summaryrefslogtreecommitdiff
path: root/src/mesa/drivers/dri/i965/brw_queryobj.c
blob: 1b183735d756a9d19d7be8475baf4cf9ce461199 (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
/*
 * Copyright © 2008 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.
 *
 * Authors:
 *    Eric Anholt <eric@anholt.net>
 *
 */

/** @file support for ARB_query_object
 *
 * ARB_query_object is implemented by using the PIPE_CONTROL command to stall
 * execution on the completion of previous depth tests, and write the
 * current PS_DEPTH_COUNT to a buffer object.
 *
 * We use before and after counts when drawing during a query so that
 * we don't pick up other clients' query data in ours.  To reduce overhead,
 * a single BO is used to record the query data for all active queries at
 * once.  This also gives us a simple bound on how much batchbuffer space is
 * required for handling queries, so that we can be sure that we won't
 * have to emit a batchbuffer without getting the ending PS_DEPTH_COUNT.
 */
#include "main/imports.h"

#include "brw_context.h"
#include "brw_state.h"
#include "intel_batchbuffer.h"
#include "intel_reg.h"

/** Waits on the query object's BO and totals the results for this query */
static void
brw_queryobj_get_results(struct brw_query_object *query)
{
   int i;
   uint64_t *results;

   if (query->bo == NULL)
      return;

   drm_intel_bo_map(query->bo, GL_FALSE);
   results = query->bo->virtual;
   if (query->Base.Target == GL_TIME_ELAPSED_EXT) {
      query->Base.Result += 1000 * ((results[1] >> 32) - (results[0] >> 32));
   } else {
      /* Map and count the pixels from the current query BO */
      for (i = query->first_index; i <= query->last_index; i++) {
	 query->Base.Result += results[i * 2 + 1] - results[i * 2];
      }
   }
   drm_intel_bo_unmap(query->bo);

   drm_intel_bo_unreference(query->bo);
   query->bo = NULL;
}

static struct gl_query_object *
brw_new_query_object(GLcontext *ctx, GLuint id)
{
   struct brw_query_object *query;

   query = calloc(1, sizeof(struct brw_query_object));

   query->Base.Id = id;
   query->Base.Result = 0;
   query->Base.Active = GL_FALSE;
   query->Base.Ready = GL_TRUE;

   return &query->Base;
}

static void
brw_delete_query(GLcontext *ctx, struct gl_query_object *q)
{
   struct brw_query_object *query = (struct brw_query_object *)q;

   drm_intel_bo_unreference(query->bo);
   free(query);
}

static void
brw_begin_query(GLcontext *ctx, struct gl_query_object *q)
{
   struct brw_context *brw = brw_context(ctx);
   struct intel_context *intel = intel_context(ctx);
   struct brw_query_object *query = (struct brw_query_object *)q;

   if (query->Base.Target == GL_TIME_ELAPSED_EXT) {
      drm_intel_bo_unreference(query->bo);
      query->bo = drm_intel_bo_alloc(intel->bufmgr, "timer query",
				     4096, 4096);

      if (intel->gen >= 6) {
	  BEGIN_BATCH(4);
	  OUT_BATCH(_3DSTATE_PIPE_CONTROL);
	  OUT_BATCH(PIPE_CONTROL_WRITE_TIMESTAMP);
	  OUT_RELOC(query->bo,
		  I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
		  PIPE_CONTROL_GLOBAL_GTT_WRITE |
		  0);
	  OUT_BATCH(0);
	  ADVANCE_BATCH();
      
      } else {
	  BEGIN_BATCH(4);
	  OUT_BATCH(_3DSTATE_PIPE_CONTROL |
		  PIPE_CONTROL_WRITE_TIMESTAMP);
	  OUT_RELOC(query->bo,
		  I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
		  PIPE_CONTROL_GLOBAL_GTT_WRITE |
		  0);
	  OUT_BATCH(0);
	  OUT_BATCH(0);
	  ADVANCE_BATCH();
      }
   } else {
      /* Reset our driver's tracking of query state. */
      drm_intel_bo_unreference(query->bo);
      query->bo = NULL;
      query->first_index = -1;
      query->last_index = -1;

      brw->query.obj = query;
      intel->stats_wm++;
   }
}

/**
 * Begin the ARB_occlusion_query query on a query object.
 */
static void
brw_end_query(GLcontext *ctx, struct gl_query_object *q)
{
   struct brw_context *brw = brw_context(ctx);
   struct intel_context *intel = intel_context(ctx);
   struct brw_query_object *query = (struct brw_query_object *)q;

   if (query->Base.Target == GL_TIME_ELAPSED_EXT) {
      if (intel->gen >= 6) {
	  BEGIN_BATCH(4);
	  OUT_BATCH(_3DSTATE_PIPE_CONTROL);
	  OUT_BATCH(PIPE_CONTROL_WRITE_TIMESTAMP);
	  OUT_RELOC(query->bo,
		  I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
		  PIPE_CONTROL_GLOBAL_GTT_WRITE |
		  8);
	  OUT_BATCH(0);
	  ADVANCE_BATCH();
      
      } else {
	  BEGIN_BATCH(4);
	  OUT_BATCH(_3DSTATE_PIPE_CONTROL |
		  PIPE_CONTROL_WRITE_TIMESTAMP);
	  OUT_RELOC(query->bo,
		  I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
		  PIPE_CONTROL_GLOBAL_GTT_WRITE |
		  8);
	  OUT_BATCH(0);
	  OUT_BATCH(0);
	  ADVANCE_BATCH();
      }

      intel_batchbuffer_flush(intel->batch);
   } else {
      /* Flush the batchbuffer in case it has writes to our query BO.
       * Have later queries write to a new query BO so that further rendering
       * doesn't delay the collection of our results.
       */
      if (query->bo) {
	 brw_emit_query_end(brw);
	 intel_batchbuffer_flush(intel->batch);

	 drm_intel_bo_unreference(brw->query.bo);
	 brw->query.bo = NULL;
      }

      brw->query.obj = NULL;

      intel->stats_wm--;
   }
}

static void brw_wait_query(GLcontext *ctx, struct gl_query_object *q)
{
   struct brw_query_object *query = (struct brw_query_object *)q;

   brw_queryobj_get_results(query);
   query->Base.Ready = GL_TRUE;
}

static void brw_check_query(GLcontext *ctx, struct gl_query_object *q)
{
   struct brw_query_object *query = (struct brw_query_object *)q;

   if (query->bo == NULL || !drm_intel_bo_busy(query->bo)) {
      brw_queryobj_get_results(query);
      query->Base.Ready = GL_TRUE;
   }
}

/** Called to set up the query BO and account for its aperture space */
void
brw_prepare_query_begin(struct brw_context *brw)
{
   struct intel_context *intel = &brw->intel;

   /* Skip if we're not doing any queries. */
   if (!brw->query.obj)
      return;

   /* Get a new query BO if we're going to need it. */
   if (brw->query.bo == NULL ||
       brw->query.index * 2 + 1 >= 4096 / sizeof(uint64_t)) {
      drm_intel_bo_unreference(brw->query.bo);
      brw->query.bo = NULL;

      brw->query.bo = drm_intel_bo_alloc(intel->bufmgr, "query", 4096, 1);
      brw->query.index = 0;
   }

   brw_add_validated_bo(brw, brw->query.bo);
}

/** Called just before primitive drawing to get a beginning PS_DEPTH_COUNT. */
void
brw_emit_query_begin(struct brw_context *brw)
{
   struct intel_context *intel = &brw->intel;
   struct brw_query_object *query = brw->query.obj;

   /* Skip if we're not doing any queries, or we've emitted the start. */
   if (!query || brw->query.active)
      return;

   if (intel->gen >= 6) {
       BEGIN_BATCH(8);

       /* workaround: CS stall required before depth stall. */
       OUT_BATCH(_3DSTATE_PIPE_CONTROL);
       OUT_BATCH(PIPE_CONTROL_CS_STALL);
       OUT_BATCH(0); /* write address */
       OUT_BATCH(0); /* write data */

       OUT_BATCH(_3DSTATE_PIPE_CONTROL);
       OUT_BATCH(PIPE_CONTROL_DEPTH_STALL |
	         PIPE_CONTROL_WRITE_DEPTH_COUNT);
       OUT_RELOC(brw->query.bo,
	         I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
		 PIPE_CONTROL_GLOBAL_GTT_WRITE |
		 ((brw->query.index * 2) * sizeof(uint64_t)));
       OUT_BATCH(0);
       ADVANCE_BATCH();
       
   } else {
       BEGIN_BATCH(4);
       OUT_BATCH(_3DSTATE_PIPE_CONTROL |
	       PIPE_CONTROL_DEPTH_STALL |
	       PIPE_CONTROL_WRITE_DEPTH_COUNT);
       /* This object could be mapped cacheable, but we don't have an exposed
	* mechanism to support that.  Since it's going uncached, tell GEM that
	* we're writing to it.  The usual clflush should be all that's required
	* to pick up the results.
	*/
       OUT_RELOC(brw->query.bo,
	       I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
	       PIPE_CONTROL_GLOBAL_GTT_WRITE |
	       ((brw->query.index * 2) * sizeof(uint64_t)));
       OUT_BATCH(0);
       OUT_BATCH(0);
       ADVANCE_BATCH();
   }

   if (query->bo != brw->query.bo) {
      if (query->bo != NULL)
	 brw_queryobj_get_results(query);
      drm_intel_bo_reference(brw->query.bo);
      query->bo = brw->query.bo;
      query->first_index = brw->query.index;
   }
   query->last_index = brw->query.index;
   brw->query.active = GL_TRUE;
}

/** Called at batchbuffer flush to get an ending PS_DEPTH_COUNT */
void
brw_emit_query_end(struct brw_context *brw)
{
   struct intel_context *intel = &brw->intel;

   if (!brw->query.active)
      return;

   if (intel->gen >= 6) {
       BEGIN_BATCH(8);
       /* workaround: CS stall required before depth stall. */
       OUT_BATCH(_3DSTATE_PIPE_CONTROL);
       OUT_BATCH(PIPE_CONTROL_CS_STALL);
       OUT_BATCH(0); /* write address */
       OUT_BATCH(0); /* write data */

       OUT_BATCH(_3DSTATE_PIPE_CONTROL);
       OUT_BATCH(PIPE_CONTROL_DEPTH_STALL |
	         PIPE_CONTROL_WRITE_DEPTH_COUNT);
       OUT_RELOC(brw->query.bo,
	         I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
		 PIPE_CONTROL_GLOBAL_GTT_WRITE |
		 ((brw->query.index * 2 + 1) * sizeof(uint64_t)));
       OUT_BATCH(0);
       ADVANCE_BATCH();
   
   } else {
       BEGIN_BATCH(4);
       OUT_BATCH(_3DSTATE_PIPE_CONTROL |
	       PIPE_CONTROL_DEPTH_STALL |
	       PIPE_CONTROL_WRITE_DEPTH_COUNT);
       OUT_RELOC(brw->query.bo,
	       I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
	       PIPE_CONTROL_GLOBAL_GTT_WRITE |
	       ((brw->query.index * 2 + 1) * sizeof(uint64_t)));
       OUT_BATCH(0);
       OUT_BATCH(0);
       ADVANCE_BATCH();
   }

   brw->query.active = GL_FALSE;
   brw->query.index++;
}

void brw_init_queryobj_functions(struct dd_function_table *functions)
{
   functions->NewQueryObject = brw_new_query_object;
   functions->DeleteQuery = brw_delete_query;
   functions->BeginQuery = brw_begin_query;
   functions->EndQuery = brw_end_query;
   functions->CheckQuery = brw_check_query;
   functions->WaitQuery = brw_wait_query;
}