OpenRadioss 2025.1.11
OpenRadioss project
Loading...
Searching...
No Matches
pslared2d.f File Reference

Go to the source code of this file.

Functions/Subroutines

subroutine pslared2d (n, ia, ja, desc, byrow, byall, work, lwork)

Function/Subroutine Documentation

◆ pslared2d()

subroutine pslared2d ( integer n,
integer ia,
integer ja,
integer, dimension( * ) desc,
real, dimension( * ) byrow,
real, dimension( * ) byall,
real, dimension( lwork ) work,
integer lwork )

Definition at line 1 of file pslared2d.f.

2*
3* -- ScaLAPACK routine (version 1.7) --
4* University of Tennessee, Knoxville, Oak Ridge National Laboratory,
5* and University of California, Berkeley.
6* December 12, 2005
7*
8* .. Scalar Arguments ..
9 INTEGER IA, JA, LWORK, N
10* ..
11* .. Array Arguments ..
12 INTEGER DESC( * )
13 REAL BYALL( * ), BYROW( * ), WORK( LWORK )
14* ..
15*
16* Purpose
17* =======
18*
19* PSLARED2D redistributes a 1D array
20*
21* It assumes that the input array, BYROW, is distributed across
22* columns and that all process rows contain the same copy of
23* BYROW. The output array, BYALL, will be identical on all processes
24* and will contain the entire array.
25*
26* Notes
27* =====
28*
29* Each global data object is described by an associated description
30* vector. This vector stores the information required to establish
31* the mapping between an object element and its corresponding process
32* and memory location.
33*
34* Let A be a generic term for any 2D block cyclicly distributed array.
35* Such a global array has an associated description vector DESCA.
36* In the following comments, the character _ should be read as
37* "of the global array".
38*
39* NOTATION STORED IN EXPLANATION
40* --------------- -------------- --------------------------------------
41* DTYPE_A(global) DESCA( DTYPE_ )The descriptor type. In this case,
42* DTYPE_A = 1.
43* CTXT_A (global) DESCA( CTXT_ ) The BLACS context handle, indicating
44* the BLACS process grid A is distribu-
45* ted over. The context itself is glo-
46* bal, but the handle (the integer
47* value) may vary.
48* M_A (global) DESCA( M_ ) The number of rows in the global
49* array A.
50* N_A (global) DESCA( N_ ) The number of columns in the global
51* array A.
52* MB_A (global) DESCA( MB_ ) The blocking factor used to distribute
53* the rows of the array.
54* NB_A (global) DESCA( NB_ ) The blocking factor used to distribute
55* the columns of the array.
56* RSRC_A (global) DESCA( RSRC_ ) The process row over which the first
57* row of the array A is distributed.
58* CSRC_A (global) DESCA( CSRC_ ) The process column over which the
59* first column of the array A is
60* distributed.
61* LLD_A (local) DESCA( LLD_ ) The leading dimension of the local
62* array. LLD_A >= MAX(1,LOCr(M_A)).
63*
64* Let K be the number of rows or columns of a distributed matrix,
65* and assume that its process grid has dimension p x q.
66* LOCr( K ) denotes the number of elements of K that a process
67* would receive if K were distributed over the p processes of its
68* process column.
69* Similarly, LOCc( K ) denotes the number of elements of K that a
70* process would receive if K were distributed over the q processes of
71* its process row.
72* The values of LOCr() and LOCc() may be determined via a call to the
73* ScaLAPACK tool function, NUMROC:
74* LOCr( M ) = NUMROC( M, MB_A, MYROW, RSRC_A, NPROW ),
75* LOCc( N ) = NUMROC( N, NB_A, MYCOL, CSRC_A, NPCOL ).
76* An upper bound for these quantities may be computed by:
77* LOCr( M ) <= ceil( ceil(M/MB_A)/NPROW )*MB_A
78* LOCc( N ) <= ceil( ceil(N/NB_A)/NPCOL )*NB_A
79*
80* Arguments
81* =========
82*
83* NP = Number of local rows in BYROW()
84*
85* N (global input) INTEGER
86* The number of elements to be redistributed. N >= 0.
87*
88* IA (global input) INTEGER
89* IA must be equal to 1
90*
91* JA (global input) INTEGER
92* JA must be equal to 1
93*
94* DESC (global/local input) INTEGER Array of dimension DLEN_
95* A 2D array descriptor, which describes BYROW
96*
97* BYROW (local input) distributed block cyclic REAL array
98* global dimension (N), local dimension (NP)
99* BYROW is distributed across the process columns
100* All process rows are assumed to contain the same value
101*
102* BYALL (global output) REAL global dimension( N )
103* local dimension (N)
104* BYALL is exactly duplicated on all processes
105* It contains the same values as BYROW, but it is replicated
106* across all processes rather than being distributed
107*
108* BYALL(i) = BYROW( NUMROC(i,DESC( MB_ ),MYCOL,0,NPCOL ) on the procs
109* whose MYCOL == mod((i-1)/DESC( MB_ ),NPCOL)
110*
111* WORK (local workspace) REAL dimension (LWORK)
112* Used to hold the buffers sent from one process to another
113*
114* LWORK (local input) INTEGER size of WORK array
115* LWORK >= NUMROC(N, DESC( MB_ ), 0, 0, NPROW)
116*
117* .. Parameters ..
118 INTEGER BLOCK_CYCLIC_2D, DLEN_, DTYPE_, CTXT_, M_, N_,
119 $ MB_, NB_, RSRC_, CSRC_, LLD_
120 parameter( block_cyclic_2d = 1, dlen_ = 9, dtype_ = 1,
121 $ ctxt_ = 2, m_ = 3, n_ = 4, mb_ = 5, nb_ = 6,
122 $ rsrc_ = 7, csrc_ = 8, lld_ = 9 )
123* ..
124* .. Local Scalars ..
125 INTEGER ALLI, BUFLEN, I, II, MB, MYCOL, MYROW, NPCOL,
126 $ NPROW, PROW
127* ..
128* .. External Functions ..
129 INTEGER NUMROC
130 EXTERNAL numroc
131* ..
132* .. External Subroutines ..
133 EXTERNAL blacs_gridinfo, scopy, sgebr2d, sgebs2d
134* ..
135* .. Intrinsic Functions ..
136 INTRINSIC min
137* ..
138* .. Executable Statements ..
139* This is just to keep ftnchek happy
140 IF( block_cyclic_2d*csrc_*ctxt_*dlen_*dtype_*lld_*mb_*m_*nb_*n_*
141 $ rsrc_.LT.0 )RETURN
142*
143 CALL blacs_gridinfo( desc( ctxt_ ), nprow, npcol, myrow, mycol )
144 mb = desc( mb_ )
145*
146 DO 30 prow = 0, nprow - 1
147 buflen = numroc( n, mb, prow, 0, nprow )
148 IF( myrow.EQ.prow ) THEN
149 CALL scopy( buflen, byrow, 1, work, 1 )
150 CALL sgebs2d( desc( ctxt_ ), 'C', ' ', buflen, 1, work,
151 $ buflen )
152 ELSE
153 CALL sgebr2d( desc( ctxt_ ), 'C', ' ', buflen, 1, work,
154 $ buflen, prow, mycol )
155 END IF
156*
157 alli = prow*mb
158 DO 20 ii = 1, buflen, mb
159 DO 10 i = 1, min( mb, buflen-ii+1 )
160 byall( alli+i ) = work( ii-1+i )
161 10 CONTINUE
162 alli = alli + mb*nprow
163 20 CONTINUE
164 30 CONTINUE
165*
166 RETURN
167*
168* End of PSLARED2D
169*
scopy
subroutine scopy(n, sx, incx, sy, incy)
SCOPY
Definition scopy.f:82
min
#define min(a, b)
Definition macros.h:20
sgebs2d
subroutine sgebs2d(contxt, scope, top, m, n, a, lda)
Definition mpi.f:1072
sgebr2d
subroutine sgebr2d(contxt, scope, top, m, n, a, lda)
Definition mpi.f:1113
blacs_gridinfo
subroutine blacs_gridinfo(cntxt, nprow, npcol, myrow, mycol)
Definition mpi.f:754
numroc
integer function numroc(n, nb, iproc, isrcproc, nprocs)
Definition mpi.f:786