aboutsummaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/iris/iris_monitor.c
blob: 80bddcc3b316363c1dea3cf9061e8bcd4c320b0f (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
/*
 * Copyright © 2019 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 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.
 */

#include "iris_monitor.h"

#include <xf86drm.h>

#include "iris_screen.h"
#include "iris_context.h"
#include "iris_perf.h"

struct iris_monitor_object {
   int num_active_counters;
   int *active_counters;

   size_t result_size;
   unsigned char *result_buffer;

   struct gen_perf_query_object *query;
};

int
iris_get_monitor_info(struct pipe_screen *pscreen, unsigned index,
                      struct pipe_driver_query_info *info)
{
   const struct iris_screen *screen = (struct iris_screen *)pscreen;
   const struct gen_perf_config *perf_cfg = screen->perf_cfg;
   assert(perf_cfg);
   if (!perf_cfg)
      return 0;

   if (!info) {
      /* return the number of metrics */
      return perf_cfg->n_counters;
   }

   struct gen_perf_query_counter *counter = perf_cfg->counters[index];

   info->group_id = counter->location.group_idx;
   info->name = counter->name;
   info->query_type = PIPE_QUERY_DRIVER_SPECIFIC + index;

   if (counter->type == GEN_PERF_COUNTER_TYPE_THROUGHPUT)
      info->result_type = PIPE_DRIVER_QUERY_RESULT_TYPE_AVERAGE;
   else
      info->result_type = PIPE_DRIVER_QUERY_RESULT_TYPE_CUMULATIVE;
   switch (counter->data_type) {
   case GEN_PERF_COUNTER_DATA_TYPE_BOOL32:
   case GEN_PERF_COUNTER_DATA_TYPE_UINT32:
      info->type = PIPE_DRIVER_QUERY_TYPE_UINT;
      assert(counter->raw_max <= UINT32_MAX);
      info->max_value.u32 = (uint32_t)counter->raw_max;
      break;
   case GEN_PERF_COUNTER_DATA_TYPE_UINT64:
      info->type = PIPE_DRIVER_QUERY_TYPE_UINT64;
      info->max_value.u64 = counter->raw_max;
      break;
   case GEN_PERF_COUNTER_DATA_TYPE_FLOAT:
   case GEN_PERF_COUNTER_DATA_TYPE_DOUBLE:
      info->type = PIPE_DRIVER_QUERY_TYPE_FLOAT;
      info->max_value.f = counter->raw_max;
      break;
   default:
      assert(false);
      break;
   }

   /* indicates that this is an OA query, not a pipeline statistics query */
   info->flags = PIPE_DRIVER_QUERY_FLAG_BATCH;
   return 1;
}

static bool
iris_monitor_init_metrics(struct iris_screen *screen)
{
   struct gen_perf_config *perf_cfg = gen_perf_new(screen);
   if (unlikely(!perf_cfg))
      return false;

   screen->perf_cfg = perf_cfg;

   iris_perf_init_vtbl(perf_cfg);

   gen_perf_init_metrics(perf_cfg, &screen->devinfo, screen->fd,
                         true /* pipeline stats*/);

   return perf_cfg->n_counters > 0;
}

int
iris_get_monitor_group_info(struct pipe_screen *pscreen,
                            unsigned group_index,
                            struct pipe_driver_query_group_info *info)
{
   struct iris_screen *screen = (struct iris_screen *)pscreen;
   if (!screen->perf_cfg) {
      if (!iris_monitor_init_metrics(screen))
         return 0;
   }

   const struct gen_perf_config *perf_cfg = screen->perf_cfg;

   if (!info) {
      /* return the count that can be queried */
      return perf_cfg->n_queries;
   }

   if (group_index >= perf_cfg->n_queries) {
      /* out of range */
      return 0;
   }

   struct gen_perf_query_info *query = &perf_cfg->queries[group_index];

   info->name = query->name;
   info->max_active_queries = query->n_counters;
   info->num_queries = query->n_counters;

   return 1;
}

static void
iris_init_monitor_ctx(struct iris_context *ice)
{
   struct iris_screen *screen = (struct iris_screen *) ice->ctx.screen;

   ice->perf_ctx = gen_perf_new_context(ice);
   if (unlikely(!ice->perf_ctx))
      return;

   struct gen_perf_context *perf_ctx = ice->perf_ctx;
   struct gen_perf_config *perf_cfg = screen->perf_cfg;
   gen_perf_init_context(perf_ctx,
                         perf_cfg,
                         ice,
                         screen->bufmgr,
                         &screen->devinfo,
                         ice->batches[IRIS_BATCH_RENDER].hw_ctx_id,
                         screen->fd);
}

/* entry point for GenPerfMonitorsAMD */
struct iris_monitor_object *
iris_create_monitor_object(struct iris_context *ice,
                           unsigned num_queries,
                           unsigned *query_types)
{
   struct iris_screen *screen = (struct iris_screen *) ice->ctx.screen;
   struct gen_perf_config *perf_cfg = screen->perf_cfg;
   struct gen_perf_query_object *query_obj = NULL;

   /* initialize perf context if this has not already been done.  This
    * function is the first entry point that carries the gl context.
    */
   if (ice->perf_ctx == NULL) {
      iris_init_monitor_ctx(ice);
   }
   struct gen_perf_context *perf_ctx = ice->perf_ctx;

   assert(num_queries > 0);
   int query_index = query_types[0] - PIPE_QUERY_DRIVER_SPECIFIC;
   assert(query_index <= perf_cfg->n_counters);
   const int group = perf_cfg->counters[query_index]->location.group_idx;

   struct iris_monitor_object *monitor =
      calloc(1, sizeof(struct iris_monitor_object));
   if (unlikely(!monitor))
      goto allocation_failure;

   monitor->num_active_counters = num_queries;
   monitor->active_counters = calloc(num_queries, sizeof(int));
   if (unlikely(!monitor->active_counters))
      goto allocation_failure;

   for (int i = 0; i < num_queries; ++i) {
      unsigned current_query = query_types[i];
      unsigned current_query_index = current_query - PIPE_QUERY_DRIVER_SPECIFIC;

      /* all queries must be in the same group */
      assert(current_query_index <= perf_cfg->n_counters);
      assert(perf_cfg->counters[current_query_index]->location.group_idx == group);
      monitor->active_counters[i] =
         perf_cfg->counters[current_query_index]->location.counter_idx;
   }

   /* create the gen_perf_query */
   query_obj = gen_perf_new_query(perf_ctx, group);
   if (unlikely(!query_obj))
      goto allocation_failure;

   monitor->query = query_obj;
   monitor->result_size = perf_cfg->queries[group].data_size;
   monitor->result_buffer = calloc(1, monitor->result_size);
   if (unlikely(!monitor->result_buffer))
      goto allocation_failure;

   return monitor;

allocation_failure:
   if (monitor) {
      free(monitor->active_counters);
      free(monitor->result_buffer);
   }
   free(query_obj);
   free(monitor);
   return NULL;
}

void
iris_destroy_monitor_object(struct pipe_context *ctx,
                            struct iris_monitor_object *monitor)
{
   struct iris_context *ice = (struct iris_context *)ctx;

   gen_perf_delete_query(ice->perf_ctx, monitor->query);
   free(monitor->result_buffer);
   monitor->result_buffer = NULL;
   free(monitor->active_counters);
   monitor->active_counters = NULL;
   free(monitor);
}

bool
iris_begin_monitor(struct pipe_context *ctx,
                   struct iris_monitor_object *monitor)
{
   struct iris_context *ice = (void *) ctx;
   struct gen_perf_context *perf_ctx = ice->perf_ctx;

   return gen_perf_begin_query(perf_ctx, monitor->query);
}

bool
iris_end_monitor(struct pipe_context *ctx,
                 struct iris_monitor_object *monitor)
{
   struct iris_context *ice = (void *) ctx;
   struct gen_perf_context *perf_ctx = ice->perf_ctx;

   gen_perf_end_query(perf_ctx, monitor->query);
   return true;
}

bool
iris_get_monitor_result(struct pipe_context *ctx,
                        struct iris_monitor_object *monitor,
                        bool wait,
                        union pipe_numeric_type_union *result)
{
   struct iris_context *ice = (void *) ctx;
   struct gen_perf_context *perf_ctx = ice->perf_ctx;
   struct iris_batch *batch = &ice->batches[IRIS_BATCH_RENDER];

   bool monitor_ready =
      gen_perf_is_query_ready(perf_ctx, monitor->query, batch);

   if (!monitor_ready) {
      if (!wait)
         return false;
      gen_perf_wait_query(perf_ctx, monitor->query, batch);
   }

   assert(gen_perf_is_query_ready(perf_ctx, monitor->query, batch));

   unsigned bytes_written;
   gen_perf_get_query_data(perf_ctx, monitor->query,
                           monitor->result_size,
                           (unsigned*) monitor->result_buffer,
                           &bytes_written);
   if (bytes_written != monitor->result_size)
      return false;

   /* copy metrics into the batch result */
   for (int i = 0; i < monitor->num_active_counters; ++i) {
      int current_counter = monitor->active_counters[i];
      const struct gen_perf_query_info *info =
         gen_perf_query_info(monitor->query);
      const struct gen_perf_query_counter *counter =
         &info->counters[current_counter];
      assert(gen_perf_query_counter_get_size(counter));
      switch (counter->data_type) {
      case GEN_PERF_COUNTER_DATA_TYPE_UINT64:
         result[i].u64 = *(uint64_t*)(monitor->result_buffer + counter->offset);
         break;
      case GEN_PERF_COUNTER_DATA_TYPE_FLOAT:
         result[i].f = *(float*)(monitor->result_buffer + counter->offset);
         break;
      case GEN_PERF_COUNTER_DATA_TYPE_UINT32:
      case GEN_PERF_COUNTER_DATA_TYPE_BOOL32:
         result[i].u64 = *(uint32_t*)(monitor->result_buffer + counter->offset);
         break;
      case GEN_PERF_COUNTER_DATA_TYPE_DOUBLE: {
         double v = *(double*)(monitor->result_buffer + counter->offset);
         result[i].f = v;
         break;
      }
      default:
         unreachable("unexpected counter data type");
      }
   }
   return true;
}