-
Notifications
You must be signed in to change notification settings - Fork 0
/
som.c
1464 lines (1151 loc) · 38.3 KB
/
som.c
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
#include "pmsis.h"
#include "config.h"
#include <math.h>
#include <stdbool.h>
#define ABS(x) (((x) < 0)? (-(x)) : (x))
#ifdef FP32
Pixel FABS(Pixel x){
int temp = (*((int *)(&x)) & 0X7FFFFFFF);
return *((Pixel *)(&temp));
}
#else
#ifdef VECTORIAL
#ifdef FP8
VPixel FABS(VPixel x){
int temp = (*((int *)(&x)) & 0X7F7F7F7F);
return *((VPixel *)(&temp));
}
#else
VPixel FABS(VPixel x){
int temp = (*((int *)(&x)) & 0X7FFF7FFF);
return *((VPixel *)(&temp));
}
#endif
#else
#ifdef FP8
Pixel FABS(Pixel x){
char temp = (*((char *)(&x)) & 0X7F);
return *((Pixel *)(&temp));
}
#else
Pixel FABS(Pixel x){
short temp = (*((short *)(&x)) & 0X7FFF);
return *((Pixel *)(&temp));
}
#endif
#endif
#endif
#ifdef FP32
#define MANTISSA_BITS 23
#define EXPONENT_BITS 8
#define EXPONENT_BIAS 127
#define INT_TYPE int
#endif
#ifdef FP16
#define MANTISSA_BITS 10
#define EXPONENT_BITS 5
#define EXPONENT_BIAS 15
#define INT_TYPE short
#endif
#ifdef FP16ALT
#define MANTISSA_BITS 7
#define EXPONENT_BITS 8
#define EXPONENT_BIAS 127
#define INT_TYPE short
#endif
#ifdef FP8
#define MANTISSA_BITS 2
#define EXPONENT_BITS 5
#define EXPONENT_BIAS 15
#define INT_TYPE char
#endif
// Compute the value 2^(-x)
Pixel neg_power_of_two(unsigned INT_TYPE x) {
INT_TYPE val = EXPONENT_BIAS - x;
unsigned INT_TYPE exp = *((unsigned INT_TYPE *) &val);
unsigned INT_TYPE res = exp << MANTISSA_BITS;
return *((Pixel *) &res);
}
#ifndef FABRIC
DATA_LOCATION Pixel minimum[NUM_CORES];
DATA_LOCATION int myindex[NUM_CORES];
DATA_LOCATION Pixel min_start =448;
#if defined (once_load)
#ifndef IN_ORDER
DATA_LOCATION Pixel SOM[R_g * w_b * M_g];
DATA_LOCATION Pixel INP[2*R_g*inp_b*M_g];
#else
DATA_LOCATION Pixel SOM[ w_b * M_g];//2* w_b * M
DATA_LOCATION Pixel INP[2*inp_b*M_g];//2*inp_b*M
#endif
#elif defined(IN_ORDER)
DATA_LOCATION Pixel SOM[2 *w_b * M_g];//2* w_b * M
DATA_LOCATION Pixel INP[2 *inp_b*M_g];//2*inp_b*M
#else
DATA_LOCATION Pixel SOM[2 *R_g *w_b * M_g];//2* w_b * M
DATA_LOCATION Pixel INP[2* R_g *inp_b*M_g];//2*inp_b*M
#endif
int find_BMU_index(Pixel * __restrict__ x, Pixel * __restrict__ SOM_N, int N, int M, int s){
pi_cl_dma_cmd_t commandA;
pi_cl_dma_cmd_t commandB;
#ifdef VECTORIAL
VPixel temp;
VPixel Inpv0, Inpv1, Sv0, Sv1;
Pixel sum =0;
#else
Pixel temp = 0;
#endif
int location =0;
int start_par =0;
int end =0;
int ava_cores =0;
int x_offset = 0;
Pixel minu = min_start;
#ifdef once_load
int double_buffer =1;
#else
int double_buffer =2;
#endif
int core_id = pi_core_id();
minimum[core_id] =min_start;
// Load the first parts of SOMs
#ifndef once_load
if (core_id == 0) // load the first som
{
pi_cl_dma_cmd_2d((int) (SOM_N ) , (int) SOM ,w_b*M*sizeof(SOM_N[0]),
w_b*M*sizeof(SOM_N[0]),w_b*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandB);
pi_cl_dma_cmd_wait(&commandB);
}
#ifndef IN_ORDER // load the second som if we run both SOMs in parallel
if (core_id == 0)
{
pi_cl_dma_cmd_2d((int) (SOM_N + ( N * M) ), (int) (SOM + double_buffer * w_b * M ) ,w_b*M*sizeof(SOM_N[0]),
w_b*M*sizeof(SOM_N[0]),w_b*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandA);
pi_cl_dma_cmd_wait(&commandA);
}
#endif
#endif
#ifdef IN_ORDER
ava_cores = NUM_CORES;
#else
ava_cores = NUM_CORES/2;
#endif
// offset for the correct block of the L1 input for each cores
x_offset =((core_id/(ava_cores)) * 2 * inp_b * M);
#ifdef once_load
int offset =0;
int offset1 = w_b;
#endif
#ifndef once_load // we need to iterate on the whole SOM network and process wb data
for (int counter =0; counter < N / w_b; counter++)
{
#endif
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
// load the next part of SOMs with DMA while processing the current data
#ifndef once_load
if (core_id==0 && counter != (N/w_b)-1)
{
// printf("SOM data for next iter from %d of input l2 put %d l1\n", (N * M) + ((counter +1) * w_b*M),(((counter +1) % 2)+ 2)* w_b*M);
pi_cl_dma_cmd_2d((int) (SOM_N + ((counter +1) * w_b*M)) , (int) (SOM + ((counter +1) % 2)* w_b*M) ,w_b*M*sizeof(SOM[0]),
w_b*M*sizeof(SOM[0]),w_b*M*sizeof(SOM[0]),PI_CL_DMA_DIR_EXT2LOC, &commandA);
#ifndef IN_ORDER
pi_cl_dma_cmd_2d((int) (SOM_N + (N * M) + ((counter +1) * w_b*M)) , (int) (SOM + (((counter +1) % 2)+ 2)* w_b*M) ,w_b*M*sizeof(SOM[0]),
w_b*M*sizeof(SOM[0]),w_b*M*sizeof(SOM[0]),PI_CL_DMA_DIR_EXT2LOC, &commandB);
#endif
}
#endif
#ifndef once_load
#ifndef IN_ORDER
int offset =(int)(counter / 2) * 2 * w_b;
int offset1 = 2 * w_b;
#else
int offset =(int)(counter / 2) * 2 * w_b;
int offset1 = 0;
#endif
// #else
// #ifndef IN_ORDER
// int offset =0;
// int offset1 = w_b;
// #else
// int offset =0;
// int offset1 = w_b;
// #endif
#endif
#ifndef once_load
#ifdef IN_ORDER
#if NUM_CORES > 1
int blockSize = (w_b+ava_cores-1)/ava_cores;
start_par = ((counter %2) * w_b) + core_id*blockSize ;
end = start_par + blockSize < (2*w_b)? start_par + blockSize : (2*w_b);
#else
start_par = ((counter %2) * w_b);
end = ((counter %2 ) +1) * w_b;
#endif
#else
#if NUM_CORES > 1
int blockSize = (w_b+ava_cores-1)/ava_cores;
if (core_id < (ava_cores)){
start_par = ((counter %2) * w_b) + core_id*blockSize ;
end = start_par + blockSize < (2*w_b)? start_par + blockSize : (2*w_b);
}
else
{
start_par = ((counter %2) * w_b) + (core_id -(ava_cores)) *blockSize + (2*w_b) ;
end = start_par + blockSize < (4*w_b)? start_par + blockSize : (4*w_b);
}
#else // only one core
start_par = ((counter %2) * w_b);
end = ((counter %2 ) +1) * w_b;
#endif
#endif
#else
#ifndef IN_ORDER
#if NUM_CORES > 1
int blockSize = (w_b+(ava_cores)-1)/(ava_cores);
if (core_id < (ava_cores)){
start_par = core_id*blockSize ;
end = start_par + blockSize < (w_b)? start_par + blockSize : (w_b);
}
else
{
start_par = (core_id -(ava_cores)) *blockSize + (w_b) ;
end = start_par + blockSize < (2*w_b)? start_par + blockSize : (2*w_b);
}
#else // we don't have this case, because we don't run both SOMs on one core with ndef in_order flag
start_par = 0;
end = 0;
#endif
#else // IN_ORDER
#if NUM_CORES > 1
int blockSize = (w_b+ava_cores-1)/ava_cores;
start_par = core_id*blockSize ;
end = start_par + blockSize < (w_b)? start_par + blockSize : (w_b);
#else
start_par = 0;
end = w_b;
#endif
#endif
#endif
for (int i = start_par; i < end; i++)
{
#ifdef VECTORIAL
#ifdef FP8
temp = (VPixel) {0, 0, 0, 0};
#if M_g != 8
for (int j = 0; j < M ; j+=8)
#endif
{
int j =0;
Inpv0 = *((VPixel *) &x[x_offset + j]);
Sv0 = *((VPixel *) &SOM[ i * M + j]);
Inpv1 = *((VPixel *) &x[x_offset + j+4]);
Sv1 = *((VPixel *) &SOM[ i * M + j+4]);
temp += FABS(Sv0 - Inpv0);
temp += FABS(Sv1 - Inpv1);
}
sum = temp[0] + temp[1] + temp[2] + temp[3];
if (sum < minimum[core_id]){
minimum[core_id] = sum;
myindex[core_id] = ( offset) + (i - ((core_id / (ava_cores)) * offset1)); //((int)(counter / 2) * 2 * w_b) + i;
}
#else
temp = (VPixel) {0, 0};
for (int j = 0; j < M ; j+=4)
{
Inpv0 = *((VPixel *) &x[x_offset + j]);
Sv0 = *((VPixel *) &SOM[ i * M + j]);
Inpv1 = *((VPixel *) &x[x_offset + j+2]);
Sv1 = *((VPixel *) &SOM[ i * M + j+2]);
temp += FABS(Sv0 - Inpv0);
temp += FABS(Sv1 - Inpv1);
}
sum = temp[0] + temp[1];
if (sum < minimum[core_id]){
minimum[core_id] = sum;
myindex[core_id] = ( offset) + (i - ((core_id / (ava_cores)) * offset1));
}
#endif
#else
temp = 0;
for (int j = 0; j < M ; j+=2)
{
temp += FABS(SOM[ i * M + j] - x[x_offset + j]);
temp += FABS(SOM[ i * M + j+1] - x[x_offset + j+1]);
}
if (temp < minimum[core_id]){
minimum[core_id] = temp;
myindex[core_id] = ( offset) + (i - ((core_id / (ava_cores)) * offset1));
}
#endif
// location = ((int)(counter / 2) * 2 * w_b) + i;
// }
}
#ifndef once_load
if (core_id==0)
{
pi_cl_dma_cmd_wait(&commandA);
#ifndef IN_ORDER
pi_cl_dma_cmd_wait(&commandB);
#endif
}
}
#endif
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
#ifndef FP8
for (int i = (core_id/ (ava_cores))*(ava_cores); i < ((core_id/ (ava_cores)) +1)*(ava_cores); i++)
{
if (minimum[i]< minu)
{
minu = minimum[i];
location = myindex[i];
}
}
#else
for (int i = (core_id/ (ava_cores))*(ava_cores); i < ((core_id/ (ava_cores)) +1)*(ava_cores); i++)
{
if (minimum[i]< minu)
{
minu = minimum[i];
location = myindex[i];
}
else if (minimum[i] == minu && myindex[i] < location)
{
minu = minimum[i];
location = myindex[i];
}
}
#endif
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
return location;
}
/*Pixel find_BMU_value(Pixel * __restrict__ x, Pixel * __restrict__ SOM, int N, int M){
Pixel min =min_start;
for (int i = 0; i < N; i++) {
Pixel temp = 0;
for (int j = 0; j < M; j++) {
//printf("i %d j %d som %f x %f\n", i, j, SOM[ i * M + j], x[j] );
temp += FABS(SOM[ i * M + j] - x[j]);
}
// printf("temp is %f \n", temp);
if (temp < min){
min = temp;
}
}
// printf("min is %f \n", min);
return min;
}*/
void __attribute__ ((noinline)) update_weights(Pixel * __restrict__ x, Pixel * __restrict__ SOM_N, int N, int M, Pixel beta, int location, int s) {
int dist =0;
Pixel power=0;
#ifdef FP32
int thereshold =19;
#else
int thereshold = 14;
#endif
#ifdef VECTORIAL
VPixel *output0, *output1;
VPixel Inpv0, Inpv1, Sv0, Sv1;
Pixel sum =0;
VPixel temp0,temp1;
VPixel pow_vec;
#else
Pixel temp = 0;
#endif
Pixel dist_func = 0.0;
Pixel dist_func1 = 0.0;
Pixel pow_temp=0;
int start =0;//0
int length = 127;
int from =0;
int extra = 0;
int adder =0;
int base =0;
int start_par = 0;
int end=0;
int ava_cores =0;
int x_offset = 0;
bool flag_begin = false;
bool flag_end = false;
pi_cl_dma_cmd_t commandB;
pi_cl_dma_cmd_t commandA;
int core_id = pi_core_id();
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
#ifdef IN_ORDER
ava_cores = NUM_CORES;
#else
ava_cores = NUM_CORES/2;
#endif
x_offset =((core_id/(ava_cores)) * 2 * inp_b * M);
if (location <thereshold){
// from .... N, start(0) ..location ... length
// from .... N is done in the first loop
start = ((core_id / (ava_cores)) * N) + 0;
length = location + thereshold +1 ;//63+1;
from = ((core_id / (ava_cores)) * N) + N - (thereshold - location);
extra =(thereshold - location) ;
flag_begin = true;
#ifndef once_load
base =((core_id / (ava_cores)) * thereshold); // base address for the the extra part
adder = ((((core_id / (ava_cores)) *3 ) +2) * thereshold);// address for the part which is greater than location
#ifndef IN_ORDER
if (core_id == (ava_cores)-1)
{
pi_cl_dma_cmd_2d((int) (SOM_N + (from*M)) , (int) (SOM + base*M) , extra*M*sizeof(SOM_N[0]),
extra*M*sizeof(SOM_N[0]),extra*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandA);
pi_cl_dma_cmd_wait(&commandA);
}
else if (core_id == (ava_cores))
{
pi_cl_dma_cmd_2d((int) (SOM_N + ( from*M)) , (int) (SOM + base*M) , extra*M*sizeof(SOM_N[0]),
extra*M*sizeof(SOM_N[0]),extra*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandB);
pi_cl_dma_cmd_wait(&commandB);
}
#else
if (core_id == 0)
{
pi_cl_dma_cmd_2d((int) (SOM_N + (from*M)) , (int) (SOM + base*M) , extra*M*sizeof(SOM_N[0]),
extra*M*sizeof(SOM_N[0]),extra*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandA);
pi_cl_dma_cmd_wait(&commandA);
}
#endif
#else
base = from;// the extra part
adder = ((core_id / (ava_cores)) * w_b) + 0 ; // the part > location Note: N =w_b in this case
#endif
}
else if (location > N - thereshold)
{
// start .... location .. length (N), from(0) .... extra,
//from ... extra is doen by the first loop
start = ((core_id / (ava_cores)) * N) + location -thereshold;
length = N - location +thereshold;;
from =((core_id / (ava_cores)) * N) + 0;
extra = thereshold - (N - location)+1;
flag_end = true;
#ifndef once_load
base =((core_id / (ava_cores)) * thereshold); // base address for the the extra part
adder = ((((core_id / (ava_cores)) *3 ) +2) * thereshold);// base for the main part
#ifndef IN_ORDER
if (core_id == (ava_cores)-1)
{
pi_cl_dma_cmd_2d((int) (SOM_N + (from*M)) , (int) (SOM + base*M) , extra*M*sizeof(SOM_N[0]),
extra*M*sizeof(SOM_N[0]),extra*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandA);
pi_cl_dma_cmd_wait(&commandA);
}
else if (core_id == (ava_cores)){
pi_cl_dma_cmd_2d((int) (SOM_N + (from*M)) , (int) (SOM + base*M) , extra*M*sizeof(SOM_N[0]),
extra*M*sizeof(SOM_N[0]),extra*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandB);
pi_cl_dma_cmd_wait(&commandB);
}
#else
if (core_id == 0)
{
pi_cl_dma_cmd_2d((int) (SOM_N + (from*M)) , (int) (SOM + base*M) , extra*M*sizeof(SOM_N[0]),
extra*M*sizeof(SOM_N[0]),extra*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandA);
pi_cl_dma_cmd_wait(&commandA);
}
#endif
#else
base =from;
adder = start;
#endif
}
else
{
// start ...location... lenght
start= ((core_id / (ava_cores)) * N) + location - thereshold;
length = 2 * thereshold +1;//127;
from =0;
extra =0;
#ifndef once_load
#ifdef IN_ORDER
adder = thereshold;//because of small block size, some times we don't have 2 * thereshold size. for example, block =64 and the threshold is 38 and length =39 -> invlaid access
#else
adder = ((((core_id / (ava_cores)) *3 ) +2) * thereshold);// because of double buffering
#endif
#else
adder = start;
#endif
}
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
#ifndef once_load
#ifndef IN_ORDER
if (core_id == (ava_cores)-1)
{
pi_cl_dma_cmd_2d((int) (SOM_N + (start*M)) , (int) (SOM + (adder * M )) , length*M*sizeof(SOM_N[0]),
length*M*sizeof(SOM_N[0]),length*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandB);
}
else if (core_id == (ava_cores))
{
pi_cl_dma_cmd_2d((int) (SOM_N + (start*M)) , (int) (SOM + (adder * M )) , length*M*sizeof(SOM_N[0]),
length*M*sizeof(SOM_N[0]),length*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandB);
}
#else
if (core_id == 0)
{
pi_cl_dma_cmd_2d((int) (SOM_N + (start*M)) , (int) (SOM + (adder * M )) , length*M*sizeof(SOM_N[0]),
length*M*sizeof(SOM_N[0]),length*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_EXT2LOC, &commandB);
}
#endif
#endif
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
if (flag_begin)
{
#if NUM_CORES > 1
int blockSize =(extra+(ava_cores)-1)/(ava_cores);
if (core_id < (ava_cores))
{
start_par = core_id*blockSize;
end = start_par + blockSize < extra? start_par + blockSize : extra;
}
else
{
start_par = (core_id - (ava_cores)) *blockSize ;
end = start_par + blockSize < extra? start_par + blockSize : extra;
}
#else
start_par = 0;
end = extra;
#endif
for (int i =start_par; i < end; i ++)
{
dist = thereshold - i;
pow_temp = (-beta) * neg_power_of_two(dist);
#ifdef VECTORIAL
#ifdef FP8
pow_vec = (VPixel) {pow_temp, pow_temp, pow_temp,pow_temp};
#ifndef No_for
for (int j = 0; j < M ; j+=8)
#else
int j =0;
#endif
{
Inpv0 = *((VPixel *) &x[x_offset + j]);
output0 =(VPixel *) &SOM[(base + i) * M + j];
Sv0 = *output0;
Inpv1 = *((VPixel *) &x[x_offset + j+4]);
output1 =(VPixel *) &SOM[(base + i) * M + j+4];
Sv1 = *output1;
temp0 = pow_vec * (Sv0 - Inpv0);
temp1 = pow_vec * (Sv1 - Inpv1);
Sv0 = Sv0 + temp0;
Sv1 = Sv1 + temp1;
*output0 = Sv0;
*output1 = Sv1;
}
#else // float16 or bfloat16 vectorization
pow_vec =__builtin_pulp_floats_to_vf16(pow_temp, pow_temp);
for (int j = 0; j < M ; j+=4)
{
Inpv0 = *((VPixel *) &x[x_offset + j]);
output0 =(VPixel *) &SOM[(base + i) * M + j];
Sv0 = *output0;
Inpv1 = *((VPixel *) &x[x_offset + j+2]);
output1 =(VPixel *) &SOM[(base + i) * M + j+2];
Sv1 = *output1;
temp0 = pow_vec * (Sv0 - Inpv0);
temp1 = pow_vec * (Sv1 - Inpv1);
Sv0 = Sv0 + temp0;
Sv1 = Sv1 + temp1;
*output0 = Sv0;
*output1 = Sv1;
}
#endif
#else
for (int j =0; j< M; j+=2)
{
dist_func = (pow_temp * (SOM[(base + i) * M + j] - x[x_offset + j]));
SOM[(base + i) * M + j] = SOM[(base + i )* M + j] + dist_func;
dist_func1 = (pow_temp * (SOM[(base + i) * M + j+1] - x[x_offset + j+1]));
SOM[(base + i) * M + j+1] = SOM[(base + i) * M + j+1] + dist_func1;
}
#endif
}
}
if (flag_end)
{
#if NUM_CORES > 1
int blockSize = (extra+(ava_cores)-1)/(ava_cores);
if (core_id < (ava_cores))
{
start_par = core_id*blockSize;
end = start_par + blockSize < extra? start_par + blockSize : extra;
}
else
{
start_par = (core_id -(ava_cores)) *blockSize ;
end = start_par + blockSize < extra? start_par + blockSize : extra;
}
#else
start_par = 0;
end = extra;
#endif
for (int i =start_par; i < end; i ++)
{
dist = i + (N - location);
pow_temp = (-beta) * neg_power_of_two(dist);
#ifdef VECTORIAL
#ifdef FP8
pow_vec = (VPixel) {pow_temp, pow_temp, pow_temp,pow_temp};
#ifndef No_for
for (int j = 0; j < M ; j+=8)
#else
int j =0;
#endif
{
Inpv0 = *((VPixel *) &x[x_offset + j]);
output0 =(VPixel *) &SOM[(base + i) * M + j];
Sv0 = *output0;
Inpv1 = *((VPixel *) &x[x_offset + j+4]);
output1 =(VPixel *) &SOM[(base + i) * M + j+4];
Sv1 = *output1;
temp0 = pow_vec * (Sv0 - Inpv0);
temp1 = pow_vec * (Sv1 - Inpv1);
Sv0 = Sv0 + temp0;
Sv1 = Sv1 + temp1;
*output0 = Sv0;
*output1 = Sv1;
}
#else
pow_vec =__builtin_pulp_floats_to_vf16(pow_temp, pow_temp);
for (int j = 0; j < M ; j+=4)
{
Inpv0 = *((VPixel *) &x[x_offset + j]);
output0 =(VPixel *) &SOM[(base + i) * M + j];
Sv0 = *output0;
Inpv1 = *((VPixel *) &x[x_offset + j+2]);
output1 =(VPixel *) &SOM[(base + i) * M + j+2];
Sv1 = *output1;
temp0 = pow_vec * (Sv0 - Inpv0);
temp1 = pow_vec * (Sv1 - Inpv1);
Sv0 = Sv0 + temp0;
Sv1 = Sv1 + temp1;
*output0 = Sv0;
*output1 = Sv1;
}
#endif
#else
for (int j =0; j< M; j+=2)
{
dist_func = (pow_temp * (SOM[(base + i) * M + j] - x[x_offset + j]));
SOM[(base + i) * M + j] = SOM[(base + i) * M + j] + dist_func;
dist_func1 = (pow_temp * (SOM[(base + i) * M + j+1] - x[x_offset + j+1]));
SOM[(base + i) * M + j+1] = SOM[(base + i) * M + j+1] + dist_func1;
}
#endif
}
}
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
#ifndef once_load
#ifndef IN_ORDER
if (core_id == (ava_cores)-1)
{
pi_cl_dma_cmd_wait(&commandB);
}
else if (core_id == (ava_cores))
{
pi_cl_dma_cmd_wait(&commandB);
}
#else
if (core_id == 0)
{
pi_cl_dma_cmd_wait(&commandB);
}
#endif
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
if (flag_begin || flag_end)
{
#ifndef IN_ORDER
if (core_id == (ava_cores)-1)
{
pi_cl_dma_cmd_2d((int) (SOM_N + (from*M)) , (int) (SOM+(base*M)) , extra*M*sizeof(SOM_N[0]),
extra*M*sizeof(SOM_N[0]),extra*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_LOC2EXT, &commandA);
}
else if (core_id == (ava_cores))
{
pi_cl_dma_cmd_2d((int) (SOM_N + (from*M)) , (int) (SOM+(base*M)) , extra*M*sizeof(SOM_N[0]),
extra*M*sizeof(SOM_N[0]),extra*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_LOC2EXT, &commandA);
}
#else
if (core_id == 0)
{
pi_cl_dma_cmd_2d((int) (SOM_N + (from*M)) , (int) (SOM+(base*M)) , extra*M*sizeof(SOM_N[0]),
extra*M*sizeof(SOM_N[0]),extra*M*sizeof(SOM_N[0]),PI_CL_DMA_DIR_LOC2EXT, &commandA);
}
#endif
}
#endif
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
#if NUM_CORES > 1
int blockSize = (length+(ava_cores)-1)/(ava_cores);
if (core_id < (ava_cores))
{
start_par = core_id*blockSize;
end = start_par + blockSize < length? start_par + blockSize : length;
}
else
{
start_par = (core_id -(ava_cores)) *blockSize ;
end = start_par + blockSize < length? start_par + blockSize : length;
}
#else
start_par = 0;
end = length;
#endif
for (int i = start_par; i < end; i++)
{
if( location < thereshold){
dist = ABS(i - location);
}
else{
dist = ABS(i - thereshold);
}
pow_temp = (-beta) * neg_power_of_two(dist);
#ifdef VECTORIAL
#ifdef FP8
pow_vec =__builtin_pulp_floats_to_vf8(pow_temp, pow_temp, pow_temp,pow_temp);
#ifndef No_for
for (int j = 0; j < M ; j+=8)
#else
int j =0;
#endif
{
Inpv0 = *((VPixel *) &x[x_offset + j]);
output0 =(VPixel *) &SOM[(i+adder) * M + j];
Sv0 = *output0;
Inpv1 = *((VPixel *) &x[x_offset + j+4]);
output1 =(VPixel *) &SOM[(i+adder) * M + j+4];
Sv1 = *output1;
temp0 = pow_vec * (Sv0 - Inpv0);
temp1 = pow_vec * (Sv1 - Inpv1);
Sv0 = Sv0 + temp0;
Sv1 = Sv1 + temp1;
*output0 = Sv0;
*output1 = Sv1;
}
#else
pow_vec =__builtin_pulp_floats_to_vf16(pow_temp, pow_temp);
for (int j = 0; j < M ; j+=4)
{
Inpv0 = *((VPixel *) &x[x_offset + j]);
output0 =(VPixel *) &SOM[(i+adder) * M + j];
Sv0 = *output0;
Inpv1 = *((VPixel *) &x[x_offset + j+2]);
output1 =(VPixel *) &SOM[(i+adder) * M + j+2];
Sv1 = *output1;
temp0 = pow_vec * (Sv0 - Inpv0);
temp1 = pow_vec * (Sv1 - Inpv1);
Sv0 = Sv0 + temp0;
Sv1 = Sv1 + temp1;
*output0 = Sv0;
*output1 = Sv1;
}
#endif
#else
for (int j =0; j< M; j+=2)
{
dist_func = (pow_temp * (SOM[(i+adder) * M + j] - x[x_offset + j]));
SOM[(i+adder) * M + j] = SOM[(i+adder) * M + j] + dist_func;
dist_func1 = (pow_temp * (SOM[(i+adder) * M + j+1] - x[x_offset + j+1]));
SOM[(i+adder) * M + j+1] = SOM[(i+adder) * M + j+1] + dist_func1;
}
#endif
}
#if NUM_CORES > 1
pi_cl_team_barrier();
#endif
#ifndef once_load // load and save data to L2
if (flag_begin || flag_end)
{
#ifndef IN_ORDER
if (core_id == (ava_cores)-1)
{
pi_cl_dma_cmd_wait(&commandA);
}
else if (core_id == (ava_cores))
{
pi_cl_dma_cmd_wait(&commandA);
}
#else
if (core_id == 0)
{
pi_cl_dma_cmd_wait(&commandA);
}
#endif
}
#ifndef IN_ORDER
if (core_id == (ava_cores)-1)
{
pi_cl_dma_cmd_2d((int) (SOM_N + ((start)*M)) , (int) (SOM+ (adder * M)) , length*M*sizeof(SOM[0]),
length*M*sizeof(SOM[0]),length*M*sizeof(SOM[0]),PI_CL_DMA_DIR_LOC2EXT, &commandA);
pi_cl_dma_cmd_wait(&commandA);
}
else if (core_id == (ava_cores))
{
pi_cl_dma_cmd_2d((int) (SOM_N + ((start)*M)) , (int) (SOM+ (adder * M)) , length*M*sizeof(SOM[0]),
length*M*sizeof(SOM[0]),length*M*sizeof(SOM[0]),PI_CL_DMA_DIR_LOC2EXT, &commandB);
pi_cl_dma_cmd_wait(&commandB);
}
#else
if (core_id == 0)
{
pi_cl_dma_cmd_2d((int) (SOM_N + ((start)*M)) , (int) (SOM+ (adder * M)) , length*M*sizeof(SOM[0]),
length*M*sizeof(SOM[0]),length*M*sizeof(SOM[0]),PI_CL_DMA_DIR_LOC2EXT, &commandA);