-
Notifications
You must be signed in to change notification settings - Fork 1
/
clover.f90.offload_opt1
814 lines (668 loc) · 35.1 KB
/
clover.f90.offload_opt1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
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
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
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
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
!Crown Copyright 2012 AWE.
!
! This file is part of CloverLeaf.
!
! CloverLeaf is free software: you can redistribute it and/or modify it under
! the terms of the GNU General Public License as published by the
! Free Software Foundation, either version 3 of the License, or (at your option)
! any later version.
!
! CloverLeaf is distributed in the hope that it will be useful, but
! WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
! FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
! details.
!
! You should have received a copy of the GNU General Public License along with
! CloverLeaf. If not, see http://www.gnu.org/licenses/.
!> @brief Communication Utilities
!> @author Wayne Gaudin
!> @details Contains all utilities required to run CloverLeaf in a distributed
!> environment, including initialisation, mesh decompostion, reductions and
!> halo exchange using explicit buffers.
!>
!> Note the halo exchange is currently coded as simply as possible and no
!> optimisations have been implemented, such as post receives before sends or packing
!> buffers with multiple data fields. This is intentional so the effect of these
!> optimisations can be measured on large systems, as and when they are added.
!>
!> Even without these modifications CloverLeaf weak scales well on moderately sized
!> systems of the order of 10K cores.
MODULE clover_module
USE data_module
USE definitions_module
USE MPI
IMPLICIT NONE
INTEGER :: signal_left = 1, signal_right = 2, signal_top = 3, signal_bottom = 4
!DIR$ ATTRIBUTES OFFLOAD:MIC :: signal_left, signal_right, signal_top, signal_bottom
CONTAINS
SUBROUTINE clover_barrier
INTEGER :: err
CALL MPI_BARRIER(MPI_COMM_WORLD,err)
END SUBROUTINE clover_barrier
SUBROUTINE clover_abort
INTEGER :: ierr,err
CALL MPI_ABORT(MPI_COMM_WORLD,ierr,err)
END SUBROUTINE clover_abort
SUBROUTINE clover_finalize
INTEGER :: err
CALL FLUSH(0)
CALL FLUSH(6)
CALL FLUSH(g_out)
CALL MPI_FINALIZE(err)
END SUBROUTINE clover_finalize
SUBROUTINE clover_init_comms
USE mic_lib
USE data_module
IMPLICIT NONE
INTEGER :: err,rank,size
rank=0
size=1
CALL MPI_INIT(err)
CALL MPI_COMM_RANK(MPI_COMM_WORLD,rank,err)
CALL MPI_COMM_SIZE(MPI_COMM_WORLD,size,err)
parallel%parallel=.TRUE.
parallel%task=rank
IF(rank.EQ.0) THEN
parallel%boss=.TRUE.
ENDIF
parallel%boss_task=0
parallel%max_task=size
g_device_count=OFFLOAD_NUMBER_OF_DEVICES()
g_mic_device=mod(rank,g_device_count)
END SUBROUTINE clover_init_comms
SUBROUTINE clover_get_num_chunks(count)
IMPLICIT NONE
INTEGER :: count
! Should be changed so there can be more than one chunk per mpi task
count=parallel%max_task
END SUBROUTINE clover_get_num_chunks
SUBROUTINE clover_decompose(x_cells,y_cells,left,right,bottom,top)
! This decomposes the mesh into a number of chunks.
! The number of chunks may be a multiple of the number of mpi tasks
! Doesn't always return the best split if there are few factors
! All factors need to be stored and the best picked. But its ok for now
IMPLICIT NONE
INTEGER :: x_cells,y_cells,left(:),right(:),top(:),bottom(:),dims(2)
INTEGER :: c,delta_x,delta_y
REAL(KIND=8) :: mesh_ratio,a,b,factor_x,factor_y
INTEGER :: chunk_x,chunk_y,mod_x,mod_y,split_found
INTEGER :: cx,cy,chunk,add_x,add_y,add_x_prev,add_y_prev
! 2D Decomposition of the mesh
mesh_ratio=real(x_cells)/real(y_cells)
chunk_x=number_of_chunks
chunk_y=1
split_found=0 ! Used to detect 1D decomposition
DO c=1,number_of_chunks
IF (MOD(number_of_chunks,c).EQ.0) THEN
factor_x=number_of_chunks/real(c)
factor_y=c
!Compare the factor ratio with the mesh ratio
IF(factor_x/factor_y.LE.mesh_ratio) THEN
chunk_y=c
chunk_x=number_of_chunks/c
split_found=1
EXIT
ENDIF
ENDIF
ENDDO
IF(split_found.EQ.0.OR.chunk_y.EQ.number_of_chunks) THEN ! Prime number or 1D decomp detected
IF(mesh_ratio.GE.1.0) THEN
chunk_x=number_of_chunks
chunk_y=1
ELSE
chunk_x=1
chunk_y=number_of_chunks
ENDIF
ENDIF
delta_x=x_cells/chunk_x
delta_y=y_cells/chunk_y
mod_x=MOD(x_cells,chunk_x)
mod_y=MOD(y_cells,chunk_y)
! Set up chunk mesh ranges and chunk connectivity
add_x_prev=0
add_y_prev=0
chunk=1
DO cy=1,chunk_y
DO cx=1,chunk_x
add_x=0
add_y=0
IF(cx.LE.mod_x)add_x=1
IF(cy.LE.mod_y)add_y=1
left(chunk)=(cx-1)*delta_x+1+add_x_prev
right(chunk)=left(chunk)+delta_x-1+add_x
bottom(chunk)=(cy-1)*delta_y+1+add_y_prev
top(chunk)=bottom(chunk)+delta_y-1+add_y
chunks(chunk)%chunk_neighbours(chunk_left)=chunk_x*(cy-1)+cx-1
chunks(chunk)%chunk_neighbours(chunk_right)=chunk_x*(cy-1)+cx+1
chunks(chunk)%chunk_neighbours(chunk_bottom)=chunk_x*(cy-2)+cx
chunks(chunk)%chunk_neighbours(chunk_top)=chunk_x*(cy)+cx
IF(cx.EQ.1)chunks(chunk)%chunk_neighbours(chunk_left)=external_face
IF(cx.EQ.chunk_x)chunks(chunk)%chunk_neighbours(chunk_right)=external_face
IF(cy.EQ.1)chunks(chunk)%chunk_neighbours(chunk_bottom)=external_face
IF(cy.EQ.chunk_y)chunks(chunk)%chunk_neighbours(chunk_top)=external_face
IF(cx.LE.mod_x)add_x_prev=add_x_prev+1
chunk=chunk+1
ENDDO
add_x_prev=0
IF(cy.LE.mod_y)add_y_prev=add_y_prev+1
ENDDO
IF(parallel%boss)THEN
WRITE(g_out,*)
WRITE(g_out,*)"Mesh ratio of ",mesh_ratio
WRITE(g_out,*)"Decomposing the mesh into ",chunk_x," by ",chunk_y," chunks"
WRITE(g_out,*)
ENDIF
END SUBROUTINE clover_decompose
SUBROUTINE clover_allocate_buffers(chunk)
IMPLICIT NONE
INTEGER :: chunk
! Unallocated buffers for external boundaries caused issues on some systems so they are now
! all allocated
IF(parallel%task.EQ.chunks(chunk)%task)THEN
!IF(chunks(chunk)%chunk_neighbours(chunk_left).NE.external_face) THEN
ALLOCATE(chunks(chunk)%left_snd_buffer(2*(chunks(chunk)%field%y_max+5)))
ALLOCATE(chunks(chunk)%left_rcv_buffer(2*(chunks(chunk)%field%y_max+5)))
!ENDIF
!IF(chunks(chunk)%chunk_neighbours(chunk_right).NE.external_face) THEN
ALLOCATE(chunks(chunk)%right_snd_buffer(2*(chunks(chunk)%field%y_max+5)))
ALLOCATE(chunks(chunk)%right_rcv_buffer(2*(chunks(chunk)%field%y_max+5)))
!ENDIF
!IF(chunks(chunk)%chunk_neighbours(chunk_bottom).NE.external_face) THEN
ALLOCATE(chunks(chunk)%bottom_snd_buffer(2*(chunks(chunk)%field%x_max+5)))
ALLOCATE(chunks(chunk)%bottom_rcv_buffer(2*(chunks(chunk)%field%x_max+5)))
!ENDIF
!IF(chunks(chunk)%chunk_neighbours(chunk_top).NE.external_face) THEN
ALLOCATE(chunks(chunk)%top_snd_buffer(2*(chunks(chunk)%field%x_max+5)))
ALLOCATE(chunks(chunk)%top_rcv_buffer(2*(chunks(chunk)%field%x_max+5)))
!ENDIF
ENDIF
END SUBROUTINE clover_allocate_buffers
SUBROUTINE clover_exchange(fields,depth)
IMPLICIT NONE
INTEGER :: fields(:),depth
! Assuming 1 patch per task, this will be changed
! Also, not packing all fields for each communication, doing one at a time
IF(fields(FIELD_DENSITY0).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%density0, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,CELL_DATA)
ENDIF
IF(fields(FIELD_DENSITY1).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%density1, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,CELL_DATA)
ENDIF
IF(fields(FIELD_ENERGY0).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%energy0, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,CELL_DATA)
ENDIF
IF(fields(FIELD_ENERGY1).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%energy1, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,CELL_DATA)
ENDIF
IF(fields(FIELD_PRESSURE).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%pressure, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,CELL_DATA)
ENDIF
IF(fields(FIELD_VISCOSITY).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%viscosity, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,CELL_DATA)
ENDIF
IF(fields(FIELD_SOUNDSPEED).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%soundspeed, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,CELL_DATA)
ENDIF
IF(fields(FIELD_XVEL0).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%xvel0, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,VERTEX_DATA)
ENDIF
IF(fields(FIELD_XVEL1).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%xvel1, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,VERTEX_DATA)
ENDIF
IF(fields(FIELD_YVEL0).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%yvel0, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,VERTEX_DATA)
ENDIF
IF(fields(FIELD_YVEL1).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%yvel1, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,VERTEX_DATA)
ENDIF
IF(fields(FIELD_VOL_FLUX_X).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%vol_flux_x, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,X_FACE_DATA)
ENDIF
IF(fields(FIELD_VOL_FLUX_Y).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%vol_flux_y, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,Y_FACE_DATA)
ENDIF
IF(fields(FIELD_MASS_FLUX_X).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%mass_flux_x, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,X_FACE_DATA)
ENDIF
IF(fields(FIELD_MASS_FLUX_Y).EQ.1) THEN
CALL clover_exchange_message(parallel%task+1,chunks(parallel%task+1)%field%mass_flux_y, &
chunks(parallel%task+1)%left_snd_buffer, &
chunks(parallel%task+1)%left_rcv_buffer, &
chunks(parallel%task+1)%right_snd_buffer, &
chunks(parallel%task+1)%right_rcv_buffer, &
chunks(parallel%task+1)%bottom_snd_buffer, &
chunks(parallel%task+1)%bottom_rcv_buffer, &
chunks(parallel%task+1)%top_snd_buffer, &
chunks(parallel%task+1)%top_rcv_buffer, &
depth,Y_FACE_DATA)
ENDIF
END SUBROUTINE clover_exchange
SUBROUTINE clover_exchange_message(chunk,field, &
left_snd_buffer, &
left_rcv_buffer, &
right_snd_buffer, &
right_rcv_buffer, &
bottom_snd_buffer, &
bottom_rcv_buffer, &
top_snd_buffer, &
top_rcv_buffer, &
depth,field_type)
IMPLICIT NONE
REAL(KIND=8) :: field(-1:,-1:) ! This seems to work for any type of mesh data
REAL(KIND=8) :: left_snd_buffer(:),left_rcv_buffer(:),right_snd_buffer(:),right_rcv_buffer(:)
REAL(KIND=8) :: bottom_snd_buffer(:),bottom_rcv_buffer(:),top_snd_buffer(:),top_rcv_buffer(:)
INTEGER :: chunk,depth,field_type
!dir$ options /offload_attribute_target=mic
INTEGER :: size,err,target,request(8),tag,message_count,j,k,x_inc,y_inc,index
INTEGER :: status(MPI_STATUS_SIZE,8)
INTEGER :: receiver,sender
INTEGER :: loc_x_min, loc_x_max, loc_y_min, loc_y_max
INTEGER :: chunk_neighbours_left, chunk_neighbours_right
INTEGER :: chunk_neighbours_bottom, chunk_neighbours_top
!dec$ end options
! Field type will either be cell, vertex, x_face or y_face to get the message limits correct
! I am packing my own buffers. I am sure this could be improved with MPI data types
! but this will do for now
! I am also sending buffers to chunks with the same task id for now.
! This can be improved in the future but at the moment there is just 1 chunk per task anyway
! The tag will be a function of the sending chunk and the face it is coming from
! like chunk 6 sending the left face
! No open mp in here either. May be beneficial will packing and unpacking in the future, though I am not sure.
! Change this so it will allow more than 1 chunk per task
request=0
message_count=0
! Pack and send
! These array modifications still need to be added on, plus the donor data location changes as in update_halo
IF(field_type.EQ.CELL_DATA) THEN
x_inc=0
y_inc=0
ENDIF
IF(field_type.EQ.VERTEX_DATA) THEN
x_inc=1
y_inc=1
ENDIF
IF(field_type.EQ.X_FACE_DATA) THEN
x_inc=1
y_inc=0
ENDIF
IF(field_type.EQ.Y_FACE_DATA) THEN
x_inc=0
y_inc=1
ENDIF
loc_x_min=chunks(chunk)%field%x_min
loc_x_max=chunks(chunk)%field%x_max
loc_y_min=chunks(chunk)%field%y_min
loc_y_max=chunks(chunk)%field%y_max
chunk_neighbours_left = chunks(chunk)%chunk_neighbours(chunk_left)
chunk_neighbours_right = chunks(chunk)%chunk_neighbours(chunk_right)
chunk_neighbours_top = chunks(chunk)%chunk_neighbours(chunk_top)
chunk_neighbours_bottom = chunks(chunk)%chunk_neighbours(chunk_bottom)
! Pack real data into buffers
IF(parallel%task.EQ.chunks(chunk)%task) THEN
size=(1+(chunks(chunk)%field%y_max+y_inc+depth)-(chunks(chunk)%field%y_min-depth))*depth
!DIR$ OFFLOAD BEGIN TARGET(MIC:g_mic_device) &
!DIR$ if((chunk_neighbours_left.NE.external_face) .OR. (chunk_neighbours_right.NE.external_face)) &
!DIR$ in(chunk_neighbours_left,chunk_neighbours_right) &
!DIR$ in(loc_y_min,loc_y_max,depth,loc_x_min,x_inc,loc_x_max,y_inc) &
!DIR$ nocopy(k,j,index) &
!DIR$ in(left_snd_buffer : length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(right_snd_buffer: length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(field : length(0) alloc_if(.false.) free_if(.false.))
!$OMP PARALLEL
IF(chunk_neighbours_left.NE.external_face) THEN
!$OMP DO PRIVATE(index)
DO k=loc_y_min-depth,loc_y_max+y_inc+depth
DO j=1,depth
index=j+(k+depth-1)*depth
left_snd_buffer(index)=field(loc_x_min+x_inc-1+j,k)
ENDDO
ENDDO
!$OMP END DO
ENDIF
IF(chunk_neighbours_right.NE.external_face) THEN
!$OMP DO PRIVATE(index)
DO k=loc_y_min-depth,loc_y_max+y_inc+depth
DO j=1,depth
index=j+(k+depth-1)*depth
right_snd_buffer(index)=field(loc_x_max+1-j,k)
ENDDO
ENDDO
!$OMP END DO
ENDIF
!$OMP END PARALLEL
!DIR$ END OFFLOAD
!!DIR$ signal(signal_left)
!!DIR$ signal(signal_right)
IF(chunks(chunk)%chunk_neighbours(chunk_left).NE.external_face) THEN
size=(1+(chunks(chunk)%field%y_max+y_inc+depth)-(chunks(chunk)%field%y_min-depth))*depth
!DIR$ OFFLOAD_TRANSFER TARGET(MIC:g_mic_device) &
!DIR$ out(left_snd_buffer: length(size) alloc_if(.false.) free_if(.false.))
!!dir$ offload_wait target(mic:g_mic_device) wait(signal_left)
tag=4*(chunk)+1 ! 4 because we have 4 faces, 1 because it is leaving the left face
receiver=chunks(chunks(chunk)%chunk_neighbours(chunk_left))%task
CALL MPI_ISEND(left_snd_buffer,size,MPI_DOUBLE_PRECISION,receiver,tag &
,MPI_COMM_WORLD,request(message_count+1),err)
tag=4*(chunks(chunk)%chunk_neighbours(chunk_left))+2 ! 4 because we have 4 faces, 1 because it is coming from the right face of the left neighbour
sender=chunks(chunks(chunk)%chunk_neighbours(chunk_left))%task
CALL MPI_IRECV(left_rcv_buffer,size,MPI_DOUBLE_PRECISION,sender,tag &
,MPI_COMM_WORLD,request(message_count+2),err)
message_count=message_count+2
ENDIF
IF(chunks(chunk)%chunk_neighbours(chunk_right).NE.external_face) THEN
size=(1+(chunks(chunk)%field%y_max+y_inc+depth)-(chunks(chunk)%field%y_min-depth))*depth
!DIR$ OFFLOAD_TRANSFER TARGET(MIC:g_mic_device) &
!DIR$ out(right_snd_buffer: length(size) alloc_if(.false.) free_if(.false.))
!!dir$ offload_wait target(mic:g_mic_device) wait(signal_right)
tag=4*chunk+2 ! 4 because we have 4 faces, 2 because it is leaving the right face
receiver=chunks(chunks(chunk)%chunk_neighbours(chunk_right))%task
CALL MPI_ISEND(right_snd_buffer,size,MPI_DOUBLE_PRECISION,receiver,tag &
,MPI_COMM_WORLD,request(message_count+1),err)
tag=4*(chunks(chunk)%chunk_neighbours(chunk_right))+1 ! 4 because we have 4 faces, 1 because it is coming from the left face of the right neighbour
sender=chunks(chunks(chunk)%chunk_neighbours(chunk_right))%task
CALL MPI_IRECV(right_rcv_buffer,size,MPI_DOUBLE_PRECISION,sender,tag, &
MPI_COMM_WORLD,request(message_count+2),err)
message_count=message_count+2
ENDIF
ENDIF
! Wait for the messages
CALL MPI_WAITALL(message_count,request,status,err)
!DIR$ OFFLOAD_TRANSFER TARGET(MIC:g_mic_device) &
!DIR$ if(chunk_neighbours_left.NE.external_face) &
!DIR$ in(left_rcv_buffer: length(size) alloc_if(.false.) free_if(.false.))
!DIR$ OFFLOAD_TRANSFER TARGET(MIC:g_mic_device) &
!DIR$ if(chunk_neighbours_right.NE.external_face) &
!DIR$ in(right_rcv_buffer: length(size) alloc_if(.false.) free_if(.false.))
! Unpack buffers in halo cells
IF(parallel%task.EQ.chunks(chunk)%task) THEN
!DIR$ OFFLOAD BEGIN TARGET(MIC:g_mic_device) &
!DIR$ if((chunk_neighbours_left.NE.external_face) .OR. (chunk_neighbours_right.NE.external_face)) &
!DIR$ in(field : length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(left_rcv_buffer: length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(right_rcv_buffer: length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(chunk_neighbours_left,chunk_neighbours_right) &
!DIR$ in(loc_y_min,loc_y_max,depth,loc_x_min,x_inc,loc_x_max,y_inc) &
!DIR$ nocopy(k,j,index)
!$OMP PARALLEL
IF(chunk_neighbours_left.NE.external_face) THEN
!$OMP DO PRIVATE(INDEX)
DO k=loc_y_min-depth,loc_y_max+y_inc+depth
DO j=1,depth
index=j+(k+depth-1)*depth
field(loc_x_min-j,k)=left_rcv_buffer(index)
ENDDO
ENDDO
!$OMP END DO
ENDIF
IF(chunk_neighbours_right.NE.external_face) THEN
!$OMP DO PRIVATE(INDEX)
DO k=loc_y_min-depth,loc_y_max+y_inc+depth
DO j=1,depth
index=j+(k+depth-1)*depth
field(loc_x_max+x_inc+j,k)=right_rcv_buffer(index)
ENDDO
ENDDO
!$OMP END DO
ENDIF
!$OMP END PARALLEL
!DIR$ END OFFLOAD
ENDIF
request=0
message_count=0
IF(parallel%task.EQ.chunks(chunk)%task) THEN
size=(1+(chunks(chunk)%field%x_max+x_inc+depth)-(chunks(chunk)%field%x_min-depth))*depth
!DIR$ OFFLOAD BEGIN TARGET(MIC:g_mic_device) &
!DIR$ if((chunk_neighbours_bottom.NE.external_face) .OR. (chunk_neighbours_top.NE.external_face)) &
!DIR$ in(field : length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(bottom_snd_buffer: length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(top_snd_buffer: length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(chunk_neighbours_bottom,chunk_neighbours_top) &
!DIR$ in(loc_y_min,loc_y_max,depth,loc_x_min,x_inc,loc_x_max,y_inc) &
!DIR$ nocopy(k,j,index)
!$OMP PARALLEL
IF(chunk_neighbours_bottom.NE.external_face) THEN
DO k=1,depth
!$OMP DO PRIVATE(INDEX)
DO j=loc_x_min-depth,loc_x_max+x_inc+depth
index=j+depth+(k-1)*(loc_x_max+x_inc+(2*depth))
bottom_snd_buffer(index)=field(j,loc_y_min+y_inc-1+k)
ENDDO
!$OMP END DO
ENDDO
ENDIF
IF(chunk_neighbours_top.NE.external_face) THEN
DO k=1,depth
!$OMP DO PRIVATE(INDEX)
DO j=loc_x_min-depth,loc_x_max+x_inc+depth
index=j+depth+(k-1)*(loc_x_max+x_inc+(2*depth))
top_snd_buffer(index)=field(j,loc_y_max+1-k)
ENDDO
!$OMP END DO
ENDDO
ENDIF
!$OMP END PARALLEL
!DIR$ END OFFLOAD
IF(chunks(chunk)%chunk_neighbours(chunk_bottom).NE.external_face) THEN
!DIR$ OFFLOAD_TRANSFER TARGET(MIC:g_mic_device) &
!DIR$ out(bottom_snd_buffer: length(size) alloc_if(.false.) free_if(.false.))
tag=4*(chunk)+3 ! 4 because we have 4 faces, 3 because it is leaving the bottom face
receiver=chunks(chunks(chunk)%chunk_neighbours(chunk_bottom))%task
CALL MPI_ISEND(bottom_snd_buffer,size,MPI_DOUBLE_PRECISION,receiver,tag &
,MPI_COMM_WORLD,request(message_count+1),err)
tag=4*(chunks(chunk)%chunk_neighbours(chunk_bottom))+4 ! 4 because we have 4 faces, 1 because it is coming from the top face of the bottom neighbour
sender=chunks(chunks(chunk)%chunk_neighbours(chunk_bottom))%task
CALL MPI_IRECV(bottom_rcv_buffer,size,MPI_DOUBLE_PRECISION,sender,tag &
,MPI_COMM_WORLD,request(message_count+2),err)
message_count=message_count+2
ENDIF
IF(chunks(chunk)%chunk_neighbours(chunk_top).NE.external_face) THEN
!DIR$ OFFLOAD_TRANSFER TARGET(MIC:g_mic_device) &
!DIR$ out(top_snd_buffer: length(size) alloc_if(.false.) free_if(.false.))
tag=4*(chunk)+4 ! 4 because we have 4 faces, 4 because it is leaving the top face
receiver=chunks(chunks(chunk)%chunk_neighbours(chunk_top))%task
CALL MPI_ISEND(top_snd_buffer,size,MPI_DOUBLE_PRECISION,receiver,tag &
,MPI_COMM_WORLD,request(message_count+1),err)
tag=4*(chunks(chunk)%chunk_neighbours(chunk_top))+3 ! 4 because we have 4 faces, 4 because it is coming from the left face of the top neighbour
sender=chunks(chunks(chunk)%chunk_neighbours(chunk_top))%task
CALL MPI_IRECV(top_rcv_buffer,size,MPI_DOUBLE_PRECISION,sender,tag, &
MPI_COMM_WORLD,request(message_count+2),err)
message_count=message_count+2
ENDIF
ENDIF
! Wait for the messages
CALL MPI_WAITALL(message_count,request,status,err)
!DIR$ OFFLOAD_TRANSFER TARGET(MIC:g_mic_device) &
!DIR$ if(chunk_neighbours_bottom.NE.external_face) &
!DIR$ in(bottom_rcv_buffer: length(size) alloc_if(.false.) free_if(.false.))
!DIR$ OFFLOAD_TRANSFER TARGET(MIC:g_mic_device) &
!DIR$ if(chunk_neighbours_top.NE.external_face) &
!DIR$ in(top_rcv_buffer: length(size) alloc_if(.false.) free_if(.false.))
! Unpack buffers in halo cells
IF(parallel%task.EQ.chunks(chunk)%task) THEN
!DIR$ OFFLOAD BEGIN TARGET(MIC:g_mic_device) &
!DIR$ if((chunk_neighbours_bottom.NE.external_face) .OR. (chunk_neighbours_top.NE.external_face)) &
!DIR$ in(field : length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(bottom_rcv_buffer: length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(top_rcv_buffer: length(0) alloc_if(.false.) free_if(.false.)) &
!DIR$ in(chunk_neighbours_bottom,chunk_neighbours_top) &
!DIR$ in(loc_y_min,loc_y_max,depth,loc_x_min,x_inc,loc_x_max,y_inc) &
!DIR$ nocopy(k,j,index)
!$OMP PARALLEL
IF(chunk_neighbours_bottom.NE.external_face) THEN
DO k=1,depth
!$OMP DO PRIVATE(INDEX)
DO j=loc_x_min-depth,loc_x_max+x_inc+depth
index=j+depth+(k-1)*(loc_x_max+x_inc+(2*depth))
field(j,loc_y_min-k)=bottom_rcv_buffer(index)
ENDDO
!$OMP END DO
ENDDO
ENDIF
IF(chunk_neighbours_top.NE.external_face) THEN
DO k=1,depth
!$OMP DO PRIVATE(INDEX)
DO j=loc_x_min-depth,loc_x_max+x_inc+depth
index=j+depth+(k-1)*(loc_x_max+x_inc+(2*depth))
field(j,loc_y_max+y_inc+k)=top_rcv_buffer(index)
ENDDO
!$OMP END DO
ENDDO
ENDIF
!$OMP END PARALLEL
!DIR$ END OFFLOAD
ENDIF
END SUBROUTINE clover_exchange_message
SUBROUTINE clover_sum(value)
! Only sums to the master
IMPLICIT NONE
REAL(KIND=8) :: value
REAL(KIND=8) :: total
INTEGER :: err
total=value
CALL MPI_REDUCE(value,total,1,MPI_DOUBLE_PRECISION,MPI_SUM,0,MPI_COMM_WORLD,err)
value=total
END SUBROUTINE clover_sum
SUBROUTINE clover_min(value)
IMPLICIT NONE
REAL(KIND=8) :: value
REAL(KIND=8) :: minimum
INTEGER :: err
minimum=value
CALL MPI_ALLREDUCE(value,minimum,1,MPI_DOUBLE_PRECISION,MPI_MIN,MPI_COMM_WORLD,err)
value=minimum
END SUBROUTINE clover_min
SUBROUTINE clover_check_error(error)
IMPLICIT NONE
INTEGER :: error
INTEGER :: maximum
INTEGER :: err
maximum=error
CALL MPI_ALLREDUCE(error,maximum,1,MPI_INTEGER,MPI_MAX,MPI_COMM_WORLD,err)
error=maximum
END SUBROUTINE clover_check_error
END MODULE clover_module