-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathm_HACApK_solve_cuda.cuf
2070 lines (1809 loc) · 71.2 KB
/
m_HACApK_solve_cuda.cuf
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
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
module m_HACApK_solve_cuda
use cudafor
implicit none
contains
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low0 &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,la1,la2,la3)
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp1,idxp2,idxp3
real*8, dimension(nd) :: zau,zu
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value :: nd,nlf,n_low,la1,la2,la3
integer*4 :: ii,tx,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt
real*8 :: zbutt,zautt
integer*4 :: istat
real*8, dimension(32) :: zbuttt
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
ii = blockIdx%x
tx = threadIdx%x
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx1 = idxp1(ii) ; idx2 = idxp2(ii)
kt=ktp(ii)
do il=1,kt
zbutt = 0.0d0
do it=tx,ndt,32
itt=it+nstrtt-1
zbutt=zbutt+a1(it+(il-1)*ndt+idx1-1)*zu(itt)
enddo
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
do it=tx,ndl,32
ill=it+nstrtl-1
zautt=a2(it+(il-1)*ndl+idx2-1)*zbutt
istat = atomicadd(zau(ill),zautt)
enddo
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low0
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low0_2 &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,la1,la2,la3)
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp1,idxp2,idxp3
real*8, dimension(nd) :: zau,zu
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value :: nd,nlf,n_low,la1,la2,la3
integer*4 :: ii,tx,ty,bydim,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt
real*8 :: zbutt,zautt
integer*4 :: istat
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
ii = blockIdx%x
tx = threadIdx%x
bydim = blockDim%y
ty = threadIdx%y
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx1 = idxp1(ii) ; idx2 = idxp2(ii)
kt=ktp(ii)
do il=ty,kt,bydim
zbutt = 0.0d0
do it=tx,ndt,32
itt=it+nstrtt-1
zbutt=zbutt+a1(it+(il-1)*ndt+idx1-1)*zu(itt)
enddo
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
do it=tx,ndl,32
ill=it+nstrtl-1
zautt=a2(it+(il-1)*ndl+idx2-1)*zbutt
istat = atomicadd(zau(ill),zautt)
enddo
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low0_2
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low0_3 &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,la1,la2,la3)
#if 0
integer*4, dimension(nlf),intent(in) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf),intent(in) :: idxp1,idxp2,idxp3
real*8, dimension(nd),intent(in) :: zu
real*8, dimension(nd),intent(inout) :: zau
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value,intent(in) :: nd,nlf,n_low,la1,la2,la3
#else
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp1,idxp2,idxp3
real*8, dimension(nd) :: zu
real*8, dimension(nd) :: zau
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value,intent(in) :: nd,nlf,n_low,la1,la2,la3
#endif
integer*4 :: ii,tx,ty,bydim,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt,txx
real*8 :: zbutt,zautt
integer*4 :: istat
real*8, shared :: tmp(64) ! blocksize should be (64,1,1)
ii = blockIdx%x
tx = threadIdx%x
txx = mod(tx+31,64)+1
bydim = blockDim%y
ty = threadIdx%y
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx1 = idxp1(ii) ; idx2 = idxp2(ii)
kt=ktp(ii)
do il=ty,kt,bydim
zbutt = 0.0d0
do it=tx,ndt,64
itt=it+nstrtt-1
zbutt=zbutt+a1(it+(il-1)*ndt+idx1-1)*zu(itt)
enddo
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
call syncthreads()
tmp(tx) = zbutt
call syncthreads()
zbutt = zbutt + tmp(txx)
do it=tx,ndl,64
ill=it+nstrtl-1
zautt=a2(it+(il-1)*ndl+idx2-1)*zbutt
istat = atomicadd(zau(ill),zautt)
enddo
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low0_3
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low7 &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,idxp1,idxp2, &
nd,nlf,n_low,la1,la2,kmax)
integer*4, dimension(nlf), intent(in) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
! integer*4, dimension(kmax,n_low),intent(in) :: idxp1,idxp2
integer*4, dimension(n_low,kmax),intent(in) :: idxp1,idxp2
real*8, dimension(nd),intent(inout) :: zau,zu
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
integer*4,value,intent(in) :: nd,nlf,n_low,la1,la2,kmax
integer*4 :: ii,tx,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt,txx
real*8 :: zbutt,zautt
integer*4 :: istat
real*8, shared :: tmp(64) ! blocksize should be (64,1,1)
ii = blockIdx%x
tx = threadIdx%x
txx = mod(tx+31,64)+1
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
kt=ktp(ii)
do il=1,kt
! idx1 = idxp1(il,ii) ; idx2 = idxp2(il,ii)
idx1 = idxp1(ii,il) ; idx2 = idxp2(ii,il)
! if(tx==1 .and. ii == 2) print *, il,ii,idx1,idx2
zbutt = 0.0d0
do it=tx,ndt,64
itt=it+nstrtt-1
zbutt=zbutt+a1(it+idx1-1)*zu(itt)
enddo
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
call syncthreads()
tmp(tx) = zbutt
call syncthreads()
zbutt = zbutt + tmp(txx)
! if(tx == 1 .and. ii == 1) then
! print *, ii,il,zbutt
! end if
do it=tx,ndl,64
ill=it+nstrtl-1
zautt=a2(it+idx2-1)*zbutt
istat = atomicadd(zau(ill),zautt)
enddo
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low7
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low_balanced_col &
(zbu_tmp,zu,subid,ndtp,nstrttp,ktp,a1,idxp1,kt_prefix,nd,nlf,n_low,la1,kmax,sublen,lenkt)
real*8, dimension(lenkt) ,intent(inout) :: zbu_tmp
real*8, dimension(nd) ,intent(in) :: zu
integer*4, dimension(nlf) ,intent(in) :: ndtp,nstrttp,ktp
integer*4, dimension(kmax,n_low),intent(in) :: idxp1
! integer*4, dimension(n_low,kmax),intent(in) :: idxp1
integer*4, dimension(sublen,2) ,intent(in) :: subid
integer*4, dimension(n_low) ,intent(in) :: kt_prefix
real*8, dimension(la1) :: a1
integer*4,value,intent(in) :: nd,nlf,n_low,la1,kmax,sublen,lenkt
! integer*4 :: ii,jj,tx,txx,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt
integer*4 :: ii,jj,tx,txx,it,il,itt,ndt,nstrtt,idx1,kt,ktpre
real*8 :: zbutt,zut
integer*4 :: istat
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
ii = subid(blockIdx%x,1)
jj = subid(blockIdx%x,2)
tx = threadIdx%x
txx= mod(tx+31,32)+1
it = blockDim%x * (jj-1) + tx
ndt =ndtp(ii)
zbutt = 0.0d0
! if(tx == 1) tmp(1) = 0.0d0
if(it > ndt) return
! if(blockIdx%x == 1) then
! print *, tx, ii, jj
! print *, tx, txx,it,ndt
! end if
nstrtt=nstrttp(ii)
itt =it+nstrtt-1
zut =zu(itt)
kt =ktp(ii)
ktpre =kt_prefix(ii)-1
do il=1,kt
idx1 = idxp1(il,ii)
! idx1 = idxp1(ii,il)
zbutt = a1(it+idx1-1)*zut
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
if(txx == 1) then
istat = atomicadd(zbu_tmp(ktpre+il),zbutt)
! zbu_tmp(ktpre+il) = zbutt
endif
! call syncthreads()
! if(tx == 33) tmp(1) = zbutt
! call syncthreads()
! if(tx == 1) then
! zbutt = zbutt + tmp(1)
! istat = atomicadd(zbu_tmp(ktpre+il),zbutt)
! endif
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low_balanced_col
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low_balanced_col2 &
(zbu_tmp,zu,subid,ndtp,nstrttp,ktp,a1,idxp1,kt_prefix,nd,nlf,n_low,la1,kmax,sublen,lenkt)
real*8, dimension(lenkt) ,intent(inout) :: zbu_tmp
real*8, dimension(nd) ,intent(in) :: zu
integer*4, dimension(nlf) ,intent(in) :: ndtp,nstrttp,ktp
integer*4, dimension(kmax,n_low),intent(in) :: idxp1
! integer*4, dimension(n_low,kmax),intent(in) :: idxp1
integer*4, dimension(sublen,2) ,intent(in) :: subid
integer*4, dimension(n_low) ,intent(in) :: kt_prefix
real*8, dimension(la1) :: a1
integer*4,value,intent(in) :: nd,nlf,n_low,la1,kmax,sublen,lenkt
! integer*4 :: ii,jj,tx,txx,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt
integer*4 :: ii,jj,tx,txx,it,il,itt,ndt,nstrtt,idx1,kt,ktpre,ty,dimy,dimx
real*8 :: zbutt,zut1,zut2
integer*4 :: istat
logical :: flag
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
ii = subid(blockIdx%x,1)
jj = subid(blockIdx%x,2)
tx = threadIdx%x
ty = threadIdx%y
dimx = blockDim%x
dimy = blockDim%y
txx= mod(tx+31,32)+1
! it = blockDim%x * (jj-1) + tx
it = dimx * (jj-1) + tx
ndt =ndtp(ii)
zbutt = 0.0d0
! if(tx == 1) tmp(1) = 0.0d0
if(it > ndt) return
! flag = .false.
! if(it+32 > ndt) flag = .true.
! if(blockIdx%x == 1) then
! print *, tx, ii, jj
! print *, tx, txx,it,ndt
! end if
nstrtt=nstrttp(ii)
itt =it+nstrtt-1
zut1 =zu(itt)
! if(flag) zut2 =zu(itt+32)
kt =ktp(ii)
ktpre =kt_prefix(ii)-1
do il=ty,kt,dimy
! do il=1,kt
idx1 = idxp1(il,ii)
! idx1 = idxp1(ii,il)
! if(flag) then
zbutt = a1(it+idx1-1)*zut1
! else
! zbutt = a1(it+idx1-1)*zut1 + a1(it+32+idx1-1)*zut2
! endif
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
if(txx == 1) then
istat = atomicadd(zbu_tmp(ktpre+il),zbutt)
! zbu_tmp(ktpre+il) = zbutt
endif
! call syncthreads()
! if(tx == 33) tmp(1) = zbutt
! call syncthreads()
! if(tx == 1) then
! zbutt = zbutt + tmp(1)
! istat = atomicadd(zbu_tmp(ktpre+il),zbutt)
! endif
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low_balanced_col2
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low_balanced_col3 &
(zbu_tmp,zu,subid,ndtp,nstrttp,ktp,a1,idxp1,kt_prefix,nd,nlf,n_low,la1,kmax,sublen,lenkt)
real*8, dimension(lenkt) ,intent(inout) :: zbu_tmp
real*8, dimension(nd) ,intent(in) :: zu
integer*4, dimension(nlf) ,intent(in) :: ndtp,nstrttp,ktp
integer*4, dimension(kmax,n_low),intent(in) :: idxp1
integer*4, dimension(sublen,2) ,intent(in) :: subid
integer*4, dimension(n_low) ,intent(in) :: kt_prefix
real*8, dimension(la1) :: a1
integer*4,value,intent(in) :: nd,nlf,n_low,la1,kmax,sublen,lenkt
integer*4 :: ii,jj,tx,txx,it,il,itt,ill,ndt,nstrtt,idx1,kt,ktpre,ty,dimy,dimx,bid
real*8 :: zbutt,zut1,zbuttt
integer*4 :: istat
logical :: flag
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
bid = blockIdx%x
ii = subid(bid,1)
jj = subid(bid,2)
tx = threadIdx%x
txx= mod(tx-1,32)+1
dimx = blockDim%x
it = dimx * (jj-1) + tx
ndt =ndtp(ii)
zbutt = 0.0d0
zbuttt = 0.0d0
! if(it > ndt) return
nstrtt=nstrttp(ii)
itt =it+nstrtt-1
zut1 =zu(itt)
kt =ktp(ii)
ktpre =kt_prefix(ii)-1
do ill=1,kt,32
! if(bid == 172951 .and. tx == 1) print *, kt, ill, bid
do il=ill,min(kt,ill+31)
idx1 = idxp1(il,ii)
if(it <= ndt) then
zbutt = a1(it+idx1-1)*zut1
else
zbutt = 0.0d0
end if
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
if(mod(il-1,32)+1 == txx) then
zbuttt = zbutt
! if(bid == 172951) then
! print *, tx, zbutt, il, txx
! end if
endif
enddo
if(txx <= mod(il-1-1,32)+1) then
! if(bid == 172951) then
! print *, tx, zbuttt, il, txx, ill+txx-1
! end if
istat = atomicadd(zbu_tmp(ktpre+ill+txx-1),zbuttt)
endif
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low_balanced_col3
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low_balanced_row &
(zau,zbu_tmp,subid,ndlp,nstrtlp,ktp,a2,idxp2,kt_prefix,nd,nlf,n_low,la2,kmax,sublen,lenkt)
real*8, dimension(nd) ,intent(out) :: zau
real*8, dimension(lenkt) ,intent(in) :: zbu_tmp
integer*4, dimension(nlf) ,intent(in) :: ndlp,nstrtlp,ktp
integer*4, dimension(kmax,n_low),intent(in) :: idxp2
! integer*4, dimension(n_low,kmax),intent(in) :: idxp2
integer*4, dimension(sublen,2) ,intent(in) :: subid
integer*4, dimension(n_low) ,intent(in) :: kt_prefix
real*8, dimension(la2) :: a2
integer*4,value,intent(in) :: nd,nlf,n_low,la2,kmax,sublen,lenkt
integer*4 :: ii,jj,tx,it,ndl,il,ill,nstrtl,idx2,kt,ktpre
real*8 :: zautt
integer*4 :: istat
! real*8, shared :: tmp(64) ! blocksize should be (64,1,1)
ii = subid(blockIdx%x,1)
jj = subid(blockIdx%x,2)
tx = threadIdx%x
! txx= mod(tx+31,64)+1
it = blockDim%x * (jj-1) + tx
ndl= ndlp(ii)
if(it > ndl) return
zautt = 0.0d0
nstrtl=nstrtlp(ii)
ill =it+nstrtl-1
kt =ktp(ii)
ktpre =kt_prefix(ii)-1
do il=1,kt
idx2 = idxp2(il,ii)
! idx2 = idxp2(ii,il)
zautt= zautt+a2(it+idx2-1)*zbu_tmp(ktpre+il)
enddo
istat = atomicadd(zau(ill),zautt)
end subroutine HACApK_adot_body_lfmtx_cuda_low_balanced_row
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_dense_balanced &
(zau,zu,subid,ndtp,nstrttp,ndlp,nstrtlp,a3,idxp3,nd,nlf,n_low,la3,sublen)
real*8, dimension(nd) ,intent(inout) :: zau
real*8, dimension(nd) ,intent(in) :: zu
integer*4, dimension(nlf) ,intent(in) :: ndtp,nstrttp,ndlp,nstrtlp
integer*4, dimension(nlf) ,intent(in) :: idxp3
integer*4, dimension(sublen,2) ,intent(in) :: subid
real*8, dimension(la3) :: a3
integer*4,value,intent(in) :: nd,nlf,n_low,la3,sublen
integer*4 :: ii,jj,tx,txx,it,il,itt,ill,ndt,nstrtt,ndl,nstrtl,idx3,ty,dimy,dimx,bid
real*8 :: zbutt,zut1,zut2
integer*4 :: istat
bid = blockIdx%x
ii = subid(bid,1)
jj = subid(bid,2)
tx = threadIdx%x
ty = threadIdx%y
dimx = blockDim%x
dimy = blockDim%y
! txx= mod(tx+31,32)+1
it = dimx * (jj-1) + tx
ndt =ndtp(ii)
zbutt = 0.0d0
if(it > ndt) return
nstrtt=nstrttp(ii)
nstrtl=nstrtlp(ii)
ndl =ndlp(ii)
itt =it+nstrtt-1
zut1 =zu(itt)
idx3 = idxp3(ii)
! if(bid < 10) then
! if(tx==1 .and. ty == 1) print *, blockIdx%x, ii, jj, idx3, itt
! endif
do il=ty,ndl,dimy
ill =il+nstrtl-1
zbutt = a3(it+(il-1)*ndt+idx3-1)*zut1
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
if(tx == 1) then
istat = atomicadd(zau(ill),zbutt)
endif
! if(bid < 10) then
! if(tx==1 .and. ty == 1) print *, blockIdx%x, zbutt
! endif
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_dense_balanced
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_dense_balanced_rowmajor &
(zau,zu,subid,ndtp,nstrttp,ndlp,nstrtlp,a3,idxp3,nd,nlf,n_low,n_dense,la3,sublen)
real*8, dimension(nd) ,intent(inout) :: zau
real*8, dimension(nd) ,intent(in) :: zu
integer*4, dimension(nlf) ,intent(in) :: ndtp,nstrttp,ndlp,nstrtlp
integer*4, dimension(nlf) ,intent(in) :: idxp3
integer*4, dimension(sublen) ,intent(in) :: subid
real*8, dimension(la3) :: a3
integer*4,value,intent(in) :: nd,nlf,n_low,n_dense,la3,sublen
integer*4 :: ii,jj,tid,tx,txx,it,il,itt,ill,ndt,nstrtt,ndl,nstrtl,idx3,ty,dimy,dimx,bid,loci
real*8 :: zautt
integer*4 :: istat
bid = blockIdx%x
tx = threadIdx%x
tid= blockDim%x * (bid-1) + tx
if(tid > sublen) return
ii = subid(tid)
il = tid - idxp3(ii-n_low) + 1
zautt = 0.0d0
nstrtt=nstrttp(ii)
nstrtl=nstrtlp(ii)
ill =il+nstrtl-1
ndt =ndtp(ii)
do it=1,ndt
idx3 = idxp3(ii-n_low+n_dense*(it-1))
itt = it+nstrtt-1
zautt = zautt+a3(il+idx3-1)*zu(itt)
! if(ii == n_low+n_dense .and. it > 96) print *, it, il, a3(il+idx3-1)
enddo
istat = atomicadd(zau(ill),zautt)
end subroutine HACApK_adot_body_lfmtx_cuda_dense_balanced_rowmajor
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,la1,la2,la3)
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp1,idxp2,idxp3
real*8, dimension(nd) :: zau,zu
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value :: nd,nlf,n_low,la1,la2,la3
integer*4 :: ii,tx,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt
real*8 :: zbutt,zautt
integer*4 :: istat
real*8, dimension(32) :: zbuttt
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
ii = blockIdx%x
tx = threadIdx%x
! do ii=1, n_low
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx1 = idxp1(ii) ; idx2 = idxp2(ii)
kt=ktp(ii)
do ilil=1,kt,32
do il=ilil,min(kt,ilil+31)
zbutt = 0.0d0
do it=tx,ndt,32
itt=it+nstrtt-1
zbutt=zbutt+a1(it+(il-1)*ndt+idx1-1)*zu(itt)
enddo
! call syncthreads()
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
! zbutt = __shfl(zbutt,1)
! call syncthreads()
! do it=1,ndt
! itt=it+nstrtt-1
! zbutt=zbutt+a1(it+(il-1)*ndt+idx1-1)*zu(itt)
! enddo
zbuttt(il-ilil+1) = zbutt
enddo
do il=ilil,min(kt,ilil+31)
do it=tx,ndl,32
ill=it+nstrtl-1
!zau(ill)=zau(ill)+a2(it+(il-1)*ndl+idx2-1)*zbuttt(il-ilil+1)
zautt=a2(it+(il-1)*ndl+idx2-1)*zbuttt(il-ilil+1)
istat = atomicadd(zau(ill),zautt)
enddo
! do it=1,ndl
! ill=it+nstrtl-1
! !zau(ill)=zau(ill)+a2(it+(il-1)*ndl+idx2-1)*zbuttt(il-ilil+1)
! zautt=a2(it+(il-1)*ndl+idx2-1)*zbuttt(il-ilil+1)
! istat = atomicadd(zau(ill),zautt)
! enddo
enddo
enddo
! enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low2 &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,la1,la2,la3)
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp1,idxp2,idxp3
real*8, dimension(nd) :: zau,zu
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value :: nd,nlf,n_low,la1,la2,la3
integer*4 :: ii,tx,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt
real*8 :: zbutt,zautt
integer*4 :: istat
real*8, dimension(32) :: zbuttt
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
ii = blockIdx%x
tx = threadIdx%x
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx1 = idxp1(ii) ; idx2 = idxp2(ii)
kt=ktp(ii)
do ilil=1,kt,32
do il=ilil,min(kt,ilil+31)
zbutt = 0.0d0
do it=tx,ndt,32
itt=it+nstrtt-1
zbutt=zbutt+a1(it+(il-1)*ndt+idx1-1)*zu(itt)
enddo
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
zbuttt(il-ilil+1) = zbutt
enddo
do it=tx,ndl,32
zautt=0.0d0
do il=ilil,min(kt,ilil+31)
ill=it+nstrtl-1
zautt=zautt+a2(it+(il-1)*ndl+idx2-1)*zbuttt(il-ilil+1)
enddo
istat = atomicadd(zau(ill),zautt)
enddo
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low2
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low22 &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,la1,la2,la3)
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp1,idxp2,idxp3
real*8, dimension(nd) :: zau,zu
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value :: nd,nlf,n_low,la1,la2,la3
integer*4 :: ii,tx,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt
real*8 :: zbutt,zautt
integer*4 :: istat
real*8 :: zbuttt(16)
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
ii = blockIdx%x
tx = threadIdx%x
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx1 = idxp1(ii) ; idx2 = idxp2(ii)
kt=ktp(ii)
do ilil=1,kt,16
do il=ilil,min(kt,ilil+15)
zbuttt(il-ilil+1) = 0.0d0
do it=tx,ndt,32
itt=it+nstrtt-1
zbuttt(il-ilil+1)=zbuttt(il-ilil+1)+a1(it+(il-1)*ndt+idx1-1)*zu(itt)
enddo
enddo
do il=ilil,min(kt,ilil+15)
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),16)
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),8)
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),4)
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),2)
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),1)
enddo
do it=tx,ndl,32
zautt=0.0d0
do il=ilil,min(kt,ilil+15)
ill=it+nstrtl-1
zautt=zautt+a2(it+(il-1)*ndl+idx2-1)*zbuttt(il-ilil+1)
enddo
istat = atomicadd(zau(ill),zautt)
enddo
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low22
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low23 &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,la1,la2,la3)
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp1,idxp2,idxp3
real*8, dimension(nd) :: zau,zu
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value :: nd,nlf,n_low,la1,la2,la3
integer*4 :: ii,tx,ilil,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt,ile
real*8 :: zbutt,zautt,zut
integer*4 :: istat
real*8 :: zbuttt(16)
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
ii = blockIdx%x
tx = threadIdx%x
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx1 = idxp1(ii) ; idx2 = idxp2(ii)
kt=ktp(ii)
do ilil=1,kt,16
ile = min(kt,ilil+15)
do il=ilil,ile
zbuttt(il-ilil+1) = 0.0d0
enddo
do it=tx,ndt,32
itt=it+nstrtt-1
zut = zu(itt)
do il=ilil,ile
zbuttt(il-ilil+1)=zbuttt(il-ilil+1)+a1(it+(il-1)*ndt+idx1-1)*zut
enddo
enddo
do il=ilil,ile
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),16)
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),8)
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),4)
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),2)
zbuttt(il-ilil+1) = zbuttt(il-ilil+1) + __shfl_xor(zbuttt(il-ilil+1),1)
enddo
do it=tx,ndl,32
zautt=0.0d0
do il=ilil,ile
ill=it+nstrtl-1
zautt=zautt+a2(it+(il-1)*ndl+idx2-1)*zbuttt(il-ilil+1)
enddo
istat = atomicadd(zau(ill),zautt)
enddo
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low23
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_low3 &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,la1,la2,la3)
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp1,idxp2,idxp3
real*8, dimension(nd) :: zau,zu
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value :: nd,nlf,n_low,la1,la2,la3
integer*4 :: ii,tx,ilil,il,ill,it,itt,itit,ndl,ndt,ns,nstrtl,nstrtt,idx1,idx2,kt
real*8 :: zbutt,zautt,zuu
integer*4 :: istat
real*8, dimension(32) :: zbuttt
! real*8, shared :: tmp(1) ! blocksize should be (64,1,1)
ii = blockIdx%x
tx = threadIdx%x
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx1 = idxp1(ii) ; idx2 = idxp2(ii)
kt=ktp(ii)
do ilil=1,kt,32
do il=ilil,min(kt,ilil+31)
zbutt = 0.0d0
do it=tx,ndt,32
itt=it+nstrtt-1
zbutt=zbutt+a1(it+(il-1)*ndt+idx1-1)*zu(itt)
enddo
zbutt = zbutt + __shfl_xor(zbutt,16)
zbutt = zbutt + __shfl_xor(zbutt,8)
zbutt = zbutt + __shfl_xor(zbutt,4)
zbutt = zbutt + __shfl_xor(zbutt,2)
zbutt = zbutt + __shfl_xor(zbutt,1)
zbuttt(il-ilil+1) = zbutt
enddo
do it=tx,ndl,32
zautt=0.0d0
do il=ilil,min(kt,ilil+31)
ill=it+nstrtl-1
zautt=zautt+a2(it+(il-1)*ndl+idx2-1)*zbuttt(il-ilil+1)
enddo
istat = atomicadd(zau(ill),zautt)
enddo
enddo
end subroutine HACApK_adot_body_lfmtx_cuda_low3
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_dense &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,n_dense,la1,la2,la3)
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp1,idxp2,idxp3
real*8, dimension(nd) :: zau,zu
real*8, dimension(la1) :: a1
real*8, dimension(la2) :: a2
real*8, dimension(la3) :: a3
integer*4,value :: nd,nlf,n_low,n_dense,la1,la2,la3
integer*4 :: ii,tx,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx3
integer*4 :: istat
real*8 :: zautt
ii = blockIdx%x + n_low
tx = threadIdx%x
! do ii=n_low+1,n_low+n_dense
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx3 = idxp3(ii)
do il=1,ndl
ill=il+nstrtl-1
zautt = 0.0d0
do it=tx,ndt,32
itt=it+nstrtt-1
zautt = zautt + a3(it+(il-1)*ndt+idx3-1)*zu(itt)
enddo
zautt = zautt + __shfl_xor(zautt,16)
zautt = zautt + __shfl_xor(zautt,8)
zautt = zautt + __shfl_xor(zautt,4)
zautt = zautt + __shfl_xor(zautt,2)
zautt = zautt + __shfl_xor(zautt,1)
! do it=1,ndt
! itt=it+nstrtt-1
! zautt = zautt + a3(it+(il-1)*ndt+idx3-1)*zu(itt)
! enddo
! zau(ill)=zau(ill)+zautt
if(tx==1) istat = atomicadd(zau(ill),zautt)
enddo
! enddo
end subroutine HACApK_adot_body_lfmtx_cuda_dense
attributes(global) subroutine HACApK_adot_body_lfmtx_cuda_dense2 &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a3,idxp3, &
nd,nlf,n_low,n_dense,la3)
integer*4, dimension(nlf) :: ndlp,ndtp,nstrtlp,nstrttp,ktp
integer*4, dimension(nlf) :: idxp3
real*8, dimension(nd) :: zau,zu
real*8, dimension(la3) :: a3
integer*4,value :: nd,nlf,n_low,n_dense,la3
integer*4 :: ii,tx,ty,il,ill,it,itt,ndl,ndt,ns,nstrtl,nstrtt,idx3,bydim,bid
integer*4 :: istat
real*8 :: zautt
bid = blockIdx%x
ii = blockIdx%x + n_low
bydim = blockDim%y
tx = threadIdx%x
ty = threadIdx%y
! do ii=n_low+1,n_low+n_dense
ndl =ndlp(ii) ; ndt =ndtp(ii) ; ns=ndl*ndt
nstrtl=nstrtlp(ii); nstrtt=nstrttp(ii)
idx3 = idxp3(ii)
do il=ty,ndl,bydim
ill=il+nstrtl-1
zautt = 0.0d0
do it=tx,ndt,32
itt=it+nstrtt-1
zautt = zautt + a3(it+(il-1)*ndt+idx3-1)*zu(itt)
enddo
zautt = zautt + __shfl_xor(zautt,16)
zautt = zautt + __shfl_xor(zautt,8)
zautt = zautt + __shfl_xor(zautt,4)
zautt = zautt + __shfl_xor(zautt,2)
zautt = zautt + __shfl_xor(zautt,1)
! do it=1,ndt
! itt=it+nstrtt-1
! zautt = zautt + a3(it+(il-1)*ndt+idx3-1)*zu(itt)
! enddo
! zau(ill)=zau(ill)+zautt
if(tx==1) istat = atomicadd(zau(ill),zautt)
! if(bid < 10) then
! if(tx==1 .and. ty == 1) print *, blockIdx%x, zautt
! endif
enddo
! enddo
end subroutine HACApK_adot_body_lfmtx_cuda_dense2
subroutine HACApK_adot_body_lfmtx_cuda_wrapper &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,n_dense)
integer*4, dimension(:),device :: ndlp,ndtp,nstrtlp,nstrttp,ktp
#if OPT>=7
integer*4, dimension(:,:),device :: idxp1,idxp2
integer*4, dimension(:),device :: idxp3
#else
integer*4, dimension(:),device :: idxp1,idxp2,idxp3
#endif
real*8, device :: zau(*),zu(*)
real*8, dimension(:),device :: a1,a2,a3
integer*4 :: nd,nlf,n_low,n_dense
integer*4 :: istat
type(dim3) :: dimGrid, dimBlock, dimBlock2, dimBlock3
integer*4 :: la1,la2,la3,kmax
! real*8 :: zau_tmp(nd)
logical,save :: flag = .true.
integer(kind=cuda_stream_kind) :: stream1, stream2
la1 = size(a1); la2 = size(a2); la3 = size(a3)
kmax = size(idxp1,1)
! print *, "kmax",kmax
! if(flag == .true.) then
! print *, "flop1 = ", (4.0d0*(la1+la2))/1000000000.0d0
! print *, "byte1 = ", (8.0d0*(la1+la2)+3.0d0*nd)/1000000000.0d0
! print *, "flop2 = ", (4.0d0*(la3))/1000000000.0d0
! print *, "byte2 = ", (8.0d0*(la3)+3.0d0*nd)/1000000000.0d0
! flag = .false.
! endif
istat = cudaStreamCreate(stream1)
istat = cudaStreamCreate(stream2)
dimGrid = dim3(n_low, 1, 1)
! dimGrid = dim3(1, 1, 1)
dimBlock = dim3(32, 1, 1)
dimBlock2 = dim3(32, 2, 1)
dimBlock3 = dim3(64, 1, 1)
! dimBlock = dim3(1, 1, 1)
#if OPT>=7
call HACApK_adot_body_lfmtx_cuda_low7<<<dimGrid,dimBlock3>>> &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,idxp1,idxp2, &
nd,nlf,n_low,la1,la2,kmax)
! stop
#else
! call HACApK_adot_body_lfmtx_cuda_low0<<<dimGrid,dimBlock>>> &
! call HACApK_adot_body_lfmtx_cuda_low0_2<<<dimGrid,dimBlock2>>> &
call HACApK_adot_body_lfmtx_cuda_low0_3<<<dimGrid,dimBlock3>>> &
! call HACApK_adot_body_lfmtx_cuda_low2<<<dimGrid,dimBlock>>> &
! call HACApK_adot_body_lfmtx_cuda_low23<<<dimGrid,dimBlock>>> &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a1,a2,a3,idxp1,idxp2,idxp3, &
nd,nlf,n_low,la1,la2,la3)
#endif
dimGrid = dim3(n_dense, 1, 1)
! dimBlock = dim3(1, 1, 1)
! call HACApK_adot_body_lfmtx_cuda_dense<<<dimGrid,dimBlock>>> &
call HACApK_adot_body_lfmtx_cuda_dense2<<<dimGrid,dimBlock2>>> &
(zau,zu,ndlp,ndtp,nstrtlp,nstrttp,ktp,a3,idxp3, &
nd,nlf,n_low,n_dense,la3)
istat = cudaDeviceSynchronize()