| 1 | subroutine setupds(gs_handle,nx,ny,nz,nel,melg,vertex,glo_num)
|
|---|
| 2 | include 'SIZE'
|
|---|
| 3 | include 'INPUT'
|
|---|
| 4 | include 'PARALLEL'
|
|---|
| 5 | include 'NONCON'
|
|---|
| 6 |
|
|---|
| 7 | integer gs_handle
|
|---|
| 8 | integer vertex(1)
|
|---|
| 9 | integer*8 glo_num(1),ngv
|
|---|
| 10 |
|
|---|
| 11 | common /nekmpi/ mid,mp,nekcomm,nekgroup,nekreal
|
|---|
| 12 |
|
|---|
| 13 | t0 = dnekclock()
|
|---|
| 14 |
|
|---|
| 15 | c Global-to-local mapping for gs
|
|---|
| 16 | call set_vert(glo_num,ngv,nx,nel,vertex,.false.)
|
|---|
| 17 |
|
|---|
| 18 | c Initialize gather-scatter code
|
|---|
| 19 | ntot = nx*ny*nz*nel
|
|---|
| 20 | call fgslib_gs_setup(gs_handle,glo_num,ntot,nekcomm,mp)
|
|---|
| 21 |
|
|---|
| 22 | c call gs_chkr(glo_num)
|
|---|
| 23 |
|
|---|
| 24 | t1 = dnekclock() - t0
|
|---|
| 25 | if (nio.eq.0) then
|
|---|
| 26 | write(6,1) t1,gs_handle,nx,ngv,melg
|
|---|
| 27 | 1 format(' setupds time',1pe11.4,' seconds ',2i3,2i12)
|
|---|
| 28 | endif
|
|---|
| 29 | c
|
|---|
| 30 | return
|
|---|
| 31 | end
|
|---|
| 32 | c-----------------------------------------------------------------------
|
|---|
| 33 | subroutine dssum(u,nx,ny,nz)
|
|---|
| 34 | include 'SIZE'
|
|---|
| 35 | include 'CTIMER'
|
|---|
| 36 | include 'INPUT'
|
|---|
| 37 | include 'NONCON'
|
|---|
| 38 | include 'PARALLEL'
|
|---|
| 39 | include 'TSTEP'
|
|---|
| 40 | real u(1)
|
|---|
| 41 |
|
|---|
| 42 | parameter (lface=lx1*ly1)
|
|---|
| 43 | common /nonctmp/ uin(lface,2*ldim),uout(lface)
|
|---|
| 44 |
|
|---|
| 45 | ifldt = ifield
|
|---|
| 46 | c if (ifldt.eq.0) ifldt = 1
|
|---|
| 47 | if (ifldt.eq.ifldmhd) ifldt = 1
|
|---|
| 48 | c write(6,*) ifldt,ifield,gsh_fld(ifldt),imesh,' ifldt'
|
|---|
| 49 |
|
|---|
| 50 | if (ifsync) call nekgsync()
|
|---|
| 51 |
|
|---|
| 52 | #ifdef TIMER
|
|---|
| 53 | if (icalld.eq.0) then
|
|---|
| 54 | tdsmx=0.
|
|---|
| 55 | tdsmn=0.
|
|---|
| 56 | endif
|
|---|
| 57 | icalld=icalld+1
|
|---|
| 58 | etime1=dnekclock()
|
|---|
| 59 | #endif
|
|---|
| 60 |
|
|---|
| 61 | c
|
|---|
| 62 | c T ~ ~T T
|
|---|
| 63 | c Implement QQ := J Q Q J
|
|---|
| 64 | c
|
|---|
| 65 | c
|
|---|
| 66 | c T
|
|---|
| 67 | c This is the J part, translating child data
|
|---|
| 68 | c
|
|---|
| 69 | c call apply_Jt(u,nx,ny,nz,nel)
|
|---|
| 70 | c
|
|---|
| 71 | c
|
|---|
| 72 | c
|
|---|
| 73 | c ~ ~T
|
|---|
| 74 | c This is the Q Q part
|
|---|
| 75 | c
|
|---|
| 76 | if (gsh_fld(ifldt).ge.0) then
|
|---|
| 77 | if (nio.eq.0.and.loglevel.gt.5)
|
|---|
| 78 | $ write(6,*) 'dssum', ifldt
|
|---|
| 79 | call fgslib_gs_op(gsh_fld(ifldt),u,1,1,0) ! 1 ==> +
|
|---|
| 80 | endif
|
|---|
| 81 | c
|
|---|
| 82 | c
|
|---|
| 83 | c
|
|---|
| 84 | c This is the J part, interpolating parent solution onto child
|
|---|
| 85 | c
|
|---|
| 86 | c call apply_J(u,nx,ny,nz,nel)
|
|---|
| 87 | c
|
|---|
| 88 | c
|
|---|
| 89 | #ifdef TIMER
|
|---|
| 90 | timee=(dnekclock()-etime1)
|
|---|
| 91 | tdsum=tdsum+timee
|
|---|
| 92 | ndsum=icalld
|
|---|
| 93 | tdsmx=max(timee,tdsmx)
|
|---|
| 94 | tdsmn=min(timee,tdsmn)
|
|---|
| 95 | #endif
|
|---|
| 96 | c
|
|---|
| 97 | return
|
|---|
| 98 | end
|
|---|
| 99 | c-----------------------------------------------------------------------
|
|---|
| 100 | subroutine dsop(u,op,nx,ny,nz)
|
|---|
| 101 | include 'SIZE'
|
|---|
| 102 | include 'PARALLEL'
|
|---|
| 103 | include 'INPUT'
|
|---|
| 104 | include 'TSTEP'
|
|---|
| 105 | include 'CTIMER'
|
|---|
| 106 |
|
|---|
| 107 | real u(1)
|
|---|
| 108 | character*3 op
|
|---|
| 109 | character*10 s1,s2
|
|---|
| 110 | c
|
|---|
| 111 | c o gs recognized operations:
|
|---|
| 112 | c
|
|---|
| 113 | c o "+" ==> addition.
|
|---|
| 114 | c o "*" ==> multiplication.
|
|---|
| 115 | c o "M" ==> maximum.
|
|---|
| 116 | c o "m" ==> minimum.
|
|---|
| 117 | c o "A" ==> (fabs(x)>fabs(y)) ? (x) : (y), ident=0.0.
|
|---|
| 118 | c o "a" ==> (fabs(x)<fabs(y)) ? (x) : (y), ident=MAX_DBL
|
|---|
| 119 | c o "e" ==> ((x)==0.0) ? (y) : (x), ident=0.0.
|
|---|
| 120 | c
|
|---|
| 121 | c o note: a binary function pointer flavor exists.
|
|---|
| 122 | c
|
|---|
| 123 | c
|
|---|
| 124 | c o gs level:
|
|---|
| 125 | c
|
|---|
| 126 | c o level=0 ==> pure tree
|
|---|
| 127 | c o level>=num_nodes-1 ==> pure pairwise
|
|---|
| 128 | c o level = 1,...num_nodes-2 ==> mix tree/pairwise.
|
|---|
| 129 | c
|
|---|
| 130 | c
|
|---|
| 131 | ifldt = ifield
|
|---|
| 132 | c if (ifldt.eq.0) ifldt = 1
|
|---|
| 133 | if (ifldt.eq.ifldmhd) ifldt = 1
|
|---|
| 134 |
|
|---|
| 135 | c if (nio.eq.0)
|
|---|
| 136 | c $ write(6,*) istep,' dsop: ',op,ifield,ifldt,gsh_fld(ifldt)
|
|---|
| 137 |
|
|---|
| 138 | if(ifsync) call nekgsync()
|
|---|
| 139 |
|
|---|
| 140 | if (op.eq.'+ ') call fgslib_gs_op(gsh_fld(ifldt),u,1,1,0)
|
|---|
| 141 | if (op.eq.'sum') call fgslib_gs_op(gsh_fld(ifldt),u,1,1,0)
|
|---|
| 142 | if (op.eq.'SUM') call fgslib_gs_op(gsh_fld(ifldt),u,1,1,0)
|
|---|
| 143 |
|
|---|
| 144 | if (op.eq.'* ') call fgslib_gs_op(gsh_fld(ifldt),u,1,2,0)
|
|---|
| 145 | if (op.eq.'mul') call fgslib_gs_op(gsh_fld(ifldt),u,1,2,0)
|
|---|
| 146 | if (op.eq.'MUL') call fgslib_gs_op(gsh_fld(ifldt),u,1,2,0)
|
|---|
| 147 |
|
|---|
| 148 | if (op.eq.'m ') call fgslib_gs_op(gsh_fld(ifldt),u,1,3,0)
|
|---|
| 149 | if (op.eq.'min') call fgslib_gs_op(gsh_fld(ifldt),u,1,3,0)
|
|---|
| 150 | if (op.eq.'mna') call fgslib_gs_op(gsh_fld(ifldt),u,1,3,0)
|
|---|
| 151 | if (op.eq.'MIN') call fgslib_gs_op(gsh_fld(ifldt),u,1,3,0)
|
|---|
| 152 | if (op.eq.'MNA') call fgslib_gs_op(gsh_fld(ifldt),u,1,3,0)
|
|---|
| 153 |
|
|---|
| 154 | if (op.eq.'M ') call fgslib_gs_op(gsh_fld(ifldt),u,1,4,0)
|
|---|
| 155 | if (op.eq.'max') call fgslib_gs_op(gsh_fld(ifldt),u,1,4,0)
|
|---|
| 156 | if (op.eq.'mxa') call fgslib_gs_op(gsh_fld(ifldt),u,1,4,0)
|
|---|
| 157 | if (op.eq.'MAX') call fgslib_gs_op(gsh_fld(ifldt),u,1,4,0)
|
|---|
| 158 | if (op.eq.'MXA') call fgslib_gs_op(gsh_fld(ifldt),u,1,4,0)
|
|---|
| 159 | c
|
|---|
| 160 | return
|
|---|
| 161 | end
|
|---|
| 162 | c-----------------------------------------------------------------------
|
|---|
| 163 | subroutine vec_dssum(u,v,w,nx,ny,nz)
|
|---|
| 164 | c
|
|---|
| 165 | c Direct stiffness summation of the face data, for field U.
|
|---|
| 166 | c
|
|---|
| 167 | c Boundary condition data corresponds to component IFIELD of
|
|---|
| 168 | c the CBC array.
|
|---|
| 169 | c
|
|---|
| 170 | INCLUDE 'SIZE'
|
|---|
| 171 | INCLUDE 'TOPOL'
|
|---|
| 172 | INCLUDE 'INPUT'
|
|---|
| 173 | INCLUDE 'PARALLEL'
|
|---|
| 174 | INCLUDE 'TSTEP'
|
|---|
| 175 | include 'CTIMER'
|
|---|
| 176 |
|
|---|
| 177 | REAL U(1),V(1),W(1)
|
|---|
| 178 |
|
|---|
| 179 | if(ifsync) call nekgsync()
|
|---|
| 180 |
|
|---|
| 181 | #ifdef TIMER
|
|---|
| 182 | if (icalld.eq.0) tvdss=0.0d0
|
|---|
| 183 | if (icalld.eq.0) tgsum=0.0d0
|
|---|
| 184 | icalld=icalld+1
|
|---|
| 185 | nvdss=icalld
|
|---|
| 186 | etime1=dnekclock()
|
|---|
| 187 | #endif
|
|---|
| 188 |
|
|---|
| 189 | c
|
|---|
| 190 | c============================================================================
|
|---|
| 191 | c execution phase
|
|---|
| 192 | c============================================================================
|
|---|
| 193 | c
|
|---|
| 194 | ifldt = ifield
|
|---|
| 195 | c if (ifldt.eq.0) ifldt = 1
|
|---|
| 196 | if (ifldt.eq.ifldmhd) ifldt = 1
|
|---|
| 197 |
|
|---|
| 198 | call fgslib_gs_op_many(gsh_fld(ifldt),u,v,w,u,u,u,ldim,1,1,0)
|
|---|
| 199 |
|
|---|
| 200 | #ifdef TIMER
|
|---|
| 201 | timee=(dnekclock()-etime1)
|
|---|
| 202 | tvdss=tvdss+timee
|
|---|
| 203 | tdsmx=max(timee,tdsmx)
|
|---|
| 204 | tdsmn=min(timee,tdsmn)
|
|---|
| 205 | #endif
|
|---|
| 206 |
|
|---|
| 207 | return
|
|---|
| 208 | end
|
|---|
| 209 |
|
|---|
| 210 | c-----------------------------------------------------------------------
|
|---|
| 211 | subroutine vec_dsop(u,v,w,nx,ny,nz,op)
|
|---|
| 212 | c
|
|---|
| 213 | c Direct stiffness summation of the face data, for field U.
|
|---|
| 214 | c
|
|---|
| 215 | c Boundary condition data corresponds to component IFIELD of
|
|---|
| 216 | c the CBC array.
|
|---|
| 217 | c
|
|---|
| 218 | INCLUDE 'SIZE'
|
|---|
| 219 | INCLUDE 'TOPOL'
|
|---|
| 220 | INCLUDE 'INPUT'
|
|---|
| 221 | INCLUDE 'PARALLEL'
|
|---|
| 222 | INCLUDE 'TSTEP'
|
|---|
| 223 | include 'CTIMER'
|
|---|
| 224 | c
|
|---|
| 225 | real u(1),v(1),w(1)
|
|---|
| 226 | character*3 op
|
|---|
| 227 |
|
|---|
| 228 | c============================================================================
|
|---|
| 229 | c execution phase
|
|---|
| 230 | c============================================================================
|
|---|
| 231 |
|
|---|
| 232 | ifldt = ifield
|
|---|
| 233 | c if (ifldt.eq.0) ifldt = 1
|
|---|
| 234 | if (ifldt.eq.ifldmhd) ifldt = 1
|
|---|
| 235 |
|
|---|
| 236 | c write(6,*) 'opdsop: ',op,ifldt,ifield
|
|---|
| 237 | if(ifsync) call nekgsync()
|
|---|
| 238 |
|
|---|
| 239 | if (op.eq.'+ ' .or. op.eq.'sum' .or. op.eq.'SUM')
|
|---|
| 240 | $ call fgslib_gs_op_many(gsh_fld(ifldt),u,v,w,u,u,u,ldim,1,1,0)
|
|---|
| 241 |
|
|---|
| 242 |
|
|---|
| 243 | if (op.eq.'* ' .or. op.eq.'mul' .or. op.eq.'MUL')
|
|---|
| 244 | $ call fgslib_gs_op_many(gsh_fld(ifldt),u,v,w,u,u,u,ldim,1,2,0)
|
|---|
| 245 |
|
|---|
| 246 |
|
|---|
| 247 | if (op.eq.'m ' .or. op.eq.'min' .or. op.eq.'mna'
|
|---|
| 248 | $ .or. op.eq.'MIN' .or. op.eq.'MNA')
|
|---|
| 249 | $ call fgslib_gs_op_many(gsh_fld(ifldt),u,v,w,u,u,u,ldim,1,3,0)
|
|---|
| 250 |
|
|---|
| 251 |
|
|---|
| 252 | if (op.eq.'M ' .or. op.eq.'max' .or. op.eq.'mxa'
|
|---|
| 253 | $ .or. op.eq.'MAX' .or. op.eq.'MXA')
|
|---|
| 254 | $ call fgslib_gs_op_many(gsh_fld(ifldt),u,v,w,u,u,u,ldim,1,4,0)
|
|---|
| 255 |
|
|---|
| 256 |
|
|---|
| 257 | return
|
|---|
| 258 | end
|
|---|
| 259 | c-----------------------------------------------------------------------
|
|---|
| 260 | subroutine nvec_dssum(u,stride,n,gs_handle)
|
|---|
| 261 |
|
|---|
| 262 | c Direct stiffness summation of the array u for n fields
|
|---|
| 263 | c
|
|---|
| 264 | include 'SIZE'
|
|---|
| 265 | include 'CTIMER'
|
|---|
| 266 |
|
|---|
| 267 | real u(1)
|
|---|
| 268 | integer n,stride,gs_handle
|
|---|
| 269 |
|
|---|
| 270 | if(ifsync) call nekgsync()
|
|---|
| 271 |
|
|---|
| 272 | #ifdef TIMER
|
|---|
| 273 | icalld=icalld+1
|
|---|
| 274 | nvdss=icalld
|
|---|
| 275 | etime1=dnekclock()
|
|---|
| 276 | #endif
|
|---|
| 277 | call fgslib_gs_op_fields(gs_handle,u,stride,n,1,1,0)
|
|---|
| 278 |
|
|---|
| 279 | #ifdef TIMER
|
|---|
| 280 | timee=(dnekclock()-etime1)
|
|---|
| 281 | tvdss=tvdss+timee
|
|---|
| 282 | tdsmx=max(timee,tdsmx)
|
|---|
| 283 | tdsmn=min(timee,tdsmn)
|
|---|
| 284 | #endif
|
|---|
| 285 |
|
|---|
| 286 | return
|
|---|
| 287 | end
|
|---|
| 288 |
|
|---|
| 289 | c----------------------------------------------------------------------
|
|---|
| 290 | subroutine matvec3(uout,Jmat,uin,iftrsp,n1,n2)
|
|---|
| 291 | c
|
|---|
| 292 | include 'SIZE'
|
|---|
| 293 | c
|
|---|
| 294 | real Jmat (n1,n1,2)
|
|---|
| 295 | real uin (1)
|
|---|
| 296 | real uout (1)
|
|---|
| 297 | logical iftrsp
|
|---|
| 298 | c
|
|---|
| 299 | common /matvtmp/ utmp(lx1,ly1)
|
|---|
| 300 | c
|
|---|
| 301 | if (ldim.eq.2) then
|
|---|
| 302 | call mxm (Jmat(1,1,1),n1,uin,n1,uout,n2)
|
|---|
| 303 | else
|
|---|
| 304 | if (iftrsp) then
|
|---|
| 305 | call transpose(uout,n2,uin,n1)
|
|---|
| 306 | else
|
|---|
| 307 | call copy (uout,uin,n1*n2)
|
|---|
| 308 | endif
|
|---|
| 309 | call mxm (Jmat(1,1,1),n1,uout,n1,utmp,n2)
|
|---|
| 310 | call mxm (utmp,n2,Jmat(1,1,2),n1,uout,n1)
|
|---|
| 311 | endif
|
|---|
| 312 | c
|
|---|
| 313 | return
|
|---|
| 314 | end
|
|---|
| 315 | c-----------------------------------------------------------------------
|
|---|
| 316 | subroutine matvec3t(uout,Jmat,uin,iftrsp,n1,n2)
|
|---|
| 317 | c
|
|---|
| 318 | include 'SIZE'
|
|---|
| 319 | c
|
|---|
| 320 | real Jmat (n1,n1,2)
|
|---|
| 321 | real uin (n1,n2)
|
|---|
| 322 | real uout (n1,n2)
|
|---|
| 323 | logical iftrsp
|
|---|
| 324 | c
|
|---|
| 325 | common /matvtmp/ utmp(lx1*ly1)
|
|---|
| 326 | c
|
|---|
| 327 | call transpose(utmp,n2,uin,n1)
|
|---|
| 328 | call mxm (Jmat(1,1,2),n1,utmp,n1,uout,n2)
|
|---|
| 329 | call mxm (uout,n2,Jmat(1,1,1),n1,utmp,n1)
|
|---|
| 330 | if (iftrsp) then
|
|---|
| 331 | call copy (uout,utmp,n1*n2)
|
|---|
| 332 | else
|
|---|
| 333 | call transpose(uout,n2,utmp,n1)
|
|---|
| 334 | endif
|
|---|
| 335 | c
|
|---|
| 336 | return
|
|---|
| 337 | end
|
|---|
| 338 | c-----------------------------------------------------------------------
|
|---|
| 339 | subroutine matvect (out,d,vec,n1,n2)
|
|---|
| 340 | dimension d(n1,n2),out(1),vec(1)
|
|---|
| 341 | c
|
|---|
| 342 | c handle non-square matrix in mat-vec mult -- TRANSPOSE
|
|---|
| 343 | c N1 is still the number of rows
|
|---|
| 344 | c N2 is still the number of cols
|
|---|
| 345 | c
|
|---|
| 346 | c
|
|---|
| 347 | call mxm(vec,1,d,n1,out,n2)
|
|---|
| 348 | c
|
|---|
| 349 | return
|
|---|
| 350 | end
|
|---|
| 351 | c-----------------------------------------------------------------------
|
|---|
| 352 | c subroutine opq_in_place(a,b,c)
|
|---|
| 353 | c include 'SIZE'
|
|---|
| 354 | c real a(1),b(1),c(1)
|
|---|
| 355 | c
|
|---|
| 356 | c call q_in_place(a)
|
|---|
| 357 | c call q_in_place(b)
|
|---|
| 358 | c if (ldim .eq.3) call q_in_place(c)
|
|---|
| 359 | c
|
|---|
| 360 | c return
|
|---|
| 361 | c end
|
|---|
| 362 | c-----------------------------------------------------------------------
|
|---|
| 363 | subroutine vectof_add(b,a,ie,iface,nx,ny,nz)
|
|---|
| 364 | C
|
|---|
| 365 | C Copy vector A to the face (IFACE) of B
|
|---|
| 366 | C IFACE is the input in the pre-processor ordering scheme.
|
|---|
| 367 | C
|
|---|
| 368 | DIMENSION A(NX,NY)
|
|---|
| 369 | DIMENSION B(NX,NY,NZ,1)
|
|---|
| 370 | CALL FACIND (KX1,KX2,KY1,KY2,KZ1,KZ2,NX,NY,NZ,IFACE)
|
|---|
| 371 | k = 0
|
|---|
| 372 | DO 100 IZ=KZ1,KZ2
|
|---|
| 373 | DO 100 IY=KY1,KY2
|
|---|
| 374 | DO 100 IX=KX1,KX2
|
|---|
| 375 | k = k + 1
|
|---|
| 376 | B(IX,IY,IZ,IE) = B(IX,IY,IZ,IE) + A(k,1)
|
|---|
| 377 | 100 CONTINUE
|
|---|
| 378 | return
|
|---|
| 379 | END
|
|---|
| 380 | c-----------------------------------------------------------------------
|
|---|
| 381 | subroutine zero_f(b,ie,iface,nx,ny,nz)
|
|---|
| 382 | C
|
|---|
| 383 | C ZERO the face (IFACE) of B
|
|---|
| 384 | C IFACE is the input in the pre-processor ordering scheme.
|
|---|
| 385 | C
|
|---|
| 386 | DIMENSION B(NX,NY,NZ,1)
|
|---|
| 387 | CALL FACIND (KX1,KX2,KY1,KY2,KZ1,KZ2,NX,NY,NZ,IFACE)
|
|---|
| 388 | c
|
|---|
| 389 | DO 100 IZ=KZ1,KZ2
|
|---|
| 390 | DO 100 IY=KY1,KY2
|
|---|
| 391 | DO 100 IX=KX1,KX2
|
|---|
| 392 | B(IX,IY,IZ,IE) = 0.
|
|---|
| 393 | 100 CONTINUE
|
|---|
| 394 | return
|
|---|
| 395 | END
|
|---|
| 396 | c-----------------------------------------------------------------------
|
|---|
| 397 | subroutine ftovec_0(a,b,ie,iface,nx,ny,nz)
|
|---|
| 398 | C
|
|---|
| 399 | C Copy the face (IFACE) of B to vector A.
|
|---|
| 400 | C IFACE is the input in the pre-processor ordering scheme.
|
|---|
| 401 | C
|
|---|
| 402 | DIMENSION A(NX,NY)
|
|---|
| 403 | DIMENSION B(NX,NY,NZ,1)
|
|---|
| 404 | CALL FACIND (KX1,KX2,KY1,KY2,KZ1,KZ2,NX,NY,NZ,IFACE)
|
|---|
| 405 | k = 0
|
|---|
| 406 | DO 100 IZ=KZ1,KZ2
|
|---|
| 407 | DO 100 IY=KY1,KY2
|
|---|
| 408 | DO 100 IX=KX1,KX2
|
|---|
| 409 | k = k + 1
|
|---|
| 410 | A(k,1)=B(IX,IY,IZ,IE)
|
|---|
| 411 | B(IX,IY,IZ,IE)=0.0
|
|---|
| 412 | 100 CONTINUE
|
|---|
| 413 | return
|
|---|
| 414 | END
|
|---|
| 415 | c-----------------------------------------------------------------------
|
|---|
| 416 | subroutine ftovec(a,b,ie,iface,nx,ny,nz)
|
|---|
| 417 | C
|
|---|
| 418 | C Copy the face (IFACE) of B to vector A.
|
|---|
| 419 | C IFACE is the input in the pre-processor ordering scheme.
|
|---|
| 420 | C
|
|---|
| 421 | real A(NX,NY)
|
|---|
| 422 | real B(NX,NY,NZ,1)
|
|---|
| 423 | CALL FACIND (KX1,KX2,KY1,KY2,KZ1,KZ2,NX,NY,NZ,IFACE)
|
|---|
| 424 | k = 0
|
|---|
| 425 | DO 100 IZ=KZ1,KZ2
|
|---|
| 426 | DO 100 IY=KY1,KY2
|
|---|
| 427 | DO 100 IX=KX1,KX2
|
|---|
| 428 | k = k + 1
|
|---|
| 429 | A(k,1)=B(IX,IY,IZ,IE)
|
|---|
| 430 | 100 CONTINUE
|
|---|
| 431 | return
|
|---|
| 432 | END
|
|---|
| 433 | c-----------------------------------------------------------------------
|
|---|
| 434 | subroutine vectof(b,a,ie,iface,nx,ny,nz)
|
|---|
| 435 | C
|
|---|
| 436 | C Copy vector A to the face (IFACE) of B
|
|---|
| 437 | C IFACE is the input in the pre-processor ordering scheme.
|
|---|
| 438 | C
|
|---|
| 439 | real A(NX,NY)
|
|---|
| 440 | real B(NX,NY,NZ,1)
|
|---|
| 441 | CALL FACIND (KX1,KX2,KY1,KY2,KZ1,KZ2,NX,NY,NZ,IFACE)
|
|---|
| 442 | k = 0
|
|---|
| 443 | DO 100 IZ=KZ1,KZ2
|
|---|
| 444 | DO 100 IY=KY1,KY2
|
|---|
| 445 | DO 100 IX=KX1,KX2
|
|---|
| 446 | k = k + 1
|
|---|
| 447 | B(IX,IY,IZ,IE) = A(k,1)
|
|---|
| 448 | 100 CONTINUE
|
|---|
| 449 | return
|
|---|
| 450 | END
|
|---|
| 451 | c-----------------------------------------------------------------------
|
|---|
| 452 | subroutine ftoveci(a,b,ie,iface,nx,ny,nz)
|
|---|
| 453 | C
|
|---|
| 454 | C Copy the face (IFACE) of B to vector A.
|
|---|
| 455 | C IFACE is the input in the pre-processor ordering scheme.
|
|---|
| 456 | C
|
|---|
| 457 | integer A(NX,NY)
|
|---|
| 458 | integer B(NX,NY,NZ,1)
|
|---|
| 459 | CALL FACIND (KX1,KX2,KY1,KY2,KZ1,KZ2,NX,NY,NZ,IFACE)
|
|---|
| 460 | k = 0
|
|---|
| 461 | DO 100 IZ=KZ1,KZ2
|
|---|
| 462 | DO 100 IY=KY1,KY2
|
|---|
| 463 | DO 100 IX=KX1,KX2
|
|---|
| 464 | k = k + 1
|
|---|
| 465 | A(k,1)=B(IX,IY,IZ,IE)
|
|---|
| 466 | 100 CONTINUE
|
|---|
| 467 | return
|
|---|
| 468 | END
|
|---|
| 469 | c-----------------------------------------------------------------------
|
|---|
| 470 | subroutine vectofi(b,a,ie,iface,nx,ny,nz)
|
|---|
| 471 | C
|
|---|
| 472 | C Copy vector A to the face (IFACE) of B
|
|---|
| 473 | C IFACE is the input in the pre-processor ordering scheme.
|
|---|
| 474 | C
|
|---|
| 475 | integer A(NX,NY)
|
|---|
| 476 | integer B(NX,NY,NZ,1)
|
|---|
| 477 | CALL FACIND (KX1,KX2,KY1,KY2,KZ1,KZ2,NX,NY,NZ,IFACE)
|
|---|
| 478 | k = 0
|
|---|
| 479 | DO 100 IZ=KZ1,KZ2
|
|---|
| 480 | DO 100 IY=KY1,KY2
|
|---|
| 481 | DO 100 IX=KX1,KX2
|
|---|
| 482 | k = k + 1
|
|---|
| 483 | B(IX,IY,IZ,IE) = A(k,1)
|
|---|
| 484 | 100 CONTINUE
|
|---|
| 485 | return
|
|---|
| 486 | END
|
|---|
| 487 | c-----------------------------------------------------------------------
|
|---|
| 488 | subroutine apply_Jt(u,nx,ny,nz,nel)
|
|---|
| 489 | include 'SIZE'
|
|---|
| 490 | include 'CTIMER'
|
|---|
| 491 | include 'INPUT'
|
|---|
| 492 | include 'NONCON'
|
|---|
| 493 | include 'PARALLEL'
|
|---|
| 494 | include 'TSTEP'
|
|---|
| 495 | real u(1)
|
|---|
| 496 | c
|
|---|
| 497 | parameter (lface=lx1*ly1)
|
|---|
| 498 | common /nonctmp/ uin(lface,2*ldim),uout(lface)
|
|---|
| 499 | c
|
|---|
| 500 | c
|
|---|
| 501 | c T
|
|---|
| 502 | c This is the J part, translating child data
|
|---|
| 503 | c
|
|---|
| 504 | do ie = 1 , nel
|
|---|
| 505 | c Note, we zero out u() on this face after extracting, for
|
|---|
| 506 | c consistency reasons discovered during Jerry's thesis.
|
|---|
| 507 | c Thus, "ftovec_0" rather than ftovec(). (iface -- Ed notation)
|
|---|
| 508 | do iface = 1 , 2*ldim
|
|---|
| 509 | im = mortar(iface,ie)
|
|---|
| 510 | if (im.ne.0) then
|
|---|
| 511 | call ftovec_0(uin(1,iface),u,ie,iface,nx,ny,nz)
|
|---|
| 512 | endif
|
|---|
| 513 | enddo
|
|---|
| 514 | do iface=1,2*ldim
|
|---|
| 515 | im = mortar(iface,ie)
|
|---|
| 516 | if (im.ne.0) then
|
|---|
| 517 | if (if3d) then
|
|---|
| 518 | call matvec3t
|
|---|
| 519 | $ (uout,Jmat(1,1,1,im),uin(1,iface),ifJt(im),nx,nx)
|
|---|
| 520 | else
|
|---|
| 521 | call matvect (uout,Jmat(1,1,1,im),uin(1,iface),nx,nx)
|
|---|
| 522 | endif
|
|---|
| 523 | call vectof_add(u,uout,ie,iface,nx,ny,nz)
|
|---|
| 524 | endif
|
|---|
| 525 | enddo
|
|---|
| 526 | enddo
|
|---|
| 527 | c
|
|---|
| 528 | return
|
|---|
| 529 | end
|
|---|
| 530 | c-----------------------------------------------------------------------
|
|---|
| 531 | subroutine apply_J(u,nx,ny,nz,nel)
|
|---|
| 532 | include 'SIZE'
|
|---|
| 533 | include 'CTIMER'
|
|---|
| 534 | include 'INPUT'
|
|---|
| 535 | include 'NONCON'
|
|---|
| 536 | include 'PARALLEL'
|
|---|
| 537 | include 'TSTEP'
|
|---|
| 538 | real u(1)
|
|---|
| 539 | c
|
|---|
| 540 | parameter (lface=lx1*ly1)
|
|---|
| 541 | common /nonctmp/ uin(lface,2*ldim),uout(lface)
|
|---|
| 542 | c
|
|---|
| 543 | c This is the J part, interpolating parent solution onto child
|
|---|
| 544 | c
|
|---|
| 545 | c
|
|---|
| 546 | do ie = 1 , nel
|
|---|
| 547 | do iface = 1 , 2*ldim
|
|---|
| 548 | im = mortar(iface,ie)
|
|---|
| 549 | if (im.ne.0) then
|
|---|
| 550 | call ftovec(uin(1,iface),u,ie,iface,nx,ny,nz)
|
|---|
| 551 | endif
|
|---|
| 552 | enddo
|
|---|
| 553 | do iface=1,2*ldim
|
|---|
| 554 | im = mortar(iface,ie)
|
|---|
| 555 | if (im.ne.0) then
|
|---|
| 556 | call matvec3
|
|---|
| 557 | $ (uout,Jmat(1,1,1,im),uin(1,iface),ifJt(im),nx,nz)
|
|---|
| 558 | call vectof (u,uout,ie,iface,nx,ny,nz)
|
|---|
| 559 | endif
|
|---|
| 560 | enddo
|
|---|
| 561 | enddo
|
|---|
| 562 | c
|
|---|
| 563 | return
|
|---|
| 564 | end
|
|---|
| 565 | c-----------------------------------------------------------------------
|
|---|
| 566 | subroutine h1_proj(u,nx,ny,nz)
|
|---|
| 567 | include 'SIZE'
|
|---|
| 568 | include 'CTIMER'
|
|---|
| 569 | include 'INPUT'
|
|---|
| 570 | include 'NONCON'
|
|---|
| 571 | include 'PARALLEL'
|
|---|
| 572 | include 'TSTEP'
|
|---|
| 573 | real u(1)
|
|---|
| 574 | c
|
|---|
| 575 | parameter (lface=lx1*ly1)
|
|---|
| 576 | common /nonctmp/ uin(lface,2*ldim),uout(lface)
|
|---|
| 577 |
|
|---|
| 578 | if(ifsync) call nekgsync()
|
|---|
| 579 |
|
|---|
| 580 | #ifdef TIMER
|
|---|
| 581 | if (icalld.eq.0) then
|
|---|
| 582 | tdsmx=0.
|
|---|
| 583 | tdsmn=0.
|
|---|
| 584 | endif
|
|---|
| 585 | icalld=icalld+1
|
|---|
| 586 | etime1=dnekclock()
|
|---|
| 587 | #endif
|
|---|
| 588 | c
|
|---|
| 589 | ifldt = ifield
|
|---|
| 590 | c if (ifldt.eq.0) ifldt = 1
|
|---|
| 591 | nel = nelv
|
|---|
| 592 | if (ifield.ge.2) nel=nelt
|
|---|
| 593 | ntot = nx*ny*nz*nel
|
|---|
| 594 |
|
|---|
| 595 |
|
|---|
| 596 | c
|
|---|
| 597 | c
|
|---|
| 598 | c ~ ~T
|
|---|
| 599 | c Implement := J Q Q Mu
|
|---|
| 600 | c
|
|---|
| 601 | c
|
|---|
| 602 | c T
|
|---|
| 603 | c
|
|---|
| 604 | call col2 (u,umult,ntot)
|
|---|
| 605 | c
|
|---|
| 606 | c ~ ~T
|
|---|
| 607 | c This is the Q Q part
|
|---|
| 608 | c
|
|---|
| 609 | call fgslib_gs_op(gsh_fld(ifldt),u,1,1,0)
|
|---|
| 610 | c
|
|---|
| 611 | c
|
|---|
| 612 | c This is the J part, interpolating parent solution onto child
|
|---|
| 613 | c
|
|---|
| 614 | call apply_J(u,nx,ny,nz,nel)
|
|---|
| 615 | c
|
|---|
| 616 | c
|
|---|
| 617 | #ifdef TIMER
|
|---|
| 618 | timee=(dnekclock()-etime1)
|
|---|
| 619 | tdsum=tdsum+timee
|
|---|
| 620 | ndsum=icalld
|
|---|
| 621 | tdsmx=max(timee,tdsmx)
|
|---|
| 622 | tdsmn=min(timee,tdsmn)
|
|---|
| 623 | #endif
|
|---|
| 624 | c
|
|---|
| 625 | return
|
|---|
| 626 | end
|
|---|
| 627 | c-----------------------------------------------------------------------
|
|---|
| 628 | subroutine dssum_msk(u,mask,nx,ny,nz)
|
|---|
| 629 | include 'SIZE'
|
|---|
| 630 | include 'CTIMER'
|
|---|
| 631 | include 'INPUT'
|
|---|
| 632 | include 'NONCON'
|
|---|
| 633 | include 'PARALLEL'
|
|---|
| 634 | include 'TSTEP'
|
|---|
| 635 | real u(1),mask(1)
|
|---|
| 636 | c
|
|---|
| 637 | parameter (lface=lx1*ly1)
|
|---|
| 638 | common /nonctmp/ uin(lface,2*ldim),uout(lface)
|
|---|
| 639 |
|
|---|
| 640 | if(ifsync) call nekgsync()
|
|---|
| 641 |
|
|---|
| 642 | #ifdef TIMER
|
|---|
| 643 | if (icalld.eq.0) then
|
|---|
| 644 | tdsmx=0.
|
|---|
| 645 | tdsmn=0.
|
|---|
| 646 | endif
|
|---|
| 647 | icalld=icalld+1
|
|---|
| 648 | etime1=dnekclock()
|
|---|
| 649 | #endif
|
|---|
| 650 | c
|
|---|
| 651 | ifldt = ifield
|
|---|
| 652 | c if (ifldt.eq.0) ifldt = 1
|
|---|
| 653 | nel = nelv
|
|---|
| 654 | if (ifield.ge.2) nel=nelt
|
|---|
| 655 | ntot = nx*ny*nz*nel
|
|---|
| 656 |
|
|---|
| 657 |
|
|---|
| 658 | c
|
|---|
| 659 | c T ~ ~T T
|
|---|
| 660 | c Implement Q M Q := J M Q Q J
|
|---|
| 661 | c
|
|---|
| 662 | c
|
|---|
| 663 | c T
|
|---|
| 664 | c This is the J part, translating child data
|
|---|
| 665 | c
|
|---|
| 666 | call apply_Jt(u,nx,ny,nz,nel)
|
|---|
| 667 | c
|
|---|
| 668 | c
|
|---|
| 669 | c
|
|---|
| 670 | c ~ ~T
|
|---|
| 671 | c This is the Q Q part
|
|---|
| 672 | c
|
|---|
| 673 | call fgslib_gs_op(gsh_fld(ifldt),u,1,1,0)
|
|---|
| 674 | call col2 (u,mask,ntot)
|
|---|
| 675 | c
|
|---|
| 676 | c
|
|---|
| 677 | c This is the J part, interpolating parent solution onto child
|
|---|
| 678 | c
|
|---|
| 679 | call apply_J(u,nx,ny,nz,nel)
|
|---|
| 680 | c
|
|---|
| 681 | c
|
|---|
| 682 | #ifdef TIMER
|
|---|
| 683 | timee=(dnekclock()-etime1)
|
|---|
| 684 | tdsum=tdsum+timee
|
|---|
| 685 | ndsum=icalld
|
|---|
| 686 | tdsmx=max(timee,tdsmx)
|
|---|
| 687 | tdsmn=min(timee,tdsmn)
|
|---|
| 688 | #endif
|
|---|
| 689 | c
|
|---|
| 690 | return
|
|---|
| 691 | end
|
|---|
| 692 | c-----------------------------------------------------------------------
|
|---|
| 693 | subroutine dssum_msk2(u,mask,binv,nx,ny,nz)
|
|---|
| 694 | include 'SIZE'
|
|---|
| 695 | include 'CTIMER'
|
|---|
| 696 | include 'INPUT'
|
|---|
| 697 | include 'NONCON'
|
|---|
| 698 | include 'PARALLEL'
|
|---|
| 699 | include 'TSTEP'
|
|---|
| 700 | real u(1),mask(1),binv(1)
|
|---|
| 701 | c
|
|---|
| 702 | parameter (lface=lx1*ly1)
|
|---|
| 703 | common /nonctmp/ uin(lface,2*ldim),uout(lface)
|
|---|
| 704 |
|
|---|
| 705 | if(ifsync) call nekgsync()
|
|---|
| 706 |
|
|---|
| 707 | #ifdef TIMER
|
|---|
| 708 | if (icalld.eq.0) then
|
|---|
| 709 | tdsmx=0.
|
|---|
| 710 | tdsmn=0.
|
|---|
| 711 | endif
|
|---|
| 712 | icalld=icalld+1
|
|---|
| 713 | etime1=dnekclock()
|
|---|
| 714 | #endif
|
|---|
| 715 |
|
|---|
| 716 | c
|
|---|
| 717 | ifldt = ifield
|
|---|
| 718 | c if (ifldt.eq.0) ifldt = 1
|
|---|
| 719 | nel = nelv
|
|---|
| 720 | if (ifield.ge.2) nel=nelt
|
|---|
| 721 | ntot = nx*ny*nz*nel
|
|---|
| 722 |
|
|---|
| 723 |
|
|---|
| 724 | c
|
|---|
| 725 | c
|
|---|
| 726 | c T ~ ~T T
|
|---|
| 727 | c Implement Q M Q := J M Q Q J
|
|---|
| 728 | c
|
|---|
| 729 | c
|
|---|
| 730 | c T
|
|---|
| 731 | c This is the J part, translating child data
|
|---|
| 732 | c
|
|---|
| 733 | call apply_Jt(u,nx,ny,nz,nel)
|
|---|
| 734 | c
|
|---|
| 735 | c
|
|---|
| 736 | c
|
|---|
| 737 | c ~ ~T
|
|---|
| 738 | c This is the Q Q part
|
|---|
| 739 | c
|
|---|
| 740 | call fgslib_gs_op(gsh_fld(ifldt),u,1,1,0)
|
|---|
| 741 | call col3 (u,mask,binv,ntot)
|
|---|
| 742 | c
|
|---|
| 743 | c
|
|---|
| 744 | c This is the J part, interpolating parent solution onto child
|
|---|
| 745 | c
|
|---|
| 746 | call apply_J(u,nx,ny,nz,nel)
|
|---|
| 747 | c
|
|---|
| 748 | c
|
|---|
| 749 | #ifdef TIMER
|
|---|
| 750 | timee=(dnekclock()-etime1)
|
|---|
| 751 | tdsum=tdsum+timee
|
|---|
| 752 | ndsum=icalld
|
|---|
| 753 | tdsmx=max(timee,tdsmx)
|
|---|
| 754 | tdsmn=min(timee,tdsmn)
|
|---|
| 755 | #endif
|
|---|
| 756 | c
|
|---|
| 757 | return
|
|---|
| 758 | end
|
|---|
| 759 | c-----------------------------------------------------------------------
|
|---|
| 760 | subroutine gtpp_gs_op(u,op,hndl)
|
|---|
| 761 | c
|
|---|
| 762 | c gather-scatter operation across global tensor product planes
|
|---|
| 763 | c
|
|---|
| 764 | include 'SIZE'
|
|---|
| 765 | include 'TOTAL'
|
|---|
| 766 |
|
|---|
| 767 | real u(*)
|
|---|
| 768 | character*3 op
|
|---|
| 769 | integer hndl
|
|---|
| 770 |
|
|---|
| 771 | if (op.eq.'+ ' .or. op.eq.'sum' .or. op.eq.'SUM') then
|
|---|
| 772 | call fgslib_gs_op(hndl,u,1,1,0)
|
|---|
| 773 | elseif (op.eq.'* ' .or. op.eq.'mul' .or. op.eq.'MUL') then
|
|---|
| 774 | call fgslib_gs_op(hndl,u,1,2,0)
|
|---|
| 775 | elseif (op.eq.'m ' .or. op.eq.'min' .or. op.eq.'mna'
|
|---|
| 776 | & .or. op.eq.'MIN' .or. op.eq.'MNA') then
|
|---|
| 777 | call fgslib_gs_op(hndl,u,1,3,0)
|
|---|
| 778 | elseif (op.eq.'M ' .or. op.eq.'max' .or. op.eq.'mxa'
|
|---|
| 779 | & .or. op.eq.'MAX' .or. op.eq.'MXA') then
|
|---|
| 780 | call fgslib_gs_op(hndl,u,1,4,0)
|
|---|
| 781 | else
|
|---|
| 782 | call exitti('gtpp_gs_op: invalid operation!$',1)
|
|---|
| 783 | endif
|
|---|
| 784 |
|
|---|
| 785 | return
|
|---|
| 786 | end
|
|---|
| 787 | c-----------------------------------------------------------------------
|
|---|
| 788 | subroutine gtpp_gs_setup(hndl,nelgx,nelgy,nelgz,idir)
|
|---|
| 789 |
|
|---|
| 790 | include 'SIZE'
|
|---|
| 791 | include 'TOTAL'
|
|---|
| 792 |
|
|---|
| 793 | integer hndl,nelgx,nelgy,nelgz,idir
|
|---|
| 794 |
|
|---|
| 795 | common /nekmpi/ mid,mp,nekcomm,nekgroup,nekreal
|
|---|
| 796 | common /c_is1/ glo_num(lx1,ly1,lz1,lelv)
|
|---|
| 797 | integer e,ex,ey,ez,eg
|
|---|
| 798 | integer*8 glo_num,ex_g
|
|---|
| 799 |
|
|---|
| 800 | nelgxyz = nelgx*nelgy*nelgz
|
|---|
| 801 | if (nelgxyz .ne. nelgv)
|
|---|
| 802 | $ call exitti('gtpp_gs_setup: invalid gtp mesh dimensions!$',
|
|---|
| 803 | $ nelgxyz)
|
|---|
| 804 |
|
|---|
| 805 | nel = nelv
|
|---|
| 806 | nelgxy = nelgx*nelgy
|
|---|
| 807 | nelgyz = nelgy*nelgz
|
|---|
| 808 | nelgzx = nelgz*nelgx
|
|---|
| 809 |
|
|---|
| 810 | if (idir.eq.1) then
|
|---|
| 811 | ! x-direction
|
|---|
| 812 | do e=1,nel
|
|---|
| 813 | eg = lglel(e)
|
|---|
| 814 | call get_exyz(ex,ey,ez,eg,nelgx,nelgyz,1)
|
|---|
| 815 | ex_g = ey
|
|---|
| 816 | do k=1,nz1 ! Enumerate points in the y-z plane
|
|---|
| 817 | do j=1,ny1
|
|---|
| 818 | do i=1,nx1
|
|---|
| 819 | glo_num(i,j,k,e) = j+ny1*(k-1) + ny1*nz1*(ex_g-1)
|
|---|
| 820 | enddo
|
|---|
| 821 | enddo
|
|---|
| 822 | enddo
|
|---|
| 823 | enddo
|
|---|
| 824 | elseif (idir.eq.2) then
|
|---|
| 825 | ! y-direction
|
|---|
| 826 | do e=1,nel
|
|---|
| 827 | eg = lglel(e)
|
|---|
| 828 | call get_exyz(ex,ey,ez,eg,nelgx,nelgy,nelgz)
|
|---|
| 829 | ex_g = (ez-1)*nelgx+ex
|
|---|
| 830 | do k=1,nz1 ! Enumerate points in the x-z plane
|
|---|
| 831 | do j=1,ny1
|
|---|
| 832 | do i=1,nx1
|
|---|
| 833 | glo_num(i,j,k,e) = k+nz1*(i-1) + nx1*nz1*(ex_g-1)
|
|---|
| 834 | enddo
|
|---|
| 835 | enddo
|
|---|
| 836 | enddo
|
|---|
| 837 | enddo
|
|---|
| 838 | elseif (idir.eq.3) then
|
|---|
| 839 | ! z-direction
|
|---|
| 840 | do e=1,nel
|
|---|
| 841 | eg = lglel(e)
|
|---|
| 842 | call get_exyz(ex,ey,ez,eg,nelgxy,1,1)
|
|---|
| 843 | ex_g = ex
|
|---|
| 844 | do k=1,nz1 ! Enumerate points in the x-y plane
|
|---|
| 845 | do j=1,ny1
|
|---|
| 846 | do i=1,nx1
|
|---|
| 847 | glo_num(i,j,k,e) = i+nx1*(j-1) + nx1*ny1*(ex_g-1)
|
|---|
| 848 | enddo
|
|---|
| 849 | enddo
|
|---|
| 850 | enddo
|
|---|
| 851 | enddo
|
|---|
| 852 | endif
|
|---|
| 853 |
|
|---|
| 854 | n = nel*nx1*ny1*nz1
|
|---|
| 855 | call fgslib_gs_setup(hndl,glo_num,n,nekcomm,np)
|
|---|
| 856 |
|
|---|
| 857 | return
|
|---|
| 858 | end
|
|---|
| 859 | c-----------------------------------------------------------------------
|
|---|
| 860 | subroutine gs_setup_ms(hndl,nel,nx,ny,nz)
|
|---|
| 861 |
|
|---|
| 862 | include 'SIZE'
|
|---|
| 863 | include 'TOTAL'
|
|---|
| 864 | include 'mpif.h'
|
|---|
| 865 |
|
|---|
| 866 | integer hndl
|
|---|
| 867 | integer e,eg
|
|---|
| 868 |
|
|---|
| 869 | common /c_is1/ glo_num(lx1*ly1*lz1*lelt)
|
|---|
| 870 | integer*8 glo_num
|
|---|
| 871 |
|
|---|
| 872 | do e=1,nel
|
|---|
| 873 | eg = lglel(e)
|
|---|
| 874 | do k=1,nz
|
|---|
| 875 | do j=1,ny
|
|---|
| 876 | do i=1,nx
|
|---|
| 877 | ii = i + nx*(j-1) + nx*ny*(k-1) + nx*ny*nz*(e-1)
|
|---|
| 878 | glo_num(ii) = i + nx*(j-1) + nx*ny*(k-1) +
|
|---|
| 879 | $ nx*ny*nz*(eg-1)
|
|---|
| 880 | enddo
|
|---|
| 881 | enddo
|
|---|
| 882 | enddo
|
|---|
| 883 | enddo
|
|---|
| 884 |
|
|---|
| 885 | n = nel*nx*ny*nz
|
|---|
| 886 | call fgslib_gs_setup(hndl,glo_num,n,iglobalcomm,np_global)
|
|---|
| 887 |
|
|---|
| 888 | return
|
|---|
| 889 | end
|
|---|
| 890 | c-----------------------------------------------------------------------
|
|---|
| 891 | subroutine gs_op_ms(u,op,hndl)
|
|---|
| 892 | c
|
|---|
| 893 | c gather-scatter operation across sessions
|
|---|
| 894 | c
|
|---|
| 895 | include 'SIZE'
|
|---|
| 896 | include 'PARALLEL'
|
|---|
| 897 | include 'INPUT'
|
|---|
| 898 | include 'TSTEP'
|
|---|
| 899 | include 'CTIMER'
|
|---|
| 900 |
|
|---|
| 901 | real u(*)
|
|---|
| 902 | character*3 op
|
|---|
| 903 |
|
|---|
| 904 | if(ifsync) call nekgsync()
|
|---|
| 905 |
|
|---|
| 906 | if (op.eq.'+ ' .or. op.eq.'sum' .or. op.eq.'SUM') then
|
|---|
| 907 | call fgslib_gs_op(hndl,u,1,1,0)
|
|---|
| 908 | else if (op.eq.'* ' .or. op.eq.'mul' .or. op.eq.'MUL') then
|
|---|
| 909 | call fgslib_gs_op(hndl,u,1,2,0)
|
|---|
| 910 | else if (op.eq.'m ' .or. op.eq.'min' .or. op.eq.'mna'
|
|---|
| 911 | & .or. op.eq.'MIN' .or. op.eq.'MNA') then
|
|---|
| 912 | call fgslib_gs_op(hndl,u,1,3,0)
|
|---|
| 913 | else if (op.eq.'M ' .or. op.eq.'max' .or. op.eq.'mxa'
|
|---|
| 914 | & .or. op.eq.'MAX' .or. op.eq.'MXA') then
|
|---|
| 915 | call fgslib_gs_op(hndl,u,1,4,0)
|
|---|
| 916 | else
|
|---|
| 917 | call exitti('gs_op_ms: invalid operation!$',1)
|
|---|
| 918 | endif
|
|---|
| 919 |
|
|---|
| 920 | return
|
|---|
| 921 | end
|
|---|
