2 * Copyright © 2008 Intel Corporation
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
24 * Eric Anholt <eric@anholt.net>
28 /** @file support for ARB_query_object
30 * ARB_query_object is implemented by using the PIPE_CONTROL command to stall
31 * execution on the completion of previous depth tests, and write the
32 * current PS_DEPTH_COUNT to a buffer object.
34 * We use before and after counts when drawing during a query so that
35 * we don't pick up other clients' query data in ours. To reduce overhead,
36 * a single BO is used to record the query data for all active queries at
37 * once. This also gives us a simple bound on how much batchbuffer space is
38 * required for handling queries, so that we can be sure that we won't
39 * have to emit a batchbuffer without getting the ending PS_DEPTH_COUNT.
41 #include "main/simple_list.h"
42 #include "main/imports.h"
44 #include "brw_context.h"
45 #include "intel_batchbuffer.h"
46 #include "intel_reg.h"
48 /** Waits on the query object's BO and totals the results for this query */
50 brw_queryobj_get_results(struct brw_query_object
*query
)
55 if (query
->bo
== NULL
)
58 /* Map and count the pixels from the current query BO */
59 dri_bo_map(query
->bo
, GL_FALSE
);
60 results
= query
->bo
->virtual;
61 for (i
= query
->first_index
; i
<= query
->last_index
; i
++) {
62 query
->Base
.Result
+= results
[i
* 2 + 1] - results
[i
* 2];
64 dri_bo_unmap(query
->bo
);
66 dri_bo_unreference(query
->bo
);
70 static struct gl_query_object
*
71 brw_new_query_object(GLcontext
*ctx
, GLuint id
)
73 struct brw_query_object
*query
;
75 query
= _mesa_calloc(sizeof(struct brw_query_object
));
78 query
->Base
.Result
= 0;
79 query
->Base
.Active
= GL_FALSE
;
80 query
->Base
.Ready
= GL_TRUE
;
86 brw_delete_query(GLcontext
*ctx
, struct gl_query_object
*q
)
88 struct brw_query_object
*query
= (struct brw_query_object
*)q
;
90 dri_bo_unreference(query
->bo
);
95 brw_begin_query(GLcontext
*ctx
, struct gl_query_object
*q
)
97 struct brw_context
*brw
= brw_context(ctx
);
98 struct intel_context
*intel
= intel_context(ctx
);
99 struct brw_query_object
*query
= (struct brw_query_object
*)q
;
101 /* Reset our driver's tracking of query state. */
102 dri_bo_unreference(query
->bo
);
104 query
->first_index
= -1;
105 query
->last_index
= -1;
107 insert_at_head(&brw
->query
.active_head
, query
);
112 * Begin the ARB_occlusion_query query on a query object.
115 brw_end_query(GLcontext
*ctx
, struct gl_query_object
*q
)
117 struct brw_context
*brw
= brw_context(ctx
);
118 struct intel_context
*intel
= intel_context(ctx
);
119 struct brw_query_object
*query
= (struct brw_query_object
*)q
;
121 /* Flush the batchbuffer in case it has writes to our query BO.
122 * Have later queries write to a new query BO so that further rendering
123 * doesn't delay the collection of our results.
126 brw_emit_query_end(brw
);
127 intel_batchbuffer_flush(intel
->batch
);
129 dri_bo_unreference(brw
->query
.bo
);
130 brw
->query
.bo
= NULL
;
133 remove_from_list(query
);
138 static void brw_wait_query(GLcontext
*ctx
, struct gl_query_object
*q
)
140 struct brw_query_object
*query
= (struct brw_query_object
*)q
;
142 brw_queryobj_get_results(query
);
143 query
->Base
.Ready
= GL_TRUE
;
146 static void brw_check_query(GLcontext
*ctx
, struct gl_query_object
*q
)
148 /* XXX: Need to expose dri_bo_is_idle from bufmgr. */
150 struct brw_query_object
*query
= (struct brw_query_object
*)q
;
152 if (dri_bo_is_idle(query
->bo
)) {
153 brw_queryobj_get_results(query
);
154 query
->Base
.Ready
= GL_TRUE
;
157 brw_wait_query(ctx
, q
);
161 /** Called to set up the query BO and account for its aperture space */
163 brw_prepare_query_begin(struct brw_context
*brw
)
165 struct intel_context
*intel
= &brw
->intel
;
166 dri_bo
*aper_array
[] = {
171 /* Skip if we're not doing any queries. */
172 if (is_empty_list(&brw
->query
.active_head
))
175 /* Get a new query BO if we're going to need it. */
176 if (brw
->query
.bo
== NULL
||
177 brw
->query
.index
* 2 + 1 >= 4096 / sizeof(uint64_t)) {
178 dri_bo_unreference(brw
->query
.bo
);
179 brw
->query
.bo
= NULL
;
181 brw
->query
.bo
= dri_bo_alloc(intel
->bufmgr
, "query", 4096, 1);
182 brw
->query
.index
= 0;
185 if (dri_bufmgr_check_aperture_space(aper_array
, ARRAY_SIZE(aper_array
)))
186 intel_batchbuffer_flush(intel
->batch
);
189 /** Called just before primitive drawing to get a beginning PS_DEPTH_COUNT. */
191 brw_emit_query_begin(struct brw_context
*brw
)
193 struct intel_context
*intel
= &brw
->intel
;
194 struct brw_query_object
*query
;
196 /* Skip if we're not doing any queries, or we've emitted the start. */
197 if (brw
->query
.active
|| is_empty_list(&brw
->query
.active_head
))
200 BEGIN_BATCH(4, IGNORE_CLIPRECTS
);
201 OUT_BATCH(_3DSTATE_PIPE_CONTROL
|
202 PIPE_CONTROL_DEPTH_STALL
|
203 PIPE_CONTROL_WRITE_DEPTH_COUNT
);
204 /* This object could be mapped cacheable, but we don't have an exposed
205 * mechanism to support that. Since it's going uncached, tell GEM that
206 * we're writing to it. The usual clflush should be all that's required
207 * to pick up the results.
209 OUT_RELOC(brw
->query
.bo
,
210 I915_GEM_DOMAIN_INSTRUCTION
, I915_GEM_DOMAIN_INSTRUCTION
,
211 PIPE_CONTROL_GLOBAL_GTT_WRITE
|
212 ((brw
->query
.index
* 2) * sizeof(uint64_t)));
217 foreach(query
, &brw
->query
.active_head
) {
218 if (query
->bo
!= brw
->query
.bo
) {
219 if (query
->bo
!= NULL
)
220 brw_queryobj_get_results(query
);
221 dri_bo_reference(brw
->query
.bo
);
222 query
->bo
= brw
->query
.bo
;
223 query
->first_index
= brw
->query
.index
;
225 query
->last_index
= brw
->query
.index
;
227 brw
->query
.active
= GL_TRUE
;
230 /** Called at batchbuffer flush to get an ending PS_DEPTH_COUNT */
232 brw_emit_query_end(struct brw_context
*brw
)
234 struct intel_context
*intel
= &brw
->intel
;
236 if (!brw
->query
.active
)
239 BEGIN_BATCH(4, IGNORE_CLIPRECTS
);
240 OUT_BATCH(_3DSTATE_PIPE_CONTROL
|
241 PIPE_CONTROL_DEPTH_STALL
|
242 PIPE_CONTROL_WRITE_DEPTH_COUNT
);
243 OUT_RELOC(brw
->query
.bo
,
244 I915_GEM_DOMAIN_INSTRUCTION
, I915_GEM_DOMAIN_INSTRUCTION
,
245 PIPE_CONTROL_GLOBAL_GTT_WRITE
|
246 ((brw
->query
.index
* 2 + 1) * sizeof(uint64_t)));
251 brw
->query
.active
= GL_FALSE
;
255 void brw_init_queryobj_functions(struct dd_function_table
*functions
)
257 functions
->NewQueryObject
= brw_new_query_object
;
258 functions
->DeleteQuery
= brw_delete_query
;
259 functions
->BeginQuery
= brw_begin_query
;
260 functions
->EndQuery
= brw_end_query
;
261 functions
->CheckQuery
= brw_check_query
;
262 functions
->WaitQuery
= brw_wait_query
;