Merge branch 'develop' of iffgit.fz-juelich.de:fleur/fleur into develop

force/force_a21_U.f90

@@ -104,6 +104,7 @@ CONTAINS
        END DO     ! lo = 1,atoms%nlo
 
        i_u = i_u + 1
+       IF(i_u.GT.atoms%n_u) EXIT
     END DO
 
   END SUBROUTINE force_a21_U

force/geo.f90

@@ -41,6 +41,7 @@ CONTAINS
     USE m_constants
     USE m_rinpXML
     USE m_winpXML
+    USE m_init_wannier_defaults
     IMPLICIT NONE
     TYPE(t_oneD),INTENT(IN)   :: oneD
     TYPE(t_cell),INTENT(IN)   :: cell
@@ -228,11 +229,13 @@ CONTAINS
                input_temp%comment)
     
        ELSE
+          kpts_temp%ntet = 1
           kpts_temp%numSpecialPoints = 1
           ALLOCATE(kpts_temp%specialPoints(3,kpts_temp%numSpecialPoints))
           ALLOCATE(noel_temp(1),atomTypeSpecies(1),speciesRepAtomType(1))
           ALLOCATE(xmlElectronStates(1,1),xmlPrintCoreStates(1,1))
           ALLOCATE(xmlCoreOccs(1,1,1))
+          CALL initWannierDefaults(wann_temp)
           CALL r_inpXML(atoms_temp,obsolete_temp,vacuum_temp,input_temp,stars_temp,sliceplot_temp,&
                         banddos_temp,dimension_temp,cell_temp,sym_temp,xcpot_temp,noco_temp,Jij_temp,&
                         oneD_temp,hybrid_temp,kpts_temp,enpara_temp,coreSpecInput_temp,wann_temp,noel_temp,&
@@ -240,7 +243,8 @@ CONTAINS
                         xmlPrintCoreStates,xmlCoreOccs,atomTypeSpecies,speciesRepAtomType,l_kpts_temp)
           numSpecies = SIZE(speciesRepAtomType)
           filename = 'inp_new.xml'
-          input_temp%l_f = input%l_f
+          input_temp = input
+          !input_temp%l_f = input%l_f
           input_temp%gw_neigd = dimension_temp%neigd
           div(:) = MIN(kpts_temp%nkpt3(:),1)
           stars_temp%gmax = stars_temp%gmaxInit

global/types.F90

@@ -45,6 +45,7 @@ MODULE m_types
      REAL :: z
      COMPLEX :: p,m
   END TYPE t_orblo
+
   TYPE t_lapw
      INTEGER :: nv(2)
      INTEGER :: nv_tot

io/writeOutParameters.f90

@@ -53,6 +53,10 @@ SUBROUTINE writeOutParameters(mpi,input,sym,stars,atoms,vacuum,obsolete,kpts,&
    CALL writeXMLElementFormPoly('density',(/'ng3','ng2'/),&
                                 attributes(:2),reshape((/7,6,8,8/),(/2,2/)))
 
+   WRITE(attributes(1),'(i0)') dimension%neigd
+   CALL writeXMLElementFormPoly('bands',(/'numbands'/),&
+                                attributes(:1),reshape((/9,8/),(/1,2/)))
+
    CALL closeXMLElement('numericalParameters')
 
 END SUBROUTINE writeOutParameters

io/xmlOutput.F90

@@ -48,6 +48,7 @@ MODULE m_xmlOutput
       USE m_constants
       USE m_utility
       USE m_compile_descr
+!$    use omp_lib
       
       IMPLICIT NONE
 
@@ -56,6 +57,7 @@ MODULE m_xmlOutput
       INTEGER           :: err, isize
 #endif
       INTEGER           :: numFlags
+      INTEGER           :: nOMPThreads
       CHARACTER(LEN=8)  :: date
       CHARACTER(LEN=10) :: time
       CHARACTER(LEN=10) :: zone
@@ -66,6 +68,7 @@ MODULE m_xmlOutput
       CHARACTER(LEN=20) :: structureSpecifiers(11)
       CHARACTER(LEN=50) :: gitdesc,githash,gitbranch,compile_date,compile_user,compile_host
       CHARACTER(LEN=50) :: compile_flags,link_flags
+      CHARACTER(LEN=20) :: attributes(7)
       
       maxNumElements = 10
       ALLOCATE(elementList(maxNumElements))
@@ -104,10 +107,24 @@ MODULE m_xmlOutput
          CALL writeXMLElementNoAttributes('additionalCompilerFlags',flags(1:numFlags))
       END IF
       CALL closeXMLElement('programVersion')
+
+      CALL openXMLElementNoAttributes('parallelSetup')
+      nOMPThreads = -1
+      !$ nOMPThreads=omp_get_max_threads()
+      IF(nOMPThreads.NE.-1) THEN
+         WRITE(attributes(1),'(i0)') nOMPThreads
+         CALL writeXMLElementFormPoly('openMP',(/'ompThreads'/),&
+                                      attributes(:1),reshape((/10,8/),(/1,2/)))
+      END IF
+
 #ifdef CPP_MPI
       CALL MPI_COMM_SIZE(MPI_COMM_WORLD,isize,err)
-      CALL writeXMLElementPoly('parallelizationParameters',(/'mpiPEs'/),(/isize/))
+      WRITE(attributes(1),'(i0)') isize
+      CALL writeXMLElementFormPoly('mpi',(/'mpiProcesses'/),&
+                                   attributes(:1),reshape((/13,8/),(/1,2/)))
 #endif
+      CALL closeXMLElement('parallelSetup')
+
       CALL writeXMLElement('startDateAndTime',(/'date','time','zone'/),(/dateString,timeString,zone/))
    END SUBROUTINE startXMLOutput
 

mix/broyden2.F90

@@ -63,6 +63,7 @@ CONTAINS
 
     ! Local Scalars
     INTEGER         :: i,j,it,k,nit,iread,nmaph, mit, historyLength
+    INTEGER         :: relIndex
     REAL            :: vFMetProd,alphan,coeff,vNorm
     LOGICAL         :: l_pot, l_exist
 
@@ -128,6 +129,12 @@ CONTAINS
 
     ! save F_m and rho_m for next iteration
     nit = mit +1
+
+    ! Comment out the following code line to switch to continuous restart mode
+    ! Note: The continuous restart mode is not good at the moment. It produces undesired and
+    !       bad convergence behavior. But it is tested and seems to be correct.
+    IF (nit > input%maxiter+1) nit = 1
+
     CALL writeLastIterInAndDiffDen(hybrid,nmap,nit,input%alpha,sm,fm,smMet,fmMet)
 
     IF (mit.EQ.1) THEN 
@@ -158,18 +165,10 @@ CONTAINS
        dNdFLast = 0.0
        dFdNLast = 0.0
 
-!       WRITE(1400,*) '========================================'
-!       WRITE(1400,*) '========================================'
-!       WRITE(1400,*) '========================================'
-!       WRITE(1400,*) 'mit: ', mit
-!       WRITE(1400,*) 'iread, historyLength: ', iread, historyLength
        DO it = 2, iread
-          CALL readDeltaNVec(input,hybrid,nmap,it-iread-1,mit,deltaN_i)
-          CALL readDeltaFVec(input,hybrid,nmap,it-iread-1,mit,deltaF_i)
-
-!          WRITE(1400,'(4i7)') it,mit,it-iread-1,mit+(it-iread-1)
-!          WRITE(1400,'(a,5f15.8)') 'deltaN_i: ', deltaN_i(1), deltaN_i(2), deltaN_i(3), deltaN_i(4), deltaN_i(5)
-!          WRITE(1400,'(a,5f15.8)') 'deltaF_i: ', deltaF_i(1), deltaF_i(2), deltaF_i(3), deltaF_i(4), deltaF_i(5)
+          relIndex = it-iread-1
+          CALL readDeltaNVec(input,hybrid,nmap,relIndex,mit,deltaN_i)
+          CALL readDeltaFVec(input,hybrid,nmap,relIndex,mit,deltaF_i)
 
           dNdNLast(it-1) = CPP_BLAS_sdot(nmap,deltaN_i,1,dNMet,1)
           dFdFLast(it-1) = CPP_BLAS_sdot(nmap,deltaF_i,1,dFMet,1)
@@ -182,11 +181,6 @@ CONTAINS
        dNdFLast(historyLength) = CPP_BLAS_sdot(nmap,dNVec,1,dFMet,1)
        dFdNLast(historyLength) = CPP_BLAS_sdot(nmap,dFVec,1,dNMet,1)
 
-!       WRITE(1400,*) 'last overlaps:'
-!       DO i = 1, historyLength
-!          WRITE(1400,'(i7,4f20.13)') i,dNdNLast(i),dFdFLast(i),dNdFLast(i),dFdNLast(i)
-!       END DO
-
        CALL writeBroydenOverlapExt(input,hybrid,mit,historyLength,&
                                    dNdNLast,dFdFLast,dNdFLast,dFdNLast)
 
@@ -198,14 +192,6 @@ CONTAINS
        CALL readBroydenOverlaps(input,hybrid,mit,historyLength,&
                                 dNdNMat,dFdFMat,dNdFMat,dFdNMat)
 
-!       WRITE(1400,*) 'all overlaps'
-!       DO i = 1, historyLength
-!          DO j = 1, historyLength
-!             WRITE(1400,'(2i7,4f20.13)') i,j,dNdNMat(j,i),dFdFMat(j,i),dNdFMat(j,i),dFdNMat(j,i)
-!          END DO
-!       END DO
-!       WRITE(1400,*) '-----------------------------'
-
        ! Extend overlap matrices <delta n(i) | delta n(j)>, <delta F(i) | delta F(j)>,
        !                         <delta n(i) | delta F(j)>, <delta F(i) | delta n(j)> -end-
 
@@ -347,8 +333,9 @@ CONTAINS
        vVec = 0.0
 
        DO it = 2, iread
-          CALL readDeltaNVec(input,hybrid,nmap,it-iread-1,mit,deltaN_i)
-          CALL readDeltaFVec(input,hybrid,nmap,it-iread-1,mit,deltaF_i)
+          relIndex = it-iread-1
+          CALL readDeltaNVec(input,hybrid,nmap,relIndex,mit,deltaN_i)
+          CALL readDeltaFVec(input,hybrid,nmap,relIndex,mit,deltaF_i)
 
           DO k = 1, nmap
              uVec(k) = uVec(k) + uDNTableau(it-1,historyLength)*deltaN_i(k)

vgen/vmts.F90

@@ -78,8 +78,10 @@ CONTAINS
        ENDDO
     ENDIF
     !           ----> g.ne.0 components
-    !$OMP PARALLEL DO DEFAULT(SHARED) PRIVATE(cp,pylm,nat,n,sbf,nd,lh,&
-    !$OMP& sm,jm,m,lm,l) REDUCTION(+:vtl)
+    ! I commented out the OMP parallelization for the following loop since
+    ! it produced wrong results with the ifort 18 compiler.
+    !!$OMP PARALLEL DO DEFAULT(SHARED) PRIVATE(cp,pylm,nat,n,sbf,nd,lh,&
+    !!$OMP& sm,jm,m,lm,l) REDUCTION(+:vtl)
     DO k = mpi%irank+2, stars%ng3, mpi%isize
        cp = vpw(k,1)*stars%nstr(k)
        IF (.NOT.oneD%odi%d1) THEN
@@ -113,7 +115,7 @@ CALL od_phasy(&
           nat = nat + atoms%neq(n)
        ENDDO
     ENDDO
-    !$OMP END PARALLEL DO
+    !!$OMP END PARALLEL DO
 #ifdef CPP_MPI
     n1 = (sphhar%nlhd+1)*atoms%ntype
     ALLOCATE(c_b(n1))

vgen/vmtxcg.F90

@@ -152,7 +152,7 @@ CONTAINS
 
     DO 200 n = n_start,atoms%ntype,n_stride
        IF (xcpot%lda_atom(n))THEN
-          IF((.NOT.xcpot%is_name("pw91"))) CALL judft_error("Using locally LDA only possible with pw91 functional")
+          IF((.NOT.xcpot%is_name("pw91"))) CALL judft_warn("Using locally LDA only possible with pw91 functional")
           CALL xcpot_tmp%init("l91",.FALSE.)
        ENDIF
        nat=sum(atoms%neq(:n-1))+1