!-------------------------------------- LICENCE BEGIN ------------------------------------
!Environment Canada - Atmospheric Science and Technology License/Disclaimer,
! version 3; Last Modified: May 7, 2008.
!This is free but copyrighted software; you can use/redistribute/modify it under the terms
!of the Environment Canada - Atmospheric Science and Technology License/Disclaimer
!version 3 or (at your option) any later version that should be found at:
!http://collaboration.cmc.ec.gc.ca/science/rpn.comm/license.html
!
!This software is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY;
!without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
!See the above mentioned License/Disclaimer for more details.
!You should have received a copy of the License/Disclaimer along with this software;
!if not, you can write to: EC-RPN COMM Group, 2121 TransCanada, suite 500, Dorval (Quebec),
!CANADA, H9P 1J3; or send e-mail to service.rpn@ec.gc.ca
!-------------------------------------- LICENCE END --------------------------------------
***s/r sol_fft8_lam - parallel direct solution of horizontal Helmholtz
* problem. With ffft8 for LAM
*
#include "model_macros_f.h"
*
subroutine sol_fft8_lam( sol, Rhs, pri, 1,2
$ Minx, Maxx, Miny, Maxy, njl,
$ Minz, Maxz, Nk, Nkl, Gni, Gnj,
$ Minij, Maxij, L_nij,
$ minx1, maxx1, minx2, maxx2,nx3,
$ F_npex1, F_npey1, ai, bi, ci,
$ fdg2,fdwfft)
implicit none
#include "ptopo.cdk"
#include "glb_ld.cdk"
#include "glb_pil.cdk"
*
*
*
*author Abdessamad Qaddouri- JULY 1999
*
*revision
* v3_10 - Corbeil & Desgagne & Lee - AIXport+Opti+OpenMP
*
*arguments
* Name I/O Description
*----------------------------------------------------------------
* Sol O - result of solver
* Rhs I - r.h.s. of elliptic equation
* Pri I - inverse projector in Fourier space
* Minx I - minimum index on X for Rhs,Sol
* Maxx I - maximum index on X for Rhs,Sol
* Miny I - minimum index on Y for Rhs,Sol
* Maxy I - maximum index on Y for Rhs,Sol
* Njl I - number of points on local PEy for J (ldnh_nj)
* Minz I - minimum index on local PEx for K (trp_12smin)
* Maxz I - maximum index on local PEx for K (trp_12smax)
* Nk I - G_nk-1 points in z direction globally (Schm_nith)
* Nkl I - number of points on local PEx for K (trp_12sn)
* Gni I - number of points in x direction globally (G_ni)
* Gnj I - number of points in y direction globally (G_nj)
* Minij I - minimum index on local PEy for I (trp_22min)
* Maxij I - maximum index on local PEy for I (trp_22max)
* L_nij I - number of points on local PEy for I (trp_22n)
* Minx1 I - minimum index on local PEx for K (trp_12smin)
* Maxx1 I - maximum index on local PEx for K (trp_12smax)
* Minx2 I - minimum index on local PEy for I (trp_22min)
* Maxx2 I - maximum index on local PEy for I (trp_22max)
* Nx3 I - number of points along J globally (G_nj)
* F_npex1 I - number of processors on X
* F_npey1 I - number of processors on Y
* ai I - sub diagonal of LU factorization
* bi I - diagonal of LU factorization
* ci I - super diagonal of LU factorization
* fdg2 I - work field
* fdwfft I - work field
*
*modules
external ffft8, rpn_comm_transpose
integer F_npex1, F_npey1
integer minx1, maxx1, minx2, maxx2,nx3
Real*8 ai(minx1:maxx1,minx2:maxx2,nx3),
$ bi(minx1:maxx1,minx2:maxx2,nx3),
$ ci(minx1:maxx1,minx2:maxx2,nx3)
integer Minx, Maxx, Miny, Maxy, njl
integer Minz, Maxz, Nk, Nkl
integer Gni, Gnj
integer Minij, Maxij, L_nij
real*8 Sol(Minx:Maxx,Miny:Maxy,Nk), Rhs(Minx:Maxx,Miny:Maxy,Nk)
real*8 pri
**
real*8 fdwfft(Miny:Maxy,Minz:Maxz,Gni+2+F_npex1)
real*8 fdg2(Minz:Maxz,Minij:Maxij,Gnj+F_npey1)
*
integer i, j, k, ki, jr, err,l_pil_w,l_pil_e
integer piece,p0,pn,plon,ptotal
real*8 zero, one
parameter( zero = 0.0 )
parameter( one = 1.0 )
*
* The I vector lies on the Y processor so, l_pil_w and l_pil_e will
* represent the pilot region along I
c call tmg_start(89,'sol_fft_lam')
l_pil_w=0
l_pil_e=0
if (l_south) l_pil_w= Lam_pil_w
if (l_north) l_pil_e= Lam_pil_e
call rpn_comm_transpose( Rhs, Minx, Maxx, Gni, (Maxy-Miny+1),
% Minz, Maxz, Nk, fdwfft, 1, 2 )
* projection ( wfft = x transposed * g )
!$omp parallel private(jr,p0,pn,piece) shared(plon,ptotal)
!$omp do
do i= 1,Gni
do k= Minz, nkl
do j= njl+1-pil_n,Maxy
fdwfft(j,k,i)=zero
enddo
enddo
do k= Minz, nkl
do j= Miny, pil_s
fdwfft(j,k,i)=zero
enddo
enddo
do k= Nkl+1,Maxz
do j= Miny,Maxy
fdwfft(j,k,i)=zero
enddo
enddo
do k= Minz, 0
do j= Miny,Maxy
fdwfft(j,k,i)=zero
enddo
enddo
enddo
!$omp enddo
*
!$omp do
do k=1,Nkl
call qcfft8(fdwfft(1+pil_s,k,1+Lam_pil_w),
% (Maxy-Miny+1)*(Maxz-Minz+1),1,
% (Maxy-Miny+1-pil_s-pil_n), -1 )
enddo
!$omp enddo
c call qcfft8 ( fdwfft,(Maxy-Miny+1)*(Maxz-Minz+1), 1,
c % (Maxy-Miny+1) * Nkl, -1 )
!$omp do
do i = 0+Lam_pil_w, Gni-1-Lam_pil_e
do k = 1, Nkl
do j = 1+pil_s, (Maxy-Miny+1)-pil_n
fdwfft(j,k,i+1) = pri * fdwfft(j,k,i+1)
enddo
enddo
enddo
!$omp enddo
*
!$omp single
call rpn_comm_transpose
$ ( fdwfft, Miny, Maxy, Gnj, (Maxz-Minz+1),
$ Minij, Maxij, Gni, fdg2, 2, 2 )
!$omp end single
*
ptotal = L_nij-l_pil_e-l_pil_w-1
plon = (ptotal+Ptopo_npeOpenMP)/ Ptopo_npeOpenMP
!$omp do
do piece=1,Ptopo_npeOpenMP
p0 = 1+l_pil_w + plon*(piece-1)
pn = min(L_nij-l_pil_e,plon*piece+l_pil_w)
j =1+Lam_pil_s
c do ki= 1, (Maxz-Minz+1) *L_nij
c fdg2(ki,1,j) = bi(ki,1,j)*fdg2(ki,1,j)
c enddo
! do i=1+l_pil_w,L_nij-l_pil_e
do i=p0,pn
do k=1,Nkl
fdg2(k,i,j) = bi(k,i,j)*fdg2(k,i,j)
enddo
enddo
do j =2+Lam_pil_s, Gnj-Lam_pil_n
jr = j - 1
! do i=1+l_pil_w,L_nij-l_pil_e
do i=p0,pn
do k=1,Nkl
fdg2(k,i,j) = bi(k,i,j)*fdg2(k,i,j) - ai(k,i,j)
$ * fdg2(k,i,jr)
enddo
enddo
enddo
c do j = Gnj-1, 1, -1
c jr = j + 1
c do ki= 1, (Maxz-Minz+1)*L_nij
c fdg2(ki,1,j) = fdg2(ki,1,j) - ci(ki,1,j) * fdg2(ki,1,jr)
c enddo
c enddo
do j = Gnj-1-Lam_pil_n, 1+Lam_pil_s, -1
jr = j + 1
! do i=1+l_pil_w,L_nij-l_pil_e
do i=p0,pn
do k=1,Nkl
fdg2(k,i,j) = fdg2(k,i,j) - ci(k,i,j) * fdg2(k,i,jr)
enddo
enddo
enddo
enddo
!$omp enddo
*
!$omp single
call rpn_comm_transpose
$ ( fdwfft, Miny, Maxy, Gnj, (Maxz-Minz+1),
$ Minij, Maxij, Gni, fdg2,- 2, 2 )
!$omp end single
* inverse projection ( r = x * w )
!$omp do
do k=1, Nkl
call qcfft8 ( fdwfft(1+pil_s,k,1+Lam_pil_w),
% (Maxy-Miny+1)*(Maxz-Minz+1),1,
% (Maxy-Miny+1-pil_s-pil_n), +1 )
enddo
!$omp enddo
!$omp end parallel
c call qcfft8 ( fdwfft,(Maxy-Miny+1)*(Maxz-Minz+1), 1,
c % (Maxy-Miny+1) * Nkl, +1 )
*
call rpn_comm_transpose( Sol, Minx, Maxx, Gni, (Maxy-Miny+1),
% Minz, Maxz, Nk, fdwfft, -1, 2 )
c call tmg_stop(89)
return
end