDA+, 4A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*LDA+, 5A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 6A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*}, T2A( ITMP1+1+J ) = A( ITMP1+1+J ) - SUM*T2{2A( ( J-1 )*LDA+ITMP1 ) = A( ( J-1 )*, 3A( ITMP1*LDA+J ) = A( ITMP1*LDA+J ) -, 4A( ( ITMP1-1 )*LDA+J ), 5A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*LDA+, 6A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) - SUM*T2, 7A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*, 8A( ( J-1 )*LDA+IROW1 ) = A( ( J-1 )*, 9A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -, 10A( ( ICOL1-1 )*LDA+J ) = A( ( ICOL1-1 )*}, T3A( ( J-1 )*LDA+IROW1+2 ) = A( ( J-1 )*LDA+{2A( ( ICOL1+1 )*LDA+J ) = A( ( ICOL1+1 )*LDA+, 3A( ( J-1 )*LDA+IROW1+1 ) = A( ( J-1 )*, 4A( ICOL1*LDA+J ) = A( ICOL1*LDA+J ) -} |
| H00 | <--- | H11H00 = HALF*( H11+H22 ), H22H00 = HALF*( H11+H22 ), HALFH00 = HALF*( H11+H22 ), ZEROH00 = ZERO |
| H10 | <--- | H11H10 = H11*H22 - H12*H21, H12H10 = H11*H22 - H12*H21, H21H10 = H11*H22 - H12*H21, H22H10 = H11*H22 - H12*H21, ZEROH10 = ZERO |
| H11 | <--- | AH11 = A( ( ICOL1-1 )*LDA+IROW1 ), ICOL1H11 = A( ( ICOL1-1 )*LDA+IROW1 ), IROW1H11 = A( ( ICOL1-1 )*LDA+IROW1 ) |
| H12 | <--- | AH12 = A( ICOL1*LDA+IROW1 ){2H12 = A( ICOL1*LDA+IROW1 )}, ICOL1H12 = A( ICOL1*LDA+IROW1 ){2H12 = A( ICOL1*LDA+IROW1 )}, IROW1H12 = A( ICOL1*LDA+IROW1 ){2H12 = A( ICOL1*LDA+IROW1 )} |
| H21 | <--- | H00H21 = H00*H00 - H10, H10H21 = H00*H00 - H10, H21H21 = SQRT( H21 ){2H21 = SQRT( ABS( H21 ) )}, AH21 = A( ( ICOL1-1 )*LDA+IROW1+1 ){2H21 = A( ( ICOL1-1 )*LDA+IROW1+1 )}, ICOL1H21 = A( ( ICOL1-1 )*LDA+IROW1+1 ){2H21 = A( ( ICOL1-1 )*LDA+IROW1+1 )}, IROW1H21 = A( ( ICOL1-1 )*LDA+IROW1+1 ){2H21 = A( ( ICOL1-1 )*LDA+IROW1+1 )} |
| H22 | <--- | AH22 = A( ICOL1*LDA+IROW1+1 ){2H22 = A( ICOL1*LDA+IROW1+1 )}, ICOL1H22 = A( ICOL1*LDA+IROW1+1 ){2H22 = A( ICOL1*LDA+IROW1+1 )}, IROW1H22 = A( ICOL1*LDA+IROW1+1 ){2H22 = A( ICOL1*LDA+IROW1+1 )} |
| ICOL1 | <--- | ICOL1ICOL1 = ( ICOL1-1 )*LDZ{2ICOL1 = ( ICOL1-1 )*LDZ, 3ICOL1 = ICOL1 + 1, 4ICOL1 = ICOL1 + 1}, ISTARTICOL1 = KCOL( KI ) + ISTOP - ISTART{2ICOL1 = KCOL( KI ) + K - ISTART}, ISTOPICOL1 = KCOL( KI ) + ISTOP - ISTART, KICOL1 = KCOL( KI ) + K - ISTART, KCOLICOL1 = KCOL( KI ){2ICOL1 = KCOL( KI ) + ISTOP - ISTART, 3ICOL1 = KCOL( KI ) + K - ISTART}, KIICOL1 = LOCALK2( KI ){2ICOL1 = KCOL( KI ), 3ICOL1 = KCOL( KI ) + ISTOP - ISTART, 4ICOL1 = KCOL( KI ) + K - ISTART}, LOCALK2ICOL1 = LOCALK2( KI ) |
| ICOL2 | <--- | ISTARTICOL2 = KP2COL( KI ) + K - ISTART, KICOL2 = KP2COL( KI ) + K - ISTART, KIICOL2 = KP2COL( KI ) + K - ISTART, KP2COLICOL2 = KP2COL( KI ) + K - ISTART |
| ICURCOL | <--- | ICURCOLICURCOL( KI ) = MOD( ICURCOL( KI )+1, NPCOL ), KIICURCOL( KI ) = MOD( ICURCOL( KI )+1, NPCOL ) |
| ICURROW | <--- | ICURROWICURROW( KI ) = MOD( ICURROW( KI )+1, NPROW ), KIICURROW( KI ) = MOD( ICURROW( KI )+1, NPROW ) |
| II | <--- | KIII = KROW( KI ), KROWII = KROW( KI ) |
| INFO | <--- | IINFO = I |
| IROW1 | <--- | IIROW1 = MIN( K2( KI )+1, I-1 ) + 1, IROW1IROW1 = IROW1 + 1{2IROW1 = IROW1 + 1}, ISTARTIROW1 = KROW( KI ) + K - ISTART{2IROW1 = KROW( KI ) + K - ISTART}, ITMP2IROW1 = ITMP2 + 1, KIROW1 = KROW( KI ) + K - ISTART{2IROW1 = KROW( KI ) + K - ISTART}, K2IROW1 = MIN( K2( KI )+1, I-1 ) + 1, KIIROW1 = KROW( KI ){2IROW1 = MIN( K2( KI )+1, I-1 ) + 1, 3IROW1 = KROW( KI ), 4IROW1 = KROW( KI ) + K - ISTART, 5IROW1 = KROW( KI ) + K - ISTART}, KROWIROW1 = KROW( KI ){2IROW1 = KROW( KI ), 3IROW1 = KROW( KI ) + K - ISTART, 4IROW1 = KROW( KI ) + K - ISTART} |
| IROW2 | <--- | IIROW2 = ITMP2 + MIN( K+3, I ) - ITMP1{2IROW2 = NUMROC( MIN( K+3, I ), HBL, MYROW, 0,}, IROW1IROW2 = IROW1 - 1{2IROW2 = IROW1 - 1}, ISTARTIROW2 = KP2ROW( KI ) + K - ISTART{2IROW2 = KP2ROW( KI ) + K - ISTART}, ITMP1IROW2 = NUMROC( ITMP1, HBL, MYROW, 0, NPROW ){2IROW2 = ITMP2 + MIN( K+3, I ) - ITMP1, 3IROW2 = NUMROC( ITMP1, HBL, MYROW, 0, NPROW )}, ITMP2IROW2 = ITMP2 + MIN( K+3, I ) - ITMP1, KIROW2 = KP2ROW( KI ) + K - ISTART{2IROW2 = KP2ROW( KI ) + K - ISTART, 3IROW2 = ITMP2 + MIN( K+3, I ) - ITMP1, 4IROW2 = NUMROC( MIN( K+3, I ), HBL, MYROW, 0,}, KIIROW2 = KP2ROW( KI ){2IROW2 = KP2ROW( KI ) + K - ISTART, 3IROW2 = KP2ROW( KI ) + K - ISTART}, KP2ROWIROW2 = KP2ROW( KI ){2IROW2 = KP2ROW( KI ) + K - ISTART, 3IROW2 = KP2ROW( KI ) + K - ISTART} |
| ISTART | <--- | K1ISTART = MAX( K1( KI ), M ){2ISTART = MAX( K1( KI ), M ), 3ISTART = MAX( K1( KI ), M ), 4ISTART = MAX( K1( KI ), M ), 5ISTART = MAX( K1( KI ), M ), 6ISTART = MAX( K1( KI ), M ), 7ISTART = MAX( K1( KI ), M ), 8ISTART = MAX( K1( KI ), M ), 9ISTART = MAX( K1( KI ), M )}, KIISTART = MAX( K1( KI ), M ){2ISTART = MAX( K1( KI ), M ), 3ISTART = MAX( K1( KI ), M ), 4ISTART = MAX( K1( KI ), M ), 5ISTART = MAX( K1( KI ), M ), 6ISTART = MAX( K1( KI ), M ), 7ISTART = MAX( K1( KI ), M ), 8ISTART = MAX( K1( KI ), M ), 9ISTART = MAX( K1( KI ), M )} |
| ISTOP | <--- | IISTOP = MIN( K2( KI ), I-1 ){2ISTOP = MIN( K1( KI )+ROTN-MOD( K1( KI ), ROTN ), I-2 ), 3ISTOP = MIN( K2( KI ), I-1 ), 4ISTOP = MIN( K2( KI ), I-1 ), 5ISTOP = MIN( K2( KI ), I-1 ), 6ISTOP = MIN( K2( KI ), I-1 ), 7ISTOP = MIN( K2( KI ), I-1 ), 8ISTOP = MIN( K2( KI ), I-1 ), 9ISTOP = MIN( K2( KI ), I-1 ), 10ISTOP = MIN( K2( KI ), I-1 )}, ISTOPISTOP = MIN( ISTOP, K1( KI )+HBL-3-{2ISTOP = MIN( ISTOP, I2-2 ), 3ISTOP = MAX( ISTOP, K1( KI ) )}, K1ISTOP = MIN( K1( KI )+ROTN-MOD( K1( KI ), ROTN ), I-2 ){2ISTOP = MIN( ISTOP, K1( KI )+HBL-3-, 3ISTOP = MAX( ISTOP, K1( KI ) )}, K2ISTOP = MIN( K2( KI ), I-1 ){2ISTOP = MIN( K2( KI ), I-1 ), 3ISTOP = MIN( K2( KI ), I-1 ), 4ISTOP = MIN( K2( KI ), I-1 ), 5ISTOP = MIN( K2( KI ), I-1 ), 6ISTOP = MIN( K2( KI ), I-1 ), 7ISTOP = MIN( K2( KI ), I-1 ), 8ISTOP = MIN( K2( KI ), I-1 ), 9ISTOP = MIN( K2( KI ), I-1 )}, KIISTOP = MIN( K2( KI ), I-1 ){2ISTOP = MIN( K1( KI )+ROTN-MOD( K1( KI ), ROTN ), I-2 ), 3ISTOP = MIN( ISTOP, K1( KI )+HBL-3-, 4ISTOP = MAX( ISTOP, K1( KI ) ), 5ISTOP = MIN( K2( KI ), I-1 ), 6ISTOP = MIN( K2( KI ), I-1 ), 7ISTOP = MIN( K2( KI ), I-1 ), 8ISTOP = MIN( K2( KI ), I-1 ), 9ISTOP = MIN( K2( KI ), I-1 ), 10ISTOP = MIN( K2( KI ), I-1 ), 11ISTOP = MIN( K2( KI ), I-1 ), 12ISTOP = MIN( K2( KI ), I-1 )} |
| ITMP1 | <--- | IITMP1 = MIN( K2( KI )+1, I-1 ) + 1{2ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 3ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 4ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 5ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 6ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 7ITMP1 = MIN( ISTART+1, I ) - 1, 8ITMP1 = MIN( K+3, I ), 9ITMP1 = MIN( ISTART+1, I ) - 1, 10ITMP1 = MIN( K+3, I )}, ISTARTITMP1 = KCOL( KI ) + K - ISTART{2ITMP1 = MIN( ISTART+1, I ) - 1, 3ITMP1 = MIN( ISTART+1, I ) - 1}, ITMP1ITMP1 = ITMP1 - 1{2ITMP1 = ( ITMP1-1 )*LDZ, 3ITMP1 = ( ITMP1-1 )*LDZ, 4ITMP1 = ( ITMP1-1 )*LDZ}, KITMP1 = NUMROC( K+1, HBL, MYROW, 0, NPROW ){2ITMP1 = KCOL( KI ) + K - ISTART, 3ITMP1 = MIN( 6, I2+2-K ), 4ITMP1 = MIN( 6, I2-K+3 ), 5ITMP1 = MIN( K+4, I2 ) + 1, 6ITMP1 = MIN( K+3, I2 ) + 1, 7ITMP1 = MIN( K+4, I2 ) + 1, 8ITMP1 = MIN( K+3, I2 ) + 1, 9ITMP1 = MIN( K+4, I2 ) + 1, 10ITMP1 = MIN( K+3, I2 ) + 1, 11ITMP1 = MIN( K+3, I ), 12ITMP1 = MAX( I1, K-1 ) - 1, 13ITMP1 = MAX( I1, K-2 ) - 1, 14ITMP1 = MIN( K+3, I ), 15ITMP1 = MAX( I1, K-1 ) - 1, 16ITMP1 = MAX( I1, K-2 ) - 1}, K2ITMP1 = MIN( K2( KI )+1, I-1 ) + 1{2ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 3ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 4ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 5ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 6ITMP1 = MIN( K2( KI )+1, I-1 ) + 1}, KCOLITMP1 = KCOL( KI ){2ITMP1 = KCOL( KI ) + K - ISTART, 3ITMP1 = KCOL( KI )}, KIITMP1 = LOCALK2( KI ){2ITMP1 = KCOL( KI ), 3ITMP1 = KCOL( KI ) + K - ISTART, 4ITMP1 = LOCALK2( KI ), 5ITMP1 = KCOL( KI ), 6ITMP1 = LOCALK2( KI ), 7ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 8ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 9ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 10ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 11ITMP1 = MIN( K2( KI )+1, I-1 ) + 1, 12ITMP1 = MIN( K2( KI )+1, I-1 ) + 1}, LOCALK2ITMP1 = LOCALK2( KI ){2ITMP1 = LOCALK2( KI ), 3ITMP1 = LOCALK2( KI )} |
| ITMP2 | <--- | IROW2ITMP2 = IROW2, KITMP2 = NUMROC( K+1, HBL, MYCOL, 0, NPCOL ){2ITMP2 = NUMROC( K+1, HBL, MYCOL, 0, NPCOL ), 3ITMP2 = MAX( I1-K+2, 1 ), 4ITMP2 = MAX( I1-K+3, 1 )}, NITMP2 = NUMROC( N, HBL, MYROW, 0, NPROW ){2ITMP2 = NUMROC( N, HBL, MYCOL, 0, NPCOL )} |
| ITN | <--- | ITNITN = ITN - ITS |
| J | <--- | IDO 50 J = ICOL1, MIN( K2( KI )+1, I-1 ) +, ICOL1DO 50 J = ICOL1, MIN( K2( KI )+1, I-1 ) +{2DO 190 J = ICOL1, ICOL2, 3DO 200 J = ICOL1, ICOL2}, ICOL2DO 190 J = ICOL1, ICOL2{2DO 200 J = ICOL1, ICOL2}, IROW1DO 290 J = IROW1, IROW2{2DO 300 J = IROW1, IROW2, 3DO 60 J = IROW1 + 1, IROW1 + 3, 4DO 270 J = IROW1, IROW2, 5DO 280 J = IROW1, IROW2}, IROW2DO 290 J = IROW1, IROW2{2DO 300 J = IROW1, IROW2, 3DO 270 J = IROW1, IROW2, 4DO 280 J = IROW1, IROW2}, ITMP1DO 100 J = 2, ITMP1{2DO 120 J = 3, ITMP1}, ITMP2DO 110 J = ITMP2, 5{2DO 130 J = ITMP2, 6}, KDO 50 J = ICOL1, MIN( K2( KI )+1, I-1 ) +, K2DO 50 J = ICOL1, MIN( K2( KI )+1, I-1 ) +, KIDO 50 J = ICOL1, MIN( K2( KI )+1, I-1 ) +, LIHIHDO 340 J = ( LILOH-1 )*LDA,{2DO 350 J = LILOH, LIHIH, 3DO 360 J = LILOH, LIHIH, 4DO 370 J = LILOH, LIHIH} |
| JBLK | <--- | IJBLK = I - L + 1 |
| JJ | <--- | ICURCOLJJ = MOD( ICURCOL( KI )+NPCOL-1, NPCOL ), ICURROWJJ = MOD( ICURROW( KI )+NPROW-1, NPROW ), KIJJ = MOD( ICURCOL( KI )+NPCOL-1, NPCOL ){2JJ = MOD( ICURROW( KI )+NPROW-1, NPROW )} |
| K | <--- | IDO 440 K = L, I, ISTARTDO 400 K = ISTART, ISTOP{2K = ISTART, 3DO 140 K = ISTART, ISTOP, 4K = ISTART, 5DO 170 K = ISTART, ISTOP, 6DO 210 K = ISTART, ISTOP, 7DO 230 K = ISTART, ISTOP, 8K = ISTART, 9DO 310 K = ISTART, ISTOP}, ISTOPDO 400 K = ISTART, ISTOP{2K = ISTOP, 3DO 140 K = ISTART, ISTOP, 4DO 170 K = ISTART, ISTOP, 5DO 210 K = ISTART, ISTOP, 6DO 230 K = ISTART, ISTOP, 7K = ISTOP, 8DO 310 K = ISTART, ISTOP} |
| K1 | <--- | K2K1( KI ) = K2( KI ) + 1, KIK1( KI ) = K2( KI ) + 1 |
| K2 | <--- | ISTOPK2( KI ) = ISTOP |
| KCOL | <--- | K1KCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1{2KCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1}, K2KCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1{2KCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1}, KCOLKCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1{2KCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1}, KIKCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1{2KCOL( KI ) = KCOL( KI ) + K2( KI ) - K1( KI ) + 1} |
| KP2COL | <--- | K1KP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1{2KP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1}, K2KP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1{2KP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1, 3KP2COL( KI ) = NUMROC( K2( KI )+3, HBL, MYCOL, 0,}, KIKP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1{2KP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1, 3KP2COL( KI ) = NUMROC( K2( KI )+3, HBL, MYCOL, 0,}, KP2COLKP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1{2KP2COL( KI ) = KP2COL( KI ) + K2( KI ) - K1( KI ) + 1} |
| KP2ROW | <--- | K1KP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1{2KP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1}, K2KP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1{2KP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1, 3KP2ROW( KI ) = NUMROC( K2( KI )+3, HBL, MYROW, 0,}, KIKP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1{2KP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1, 3KP2ROW( KI ) = NUMROC( K2( KI )+3, HBL, MYROW, 0,}, KP2ROWKP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1{2KP2ROW( KI ) = KP2ROW( KI ) + K2( KI ) - K1( KI ) + 1} |
| KROW | <--- | K1KROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1{2KROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1}, K2KROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1{2KROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1}, KIKROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1{2KROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1}, KROWKROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1{2KROW( KI ) = KROW( KI ) + K2( KI ) - K1( KI ) + 1} |
| LIHIH | <--- | ILIHIH = NUMROC( I, HBL, MYROW, 0, NPROW ) |
| LOCALK2 | <--- | KILOCALK2( KI ) = LOCALK2( KI ) + 1{2LOCALK2( KI ) = LOCALK2( KI ) + 2, 3LOCALK2( KI ) = LOCALK2( KI ) + 1, 4LOCALK2( KI ) = LOCALK2( KI ) + 2, 5LOCALK2( KI ) = LOCALK2( KI ) + 1, 6LOCALK2( KI ) = LOCALK2( KI ) + 1, 7LOCALK2( KI ) = LOCALK2( KI ) + 1}, LOCALK2LOCALK2( KI ) = LOCALK2( KI ) + 1{2LOCALK2( KI ) = LOCALK2( KI ) + 2, 3LOCALK2( KI ) = LOCALK2( KI ) + 1, 4LOCALK2( KI ) = LOCALK2( KI ) + 2, 5LOCALK2( KI ) = LOCALK2( KI ) + 1, 6LOCALK2( KI ) = LOCALK2( KI ) + 1, 7LOCALK2( KI ) = LOCALK2( KI ) + 1} |
| MODKM1 | <--- | IMODKM1 = MOD( I-1+HBL, HBL ), ISTARTMODKM1 = MOD( ISTART-1, HBL ){2MODKM1 = MOD( ISTART-1, HBL )}, KMODKM1 = MOD( K-1, HBL ) |
| NR | <--- | INR = MIN( 3, I-K+1 ){2NR = MIN( 3, I-K+1 ), 3NR = MIN( 3, I-K+1 ), 4NR = MIN( 3, I-K+1 ), 5NR = MIN( 3, I-K+1 ), 6NR = MIN( 3, I-K+1 ), 7NR = MIN( 3, I-K+1 ), 8NR = MIN( 3, I-K+1 )}, KNR = MIN( 3, I-K+1 ){2NR = MIN( 3, I-K+1 ), 3NR = MIN( 3, I-K+1 ), 4NR = MIN( 3, I-K+1 ), 5NR = MIN( 3, I-K+1 ), 6NR = MIN( 3, I-K+1 ), 7NR = MIN( 3, I-K+1 ), 8NR = MIN( 3, I-K+1 )} |
| SMALLA | <--- | JSMALLA( 2, J, KI ) = SMALLA( 2, J, KI ) - SUM*T1{2SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T2, 3SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T3, 4SMALLA( J, 2, KI ) = SMALLA( J, 2, KI ) - SUM*T1, 5SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T2, 6SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T3, 7SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T1, 8SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T2, 9SMALLA( 5, J, KI ) = SMALLA( 5, J, KI ) - SUM*T3, 10SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T1, 11SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T2, 12SMALLA( J, 5, KI ) = SMALLA( J, 5, KI ) - SUM*T3}, KISMALLA( 2, J, KI ) = SMALLA( 2, J, KI ) - SUM*T1{2SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T2, 3SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T3, 4SMALLA( J, 2, KI ) = SMALLA( J, 2, KI ) - SUM*T1, 5SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T2, 6SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T3, 7SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T1, 8SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T2, 9SMALLA( 5, J, KI ) = SMALLA( 5, J, KI ) - SUM*T3, 10SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T1, 11SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T2, 12SMALLA( J, 5, KI ) = SMALLA( J, 5, KI ) - SUM*T3}, SMALLASMALLA( 2, J, KI ) = SMALLA( 2, J, KI ) - SUM*T1{2SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T2, 3SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T3, 4SMALLA( J, 2, KI ) = SMALLA( J, 2, KI ) - SUM*T1, 5SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T2, 6SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T3, 7SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T1, 8SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T2, 9SMALLA( 5, J, KI ) = SMALLA( 5, J, KI ) - SUM*T3, 10SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T1, 11SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T2, 12SMALLA( J, 5, KI ) = SMALLA( J, 5, KI ) - SUM*T3}, SUMSMALLA( 2, J, KI ) = SMALLA( 2, J, KI ) - SUM*T1{2SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T2, 3SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T3, 4SMALLA( J, 2, KI ) = SMALLA( J, 2, KI ) - SUM*T1, 5SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T2, 6SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T3, 7SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T1, 8SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T2, 9SMALLA( 5, J, KI ) = SMALLA( 5, J, KI ) - SUM*T3, 10SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T1, 11SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T2, 12SMALLA( J, 5, KI ) = SMALLA( J, 5, KI ) - SUM*T3}, T1SMALLA( 2, J, KI ) = SMALLA( 2, J, KI ) - SUM*T1{2SMALLA( J, 2, KI ) = SMALLA( J, 2, KI ) - SUM*T1, 3SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T1, 4SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T1}, T2SMALLA( 3, J, KI ) = SMALLA( 3, J, KI ) - SUM*T2{2SMALLA( J, 3, KI ) = SMALLA( J, 3, KI ) - SUM*T2, 3SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T2, 4SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T2}, T3SMALLA( 4, J, KI ) = SMALLA( 4, J, KI ) - SUM*T3{2SMALLA( J, 4, KI ) = SMALLA( J, 4, KI ) - SUM*T3, 3SMALLA( 5, J, KI ) = SMALLA( 5, J, KI ) - SUM*T3, 4SMALLA( J, 5, KI ) = SMALLA( J, 5, KI ) - SUM*T3} |
| SUM | <--- | ASUM = A( ITMP1+J ) + V2*A( ITMP1+1+J ){2SUM = WORK( IRBUF+J-LILOH+1 ) +, 3SUM = A( ( ITMP1-1 )*LDA+J ) +, 4SUM = WORK( ICBUF+J ) +, 5SUM = A( ( J-1 )*LDA+IROW1 ) +, 6SUM = A( ( ICOL1-1 )*LDA+J ) +, 7SUM = A( ( J-1 )*LDA+IROW1 ) +, 8SUM = WORK( IRBUF+J-ICOL1+1 ) +, 9SUM = A( ( ICOL1-1 )*LDA+J ) +, 10SUM = WORK( ICBUF+J-IROW1+1 ) +}, ICOL1SUM = Z( ICOL1+J ) +{2SUM = WORK( ICBUF+J-IROW1+1 ) +, 3SUM = A( ( ICOL1-1 )*LDA+J ) +, 4SUM = A( ( J-1 )*LDA+IROW1 ) +, 5SUM = WORK( IRBUF+J-ICOL1+1 ) +, 6SUM = A( ( ICOL1-1 )*LDA+J ) +, 7SUM = WORK( ICBUF+J-IROW1+1 ) +}, IROW1SUM = Z( ICOL1+J ) +{2SUM = WORK( ICBUF+J-IROW1+1 ) +, 3SUM = A( ( J-1 )*LDA+IROW1 ) +, 4SUM = A( ( J-1 )*LDA+IROW1 ) +, 5SUM = WORK( IRBUF+J-ICOL1+1 ) +, 6SUM = A( ( ICOL1-1 )*LDA+J ) +, 7SUM = WORK( ICBUF+J-IROW1+1 ) +}, ITMP1SUM = A( ITMP1+J ) + V2*A( ITMP1+1+J ){2SUM = WORK( IRBUF+J-LILOH+1 ) +, 3SUM = A( ( ITMP1-1 )*LDA+J ) +, 4SUM = WORK( ICBUF+J ) +, 5SUM = Z( J+ITMP1 ) +, 6SUM = WORK( ICBUF+J ) +}, JSUM = Z( ICOL1+J ) +{2SUM = WORK( ICBUF+J-IROW1+1 ) +, 3SUM = A( ITMP1+J ) + V2*A( ITMP1+1+J ), 4SUM = WORK( IRBUF+J-LILOH+1 ) +, 5SUM = A( ( ITMP1-1 )*LDA+J ) +, 6SUM = WORK( ICBUF+J ) +, 7SUM = Z( J+ITMP1 ) +, 8SUM = WORK( ICBUF+J ) +, 9SUM = A( ( J-1 )*LDA+IROW1 ) +, 10SUM = A( ( ICOL1-1 )*LDA+J ) +, 11SUM = SMALLA( 2, J, KI ) +, 12SUM = SMALLA( J, 2, KI ) +, 13SUM = SMALLA( 3, J, KI ) +, 14SUM = SMALLA( J, 3, KI ) +, 15SUM = A( ( J-1 )*LDA+IROW1 ) +, 16SUM = WORK( IRBUF+J-ICOL1+1 ) +, 17SUM = A( ( ICOL1-1 )*LDA+J ) +, 18SUM = WORK( ICBUF+J-IROW1+1 ) +}, KISUM = SMALLA( 2, J, KI ) +{2SUM = SMALLA( J, 2, KI ) +, 3SUM = SMALLA( 3, J, KI ) +, 4SUM = SMALLA( J, 3, KI ) +}, SMALLASUM = SMALLA( 2, J, KI ) +{2SUM = SMALLA( J, 2, KI ) +, 3SUM = SMALLA( 3, J, KI ) +, 4SUM = SMALLA( J, 3, KI ) +}, V2SUM = Z( ICOL1+J ) +{2SUM = WORK( ICBUF+J-IROW1+1 ) +, 3SUM = A( ITMP1+J ) + V2*A( ITMP1+1+J ), 4SUM = WORK( IRBUF+J-LILOH+1 ) +, 5SUM = A( ( ITMP1-1 )*LDA+J ) +, 6SUM = WORK( ICBUF+J ) +, 7SUM = Z( J+ITMP1 ) +, 8SUM = WORK( ICBUF+J ) +, 9SUM = A( ( J-1 )*LDA+IROW1 ) +, 10SUM = A( ( ICOL1-1 )*LDA+J ) +, 11SUM = SMALLA( 2, J, KI ) +, 12SUM = SMALLA( J, 2, KI ) +, 13SUM = SMALLA( 3, J, KI ) +, 14SUM = SMALLA( J, 3, KI ) +, 15SUM = A( ( J-1 )*LDA+IROW1 ) +, 16SUM = WORK( IRBUF+J-ICOL1+1 ) +, 17SUM = A( ( ICOL1-1 )*LDA+J ) +, 18SUM = WORK( ICBUF+J-IROW1+1 ) +}, V3SUM = Z( ICOL1+J ) +{2SUM = WORK( ICBUF+J-IROW1+1 ) +, 3SUM = A( ( J-1 )*LDA+IROW1 ) +, 4SUM = A( ( ICOL1-1 )*LDA+J ) +, 5SUM = SMALLA( 2, J, KI ) +, 6SUM = SMALLA( J, 2, KI ) +, 7SUM = SMALLA( 3, J, KI ) +, 8SUM = SMALLA( J, 3, KI ) +, 9SUM = A( ( J-1 )*LDA+IROW1 ) +, 10SUM = WORK( IRBUF+J-ICOL1+1 ) +, 11SUM = A( ( ICOL1-1 )*LDA+J ) +, 12SUM = WORK( ICBUF+J-IROW1+1 ) +}, WORKSUM = Z( ICOL1+J ) +{2SUM = WORK( ICBUF+J-IROW1+1 ) +, 3SUM = WORK( IRBUF+J-LILOH+1 ) +, 4SUM = WORK( ICBUF+J ) +, 5SUM = WORK( ICBUF+J ) +, 6SUM = A( ( J-1 )*LDA+IROW1 ) +, 7SUM = WORK( IRBUF+J-ICOL1+1 ) +, 8SUM = A( ( ICOL1-1 )*LDA+J ) +, 9SUM = WORK( ICBUF+J-IROW1+1 ) +}, ZSUM = Z( ICOL1+J ) +{2SUM = WORK( ICBUF+J-IROW1+1 ) +, 3SUM = Z( J+ITMP1 ) +, 4SUM = WORK( ICBUF+J ) +} |
| T1 | <--- | KT1 = WORK( VECSIDX+( K-1 )*3+3 ){2T1 = WORK( VECSIDX+( K-1 )*3+3 ), 3T1 = WORK( VECSIDX+( K-1 )*3+3 ), 4T1 = WORK( VECSIDX+( K-1 )*3+3 ), 5T1 = WORK( VECSIDX+( K-1 )*3+3 ), 6T1 = WORK( VECSIDX+( K-1 )*3+3 ), 7T1 = WORK( VECSIDX+( K-1 )*3+3 ), 8T1 = WORK( VECSIDX+( K-1 )*3+3 ), 9T1 = WORK( VECSIDX+( K-1 )*3+3 )}, WORKT1 = WORK( VECSIDX+( K-1 )*3+3 ){2T1 = WORK( VECSIDX+( K-1 )*3+3 ), 3T1 = WORK( VECSIDX+( K-1 )*3+3 ), 4T1 = WORK( VECSIDX+( K-1 )*3+3 ), 5T1 = WORK( VECSIDX+( K-1 )*3+3 ), 6T1 = WORK( VECSIDX+( K-1 )*3+3 ), 7T1 = WORK( VECSIDX+( K-1 )*3+3 ), 8T1 = WORK( VECSIDX+( K-1 )*3+3 ), 9T1 = WORK( VECSIDX+( K-1 )*3+3 )} |
| T2 | <--- | T1T2 = T1*V2{2T2 = T1*V2, 3T2 = T1*V2, 4T2 = T1*V2, 5T2 = T1*V2, 6T2 = T1*V2, 7T2 = T1*V2, 8T2 = T1*V2, 9T2 = T1*V2, 10T2 = T1*V2, 11T2 = T1*V2, 12T2 = T1*V2, 13T2 = T1*V2, 14T2 = T1*V2}, V2T2 = T1*V2{2T2 = T1*V2, 3T2 = T1*V2, 4T2 = T1*V2, 5T2 = T1COPY*V2, 6T2 = T1*V2, 7T2 = T1*V2, 8T2 = T1*V2, 9T2 = T1*V2, 10T2 = T1*V2, 11T2 = T1*V2, 12T2 = T1*V2, 13T2 = T1*V2, 14T2 = T1*V2, 15T2 = T1*V2} |
| T3 | <--- | T1T3 = T1*V3{2T3 = T1*V3, 3T3 = T1*V3, 4T3 = T1*V3, 5T3 = T1*V3, 6T3 = T1*V3, 7T3 = T1*V3, 8T3 = T1*V3, 9T3 = T1*V3, 10T3 = T1*V3, 11T3 = T1*V3, 12T3 = T1*V3, 13T3 = T1*V3}, V3T3 = T1*V3{2T3 = T1*V3, 3T3 = T1*V3, 4T3 = T1*V3, 5T3 = T1*V3, 6T3 = T1*V3, 7T3 = T1*V3, 8T3 = T1*V3, 9T3 = T1*V3, 10T3 = T1*V3, 11T3 = T1*V3, 12T3 = T1*V3, 13T3 = T1*V3} |
| V2 | <--- | KV2 = WORK( VECSIDX+( K-1 )*3+1 ){2V2 = WORK( VECSIDX+( K-1 )*3+1 ), 3V2 = WORK( VECSIDX+( K-1 )*3+1 ), 4V2 = WORK( VECSIDX+( K-1 )*3+1 ), 5V2 = WORK( VECSIDX+( K-1 )*3+1 ), 6V2 = WORK( VECSIDX+( K-1 )*3+1 ), 7V2 = WORK( VECSIDX+( K-1 )*3+1 ), 8V2 = WORK( VECSIDX+( K-1 )*3+1 ), 9V2 = WORK( VECSIDX+( K-1 )*3+1 )}, WORKV2 = WORK( VECSIDX+( K-1 )*3+1 ){2V2 = WORK( VECSIDX+( K-1 )*3+1 ), 3V2 = WORK( VECSIDX+( K-1 )*3+1 ), 4V2 = WORK( VECSIDX+( K-1 )*3+1 ), 5V2 = WORK( VECSIDX+( K-1 )*3+1 ), 6V2 = WORK( VECSIDX+( K-1 )*3+1 ), 7V2 = WORK( VECSIDX+( K-1 )*3+1 ), 8V2 = WORK( VECSIDX+( K-1 )*3+1 ), 9V2 = WORK( VECSIDX+( K-1 )*3+1 )} |
| V3 | <--- | KV3 = WORK( VECSIDX+( K-1 )*3+2 ){2V3 = WORK( VECSIDX+( K-1 )*3+2 ), 3V3 = WORK( VECSIDX+( K-1 )*3+2 ), 4V3 = WORK( VECSIDX+( K-1 )*3+2 ), 5V3 = WORK( VECSIDX+( K-1 )*3+2 ), 6V3 = WORK( VECSIDX+( K-1 )*3+2 ), 7V3 = WORK( VECSIDX+( K-1 )*3+2 ), 8V3 = WORK( VECSIDX+( K-1 )*3+2 ), 9V3 = WORK( VECSIDX+( K-1 )*3+2 )}, WORKV3 = WORK( VECSIDX+( K-1 )*3+2 ){2V3 = WORK( VECSIDX+( K-1 )*3+2 ), 3V3 = WORK( VECSIDX+( K-1 )*3+2 ), 4V3 = WORK( VECSIDX+( K-1 )*3+2 ), 5V3 = WORK( VECSIDX+( K-1 )*3+2 ), 6V3 = WORK( VECSIDX+( K-1 )*3+2 ), 7V3 = WORK( VECSIDX+( K-1 )*3+2 ), 8V3 = WORK( VECSIDX+( K-1 )*3+2 ), 9V3 = WORK( VECSIDX+( K-1 )*3+2 )} |
| WI | <--- | H21WI( I-1 ) = H21{2WI( I ) = -H21}, ZEROWI( I ) = ZERO{2WI( I-1 ) = ZERO, 3WI( I ) = ZERO, 4WI( I-1 ) = ZERO, 5WI( I ) = ZERO, 6WI( K ) = ZERO} |
| WORK | <--- | ICOL1WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+{2WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+}, IROW1WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+{2WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+, 3WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-, 4WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-}, JWORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+{2WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+, 3WORK( IRBUF+J-LILOH+1 ) = WORK( IRBUF+, 4WORK( ICBUF+J ) = WORK( ICBUF+J ) -, 5WORK( ICBUF+J ) = WORK( ICBUF+J ) -, 6WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+, 7WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+, 8WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-, 9WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-}, SUMWORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+{2WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+, 3WORK( IRBUF+J-LILOH+1 ) = WORK( IRBUF+, 4WORK( ICBUF+J ) = WORK( ICBUF+J ) -, 5WORK( ICBUF+J ) = WORK( ICBUF+J ) -, 6WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+, 7WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+, 8WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-, 9WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-}, T1WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+{2WORK( IRBUF+J-LILOH+1 ) = WORK( IRBUF+, 3WORK( ICBUF+J ) = WORK( ICBUF+J ) -, 4WORK( ICBUF+J ) = WORK( ICBUF+J ) -, 5WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+, 6WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-}, T3WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+{2WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+, 3WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-}, WORKWORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+{2WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+, 3WORK( IRBUF+J-LILOH+1 ) = WORK( IRBUF+, 4WORK( ICBUF+J ) = WORK( ICBUF+J ) -, 5WORK( ICBUF+J ) = WORK( ICBUF+J ) -, 6WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+, 7WORK( IRBUF+J-ICOL1+1 ) = WORK( IRBUF+, 8WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-, 9WORK( ICBUF+J-IROW1+1 ) = WORK( ICBUF+J-} |
| WR | <--- | H00WR( I-1 ) = H00 + H21{2WR( I ) = H00 - H21, 3WR( I-1 ) = H00, 4WR( I ) = H00}, H21WR( I-1 ) = H00 + H21{2WR( I ) = H00 - H21}, AWR( I ) = A( ( ICOL-1 )*LDA+IROW ), ZEROWR( I ) = ZERO{2WR( I-1 ) = ZERO, 3WR( I ) = ZERO, 4WR( K ) = ZERO} |
| Z | <--- | ICOL1Z( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T1{2Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -, 3Z( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T2, 4Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -}, ITMP1Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T1{2Z( J+ITMP1+LDZ ) = Z( J+ITMP1+LDZ ) -, 3Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T2}, JZ( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T1{2Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -, 3Z( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T2, 4Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -, 5Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T1, 6Z( J+ITMP1+LDZ ) = Z( J+ITMP1+LDZ ) -, 7Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T2}, SUMZ( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T1{2Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -, 3Z( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T2, 4Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -, 5Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T1, 6Z( J+ITMP1+LDZ ) = Z( J+ITMP1+LDZ ) -, 7Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T2}, T1Z( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T1{2Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T1}, T2Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -{2Z( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T2, 3Z( J+ITMP1+LDZ ) = Z( J+ITMP1+LDZ ) -, 4Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T2}, T3Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -, ZZ( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T1{2Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -, 3Z( J+ICOL1 ) = Z( J+ICOL1 ) - SUM*T2, 4Z( J+ICOL1+LDZ ) = Z( J+ICOL1+LDZ ) -, 5Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T1, 6Z( J+ITMP1+LDZ ) = Z( J+ITMP1+LDZ ) -, 7Z( J+ITMP1 ) = Z( J+ITMP1 ) - SUM*T2} |
|
|
Analysis elements of the routine PSLAHQR()
Put the mouse over each element to display detailed matching information
 Assigned variables |
| | | BLOCK_CYCLIC_2D , CONST , CSRC_ , CTXT_ , DLEN_ , DTYPE_ , H00 , H10 , H11 , H12 , H21 , H22 , HALF , I , IBLK , ICOL1 , ICOL2 , IHI , IHIZ , II , ILO , ILOZ , INFO , IROW1 , IROW2 , ISPEC , ISTART , ISTOP , ITMP1 , ITMP2 , ITN , J , JBLK , JJ , K , K2 , KI , LIHIH , LLD_ , LWORK , M_ , MB_ , MODKM1 , N , N_ , NB_ , NR , ONE , RSRC_ , SUM , T1 , T2 , T3 , V2 , V3 , ZERO |
|
| Active variables |
| | | A , BLOCK_CYCLIC_2D , CONST , CSRC_ , CTXT_ , DESCA , DESCZ , DLEN_ , DTYPE_ , H00 , H10 , H11 , H12 , H21 , H22 , HALF , i , IBLK , ICOL1 , ICOL2 , ICURCOL , ICURROW , IHI , IHIZ , II , ILO , ILOZ , ILWORK , INFO , IROW1 , IROW2 , ISPEC , ISTART , ISTOP , ITMP1 , ITMP2 , ITN , IWORK , J , JBLK , JJ , K , K1 , K2 , KCOL , KI , KP2COL , KP2ROW , KROW , LIHIH , LLD_ , LOCALK2 , LWORK , M_ , MB_ , MODKM1 , N , N_ , NB_ , NR , one , RSRC_ , SMALLA , SUM , T1 , T2 , T3 , V2 , V3 , WANTT , WANTZ , WI , WORK , WR , Z , ZERO |
|
| Allocated variables [ statement : associated variable ] |
| |  new | : with |
|
| Accessed arrays [ array name : associated index ] |
| | A | : ( ICOL-1 )*LDA+IROW , ( ICOL1+1 )*LDA+J , ( ICOL1+1 )*LDA+J , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1 , ( ICOL1-1 )*LDA+IROW1+1 , ( ICOL1-1 )*LDA+IROW1+1 , ( ICOL1-1 )*LDA+IROW2 , ( ICOL1-1 )*LDA+IROW2 , ( ICOL1-1 )*LDA+J , ( ICOL1-1 )*LDA+J , ( ICOL1-1 )*LDA+J , ( ICOL1-1 )*LDA+J , ( ICOL1-1 )*LDA+J , ( ICOL1-1 )*LDA+J , ( ICOL1-2 )*LDA+IROW1 , ( ITMP1-1 )*LDA+J , ( ITMP1-1 )*LDA+J , ( ITMP1-1 )*LDA+J , ( ITMP1-1 )*LDA+J , ( ITMP1-1 )*LDA+J , ( ITMP1-1 )*LDA+LILOH , ( ITMP1-1 )*LDA+LILOH , ( J-1 )*LDA+IROW1 , ( J-1 )*LDA+IROW1 , ( J-1 )*LDA+IROW1 , ( J-1 )*LDA+IROW1 , ( J-1 )*LDA+IROW1 , ( J-1 )*LDA+IROW1 , ( J-1 )*LDA+IROW1+1 , ( J-1 )*LDA+IROW1+1 , ( J-1 )*LDA+IROW1+1 , ( J-1 )*LDA+IROW1+1 , ( J-1 )*LDA+IROW1+1 , ( J-1 )*LDA+IROW1+1 , ( J-1 )*LDA+IROW1+2 , ( J-1 )*LDA+IROW1+2 , ( J-1 )*LDA+ITMP1 , ( J-1 )*LDA+ITMP1 , ( LILOH-1 )*LDA+ITMP1 , ( LILOH-1 )*LDA+ITMP1 , * , ICOL1*LDA+IROW1 , ICOL1*LDA+IROW1 , ICOL1*LDA+IROW1+1 , ICOL1*LDA+IROW1+1 , ICOL1*LDA+J , ICOL1*LDA+J , ICOL1*LDA+J , ICOL1*LDA+J , ICOL1*LDA+J , ICOL1*LDA+J , ILO,ILO-1 , ITMP1*LDA+J , ITMP1*LDA+J , ITMP1+1+J , ITMP1+1+J , ITMP1+J , ITMP1+J , K-1:K+4,K-1:K+4 , K-2:K+3,K-2:K+3 , LILOZ:LIHIZ,LOCALK2:LOCALK2+2 |
| | DESCA | : * , CSRC_ , CTXT_ , DTYPE_ , LLD_ , LLD_ , LLD_ , M_ , MB_ , MB_ , N_ , NB_ , NB_ , RSRC_ |
| | DESCZ | : * , LLD_ |
| | ICURCOL | : KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI |
| | ICURROW | : KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI |
| | IWORK | : * |
| | K1 | : * , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI |
| | K2 | : * , IBULGE , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI |
| | KCOL | : KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI |
| | KP2COL | : KI , KI , KI , KI , KI , KI , KI |
| | KP2ROW | : KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI |
| | KROW | : KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI |
| | LOCALK2 | : KI , KI , KI , KI , KI , KI , KI , KI , KI , KI , KI |
| | SMALLA | : 1, 1, KI , 1, 1, KI , 2, J, KI , 2, J, KI , 3, J, KI , 3, J, KI , 3, J, KI , 3, J, KI , 4, J, KI , 4, J, KI , 4, J, KI , 4, J, KI , 5, J, KI , 5, J, KI , J, 2, KI , J, 2, KI , J, 3, KI , J, 3, KI , J, 3, KI , J, 3, KI , J, 4, KI , J, 4, KI , J, 4, KI , J, 4, KI , J, 5, KI , J, 5, KI |
| | WI | : * , i , I , I , I , I , i+1 , I-1 , I-1 , I-1 , K , L |
| | WORK | : * , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+1 , ICBUF+J , ICBUF+J , ICBUF+J , ICBUF+J , ICBUF+J-IROW1+1 , ICBUF+J-IROW1+1 , ICBUF+J-IROW1+1 , ICBUF+J-IROW1+1 , ICBUF+J-IROW1+1 , ICBUF+J-IROW1+1 , ICBUF+J-IROW1+1 , ICBUF+J-IROW1+1 , IRBUF+1 , IRBUF+1 , IRBUF+1 , IRBUF+1 , IRBUF+1 , IRBUF+1 , IRBUF+J-ICOL1+1 , IRBUF+J-ICOL1+1 , IRBUF+J-ICOL1+1 , IRBUF+J-ICOL1+1 , IRBUF+J-LILOH+1 , IRBUF+J-LILOH+1 , LWORK , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( ISTART-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+1 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+2 , VECSIDX+( K-1 )*3+3 , VECSIDX+( K-1 )*3+3 , VECSIDX+( K-1 )*3+3 , VECSIDX+( K-1 )*3+3 , VECSIDX+( K-1 )*3+3 , VECSIDX+( K-1 )*3+3 , VECSIDX+( K-1 )*3+3 , VECSIDX+( K-1 )*3+3 , VECSIDX+( K-1 )*3+3 , VECSIDX+( K-1 )*3+3 , VECSIDX+1 , VECSIDX+1 , VECSIDX+1 , VECSIDX+1 , VECSIDX+1 , VECSIDX+1 |
| | WR | : * , I , I , I , I , I , I-1 , I-1 , I-1 , K , L |
| | Z | : ( ICOL1-1 )*LDZ+IROW1 , ( ICOL1-1 )*LDZ+IROW1 , ( ICOL1-1 )*LDZ+IROW1 , ( ICOL1-1 )*LDZ+IROW1 , * , ICOL1+J , ICOL1+J+LDZ , ILOZ:IHIZ,ILO:IHI , J+ICOL1 , J+ICOL1 , J+ICOL1 , J+ICOL1+LDZ , J+ICOL1+LDZ , J+ICOL1+LDZ , J+ITMP1 , J+ITMP1 , J+ITMP1 , J+ITMP1 , J+ITMP1+LDZ , J+ITMP1+LDZ , LILOZ:LIHIZ,LOCALK2:LOCALK2+2 , LILOZ+ITMP1 , LILOZ+ITMP1 |
|
| Conditional statements [ statement : associated predicate ] |
| | do | : ( some work so next step is ready... ) , ( 50 J = ICOL1 , MIN( K2( KI ) + 1 , I - 1 ) + ) , ( 60 J = IROW1 + 1 , IROW1 + 3 ) , ( column transforms and finish work ) , ( 90 KI = 1 , IBULGE ) , ( make up work to have things in block fashion ) , ( 150 KI = 1 , IBULGE ) , ( some work so next step is ready... ) , ( 100 J = 2 , ITMP1 ) , ( 110 J = ITMP2 , 5 ) , ( some work so next step is ready... ) , ( 120 J = 3 , ITMP1 ) , ( 130 J = ITMP2 , 6 ) , ( 140 K = ISTART , ISTOP ) , ( some work so next step is ready... ) , ( 160 KI = 1 , IBULGE ) , ( 180 KI = 1 , IBULGE ) , ( 170 K = ISTART , ISTOP ) , ( 220 KI = 1 , IBULGE ) , ( 210 K = ISTART , ISTOP ) , ( 190 J = ICOL1 , ICOL2 ) , ( 200 J = ICOL1 , ICOL2 ) , ( 240 KI = 1 , IBULGE ) , ( 230 K = ISTART , ISTOP ) , ( 260 KI = 1 , IBULGE ) , ( from ITMP1 + 1 to MIN(K + 3 , I) ) , ( 320 KI = 1 , IBULGE ) , ( 310 K = ISTART , ISTOP ) , ( 270 J = IROW1 , IROW2 ) , ( 280 J = IROW1 , IROW2 ) , ( 290 J = IROW1 , IROW2 ) , ( 300 J = IROW1 , IROW2 ) , ( NR=2 work ) , ( 410 KI = 1 , IBULGE ) , ( 400 K = ISTART , ISTOP ) , ( 340 J = ( LILOH - 1 )*LDA , ) , ( 350 J = LILOH , LIHIH ) , ( 360 J = LILOH , LIHIH ) , ( 370 J = LILOH , LIHIH ) , ( 380 J = LILOZ , LIHIZ ) , ( 390 J = LILOZ , LIHIZ ) , ( 440 K = L , I ) |
| | for | : ( any 2D block cyclicly distributed array. ) , ( these quantities may be computed by : ) , ( the distributed matrix A. ) , ( the distributed matrix Z. ) , ( the next release. ) , ( any bulge are ) , ( this release , it is assumed RSRC_ = CSRC_ = 0 ) , ( this bulge ) |
| | if | : ( K were distributed over the p processes of its ) , ( K were distributed over the q processes of ) , ( WANTZ). ) , ( WANTT is .TRUE.. ) , ( WANTT is .TRUE. , A is upper quasi - triangular in ) , ( WANTT is .FALSE. , the ) , ( two eigenvalues are computed as a ) , ( WANTT is .TRUE. , the ) , ( WANTZ is .TRUE.. ) , ( WANTZ is .TRUE. , on entry Z must contain the current ) , ( WANTZ is .FALSE. , Z is not referenced. ) , ( INFO = i , ) , ( possible) ) , ( elements ) , ( M.GT.L ) , ( (MOD( K - 1 , HBL ).GT.0 ) ) , ( K.LT.ISTOP ) , ( MODKM1.EQ.HBL - 2 ) , ( (( DOWN.EQ.ICURROW( KI ) ) .AND. ) , ( (( MYROW.EQ.ICURROW( KI ) ) .AND. ) , ( (( DOWN.EQ.ICURROW( KI ) ) .AND. ) , ( MYCOL.NE.JJ ) , ( (( MYROW.EQ.ICURROW( KI ) ) .AND. ( NPCOL.GT.1 ) .AND. ) , ( (MYCOL.NE.ICURCOL( KI ) ) ) , ( (MOD( ISTART - 1 , HBL ).EQ.HBL - 2 ) ) , ( (( RIGHT.EQ.ICURCOL( KI ) ) .AND. ) , ( MYROW.NE.JJ ) , ( (( MYCOL.EQ.ICURCOL( KI ) ) .AND. ( NPROW.GT.1 ) .AND. ) , ( (MYROW.NE.ICURROW( KI ) ) ) , ( (( MYROW.EQ.ICURROW( KI ) ) .AND. ) , ( NR.EQ.3 ) , ( (( MOD( ISTART - 1 , HBL ).EQ.HBL - 1 ) .AND. ) , ( NR.EQ.3 ) , ( (( MYROW.EQ.ICURROW( KI ) ) .AND. ) , ( K.LT.ISTOP ) , ( (( NR.EQ.3 ) .AND. ( MOD( K - 1 , ) , ( (( MODKM1.GE.HBL - 2 ) .AND. ( K.LE.I - 1 ) ) ) , ( (( MODKM1.EQ.HBL - 2 ) .AND. ( K.LT.I - 1 ) ) ) , ( MODKM1.EQ.HBL - 1 ) , ( (( MYROW.NE.ICURROW( KI ) ) .AND. ) , ( (( ISTOP.GT.ISTART ) .AND. ) , ( (KROW( KI ).GT.KP2ROW( KI ) ) ) , ( (( MYROW.NE.ICURROW( KI ) ) .AND. ) , ( (( ISTART.EQ.ISTOP ) .OR. ) , ( (( NR.EQ.3 ) .AND. ( KROW( KI ).LE. ) , ( (( K.LT.ISTOP ) .AND. ) , ( (MOD( K - 1 , HBL ).LT.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 1 ) ) , ( (( MOD( K - 1 , HBL ).LT.HBL - 2 ) .OR. ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND. ) , ( IROW1.EQ.IROW2 ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND. ) , ( IROW1.EQ.IROW2 ) , ( (KROW( KI ).GT.KP2ROW( KI ) ) ) , ( (( MYROW.NE.ICURROW( KI ) ) .AND. ) , ( (( ISTART.EQ.ISTOP ) .OR. ) , ( (( NR.EQ.3 ) .AND. ( KROW( KI ).LE. ) , ( (( K.LT.ISTOP ) .AND. ) , ( (MOD( K - 1 , HBL ).LT.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 1 ) ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND. ) , ( IROW1.NE.IROW2 ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND. ) , ( IROW1.NE.IROW2 ) , ( (KROW( KI ).GT.KP2ROW( KI ) ) ) , ( (( MYROW.NE.ICURROW( KI ) ) .AND. ) , ( (( ISTART.EQ.ISTOP ) .OR. ) , ( (( NR.EQ.3 ) .AND. ( KROW( KI ).LE. ) , ( (( K.LT.ISTOP ) .AND. ) , ( (MOD( K - 1 , HBL ).LT.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 1 ) ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND. ) , ( IROW1.EQ.IROW2 ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND. ) , ( IROW1.EQ.IROW2 ) , ( (( MYCOL.NE.ICURCOL( KI ) ) .AND. ) , ( (( ( MOD( ISTART - 1 , HBL ).LT.HBL - 2 ) .OR. ( NPCOL.EQ. ) , ( (( K.LT.ISTOP ) .AND. ( MOD( K - 1 , ) , ( (MOD( K - 1 , HBL ).LT.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 1 ) ) , ( IROW1.LE.IROW2 ) , ( (MOD( K - 1 , HBL ).LT.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).LT.HBL - 3 ) ) , ( (MOD( ( ITMP1 / HBL ) , NPROW ).EQ.MYROW ) ) , ( ITMP2.GT.0 ) , ( (KCOL( KI ).GT.KP2COL( KI ) ) ) , ( (( MYCOL.NE.ICURCOL( KI ) ) .AND. ) , ( (MOD( ISTART - 1 , HBL ).GE.HBL - 2 ) ) , ( (( NR.EQ.3 ) .AND. ( KCOL( KI ).LE.KP2COL( KI ) ) ) ) , ( (( K.LT.ISTOP ) .AND. ) , ( (MOD( K - 1 , HBL ).LT.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 2 ) ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 1 ) ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND. ) , ( ICOL1.EQ.ICOL2 ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND. ) , ( ICOL1.EQ.ICOL2 ) , ( we want Z and we haven't already done any Z ) , ( (( WANTZ ) .AND. ( MOD( K - 1 , ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 2 ) ) , ( ICOL1.EQ.ICOL2 ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 1 ) ) , ( ICOL1.EQ.ICOL2 ) , ( (ICURCOL( KI ).EQ.MYCOL ) ) , ( (( ISPEC.EQ.0 ) .OR. ( NPCOL.EQ.1 ) ) ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND. ) , ( K.GT.M ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 2 ) .AND. ) , ( (MOD( ISTART - 1 , HBL ).GE.HBL - 2 ) ) , ( NR.EQ.2 ) , ( (ICURROW( KI ).EQ.MYROW ) ) , ( (ICURCOL( KI ).EQ.MYCOL ) ) , ( (ICURROW( KI ).EQ.MYROW ) ) , ( (( ISPEC.EQ.0 ) .OR. ( NPROW.EQ.1 ) .OR. ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 1 ) ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND. ) , ( (ICURCOL( KI ).EQ.MYCOL ) ) , ( (( ISPEC.EQ.0 ) .OR. ( NPCOL.EQ.1 ) .OR. ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 1 ) ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND. ) , ( WANTZ ) , ( (ICURCOL( KI ).EQ.MYCOL ) ) , ( (( ISPEC.EQ.0 ) .OR. ( NPCOL.EQ.1 ) .OR. ) , ( WE ACTUALLY OWN COLUMN K ) , ( (MOD( K - 1 , HBL ).EQ.HBL - 1 ) ) , ( (( MOD( K - 1 , HBL ).EQ.HBL - 1 ) .AND. ) , ( NPROW.EQ.1 ) , ( (( MOD( K1( KI ) - 1 , HBL ).LT.HBL - 2 ) .AND. ) , ( (( MOD( K2( KI ) , HBL ).LT.HBL - 2 ) .AND. ) , ( (( MOD( K1( KI ) - 1 , HBL ).GE.HBL - 2 ) .AND. ) , ( (( MOD( K2( KI ) , HBL ).GE.HBL - 2 ) .AND. ) , ( NPCOL.EQ.1 ) , ( (( MOD( K1( KI ) - 1 , HBL ).LT.HBL - 2 ) .AND. ) , ( (( MOD( K2( KI ) , HBL ).LT.HBL - 2 ) .AND. ) , ( (( MOD( K1( KI ) - 1 , HBL ).GE.HBL - 2 ) .AND. ) , ( (( MOD( K2( KI ) , HBL ).GE.HBL - 2 ) .AND. ) , ( (K1( KI ).EQ.ISTOP ) ) , ( (( MOD( ISTOP - 1 , HBL ).EQ.HBL - 2 ) .AND. ) , ( (K2( IBULGE ).LE.I - 1 ) ) , ( L.EQ.I ) , ( (( MYROW.EQ.ITMP1 ) .AND. ( MYCOL.EQ.ITMP2 ) ) ) , ( L.EQ.I - 1 ) , ( (( MYROW.EQ.II ) .AND. ( MYCOL.EQ.JJ ) ) ) , ( MODKM1.NE.0 ) , ( NPROW.GT.1 ) , ( NPCOL.GT.1 ) , ( NUM.GT.1 ) , ( MODKM1.EQ.0 ) , ( (( NPROW.GT.1 ) .AND. ( MYCOL.EQ.JJ ) .AND. ) , ( (( NPCOL.GT.1 ) .AND. ( LEFT.EQ.JJ ) .AND. ) , ( (( NUM.GT.1 ) .AND. ( LEFT.EQ.JJ ) .AND. ) , ( H21.GE.ZERO ) , ( JBLK.LE.2*IBLK ) , ( NODE.NE.0 ) , ( M.EQ.L - 10 ) |
| | until | : ( the next release. ) , ( all are started. To offset pipeline ) |
| | while | : ( the deflation is done ) |
|
| List of variables |
$
A( * )
BLOCK_CYCLIC_2D
CONST
CSRC_
CTXT_
DESCA( * )
| DESCZ( * )
DLEN_
DTYPE_
H00
H10
H11
H12
H21
| H22
HALF
I
IBLK
ICOL1
ICOL2
ICURCOL
ICURROW
| IHI
IHIZ
II
ILO
ILOZ
ILWORK
INFO
IROW1
| IROW2
ISPEC
ISTART
ISTOP
ITMP1
ITMP2
ITN
IWORK( * )
| J
JBLK
JJ
K
K1
K2
KCOL
KI
| KP2COL
KP2ROW
KROW
LIHIH
LLD_
LOCALK2
LWORK
M_
| MB_
MODKM1
N
N_
NB_
NR
ONE
RSRC_
| SMALLA
SUM
T1
T2
T3
V2
V3
WANTT
| WANTZ
WI( * )
WORK( * )
WR( * )
Z( * )
ZERO | | close
| |
$
A( * )
BLOCK_CYCLIC_2D
CONST
CSRC_
CTXT_
DESCA( * )
DESCZ( * )
DLEN_
DTYPE_
H00
H10
H11
H12
H21
H22
HALF
I
IBLK
ICOL1
ICOL2
ICURCOL
ICURROW
IHI
IHIZ
II
ILO
ILOZ
ILWORK
INFO
IROW1
IROW2
ISPEC
ISTART
ISTOP
ITMP1
ITMP2
ITN
IWORK( * )
J
JBLK
JJ
K
K1
K2
KCOL
KI
KP2COL
KP2ROW
KROW
LIHIH
LLD_
LOCALK2
LWORK
M_
MB_
MODKM1
N
N_
NB_
NR
ONE
RSRC_
SMALLA
SUM
T1
T2
T3
V2
V3
WANTT
WANTZ
WI( * )
WORK( * )
WR( * )
Z( * )
ZERO
607#357
| |
