eig66_mpi.F90 27.6 KB
Newer Older
1 2
MODULE m_eig66_mpi
#include "juDFT_env.h"
3
  USE m_eig66_data
4
#ifdef CPP_MPI
5
  use mpi
6
#endif
7 8
  IMPLICIT NONE
  PRIVATE
9
  PUBLIC open_eig,read_eig,write_eig,close_eig,write_dos,read_dos
10 11
CONTAINS

12 13 14
  SUBROUTINE priv_find_data(id,d)
    INTEGER,INTENT(IN)::id
    TYPE(t_data_mpi),POINTER:: d
15

16 17 18 19 20 21 22 23 24
    CLASS(t_data),POINTER   ::dp
    CALL eig66_find_data(dp,id)
    SELECT TYPE(dp)
    TYPE is (t_data_mpi)
       d=>dp
       CLASS default
       CALL judft_error("BUG: wrong datatype in eig66_mpi")
    END SELECT
  END SUBROUTINE priv_find_data
25 26


27
  SUBROUTINE open_eig(id,mpi_comm,nmat,neig,nkpts,jspins,lmax,nlo,ntype,create,l_real,l_soc,nlotot,l_noco,n_size_opt,l_dos,l_mcd,l_orb,filename,layers,nstars,ncored,nsld,nat)
28 29 30
    USE,INTRINSIC::iso_c_binding
    IMPLICIT NONE
    INTEGER, INTENT(IN) :: id,mpi_comm,nmat,neig,nkpts,jspins,nlo,ntype,lmax,nlotot
31
    LOGICAL, INTENT(IN) :: l_noco,create,l_real,l_soc
32
    INTEGER,INTENT(IN),OPTIONAL:: n_size_opt
33
    LOGICAL,INTENT(IN),OPTIONAL ::l_dos,l_mcd,l_orb
34 35 36 37
    CHARACTER(LEN=*),INTENT(IN),OPTIONAL :: filename
    INTEGER,INTENT(IN),OPTIONAL :: layers,nstars,ncored,nsld,nat
#ifdef CPP_MPI
    INTEGER:: isize,e,slot_size,local_slots
38
    INTEGER,PARAMETER::mcored=27 !there should not be more that 27 core states
39 40 41
    TYPE(t_data_MPI),POINTER :: d

    CALL priv_find_data(id,d)
42
    CALL eig66_data_storedefault(d,jspins,nkpts,nmat,neig,lmax,nlotot,nlo,ntype,l_real,l_soc,l_dos,l_mcd,l_orb)
43 44 45 46 47 48 49 50

    IF (PRESENT(n_size_opt)) d%n_size=n_size_opt
    IF (ALLOCATED(d%pe_ev)) THEN
       IF (create) THEN
          d%neig_data=0
          d%eig_data=1E99
          d%int_data=9999999
          d%real_data=1E99
51 52 53 54 55
          if (l_real.and..not.l_soc) THEN
             d%zr_data=0.0
          else
             d%zc_data=0.0
          endif
56 57 58 59 60 61 62 63 64 65 66 67 68 69 70
          d%qal_data=0.0
          d%qvac_data=0.0
          d%qvlay_data=0.0
          d%qstars_data=0.0
          d%ksym_data=0.0
          d%jsym_data=0.0
          d%mcd_data=0.0
          d%qintsl_data=0.0
          d%qmtsl_data=0.0
          d%qmtp_data=0.0
          d%orbcomp_data=0.0
       ENDIF
       IF (PRESENT(filename)) CALL priv_readfromfile()
       RETURN !everything already done!
    ENDIF
71

72 73 74 75 76 77 78 79 80 81
    CALL timestart("create data spaces in ei66_mpi")
    CALL MPI_COMM_RANK(MPI_COMM,d%irank,e)
    CALL MPI_COMM_SIZE(MPI_COMM,isize,e)

    CALL create_maps(d,isize,nkpts,jspins,neig,d%n_size)
    local_slots=COUNT(d%pe_basis==d%irank)
    !Now create the windows

    !Window for neig
    slot_size=1
82
    CALL priv_create_memory(1,local_slots,d%neig_handle,d%neig_data)
83 84 85 86 87
    d%neig_data=0

    !The integer values
    d%size_k=nmat
    slot_size=(5+3*d%size_k+1+nlotot)
88
    CALL priv_create_memory(slot_size,local_slots,d%int_handle,d%int_data)
89 90 91 92 93 94
    d%int_data=9999999

    !The real values
    d%size_el=(1+lmax)*ntype
    d%size_ello=nlo*ntype
    slot_size=(6+d%size_el+d%size_ello)
95
    CALL priv_create_memory(slot_size,local_slots,d%real_handle,real_data_ptr=d%real_data)
96 97 98 99
    d%real_data=1E99

    !The eigenvalues
    d%size_eig=neig
100
    CALL priv_create_memory(d%size_eig,local_slots,d%eig_handle,real_data_ptr=d%eig_data)
101 102 103 104 105
    d%eig_data=1E99
    !The eigenvectors
    local_slots=COUNT(d%pe_ev==d%irank)
    slot_size=nmat

106 107 108 109 110
    if (l_real.and..not.l_soc) THEN
       CALL priv_create_memory(slot_size,local_slots,d%zr_handle,real_data_ptr=d%zr_data)
    else
       CALL priv_create_memory(slot_size,local_slots,d%zc_handle,cmplx_data_ptr=d%zc_data)
    endif
111
    !Data for DOS etc
112
    IF (d%l_dos) THEN
113
       local_slots=COUNT(d%pe_basis==d%irank)
114 115 116
       CALL priv_create_memory(4*ntype*neig,local_slots,d%qal_handle,real_data_ptr=d%qal_data)
       CALL priv_create_memory(neig*2,local_slots,d%qvac_handle,real_data_ptr=d%qvac_data)
       CALL priv_create_memory(neig,local_slots,d%qis_handle,real_data_ptr=d%qis_data)
117 118
       CALL priv_create_memory(neig*max(1,layers)*2,local_slots,d%qvlay_handle,real_data_ptr=d%qvlay_data)
       CALL priv_create_memory(max(1,nstars)*neig*max(1,layers)*2,local_slots,d%qstars_handle,cmplx_data_ptr=d%qstars_data)
119 120 121
       CALL priv_create_memory(neig,local_slots,d%jsym_handle,d%jsym_data)
       CALL priv_create_memory(neig,local_slots,d%ksym_handle,d%ksym_data)
       IF (l_mcd) CALL priv_create_memory(3*ntype*mcored*neig,local_slots,d%mcd_handle,real_data_ptr=d%mcd_data)
122
       IF (l_orb) THEN
123 124 125 126
          CALL priv_create_memory(nsld*neig,local_slots,d%qintsl_handle,real_data_ptr=d%qintsl_data)
          CALL priv_create_memory(nsld*neig,local_slots,d%qmtsl_handle,real_data_ptr=d%qmtsl_data)
          CALL priv_create_memory(nat*neig,local_slots,d%qmtp_handle,real_data_ptr=d%qmtp_data)
          CALL priv_create_memory(23*nat*neig,local_slots,d%orbcomp_handle,real_data_ptr=d%orbcomp_data)
127 128 129 130 131 132 133 134 135
       ENDIF
    ELSE
       ALLOCATE(d%qal_data(1),d%qvac_data(1),d%qis_data(1),d%qvlay_data(1),d%qstars_data(1),&
            d%jsym_data(1),d%ksym_data(1),d%mcd_data(1),d%qintsl_data(1),d%qmtsl_data(1),&
            d%qmtp_data(1),d%orbcomp_data(1))
    ENDIF
    IF (PRESENT(filename).AND..NOT.create) CALL priv_readfromfile()
    CALL timestop("create data spaces in ei66_mpi")
  CONTAINS
136
    SUBROUTINE priv_create_memory(slot_size,local_slots,handle,int_data_ptr,real_data_ptr,cmplx_data_ptr)
137 138
      IMPLICIT NONE
      INTEGER,INTENT(IN)           :: slot_size,local_slots
139 140 141
      INTEGER,POINTER,INTENT(OUT),OPTIONAL  :: int_data_ptr(:)
      REAL   ,POINTER,INTENT(OUT),OPTIONAL  :: real_data_ptr(:)
      COMPLEX,POINTER,INTENT(OUT),OPTIONAL  :: cmplx_data_ptr(:)
142
      INTEGER,INTENT(OUT)          :: handle
143
#ifdef CPP_MPI
144 145 146
      TYPE(c_ptr)::ptr
      INTEGER:: e
      INTEGER(MPI_ADDRESS_KIND) :: length
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162
      INTEGER                   :: type_size

      length=0   
      IF (present(real_data_ptr)) THEN
          length=length+1
          CALL MPI_TYPE_SIZE(MPI_DOUBLE_PRECISION,type_size,e)
      ENDIF
      IF (present(cmplx_data_ptr)) THEN
          length=length+1
          CALL MPI_TYPE_SIZE(MPI_DOUBLE_COMPLEX,type_size,e)
      ENDIF
      IF (present(int_data_ptr)) THEN 
          length=length+1
          CALL MPI_TYPE_SIZE(MPI_INTEGER,type_size,e)
      ENDIF
      if (length.ne.1) call judft_error("Bug in eig66_mpi:create_memory") 
163
      length=slot_size*local_slots
164
 
165 166 167 168
      length=length*type_size

      CALL MPI_ALLOC_MEM(length,MPI_INFO_NULL,ptr,e)
      IF (e.NE.0) CPP_error("Could not allocated MPI-Data in eig66_mpi")
169 170 171
	
      IF (present(real_data_ptr)) THEN	
      	CALL C_F_POINTER(ptr,real_data_ptr,(/length/type_size/))
172
      	CALL MPI_WIN_CREATE(real_data_ptr, length,slot_size*type_size,Mpi_INFO_NULL, MPI_COMM,handle, e)
173 174
      ELSEIF(present(int_data_ptr)) THEN
      	CALL C_F_POINTER(ptr,int_data_ptr,(/length/type_size/))
175
      	CALL MPI_WIN_CREATE(int_data_ptr, length,slot_size*type_size,Mpi_INFO_NULL, MPI_COMM,handle, e)
176 177
      ELSE
      	CALL C_F_POINTER(ptr,cmplx_data_ptr,(/length/type_size/))
178
      	CALL MPI_WIN_CREATE(cmplx_data_ptr, length,slot_size*type_size,Mpi_INFO_NULL, MPI_COMM,handle, e)
179
      ENDIF
180
#endif
181
    END SUBROUTINE priv_create_memory
182 183


184 185 186 187 188 189
    SUBROUTINE priv_readfromfile()
      USE m_eig66_DA,ONLY:open_eig_DA=>open_eig,read_eig_DA=>read_eig,close_eig_da=>close_eig
      INTEGER:: jspin,nk,i,ii,iii,nv,tmp_id
      REAL   :: wk,bk3(3),evac(2)
      INTEGER :: k1(nmat),k2(nmat),k3(nmat),kveclo(nlotot)
      REAL    :: eig(neig),ello(nlo,ntype),el(lmax,ntype)
190 191
      REAL    :: z_r(nmat,neig)
      COMPLEX :: z_c(nmat,neig)
192 193 194
      !only do this with PE=0
      IF (d%irank==0) THEN
         tmp_id=eig66_data_newid(DA_mode)
195
         IF (d%l_dos) CPP_error("Could not read DOS data")
196
         CALL open_eig_DA(tmp_id,nmat,neig,nkpts,jspins,lmax,nlo,ntype,nlotot,.FALSE.,.FALSE.,l_real,l_soc,.FALSE.,.FALSE.,filename)
197 198
         DO jspin=1,jspins
            DO nk=1,nkpts
199 200 201 202 203 204 205
               if (l_real) THEN
                  CALL read_eig_DA(tmp_id,nk,jspin,nv,i,k1,k2,k3,bk3,wk,ii,eig,el,ello,evac,kveclo,z=z_r)
                  CALL write_eig(id,nk,jspin,ii,ii,nv,nmat,k1,k2,k3,bk3,wk,eig,el,ello,evac,nlotot,kveclo,z=z_r)
               ELSE
                  CALL read_eig_DA(tmp_id,nk,jspin,nv,i,k1,k2,k3,bk3,wk,ii,eig,el,ello,evac,kveclo,z=z_c)
                  CALL write_eig(id,nk,jspin,ii,ii,nv,nmat,k1,k2,k3,bk3,wk,eig,el,ello,evac,nlotot,kveclo,z=z_c)
               end if
206
            ENDDO
207 208 209 210
         ENDDO
         CALL close_eig_DA(tmp_id)
      ENDIF
    END SUBROUTINE priv_readfromfile
211
#endif
212 213
  END SUBROUTINE open_eig
  SUBROUTINE close_eig(id,delete,filename)
214 215 216
    INTEGER,INTENT(IN)         :: id
    LOGICAL,INTENT(IN),OPTIONAL:: delete
    CHARACTER(LEN=*),INTENT(IN),OPTIONAL::filename
217 218
    TYPE(t_data_MPI),POINTER :: d
    CALL priv_find_data(id,d)
219

220 221
    IF (PRESENT(delete)) THEN
       IF (delete) WRITE(*,*) "No deallocation of memory implemented in eig66_mpi"
222
    ENDIF
223 224 225 226 227 228 229 230 231 232
    IF (PRESENT(filename)) CALL priv_writetofile()
  CONTAINS
    SUBROUTINE priv_writetofile()
      USE m_eig66_DA,ONLY:open_eig_DA=>open_eig,write_eig_DA=>write_eig,close_eig_DA=>close_eig
      IMPLICIT NONE

      INTEGER:: nlotot,nk,jspin,nv,i,ii,tmp_id
      REAL   :: wk,bk3(3),evac(2)
      INTEGER :: k1(d%nmat),k2(d%nmat),k3(d%nmat),kveclo(d%nlotot)
      REAL    :: eig(d%neig),ello(d%nlo,d%ntype),el(d%lmax,d%ntype)
233 234
      REAL    :: z_r(d%nmat,d%neig)
      COMPLEX :: z_c(d%nmat,d%neig)
235
      nlotot=d%nlotot
236

237 238 239
      IF (d%irank==0) THEN
         tmp_id=eig66_data_newid(DA_mode)
         IF (d%l_dos) CPP_error("Could not write DOS data")
240
         CALL open_eig_DA(tmp_id,d%nmat,d%neig,d%nkpts,d%jspins,d%lmax,d%nlo,d%ntype,d%nlotot,.FALSE.,.FALSE.,d%l_real,d%l_soc,.FALSE.,.FALSE.,filename)
241 242
         DO jspin=1,d%jspins
            DO nk=1,d%nkpts
243 244 245 246 247 248 249
               if (d%l_real) THEN
                  CALL read_eig(id,nk,jspin,nv,i,k1,k2,k3,bk3,wk,ii,eig,el,ello,evac,kveclo,z=z_r)
                  CALL write_eig_DA(tmp_id,nk,jspin,ii,ii,nv,i,k1,k2,k3,bk3,wk,eig,el,ello,evac,nlotot,kveclo,z=z_r)
               else
                  CALL read_eig(id,nk,jspin,nv,i,k1,k2,k3,bk3,wk,ii,eig,el,ello,evac,kveclo,z=z_c)
                  CALL write_eig_DA(tmp_id,nk,jspin,ii,ii,nv,i,k1,k2,k3,bk3,wk,eig,el,ello,evac,nlotot,kveclo,z=z_c)
               end if
250
            ENDDO
251
         ENDDO
252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
         CALL close_eig_DA(tmp_id)
      ENDIF
      CALL eig66_remove_data(id)
    END SUBROUTINE priv_writetofile

  END SUBROUTINE close_eig

  SUBROUTINE read_eig(id,nk,jspin,nv,nmat,k1,k2,k3,bk3,wk,neig,eig,el,&
       ello,evac,kveclo,n_start,n_end,z)
    IMPLICIT NONE
    INTEGER, INTENT(IN)            :: id,nk,jspin
    INTEGER, INTENT(OUT),OPTIONAL  :: nv,nmat
    INTEGER, INTENT(OUT),OPTIONAL  :: neig
    REAL,    INTENT(OUT),OPTIONAL  :: eig(:)
    INTEGER, INTENT(OUT),OPTIONAL  :: k1(:),k2(:),k3(:),kveclo(:)
    REAL,    INTENT(OUT),OPTIONAL  :: evac(:),ello(:,:),el(:,:)
    REAL,    INTENT(OUT),OPTIONAL  :: bk3(:),wk
    INTEGER, INTENT(IN),OPTIONAL   :: n_start,n_end
    CLASS(*),TARGET,INTENT(OUT),OPTIONAL  :: z(:,:)
271 272

#ifdef CPP_MPI
273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348
    INTEGER                   :: pe,tmp_size,e
    INTEGER(MPI_ADDRESS_KIND) :: slot
    INTEGER                   :: n1,n2,n3,n
    INTEGER,ALLOCATABLE       :: tmp_int(:)
    REAL,ALLOCATABLE          :: tmp_real(:)
    COMPLEX,ALLOCATABLE       :: tmp_cmplx(:)
    TYPE(t_data_MPI),POINTER :: d
    CALL priv_find_data(id,d)
    pe=d%pe_basis(nk,jspin)
    slot=d%slot_basis(nk,jspin)
    IF (PRESENT(neig))THEN
       CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,d%neig_handle,e)
       ! Get current values
       CALL  MPI_GET(neig,1,MPI_INTEGER,pe,slot,1,MPI_INTEGER,d%neig_handle,e)
       CALL MPI_WIN_UNLOCK(pe,d%neig_handle,e)

    ENDIF
    !read the integer values
    IF (ANY((/PRESENT(nv),PRESENT(nmat),PRESENT(k1),PRESENT(k2),PRESENT(k3),PRESENT(kveclo)/))) THEN
       tmp_size=4+3*d%size_k
       IF (PRESENT(kveclo)) tmp_size=tmp_size+SIZE(kveclo)
       ALLOCATE(tmp_int(tmp_size))
       CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,d%int_handle,e)
       ! Get current values
       CALL  MPI_GET(tmp_int,tmp_size,MPI_INTEGER,pe,slot,tmp_size,MPI_INTEGER,d%int_handle,e)
       CALL MPI_WIN_UNLOCK(pe,d%int_handle,e)
       !IF (present(neig)) neig=tmp_int(1)
       IF (PRESENT(nv))   nv=tmp_int(2)
       IF (PRESENT(nmat)) nmat=tmp_int(3)
       IF (PRESENT(k1))   k1=tmp_int(4+1:4+SIZE(k1))
       IF (PRESENT(k2))   k2=tmp_int(4+d%size_k+1:4+d%size_k+SIZE(k2))
       IF (PRESENT(k3))   k3=tmp_int(4+2*d%size_k+1:4+2*d%size_k+SIZE(k3))
       IF (PRESENT(kveclo)) kveclo=tmp_int(4+3*d%size_k+1:4+3*d%size_k+SIZE(kveclo))

    ENDIF
    !read the real-values
    IF (ANY((/PRESENT(wk),PRESENT(bk3),PRESENT(el),PRESENT(ello),PRESENT(evac)/))) THEN
       tmp_size=6+d%size_el+d%size_ello
       ALLOCATE(tmp_real(tmp_size))
       CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,d%real_handle,e)
       ! Get current values
       CALL  MPI_GET(tmp_real,tmp_size,MPI_DOUBLE_PRECISION,pe,slot,tmp_size,MPI_DOUBLE_PRECISION,d%real_handle,e)
       CALL MPI_WIN_UNLOCK(pe,d%real_handle,e)
       IF (PRESENT(wk))   wk=tmp_real(1)
       IF (PRESENT(bk3))  bk3=tmp_real(2:4)
       IF (PRESENT(evac)) evac=tmp_real(5:6)
       IF (PRESENT(el))   el=RESHAPE(tmp_real(6+1:6+SIZE(el)),SHAPE(el))
       IF (PRESENT(ello)) ello=RESHAPE(tmp_real(6+d%size_el+1:6+d%size_el+SIZE(ello)),SHAPE(ello))
       DEALLOCATE(tmp_real)
    ENDIF
    IF (PRESENT(eig)) THEN
       CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,d%eig_handle,e)
       ALLOCATE(tmp_real(d%size_eig))
       CALL MPI_GET(tmp_real,d%size_eig,MPI_DOUBLE_PRECISION,pe,slot,d%size_eig,MPI_DOUBLE_PRECISION,d%eig_handle,e)
       CALL MPI_WIN_UNLOCK(pe,d%eig_handle,e)
       n1=1;n3=1;n2=SIZE(eig)
       IF (PRESENT(n_start)) n1=n_start
       IF (PRESENT(n_end)) n2=n_end
       eig(:n2-n1+1)=tmp_real(n1:n2)
       DEALLOCATE(tmp_real)
    ENDIF

    IF (PRESENT(z)) THEN
       tmp_size=SIZE(z,1)
       ALLOCATE(tmp_real(tmp_size))
       ALLOCATE(tmp_cmplx(tmp_size))
       DO n=1,SIZE(z,2)
          n1=n
          IF (PRESENT(n_start)) n1=n_start+n-1
          IF (PRESENT(n_end)) THEN
             IF (n1>n_end) CYCLE
          ENDIF
          slot=d%slot_ev(nk,jspin,n1)
          pe=d%pe_ev(nk,jspin,n1)
          SELECT TYPE(z)
          TYPE IS(REAL)
349 350 351 352 353 354 355 356 357 358 359 360 361
             if (.not.d%l_real) THEN
                CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,d%zc_handle,e)
                CALL MPI_GET(tmp_cmplx,tmp_size,MPI_DOUBLE_COMPLEX,pe,slot,tmp_size,MPI_DOUBLE_COMPLEX,d%zc_handle,e)
                CALL MPI_WIN_UNLOCK(pe,d%zc_handle,e)
                !print *, nk,jspin,n1,"r PE:",pe," Slot: ",slot," Size:",tmp_size,tmp_cmplx(1)
                z(:,n)=REAL(tmp_cmplx)
             else
                CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,d%zr_handle,e)
                CALL MPI_GET(tmp_real,tmp_size,MPI_DOUBLE_PRECISION,pe,slot,tmp_size,MPI_DOUBLE_PRECISION,d%zr_handle,e)
                CALL MPI_WIN_UNLOCK(pe,d%zr_handle,e)
                !print *, nk,jspin,n1,"r PE:",pe," Slot: ",slot," Size:",tmp_size,tmp_real(1)
                z(:,n)=tmp_real
             endif
362
          TYPE IS (COMPLEX)
363
             if (.not.d%l_real) call judft_error("Could not real complex data, only real data is stored",calledby="eig66_mpi%read_eig")
364 365 366 367 368 369 370 371
             CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,d%zc_handle,e)
             CALL MPI_GET(tmp_cmplx,tmp_size,MPI_DOUBLE_COMPLEX,pe,slot,tmp_size,MPI_DOUBLE_COMPLEX,d%zc_handle,e)
             CALL MPI_WIN_UNLOCK(pe,d%zc_handle,e)
             !print *, nk,jspin,n1,"r PE:",pe," Slot: ",slot," Size:",tmp_size,tmp_cmplx(1)
             z(:,n)=tmp_cmplx
          END SELECT
       ENDDO
    ENDIF
372 373

#endif
374 375 376 377 378 379 380 381 382 383 384 385 386
  END SUBROUTINE read_eig

  SUBROUTINE write_eig(id,nk,jspin,neig,neig_total,nv,nmat,k1,k2,k3,bk3,wk, &
       eig,el,ello,evac,                     &
       nlotot,kveclo,n_size,n_rank,z)
    INTEGER, INTENT(IN)          :: id,nk,jspin
    INTEGER, INTENT(IN),OPTIONAL :: n_size,n_rank
    REAL,    INTENT(IN),OPTIONAL :: wk
    INTEGER, INTENT(IN),OPTIONAL :: neig,nv,nmat,nlotot,neig_total
    INTEGER, INTENT(IN),OPTIONAL :: k1(:),k2(:),k3(:),kveclo(:)
    REAL,    INTENT(IN),OPTIONAL :: bk3(3),eig(:),el(:,:)
    REAL,    INTENT(IN),OPTIONAL :: evac(:),ello(:,:)
    CLASS(*),INTENT(IN),OPTIONAL :: z(:,:)
387 388

#ifdef CPP_MPI
389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453
    INTEGER                   :: pe,tmp_size,e
    INTEGER(MPI_ADDRESS_KIND) :: slot
    INTEGER                   :: n1,n2,n3,n
    INTEGER,ALLOCATABLE       :: tmp_int(:)
    REAL,ALLOCATABLE          :: tmp_real(:)
    COMPLEX,ALLOCATABLE       :: tmp_cmplx(:)
    LOGICAL                   :: acc
    TYPE(t_data_MPI),POINTER :: d

    CALL priv_find_data(id,d)

    pe=d%pe_basis(nk,jspin)
    slot=d%slot_basis(nk,jspin)
    !write the number of eigenvalues values
    IF (PRESENT(neig_total)) THEN
       CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,d%neig_handle,e)
       ALLOCATE(tmp_int(1))
       tmp_int(1)=neig_total
       CALL MPI_PUT(tmp_int,1,MPI_INTEGER,pe,slot,1,MPI_INTEGER,d%neig_handle,e)
       CALL MPI_WIN_UNLOCK(pe,d%neig_handle,e)
       DEALLOCATE(tmp_int)
    ENDIF

    IF (ANY((/PRESENT(nv),PRESENT(nmat),PRESENT(nlotot),PRESENT(k1),PRESENT(k2),PRESENT(k3),PRESENT(kveclo)/))) THEN
       tmp_size=5+3*d%size_k
       IF (PRESENT(kveclo)) tmp_size=tmp_size+SIZE(kveclo)
       ALLOCATE(tmp_int(tmp_size))
       tmp_int=9999999
       tmp_int(1)=0
       IF (PRESENT(nv))   tmp_int(2)=nv
       IF (PRESENT(nmat)) tmp_int(3)=nmat
       IF (PRESENT(nlotot)) tmp_int(4)=nlotot
       IF (PRESENT(k1))   tmp_int(4+1:4+SIZE(k1))=k1
       IF (PRESENT(k2))   tmp_int(4+d%size_k+1:4+d%size_k+SIZE(k2))=k2
       IF (PRESENT(k3))   tmp_int(4+2*d%size_k+1:4+2*d%size_k+SIZE(k3))=k3
       IF (PRESENT(kveclo)) tmp_int(4+3*d%size_k+1:4+3*d%size_k+SIZE(kveclo))=kveclo
       CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,d%int_handle,e)
       CALL MPI_ACCUMULATE(tmp_int,tmp_size,MPI_INTEGER,pe,slot,tmp_size,MPI_INTEGER,MPI_MIN,d%int_handle,e)
       CALL MPI_WIN_UNLOCK(pe,d%int_handle,e)
    ENDIF
    !write the real-values
    IF (ANY((/PRESENT(wk),PRESENT(bk3),PRESENT(el),PRESENT(ello),PRESENT(evac)/))) THEN
       tmp_size=6+d%size_el+d%size_ello
       ALLOCATE(tmp_real(tmp_size))
       tmp_real=1E99
       IF (PRESENT(wk))   tmp_real(1)=wk
       IF (PRESENT(bk3))  tmp_real(2:4)=bk3
       IF (PRESENT(evac)) tmp_real(5:6)=evac
       IF (PRESENT(el))   tmp_real(6+1:6+SIZE(el))=RESHAPE(el,(/SIZE(el)/))
       IF (PRESENT(ello)) tmp_real(6+d%size_el+1:6+d%size_el+SIZE(ello))=RESHAPE(ello,(/SIZE(ello)/))

       CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,d%real_handle,e)
       CALL MPI_ACCUMULATE(tmp_real,tmp_size,MPI_DOUBLE_PRECISION,pe,slot,tmp_size,MPI_DOUBLE_PRECISION,MPI_MIN,d%real_handle,e)
       CALL MPI_WIN_UNLOCK(pe,d%real_handle,e)
       DEALLOCATE(tmp_real)
    ENDIF
    IF (PRESENT(eig)) THEN
       ALLOCATE(tmp_real(d%size_eig))
       tmp_real=1E99
       n1=1;n3=1
       IF (PRESENT(n_rank)) n1=n_rank+1
       IF (PRESENT(n_size)) n3=n_size
       n2=SIZE(eig)*n3+n1-1
       tmp_real(n1:n2:n3)=eig
       CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,d%eig_handle,e)
454 455 456 457 458
       IF (n3.ne.1) THEN
          CALL MPI_ACCUMULATE(tmp_real,d%size_eig,MPI_DOUBLE_PRECISION,pe,slot,d%size_eig,MPI_DOUBLE_PRECISION,MPI_MIN,d%eig_handle,e)
       ELSE
          CALL MPI_PUT(tmp_real,d%size_eig,MPI_DOUBLE_PRECISION,pe,slot,d%size_eig,MPI_DOUBLE_PRECISION,d%eig_handle,e)
       ENDIF
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
       CALL MPI_WIN_UNLOCK(pe,d%eig_handle,e)
       DEALLOCATE(tmp_real)
    ENDIF
    IF (PRESENT(z)) THEN
       tmp_size=SIZE(z,1)
       ALLOCATE(tmp_real(tmp_size))
       ALLOCATE(tmp_cmplx(tmp_size))
       DO n=1,SIZE(z,2)
          n1=n-1
          IF (PRESENT(n_size)) n1=n_size*n1
          IF (PRESENT(n_rank)) n1=n1+n_rank
          slot=d%slot_ev(nk,jspin,n1+1)
          pe=d%pe_ev(nk,jspin,n1+1)
          !print *, "PE:",pe," Slot: ",slot," Size:",tmp_size,tmp_real(1)
          SELECT TYPE(z)
          TYPE IS(REAL)
475 476 477 478 479 480 481 482 483 484 485
             if (.not.d%l_real) THEN
                tmp_cmplx=z(:,n)
                CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,d%zc_handle,e)
                CALL MPI_PUT(tmp_cmplx,tmp_size,MPI_DOUBLE_COMPLEX,pe,slot,tmp_size,MPI_DOUBLE_COMPLEX,d%zc_handle,e)
                CALL MPI_WIN_UNLOCK(pe,d%zc_handle,e)
             else
                tmp_real=z(:,n)
                CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,d%zr_handle,e)
                CALL MPI_PUT(tmp_real,tmp_size,MPI_DOUBLE_PRECISION,pe,slot,tmp_size,MPI_DOUBLE_PRECISION,d%zr_handle,e)
                CALL MPI_WIN_UNLOCK(pe,d%zr_handle,e)
             endif
486
          TYPE IS(COMPLEX)
487
             if (d%l_real) CALL juDFT_error("Could not write complex data to file prepared for real data",calledby="eig66_mpi%write_eig")
488 489 490 491 492 493 494
             tmp_cmplx=z(:,n)
             CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,d%zc_handle,e)
             CALL MPI_PUT(tmp_cmplx,tmp_size,MPI_DOUBLE_COMPLEX,pe,slot,tmp_size,MPI_DOUBLE_COMPLEX,d%zc_handle,e)
             CALL MPI_WIN_UNLOCK(pe,d%zc_handle,e)
          END SELECT
       ENDDO
    ENDIF
495 496

#endif
497 498
  END SUBROUTINE write_eig

499
#ifdef CPP_MPI
500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516
  SUBROUTINE priv_put_data(pe,slot,DATA,handle)
    IMPLICIT NONE
    INTEGER,INTENT(IN)  :: pe,slot
    CLASS(*),INTENT(IN) :: DATA(:)
    INTEGER,INTENT(IN)  :: handle

    INTEGER             :: len,e
    INTEGER,ALLOCATABLE :: int_tmp(:)
    REAL,ALLOCATABLE    :: real_tmp(:)
    COMPLEX,ALLOCATABLE:: cmplx_tmp(:)
    INCLUDE 'mpif.h'
    len=SIZE(DATA)
    SELECT TYPE(DATA)
    TYPE IS (INTEGER)
       ALLOCATE(int_tmp(len))
       int_tmp=DATA
       CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,handle,e)
517
       CALL MPI_PUT(int_tmp,len,MPI_INTEGER,pe,int(slot,MPI_ADDRESS_KIND),len,MPI_INTEGER,handle,e)
518 519 520 521 522
       CALL MPI_WIN_UNLOCK(pe,handle,e)
    TYPE is (REAL)
       ALLOCATE(real_tmp(len))
       real_tmp=DATA
       CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,handle,e)
523
       CALL MPI_PUT(real_tmp,len,MPI_DOUBLE_PRECISION,pe,int(slot,MPI_ADDRESS_KIND),len,MPI_DOUBLE_PRECISION,handle,e)
524 525 526 527 528
       CALL MPI_WIN_UNLOCK(pe,handle,e)
    TYPE is (COMPLEX)
       ALLOCATE(cmplx_tmp(len))
       cmplx_tmp=DATA
       CALL MPI_WIN_LOCK(MPI_LOCK_EXCLUSIVE,pe,0,handle,e)
529
       CALL MPI_PUT(cmplx_tmp,len,MPI_DOUBLE_COMPLEX,pe,int(slot,MPI_ADDRESS_KIND),len,MPI_DOUBLE_COMPLEX,handle,e)
530 531 532 533
       CALL MPI_WIN_UNLOCK(pe,handle,e)
    END SELECT
  END SUBROUTINE priv_put_data

534
  SUBROUTINE priv_get_data(pe,slot,len,handle,idata,rdata,cdata)
535 536
    IMPLICIT NONE
    INTEGER,INTENT(IN)  :: pe,slot,len
537 538 539
    INTEGER,INTENT(OUT),optional :: iDATA(len)
    REAL,INTENT(OUT),optional    :: rDATA(len)
    COMPLEX,INTENT(OUT),optional :: cDATA(len)
540 541 542 543 544 545 546 547
    INTEGER,INTENT(IN)  :: handle

    INTEGER             :: e
    INTEGER,ALLOCATABLE :: int_tmp(:)
    REAL,ALLOCATABLE    :: real_tmp(:)
    COMPLEX,ALLOCATABLE:: cmplx_tmp(:)
    INCLUDE 'mpif.h'

548
    IF (present(idata)) THEN
549 550
       ALLOCATE(int_tmp(len))
       CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,handle,e)
551
       CALL MPI_GET(int_tmp,len,MPI_INTEGER,pe,int(slot,MPI_ADDRESS_KIND),len,MPI_INTEGER,handle,e)
552
       CALL MPI_WIN_UNLOCK(pe,handle,e)
553 554
       iDATA=int_tmp
    ELSE IF (PRESENT(rdata)) THEN
555 556
       ALLOCATE(real_tmp(len))
       CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,handle,e)
557
       CALL MPI_GET(real_tmp,len,MPI_DOUBLE_PRECISION,pe,int(slot,MPI_ADDRESS_KIND),len,MPI_DOUBLE_PRECISION,handle,e)
558
       CALL MPI_WIN_UNLOCK(pe,handle,e)
559 560
       rDATA=real_tmp
    ELSE IF (PRESENT(cdata)) THEN
561 562
       ALLOCATE(cmplx_tmp(len))
       CALL MPI_WIN_LOCK(MPI_LOCK_SHARED,pe,0,handle,e)
563
       CALL MPI_GET(cmplx_tmp,len,MPI_DOUBLE_COMPLEX,pe,int(slot,MPI_ADDRESS_KIND),len,MPI_DOUBLE_COMPLEX,handle,e)
564
       CALL MPI_WIN_UNLOCK(pe,handle,e)
565 566 567 568 569
       cDATA=cmplx_tmp
    ELSE
       call judft_error("BUG in priv_get_data")
    ENDIF

570
  END SUBROUTINE priv_get_data
571
#endif
572 573 574 575 576 577 578 579 580

  SUBROUTINE write_dos(id,nk,jspin,qal,qvac,qis,qvlay,qstars,ksym,jsym,mcd,qintsl,qmtsl,qmtp,orbcomp)
    IMPLICIT NONE
    INTEGER, INTENT(IN)          :: id,nk,jspin
    REAL,INTENT(IN)              :: qal(:,:,:),qvac(:,:),qis(:),qvlay(:,:,:)
    COMPLEX,INTENT(IN)           :: qstars(:,:,:,:)
    INTEGER,INTENT(IN)           :: ksym(:),jsym(:)
    REAL,INTENT(IN),OPTIONAL     :: mcd(:,:,:)
    REAL,INTENT(IN),OPTIONAL     :: qintsl(:,:),qmtsl(:,:),qmtp(:,:),orbcomp(:,:,:)
581
#ifdef CPP_MPI
582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602
    TYPE(t_data_MPI),POINTER :: d
    INTEGER:: pe,slot

    CALL priv_find_data(id,d)
    pe=d%pe_basis(nk,jspin)
    slot=d%slot_basis(nk,jspin)

    CALL priv_put_data(pe,slot,RESHAPE(qal,(/SIZE(qal)/)),d%qal_handle)
    CALL priv_put_data(pe,slot,RESHAPE(qvac,(/SIZE(qvac)/)),d%qvac_handle)
    CALL priv_put_data(pe,slot,RESHAPE(qis,(/SIZE(qis)/)),d%qis_handle)
    CALL priv_put_data(pe,slot,RESHAPE(qvlay,(/SIZE(qvlay)/)),d%qvlay_handle)
    CALL priv_put_data(pe,slot,RESHAPE(qstars,(/SIZE(qstars)/)),d%qstars_handle)
    CALL priv_put_data(pe,slot,RESHAPE(ksym,(/SIZE(ksym)/)),d%ksym_handle)
    CALL priv_put_data(pe,slot,RESHAPE(jsym,(/SIZE(jsym)/)),d%jsym_handle)
    IF (d%l_mcd.AND.PRESENT(mcd))  CALL priv_put_data(pe,slot,RESHAPE(mcd,(/SIZE(mcd)/)),d%mcd_handle)
    IF (d%l_orb.AND.PRESENT(qintsl)) THEN
       CALL priv_put_data(pe,slot,RESHAPE(qintsl,(/SIZE(qintsl)/)),d%qintsl_handle)
       CALL priv_put_data(pe,slot,RESHAPE(qmtsl,(/SIZE(qmtsl)/)),d%qmtsl_handle)
       CALL priv_put_data(pe,slot,RESHAPE(qmtp,(/SIZE(qmtp)/)),d%qmtp_handle)
       CALL priv_put_data(pe,slot,RESHAPE(orbcomp,(/SIZE(orbcomp)/)),d%orbcomp_handle)
    ENDIF
603
#endif
604 605 606 607 608 609 610 611 612 613
  END SUBROUTINE write_dos

  SUBROUTINE read_dos(id,nk,jspin,qal,qvac,qis,qvlay,qstars,ksym,jsym,mcd,qintsl,qmtsl,qmtp,orbcomp)
    IMPLICIT NONE
    INTEGER, INTENT(IN)          :: id,nk,jspin
    REAL,INTENT(out)              :: qal(:,:,:),qvac(:,:),qis(:),qvlay(:,:,:)
    COMPLEX,INTENT(out)           :: qstars(:,:,:,:)
    INTEGER,INTENT(out)           :: ksym(:),jsym(:)
    REAL,INTENT(out),OPTIONAL     :: mcd(:,:,:)
    REAL,INTENT(out),OPTIONAL     :: qintsl(:,:),qmtsl(:,:),qmtp(:,:),orbcomp(:,:,:)
614
#ifdef CPP_MPI
615 616 617 618 619 620 621
    TYPE(t_data_MPI),POINTER :: d
    INTEGER:: pe,slot

    CALL priv_find_data(id,d)
    pe=d%pe_basis(nk,jspin)
    slot=d%slot_basis(nk,jspin)

622 623 624 625 626 627 628 629
    CALL priv_get_data(pe,slot,SIZE(qal),d%qal_handle,rdata=qal)
    CALL priv_get_data(pe,slot,SIZE(qvac),d%qvac_handle,rdata=qvac)
    CALL priv_get_data(pe,slot,SIZE(qis),d%qis_handle,rdata=qis)
    CALL priv_get_data(pe,slot,SIZE(qvlay),d%qvlay_handle,rdata=qvlay)
    CALL priv_get_data(pe,slot,SIZE(qstars),d%qstars_handle,cdata=qstars)
    CALL priv_get_data(pe,slot,SIZE(ksym),d%ksym_handle,idata=ksym)
    CALL priv_get_data(pe,slot,SIZE(jsym),d%jsym_handle,idata=jsym)
    IF (d%l_mcd.AND.PRESENT(mcd))  CALL priv_get_data(pe,slot,SIZE(mcd),d%mcd_handle,rdata=mcd)
630
    IF (d%l_orb.AND.PRESENT(qintsl)) THEN
631 632 633 634
       CALL priv_get_data(pe,slot,SIZE(qintsl),d%qintsl_handle,rdata=qintsl)
       CALL priv_get_data(pe,slot,SIZE(qmtsl),d%qmtsl_handle,rdata=qmtsl)
       CALL priv_get_data(pe,slot,SIZE(qmtp),d%qmtp_handle,rdata=qmtp)
       CALL priv_get_data(pe,slot,SIZE(orbcomp),d%orbcomp_handle,rdata=orbcomp)
635
    ENDIF
636
#endif
637 638
  END SUBROUTINE read_dos

639 640

#ifdef CPP_MPI
641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678
  SUBROUTINE create_maps(d,isize,nkpts,jspins,neig,n_size)
    IMPLICIT NONE
    TYPE(t_data_MPI),INTENT(INOUT):: d
    INTEGER,INTENT(IN):: isize,nkpts,jspins,neig,n_size

    INTEGER:: nk,j,n1,n2,n,pe,n_members
    INTEGER::used(0:isize)

    ALLOCATE(d%pe_basis(nkpts,jspins),d%slot_basis(nkpts,jspins))
    ALLOCATE(d%pe_ev(nkpts,jspins,neig),d%slot_ev(nkpts,jspins,neig))

    !basis contains a total of nkpts*jspins entries
    d%pe_basis=-1
    d%pe_ev=-1
    used=0
    n_members=isize/n_size !no of k-points in parallel
    DO j=1,jspins
       DO nk=1,nkpts
          n1=nk+(j-1)*nkpts-1
          pe=MOD(n1,n_members)*n_size
          d%pe_basis(nk,j)=pe
          d%slot_basis(nk,j)=used(pe)
          used(pe)=used(pe)+1
       ENDDO
    ENDDO
    used=0
    DO n=1,neig
       DO j=1,jspins
          DO nk=1,nkpts
             n1=nk+(j-1)*nkpts-1
             !eigenvectors have more entries
             pe=MOD(n1,n_members)*n_size+MOD(n,n_size)
             d%pe_ev(nk,j,n)=pe
             d%slot_ev(nk,j,n)=used(pe)
             used(pe)=used(pe)+1
          ENDDO
       ENDDO
    ENDDO
679

680
  END SUBROUTINE create_maps
681 682 683 684


#endif
END MODULE m_eig66_mpi