1 | SUBROUTINE transpose_xy( f_in, work, f_out ) |
---|
2 | |
---|
3 | !--------------------------------------------------------------------------------! |
---|
4 | ! This file is part of PALM. |
---|
5 | ! |
---|
6 | ! PALM is free software: you can redistribute it and/or modify it under the terms |
---|
7 | ! of the GNU General Public License as published by the Free Software Foundation, |
---|
8 | ! either version 3 of the License, or (at your option) any later version. |
---|
9 | ! |
---|
10 | ! PALM is distributed in the hope that it will be useful, but WITHOUT ANY |
---|
11 | ! WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR |
---|
12 | ! A PARTICULAR PURPOSE. See the GNU General Public License for more details. |
---|
13 | ! |
---|
14 | ! You should have received a copy of the GNU General Public License along with |
---|
15 | ! PALM. If not, see <http://www.gnu.org/licenses/>. |
---|
16 | ! |
---|
17 | ! Copyright 1997-2012 Leibniz University Hannover |
---|
18 | !--------------------------------------------------------------------------------! |
---|
19 | ! |
---|
20 | ! Current revisions: |
---|
21 | ! ----------------- |
---|
22 | ! openACC directives added, |
---|
23 | ! resorting data from/to work changed, work got 4 dimensions instead of 1 |
---|
24 | ! |
---|
25 | ! Former revisions: |
---|
26 | ! ----------------- |
---|
27 | ! $Id: transpose.f90 1111 2013-03-08 23:54:10Z raasch $ |
---|
28 | ! |
---|
29 | ! 1106 2013-03-04 05:31:38Z raasch |
---|
30 | ! preprocessor lines rearranged so that routines can also be used in serial |
---|
31 | ! (non-parallel) mode |
---|
32 | ! |
---|
33 | ! 1092 2013-02-02 11:24:22Z raasch |
---|
34 | ! unused variables removed |
---|
35 | ! |
---|
36 | ! 1036 2012-10-22 13:43:42Z raasch |
---|
37 | ! code put under GPL (PALM 3.9) |
---|
38 | ! |
---|
39 | ! 1003 2012-09-14 14:35:53Z raasch |
---|
40 | ! indices nxa, nya, etc. replaced by nx, ny, etc. |
---|
41 | ! |
---|
42 | ! 683 2011-02-09 14:25:15Z raasch |
---|
43 | ! openMP parallelization of transpositions for 2d-domain-decomposition |
---|
44 | ! |
---|
45 | ! 622 2010-12-10 08:08:13Z raasch |
---|
46 | ! optional barriers included in order to speed up collective operations |
---|
47 | ! |
---|
48 | ! 164 2008-05-15 08:46:15Z raasch |
---|
49 | ! f_inv changed from subroutine argument to automatic array in order to do |
---|
50 | ! re-ordering from f_in to f_inv in one step, one array work is needed instead |
---|
51 | ! of work1 and work2 |
---|
52 | ! |
---|
53 | ! February 2007 |
---|
54 | ! RCS Log replace by Id keyword, revision history cleaned up |
---|
55 | ! |
---|
56 | ! Revision 1.2 2004/04/30 13:12:17 raasch |
---|
57 | ! Switched from mpi_alltoallv to the simpler mpi_alltoall, |
---|
58 | ! all former transpose-routine files collected in this file, enlarged |
---|
59 | ! transposition arrays introduced |
---|
60 | ! |
---|
61 | ! Revision 1.1 2004/04/30 13:08:16 raasch |
---|
62 | ! Initial revision (collection of former routines transpose_xy, transpose_xz, |
---|
63 | ! transpose_yx, transpose_yz, transpose_zx, transpose_zy) |
---|
64 | ! |
---|
65 | ! Revision 1.1 1997/07/24 11:25:18 raasch |
---|
66 | ! Initial revision |
---|
67 | ! |
---|
68 | ! |
---|
69 | ! Description: |
---|
70 | ! ------------ |
---|
71 | ! Transposition of input array (f_in) from x to y. For the input array, all |
---|
72 | ! elements along x reside on the same PE, while after transposition, all |
---|
73 | ! elements along y reside on the same PE. |
---|
74 | !------------------------------------------------------------------------------! |
---|
75 | |
---|
76 | USE cpulog |
---|
77 | USE indices |
---|
78 | USE interfaces |
---|
79 | USE pegrid |
---|
80 | USE transpose_indices |
---|
81 | |
---|
82 | IMPLICIT NONE |
---|
83 | |
---|
84 | INTEGER :: i, j, k, l, ys |
---|
85 | |
---|
86 | REAL :: f_in(0:nx,nys_x:nyn_x,nzb_x:nzt_x), f_out(0:ny,nxl_y:nxr_y,nzb_y:nzt_y) |
---|
87 | |
---|
88 | REAL, DIMENSION(nyn_x-nys_x+1,nzb_y:nzt_y,nxl_y:nxr_y,0:pdims(2)-1) :: work |
---|
89 | |
---|
90 | !$acc declare create( f_inv ) |
---|
91 | REAL :: f_inv(nys_x:nyn_x,nzb_x:nzt_x,0:nx) |
---|
92 | |
---|
93 | |
---|
94 | ! |
---|
95 | !-- Rearrange indices of input array in order to make data to be send |
---|
96 | !-- by MPI contiguous |
---|
97 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
98 | !$OMP DO |
---|
99 | !$acc kernels present( f_in ) |
---|
100 | !$acc loop |
---|
101 | DO i = 0, nx |
---|
102 | DO k = nzb_x, nzt_x |
---|
103 | !$acc loop vector( 32 ) |
---|
104 | DO j = nys_x, nyn_x |
---|
105 | f_inv(j,k,i) = f_in(i,j,k) |
---|
106 | ENDDO |
---|
107 | ENDDO |
---|
108 | ENDDO |
---|
109 | !$acc end kernels |
---|
110 | !$OMP END PARALLEL |
---|
111 | |
---|
112 | IF ( numprocs /= 1 ) THEN |
---|
113 | |
---|
114 | #if defined( __parallel ) |
---|
115 | ! |
---|
116 | !-- Transpose array |
---|
117 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
---|
118 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
119 | !$acc update host( f_inv ) |
---|
120 | CALL MPI_ALLTOALL( f_inv(nys_x,nzb_x,0), sendrecvcount_xy, MPI_REAL, & |
---|
121 | work(1,nzb_y,nxl_y,0), sendrecvcount_xy, MPI_REAL, & |
---|
122 | comm1dy, ierr ) |
---|
123 | !$acc update device( work ) |
---|
124 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
---|
125 | |
---|
126 | ! |
---|
127 | !-- Reorder transposed array |
---|
128 | !$OMP PARALLEL PRIVATE ( i, j, k, l, ys ) |
---|
129 | !$OMP DO |
---|
130 | DO l = 0, pdims(2) - 1 |
---|
131 | ys = 0 + l * ( nyn_x - nys_x + 1 ) |
---|
132 | !$acc kernels present( f_out, work ) |
---|
133 | !$acc loop |
---|
134 | DO i = nxl_y, nxr_y |
---|
135 | DO k = nzb_y, nzt_y |
---|
136 | !$acc loop vector( 32 ) |
---|
137 | DO j = ys, ys + nyn_x - nys_x |
---|
138 | f_out(j,i,k) = work(j-ys+1,k,i,l) |
---|
139 | ENDDO |
---|
140 | ENDDO |
---|
141 | ENDDO |
---|
142 | !$acc end kernels |
---|
143 | ENDDO |
---|
144 | !$OMP END PARALLEL |
---|
145 | #endif |
---|
146 | |
---|
147 | ELSE |
---|
148 | |
---|
149 | ! |
---|
150 | !-- Reorder transposed array |
---|
151 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
152 | !$OMP DO |
---|
153 | !$acc kernels present( f_out ) |
---|
154 | !$acc loop |
---|
155 | DO k = nzb_y, nzt_y |
---|
156 | DO i = nxl_y, nxr_y |
---|
157 | !$acc loop vector( 32 ) |
---|
158 | DO j = 0, ny |
---|
159 | f_out(j,i,k) = f_inv(j,k,i) |
---|
160 | ENDDO |
---|
161 | ENDDO |
---|
162 | ENDDO |
---|
163 | !$acc end kernels |
---|
164 | !$OMP END PARALLEL |
---|
165 | |
---|
166 | ENDIF |
---|
167 | |
---|
168 | END SUBROUTINE transpose_xy |
---|
169 | |
---|
170 | |
---|
171 | SUBROUTINE transpose_xz( f_in, work, f_out ) |
---|
172 | |
---|
173 | !------------------------------------------------------------------------------! |
---|
174 | ! Description: |
---|
175 | ! ------------ |
---|
176 | ! Transposition of input array (f_in) from x to z. For the input array, all |
---|
177 | ! elements along x reside on the same PE, while after transposition, all |
---|
178 | ! elements along z reside on the same PE. |
---|
179 | !------------------------------------------------------------------------------! |
---|
180 | |
---|
181 | USE cpulog |
---|
182 | USE indices |
---|
183 | USE interfaces |
---|
184 | USE pegrid |
---|
185 | USE transpose_indices |
---|
186 | |
---|
187 | IMPLICIT NONE |
---|
188 | |
---|
189 | INTEGER :: i, j, k, l, xs |
---|
190 | |
---|
191 | REAL :: f_in(0:nx,nys_x:nyn_x,nzb_x:nzt_x), f_out(1:nz,nys:nyn,nxl:nxr) |
---|
192 | |
---|
193 | REAL, DIMENSION(nys_x:nyn_x,nnx,nzb_x:nzt_x,0:pdims(1)-1) :: work |
---|
194 | |
---|
195 | !$acc declare create( f_inv ) |
---|
196 | REAL :: f_inv(nys:nyn,nxl:nxr,1:nz) |
---|
197 | |
---|
198 | |
---|
199 | ! |
---|
200 | !-- If the PE grid is one-dimensional along y, the array has only to be |
---|
201 | !-- reordered locally and therefore no transposition has to be done. |
---|
202 | IF ( pdims(1) /= 1 ) THEN |
---|
203 | |
---|
204 | #if defined( __parallel ) |
---|
205 | ! |
---|
206 | !-- Reorder input array for transposition |
---|
207 | !$OMP PARALLEL PRIVATE ( i, j, k, l, xs ) |
---|
208 | !$OMP DO |
---|
209 | DO l = 0, pdims(1) - 1 |
---|
210 | xs = 0 + l * nnx |
---|
211 | !$acc kernels present( f_in, work ) |
---|
212 | !$acc loop |
---|
213 | DO k = nzb_x, nzt_x |
---|
214 | DO i = xs, xs + nnx - 1 |
---|
215 | !$acc loop vector( 32 ) |
---|
216 | DO j = nys_x, nyn_x |
---|
217 | work(j,i-xs+1,k,l) = f_in(i,j,k) |
---|
218 | ENDDO |
---|
219 | ENDDO |
---|
220 | ENDDO |
---|
221 | !$acc end kernels |
---|
222 | ENDDO |
---|
223 | !$OMP END PARALLEL |
---|
224 | |
---|
225 | ! |
---|
226 | !-- Transpose array |
---|
227 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
---|
228 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
229 | !$acc update host( work ) |
---|
230 | CALL MPI_ALLTOALL( work(nys_x,1,nzb_x,0), sendrecvcount_zx, MPI_REAL, & |
---|
231 | f_inv(nys,nxl,1), sendrecvcount_zx, MPI_REAL, & |
---|
232 | comm1dx, ierr ) |
---|
233 | !$acc update device( f_inv ) |
---|
234 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
---|
235 | |
---|
236 | ! |
---|
237 | !-- Reorder transposed array in a way that the z index is in first position |
---|
238 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
239 | !$OMP DO |
---|
240 | !$acc kernels present( f_out ) |
---|
241 | !$acc loop |
---|
242 | DO k = 1, nz |
---|
243 | DO i = nxl, nxr |
---|
244 | !$acc loop vector( 32 ) |
---|
245 | DO j = nys, nyn |
---|
246 | f_out(k,j,i) = f_inv(j,i,k) |
---|
247 | ENDDO |
---|
248 | ENDDO |
---|
249 | ENDDO |
---|
250 | !$acc end kernels |
---|
251 | !$OMP END PARALLEL |
---|
252 | #endif |
---|
253 | |
---|
254 | ELSE |
---|
255 | |
---|
256 | ! |
---|
257 | !-- Reorder the array in a way that the z index is in first position |
---|
258 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
259 | !$OMP DO |
---|
260 | !$acc kernels present( f_in ) |
---|
261 | !$acc loop |
---|
262 | DO i = nxl, nxr |
---|
263 | DO j = nys, nyn |
---|
264 | !$acc loop vector( 32 ) |
---|
265 | DO k = 1, nz |
---|
266 | f_inv(j,i,k) = f_in(i,j,k) |
---|
267 | ENDDO |
---|
268 | ENDDO |
---|
269 | ENDDO |
---|
270 | !$acc end kernels |
---|
271 | !$OMP END PARALLEL |
---|
272 | |
---|
273 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
274 | !$OMP DO |
---|
275 | !$acc kernels present( f_out ) |
---|
276 | !$acc loop |
---|
277 | DO k = 1, nz |
---|
278 | DO i = nxl, nxr |
---|
279 | !$acc loop vector( 32 ) |
---|
280 | DO j = nys, nyn |
---|
281 | f_out(k,j,i) = f_inv(j,i,k) |
---|
282 | ENDDO |
---|
283 | ENDDO |
---|
284 | ENDDO |
---|
285 | !$acc end kernels |
---|
286 | !$OMP END PARALLEL |
---|
287 | |
---|
288 | ENDIF |
---|
289 | |
---|
290 | END SUBROUTINE transpose_xz |
---|
291 | |
---|
292 | |
---|
293 | SUBROUTINE transpose_yx( f_in, work, f_out ) |
---|
294 | |
---|
295 | !------------------------------------------------------------------------------! |
---|
296 | ! Description: |
---|
297 | ! ------------ |
---|
298 | ! Transposition of input array (f_in) from y to x. For the input array, all |
---|
299 | ! elements along y reside on the same PE, while after transposition, all |
---|
300 | ! elements along x reside on the same PE. |
---|
301 | !------------------------------------------------------------------------------! |
---|
302 | |
---|
303 | USE cpulog |
---|
304 | USE indices |
---|
305 | USE interfaces |
---|
306 | USE pegrid |
---|
307 | USE transpose_indices |
---|
308 | |
---|
309 | IMPLICIT NONE |
---|
310 | |
---|
311 | INTEGER :: i, j, k, l, ys |
---|
312 | |
---|
313 | REAL :: f_in(0:ny,nxl_y:nxr_y,nzb_y:nzt_y), f_out(0:nx,nys_x:nyn_x,nzb_x:nzt_x) |
---|
314 | |
---|
315 | REAL, DIMENSION(nyn_x-nys_x+1,nzb_y:nzt_y,nxl_y:nxr_y,0:pdims(2)-1) :: work |
---|
316 | |
---|
317 | !$acc declare create( f_inv ) |
---|
318 | REAL :: f_inv(nys_x:nyn_x,nzb_x:nzt_x,0:nx) |
---|
319 | |
---|
320 | |
---|
321 | IF ( numprocs /= 1 ) THEN |
---|
322 | |
---|
323 | #if defined( __parallel ) |
---|
324 | ! |
---|
325 | !-- Reorder input array for transposition |
---|
326 | !$OMP PARALLEL PRIVATE ( i, j, k, l, ys ) |
---|
327 | !$OMP DO |
---|
328 | DO l = 0, pdims(2) - 1 |
---|
329 | ys = 0 + l * ( nyn_x - nys_x + 1 ) |
---|
330 | !$acc kernels present( f_in, work ) |
---|
331 | !$acc loop |
---|
332 | DO i = nxl_y, nxr_y |
---|
333 | DO k = nzb_y, nzt_y |
---|
334 | !$acc loop vector( 32 ) |
---|
335 | DO j = ys, ys + nyn_x - nys_x |
---|
336 | work(j-ys+1,k,i,l) = f_in(j,i,k) |
---|
337 | ENDDO |
---|
338 | ENDDO |
---|
339 | ENDDO |
---|
340 | !$acc end kernels |
---|
341 | ENDDO |
---|
342 | !$OMP END PARALLEL |
---|
343 | |
---|
344 | ! |
---|
345 | !-- Transpose array |
---|
346 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
---|
347 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
348 | !$acc update host( work ) |
---|
349 | CALL MPI_ALLTOALL( work(1,nzb_y,nxl_y,0), sendrecvcount_xy, MPI_REAL, & |
---|
350 | f_inv(nys_x,nzb_x,0), sendrecvcount_xy, MPI_REAL, & |
---|
351 | comm1dy, ierr ) |
---|
352 | !$acc update device( f_inv ) |
---|
353 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
---|
354 | #endif |
---|
355 | |
---|
356 | ELSE |
---|
357 | |
---|
358 | ! |
---|
359 | !-- Reorder array f_in the same way as ALLTOALL did it |
---|
360 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
361 | !$OMP DO |
---|
362 | !$acc kernels present( f_in ) |
---|
363 | !$acc loop |
---|
364 | DO i = nxl_y, nxr_y |
---|
365 | DO k = nzb_y, nzt_y |
---|
366 | !$acc loop vector( 32 ) |
---|
367 | DO j = 0, ny |
---|
368 | f_inv(j,k,i) = f_in(j,i,k) |
---|
369 | ENDDO |
---|
370 | ENDDO |
---|
371 | ENDDO |
---|
372 | !$acc end kernels |
---|
373 | !$OMP END PARALLEL |
---|
374 | |
---|
375 | ENDIF |
---|
376 | |
---|
377 | ! |
---|
378 | !-- Reorder transposed array in a way that the x index is in first position |
---|
379 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
380 | !$OMP DO |
---|
381 | !$acc kernels present( f_out ) |
---|
382 | !$acc loop |
---|
383 | DO i = 0, nx |
---|
384 | DO k = nzb_x, nzt_x |
---|
385 | !$acc loop vector( 32 ) |
---|
386 | DO j = nys_x, nyn_x |
---|
387 | f_out(i,j,k) = f_inv(j,k,i) |
---|
388 | ENDDO |
---|
389 | ENDDO |
---|
390 | ENDDO |
---|
391 | !$acc end kernels |
---|
392 | !$OMP END PARALLEL |
---|
393 | |
---|
394 | END SUBROUTINE transpose_yx |
---|
395 | |
---|
396 | |
---|
397 | SUBROUTINE transpose_yxd( f_in, work, f_out ) |
---|
398 | |
---|
399 | !------------------------------------------------------------------------------! |
---|
400 | ! Description: |
---|
401 | ! ------------ |
---|
402 | ! Transposition of input array (f_in) from y to x. For the input array, all |
---|
403 | ! elements along y reside on the same PE, while after transposition, all |
---|
404 | ! elements along x reside on the same PE. |
---|
405 | ! This is a direct transposition for arrays with indices in regular order |
---|
406 | ! (k,j,i) (cf. transpose_yx). |
---|
407 | !------------------------------------------------------------------------------! |
---|
408 | |
---|
409 | USE cpulog |
---|
410 | USE indices |
---|
411 | USE interfaces |
---|
412 | USE pegrid |
---|
413 | USE transpose_indices |
---|
414 | |
---|
415 | IMPLICIT NONE |
---|
416 | |
---|
417 | INTEGER :: i, j, k, l, m, xs |
---|
418 | |
---|
419 | REAL :: f_in(1:nz,nys:nyn,nxl:nxr), f_inv(nxl:nxr,1:nz,nys:nyn), & |
---|
420 | f_out(0:nx,nys_x:nyn_x,nzb_x:nzt_x), & |
---|
421 | work(nnx*nny*nnz) |
---|
422 | |
---|
423 | #if defined( __parallel ) |
---|
424 | |
---|
425 | ! |
---|
426 | !-- Rearrange indices of input array in order to make data to be send |
---|
427 | !-- by MPI contiguous |
---|
428 | DO k = 1, nz |
---|
429 | DO j = nys, nyn |
---|
430 | DO i = nxl, nxr |
---|
431 | f_inv(i,k,j) = f_in(k,j,i) |
---|
432 | ENDDO |
---|
433 | ENDDO |
---|
434 | ENDDO |
---|
435 | |
---|
436 | ! |
---|
437 | !-- Transpose array |
---|
438 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
---|
439 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
440 | CALL MPI_ALLTOALL( f_inv(nxl,1,nys), sendrecvcount_xy, MPI_REAL, & |
---|
441 | work(1), sendrecvcount_xy, MPI_REAL, & |
---|
442 | comm1dx, ierr ) |
---|
443 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
---|
444 | |
---|
445 | ! |
---|
446 | !-- Reorder transposed array |
---|
447 | m = 0 |
---|
448 | DO l = 0, pdims(1) - 1 |
---|
449 | xs = 0 + l * nnx |
---|
450 | DO j = nys_x, nyn_x |
---|
451 | DO k = 1, nz |
---|
452 | DO i = xs, xs + nnx - 1 |
---|
453 | m = m + 1 |
---|
454 | f_out(i,j,k) = work(m) |
---|
455 | ENDDO |
---|
456 | ENDDO |
---|
457 | ENDDO |
---|
458 | ENDDO |
---|
459 | |
---|
460 | #endif |
---|
461 | |
---|
462 | END SUBROUTINE transpose_yxd |
---|
463 | |
---|
464 | |
---|
465 | SUBROUTINE transpose_yz( f_in, work, f_out ) |
---|
466 | |
---|
467 | !------------------------------------------------------------------------------! |
---|
468 | ! Description: |
---|
469 | ! ------------ |
---|
470 | ! Transposition of input array (f_in) from y to z. For the input array, all |
---|
471 | ! elements along y reside on the same PE, while after transposition, all |
---|
472 | ! elements along z reside on the same PE. |
---|
473 | !------------------------------------------------------------------------------! |
---|
474 | |
---|
475 | USE cpulog |
---|
476 | USE indices |
---|
477 | USE interfaces |
---|
478 | USE pegrid |
---|
479 | USE transpose_indices |
---|
480 | |
---|
481 | IMPLICIT NONE |
---|
482 | |
---|
483 | INTEGER :: i, j, k, l, zs |
---|
484 | |
---|
485 | REAL :: f_in(0:ny,nxl_y:nxr_y,nzb_y:nzt_y), f_out(nxl_z:nxr_z,nys_z:nyn_z,1:nz) |
---|
486 | |
---|
487 | REAL, DIMENSION(nxl_z:nxr_z,nzt_y-nzb_y+1,nys_z:nyn_z,0:pdims(1)-1) :: work |
---|
488 | |
---|
489 | !$acc declare create( f_inv ) |
---|
490 | REAL :: f_inv(nxl_y:nxr_y,nzb_y:nzt_y,0:ny) |
---|
491 | |
---|
492 | |
---|
493 | ! |
---|
494 | !-- Rearrange indices of input array in order to make data to be send |
---|
495 | !-- by MPI contiguous |
---|
496 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
497 | !$OMP DO |
---|
498 | !$acc kernels present( f_in ) |
---|
499 | !$acc loop |
---|
500 | DO j = 0, ny |
---|
501 | DO k = nzb_y, nzt_y |
---|
502 | !$acc loop vector( 32 ) |
---|
503 | DO i = nxl_y, nxr_y |
---|
504 | f_inv(i,k,j) = f_in(j,i,k) |
---|
505 | ENDDO |
---|
506 | ENDDO |
---|
507 | ENDDO |
---|
508 | !$acc end kernels |
---|
509 | !$OMP END PARALLEL |
---|
510 | |
---|
511 | ! |
---|
512 | !-- Move data to different array, because memory location of work1 is |
---|
513 | !-- needed further below (work1 = work2). |
---|
514 | !-- If the PE grid is one-dimensional along y, only local reordering |
---|
515 | !-- of the data is necessary and no transposition has to be done. |
---|
516 | IF ( pdims(1) == 1 ) THEN |
---|
517 | |
---|
518 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
519 | !$OMP DO |
---|
520 | !$acc kernels present( f_out ) |
---|
521 | !$acc loop |
---|
522 | DO j = 0, ny |
---|
523 | DO k = nzb_y, nzt_y |
---|
524 | !$acc loop vector( 32 ) |
---|
525 | DO i = nxl_y, nxr_y |
---|
526 | f_out(i,j,k) = f_inv(i,k,j) |
---|
527 | ENDDO |
---|
528 | ENDDO |
---|
529 | ENDDO |
---|
530 | !$acc end kernels |
---|
531 | !$OMP END PARALLEL |
---|
532 | |
---|
533 | ELSE |
---|
534 | |
---|
535 | #if defined( __parallel ) |
---|
536 | ! |
---|
537 | !-- Transpose array |
---|
538 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
---|
539 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
540 | !$acc update host( f_inv ) |
---|
541 | CALL MPI_ALLTOALL( f_inv(nxl_y,nzb_y,0), sendrecvcount_yz, MPI_REAL, & |
---|
542 | work(nxl_z,1,nys_z,0), sendrecvcount_yz, MPI_REAL, & |
---|
543 | comm1dx, ierr ) |
---|
544 | !$acc update device( work ) |
---|
545 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
---|
546 | |
---|
547 | ! |
---|
548 | !-- Reorder transposed array |
---|
549 | !$OMP PARALLEL PRIVATE ( i, j, k, l, zs ) |
---|
550 | !$OMP DO |
---|
551 | DO l = 0, pdims(1) - 1 |
---|
552 | zs = 1 + l * ( nzt_y - nzb_y + 1 ) |
---|
553 | !$acc kernels present( f_out, work ) |
---|
554 | !$acc loop |
---|
555 | DO j = nys_z, nyn_z |
---|
556 | DO k = zs, zs + nzt_y - nzb_y |
---|
557 | !$acc loop vector( 32 ) |
---|
558 | DO i = nxl_z, nxr_z |
---|
559 | f_out(i,j,k) = work(i,k-zs+1,j,l) |
---|
560 | ENDDO |
---|
561 | ENDDO |
---|
562 | ENDDO |
---|
563 | !$acc end kernels |
---|
564 | ENDDO |
---|
565 | !$OMP END PARALLEL |
---|
566 | #endif |
---|
567 | |
---|
568 | ENDIF |
---|
569 | |
---|
570 | END SUBROUTINE transpose_yz |
---|
571 | |
---|
572 | |
---|
573 | SUBROUTINE transpose_zx( f_in, work, f_out ) |
---|
574 | |
---|
575 | !------------------------------------------------------------------------------! |
---|
576 | ! Description: |
---|
577 | ! ------------ |
---|
578 | ! Transposition of input array (f_in) from z to x. For the input array, all |
---|
579 | ! elements along z reside on the same PE, while after transposition, all |
---|
580 | ! elements along x reside on the same PE. |
---|
581 | !------------------------------------------------------------------------------! |
---|
582 | |
---|
583 | USE cpulog |
---|
584 | USE indices |
---|
585 | USE interfaces |
---|
586 | USE pegrid |
---|
587 | USE transpose_indices |
---|
588 | |
---|
589 | IMPLICIT NONE |
---|
590 | |
---|
591 | INTEGER :: i, j, k, l, xs |
---|
592 | |
---|
593 | REAL :: f_in(1:nz,nys:nyn,nxl:nxr), f_out(0:nx,nys_x:nyn_x,nzb_x:nzt_x) |
---|
594 | |
---|
595 | REAL, DIMENSION(nys_x:nyn_x,nnx,nzb_x:nzt_x,0:pdims(1)-1) :: work |
---|
596 | |
---|
597 | !$acc declare create( f_inv ) |
---|
598 | REAL :: f_inv(nys:nyn,nxl:nxr,1:nz) |
---|
599 | |
---|
600 | |
---|
601 | ! |
---|
602 | !-- Rearrange indices of input array in order to make data to be send |
---|
603 | !-- by MPI contiguous |
---|
604 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
605 | !$OMP DO |
---|
606 | !$acc kernels present( f_in ) |
---|
607 | !$acc loop |
---|
608 | DO k = 1,nz |
---|
609 | DO i = nxl, nxr |
---|
610 | !$acc loop vector( 32 ) |
---|
611 | DO j = nys, nyn |
---|
612 | f_inv(j,i,k) = f_in(k,j,i) |
---|
613 | ENDDO |
---|
614 | ENDDO |
---|
615 | ENDDO |
---|
616 | !$acc end kernels |
---|
617 | !$OMP END PARALLEL |
---|
618 | |
---|
619 | ! |
---|
620 | !-- Move data to different array, because memory location of work1 is |
---|
621 | !-- needed further below (work1 = work2). |
---|
622 | !-- If the PE grid is one-dimensional along y, only local reordering |
---|
623 | !-- of the data is necessary and no transposition has to be done. |
---|
624 | IF ( pdims(1) == 1 ) THEN |
---|
625 | |
---|
626 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
627 | !$OMP DO |
---|
628 | !$acc kernels present( f_out ) |
---|
629 | !$acc loop |
---|
630 | DO k = 1, nz |
---|
631 | DO i = nxl, nxr |
---|
632 | !$acc loop vector( 32 ) |
---|
633 | DO j = nys, nyn |
---|
634 | f_out(i,j,k) = f_inv(j,i,k) |
---|
635 | ENDDO |
---|
636 | ENDDO |
---|
637 | ENDDO |
---|
638 | !$acc end kernels |
---|
639 | !$OMP END PARALLEL |
---|
640 | |
---|
641 | ELSE |
---|
642 | |
---|
643 | #if defined( __parallel ) |
---|
644 | ! |
---|
645 | !-- Transpose array |
---|
646 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
---|
647 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
648 | !$acc update host( f_inv ) |
---|
649 | CALL MPI_ALLTOALL( f_inv(nys,nxl,1), sendrecvcount_zx, MPI_REAL, & |
---|
650 | work(nys_x,1,nzb_x,0), sendrecvcount_zx, MPI_REAL, & |
---|
651 | comm1dx, ierr ) |
---|
652 | !$acc update device( work ) |
---|
653 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
---|
654 | |
---|
655 | ! |
---|
656 | !-- Reorder transposed array |
---|
657 | !$OMP PARALLEL PRIVATE ( i, j, k, l, xs ) |
---|
658 | !$OMP DO |
---|
659 | DO l = 0, pdims(1) - 1 |
---|
660 | xs = 0 + l * nnx |
---|
661 | !$acc kernels present( f_out, work ) |
---|
662 | !$acc loop |
---|
663 | DO k = nzb_x, nzt_x |
---|
664 | DO i = xs, xs + nnx - 1 |
---|
665 | !$acc loop vector( 32 ) |
---|
666 | DO j = nys_x, nyn_x |
---|
667 | f_out(i,j,k) = work(j,i-xs+1,k,l) |
---|
668 | ENDDO |
---|
669 | ENDDO |
---|
670 | ENDDO |
---|
671 | !$acc end kernels |
---|
672 | ENDDO |
---|
673 | !$OMP END PARALLEL |
---|
674 | #endif |
---|
675 | |
---|
676 | ENDIF |
---|
677 | |
---|
678 | END SUBROUTINE transpose_zx |
---|
679 | |
---|
680 | |
---|
681 | SUBROUTINE transpose_zy( f_in, work, f_out ) |
---|
682 | |
---|
683 | !------------------------------------------------------------------------------! |
---|
684 | ! Description: |
---|
685 | ! ------------ |
---|
686 | ! Transposition of input array (f_in) from z to y. For the input array, all |
---|
687 | ! elements along z reside on the same PE, while after transposition, all |
---|
688 | ! elements along y reside on the same PE. |
---|
689 | !------------------------------------------------------------------------------! |
---|
690 | |
---|
691 | USE cpulog |
---|
692 | USE indices |
---|
693 | USE interfaces |
---|
694 | USE pegrid |
---|
695 | USE transpose_indices |
---|
696 | |
---|
697 | IMPLICIT NONE |
---|
698 | |
---|
699 | INTEGER :: i, j, k, l, zs |
---|
700 | |
---|
701 | REAL :: f_in(nxl_z:nxr_z,nys_z:nyn_z,1:nz), f_out(0:ny,nxl_y:nxr_y,nzb_y:nzt_y) |
---|
702 | |
---|
703 | REAL, DIMENSION(nxl_z:nxr_z,nzt_y-nzb_y+1,nys_z:nyn_z,0:pdims(1)-1) :: work |
---|
704 | |
---|
705 | !$acc declare create( f_inv ) |
---|
706 | REAL :: f_inv(nxl_y:nxr_y,nzb_y:nzt_y,0:ny) |
---|
707 | |
---|
708 | |
---|
709 | ! |
---|
710 | !-- If the PE grid is one-dimensional along y, the array has only to be |
---|
711 | !-- reordered locally and therefore no transposition has to be done. |
---|
712 | IF ( pdims(1) /= 1 ) THEN |
---|
713 | |
---|
714 | #if defined( __parallel ) |
---|
715 | ! |
---|
716 | !-- Reorder input array for transposition |
---|
717 | !$OMP PARALLEL PRIVATE ( i, j, k, l, zs ) |
---|
718 | !$OMP DO |
---|
719 | DO l = 0, pdims(1) - 1 |
---|
720 | zs = 1 + l * ( nzt_y - nzb_y + 1 ) |
---|
721 | !$acc kernels present( f_in, work ) |
---|
722 | !$acc loop |
---|
723 | DO j = nys_z, nyn_z |
---|
724 | DO k = zs, zs + nzt_y - nzb_y |
---|
725 | !$acc loop vector( 32 ) |
---|
726 | DO i = nxl_z, nxr_z |
---|
727 | work(i,k-zs+1,j,l) = f_in(i,j,k) |
---|
728 | ENDDO |
---|
729 | ENDDO |
---|
730 | ENDDO |
---|
731 | !$acc end kernels |
---|
732 | ENDDO |
---|
733 | !$OMP END PARALLEL |
---|
734 | |
---|
735 | ! |
---|
736 | !-- Transpose array |
---|
737 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
---|
738 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
739 | !$acc update host( work ) |
---|
740 | CALL MPI_ALLTOALL( work(nxl_z,1,nys_z,0), sendrecvcount_yz, MPI_REAL, & |
---|
741 | f_inv(nxl_y,nzb_y,0), sendrecvcount_yz, MPI_REAL, & |
---|
742 | comm1dx, ierr ) |
---|
743 | !$acc update device( f_inv ) |
---|
744 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
---|
745 | #endif |
---|
746 | |
---|
747 | ELSE |
---|
748 | ! |
---|
749 | !-- Reorder the array in the same way like ALLTOALL did it |
---|
750 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
751 | !$OMP DO |
---|
752 | !$acc kernels present( f_in ) |
---|
753 | !$acc loop |
---|
754 | DO k = nzb_y, nzt_y |
---|
755 | DO j = 0, ny |
---|
756 | !$acc loop vector( 32 ) |
---|
757 | DO i = nxl_y, nxr_y |
---|
758 | f_inv(i,k,j) = f_in(i,j,k) |
---|
759 | ENDDO |
---|
760 | ENDDO |
---|
761 | ENDDO |
---|
762 | !$acc end kernels |
---|
763 | !$OMP END PARALLEL |
---|
764 | |
---|
765 | ENDIF |
---|
766 | |
---|
767 | ! |
---|
768 | !-- Reorder transposed array in a way that the y index is in first position |
---|
769 | !$OMP PARALLEL PRIVATE ( i, j, k ) |
---|
770 | !$OMP DO |
---|
771 | !$acc kernels present( f_out ) |
---|
772 | !$acc loop |
---|
773 | DO k = nzb_y, nzt_y |
---|
774 | DO i = nxl_y, nxr_y |
---|
775 | !$acc loop vector( 32 ) |
---|
776 | DO j = 0, ny |
---|
777 | f_out(j,i,k) = f_inv(i,k,j) |
---|
778 | ENDDO |
---|
779 | ENDDO |
---|
780 | ENDDO |
---|
781 | !$acc end kernels |
---|
782 | !$OMP END PARALLEL |
---|
783 | |
---|
784 | END SUBROUTINE transpose_zy |
---|
785 | |
---|
786 | |
---|
787 | SUBROUTINE transpose_zyd( f_in, work, f_out ) |
---|
788 | |
---|
789 | !------------------------------------------------------------------------------! |
---|
790 | ! Description: |
---|
791 | ! ------------ |
---|
792 | ! Transposition of input array (f_in) from z to y. For the input array, all |
---|
793 | ! elements along z reside on the same PE, while after transposition, all |
---|
794 | ! elements along y reside on the same PE. |
---|
795 | ! This is a direct transposition for arrays with indices in regular order |
---|
796 | ! (k,j,i) (cf. transpose_zy). |
---|
797 | !------------------------------------------------------------------------------! |
---|
798 | |
---|
799 | USE cpulog |
---|
800 | USE indices |
---|
801 | USE interfaces |
---|
802 | USE pegrid |
---|
803 | USE transpose_indices |
---|
804 | |
---|
805 | IMPLICIT NONE |
---|
806 | |
---|
807 | INTEGER :: i, j, k, l, m, ys |
---|
808 | |
---|
809 | REAL :: f_in(1:nz,nys:nyn,nxl:nxr), f_inv(nys:nyn,nxl:nxr,1:nz), & |
---|
810 | f_out(0:ny,nxl_yd:nxr_yd,nzb_yd:nzt_yd), & |
---|
811 | work(nnx*nny*nnz) |
---|
812 | |
---|
813 | #if defined( __parallel ) |
---|
814 | |
---|
815 | ! |
---|
816 | !-- Rearrange indices of input array in order to make data to be send |
---|
817 | !-- by MPI contiguous |
---|
818 | DO i = nxl, nxr |
---|
819 | DO j = nys, nyn |
---|
820 | DO k = 1, nz |
---|
821 | f_inv(j,i,k) = f_in(k,j,i) |
---|
822 | ENDDO |
---|
823 | ENDDO |
---|
824 | ENDDO |
---|
825 | |
---|
826 | ! |
---|
827 | !-- Move data to different array, because memory location of work1 is |
---|
828 | !-- needed further below (work1 = work2). |
---|
829 | !-- If the PE grid is one-dimensional along x, only local reordering |
---|
830 | !-- of the data is necessary and no transposition has to be done. |
---|
831 | IF ( pdims(2) == 1 ) THEN |
---|
832 | DO k = 1, nz |
---|
833 | DO i = nxl, nxr |
---|
834 | DO j = nys, nyn |
---|
835 | f_out(j,i,k) = f_inv(j,i,k) |
---|
836 | ENDDO |
---|
837 | ENDDO |
---|
838 | ENDDO |
---|
839 | RETURN |
---|
840 | ENDIF |
---|
841 | |
---|
842 | ! |
---|
843 | !-- Transpose array |
---|
844 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'start' ) |
---|
845 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
846 | CALL MPI_ALLTOALL( f_inv(nys,nxl,1), sendrecvcount_zyd, MPI_REAL, & |
---|
847 | work(1), sendrecvcount_zyd, MPI_REAL, & |
---|
848 | comm1dy, ierr ) |
---|
849 | CALL cpu_log( log_point_s(32), 'mpi_alltoall', 'stop' ) |
---|
850 | |
---|
851 | ! |
---|
852 | !-- Reorder transposed array |
---|
853 | m = 0 |
---|
854 | DO l = 0, pdims(2) - 1 |
---|
855 | ys = 0 + l * nny |
---|
856 | DO k = nzb_yd, nzt_yd |
---|
857 | DO i = nxl_yd, nxr_yd |
---|
858 | DO j = ys, ys + nny - 1 |
---|
859 | m = m + 1 |
---|
860 | f_out(j,i,k) = work(m) |
---|
861 | ENDDO |
---|
862 | ENDDO |
---|
863 | ENDDO |
---|
864 | ENDDO |
---|
865 | |
---|
866 | #endif |
---|
867 | |
---|
868 | END SUBROUTINE transpose_zyd |
---|