forked from KlausT/ccminer
-
Notifications
You must be signed in to change notification settings - Fork 0
/
cuda_helper.h.orig
1115 lines (1015 loc) · 28.7 KB
/
cuda_helper.h.orig
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
#ifndef CUDA_HELPER_H
#define CUDA_HELPER_H
#include <cuda.h>
#include <cuda_runtime.h>
#ifdef __cplusplus
#include <cstdint>
#include <cstdio>
using namespace std;
#else
#include <stdint.h>
#endif
#ifdef __INTELLISENSE__
#define NOASM
/* reduce vstudio warnings (__byteperm, blockIdx...) */
#include <device_functions.h>
#include <device_launch_parameters.h>
#define __launch_bounds__(max_tpb, min_blocks)
uint32_t __byte_perm(uint32_t x, uint32_t y, uint32_t z);
uint32_t __shfl(uint32_t x, uint32_t y, uint32_t z);
uint32_t atomicExch(uint32_t *x, uint32_t y);
uint32_t atomicAdd(uint32_t *x, uint32_t y);
void __syncthreads(void);
void __threadfence(void);
#define __ldg(x) (*(x))
#endif
#ifndef MAX_GPUS
#define MAX_GPUS 8
#endif
extern int device_map[MAX_GPUS];
extern long device_sm[MAX_GPUS];
extern cudaStream_t gpustream[MAX_GPUS];
extern bool stop_mining;
extern volatile bool mining_has_stopped[MAX_GPUS];
// common functions
extern void cuda_check_cpu_init(int thr_id, uint32_t threads);
extern void cuda_check_cpu_setTarget(const void *ptarget, int thr_id);
extern void cuda_check_cpu_setTarget_mod(const void *ptarget, const void *ptarget2);
extern uint32_t cuda_check_hash(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_inputHash);
extern uint32_t cuda_check_hash_suppl(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_inputHash, uint32_t foundnonce);
extern void cudaReportHardwareFailure(int thr_id, cudaError_t error, const char* func);
#ifndef __CUDA_ARCH__
// define blockDim and threadIdx for host
extern const dim3 blockDim;
extern const uint3 threadIdx;
#endif
extern cudaError_t MyStreamSynchronize(cudaStream_t stream, int situation, int thr_id);
#ifndef SPH_C32
#define SPH_C32(x) ((x ## U))
// #define SPH_C32(x) ((uint32_t)(x ## U))
#endif
#ifndef SPH_C64
#define SPH_C64(x) ((x ## ULL))
// #define SPH_C64(x) ((uint64_t)(x ## ULL))
#endif
#ifndef SPH_T32
#define SPH_T32(x) (x)
// #define SPH_T32(x) ((x) & SPH_C32(0xFFFFFFFF))
#endif
#ifndef SPH_T64
#define SPH_T64(x) (x)
// #define SPH_T64(x) ((x) & SPH_C64(0xFFFFFFFFFFFFFFFF))
#endif
#if defined _MSC_VER && !defined __CUDA_ARCH__
#define ROTL32c(x, n) _rotl(x, n)
#define ROTR32c(x, n) _rotr(x, n)
#else
#define ROTL32c(x, n) ((x) << (n)) | ((x) >> (32 - (n)))
#define ROTR32c(x, n) ((x) >> (n)) | ((x) << (32 - (n)))
#endif
#ifndef __CUDA_ARCH__
#define ROTR32(x, n) ROTR32c(x, n)
#define ROTL32(x, n) ROTL32c(x, n)
#else
#if __CUDA_ARCH__ < 320
// Kepler (Compute 3.0)
__device__ __forceinline__ uint32_t ROTR32(const uint32_t x, const uint32_t n)
{
return (x >> n) | (x << (32 - n));
}
__device__ __forceinline__ uint32_t ROTL32(const uint32_t x, const uint32_t n)
{
return (x << n) | (x >> (32 - n));
}
#else
__device__ __forceinline__ uint32_t ROTR32(const uint32_t x, const uint32_t n)
{
return __funnelshift_r(x, x, n);
}
__device__ __forceinline__ uint32_t ROTL32(const uint32_t x, const uint32_t n)
{
return __funnelshift_l(x, x, n);
}
#endif
#endif
// #define NOASM here if you don't want asm
#ifndef __CUDA_ARCH__
#define NOASM
#endif
#define MAKE_ULONGLONG(lo, hi) MAKE_UINT64(lo, hi)
__device__ __forceinline__ uint64_t MAKE_UINT64(uint32_t LO, uint32_t HI)
{
#ifndef NOASM
uint64_t result;
asm("mov.b64 %0,{%1,%2}; \n\t"
: "=l"(result) : "r"(LO), "r"(HI));
return result;
#else
return LO + (uint64_t)HI << 32;
#endif
}
__device__ __forceinline__ uint64_t REPLACE_HIWORD(const uint64_t x, const uint32_t y)
{
#ifndef NOASM
uint64_t result;
asm(
"{\n\t"
".reg .u32 t,t2; \n\t"
"mov.b64 {t2,t},%1; \n\t"
"mov.b64 %0,{t2,%2}; \n\t"
"}" : "=l"(result) : "l"(x), "r"(y)
);
return result;
#else
return (x & 0xffffffff) + ((uint64_t)y << 32);
#endif
}
__device__ __forceinline__ uint64_t REPLACE_LOWORD(const uint64_t x, const uint32_t y)
{
#ifndef NOASM
uint64_t result;
asm(
"{\n\t"
".reg .u32 t,t2; \n\t"
"mov.b64 {t2,t},%1; \n\t"
"mov.b64 %0,{%2,t}; \n\t"
"}" : "=l"(result) : "l"(x), "r"(y)
);
return result;
#else
return (x & 0xffffffff00000000) + y;
#endif
}
// endian change for 32bit
#ifdef __CUDA_ARCH__
__device__ __forceinline__ uint32_t cuda_swab32(const uint32_t x)
{
/* device */
return __byte_perm(x, x, 0x0123);
}
#else
/* host */
#define cuda_swab32(x) \
( ((x) << 24) | (((x) << 8) & 0x00ff0000u) | \
(((x) >> 8) & 0x0000ff00u) | ((x) >> 24))
#endif
static __device__ uint32_t _HIWORD(const uint64_t x)
{
#ifndef NOASM
uint32_t result;
asm(
"{\n\t"
".reg .u32 xl; \n\t"
"mov.b64 {xl,%0},%1; \n\t"
"}" : "=r"(result) : "l"(x)
);
return result;
#else
return x >> 32;
#endif
}
static __device__ uint32_t _LOWORD(const uint64_t x)
{
#ifndef NOASM
uint32_t result;
asm(
"{\n\t"
".reg .u32 xh; \n\t"
"mov.b64 {%0,xh},%1; \n\t"
"}" : "=r"(result) : "l"(x)
);
return result;
#else
return x & 0xffffffff;
#endif
}
// endian change for 64bit
#if (defined __CUDA_ARCH__ && !defined NOASM)
__device__ __forceinline__ uint64_t cuda_swab64(uint64_t x)
{
uint64_t result;
uint2 t;
asm("mov.b64 {%0,%1},%2; \n\t"
: "=r"(t.x), "=r"(t.y) : "l"(x));
t.x=__byte_perm(t.x, 0, 0x0123);
t.y=__byte_perm(t.y, 0, 0x0123);
asm("mov.b64 %0,{%1,%2}; \n\t"
: "=l"(result) : "r"(t.y), "r"(t.x));
return result;
}
#else
/* host */
#define cuda_swab64(x) \
((uint64_t)((((uint64_t)(x)) >> 56) | \
(((uint64_t)(x) & 0x00ff000000000000ULL) >> 40) | \
(((uint64_t)(x) & 0x0000ff0000000000ULL) >> 24) | \
(((uint64_t)(x) & 0x000000ff00000000ULL) >> 8) | \
(((uint64_t)(x) & 0x00000000ff000000ULL) << 8) | \
(((uint64_t)(x) & 0x0000000000ff0000ULL) << 24) | \
(((uint64_t)(x) & 0x000000000000ff00ULL) << 40) | \
(((uint64_t)(x)) << 56)))
#endif
/*********************************************************************/
// Macros to catch CUDA errors in CUDA runtime calls
#define CUDA_SAFE_CALL(call) \
do { \
cudaError_t err = call; \
if (cudaSuccess != err) { \
fprintf(stderr, "Cuda error in func '%s' at line %i : %s.\n", \
__FUNCTION__, __LINE__, cudaGetErrorString(err) ); \
exit(EXIT_FAILURE); \
} \
} while (0)
#define CUDA_CALL_OR_RET(call) do { \
cudaError_t err = call; \
if (cudaSuccess != err) { \
cudaReportHardwareFailure(thr_id, err, __FUNCTION__); \
return; \
} \
} while (0)
#define CUDA_CALL_OR_RET_X(call, ret) do { \
cudaError_t err = call; \
if (cudaSuccess != err) { \
cudaReportHardwareFailure(thr_id, err, __FUNCTION__); \
return ret; \
} \
} while (0)
/*********************************************************************/
#if (defined _WIN64 || defined NOASM)
#define USE_XOR_ASM_OPTS 0
#else
#define USE_XOR_ASM_OPTS 1
#endif
#if USE_XOR_ASM_OPTS
// device asm for whirpool
__device__ __forceinline__
uint64_t xor1(const uint64_t a, const uint64_t b)
{
uint64_t result;
asm("xor.b64 %0, %1, %2;" : "=l"(result) : "l"(a), "l"(b));
return result;
}
#else
#define xor1(a,b) ((a) ^ (b))
#endif
#if USE_XOR_ASM_OPTS
// device asm for whirpool
__device__ __forceinline__
uint64_t xor3(const uint64_t a, const uint64_t b, const uint64_t c)
{
uint64_t result;
asm("xor.b64 %0, %2, %3;\n\t"
"xor.b64 %0, %0, %1;\n\t"
/* output : input registers */
: "=l"(result) : "l"(a), "l"(b), "l"(c));
return result;
}
#else
#define xor3(a,b,c) ((a) ^ (b) ^ (c))
#endif
#if USE_XOR_ASM_OPTS
// device asm for whirpool
__device__ __forceinline__
uint64_t xor8(const uint64_t a, const uint64_t b, const uint64_t c, const uint64_t d, const uint64_t e, const uint64_t f, const uint64_t g, const uint64_t h)
{
uint64_t result;
asm("xor.b64 %0, %1, %2;" : "=l"(result) : "l"(g) ,"l"(h));
asm("xor.b64 %0, %0, %1;" : "+l"(result) : "l"(f));
asm("xor.b64 %0, %0, %1;" : "+l"(result) : "l"(e));
asm("xor.b64 %0, %0, %1;" : "+l"(result) : "l"(d));
asm("xor.b64 %0, %0, %1;" : "+l"(result) : "l"(c));
asm("xor.b64 %0, %0, %1;" : "+l"(result) : "l"(b));
asm("xor.b64 %0, %0, %1;" : "+l"(result) : "l"(a));
return result;
}
#else
#define xor8(a,b,c,d,e,f,g,h) ((a)^(b)^(c)^(d)^(e)^(f)^(g)^(h))
#endif
// device asm for x17
__device__ __forceinline__
uint64_t xandx(const uint64_t a, const uint64_t b, const uint64_t c)
{
uint64_t result;
#ifndef NOASM
asm("{\n\t"
".reg .u64 n;\n\t"
"xor.b64 %0, %2, %3;\n\t"
"and.b64 n, %0, %1;\n\t"
"xor.b64 %0, n, %3;"
"}\n"
: "=l"(result) : "l"(a), "l"(b), "l"(c));
#else
result = ((((b) ^ (c)) & (a)) ^ (c));
#endif
return result;
}
// device asm for x17
__device__ __forceinline__
uint64_t andor(uint64_t a, uint64_t b, uint64_t c)
{
uint64_t result;
#ifndef NOASM
asm("{\n\t"
".reg .u64 m,n;\n\t"
"and.b64 m, %1, %2;\n\t"
" or.b64 n, %1, %2;\n\t"
"and.b64 %0, n, %3;\n\t"
" or.b64 %0, %0, m ;\n\t"
"}\n"
: "=l"(result) : "l"(a), "l"(b), "l"(c));
#else
result = (((a) & (b)) | (((a) | (b)) & (c)));
#endif
return result;
}
// device asm for x17
__device__ __forceinline__
uint64_t shr_t64(uint64_t x, uint32_t n)
{
uint64_t result;
#ifndef NOASM
asm("shr.b64 %0,%1,%2;\n\t"
: "=l"(result) : "l"(x), "r"(n));
#else
result = x >> n;
#endif
return result;
}
// device asm for ?
__device__ __forceinline__
uint64_t shl_t64(uint64_t x, uint32_t n)
{
uint64_t result;
#ifndef NOASM
asm("shl.b64 %0,%1,%2;\n\t"
: "=l"(result) : "l"(x), "r"(n));
#else
result = x << n;
#endif
return result;
}
#ifndef USE_ROT_ASM_OPT
#define USE_ROT_ASM_OPT 1
#endif
#ifdef NOASM
#undef USE_ROT_ASM_OPT
#endif
// 64-bit ROTATE RIGHT
#if __CUDA_ARCH__ >= 320 && USE_ROT_ASM_OPT == 1
/* complicated sm >= 3.5 one (with Funnel Shifter beschleunigt), to bench */
__device__ __forceinline__
uint64_t ROTR64(const uint64_t value, const int offset) {
uint2 result;
if(offset < 32) {
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
} else {
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
}
return __double_as_longlong(__hiloint2double(result.y, result.x));
}
#elif __CUDA_ARCH__ >= 120 && USE_ROT_ASM_OPT == 2
__device__ __forceinline__
uint64_t ROTR64(const uint64_t x, const int offset)
{
uint64_t result;
asm("{\n\t"
".reg .b64 lhs;\n\t"
".reg .u32 roff;\n\t"
"shr.b64 lhs, %1, %2;\n\t"
"sub.u32 roff, 64, %2;\n\t"
"shl.b64 %0, %1, roff;\n\t"
"add.u64 %0, %0, lhs;\n\t"
"}\n"
: "=l"(result) : "l"(x), "r"(offset));
return result;
}
#else
/* host */
#if defined _MSC_VER && !defined __CUDA_ARCH__
#define ROTR64(x, n) _rotr64(x, n)
#else
#define ROTR64(x, n) (((x) >> (n)) | ((x) << (64 - (n))))
#endif
#endif
// 64-bit ROTATE LEFT
#if __CUDA_ARCH__ >= 320 && USE_ROT_ASM_OPT == 1
__device__ __forceinline__
uint64_t ROTL64(const uint64_t value, const int offset) {
uint2 result;
if(offset >= 32) {
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
} else {
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
}
return __double_as_longlong(__hiloint2double(result.y, result.x));
}
#elif __CUDA_ARCH__ >= 120 && USE_ROT_ASM_OPT == 2
__device__ __forceinline__
uint64_t ROTL64(const uint64_t x, const int offset)
{
uint64_t result;
asm("{\n\t"
".reg .b64 lhs;\n\t"
".reg .u32 roff;\n\t"
"shl.b64 lhs, %1, %2;\n\t"
"sub.u32 roff, 64, %2;\n\t"
"shr.b64 %0, %1, roff;\n\t"
"add.u64 %0, lhs, %0;\n\t"
"}\n"
: "=l"(result) : "l"(x), "r"(offset));
return result;
}
#elif __CUDA_ARCH__ >= 320 && USE_ROT_ASM_OPT == 3
__device__
uint64_t ROTL64(const uint64_t x, const int offset)
{
uint64_t res;
asm("{\n\t"
".reg .u32 tl,th,vl,vh;\n\t"
".reg .pred p;\n\t"
"mov.b64 {tl,th}, %1;\n\t"
"shf.l.wrap.b32 vl, tl, th, %2;\n\t"
"shf.l.wrap.b32 vh, th, tl, %2;\n\t"
"setp.lt.u32 p, %2, 32;\n\t"
"@!p mov.b64 %0, {vl,vh};\n\t"
"@p mov.b64 %0, {vh,vl};\n\t"
"}"
: "=l"(res) : "l"(x) , "r"(offset)
);
return res;
}
#else
/* host */
#if defined _MSC_VER && !defined __CUDA_ARCH__
#define ROTL64(x, n) _rotr64(x, n)
#else
#define ROTL64(x, n) (((x) << (n)) | ((x) >> (64 - (n))))
#endif
#endif
__device__ __forceinline__
uint64_t SWAPDWORDS(uint64_t value)
{
#if __CUDA_ARCH__ >= 320 && !defined NOASM
uint2 temp;
asm("mov.b64 {%0, %1}, %2; ": "=r"(temp.x), "=r"(temp.y) : "l"(value));
asm("mov.b64 %0, {%1, %2}; ": "=l"(value) : "r"(temp.y), "r"(temp.x));
return value;
#else
return ROTL64(value, 32);
#endif
}
/* lyra2 - int2 operators */
__device__ __forceinline__
void LOHI(uint32_t &lo, uint32_t &hi, uint64_t x)
{
#ifndef NOASM
asm("mov.b64 {%0,%1},%2; \n\t"
: "=r"(lo), "=r"(hi) : "l"(x));
#else
lo = x & 0xffffffff;
hi = x >> 32;
#endif
}
__device__ __forceinline__ uint64_t devectorize(uint2 x)
{
#ifndef NOASM
uint64_t result;
asm("mov.b64 %0,{%1,%2}; \n\t"
: "=l"(result) : "r"(x.x), "r"(x.y));
return result;
#else
return x.x + ((uint64_t)x.y << 32);
#endif
}
<<<<<<< HEAD
__device__ __forceinline__ uint2 vectorize(uint64_t x)
=======
__device__ __forceinline__ uint2 vectorize(const uint64_t x)
>>>>>>> 6506ecf... fixed build
{
#ifndef NOASM
uint2 result;
asm("mov.b64 {%0,%1},%2; \n\t"
: "=r"(result.x), "=r"(result.y) : "l"(x));
return result;
#else
return make_uint2(x & 0xffffffff, x >> 32);
#endif
}
static __device__ __forceinline__ uint2 vectorizelow(uint32_t v) {
uint2 result;
result.x = v;
result.y = 0;
return result;
}
static __device__ __forceinline__ uint2 vectorizehigh(uint32_t v) {
uint2 result;
result.x = 0;
result.y = v;
return result;
}
static __device__ __forceinline__ uint2 operator^ (uint2 a, uint32_t b) { return make_uint2(a.x^ b, a.y); }
static __device__ __forceinline__ uint2 operator^ (uint2 a, uint2 b) { return make_uint2(a.x ^ b.x, a.y ^ b.y); }
static __device__ __forceinline__ uint2 operator& (uint2 a, uint2 b) { return make_uint2(a.x & b.x, a.y & b.y); }
static __device__ __forceinline__ uint2 operator| (uint2 a, uint2 b) { return make_uint2(a.x | b.x, a.y | b.y); }
static __device__ __forceinline__ uint2 operator~ (uint2 a) { return make_uint2(~a.x, ~a.y); }
static __device__ __forceinline__ void operator^= (uint2 &a, uint2 b) { a = a ^ b; }
static __device__ __forceinline__ uint2 operator+ (uint2 a, uint2 b)
{
#ifndef NOASM
uint2 result;
asm("{\n\t"
"add.cc.u32 %0,%2,%4; \n\t"
"addc.u32 %1,%3,%5; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(b.x), "r"(b.y));
return result;
#else
return make_uint2(a.x + b.x, a.y + b.y);
#endif
}
<<<<<<< HEAD
=======
static __device__ __forceinline__ uint2 operator+ (uint2 a, uint32_t b)
{
uint2 result;
asm("{\n\t"
"add.cc.u32 %0,%2,%4; \n\t"
"addc.u32 %1,%3,%5; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(b), "r"(0));
return result;
}
static __device__ __forceinline__ uint2 operator- (uint2 a, uint32_t b)
{
uint2 result;
asm("{\n\t"
"sub.cc.u32 %0,%2,%4; \n\t"
"subc.u32 %1,%3,%5; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(b), "r"(0));
return result;
}
>>>>>>> 6506ecf... fixed build
static __device__ __forceinline__ uint2 operator- (uint2 a, uint2 b)
{
#ifndef NOASM
uint2 result;
asm("{\n\t"
"sub.cc.u32 %0,%2,%4; \n\t"
"subc.u32 %1,%3,%5; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(b.x), "r"(b.y));
return result;
#else
return make_uint2(a.x - b.x, a.y - b.y);
#endif
}
static __device__ __forceinline__ uint4 operator^ (uint4 a, uint4 b) { return make_uint4(a.x ^ b.x, a.y ^ b.y, a.z ^ b.z, a.w ^ b.w); }
static __device__ __forceinline__ uint4 operator& (uint4 a, uint4 b) { return make_uint4(a.x & b.x, a.y & b.y, a.z & b.z, a.w & b.w); }
static __device__ __forceinline__ uint4 operator| (uint4 a, uint4 b) { return make_uint4(a.x | b.x, a.y | b.y, a.z | b.z, a.w | b.w); }
static __device__ __forceinline__ uint4 operator~ (uint4 a) { return make_uint4(~a.x, ~a.y, ~a.z, ~a.w); }
static __device__ __forceinline__ void operator^= (uint4 &a, uint4 b) { a = a ^ b; }
static __device__ __forceinline__ uint4 operator^ (uint4 a, uint2 b) { return make_uint4(a.x ^ b.x, a.y ^ b.y, a.z ^ b.x, a.w ^ b.y); }
static __device__ __forceinline__ void operator+= (uint2 &a, uint2 b) { a = a + b; }
/**
* basic multiplication between 64bit no carry outside that range (ie mul.lo.b64(a*b))
* (what does uint64 "*" operator)
*/
static __device__ __forceinline__ uint2 operator* (uint2 a, uint2 b)
{
#ifndef NOASM
uint2 result;
asm("{\n\t"
"mul.lo.u32 %0,%2,%4; \n\t"
"mul.hi.u32 %1,%2,%4; \n\t"
"mad.lo.cc.u32 %1,%3,%4,%1; \n\t"
"madc.lo.u32 %1,%3,%5,%1; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(b.x), "r"(b.y));
return result;
#else
return vectorize(devectorize(a)*devectorize(b));
#endif
}
// uint2 method
#if __CUDA_ARCH__ >= 320 && !defined NOASM
__device__ __inline__ uint2 ROR2(const uint2 a, const int offset)
{
uint2 result;
if (offset < 32) {
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(a.x), "r"(a.y), "r"(offset));
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(a.y), "r"(a.x), "r"(offset));
}
else {
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(a.y), "r"(a.x), "r"(offset));
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(offset));
}
return result;
}
#else
__device__ __inline__ uint2 ROR2(const uint2 v, const int n)
{
uint2 result;
if (n <= 32)
{
result.y = ((v.y >> (n)) | (v.x << (32 - n)));
result.x = ((v.x >> (n)) | (v.y << (32 - n)));
}
else
{
result.y = ((v.x >> (n - 32)) | (v.y << (64 - n)));
result.x = ((v.y >> (n - 32)) | (v.x << (64 - n)));
}
return result;
}
#endif
__device__ __inline__ uint32_t ROL8(const uint32_t x)
{
return __byte_perm(x, x, 0x2103);
}
__device__ __inline__ uint32_t ROL16(const uint32_t x)
{
return __byte_perm(x, x, 0x1032);
}
__device__ __inline__ uint32_t ROL24(const uint32_t x)
{
return __byte_perm(x, x, 0x0321);
}
__device__ __inline__ uint2 ROR8(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x0765);
result.y = __byte_perm(a.y, a.x, 0x4321);
return result;
}
__device__ __inline__ uint2 ROR16(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x1076);
result.y = __byte_perm(a.y, a.x, 0x5432);
return result;
}
__device__ __inline__ uint2 ROR24(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x2107);
result.y = __byte_perm(a.y, a.x, 0x6543);
return result;
}
__device__ __inline__ uint2 ROL8(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x6543);
result.y = __byte_perm(a.y, a.x, 0x2107);
return result;
}
__device__ __inline__ uint2 ROL16(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x5432);
result.y = __byte_perm(a.y, a.x, 0x1076);
return result;
}
__device__ __inline__ uint2 ROL24(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x4321);
result.y = __byte_perm(a.y, a.x, 0x0765);
return result;
}
#if __CUDA_ARCH__ >= 320 && !defined NOASM
__inline__ __device__ uint2 ROL2(const uint2 a, const int offset) {
uint2 result;
if (offset >= 32) {
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(a.x), "r"(a.y), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(a.y), "r"(a.x), "r"(offset));
}
else {
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(a.y), "r"(a.x), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(offset));
}
return result;
}
#else
__inline__ __device__ uint2 ROL2(const uint2 v, const int n)
{
uint2 result;
if (n <= 32)
{
result.y = ((v.y << (n)) | (v.x >> (32 - n)));
result.x = ((v.x << (n)) | (v.y >> (32 - n)));
}
else
{
result.y = ((v.x << (n - 32)) | (v.y >> (64 - n)));
result.x = ((v.y << (n - 32)) | (v.x >> (64 - n)));
}
return result;
}
#endif
__device__ __forceinline__
uint64_t ROTR16(uint64_t x)
{
#if __CUDA_ARCH__ > 500 && !defined NOASM
short4 temp;
asm("mov.b64 { %0, %1, %2, %3 }, %4; ": "=h"(temp.x), "=h"(temp.y), "=h"(temp.z), "=h"(temp.w) : "l"(x));
asm("mov.b64 %0, {%1, %2, %3 , %4}; ": "=l"(x) : "h"(temp.y), "h"(temp.z), "h"(temp.w), "h"(temp.x));
return x;
#else
return ROTR64(x, 16);
#endif
}
__device__ __forceinline__
uint64_t ROTL16(uint64_t x)
{
#if __CUDA_ARCH__ > 500 && !defined NOASM
short4 temp;
asm("mov.b64 { %0, %1, %2, %3 }, %4; ": "=h"(temp.x), "=h"(temp.y), "=h"(temp.z), "=h"(temp.w) : "l"(x));
asm("mov.b64 %0, {%1, %2, %3 , %4}; ": "=l"(x) : "h"(temp.w), "h"(temp.x), "h"(temp.y), "h"(temp.z));
return x;
#else
return ROTL64(x, 16);
#endif
}
__device__ __forceinline__
uint2 SWAPINT2(uint2 x)
{
return(make_uint2(x.y, x.x));
}
__device__ __forceinline__ bool cuda_hashisbelowtarget(const uint32_t *const __restrict__ hash, const uint32_t *const __restrict__ target)
{
if (hash[7] > target[7])
return false;
if (hash[7] < target[7])
return true;
if (hash[6] > target[6])
return false;
if (hash[6] < target[6])
return true;
if (hash[5] > target[5])
return false;
if (hash[5] < target[5])
return true;
if (hash[4] > target[4])
return false;
if (hash[4] < target[4])
return true;
if (hash[3] > target[3])
return false;
if (hash[3] < target[3])
return true;
if (hash[2] > target[2])
return false;
if (hash[2] < target[2])
return true;
if (hash[1] > target[1])
return false;
if (hash[1] < target[1])
return true;
if (hash[0] > target[0])
return false;
return true;
}
__device__ __forceinline__
uint2 SWAPDWORDS2(uint2 value)
{
return make_uint2(value.y, value.x);
}
static __forceinline__ __device__ uint2 SHL2(const uint2 a, int offset)
{
uint2 result;
#if __CUDA_ARCH__ > 300 && !defined NOASM
if (offset<32)
{
asm("{\n\t"
"shf.l.clamp.b32 %1,%2,%3,%4; \n\t"
"shl.b32 %0,%2,%4; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(offset));
}
else {
asm("{\n\t"
"shf.l.clamp.b32 %1,%2,%3,%4; \n\t"
"shl.b32 %0,%2,%4; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.y), "r"(a.x), "r"(offset));
}
#else
if (offset<=32)
{
result.y = (a.y << offset) | (a.x >> (32 - offset));
result.x = (a.x << offset);
}
else
{
result.y = (a.x << (offset - 32));
result.x = 0;
}
#endif
return result;
}
static __forceinline__ __device__ uint2 SHR2(const uint2 a, int offset)
{
uint2 result;
#if __CUDA_ARCH__ >= 320 && !defined NOASM
if (offset<32) {
asm("{\n\t"
"shf.r.clamp.b32 %0,%2,%3,%4; \n\t"
"shr.b32 %1,%3,%4; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(offset));
}
else {
asm("{\n\t"
"shf.l.clamp.b32 %0,%2,%3,%4; \n\t"
"shl.b32 %1,%3,%4; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.y), "r"(a.x), "r"(offset));
}
#else
if (offset<=32)
{
result.x = (a.x >> offset) | (a.y << (32 - offset));
result.y = (a.y >> offset);
}
else
{
result.x = (a.y >> (offset - 32));
result.y = 0;
}
#endif
return result;
}
static __device__ __forceinline__ uint64_t devectorizeswap(uint2 v) { return MAKE_UINT64(cuda_swab32(v.y), cuda_swab32(v.x)); }
static __device__ __forceinline__ uint2 vectorizeswap(uint64_t v)
{
uint2 result;
LOHI(result.y, result.x, v);
result.x = cuda_swab32(result.x);
result.y = cuda_swab32(result.y);
return result;
}
__device__ __forceinline__ uint32_t devectorize16(ushort2 x)
{
uint32_t result;
#ifndef NOASM
asm("mov.b32 %0,{%1,%2}; \n\t"
: "=r"(result) : "h"(x.x) , "h"(x.y));
#else
result = x.x + (x.y << 16);
#endif
return result;
}
__device__ __forceinline__ ushort2 vectorize16(uint32_t x)
{
ushort2 result;
#ifndef NOASM
asm("mov.b32 {%0,%1},%2; \n\t"
: "=h"(result.x), "=h"(result.y) : "r"(x));
#else
result.x = x & 0xffff;
result.y = x >> 16;
#endif
return result;
}
extern int cuda_arch[MAX_GPUS];
extern void get_cuda_arch(int *);
/*
static __device__ __forceinline__ uint4 mul4(uint4 a)
{
uint4 result;
asm("{\n\t"
"mul.lo.u32 %0,%4,%5; \n\t"
"mul.hi.u32 %1,%4,%5; \n\t"
"mul.lo.u32 %2,%6,%7; \n\t"
"mul.hi.u32 %3,%6,%7; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y), "=r"(result.z), "=r"(result.w) : "r"(a.x), "r"(a.y), "r"(a.z), "r"(a.w));
return result;
}
static __device__ __forceinline__ uint4 add4(uint4 a, uint4 b)
{
uint4 result;
asm("{\n\t"
"add.cc.u32 %0,%4,%8; \n\t"
"addc.u32 %1,%5,%9; \n\t"
"add.cc.u32 %2,%6,%10; \n\t"
"addc.u32 %3,%7,%11; \n\t"
"}\n\t"
: "=r"(result.x), "=r"(result.y), "=r"(result.z), "=r"(result.w) : "r"(a.x), "r"(a.y), "r"(a.z), "r"(a.w), "r"(b.x), "r"(b.y), "r"(b.z), "r"(b.w));
return result;
}
static __device__ __forceinline__ uint4 madd4(uint4 a, uint4 b)
{