aboutsummaryrefslogtreecommitdiff
path: root/mesalib/src/mesa/main/performance_monitor.c
blob: 2d740daf019d1d2b9ac818bc5d5d03353107b259 (plain)
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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
/*
 * Copyright © 2012 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

/**
 * \file performance_monitor.c
 * Core Mesa support for the AMD_performance_monitor extension.
 *
 * In order to implement this extension, start by defining two enums:
 * one for Groups, and one for Counters.  These will be used as indexes into
 * arrays, so they should start at 0 and increment from there.
 *
 * Counter IDs need to be globally unique.  That is, you can't have counter 7
 * in group A and counter 7 in group B.  A global enum of all available
 * counters is a convenient way to guarantee this.
 */

#include <stdbool.h>
#include "glheader.h"
#include "context.h"
#include "enums.h"
#include "hash.h"
#include "macros.h"
#include "mtypes.h"
#include "performance_monitor.h"
#include "util/bitset.h"
#include "util/ralloc.h"

void
_mesa_init_performance_monitors(struct gl_context *ctx)
{
   ctx->PerfMonitor.Monitors = _mesa_NewHashTable();
   ctx->PerfMonitor.NumGroups = 0;
   ctx->PerfMonitor.Groups = NULL;
}

static struct gl_perf_monitor_object *
new_performance_monitor(struct gl_context *ctx, GLuint index)
{
   unsigned i;
   struct gl_perf_monitor_object *m = ctx->Driver.NewPerfMonitor(ctx);

   if (m == NULL)
      return NULL;

   m->Name = index;

   m->Active = false;

   m->ActiveGroups =
      rzalloc_array(NULL, unsigned, ctx->PerfMonitor.NumGroups);

   m->ActiveCounters =
      ralloc_array(NULL, BITSET_WORD *, ctx->PerfMonitor.NumGroups);

   if (m->ActiveGroups == NULL || m->ActiveCounters == NULL)
      goto fail;

   for (i = 0; i < ctx->PerfMonitor.NumGroups; i++) {
      const struct gl_perf_monitor_group *g = &ctx->PerfMonitor.Groups[i];

      m->ActiveCounters[i] = rzalloc_array(m->ActiveCounters, BITSET_WORD,
                                           BITSET_WORDS(g->NumCounters));
      if (m->ActiveCounters[i] == NULL)
         goto fail;
   }

   return m;

fail:
   ralloc_free(m->ActiveGroups);
   ralloc_free(m->ActiveCounters);
   ctx->Driver.DeletePerfMonitor(ctx, m);
   return NULL;
}

static void
free_performance_monitor(GLuint key, void *data, void *user)
{
   struct gl_perf_monitor_object *m = data;
   struct gl_context *ctx = user;

   ralloc_free(m->ActiveGroups);
   ralloc_free(m->ActiveCounters);
   ctx->Driver.DeletePerfMonitor(ctx, m);
}

void
_mesa_free_performance_monitors(struct gl_context *ctx)
{
   _mesa_HashDeleteAll(ctx->PerfMonitor.Monitors,
                       free_performance_monitor, ctx);
   _mesa_DeleteHashTable(ctx->PerfMonitor.Monitors);
}

static inline struct gl_perf_monitor_object *
lookup_monitor(struct gl_context *ctx, GLuint id)
{
   return (struct gl_perf_monitor_object *)
      _mesa_HashLookup(ctx->PerfMonitor.Monitors, id);
}

static inline const struct gl_perf_monitor_group *
get_group(const struct gl_context *ctx, GLuint id)
{
   if (id >= ctx->PerfMonitor.NumGroups)
      return NULL;

   return &ctx->PerfMonitor.Groups[id];
}

static inline const struct gl_perf_monitor_counter *
get_counter(const struct gl_perf_monitor_group *group_obj, GLuint id)
{
   if (id >= group_obj->NumCounters)
      return NULL;

   return &group_obj->Counters[id];
}

/* For INTEL_performance_query, query id 0 is reserved to be invalid. We use
 * index to Groups array + 1 as the query id. Same applies to counter id.
 */
static inline GLuint
queryid_to_index(GLuint queryid)
{
   return queryid - 1;
}

static inline GLuint
index_to_queryid(GLuint index)
{
   return index + 1;
}

static inline bool
queryid_valid(const struct gl_context *ctx, GLuint queryid)
{
   return get_group(ctx, queryid_to_index(queryid)) != NULL;
}

static inline GLuint
counterid_to_index(GLuint counterid)
{
   return counterid - 1;
}

/*****************************************************************************/

void GLAPIENTRY
_mesa_GetPerfMonitorGroupsAMD(GLint *numGroups, GLsizei groupsSize,
                              GLuint *groups)
{
   GET_CURRENT_CONTEXT(ctx);

   if (numGroups != NULL)
      *numGroups = ctx->PerfMonitor.NumGroups;

   if (groupsSize > 0 && groups != NULL) {
      unsigned i;
      unsigned n = MIN2((GLuint) groupsSize, ctx->PerfMonitor.NumGroups);

      /* We just use the index in the Groups array as the ID. */
      for (i = 0; i < n; i++)
         groups[i] = i;
   }
}

void GLAPIENTRY
_mesa_GetPerfMonitorCountersAMD(GLuint group, GLint *numCounters,
                                GLint *maxActiveCounters,
                                GLsizei countersSize, GLuint *counters)
{
   GET_CURRENT_CONTEXT(ctx);
   const struct gl_perf_monitor_group *group_obj = get_group(ctx, group);
   if (group_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfMonitorCountersAMD(invalid group)");
      return;
   }

   if (maxActiveCounters != NULL)
      *maxActiveCounters = group_obj->MaxActiveCounters;

   if (numCounters != NULL)
      *numCounters = group_obj->NumCounters;

   if (counters != NULL) {
      unsigned i;
      unsigned n = MIN2(group_obj->NumCounters, (GLuint) countersSize);
      for (i = 0; i < n; i++) {
         /* We just use the index in the Counters array as the ID. */
         counters[i] = i;
      }
   }
}

void GLAPIENTRY
_mesa_GetPerfMonitorGroupStringAMD(GLuint group, GLsizei bufSize,
                                   GLsizei *length, GLchar *groupString)
{
   GET_CURRENT_CONTEXT(ctx);

   const struct gl_perf_monitor_group *group_obj = get_group(ctx, group);

   if (group_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE, "glGetPerfMonitorGroupStringAMD");
      return;
   }

   if (bufSize == 0) {
      /* Return the number of characters that would be required to hold the
       * group string, excluding the null terminator.
       */
      if (length != NULL)
         *length = strlen(group_obj->Name);
   } else {
      if (length != NULL)
         *length = MIN2(strlen(group_obj->Name), bufSize);
      if (groupString != NULL)
         strncpy(groupString, group_obj->Name, bufSize);
   }
}

void GLAPIENTRY
_mesa_GetPerfMonitorCounterStringAMD(GLuint group, GLuint counter,
                                     GLsizei bufSize, GLsizei *length,
                                     GLchar *counterString)
{
   GET_CURRENT_CONTEXT(ctx);

   const struct gl_perf_monitor_group *group_obj;
   const struct gl_perf_monitor_counter *counter_obj;

   group_obj = get_group(ctx, group);

   if (group_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfMonitorCounterStringAMD(invalid group)");
      return;
   }

   counter_obj = get_counter(group_obj, counter);

   if (counter_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfMonitorCounterStringAMD(invalid counter)");
      return;
   }

   if (bufSize == 0) {
      /* Return the number of characters that would be required to hold the
       * counter string, excluding the null terminator.
       */
      if (length != NULL)
         *length = strlen(counter_obj->Name);
   } else {
      if (length != NULL)
         *length = MIN2(strlen(counter_obj->Name), bufSize);
      if (counterString != NULL)
         strncpy(counterString, counter_obj->Name, bufSize);
   }
}

void GLAPIENTRY
_mesa_GetPerfMonitorCounterInfoAMD(GLuint group, GLuint counter, GLenum pname,
                                   GLvoid *data)
{
   GET_CURRENT_CONTEXT(ctx);

   const struct gl_perf_monitor_group *group_obj;
   const struct gl_perf_monitor_counter *counter_obj;

   group_obj = get_group(ctx, group);

   if (group_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfMonitorCounterInfoAMD(invalid group)");
      return;
   }

   counter_obj = get_counter(group_obj, counter);

   if (counter_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfMonitorCounterInfoAMD(invalid counter)");
      return;
   }

   switch (pname) {
   case GL_COUNTER_TYPE_AMD:
      *((GLenum *) data) = counter_obj->Type;
      break;

   case GL_COUNTER_RANGE_AMD:
      switch (counter_obj->Type) {
      case GL_FLOAT:
      case GL_PERCENTAGE_AMD: {
         float *f_data = data;
         f_data[0] = counter_obj->Minimum.f;
         f_data[1] = counter_obj->Maximum.f;
         break;
      }
      case GL_UNSIGNED_INT: {
         uint32_t *u32_data = data;
         u32_data[0] = counter_obj->Minimum.u32;
         u32_data[1] = counter_obj->Maximum.u32;
         break;
      }
      case GL_UNSIGNED_INT64_AMD: {
         uint64_t *u64_data = data;
         u64_data[0] = counter_obj->Minimum.u64;
         u64_data[1] = counter_obj->Maximum.u64;
         break;
      }
      default:
         assert(!"Should not get here: invalid counter type");
      }
      break;

   default:
      _mesa_error(ctx, GL_INVALID_ENUM,
                  "glGetPerfMonitorCounterInfoAMD(pname)");
      return;
   }
}

void GLAPIENTRY
_mesa_GenPerfMonitorsAMD(GLsizei n, GLuint *monitors)
{
   GLuint first;
   GET_CURRENT_CONTEXT(ctx);

   if (MESA_VERBOSE & VERBOSE_API)
      _mesa_debug(ctx, "glGenPerfMonitorsAMD(%d)\n", n);

   if (n < 0) {
      _mesa_error(ctx, GL_INVALID_VALUE, "glGenPerfMonitorsAMD(n < 0)");
      return;
   }

   if (monitors == NULL)
      return;

   /* We don't actually need them to be contiguous, but this is what
    * the rest of Mesa does, so we may as well.
    */
   first = _mesa_HashFindFreeKeyBlock(ctx->PerfMonitor.Monitors, n);
   if (first) {
      GLsizei i;
      for (i = 0; i < n; i++) {
         struct gl_perf_monitor_object *m =
            new_performance_monitor(ctx, first + i);
         if (!m) {
            _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGenPerfMonitorsAMD");
            return;
         }
         monitors[i] = first + i;
         _mesa_HashInsert(ctx->PerfMonitor.Monitors, first + i, m);
      }
   } else {
      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGenPerfMonitorsAMD");
      return;
   }
}

void GLAPIENTRY
_mesa_DeletePerfMonitorsAMD(GLsizei n, GLuint *monitors)
{
   GLint i;
   GET_CURRENT_CONTEXT(ctx);

   if (MESA_VERBOSE & VERBOSE_API)
      _mesa_debug(ctx, "glDeletePerfMonitorsAMD(%d)\n", n);

   if (n < 0) {
      _mesa_error(ctx, GL_INVALID_VALUE, "glDeletePerfMonitorsAMD(n < 0)");
      return;
   }

   if (monitors == NULL)
      return;

   for (i = 0; i < n; i++) {
      struct gl_perf_monitor_object *m = lookup_monitor(ctx, monitors[i]);

      if (m) {
         /* Give the driver a chance to stop the monitor if it's active. */
         if (m->Active) {
            ctx->Driver.ResetPerfMonitor(ctx, m);
            m->Ended = false;
         }

         _mesa_HashRemove(ctx->PerfMonitor.Monitors, monitors[i]);
         ralloc_free(m->ActiveGroups);
         ralloc_free(m->ActiveCounters);
         ctx->Driver.DeletePerfMonitor(ctx, m);
      } else {
         /* "INVALID_VALUE error will be generated if any of the monitor IDs
          *  in the <monitors> parameter to DeletePerfMonitorsAMD do not
          *  reference a valid generated monitor ID."
          */
         _mesa_error(ctx, GL_INVALID_VALUE,
                     "glDeletePerfMonitorsAMD(invalid monitor)");
      }
   }
}

void GLAPIENTRY
_mesa_SelectPerfMonitorCountersAMD(GLuint monitor, GLboolean enable,
                                   GLuint group, GLint numCounters,
                                   GLuint *counterList)
{
   GET_CURRENT_CONTEXT(ctx);
   int i;
   struct gl_perf_monitor_object *m;
   const struct gl_perf_monitor_group *group_obj;

   m = lookup_monitor(ctx, monitor);

   /* "INVALID_VALUE error will be generated if the <monitor> parameter to
    *  SelectPerfMonitorCountersAMD does not reference a monitor created by
    *  GenPerfMonitorsAMD."
    */
   if (m == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glSelectPerfMonitorCountersAMD(invalid monitor)");
      return;
   }

   group_obj = get_group(ctx, group);

   /* "INVALID_VALUE error will be generated if the <group> parameter to
    *  GetPerfMonitorCountersAMD, GetPerfMonitorCounterStringAMD,
    *  GetPerfMonitorCounterStringAMD, GetPerfMonitorCounterInfoAMD, or
    *  SelectPerfMonitorCountersAMD does not reference a valid group ID."
    */
   if (group_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glSelectPerfMonitorCountersAMD(invalid group)");
      return;
   }

   /* "INVALID_VALUE error will be generated if the <numCounters> parameter to
    *  SelectPerfMonitorCountersAMD is less than 0."
    */
   if (numCounters < 0) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glSelectPerfMonitorCountersAMD(numCounters < 0)");
      return;
   }

   /* "When SelectPerfMonitorCountersAMD is called on a monitor, any outstanding
    *  results for that monitor become invalidated and the result queries
    *  PERFMON_RESULT_SIZE_AMD and PERFMON_RESULT_AVAILABLE_AMD are reset to 0."
    */
   ctx->Driver.ResetPerfMonitor(ctx, m);

   /* Sanity check the counter ID list. */
   for (i = 0; i < numCounters; i++) {
      if (counterList[i] >= group_obj->NumCounters) {
         _mesa_error(ctx, GL_INVALID_VALUE,
                     "glSelectPerfMonitorCountersAMD(invalid counter ID)");
         return;
      }
   }

   if (enable) {
      /* Enable the counters */
      for (i = 0; i < numCounters; i++) {
         ++m->ActiveGroups[group];
         BITSET_SET(m->ActiveCounters[group], counterList[i]);
      }
   } else {
      /* Disable the counters */
      for (i = 0; i < numCounters; i++) {
         --m->ActiveGroups[group];
         BITSET_CLEAR(m->ActiveCounters[group], counterList[i]);
      }
   }
}

void GLAPIENTRY
_mesa_BeginPerfMonitorAMD(GLuint monitor)
{
   GET_CURRENT_CONTEXT(ctx);

   struct gl_perf_monitor_object *m = lookup_monitor(ctx, monitor);

   if (m == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glBeginPerfMonitorAMD(invalid monitor)");
      return;
   }

   /* "INVALID_OPERATION error will be generated if BeginPerfMonitorAMD is
    *  called when a performance monitor is already active."
    */
   if (m->Active) {
      _mesa_error(ctx, GL_INVALID_OPERATION,
                  "glBeginPerfMonitor(already active)");
      return;
   }

   /* The driver is free to return false if it can't begin monitoring for
    * any reason.  This translates into an INVALID_OPERATION error.
    */
   if (ctx->Driver.BeginPerfMonitor(ctx, m)) {
      m->Active = true;
      m->Ended = false;
   } else {
      _mesa_error(ctx, GL_INVALID_OPERATION,
                  "glBeginPerfMonitor(driver unable to begin monitoring)");
   }
}

void GLAPIENTRY
_mesa_EndPerfMonitorAMD(GLuint monitor)
{
   GET_CURRENT_CONTEXT(ctx);

   struct gl_perf_monitor_object *m = lookup_monitor(ctx, monitor);

   if (m == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE, "glEndPerfMonitorAMD(invalid monitor)");
      return;
   }

   /* "INVALID_OPERATION error will be generated if EndPerfMonitorAMD is called
    *  when a performance monitor is not currently started."
    */
   if (!m->Active) {
      _mesa_error(ctx, GL_INVALID_OPERATION, "glBeginPerfMonitor(not active)");
      return;
   }

   ctx->Driver.EndPerfMonitor(ctx, m);

   m->Active = false;
   m->Ended = true;
}

/**
 * Return the number of bytes needed to store a monitor's result.
 */
static unsigned
perf_monitor_result_size(const struct gl_context *ctx,
                         const struct gl_perf_monitor_object *m)
{
   unsigned group, counter;
   unsigned size = 0;

   for (group = 0; group < ctx->PerfMonitor.NumGroups; group++) {
      const struct gl_perf_monitor_group *g = &ctx->PerfMonitor.Groups[group];
      for (counter = 0; counter < g->NumCounters; counter++) {
         const struct gl_perf_monitor_counter *c = &g->Counters[counter];

         if (!BITSET_TEST(m->ActiveCounters[group], counter))
            continue;

         size += sizeof(uint32_t); /* Group ID */
         size += sizeof(uint32_t); /* Counter ID */
         size += _mesa_perf_monitor_counter_size(c);
      }
   }
   return size;
}

void GLAPIENTRY
_mesa_GetPerfMonitorCounterDataAMD(GLuint monitor, GLenum pname,
                                   GLsizei dataSize, GLuint *data,
                                   GLint *bytesWritten)
{
   GET_CURRENT_CONTEXT(ctx);

   struct gl_perf_monitor_object *m = lookup_monitor(ctx, monitor);
   bool result_available;

   if (m == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfMonitorCounterDataAMD(invalid monitor)");
      return;
   }

   /* "It is an INVALID_OPERATION error for <data> to be NULL." */
   if (data == NULL) {
      _mesa_error(ctx, GL_INVALID_OPERATION,
                  "glGetPerfMonitorCounterDataAMD(data == NULL)");
      return;
   }

   /* We need at least enough room for a single value. */
   if (dataSize < sizeof(GLuint)) {
      if (bytesWritten != NULL)
         *bytesWritten = 0;
      return;
   }

   /* If the monitor has never ended, there is no result. */
   result_available = m->Ended &&
      ctx->Driver.IsPerfMonitorResultAvailable(ctx, m);

   /* AMD appears to return 0 for all queries unless a result is available. */
   if (!result_available) {
      *data = 0;
      if (bytesWritten != NULL)
         *bytesWritten = sizeof(GLuint);
      return;
   }

   switch (pname) {
   case GL_PERFMON_RESULT_AVAILABLE_AMD:
      *data = 1;
      if (bytesWritten != NULL)
         *bytesWritten = sizeof(GLuint);
      break;
   case GL_PERFMON_RESULT_SIZE_AMD:
      *data = perf_monitor_result_size(ctx, m);
      if (bytesWritten != NULL)
         *bytesWritten = sizeof(GLuint);
      break;
   case GL_PERFMON_RESULT_AMD:
      ctx->Driver.GetPerfMonitorResult(ctx, m, dataSize, data, bytesWritten);
      break;
   default:
      _mesa_error(ctx, GL_INVALID_ENUM,
                  "glGetPerfMonitorCounterDataAMD(pname)");
   }
}

/**
 * Returns how many bytes a counter's value takes up.
 */
unsigned
_mesa_perf_monitor_counter_size(const struct gl_perf_monitor_counter *c)
{
   switch (c->Type) {
   case GL_FLOAT:
   case GL_PERCENTAGE_AMD:
      return sizeof(GLfloat);
   case GL_UNSIGNED_INT:
      return sizeof(GLuint);
   case GL_UNSIGNED_INT64_AMD:
      return sizeof(uint64_t);
   default:
      assert(!"Should not get here: invalid counter type");
      return 0;
   }
}

extern void GLAPIENTRY
_mesa_GetFirstPerfQueryIdINTEL(GLuint *queryId)
{
   GET_CURRENT_CONTEXT(ctx);
   unsigned numGroups;

   /* The GL_INTEL_performance_query spec says:
    *
    *    "If queryId pointer is equal to 0, INVALID_VALUE error is generated."
    */
   if (!queryId) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetFirstPerfQueryIdINTEL(queryId == NULL)");
      return;
   }

   numGroups = ctx->PerfMonitor.NumGroups;

   /* The GL_INTEL_performance_query spec says:
    *
    *    "If the given hardware platform doesn't support any performance
    *    queries, then the value of 0 is returned and INVALID_OPERATION error
    *    is raised."
    */
   if (numGroups == 0) {
      *queryId = 0;
      _mesa_error(ctx, GL_INVALID_OPERATION,
                  "glGetFirstPerfQueryIdINTEL(no queries supported)");
      return;
   }

   *queryId = index_to_queryid(0);
}

extern void GLAPIENTRY
_mesa_GetNextPerfQueryIdINTEL(GLuint queryId, GLuint *nextQueryId)
{
   GET_CURRENT_CONTEXT(ctx);

   /* The GL_INTEL_performance_query spec says:
    *
    *    "The result is passed in location pointed by nextQueryId. If query
    *    identified by queryId is the last query available the value of 0 is
    *    returned. If the specified performance query identifier is invalid
    *    then INVALID_VALUE error is generated. If nextQueryId pointer is
    *    equal to 0, an INVALID_VALUE error is generated.  Whenever error is
    *    generated, the value of 0 is returned."
    */

   if (!nextQueryId) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetNextPerfQueryIdINTEL(nextQueryId == NULL)");
      return;
   }

   if (!queryid_valid(ctx, queryId)) {
      *nextQueryId = 0;
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetNextPerfQueryIdINTEL(invalid query)");
      return;
   }

   ++queryId;

   if (!queryid_valid(ctx, queryId)) {
      *nextQueryId = 0;
   } else {
      *nextQueryId = queryId;
   }
}

extern void GLAPIENTRY
_mesa_GetPerfQueryIdByNameINTEL(char *queryName, GLuint *queryId)
{
   GET_CURRENT_CONTEXT(ctx);
   unsigned i;

   /* The GL_INTEL_performance_query spec says:
    *
    *    "If queryName does not reference a valid query name, an INVALID_VALUE
    *    error is generated."
    */
   if (!queryName) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfQueryIdByNameINTEL(queryName == NULL)");
      return;
   }

   /* The specification does not state that this produces an error. */
   if (!queryId) {
      _mesa_warning(ctx, "glGetPerfQueryIdByNameINTEL(queryId == NULL)");
      return;
   }

   for (i = 0; i < ctx->PerfMonitor.NumGroups; ++i) {
      const struct gl_perf_monitor_group *group_obj = get_group(ctx, i);
      if (strcmp(group_obj->Name, queryName) == 0) {
         *queryId = index_to_queryid(i);
         return;
      }
   }

   _mesa_error(ctx, GL_INVALID_VALUE,
               "glGetPerfQueryIdByNameINTEL(invalid query name)");
}

extern void GLAPIENTRY
_mesa_GetPerfQueryInfoINTEL(GLuint queryId,
                            GLuint queryNameLength, char *queryName,
                            GLuint *dataSize, GLuint *noCounters,
                            GLuint *noActiveInstances,
                            GLuint *capsMask)
{
   GET_CURRENT_CONTEXT(ctx);
   unsigned i;

   const struct gl_perf_monitor_group *group_obj =
      get_group(ctx, queryid_to_index(queryId));

   if (group_obj == NULL) {
      /* The GL_INTEL_performance_query spec says:
       *
       *    "If queryId does not reference a valid query type, an
       *    INVALID_VALUE error is generated."
       */
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfQueryInfoINTEL(invalid query)");
      return;
   }

   if (queryName) {
      strncpy(queryName, group_obj->Name, queryNameLength);

      /* No specification given about whether the string needs to be
       * zero-terminated. Zero-terminate the string always as we don't
       * otherwise communicate the length of the returned string.
       */
      if (queryNameLength > 0) {
         queryName[queryNameLength - 1] = '\0';
      }
   }

   if (dataSize) {
      unsigned size = 0;
      for (i = 0; i < group_obj->NumCounters; ++i) {
         /* What we get from the driver is group id (uint32_t) + counter id
          * (uint32_t) + value.
          */
         size += 2 * sizeof(uint32_t) + _mesa_perf_monitor_counter_size(&group_obj->Counters[i]);
      }
      *dataSize = size;
   }

   if (noCounters) {
      *noCounters = group_obj->NumCounters;
   }

   /* The GL_INTEL_performance_query spec says:
    *
    *    "-- the actual number of already created query instances in
    *    maxInstances location"
    *
    * 1) Typo in the specification, should be noActiveInstances.
    * 2) Another typo in the specification, maxInstances parameter is not listed
    *    in the declaration of this function in the list of new functions.
    */
   if (noActiveInstances) {
      *noActiveInstances = _mesa_HashNumEntries(ctx->PerfMonitor.Monitors);
   }

   if (capsMask) {
      /* TODO: This information not yet available in the monitor structs. For
       * now, we hardcode SINGLE_CONTEXT, since that's what the implementation
       * currently tries very hard to do.
       */
      *capsMask = GL_PERFQUERY_SINGLE_CONTEXT_INTEL;
   }
}

extern void GLAPIENTRY
_mesa_GetPerfCounterInfoINTEL(GLuint queryId, GLuint counterId,
                              GLuint counterNameLength, char *counterName,
                              GLuint counterDescLength, char *counterDesc,
                              GLuint *counterOffset, GLuint *counterDataSize, GLuint *counterTypeEnum,
                              GLuint *counterDataTypeEnum, GLuint64 *rawCounterMaxValue)
{
   GET_CURRENT_CONTEXT(ctx);

   const struct gl_perf_monitor_group *group_obj;
   const struct gl_perf_monitor_counter *counter_obj;
   unsigned counterIndex;
   unsigned i;

   group_obj = get_group(ctx, queryid_to_index(queryId));

   /* The GL_INTEL_performance_query spec says:
    *
    *    "If the pair of queryId and counterId does not reference a valid
    *    counter, an INVALID_VALUE error is generated."
    */
   if (group_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfCounterInfoINTEL(invalid queryId)");
      return;
   }

   counterIndex = counterid_to_index(counterId);
   counter_obj = get_counter(group_obj, counterIndex);

   if (counter_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfCounterInfoINTEL(invalid counterId)");
      return;
   }

   if (counterName) {
      strncpy(counterName, counter_obj->Name, counterNameLength);

      /* No specification given about whether the string needs to be
       * zero-terminated. Zero-terminate the string always as we don't
       * otherwise communicate the length of the returned string.
       */
      if (counterNameLength > 0) {
         counterName[counterNameLength - 1] = '\0';
      }
   }

   if (counterDesc) {
      /* TODO: No separate description text at the moment. We pass the name
       * again for the moment.
       */
      strncpy(counterDesc, counter_obj->Name, counterDescLength);

      /* No specification given about whether the string needs to be
       * zero-terminated. Zero-terminate the string always as we don't
       * otherwise communicate the length of the returned string.
       */
      if (counterDescLength > 0) {
         counterDesc[counterDescLength - 1] = '\0';
      }
   }

   if (counterOffset) {
      unsigned offset = 0;
      for (i = 0; i < counterIndex; ++i) {
         /* What we get from the driver is group id (uint32_t) + counter id
          * (uint32_t) + value.
          */
         offset += 2 * sizeof(uint32_t) + _mesa_perf_monitor_counter_size(&group_obj->Counters[i]);
      }
      *counterOffset = 2 * sizeof(uint32_t) + offset;
   }

   if (counterDataSize) {
      *counterDataSize = _mesa_perf_monitor_counter_size(counter_obj);
   }

   if (counterTypeEnum) {
      /* TODO: Different counter types (semantic type, not data type) not
       * supported as of yet.
       */
      *counterTypeEnum = GL_PERFQUERY_COUNTER_RAW_INTEL;
   }

   if (counterDataTypeEnum) {
      switch (counter_obj->Type) {
      case GL_FLOAT:
      case GL_PERCENTAGE_AMD:
         *counterDataTypeEnum = GL_PERFQUERY_COUNTER_DATA_FLOAT_INTEL;
         break;
      case GL_UNSIGNED_INT:
         *counterDataTypeEnum = GL_PERFQUERY_COUNTER_DATA_UINT32_INTEL;
         break;
      case GL_UNSIGNED_INT64_AMD:
         *counterDataTypeEnum = GL_PERFQUERY_COUNTER_DATA_UINT64_INTEL;
         break;
      default:
         assert(!"Should not get here: invalid counter type");
         return;
      }
   }

   if (rawCounterMaxValue) {
      /* This value is (implicitly) specified to be used only with
       * GL_PERFQUERY_COUNTER_RAW_INTEL counters. When semantic types for
       * counters are added, that needs to be checked.
       */

      /* The GL_INTEL_performance_query spec says:
       *
       *    "for some raw counters for which the maximal value is
       *    deterministic, the maximal value of the counter in 1 second is
       *    returned in the location pointed by rawCounterMaxValue, otherwise,
       *    the location is written with the value of 0."
       *
       * The maximum value reported by the driver at the moment is not with
       * these semantics, so write 0 always to be safe.
       */
      *rawCounterMaxValue = 0;
   }
}

extern void GLAPIENTRY
_mesa_CreatePerfQueryINTEL(GLuint queryId, GLuint *queryHandle)
{
   GET_CURRENT_CONTEXT(ctx);
   GLuint first;
   GLuint group;
   const struct gl_perf_monitor_group *group_obj;
   struct gl_perf_monitor_object *m;
   unsigned i;

   /* This is not specified in the extension, but is the only sane thing to
    * do.
    */
   if (queryHandle == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glCreatePerfQueryINTEL(queryHandle == NULL)");
      return;
   }

   group = queryid_to_index(queryId);
   group_obj = get_group(ctx, group);

   /* The GL_INTEL_performance_query spec says:
    *
    *    "If queryId does not reference a valid query type, an INVALID_VALUE
    *    error is generated."
    */
   if (group_obj == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glCreatePerfQueryINTEL(invalid queryId)");
      return;
   }

   /* The query object created here is the counterpart of a `monitor' in
    * AMD_performance_monitor. This call is equivalent to calling
    * GenPerfMonitorsAMD and SelectPerfMonitorCountersAMD with a list of all
    * counters in a group.
    */

   /* We keep the monitor ids contiguous */
   first = _mesa_HashFindFreeKeyBlock(ctx->PerfMonitor.Monitors, 1);
   if (!first) {
      /* The GL_INTEL_performance_query spec says:
       *
       *    "If the query instance cannot be created due to exceeding the
       *    number of allowed instances or driver fails query creation due to
       *    an insufficient memory reason, an OUT_OF_MEMORY error is
       *    generated, and the location pointed by queryHandle returns NULL."
      */
      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCreatePerfQueryINTEL");
      return;
   }

   m = new_performance_monitor(ctx, first);
   if (m == NULL) {
      _mesa_error_no_memory(__func__);
      return;
   }

   _mesa_HashInsert(ctx->PerfMonitor.Monitors, first, m);
   *queryHandle = first;

   ctx->Driver.ResetPerfMonitor(ctx, m);

   for (i = 0; i < group_obj->NumCounters; ++i) {
      ++m->ActiveGroups[group];
      /* Counters are a continuous range of integers, 0 to NumCounters (excl),
       * so i is the counter value to use here.
       */
      BITSET_SET(m->ActiveCounters[group], i);
   }
}

extern void GLAPIENTRY
_mesa_DeletePerfQueryINTEL(GLuint queryHandle)
{
   GET_CURRENT_CONTEXT(ctx);
   struct gl_perf_monitor_object *m;

   /* The queryHandle is the counterpart to AMD_performance_monitor's monitor
    * id.
    */
   m = lookup_monitor(ctx, queryHandle);

   /* The GL_INTEL_performance_query spec says:
    *
    *    "If a query handle doesn't reference a previously created performance
    *    query instance, an INVALID_VALUE error is generated."
    */
   if (m == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glDeletePerfQueryINTEL(invalid queryHandle)");
      return;
   }

   /* Let the driver stop the monitor if it's active. */
   if (m->Active) {
      ctx->Driver.ResetPerfMonitor(ctx, m);
      m->Ended = false;
   }

   _mesa_HashRemove(ctx->PerfMonitor.Monitors, queryHandle);
   ralloc_free(m->ActiveGroups);
   ralloc_free(m->ActiveCounters);
   ctx->Driver.DeletePerfMonitor(ctx, m);
}

extern void GLAPIENTRY
_mesa_BeginPerfQueryINTEL(GLuint queryHandle)
{
   GET_CURRENT_CONTEXT(ctx);
   struct gl_perf_monitor_object *m;

   /* The queryHandle is the counterpart to AMD_performance_monitor's monitor
    * id.
    */

   m = lookup_monitor(ctx, queryHandle);

   /* The GL_INTEL_performance_query spec says:
    *
    *    "If a query handle doesn't reference a previously created performance
    *    query instance, an INVALID_VALUE error is generated."
    */
   if (m == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glBeginPerfQueryINTEL(invalid queryHandle)");
      return;
   }

   /* The GL_INTEL_performance_query spec says:
    *
    *    "Note that some query types, they cannot be collected in the same
    *    time. Therefore calls of BeginPerfQueryINTEL() cannot be nested if
    *    they refer to queries of such different types. In such case
    *    INVALID_OPERATION error is generated."
    *
    * We also generate an INVALID_OPERATION error if the driver can't begin
    * monitoring for its own reasons, and for nesting the same query.
    */
   if (m->Active) {
      _mesa_error(ctx, GL_INVALID_OPERATION,
                  "glBeginPerfQueryINTEL(already active)");
      return;
   }

   if (ctx->Driver.BeginPerfMonitor(ctx, m)) {
      m->Active = true;
      m->Ended = false;
   } else {
      _mesa_error(ctx, GL_INVALID_OPERATION,
                  "glBeginPerfQueryINTEL(driver unable to begin monitoring)");
   }
}

extern void GLAPIENTRY
_mesa_EndPerfQueryINTEL(GLuint queryHandle)
{
   GET_CURRENT_CONTEXT(ctx);
   struct gl_perf_monitor_object *m;

   /* The queryHandle is the counterpart to AMD_performance_monitor's monitor
    * id.
    */

   m = lookup_monitor(ctx, queryHandle);

   /* The GL_INTEL_performance_query spec says:
    *
    *    "If a performance query is not currently started, an
    *    INVALID_OPERATION error will be generated."
    *
    * The specification doesn't state that an invalid handle would be an
    * INVALID_VALUE error. Regardless, query for such a handle will not be
    * started, so we generate an INVALID_OPERATION in that case too.
    */
   if (m == NULL) {
      _mesa_error(ctx, GL_INVALID_OPERATION,
                  "glEndPerfQueryINTEL(invalid queryHandle)");
      return;
   }

   if (!m->Active) {
      _mesa_error(ctx, GL_INVALID_OPERATION,
                  "glEndPerfQueryINTEL(not active)");
      return;
   }

   ctx->Driver.EndPerfMonitor(ctx, m);

   m->Active = false;
   m->Ended = true;
}

extern void GLAPIENTRY
_mesa_GetPerfQueryDataINTEL(GLuint queryHandle, GLuint flags,
                            GLsizei dataSize, void *data, GLuint *bytesWritten)
{
   GET_CURRENT_CONTEXT(ctx);
   struct gl_perf_monitor_object *m;
   bool result_available;

   /* The GL_INTEL_performance_query spec says:
    *
    *    "If bytesWritten or data pointers are NULL then an INVALID_VALUE
    *    error is generated."
    */
   if (!bytesWritten || !data) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfQueryDataINTEL(bytesWritten or data is NULL)");
      return;
   }

   /* The queryHandle is the counterpart to AMD_performance_monitor's monitor
    * id.
    */

   m = lookup_monitor(ctx, queryHandle);

   /* The specification doesn't state that an invalid handle generates an
    * error. We could interpret that to mean the case should be handled as
    * "measurement not ready for this query", but what should be done if
    * `flags' equals PERFQUERY_WAIT_INTEL?
    *
    * To resolve this, we just generate an INVALID_VALUE from an invalid query
    * handle.
    */
   if (m == NULL) {
      _mesa_error(ctx, GL_INVALID_VALUE,
                  "glGetPerfQueryDataINTEL(invalid queryHandle)");
      return;
   }

   /* We need at least enough room for a single value. */
   if (dataSize < sizeof(GLuint)) {
      *bytesWritten = 0;
      return;
   }

   /* The GL_INTEL_performance_query spec says:
    *
    *    "The call may end without returning any data if they are not ready
    *    for reading as the measurement session is still pending (the
    *    EndPerfQueryINTEL() command processing is not finished by
    *    hardware). In this case location pointed by the bytesWritten
    *    parameter will be set to 0."
    *
    * If EndPerfQueryINTEL() is not called at all, we follow this.
    */
   if (!m->Ended) {
      *bytesWritten = 0;
      return;
   }

   result_available = ctx->Driver.IsPerfMonitorResultAvailable(ctx, m);

   if (!result_available) {
      if (flags == GL_PERFQUERY_FLUSH_INTEL) {
         ctx->Driver.Flush(ctx);
      } else if (flags == GL_PERFQUERY_WAIT_INTEL) {
         /* Assume Finish() is both enough and not too much to wait for
          * results. If results are still not available after Finish(), the
          * later code automatically bails out with 0 for bytesWritten.
          */
         ctx->Driver.Finish(ctx);
         result_available =
            ctx->Driver.IsPerfMonitorResultAvailable(ctx, m);
      }
   }

   if (result_available) {
      ctx->Driver.GetPerfMonitorResult(ctx, m, dataSize, data, (GLint*)bytesWritten);
   } else {
      *bytesWritten = 0;
   }
}