forked from radifier/radiator
-
Notifications
You must be signed in to change notification settings - Fork 0
/
nvml.cpp
2307 lines (2031 loc) · 70.7 KB
/
nvml.cpp
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
/*
* A trivial little dlopen()-based wrapper library for the
* NVIDIA NVML library, to allow runtime discovery of NVML on an
* arbitrary system. This is all very hackish and simple-minded, but
* it serves my immediate needs in the short term until NVIDIA provides
* a static NVML wrapper library themselves, hopefully in
* CUDA 6.5 or maybe sometime shortly after.
*
* This trivial code is made available under the "new" 3-clause BSD license,
* and/or any of the GPL licenses you prefer.
* Feel free to use the code and modify as you see fit.
*
* John E. Stone - [email protected]
* Tanguy Pruvot - tpruvot@github
*
*/
#ifndef WIN32
#include "ccminer-config.h"
#else
#include "ccminer-config-win.h"
#endif
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include "cuda_runtime.h"
#include "nvml.h"
#include "miner.h"
#ifdef USE_WRAPNVML
extern nvml_handle *hnvml;
extern char driver_version[32];
static uint32_t device_bus_ids[MAX_GPUS] = { 0 };
extern uint32_t device_gpu_clocks[MAX_GPUS];
extern uint32_t device_mem_clocks[MAX_GPUS];
extern int32_t device_mem_offsets[MAX_GPUS];
extern uint8_t device_tlimit[MAX_GPUS];
extern int8_t device_pstate[MAX_GPUS];
extern int32_t device_led[MAX_GPUS];
int32_t device_led_state[MAX_GPUS] = { 0 };
static THREAD bool has_rgb_ok = false;
uint32_t clock_prev[MAX_GPUS] = { 0 };
uint32_t clock_prev_mem[MAX_GPUS] = { 0 };
uint32_t limit_prev[MAX_GPUS] = { 0 };
static bool nvml_plimit_set = false;
extern bool need_memclockrst;
/*
* Wrappers to emulate dlopen() on other systems like Windows
*/
#if defined(_MSC_VER) || defined(_WIN32) || defined(_WIN64)
#include <windows.h>
static void *wrap_dlopen(const char *filename) {
HMODULE h = LoadLibrary(filename);
if (!h && opt_debug) {
applog(LOG_DEBUG, "dlopen(%d): failed to load %s",
GetLastError(), filename);
}
return (void*)h;
}
static void *wrap_dlsym(void *h, const char *sym) {
return (void *)GetProcAddress((HINSTANCE)h, sym);
}
static int wrap_dlclose(void *h) {
/* FreeLibrary returns nonzero on success */
return (!FreeLibrary((HINSTANCE)h));
}
#else
/* assume we can use dlopen itself... */
#include <dlfcn.h>
#include <errno.h>
static void *wrap_dlopen(const char *filename) {
void *h = dlopen(filename, RTLD_NOW);
if (h == NULL && opt_debug) {
applog(LOG_DEBUG, "dlopen(%d): failed to load %s",
errno, filename);
}
return (void*)h;
}
static void *wrap_dlsym(void *h, const char *sym) {
return dlsym(h, sym);
}
static int wrap_dlclose(void *h) {
return dlclose(h);
}
#endif
nvml_handle * nvml_create()
{
int i=0;
nvml_handle *nvmlh = NULL;
char tmp[512];
#ifdef WIN32
ExpandEnvironmentStrings(NVML_LIBPATH, tmp, sizeof(tmp));
#else
strcpy(tmp, NVML_LIBPATH);
#endif
void *nvml_dll = wrap_dlopen(tmp);
if (nvml_dll == NULL) {
#ifdef WIN32
nvml_dll = wrap_dlopen("nvml.dll");
if (nvml_dll == NULL)
#endif
return NULL;
}
nvmlh = (nvml_handle *) calloc(1, sizeof(nvml_handle));
nvmlh->nvml_dll = nvml_dll;
nvmlh->nvmlInit = (nvmlReturn_t (*)(void)) wrap_dlsym(nvmlh->nvml_dll, "nvmlInit_v2");
if (!nvmlh->nvmlInit)
nvmlh->nvmlInit = (nvmlReturn_t (*)(void)) wrap_dlsym(nvmlh->nvml_dll, "nvmlInit");
nvmlh->nvmlDeviceGetCount = (nvmlReturn_t (*)(int *)) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCount_v2");
if (!nvmlh->nvmlDeviceGetCount)
nvmlh->nvmlDeviceGetCount = (nvmlReturn_t (*)(int *)) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCount");
nvmlh->nvmlDeviceGetHandleByIndex = (nvmlReturn_t (*)(int, nvmlDevice_t *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetHandleByIndex_v2");
nvmlh->nvmlDeviceGetAPIRestriction = (nvmlReturn_t (*)(nvmlDevice_t, nvmlRestrictedAPI_t, nvmlEnableState_t *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetAPIRestriction");
nvmlh->nvmlDeviceSetAPIRestriction = (nvmlReturn_t (*)(nvmlDevice_t, nvmlRestrictedAPI_t, nvmlEnableState_t))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceSetAPIRestriction");
nvmlh->nvmlDeviceGetDefaultApplicationsClock = (nvmlReturn_t (*)(nvmlDevice_t, nvmlClockType_t, unsigned int *clock))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetDefaultApplicationsClock");
nvmlh->nvmlDeviceGetApplicationsClock = (nvmlReturn_t (*)(nvmlDevice_t, nvmlClockType_t, unsigned int *clocks))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetApplicationsClock");
nvmlh->nvmlDeviceSetApplicationsClocks = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int mem, unsigned int gpu))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceSetApplicationsClocks");
nvmlh->nvmlDeviceResetApplicationsClocks = (nvmlReturn_t (*)(nvmlDevice_t))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceResetApplicationsClocks");
nvmlh->nvmlDeviceGetSupportedGraphicsClocks = (nvmlReturn_t (*)(nvmlDevice_t, uint32_t mem, uint32_t *num, uint32_t *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetSupportedGraphicsClocks");
nvmlh->nvmlDeviceGetSupportedMemoryClocks = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *count, unsigned int *clocksMHz))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetSupportedMemoryClocks");
nvmlh->nvmlDeviceGetClockInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlClockType_t, unsigned int *clock))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetClockInfo");
nvmlh->nvmlDeviceGetMaxClockInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlClockType_t, unsigned int *clock))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetMaxClockInfo");
nvmlh->nvmlDeviceGetPciInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlPciInfo_t *)) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPciInfo_v2");
if (!nvmlh->nvmlDeviceGetPciInfo)
nvmlh->nvmlDeviceGetPciInfo = (nvmlReturn_t (*)(nvmlDevice_t, nvmlPciInfo_t *)) wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPciInfo");
nvmlh->nvmlDeviceGetCurrPcieLinkGeneration = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *gen))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCurrPcieLinkGeneration");
nvmlh->nvmlDeviceGetCurrPcieLinkWidth = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *width))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCurrPcieLinkWidth");
nvmlh->nvmlDeviceGetMaxPcieLinkGeneration = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *gen))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetMaxPcieLinkGeneration");
nvmlh->nvmlDeviceGetMaxPcieLinkWidth = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *width))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetMaxPcieLinkWidth");
nvmlh->nvmlDeviceGetPowerUsage = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerUsage");
nvmlh->nvmlDeviceGetPowerManagementDefaultLimit = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *limit))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerManagementDefaultLimit");
nvmlh->nvmlDeviceGetPowerManagementLimit = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *limit))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerManagementLimit");
nvmlh->nvmlDeviceGetPowerManagementLimitConstraints = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *min, unsigned int *max))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPowerManagementLimitConstraints");
nvmlh->nvmlDeviceSetPowerManagementLimit = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int limit))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceSetPowerManagementLimit");
nvmlh->nvmlDeviceGetName = (nvmlReturn_t (*)(nvmlDevice_t, char *, int))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetName");
nvmlh->nvmlDeviceGetTemperature = (nvmlReturn_t (*)(nvmlDevice_t, int, unsigned int *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetTemperature");
nvmlh->nvmlDeviceGetFanSpeed = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetFanSpeed");
nvmlh->nvmlDeviceGetPerformanceState = (nvmlReturn_t (*)(nvmlDevice_t, int *))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPerformanceState"); /* or nvmlDeviceGetPowerState */
nvmlh->nvmlDeviceGetSerial = (nvmlReturn_t (*)(nvmlDevice_t, char *, unsigned int))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetSerial");
nvmlh->nvmlDeviceGetUUID = (nvmlReturn_t (*)(nvmlDevice_t, char *, unsigned int))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetUUID");
nvmlh->nvmlDeviceGetVbiosVersion = (nvmlReturn_t (*)(nvmlDevice_t, char *, unsigned int))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetVbiosVersion");
nvmlh->nvmlSystemGetDriverVersion = (nvmlReturn_t (*)(char *, unsigned int))
wrap_dlsym(nvmlh->nvml_dll, "nvmlSystemGetDriverVersion");
nvmlh->nvmlErrorString = (char* (*)(nvmlReturn_t))
wrap_dlsym(nvmlh->nvml_dll, "nvmlErrorString");
nvmlh->nvmlShutdown = (nvmlReturn_t (*)())
wrap_dlsym(nvmlh->nvml_dll, "nvmlShutdown");
// v331
nvmlh->nvmlDeviceGetEnforcedPowerLimit = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int *limit))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetEnforcedPowerLimit");
// v340
#ifdef __linux__
nvmlh->nvmlDeviceClearCpuAffinity = (nvmlReturn_t (*)(nvmlDevice_t))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceClearCpuAffinity");
nvmlh->nvmlDeviceGetCpuAffinity = (nvmlReturn_t (*)(nvmlDevice_t, unsigned int sz, unsigned long *cpuSet))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetCpuAffinity");
nvmlh->nvmlDeviceSetCpuAffinity = (nvmlReturn_t (*)(nvmlDevice_t))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceSetCpuAffinity");
#endif
// v346
nvmlh->nvmlDeviceGetPcieThroughput = (nvmlReturn_t (*)(nvmlDevice_t, nvmlPcieUtilCounter_t, unsigned int *value))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetPcieThroughput");
// v36x (API 8 / Pascal)
nvmlh->nvmlDeviceGetClock = (nvmlReturn_t (*)(nvmlDevice_t, nvmlClockType_t clockType, nvmlClockId_t clockId, unsigned int *clockMHz))
wrap_dlsym(nvmlh->nvml_dll, "nvmlDeviceGetClock");
if (nvmlh->nvmlInit == NULL ||
nvmlh->nvmlShutdown == NULL ||
nvmlh->nvmlErrorString == NULL ||
nvmlh->nvmlDeviceGetCount == NULL ||
nvmlh->nvmlDeviceGetHandleByIndex == NULL ||
nvmlh->nvmlDeviceGetPciInfo == NULL ||
nvmlh->nvmlDeviceGetName == NULL)
{
if (opt_debug)
applog(LOG_DEBUG, "Failed to obtain required NVML function pointers");
wrap_dlclose(nvmlh->nvml_dll);
free(nvmlh);
return NULL;
}
nvmlh->nvmlInit();
if (nvmlh->nvmlSystemGetDriverVersion)
nvmlh->nvmlSystemGetDriverVersion(driver_version, sizeof(driver_version));
nvmlh->nvmlDeviceGetCount(&nvmlh->nvml_gpucount);
/* Query CUDA device count, in case it doesn't agree with NVML, since */
/* CUDA will only report GPUs with compute capability greater than 1.0 */
if (cudaGetDeviceCount(&nvmlh->cuda_gpucount) != cudaSuccess) {
if (opt_debug)
applog(LOG_DEBUG, "Failed to query CUDA device count!");
wrap_dlclose(nvmlh->nvml_dll);
free(nvmlh);
return NULL;
}
nvmlh->devs = (nvmlDevice_t *) calloc(nvmlh->nvml_gpucount, sizeof(nvmlDevice_t));
nvmlh->nvml_pci_domain_id = (unsigned int*) calloc(nvmlh->nvml_gpucount, sizeof(unsigned int));
nvmlh->nvml_pci_bus_id = (unsigned int*) calloc(nvmlh->nvml_gpucount, sizeof(unsigned int));
nvmlh->nvml_pci_device_id = (unsigned int*) calloc(nvmlh->nvml_gpucount, sizeof(unsigned int));
nvmlh->nvml_pci_vendor_id = (unsigned int*) calloc(nvmlh->nvml_gpucount, sizeof(unsigned int));
nvmlh->nvml_pci_subsys_id = (unsigned int*) calloc(nvmlh->nvml_gpucount, sizeof(unsigned int));
nvmlh->nvml_cuda_device_id = (int*) calloc(nvmlh->nvml_gpucount, sizeof(int));
nvmlh->cuda_nvml_device_id = (int*) calloc(nvmlh->cuda_gpucount, sizeof(int));
nvmlh->app_clocks = (nvmlEnableState_t*) calloc(nvmlh->nvml_gpucount, sizeof(nvmlEnableState_t));
/* Obtain GPU device handles we're going to need repeatedly... */
for (i=0; i<nvmlh->nvml_gpucount; i++) {
nvmlh->nvmlDeviceGetHandleByIndex(i, &nvmlh->devs[i]);
}
/* Query PCI info for each NVML device, and build table for mapping of */
/* CUDA device IDs to NVML device IDs and vice versa */
for (i=0; i<nvmlh->nvml_gpucount; i++) {
nvmlPciInfo_t pciinfo;
nvmlh->nvmlDeviceGetPciInfo(nvmlh->devs[i], &pciinfo);
nvmlh->nvml_pci_domain_id[i] = pciinfo.domain;
nvmlh->nvml_pci_bus_id[i] = pciinfo.bus;
nvmlh->nvml_pci_device_id[i] = pciinfo.device;
nvmlh->nvml_pci_vendor_id[i] = pciinfo.pci_device_id;
nvmlh->nvml_pci_subsys_id[i] = pciinfo.pci_subsystem_id;
nvmlh->app_clocks[i] = NVML_FEATURE_UNKNOWN;
if (nvmlh->nvmlDeviceSetAPIRestriction) {
nvmlh->nvmlDeviceSetAPIRestriction(nvmlh->devs[i], NVML_RESTRICTED_API_SET_APPLICATION_CLOCKS,
NVML_FEATURE_ENABLED);
/* there is only this API_SET_APPLICATION_CLOCKS on the 750 Ti (340.58) */
}
if (nvmlh->nvmlDeviceGetAPIRestriction) {
nvmlh->nvmlDeviceGetAPIRestriction(nvmlh->devs[i], NVML_RESTRICTED_API_SET_APPLICATION_CLOCKS,
&nvmlh->app_clocks[i]);
}
}
/* build mapping of NVML device IDs to CUDA IDs */
for (i=0; i<nvmlh->nvml_gpucount; i++) {
nvmlh->nvml_cuda_device_id[i] = -1;
}
for (i=0; i<nvmlh->cuda_gpucount; i++) {
cudaDeviceProp props;
nvmlh->cuda_nvml_device_id[i] = -1;
if (cudaGetDeviceProperties(&props, i) == cudaSuccess) {
device_bus_ids[i] = props.pciBusID;
for (int j = 0; j < nvmlh->nvml_gpucount; j++) {
if ((nvmlh->nvml_pci_domain_id[j] == (uint32_t) props.pciDomainID) &&
(nvmlh->nvml_pci_bus_id[j] == (uint32_t) props.pciBusID) &&
(nvmlh->nvml_pci_device_id[j] == (uint32_t) props.pciDeviceID)) {
if (opt_debug)
applog(LOG_DEBUG, "CUDA GPU %d matches NVML GPU %d by busId %u",
i, j, (uint32_t) props.pciBusID);
nvmlh->nvml_cuda_device_id[j] = i;
nvmlh->cuda_nvml_device_id[i] = j;
}
}
}
}
return nvmlh;
}
/* apply config clocks to an used device */
int nvml_set_clocks(nvml_handle *nvmlh, int dev_id)
{
nvmlReturn_t rc;
uint32_t gpu_clk = 0, mem_clk = 0;
int n = nvmlh->cuda_nvml_device_id[dev_id];
//if (need_nvsettings) /* prefer later than init time */
// nvs_set_clocks(dev_id);
if (n < 0 || n >= nvmlh->nvml_gpucount)
return -ENODEV;
if (!device_gpu_clocks[dev_id] && !device_mem_clocks[dev_id])
return 0; // nothing to do
if (nvmlh->app_clocks[n] != NVML_FEATURE_ENABLED) {
applog(LOG_WARNING, "GPU #%d: NVML application clock feature is not allowed!", dev_id);
return -EPERM;
}
uint32_t mem_prev = clock_prev_mem[dev_id];
if (!mem_prev)
nvmlh->nvmlDeviceGetApplicationsClock(nvmlh->devs[n], NVML_CLOCK_MEM, &mem_prev);
uint32_t gpu_prev = clock_prev[dev_id];
if (!gpu_prev)
nvmlh->nvmlDeviceGetApplicationsClock(nvmlh->devs[n], NVML_CLOCK_GRAPHICS, &gpu_prev);
nvmlh->nvmlDeviceGetDefaultApplicationsClock(nvmlh->devs[n], NVML_CLOCK_MEM, &mem_clk);
rc = nvmlh->nvmlDeviceGetDefaultApplicationsClock(nvmlh->devs[n], NVML_CLOCK_GRAPHICS, &gpu_clk);
if (rc != NVML_SUCCESS) {
applog(LOG_WARNING, "GPU #%d: unable to query application clocks", dev_id);
return -EINVAL;
}
if (opt_debug)
applog(LOG_DEBUG, "GPU #%d: default application clocks are %u/%u", dev_id, mem_clk, gpu_clk);
// get application config values
if (device_mem_clocks[dev_id]) mem_clk = device_mem_clocks[dev_id];
if (device_gpu_clocks[dev_id]) gpu_clk = device_gpu_clocks[dev_id];
// these functions works for the 960 and the 970 (346.72+), and for the 750 Ti with driver ~361+
uint32_t nclocks = 0, mem_clocks[32] = { 0 };
nvmlh->nvmlDeviceGetSupportedMemoryClocks(nvmlh->devs[n], &nclocks, NULL);
nclocks = min(nclocks, 32U);
if (nclocks)
nvmlh->nvmlDeviceGetSupportedMemoryClocks(nvmlh->devs[n], &nclocks, mem_clocks);
for (uint8_t u=0; u < nclocks; u++) {
// ordered by pstate (so highest is first memory clock - P0)
if (mem_clocks[u] <= mem_clk) {
mem_clk = mem_clocks[u];
break;
}
}
uint32_t* gpu_clocks = NULL;
nclocks = 0;
nvmlh->nvmlDeviceGetSupportedGraphicsClocks(nvmlh->devs[n], mem_clk, &nclocks, NULL);
if (nclocks) {
if (opt_debug)
applog(LOG_DEBUG, "GPU #%d: %u clocks found for mem %u", dev_id, nclocks, mem_clk);
gpu_clocks = (uint32_t*) calloc(1, sizeof(uint32_t) * nclocks + 4);
nvmlh->nvmlDeviceGetSupportedGraphicsClocks(nvmlh->devs[n], mem_clk, &nclocks, gpu_clocks);
for (uint8_t u=0; u < nclocks; u++) {
// ordered desc, so get first
if (gpu_clocks[u] <= gpu_clk) {
gpu_clk = gpu_clocks[u];
break;
}
}
free(gpu_clocks);
}
rc = nvmlh->nvmlDeviceSetApplicationsClocks(nvmlh->devs[n], mem_clk, gpu_clk);
if (rc == NVML_SUCCESS)
applog(LOG_INFO, "GPU #%d: application clocks set to %u/%u", dev_id, mem_clk, gpu_clk);
else {
applog(LOG_WARNING, "GPU #%d: %u/%u - %s (NVML)", dev_id, mem_clk, gpu_clk, nvmlh->nvmlErrorString(rc));
return -1;
}
// store previous clocks for reset on exit (or during wait...)
clock_prev[dev_id] = gpu_prev;
clock_prev_mem[dev_id] = mem_prev;
return 1;
}
/* reset default app clocks and limits on exit */
int nvml_reset_clocks(nvml_handle *nvmlh, int dev_id)
{
int ret = 0;
nvmlReturn_t rc;
uint32_t gpu_clk = 0, mem_clk = 0;
int n = nvmlh->cuda_nvml_device_id[dev_id];
if (need_nvsettings)
nvs_reset_clocks(dev_id);
if (n < 0 || n >= nvmlh->nvml_gpucount)
return -ENODEV;
if (clock_prev[dev_id]) {
rc = nvmlh->nvmlDeviceResetApplicationsClocks(nvmlh->devs[n]);
if (rc != NVML_SUCCESS) {
applog(LOG_WARNING, "GPU #%d: unable to reset application clocks", dev_id);
}
clock_prev[dev_id] = 0;
ret = 1;
}
if (limit_prev[dev_id]) {
uint32_t plimit = limit_prev[dev_id];
if (nvmlh->nvmlDeviceGetPowerManagementDefaultLimit && !plimit) {
rc = nvmlh->nvmlDeviceGetPowerManagementDefaultLimit(nvmlh->devs[n], &plimit);
} else if (plimit) {
rc = NVML_SUCCESS;
}
if (rc == NVML_SUCCESS)
nvmlh->nvmlDeviceSetPowerManagementLimit(nvmlh->devs[n], plimit);
ret = 1;
}
return ret;
}
/**
* Set power state of a device (9xx)
* Code is similar as clocks one, which allow the change of the pstate
*/
int nvml_set_pstate(nvml_handle *nvmlh, int dev_id)
{
nvmlReturn_t rc;
uint32_t gpu_clk = 0, mem_clk = 0;
int n = nvmlh->cuda_nvml_device_id[dev_id];
if (n < 0 || n >= nvmlh->nvml_gpucount)
return -ENODEV;
if (device_pstate[dev_id] < 0)
return 0;
if (nvmlh->app_clocks[n] != NVML_FEATURE_ENABLED) {
applog(LOG_WARNING, "GPU #%d: NVML app. clock feature is not allowed!", dev_id);
return -EPERM;
}
nvmlh->nvmlDeviceGetDefaultApplicationsClock(nvmlh->devs[n], NVML_CLOCK_MEM, &mem_clk);
rc = nvmlh->nvmlDeviceGetDefaultApplicationsClock(nvmlh->devs[n], NVML_CLOCK_GRAPHICS, &gpu_clk);
if (rc != NVML_SUCCESS) {
applog(LOG_WARNING, "GPU #%d: unable to query application clocks", dev_id);
return -EINVAL;
}
// get application config values
if (device_mem_clocks[dev_id]) mem_clk = device_mem_clocks[dev_id];
if (device_gpu_clocks[dev_id]) gpu_clk = device_gpu_clocks[dev_id];
// these functions works for the 960 and the 970 (346.72+), and for the 750 Ti with driver ~361+
uint32_t nclocks = 0, mem_clocks[32] = { 0 };
int8_t wanted_pstate = device_pstate[dev_id];
nvmlh->nvmlDeviceGetSupportedMemoryClocks(nvmlh->devs[n], &nclocks, NULL);
nclocks = min(nclocks, 32U);
if (nclocks)
nvmlh->nvmlDeviceGetSupportedMemoryClocks(nvmlh->devs[n], &nclocks, mem_clocks);
if ((uint32_t) wanted_pstate+1 > nclocks) {
applog(LOG_WARNING, "GPU #%d: only %u mem clocks available (p-states)", dev_id, nclocks);
}
for (uint8_t u=0; u < nclocks; u++) {
// ordered by pstate (so highest P0 first)
if (u == wanted_pstate) {
mem_clk = mem_clocks[u];
break;
}
}
uint32_t* gpu_clocks = NULL;
nclocks = 0;
nvmlh->nvmlDeviceGetSupportedGraphicsClocks(nvmlh->devs[n], mem_clk, &nclocks, NULL);
if (nclocks) {
gpu_clocks = (uint32_t*) calloc(1, sizeof(uint32_t) * nclocks + 4);
rc = nvmlh->nvmlDeviceGetSupportedGraphicsClocks(nvmlh->devs[n], mem_clk, &nclocks, gpu_clocks);
if (rc == NVML_SUCCESS) {
// ordered desc, get the max app clock (do not limit)
gpu_clk = gpu_clocks[0];
}
free(gpu_clocks);
}
rc = nvmlh->nvmlDeviceSetApplicationsClocks(nvmlh->devs[n], mem_clk, gpu_clk);
if (rc != NVML_SUCCESS) {
applog(LOG_WARNING, "GPU #%d: pstate P%d (%u/%u) %s", dev_id, (int) wanted_pstate,
mem_clk, gpu_clk, nvmlh->nvmlErrorString(rc));
return -1;
}
if (!opt_quiet)
applog(LOG_INFO, "GPU #%d: app clocks set to P%d (%u/%u)", dev_id, (int) wanted_pstate, mem_clk, gpu_clk);
clock_prev[dev_id] = 1;
return 1;
}
int nvml_set_plimit(nvml_handle *nvmlh, int dev_id)
{
nvmlReturn_t rc = NVML_ERROR_UNKNOWN;
uint32_t gpu_clk = 0, mem_clk = 0;
int n = nvmlh->cuda_nvml_device_id[dev_id];
if (n < 0 || n >= nvmlh->nvml_gpucount)
return -ENODEV;
if (!device_plimit[dev_id])
return 0; // nothing to do
if (!nvmlh->nvmlDeviceSetPowerManagementLimit)
return -ENOSYS;
uint32_t plimit = device_plimit[dev_id] * 1000;
uint32_t pmin = 1000, pmax = 0, prev_limit = 0;
if (nvmlh->nvmlDeviceGetPowerManagementLimitConstraints)
rc = nvmlh->nvmlDeviceGetPowerManagementLimitConstraints(nvmlh->devs[n], &pmin, &pmax);
if (rc != NVML_SUCCESS) {
if (!nvmlh->nvmlDeviceGetPowerManagementLimit)
return -ENOSYS;
}
nvmlh->nvmlDeviceGetPowerManagementLimit(nvmlh->devs[n], &prev_limit);
if (!pmax) pmax = prev_limit;
plimit = min(plimit, pmax);
plimit = max(plimit, pmin);
rc = nvmlh->nvmlDeviceSetPowerManagementLimit(nvmlh->devs[n], plimit);
if (rc != NVML_SUCCESS) {
#ifndef WIN32
applog(LOG_WARNING, "GPU #%d: plimit %s", dev_id, nvmlh->nvmlErrorString(rc));
#endif
return -1;
} else {
device_plimit[dev_id] = plimit / 1000;
nvml_plimit_set = true;
}
if (!opt_quiet) {
applog(LOG_INFO, "GPU #%d: power limit set to %uW (allowed range is %u-%u)",
dev_id, plimit/1000U, pmin/1000U, pmax/1000U);
}
limit_prev[dev_id] = prev_limit;
return 1;
}
uint32_t nvml_get_plimit(nvml_handle *nvmlh, int dev_id)
{
uint32_t plimit = 0;
int n = nvmlh ? nvmlh->cuda_nvml_device_id[dev_id] : -1;
if (n < 0 || n >= nvmlh->nvml_gpucount)
return 0;
if (nvmlh->nvmlDeviceGetPowerManagementLimit) {
nvmlh->nvmlDeviceGetPowerManagementLimit(nvmlh->devs[n], &plimit);
}
return plimit;
}
// ccminer -D -n
#define LSTDEV_PFX " "
void nvml_print_device_info(int dev_id)
{
if (!hnvml) return;
int n = hnvml->cuda_nvml_device_id[dev_id];
if (n < 0 || n >= hnvml->nvml_gpucount)
return;
nvmlReturn_t rc;
// fprintf(stderr, "------ Hardware ------\n");
int gvid = hnvml->nvml_pci_vendor_id[n] & 0xFFFF;
int gpid = hnvml->nvml_pci_vendor_id[n] >> 16;
int svid = hnvml->nvml_pci_subsys_id[n] & 0xFFFF;
int spid = hnvml->nvml_pci_subsys_id[n] >> 16;
fprintf(stderr, LSTDEV_PFX "ID %04x:%04x/%04x:%04x BUS %04x:%02x:%02x.0\n", gvid, gpid, svid, spid,
(int) hnvml->nvml_pci_domain_id[n], (int) hnvml->nvml_pci_bus_id[n], (int) hnvml->nvml_pci_device_id[n]);
if (hnvml->nvmlDeviceGetClock) {
uint32_t gpu_clk = 0, mem_clk = 0;
// fprintf(stderr, "------- Clocks -------\n");
hnvml->nvmlDeviceGetClock(hnvml->devs[n], NVML_CLOCK_GRAPHICS, NVML_CLOCK_ID_APP_CLOCK_DEFAULT, &gpu_clk);
rc = hnvml->nvmlDeviceGetClock(hnvml->devs[n], NVML_CLOCK_MEM, NVML_CLOCK_ID_APP_CLOCK_DEFAULT, &mem_clk);
if (rc == NVML_SUCCESS) {
fprintf(stderr, LSTDEV_PFX "DEFAULT MEM %4u GPU %4u MHz\n", mem_clk, gpu_clk);
}
hnvml->nvmlDeviceGetClock(hnvml->devs[n], NVML_CLOCK_GRAPHICS, NVML_CLOCK_ID_APP_CLOCK_TARGET, &gpu_clk);
rc = hnvml->nvmlDeviceGetClock(hnvml->devs[n], NVML_CLOCK_MEM, NVML_CLOCK_ID_APP_CLOCK_TARGET, &mem_clk);
if (rc == NVML_SUCCESS) {
fprintf(stderr, LSTDEV_PFX "TARGET MEM %4u GPU %4u MHz\n", mem_clk, gpu_clk);
}
hnvml->nvmlDeviceGetClock(hnvml->devs[n], NVML_CLOCK_GRAPHICS, NVML_CLOCK_ID_CURRENT, &gpu_clk);
rc = hnvml->nvmlDeviceGetClock(hnvml->devs[n], NVML_CLOCK_MEM, NVML_CLOCK_ID_CURRENT, &mem_clk);
if (rc == NVML_SUCCESS) {
fprintf(stderr, LSTDEV_PFX "CURRENT MEM %4u GPU %4u MHz\n", mem_clk, gpu_clk);
}
}
}
int nvml_get_gpucount(nvml_handle *nvmlh, int *gpucount)
{
*gpucount = nvmlh->nvml_gpucount;
return 0;
}
int cuda_get_gpucount(nvml_handle *nvmlh, int *gpucount)
{
*gpucount = nvmlh->cuda_gpucount;
return 0;
}
int nvml_get_gpu_name(nvml_handle *nvmlh, int cudaindex, char *namebuf, int bufsize)
{
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -ENODEV;
if (!nvmlh->nvmlDeviceGetName)
return -ENOSYS;
if (nvmlh->nvmlDeviceGetName(nvmlh->devs[gpuindex], namebuf, bufsize) != NVML_SUCCESS)
return -1;
return 0;
}
int nvml_get_tempC(nvml_handle *nvmlh, int cudaindex, unsigned int *tempC)
{
nvmlReturn_t rc;
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -ENODEV;
if (!nvmlh->nvmlDeviceGetTemperature)
return -ENOSYS;
rc = nvmlh->nvmlDeviceGetTemperature(nvmlh->devs[gpuindex], 0u /* NVML_TEMPERATURE_GPU */, tempC);
if (rc != NVML_SUCCESS) {
return -1;
}
return 0;
}
int nvml_get_fanpcnt(nvml_handle *nvmlh, int cudaindex, unsigned int *fanpcnt)
{
nvmlReturn_t rc;
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -ENODEV;
if (!nvmlh->nvmlDeviceGetFanSpeed)
return -ENOSYS;
rc = nvmlh->nvmlDeviceGetFanSpeed(nvmlh->devs[gpuindex], fanpcnt);
if (rc != NVML_SUCCESS) {
return -1;
}
return 0;
}
int nvml_get_current_clocks(int cudaindex, unsigned int *graphics_clock, unsigned int *mem_clock)
{
nvmlReturn_t rc;
int gpuindex = hnvml->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= hnvml->nvml_gpucount) return -ENODEV;
if (!hnvml->nvmlDeviceGetClockInfo) return -ENOSYS;
rc = hnvml->nvmlDeviceGetClockInfo(hnvml->devs[gpuindex], NVML_CLOCK_SM, graphics_clock);
if (rc != NVML_SUCCESS) return -1;
rc = hnvml->nvmlDeviceGetClockInfo(hnvml->devs[gpuindex], NVML_CLOCK_MEM, mem_clock);
if (rc != NVML_SUCCESS) return -1;
return 0;
}
/* Not Supported on 750Ti 340.23 */
int nvml_get_power_usage(nvml_handle *nvmlh, int cudaindex, unsigned int *milliwatts)
{
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -ENODEV;
if (!nvmlh->nvmlDeviceGetPowerUsage)
return -ENOSYS;
nvmlReturn_t res = nvmlh->nvmlDeviceGetPowerUsage(nvmlh->devs[gpuindex], milliwatts);
if (res != NVML_SUCCESS) {
//if (opt_debug)
// applog(LOG_DEBUG, "nvmlDeviceGetPowerUsage: %s", nvmlh->nvmlErrorString(res));
return -1;
}
return 0;
}
/* Not Supported on 750Ti 340.23 */
int nvml_get_pstate(nvml_handle *nvmlh, int cudaindex, int *pstate)
{
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -ENODEV;
if (!nvmlh->nvmlDeviceGetPerformanceState)
return -ENOSYS;
nvmlReturn_t res = nvmlh->nvmlDeviceGetPerformanceState(nvmlh->devs[gpuindex], pstate);
if (res != NVML_SUCCESS) {
//if (opt_debug)
// applog(LOG_DEBUG, "nvmlDeviceGetPerformanceState: %s", nvmlh->nvmlErrorString(res));
return -1;
}
return 0;
}
int nvml_get_busid(nvml_handle *nvmlh, int cudaindex, int *busid)
{
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -ENODEV;
(*busid) = nvmlh->nvml_pci_bus_id[gpuindex];
return 0;
}
int nvml_get_serial(nvml_handle *nvmlh, int cudaindex, char *sn, int maxlen)
{
uint32_t subids = 0;
char uuid[NVML_DEVICE_UUID_BUFFER_SIZE];
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
nvmlReturn_t res;
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -ENODEV;
if (nvmlh->nvmlDeviceGetSerial) {
res = nvmlh->nvmlDeviceGetSerial(nvmlh->devs[gpuindex], sn, maxlen);
if (res == NVML_SUCCESS)
return 0;
}
if (!nvmlh->nvmlDeviceGetUUID)
return -ENOSYS;
// nvmlDeviceGetUUID: GPU-f2bd642c-369f-5a14-e0b4-0d22dfe9a1fc
// use a part of uuid to generate an unique serial
// todo: check if there is vendor id is inside
memset(uuid, 0, sizeof(uuid));
res = nvmlh->nvmlDeviceGetUUID(nvmlh->devs[gpuindex], uuid, sizeof(uuid)-1);
if (res != NVML_SUCCESS) {
if (opt_debug)
applog(LOG_DEBUG, "nvmlDeviceGetUUID: %s", nvmlh->nvmlErrorString(res));
return -1;
}
strncpy(sn, &uuid[4], min((int) strlen(uuid), maxlen));
sn[maxlen-1] = '\0';
return 0;
}
int nvml_get_bios(nvml_handle *nvmlh, int cudaindex, char *desc, int maxlen)
{
uint32_t subids = 0;
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -ENODEV;
if (!nvmlh->nvmlDeviceGetVbiosVersion)
return -ENOSYS;
nvmlReturn_t res = nvmlh->nvmlDeviceGetVbiosVersion(nvmlh->devs[gpuindex], desc, maxlen);
if (res != NVML_SUCCESS) {
if (opt_debug)
applog(LOG_DEBUG, "nvmlDeviceGetVbiosVersion: %s", nvmlh->nvmlErrorString(res));
return -1;
}
return 0;
}
int nvml_get_info(nvml_handle *nvmlh, int cudaindex, uint16_t &vid, uint16_t &pid)
{
uint32_t subids = 0;
int gpuindex = nvmlh->cuda_nvml_device_id[cudaindex];
if (gpuindex < 0 || gpuindex >= nvmlh->nvml_gpucount)
return -ENODEV;
subids = nvmlh->nvml_pci_subsys_id[gpuindex];
if (!subids) subids = nvmlh->nvml_pci_vendor_id[gpuindex];
pid = subids >> 16;
vid = subids & 0xFFFF;
// Colorful and Inno3D
if (pid == 0) pid = nvmlh->nvml_pci_vendor_id[gpuindex] >> 16;
return 0;
}
int nvml_destroy(nvml_handle *nvmlh)
{
nvmlh->nvmlShutdown();
wrap_dlclose(nvmlh->nvml_dll);
free(nvmlh->nvml_pci_bus_id);
free(nvmlh->nvml_pci_device_id);
free(nvmlh->nvml_pci_domain_id);
free(nvmlh->nvml_pci_vendor_id);
free(nvmlh->nvml_pci_subsys_id);
free(nvmlh->nvml_cuda_device_id);
free(nvmlh->cuda_nvml_device_id);
free(nvmlh->app_clocks);
free(nvmlh->devs);
free(nvmlh);
return 0;
}
// ----------------------------------------------------------------------------
/**
* nvapi alternative for windows x86 binaries
* nvml api doesn't exists as 32bit dll :///
*/
#ifdef WIN32
#include "nvapi/nvapi_ccminer.h"
static unsigned int nvapi_dev_map[MAX_GPUS] = { 0 };
static NvDisplayHandle hDisplay_a[NVAPI_MAX_PHYSICAL_GPUS * 2] = { 0 };
static NvPhysicalGpuHandle phys[NVAPI_MAX_PHYSICAL_GPUS] = { 0 };
static NvU32 nvapi_dev_cnt = 0;
extern bool nvapi_dll_loaded;
int nvapi_temperature(unsigned int devNum, unsigned int *temperature)
{
NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt)
return -ENODEV;
NV_GPU_THERMAL_SETTINGS thermal;
thermal.version = NV_GPU_THERMAL_SETTINGS_VER;
ret = NvAPI_GPU_GetThermalSettings(phys[devNum], 0, &thermal);
if (ret != NVAPI_OK) {
NvAPI_ShortString string;
NvAPI_GetErrorMessage(ret, string);
if (opt_debug)
applog(LOG_DEBUG, "NVAPI NvAPI_GPU_GetThermalSettings: %s", string);
return -1;
}
(*temperature) = (unsigned int) thermal.sensor[0].currentTemp;
return 0;
}
int nvapi_fanspeed(unsigned int devNum, unsigned int *speed)
{
NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt)
return -ENODEV;
NvU32 fanspeed = 0;
ret = NvAPI_GPU_GetTachReading(phys[devNum], &fanspeed);
if (ret != NVAPI_OK) {
NvAPI_ShortString string;
NvAPI_GetErrorMessage(ret, string);
if (opt_debug)
applog(LOG_DEBUG, "NVAPI NvAPI_GPU_GetTachReading: %s", string);
return -1;
}
(*speed) = (unsigned int) fanspeed;
return 0;
}
int nvapi_getpstate(unsigned int devNum, unsigned int *pstate)
{
NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt)
return -ENODEV;
NV_GPU_PERF_PSTATE_ID CurrentPstate = NVAPI_GPU_PERF_PSTATE_UNDEFINED; /* 16 */
ret = NvAPI_GPU_GetCurrentPstate(phys[devNum], &CurrentPstate);
if (ret != NVAPI_OK) {
NvAPI_ShortString string;
NvAPI_GetErrorMessage(ret, string);
if (opt_debug)
applog(LOG_DEBUG, "NVAPI NvAPI_GPU_GetCurrentPstate: %s", string);
return -1;
}
else {
// get pstate for the moment... often 0 = P0
(*pstate) = (unsigned int)CurrentPstate;
}
return 0;
}
#define UTIL_DOMAIN_GPU 0
int nvapi_getusage(unsigned int devNum, unsigned int *pct)
{
NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt)
return -ENODEV;
NV_GPU_DYNAMIC_PSTATES_INFO_EX info;
info.version = NV_GPU_DYNAMIC_PSTATES_INFO_EX_VER;
ret = NvAPI_GPU_GetDynamicPstatesInfoEx(phys[devNum], &info);
if (ret != NVAPI_OK) {
NvAPI_ShortString string;
NvAPI_GetErrorMessage(ret, string);
if (opt_debug)
applog(LOG_DEBUG, "NVAPI GetDynamicPstatesInfoEx: %s", string);
return -1;
}
else {
if (info.utilization[UTIL_DOMAIN_GPU].bIsPresent)
(*pct) = info.utilization[UTIL_DOMAIN_GPU].percentage;
}
return 0;
}
int nvapi_getinfo(unsigned int devNum, uint16_t &vid, uint16_t &pid)
{
NvAPI_Status ret;
NvU32 pDeviceId, pSubSystemId, pRevisionId, pExtDeviceId;
if (devNum >= nvapi_dev_cnt)
return -ENODEV;
ret = NvAPI_GPU_GetPCIIdentifiers(phys[devNum], &pDeviceId, &pSubSystemId, &pRevisionId, &pExtDeviceId);
if (ret != NVAPI_OK) {
NvAPI_ShortString string;
NvAPI_GetErrorMessage(ret, string);
if (opt_debug)
applog(LOG_DEBUG, "NVAPI GetPCIIdentifiers: %s", string);
return -1;
}
pid = pDeviceId >> 16;
vid = pDeviceId & 0xFFFF;
if (vid == 0x10DE && pSubSystemId) {
vid = pSubSystemId & 0xFFFF;
pid = pSubSystemId >> 16;
// Colorful and Inno3D
if (pid == 0) pid = pDeviceId >> 16;
}
return 0;
}
int nvapi_getserial(unsigned int devNum, char *serial, unsigned int maxlen)
{
NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt)
return -ENODEV;
memset(serial, 0, maxlen);
if (maxlen < 11)
return -EINVAL;
NvAPI_ShortString ser = { 0 };
ret = NvAPI_DLL_GetSerialNumber(phys[devNum], ser);
if (ret != NVAPI_OK) {
NvAPI_ShortString string;
NvAPI_GetErrorMessage(ret, string);
if (opt_debug)
applog(LOG_DEBUG, "NVAPI GetSerialNumber: %s", string);
return -1;
}
uint8_t *bytes = (uint8_t*) ser;
for (int n=0; n<5; n++) sprintf(&serial[n*2], "%02X", bytes[n]);
return 0;
}
int nvapi_getbios(unsigned int devNum, char *desc, unsigned int maxlen)
{
NvAPI_Status ret;
if (devNum >= nvapi_dev_cnt)
return -ENODEV;
if (maxlen < 64) // Short String