hsmt_sph.F90 18.2 KB
Newer Older
1 2 3 4 5 6
!--------------------------------------------------------------------------------
! Copyright (c) 2016 Peter Grünberg Institut, Forschungszentrum Jülich, Germany
! This file is part of FLEUR and available as free software under the conditions
! of the MIT license as expressed in the LICENSE file in more detail.
!--------------------------------------------------------------------------------

7
MODULE m_hsmt_sph
8 9
  USE m_juDFT
  IMPLICIT NONE
Uliana Alekseeva's avatar
Uliana Alekseeva committed
10 11 12 13 14 15 16 17 18 19 20 21

  INTERFACE hsmt_sph
    module procedure hsmt_sph_cpu
#ifdef CPP_GPU
    module procedure hsmt_sph_gpu
  END INTERFACE

  INTERFACE HsmtSphGpuKernel_noApw
    module procedure HsmtSphGpuKernel_noApw_cmplx, HsmtSphGpuKernel_noApw_real
#endif
  END INTERFACE

22
CONTAINS
Uliana Alekseeva's avatar
Uliana Alekseeva committed
23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328

#ifdef CPP_GPU
  SUBROUTINE HsmtSphGpuKernel_noApw_cmplx(iintsp,jintsp,nv,lmaxd,lmax,ki_start,ki_end,ki_step,nn_start,nn_end,&
                       lnonsph,chi,qssbti,qssbtj,gvec,gk,fleg1,fleg2,fl2p1,fl2p1bt,fj,gj,taual,ddn,el,e_shift,smat_data,hmat_data)
     INTEGER,INTENT(IN) :: iintsp,jintsp,nv(2),lmaxd,lmax,ki_start,ki_end,ki_step,nn_start,nn_end,lnonsph
     COMPLEX, INTENT(IN)  :: chi
     REAL,INTENT(IN)    :: qssbti(3),qssbtj(3)
     INTEGER,INTENT(IN) :: gvec(:,:,:) 
     REAL,INTENT(IN)    :: gk(:,:,:)
     REAL,INTENT(IN)    :: fleg1(0:lmaxd),fleg2(0:lmaxd),fl2p1(0:lmaxd)     
     REAL,INTENT(IN)    :: fl2p1bt(0:lmaxd)
     REAL,MANAGED,INTENT(IN)    :: fj(:,0:,:),gj(:,0:,:)
     REAL,INTENT(IN)    :: taual(:,:)  
     REAL,INTENT(IN)    :: ddn(0:lmaxd) 
     REAL,INTENT(IN)    :: el(0:lmaxd)
     REAL,INTENT(IN)    :: e_shift
     COMPLEX,INTENT(INOUT) :: smat_data(:,:),hmat_data(:,:)

     REAL,   PARAMETER :: tpi_const=2.*3.1415926535897932
     REAL, ALLOCATABLE :: plegend(:,:)
     COMPLEX, ALLOCATABLE :: cph(:)
     REAL tnn(3), elall,fct,fjkiln,gjkiln,ddnln,ski(3)
     INTEGER kii,ki,kj,l,nn

     ALLOCATE(cph(MAXVAL(nv)))
     ALLOCATE(plegend(MAXVAL(nv),0:lmaxd))
     plegend=0.0
     plegend(:,0)=1.0

     DO  ki =  ki_start,ki_end,ki_step 
       kii=(ki-1)/ki_step+1
       ski = gvec(:,ki,jintsp) + qssbti
       !--->       legendre polynomials
       DO kj = 1,ki
          plegend(kj,1) = DOT_PRODUCT(gk(:,kj,iintsp),gk(:,ki,jintsp))
       END DO
       DO l = 1,lmax - 1
          plegend(:ki,l+1) = fleg1(l)*plegend(:ki,1)*plegend(:ki,l) - fleg2(l)*plegend(:ki,l-1)
       END DO
       !--->             set up phase factors
       cph = 0.0
       DO nn = nn_start,nn_end
          tnn = tpi_const*taual(:,nn)
          DO kj = 1,ki
             cph(kj) = cph(kj) +&
                  CMPLX(COS(DOT_PRODUCT(ski-gvec(:,kj,iintsp)-qssbtj,tnn)),&
                  SIN(DOT_PRODUCT(gvec(:,kj,iintsp)+qssbtj-ski,tnn)))
            ! IF (iintsp.NE.jintsp) cph(kj)=CONJG(cph(kj))
          END DO
       END DO

       !--->          update overlap and l-diagonal hamiltonian matrix
       DO  l = 0,lmax
          fjkiln = fj(ki,l,jintsp)
          gjkiln = gj(ki,l,jintsp)
          ddnln =  ddn(l)
          elall = el(l)
          IF (l<=lnonsph) elall=elall-e_shift!(isp)
          DO kj = 1,MIN(ki,nv(iintsp))
              fct  = chi*plegend(kj,l)*fl2p1(l)*&
                    ( fjkiln*fj(kj,l,iintsp) + gjkiln*gj(kj,l,iintsp)*ddnln )

              smat_data(kj,kii)=smat_data(kj,kii) + cph(kj)*fct
              hmat_data(kj,kii)=hmat_data(kj,kii) + cph(kj) * ( fct*elall &
                     + chi*plegend(kj,l)*fl2p1bt(l) * ( fjkiln*gj(kj,l,iintsp) + gjkiln*fj(kj,l,iintsp) ) )
          END DO
       !--->          end loop over l
       ENDDO
     !--->    end loop over ki
     ENDDO
     DEALLOCATE(plegend)
     DEALLOCATE(cph)
  END SUBROUTINE HsmtSphGpuKernel_noApw_cmplx

  SUBROUTINE HsmtSphGpuKernel_noApw_real(iintsp,jintsp,nv,lmaxd,lmax,ki_start,ki_end,ki_step,nn_start,nn_end,&
                       lnonsph,chi,qssbti,qssbtj,gvec,gk,fleg1,fleg2,fl2p1,fl2p1bt,fj,gj,taual,ddn,el,e_shift,smat_data,hmat_data)
     INTEGER,INTENT(IN) :: iintsp,jintsp,nv(2),lmaxd,lmax,ki_start,ki_end,ki_step,nn_start,nn_end,lnonsph
     COMPLEX, INTENT(IN)  :: chi
     REAL,INTENT(IN)    :: qssbti(3),qssbtj(3)
     INTEGER,INTENT(IN) :: gvec(:,:,:) 
     REAL,INTENT(IN)    :: gk(:,:,:)
     REAL,INTENT(IN)    :: fleg1(0:lmaxd),fleg2(0:lmaxd),fl2p1(0:lmaxd)     
     REAL,INTENT(IN)    :: fl2p1bt(0:lmaxd)
     REAL,MANAGED,INTENT(IN)    :: fj(:,0:,:),gj(:,0:,:)
     REAL,INTENT(IN)    :: taual(:,:)  
     REAL,INTENT(IN)    :: ddn(0:lmaxd) 
     REAL,INTENT(IN)    :: el(0:lmaxd)
     REAL,INTENT(IN)    :: e_shift
     REAL,INTENT(INOUT) :: smat_data(:,:),hmat_data(:,:)

     REAL,   PARAMETER :: tpi_const=2.*3.1415926535897932
     REAL, ALLOCATABLE :: plegend(:,:)
     COMPLEX, ALLOCATABLE :: cph(:)
     REAL tnn(3), elall,fct,fjkiln,gjkiln,ddnln,ski(3)
     INTEGER kii,ki,kj,l,nn

     ALLOCATE(cph(MAXVAL(nv)))
     ALLOCATE(plegend(MAXVAL(nv),0:lmaxd))
     plegend=0.0
     plegend(:,0)=1.0

     DO  ki =  ki_start,ki_end,ki_step 
       kii=(ki-1)/ki_step+1
       ski = gvec(:,ki,jintsp) + qssbti
       !--->       legendre polynomials
       DO kj = 1,ki
          plegend(kj,1) = DOT_PRODUCT(gk(:,kj,iintsp),gk(:,ki,jintsp))
       END DO
       DO l = 1,lmax - 1
          plegend(:ki,l+1) = fleg1(l)*plegend(:ki,1)*plegend(:ki,l) - fleg2(l)*plegend(:ki,l-1)
       END DO
       !--->             set up phase factors
       cph = 0.0
       DO nn = nn_start,nn_end
          tnn = tpi_const*taual(:,nn)
          DO kj = 1,ki
             cph(kj) = cph(kj) +&
                  CMPLX(COS(DOT_PRODUCT(ski-gvec(:,kj,iintsp)-qssbtj,tnn)),&
                  SIN(DOT_PRODUCT(gvec(:,kj,iintsp)+qssbtj-ski,tnn)))
            ! IF (iintsp.NE.jintsp) cph(kj)=CONJG(cph(kj))
          END DO
       END DO

       !--->          update overlap and l-diagonal hamiltonian matrix
       DO  l = 0,lmax
          fjkiln = fj(ki,l,jintsp)
          gjkiln = gj(ki,l,jintsp)
          ddnln =  ddn(l)
          elall = el(l)
          IF (l<=lnonsph) elall=elall-e_shift!(isp)
          DO kj = 1,ki
              fct  = plegend(kj,l)*fl2p1(l)*&
                     ( fjkiln*fj(kj,l,iintsp) + gjkiln*gj(kj,l,iintsp)*ddnln )
              smat_data(kj,kii)=smat_data(kj,kii)+REAL(cph(kj))*fct
              hmat_data(kj,kii)=hmat_data(kj,kii) + REAL(cph(kj)) * &
                     ( fct * elall + plegend(kj,l) * fl2p1bt(l) *&
                     ( fjkiln*gj(kj,l,iintsp) + gjkiln*fj(kj,l,iintsp) ) )
          ENDDO
       !--->          end loop over l
       ENDDO
     !--->    end loop over ki
     ENDDO
     DEALLOCATE(plegend)
     DEALLOCATE(cph)
  END SUBROUTINE HsmtSphGpuKernel_noApw_real

  SUBROUTINE hsmt_sph_gpu(n,atoms,mpi,isp,input,noco,iintsp,jintsp,chi,lapw,el,e_shift,usdus,fj,gj,smat,hmat)
    USE m_constants, ONLY : fpi_const,tpi_const
    USE m_types
    USE nvtx
    IMPLICIT NONE
    TYPE(t_input),INTENT(IN)      :: input
    TYPE(t_mpi),INTENT(IN)        :: mpi
    TYPE(t_noco),INTENT(IN)       :: noco
    TYPE(t_atoms),INTENT(IN)      :: atoms
    TYPE(t_lapw),INTENT(IN)       :: lapw
    TYPE(t_usdus),INTENT(IN)      :: usdus
    CLASS(t_mat),INTENT(INOUT)     :: smat,hmat
    !     ..
    !     .. Scalar Arguments ..
    INTEGER, INTENT (IN) :: n,isp,iintsp,jintsp
    COMPLEX, INTENT(IN)  :: chi
    !     ..
    !     .. Array Arguments ..
    REAL,    INTENT (IN) :: el(0:atoms%lmaxd,atoms%ntype,input%jspins)
    REAL,    INTENT (IN) :: e_shift!(atoms%ntype,input%jspins)
#if defined CPP_GPU
    REAL,MANAGED,INTENT(IN)    :: fj(:,0:,:),gj(:,0:,:)
#else
    REAL,    INTENT (IN) :: fj(:,0:,:),gj(:,0:,:)
#endif
    !     ..
    !     .. Local Scalars ..
    REAL tnn(3), elall,fct,fjkiln,gjkiln,ddnln,ski(3)
    REAL apw_lo1,apw_lo2,apw1,w1

    COMPLEX capw1
    INTEGER kii,ki,kj,l,nn

    !     ..
    !     .. Local Arrays ..
    REAL fleg1(0:atoms%lmaxd),fleg2(0:atoms%lmaxd),fl2p1(0:atoms%lmaxd)     
    REAL fl2p1bt(0:atoms%lmaxd)
    REAL qssbti(3),qssbtj(3)
    REAL, ALLOCATABLE :: plegend(:,:)
    COMPLEX, ALLOCATABLE :: cph(:)
    LOGICAL apw(0:atoms%lmaxd)

    call nvtxStartRange("hsmt_sph",2)    
    CALL timestart("spherical setup")

    DO l = 0,atoms%lmaxd
       fleg1(l) = REAL(l+l+1)/REAL(l+1)
       fleg2(l) = REAL(l)/REAL(l+1)
       fl2p1(l) = REAL(l+l+1)/fpi_const
       fl2p1bt(l) = fl2p1(l)*0.5
    END DO
    !!$OMP PARALLEL DEFAULT(SHARED)&
    !!$OMP PRIVATE(kii,ki,ski,kj,plegend,l)&
    !!$OMP PRIVATE(cph,nn,tnn,fjkiln,gjkiln)&
    !!$OMP PRIVATE(w1,apw_lo1,apw_lo2,ddnln,elall,fct,apw1)&
    !!$OMP PRIVATE(capw1) 
    qssbti=MERGE(- noco%qss/2,+ noco%qss/2,jintsp.EQ.1)
    qssbtj=MERGE(- noco%qss/2,+ noco%qss/2,iintsp.EQ.1)

    !!$OMP  DO SCHEDULE(DYNAMIC,1)
    IF (input%l_useapw) THEN
     ALLOCATE(cph(MAXVAL(lapw%nv)))
     ALLOCATE(plegend(MAXVAL(lapw%nv),0:atoms%lmaxd))
     plegend=0.0
     plegend(:,0)=1.0
     DO  ki =  mpi%n_rank+1, lapw%nv(jintsp), mpi%n_size
       kii=(ki-1)/mpi%n_size+1
       ski = lapw%gvec(:,ki,jintsp) + qssbti
       !--->       legendre polynomials
       DO kj = 1,ki
          plegend(kj,1) = DOT_PRODUCT(lapw%gk(:,kj,iintsp),lapw%gk(:,ki,jintsp))
       END DO
       DO l = 1,atoms%lmax(n) - 1
          plegend(:ki,l+1) = fleg1(l)*plegend(:ki,1)*plegend(:ki,l) - fleg2(l)*plegend(:ki,l-1)
       END DO
       !--->             set up phase factors
       cph = 0.0
       DO nn = SUM(atoms%neq(:n-1))+1,SUM(atoms%neq(:n))
          tnn = tpi_const*atoms%taual(:,nn)
          DO kj = 1,ki
             cph(kj) = cph(kj) +&
                  CMPLX(COS(DOT_PRODUCT(ski-lapw%gvec(:,kj,iintsp)-qssbtj,tnn)),&
                  SIN(DOT_PRODUCT(lapw%gvec(:,kj,iintsp)+qssbtj-ski,tnn)))
            ! IF (iintsp.NE.jintsp) cph(kj)=CONJG(cph(kj))
          END DO
       END DO

       !--->          update overlap and l-diagonal hamiltonian matrix
       DO  l = 0,atoms%lmax(n)
          fjkiln = fj(ki,l,jintsp)
          gjkiln = gj(ki,l,jintsp)
          !+APW
          w1 = 0.5 * ( usdus%uds(l,n,isp)*usdus%dus(l,n,isp) + &
                  usdus%us(l,n,isp)*usdus%duds(l,n,isp) )
          apw_lo1 = fl2p1(l) * 0.5 * atoms%rmt(n)**2 * ( gjkiln * w1 +&
                  fjkiln * usdus%us(l,n,isp) * usdus%dus(l,n,isp) )
          apw_lo2 = fl2p1(l) * 0.5 * atoms%rmt(n)**2 * ( fjkiln * w1 +&
                  gjkiln * usdus%uds(l,n,isp) * usdus%duds(l,n,isp) )
          !-APW
          ddnln =  usdus%ddn(l,n,isp)
          elall = el(l,n,isp)
          IF (l<=atoms%lnonsph(n)) elall=elall-e_shift!(isp)
          IF (smat%l_real) THEN
             DO kj = 1,ki
                fct  = plegend(kj,l)*fl2p1(l)*&
                     ( fjkiln*fj(kj,l,iintsp) + gjkiln*gj(kj,l,iintsp)*ddnln )
                smat%data_r(kj,kii)=smat%data_r(kj,kii)+REAL(cph(kj))*fct
                hmat%data_r(kj,kii)=hmat%data_r(kj,kii) + REAL(cph(kj)) * &
                     ( fct * elall + plegend(kj,l) * fl2p1bt(l) *&
                     ( fjkiln*gj(kj,l,iintsp) + gjkiln*fj(kj,l,iintsp) ) )
                !+APW
                apw1 = REAL(cph(kj)) * plegend(kj,l)  * &
                        ( apw_lo1 * fj(kj,l,iintsp) + apw_lo2 * gj(kj,l,iintsp) )
                hmat%data_r(kj,kii)=hmat%data_r(kj,kii) + apw1
                !-APW
             ENDDO
          ELSE
             DO kj = 1,MIN(ki,lapw%nv(iintsp))
                fct  = chi*plegend(kj,l)*fl2p1(l)*&
                     ( fjkiln*fj(kj,l,iintsp) + gjkiln*gj(kj,l,iintsp)*ddnln )

                smat%data_c(kj,kii)=smat%data_c(kj,kii) + cph(kj)*fct
                hmat%data_c(kj,kii)=hmat%data_c(kj,kii) + cph(kj) * ( fct*elall &
                     + chi*plegend(kj,l)*fl2p1bt(l) * ( fjkiln*gj(kj,l,iintsp) + gjkiln*fj(kj,l,iintsp) ) )
                !+APW
                   capw1 = cph(kj)*plegend(kj,l)&
                        * ( apw_lo1 * fj(kj,l,iintsp) + apw_lo2 * gj(kj,l,iintsp) )
                   hmat%data_c(kj,kii)=hmat%data_c(kj,kii) + capw1
                !-APW
             END DO

          ENDIF
          !--->          end loop over l
       ENDDO
       !--->    end loop over ki
     ENDDO
     !--->       end loop over atom types (ntype)
     DEALLOCATE(plegend)
     DEALLOCATE(cph)
    ELSE
       IF (smat%l_real) THEN
          CALL HsmtSphGpuKernel_noApw(iintsp,jintsp,lapw%nv,atoms%lmaxd,atoms%lmax(n),mpi%n_rank+1, lapw%nv(jintsp), mpi%n_size,&
                    SUM(atoms%neq(:n-1))+1,SUM(atoms%neq(:n)),atoms%lnonsph(n),chi,qssbti,qssbtj,lapw%gvec,&
                   lapw%gk,fleg1,fleg2,fl2p1,fl2p1bt,fj,gj,atoms%taual,usdus%ddn(:,n,isp),el(:,n,isp),e_shift,smat%data_r,hmat%data_r)
       ELSE
          CALL HsmtSphGpuKernel_noApw(iintsp,jintsp,lapw%nv,atoms%lmaxd,atoms%lmax(n),mpi%n_rank+1, lapw%nv(jintsp), mpi%n_size,&
                    SUM(atoms%neq(:n-1))+1,SUM(atoms%neq(:n)),atoms%lnonsph(n),chi,qssbti,qssbtj,lapw%gvec,&
                   lapw%gk,fleg1,fleg2,fl2p1,fl2p1bt,fj,gj,atoms%taual,usdus%ddn(:,n,isp),el(:,n,isp),e_shift,smat%data_c,hmat%data_c)
       ENDIF
    ENDIF
    !!$OMP END DO
    !!$OMP END PARALLEL
    CALL timestop("spherical setup")

    call nvtxEndRange
    RETURN
  END SUBROUTINE hsmt_sph_gpu
#endif

  SUBROUTINE hsmt_sph_cpu(n,atoms,mpi,isp,input,noco,iintsp,jintsp,chi,lapw,el,e_shift,usdus,fj,gj,smat,hmat)
329 330 331
    USE m_constants, ONLY : fpi_const,tpi_const
    USE m_types
    IMPLICIT NONE
332
    TYPE(t_input),INTENT(IN)      :: input
Daniel Wortmann's avatar
Daniel Wortmann committed
333
    TYPE(t_mpi),INTENT(IN)        :: mpi
334 335
    TYPE(t_noco),INTENT(IN)       :: noco
    TYPE(t_atoms),INTENT(IN)      :: atoms
Daniel Wortmann's avatar
Daniel Wortmann committed
336 337 338
    TYPE(t_lapw),INTENT(IN)       :: lapw
    TYPE(t_usdus),INTENT(IN)      :: usdus
    CLASS(t_mat),INTENT(INOUT)     :: smat,hmat
339 340
    !     ..
    !     .. Scalar Arguments ..
Daniel Wortmann's avatar
Daniel Wortmann committed
341 342
    INTEGER, INTENT (IN) :: n,isp,iintsp,jintsp
    COMPLEX, INTENT(IN)  :: chi
343 344
    !     ..
    !     .. Array Arguments ..
Daniel Wortmann's avatar
Daniel Wortmann committed
345
    REAL,    INTENT (IN) :: el(0:atoms%lmaxd,atoms%ntype,input%jspins)
346
    REAL,    INTENT (IN) :: e_shift!(atoms%ntype,input%jspins)
Daniel Wortmann's avatar
Daniel Wortmann committed
347
    REAL,    INTENT (IN) :: fj(:,0:,:),gj(:,0:,:)
348 349
    !     ..
    !     .. Local Scalars ..
Daniel Wortmann's avatar
Daniel Wortmann committed
350 351
    REAL tnn(3), elall,fct,fjkiln,gjkiln,ddnln,ski(3)
    REAL apw_lo1,apw_lo2,apw1,w1
352

Daniel Wortmann's avatar
Daniel Wortmann committed
353 354
    COMPLEX capw1
    INTEGER kii,ki,kj,l,nn
355 356 357

    !     ..
    !     .. Local Arrays ..
Daniel Wortmann's avatar
Daniel Wortmann committed
358 359
    REAL fleg1(0:atoms%lmaxd),fleg2(0:atoms%lmaxd),fl2p1(0:atoms%lmaxd)     
    REAL fl2p1bt(0:atoms%lmaxd)
360 361
    REAL qssbti(3),qssbtj(3)
    REAL, ALLOCATABLE :: plegend(:,:)
Daniel Wortmann's avatar
Daniel Wortmann committed
362
    COMPLEX, ALLOCATABLE :: cph(:)
363 364
    LOGICAL apw(0:atoms%lmaxd)

Daniel Wortmann's avatar
Daniel Wortmann committed
365
    CALL timestart("spherical setup")
366 367

    DO l = 0,atoms%lmaxd
368 369 370
       fleg1(l) = REAL(l+l+1)/REAL(l+1)
       fleg2(l) = REAL(l)/REAL(l+1)
       fl2p1(l) = REAL(l+l+1)/fpi_const
371 372
       fl2p1bt(l) = fl2p1(l)*0.5
    END DO
Daniel Wortmann's avatar
Daniel Wortmann committed
373 374 375 376 377 378 379 380 381
    !$OMP PARALLEL DEFAULT(SHARED)&
    !$OMP PRIVATE(kii,ki,ski,kj,plegend,l)&
    !$OMP PRIVATE(cph,nn,tnn,fjkiln,gjkiln)&
    !$OMP PRIVATE(w1,apw_lo1,apw_lo2,ddnln,elall,fct,apw1)&
    !$OMP PRIVATE(capw1) 
    ALLOCATE(cph(MAXVAL(lapw%nv)))
    ALLOCATE(plegend(MAXVAL(lapw%nv),0:atoms%lmaxd))
    plegend=0.0
    plegend(:,0)=1.0
382 383
    qssbti=MERGE(- noco%qss/2,+ noco%qss/2,jintsp.EQ.1)
    qssbtj=MERGE(- noco%qss/2,+ noco%qss/2,iintsp.EQ.1)
Daniel Wortmann's avatar
Daniel Wortmann committed
384
    !$OMP  DO SCHEDULE(DYNAMIC,1)
385
    DO  ki =  mpi%n_rank+1, lapw%nv(jintsp), mpi%n_size
Daniel Wortmann's avatar
Daniel Wortmann committed
386
       kii=(ki-1)/mpi%n_size+1
387
       ski = lapw%gvec(:,ki,jintsp) + qssbti
Daniel Wortmann's avatar
Daniel Wortmann committed
388 389
       !--->       legendre polynomials
       DO kj = 1,ki
390
          plegend(kj,1) = DOT_PRODUCT(lapw%gk(:,kj,iintsp),lapw%gk(:,ki,jintsp))
Daniel Wortmann's avatar
Daniel Wortmann committed
391 392 393 394 395 396 397 398 399 400
       END DO
       DO l = 1,atoms%lmax(n) - 1
          plegend(:ki,l+1) = fleg1(l)*plegend(:ki,1)*plegend(:ki,l) - fleg2(l)*plegend(:ki,l-1)
       END DO
       !--->             set up phase factors
       cph = 0.0
       DO nn = SUM(atoms%neq(:n-1))+1,SUM(atoms%neq(:n))
          tnn = tpi_const*atoms%taual(:,nn)
          DO kj = 1,ki
             cph(kj) = cph(kj) +&
401 402 403
                  CMPLX(COS(DOT_PRODUCT(ski-lapw%gvec(:,kj,iintsp)-qssbtj,tnn)),&
                  SIN(DOT_PRODUCT(lapw%gvec(:,kj,iintsp)+qssbtj-ski,tnn)))
            ! IF (iintsp.NE.jintsp) cph(kj)=CONJG(cph(kj))
Daniel Wortmann's avatar
Daniel Wortmann committed
404 405 406 407 408
          END DO
       END DO

       !--->          update overlap and l-diagonal hamiltonian matrix
       DO  l = 0,atoms%lmax(n)
409 410
          fjkiln = fj(ki,l,jintsp)
          gjkiln = gj(ki,l,jintsp)
411
          !
Daniel Wortmann's avatar
Daniel Wortmann committed
412 413 414 415 416 417 418 419 420 421 422
          IF (input%l_useapw) THEN
             w1 = 0.5 * ( usdus%uds(l,n,isp)*usdus%dus(l,n,isp) + &
                  usdus%us(l,n,isp)*usdus%duds(l,n,isp) )
             apw_lo1 = fl2p1(l) * 0.5 * atoms%rmt(n)**2 * ( gjkiln * w1 +&
                  fjkiln * usdus%us(l,n,isp) * usdus%dus(l,n,isp) )
             apw_lo2 = fl2p1(l) * 0.5 * atoms%rmt(n)**2 * ( fjkiln * w1 +&
                  gjkiln * usdus%uds(l,n,isp) * usdus%duds(l,n,isp) )
             !
          ENDIF
          ddnln =  usdus%ddn(l,n,isp)
          elall = el(l,n,isp)
423
          IF (l<=atoms%lnonsph(n)) elall=elall-e_shift!(isp)
Daniel Wortmann's avatar
Daniel Wortmann committed
424 425 426
          IF (smat%l_real) THEN
             DO kj = 1,ki
                fct  = plegend(kj,l)*fl2p1(l)*&
427
                     ( fjkiln*fj(kj,l,iintsp) + gjkiln*gj(kj,l,iintsp)*ddnln )
Daniel Wortmann's avatar
Daniel Wortmann committed
428 429 430
                smat%data_r(kj,kii)=smat%data_r(kj,kii)+REAL(cph(kj))*fct
                hmat%data_r(kj,kii)=hmat%data_r(kj,kii) + REAL(cph(kj)) * &
                     ( fct * elall + plegend(kj,l) * fl2p1bt(l) *&
431
                     ( fjkiln*gj(kj,l,iintsp) + gjkiln*fj(kj,l,iintsp) ) )
Daniel Wortmann's avatar
Daniel Wortmann committed
432 433 434 435 436
                !+APW
                IF (input%l_useapw) THEN
                   apw1 = REAL(cph(kj)) * plegend(kj,l)  * &
                        ( apw_lo1 * fj(kj,l,iintsp) + apw_lo2 * gj(kj,l,iintsp) )
                   hmat%data_r(kj,kii)=hmat%data_r(kj,kii) + apw1
437
                ENDIF
Daniel Wortmann's avatar
Daniel Wortmann committed
438 439 440
                !-APW
             ENDDO
          ELSE
441
             DO kj = 1,MIN(ki,lapw%nv(iintsp))
Daniel Wortmann's avatar
Daniel Wortmann committed
442
                fct  = chi*plegend(kj,l)*fl2p1(l)*&
443
                     ( fjkiln*fj(kj,l,iintsp) + gjkiln*gj(kj,l,iintsp)*ddnln )
Daniel Wortmann's avatar
Daniel Wortmann committed
444 445 446

                smat%data_c(kj,kii)=smat%data_c(kj,kii) + cph(kj)*fct
                hmat%data_c(kj,kii)=hmat%data_c(kj,kii) + cph(kj) * ( fct*elall &
447
                     + chi*plegend(kj,l)*fl2p1bt(l) * ( fjkiln*gj(kj,l,iintsp) + gjkiln*fj(kj,l,iintsp) ) )
Daniel Wortmann's avatar
Daniel Wortmann committed
448 449 450 451 452

                IF (input%l_useapw) THEN
                   capw1 = cph(kj)*plegend(kj,l)&
                        * ( apw_lo1 * fj(kj,l,iintsp) + apw_lo2 * gj(kj,l,iintsp) )
                   hmat%data_c(kj,kii)=hmat%data_c(kj,kii) + capw1
453 454 455
                ENDIF
             END DO

Daniel Wortmann's avatar
Daniel Wortmann committed
456
          ENDIF
Uliana Alekseeva's avatar
Uliana Alekseeva committed
457
       !--->          end loop over l
Daniel Wortmann's avatar
Daniel Wortmann committed
458
       ENDDO
Uliana Alekseeva's avatar
Uliana Alekseeva committed
459
    !--->    end loop over ki
Daniel Wortmann's avatar
Daniel Wortmann committed
460 461
    ENDDO
    !$OMP END DO
462
    DEALLOCATE(plegend)
Daniel Wortmann's avatar
Daniel Wortmann committed
463
    DEALLOCATE(cph)
464
    !$OMP END PARALLEL
Daniel Wortmann's avatar
Daniel Wortmann committed
465
    CALL timestop("spherical setup")
466 467

    RETURN
Uliana Alekseeva's avatar
Uliana Alekseeva committed
468
  END SUBROUTINE hsmt_sph_cpu
469
END MODULE m_hsmt_sph