mirror of https://gitlab.com/QEF/q-e.git
Fixed a nasty bug in the algorithm for parallel inversion of a triangular matrix.
This was probably responsible for the crashes of the parallel version of Davidson. C.S. git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@3575 c92efa57-630b-4861-b058-cf58834340f0
This commit is contained in:
sbraccia
parent
6e55c387ee
commit
1941a5ceaa
|
|
@ -3191,7 +3191,8 @@ SUBROUTINE para_dgemm( transa, transb, m, n, k, &
|
|||
!
|
||||
#if defined (__XD1)
|
||||
!
|
||||
CALL rep_matmul_drv( TRANSA, TRANSB, M, N, K, ALPHA, A, LDA, B, LDB, BETA, C, LDC, comm )
|
||||
CALL rep_matmul_drv( transa, transb, m, n, k, &
|
||||
alpha, a, lda, b, ldb, beta, c, ldc, comm )
|
||||
RETURN
|
||||
!
|
||||
#endif
|
||||
|
|
@ -3269,19 +3270,21 @@ SUBROUTINE para_zgemm( transa, transb, m, n, k, &
|
|||
COMPLEX(DP), INTENT(INOUT) :: a(lda,*), b(ldb,*), c(ldc,*)
|
||||
INTEGER, INTENT(IN) :: comm
|
||||
!
|
||||
COMPLEX(DP), PARAMETER :: ONE=(1.0_DP, 0.0_DP), ZERO=(0.0_DP, 0.0_DP)
|
||||
INTEGER :: i, mpime, nproc, ierr
|
||||
INTEGER :: ncol, i0, i1
|
||||
INTEGER, ALLOCATABLE :: i0a(:), i1a(:)
|
||||
INTEGER :: i, mpime, nproc, ierr
|
||||
INTEGER :: ncol, i0, i1
|
||||
INTEGER, ALLOCATABLE :: i0a(:), i1a(:)
|
||||
!
|
||||
COMPLEX(DP), PARAMETER :: ONE = (1.0_DP,0.0_DP), ZERO = ( 0.0_DP, 0.0_DP )
|
||||
!
|
||||
! ... quick return if possible
|
||||
!
|
||||
IF ( m == 0 .OR. n == 0 .OR. &
|
||||
( ( alpha == 0.D0 .OR. k == 0 ) .AND. beta == ONE ) ) RETURN
|
||||
|
||||
!
|
||||
#if defined (__XD1)
|
||||
!
|
||||
CALL zrep_matmul_drv( TRANSA, TRANSB, M, N, K, ALPHA, A, LDA, B, LDB, BETA, C, LDC, comm )
|
||||
CALL zrep_matmul_drv( transa, transb, m, n, k, &
|
||||
alpha, a, lda, b, ldb, beta, c, ldc, comm )
|
||||
RETURN
|
||||
!
|
||||
#endif
|
||||
|
|
@ -3451,7 +3454,7 @@ SUBROUTINE para_dgemv( trans, m, n, alpha, &
|
|||
!
|
||||
y(j) = ydum(i)
|
||||
!
|
||||
END DO
|
||||
END DO
|
||||
!
|
||||
DEALLOCATE( ydum, i0a, i1a )
|
||||
!
|
||||
|
|
@ -3645,11 +3648,12 @@ SUBROUTINE para_zcholdc( n, a, lda, comm )
|
|||
COMPLEX(DP), INTENT(INOUT) :: a(lda,*)
|
||||
INTEGER, INTENT(IN) :: comm
|
||||
!
|
||||
COMPLEX(DP), PARAMETER :: ONE=(1.0_DP, 0.0_DP), ZERO=(0.0_DP, 0.0_DP)
|
||||
INTEGER :: i, j
|
||||
REAL(DP) :: aii
|
||||
COMPLEX(DP), EXTERNAL :: ZDOTC
|
||||
!
|
||||
COMPLEX(DP), PARAMETER :: ONE = (1.0_DP,0.0_DP), ZERO = ( 0.0_DP, 0.0_DP )
|
||||
!
|
||||
!
|
||||
DO i = 1, n
|
||||
!
|
||||
|
|
@ -3703,9 +3707,10 @@ SUBROUTINE para_dtrtri( n, a, lda, comm )
|
|||
INTEGER, INTENT(IN) :: comm
|
||||
!
|
||||
INTEGER :: i, j, k
|
||||
INTEGER :: i0, i1, dim, mpime, nproc, ierr
|
||||
INTEGER :: i0, i1, mpime, nproc, ierr
|
||||
INTEGER, ALLOCATABLE :: i0a(:), i1a(:)
|
||||
REAL(DP) :: sum, area
|
||||
REAL(DP) :: an, xfrac
|
||||
REAL(DP) :: sum
|
||||
REAL(DP), ALLOCATABLE :: inva(:,:)
|
||||
!
|
||||
!
|
||||
|
|
@ -3720,29 +3725,24 @@ SUBROUTINE para_dtrtri( n, a, lda, comm )
|
|||
mpime = 0
|
||||
!
|
||||
#endif
|
||||
!
|
||||
dim = n
|
||||
area = DBLE( n*n / 2 ) / DBLE( nproc )
|
||||
!
|
||||
ALLOCATE( i0a( 0:nproc-1 ), i1a( 0:nproc-1 ) )
|
||||
!
|
||||
i0a(0) = 1
|
||||
i1a(0) = 1 + ANINT( dim - SQRT( MAX( dim*dim - 2.D0*area, 0.D0 ) ) )
|
||||
i1a(0) = MIN( i1a(0), n )
|
||||
an = 1.D0 / DBLE( nproc )
|
||||
!
|
||||
DO i = 1, nproc - 1
|
||||
i0a(0) = 1
|
||||
!
|
||||
DO i = 0, nproc - 2
|
||||
!
|
||||
dim = n - i1a(i-1)
|
||||
xfrac = 1.D0 - SQRT( 1.D0 - DBLE( i+1 )*an )
|
||||
!
|
||||
i0a(i) = i1a(i-1) + 1
|
||||
!
|
||||
i1a(i) = i1a(i-1) + &
|
||||
ANINT( dim - SQRT( MAX( dim*dim - 2.D0*area, 0.D0 ) ) )
|
||||
!
|
||||
i1a(i) = MIN( i1a(i), n )
|
||||
i1a(i) = ANINT( xfrac*n )
|
||||
i0a(i+1) = i1a(i) + 1
|
||||
!
|
||||
END DO
|
||||
!
|
||||
i1a(nproc-1) = n
|
||||
!
|
||||
i0 = i0a(mpime)
|
||||
i1 = i1a(mpime)
|
||||
!
|
||||
|
|
@ -3777,7 +3777,11 @@ SUBROUTINE para_dtrtri( n, a, lda, comm )
|
|||
!
|
||||
DO i = 0 , nproc - 1
|
||||
!
|
||||
#if defined __XD1
|
||||
CALL BCAST_REAL( a(1,i0a(i)), i1a(i)-i0a(i)+1, i, comm, ierr )
|
||||
#else
|
||||
CALL mp_bcast( a(1:n,i0a(i):i1a(i)), i, comm )
|
||||
#endif
|
||||
!
|
||||
END DO
|
||||
!
|
||||
|
|
@ -3793,7 +3797,7 @@ SUBROUTINE para_ztrtri( n, a, lda, comm )
|
|||
!
|
||||
! ... parallel inversion of a lower trinagular matrix done distributing
|
||||
! ... by columns ( the number of columns assigned to each processor are
|
||||
! ... chosen to optimize the load balance )
|
||||
! ... chosen to optimize the load balance in the limit of large matrices )
|
||||
!
|
||||
USE kinds, ONLY : DP
|
||||
USE mp, ONLY : mp_bcast
|
||||
|
|
@ -3801,19 +3805,20 @@ SUBROUTINE para_ztrtri( n, a, lda, comm )
|
|||
!
|
||||
IMPLICIT NONE
|
||||
!
|
||||
COMPLEX(DP), PARAMETER :: ONE=(1.0_DP, 0.0_DP), ZERO=(0.0_DP, 0.0_DP)
|
||||
INTEGER, INTENT(IN) :: n
|
||||
INTEGER, INTENT(IN) :: lda
|
||||
COMPLEX(DP), INTENT(INOUT) :: a(lda,*)
|
||||
INTEGER, INTENT(IN) :: comm
|
||||
!
|
||||
INTEGER :: i, j, k
|
||||
INTEGER :: i0, i1, dim, mpime, nproc, ierr
|
||||
INTEGER, ALLOCATABLE :: i0a(:), i1a(:)
|
||||
REAL(DP) :: area
|
||||
INTEGER :: i0, i1, mpime, nproc, ierr
|
||||
INTEGER, ALLOCATABLE :: i0a(:), i1a(:)
|
||||
REAL(DP) :: an, xfrac
|
||||
COMPLEX(DP) :: sum
|
||||
COMPLEX(DP), ALLOCATABLE :: inva(:,:)
|
||||
!
|
||||
COMPLEX(DP), PARAMETER :: ONE = (1.0_DP,0.0_DP), ZERO = ( 0.0_DP, 0.0_DP )
|
||||
!
|
||||
!
|
||||
#if defined (__MPI)
|
||||
!
|
||||
|
|
@ -3826,29 +3831,24 @@ SUBROUTINE para_ztrtri( n, a, lda, comm )
|
|||
mpime = 0
|
||||
!
|
||||
#endif
|
||||
!
|
||||
dim = n
|
||||
area = DBLE( n*n / 2 ) / DBLE( nproc )
|
||||
!
|
||||
ALLOCATE( i0a( 0:nproc-1 ), i1a( 0:nproc-1 ) )
|
||||
!
|
||||
i0a(0) = 1
|
||||
i1a(0) = 1 + ANINT( dim - SQRT( MAX( dim*dim - 2.D0*area, 0.D0 ) ) )
|
||||
i1a(0) = MIN( i1a(0), n )
|
||||
an = 1.D0 / DBLE( nproc )
|
||||
!
|
||||
DO i = 1, nproc - 1
|
||||
i0a(0) = 1
|
||||
!
|
||||
DO i = 0, nproc - 2
|
||||
!
|
||||
dim = n - i1a(i-1)
|
||||
xfrac = 1.D0 - SQRT( 1.D0 - DBLE( i+1 )*an )
|
||||
!
|
||||
i0a(i) = i1a(i-1) + 1
|
||||
!
|
||||
i1a(i) = i1a(i-1) + &
|
||||
ANINT( dim - SQRT( MAX( dim*dim - 2.D0*area, 0.D0 ) ) )
|
||||
!
|
||||
i1a(i) = MIN( i1a(i), n )
|
||||
i1a(i) = ANINT( xfrac*n )
|
||||
i0a(i+1) = i1a(i) + 1
|
||||
!
|
||||
END DO
|
||||
!
|
||||
i1a(nproc-1) = n
|
||||
!
|
||||
i0 = i0a(mpime)
|
||||
i1 = i1a(mpime)
|
||||
!
|
||||
|
|
@ -3883,11 +3883,7 @@ SUBROUTINE para_ztrtri( n, a, lda, comm )
|
|||
!
|
||||
DO i = 0 , nproc - 1
|
||||
!
|
||||
#if defined __XD1
|
||||
CALL BCAST_REAL( a( 1, i0a(i) ), i1a(i)-i0a(i)+1, i, comm, ierr )
|
||||
#else
|
||||
CALL mp_bcast( a(1:n,i0a(i):i1a(i)), i, comm )
|
||||
#endif
|
||||
!
|
||||
END DO
|
||||
!
|
||||
|
|
@ -3898,9 +3894,6 @@ SUBROUTINE para_ztrtri( n, a, lda, comm )
|
|||
END SUBROUTINE para_ztrtri
|
||||
|
||||
!
|
||||
!
|
||||
!
|
||||
|
||||
SUBROUTINE sqr_mm_cannon( transa, transb, n, alpha, a, lda, b, ldb, beta, c, ldc, dims, coor, comm )
|
||||
!
|
||||
! Parallel square matrix multiplication with Cannon's algorithm
|
||||
|
|
|
|||
Loading…
Reference in New Issue