From b4ec3afe4cd463d8cd29a8aa45dc1248d4a68eea Mon Sep 17 00:00:00 2001
From: Theophile Terraz <theophile.terraz@inrae.fr>
Date: Wed, 30 Nov 2022 16:01:47 +0100
Subject: [PATCH] debug MPI
---
code/Makefile | 4 +--
code/rubar20.f90 | 69 ++++++++++++++++++++++++++++++++++++++++++------
2 files changed, 63 insertions(+), 10 deletions(-)
diff --git a/code/Makefile b/code/Makefile
index e253486..d909eda 100644
--- a/code/Makefile
+++ b/code/Makefile
@@ -36,8 +36,8 @@ USUAL_FLAG =
OMP_FLAG = -fopenmp
MPI_FLAG = mpifort -cpp -DWITH_MPI
# change to debug_flag to enable debug features
-COMPILE_FLAG = $(OPTIM_FLAG)
-# COMPILE_FLAG = $(DEBUG_FLAG)
+# COMPILE_FLAG = $(OPTIM_FLAG)
+COMPILE_FLAG = $(DEBUG_FLAG)
# METIS_FLAG = -DWITH_METIS
NB_PROC = 2
diff --git a/code/rubar20.f90 b/code/rubar20.f90
index 6087598..bea2410 100644
--- a/code/rubar20.f90
+++ b/code/rubar20.f90
@@ -283,6 +283,9 @@
#endif /* FRA */
end if ! me==scribe
call ledd(alpha,nbpt,xnb,ynb,cvi)
+#ifdef WITH_MPI
+ call mpi_bcast(ifm,1,mpi_integer,scribe,mpi_comm_world,i)
+#endif /* WITH_MPI */
! lecture/edition donnees maillage
if (me==scribe)then
@@ -613,7 +616,9 @@
endif
#ifdef WITH_MPI
+! print*,"plop1 proc",me
call message_gatherv (hae)
+! print*,"plop2 proc",me
#if TRANSPORT_SOLIDE
call message_gatherv2 (cet)
call message_gatherv3(zfn)
@@ -714,6 +719,7 @@
! enddo
write(id_tps,'(f15.3)')tm
write(id_tps,'(8f10.5)') (het(ie),ie=1,ne)
+! call write_contrainte(id_tps)
write(id_tps,'(8f10.5)') (quet(ie),ie=1,ne)
write(id_tps,'(8f10.5)') (qvet(ie),ie=1,ne)
#if WITH_CGNS
@@ -1344,9 +1350,9 @@
if(me/=scribe)then
- if (ifen > 0) then
- call lenv
- endif
+ if (ifen > 0) then
+ call lenv
+ endif
end if ! me/=scribe
@@ -1360,9 +1366,9 @@
call message_gatherv2(qvet)
end if
-
+!
if ((tm.ge.tr .or. tm+dt>=tfin) .and. ifen>0) then
- call message_gatherv2(th)
+! call message_gatherv2(th)
call message_gatherv2(hem)
call message_gatherv2(hem2)
call message_gatherv2(tu)
@@ -9864,6 +9870,9 @@
use module_tableaux
use module_loc, only:hem,qum,qvm,hem2,qum2,qvm2,th,tu
use module_mpi,only:me,scribe,ne_loc,mailles_loc
+#ifdef WITH_MPI
+ use module_messages
+#endif /* WITH_MPI */
!$ use omp_lib
implicit none
@@ -9949,7 +9958,7 @@
! if (me/=scribe)then
!$omp parallel private(vx,vy)
!$omp do schedule(runtime)
- do ieloc = 1,ne_loc(me)
+ do ieloc = 1,ne_loc(me)
ie=mailles_loc(ieloc)
if(het(ie).gt.paray)then
if (het(ie) .gt. hem(ie)) then
@@ -9969,10 +9978,22 @@
tu(ie) = tm
endif
endif
- enddo
+ enddo
!$omp end do
!$omp end parallel
! end if !if me/=scribe
+#ifdef WITH_MPI
+ if ((tm.ge.tr .or. tm.eq.tmax) .and. ifen>0) then
+ call message_gatherv2(th)
+ call message_gatherv2(hem)
+ call message_gatherv2(hem2)
+ call message_gatherv2(tu)
+ call message_gatherv2(qum)
+ call message_gatherv2(qum2)
+ call message_gatherv2(qvm)
+ call message_gatherv2(qvm2)
+ end if
+#endif /* WITH_MPI */
! ecriture sur etude.env en meme temps qu'ecriture sur tps
if (me==scribe)then
if (tm.ge.tr .or. tm.eq.tmax) then
@@ -19263,6 +19284,7 @@
use module_tableaux,only:ne,na,nn,eau,het,quet,qvet,fra,fre,cofr&
&,ieva,iae,nne,g,smbh,zfn,dt,cfl,dt0,t,eps1,eps2,paray,idl
+ use module_precision,only:wp
use module_ts,only : efficace,sigma0,rapsmbhc,nbcoun,epcoun,diamn,sigman&
&,taummn,diame2,sigmae2,exposantaucr,camenen,hansen,exptaucr,taue,rapva,smbhc&
@@ -19279,7 +19301,7 @@
#endif /* WITH_MPI */
use module_mpi,only:ne_loc,me,mailles_loc,naloc,aretes_loc
- ! implicit none
+ implicit none
!!!!! real(WP),intent(inout) :: cofr2(ne),zfn2(nn),zfa2(na),zfe2(ne)!,ceq(ne)
real(WP) :: rapve(ne),vitchut2
@@ -22851,4 +22873,35 @@ use module_precision
end
#endif /* fin section transport solide */
+! subroutine write_contrainte(id_tps)
+! use module_tableaux,only:ne,het,fra,fre,quet,qvet,paray,g,cofr
+! use module_precision,only:wp
+! implicit none
+! real(WP) :: q,vstar,fra2,vfr
+! integer,intent(in) :: id_tps
+! integer :: ie
+! real(WP), dimension(:),allocatable :: vitfro
+!
+! allocate(vitfro(ne))
+! do ie=1,ne
+! if(het(ie).gt.paray)then
+! q=sqrt(quet(ie)**2+qvet(ie)**2)
+! vstar=q/het(ie)
+! ! on limite la vitesse
+! if(vstar.gt.10.)vstar=10.
+! ! calcul du coefficient de chezy
+! fra2=max(fra(ie),fre(ie)*het(ie)**0.3333333)
+! ! on divise par cofr car la surface est integree dans fra ou fre
+! fra2=fra2/cofr(ie)
+! ! calcul de la vitesse de frottement
+! vfr= sqrt(g)*vstar/sqrt(fra2)
+! vitfro(ie)=vfr
+! else
+! vitfro(ie)=0.
+! endif
+! enddo
+! deallocate(vitfro)
+! write(id_tps,'(8f10.5)') (vitfro(ie),ie=1,ne)
+! end subroutine write_contrainte
+
--
GitLab