i965: Add support for xfb overflow query on conditional render.
[mesa.git] / src / mesa / drivers / dri / i965 / brw_conditional_render.c
1 /*
2 * Copyright © 2014 Intel Corporation
3 *
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:
10 *
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
13 * Software.
14 *
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
21 * IN THE SOFTWARE.
22 *
23 * Authors:
24 * Neil Roberts <neil@linux.intel.com>
25 */
26
27 /** @file brw_conditional_render.c
28 *
29 * Support for conditional rendering based on query objects
30 * (GL_NV_conditional_render, GL_ARB_conditional_render_inverted) on Gen7+.
31 */
32
33 #include "main/imports.h"
34 #include "main/condrender.h"
35
36 #include "brw_context.h"
37 #include "brw_defines.h"
38 #include "intel_batchbuffer.h"
39
40 static void
41 set_predicate_enable(struct brw_context *brw,
42 bool value)
43 {
44 if (value)
45 brw->predicate.state = BRW_PREDICATE_STATE_RENDER;
46 else
47 brw->predicate.state = BRW_PREDICATE_STATE_DONT_RENDER;
48 }
49
50 static void
51 set_predicate_for_overflow_query(struct brw_context *brw,
52 struct brw_query_object *query,
53 int stream_start, int count)
54 {
55 hsw_overflow_result_to_gpr0(brw, query, count);
56 brw_load_register_reg64(brw, HSW_CS_GPR(0), MI_PREDICATE_SRC0);
57 brw_load_register_imm64(brw, MI_PREDICATE_SRC1, 0ull);
58 }
59
60 static void
61 set_predicate_for_occlusion_query(struct brw_context *brw,
62 struct brw_query_object *query)
63 {
64 brw_load_register_mem64(brw,
65 MI_PREDICATE_SRC0,
66 query->bo,
67 I915_GEM_DOMAIN_INSTRUCTION,
68 0, /* write domain */
69 0 /* offset */);
70 brw_load_register_mem64(brw,
71 MI_PREDICATE_SRC1,
72 query->bo,
73 I915_GEM_DOMAIN_INSTRUCTION,
74 0, /* write domain */
75 8 /* offset */);
76 }
77
78 static void
79 set_predicate_for_result(struct brw_context *brw,
80 struct brw_query_object *query,
81 bool inverted)
82 {
83
84 int load_op;
85
86 assert(query->bo != NULL);
87
88 /* Needed to ensure the memory is coherent for the MI_LOAD_REGISTER_MEM
89 * command when loading the values into the predicate source registers for
90 * conditional rendering.
91 */
92 brw_emit_pipe_control_flush(brw, PIPE_CONTROL_FLUSH_ENABLE);
93
94 switch (query->Base.Target) {
95 case GL_TRANSFORM_FEEDBACK_STREAM_OVERFLOW_ARB:
96 set_predicate_for_overflow_query(brw, query, 0, 1);
97 break;
98 case GL_TRANSFORM_FEEDBACK_OVERFLOW_ARB:
99 set_predicate_for_overflow_query(brw, query, 0, MAX_VERTEX_STREAMS);
100 break;
101 default:
102 set_predicate_for_occlusion_query(brw, query);
103 }
104
105 if (inverted)
106 load_op = MI_PREDICATE_LOADOP_LOAD;
107 else
108 load_op = MI_PREDICATE_LOADOP_LOADINV;
109
110 BEGIN_BATCH(1);
111 OUT_BATCH(GEN7_MI_PREDICATE |
112 load_op |
113 MI_PREDICATE_COMBINEOP_SET |
114 MI_PREDICATE_COMPAREOP_SRCS_EQUAL);
115 ADVANCE_BATCH();
116
117 brw->predicate.state = BRW_PREDICATE_STATE_USE_BIT;
118 }
119
120 static void
121 brw_begin_conditional_render(struct gl_context *ctx,
122 struct gl_query_object *q,
123 GLenum mode)
124 {
125 struct brw_context *brw = brw_context(ctx);
126 struct brw_query_object *query = (struct brw_query_object *) q;
127 bool inverted;
128
129 if (!brw->predicate.supported)
130 return;
131
132 if ((query->Base.Target == GL_TRANSFORM_FEEDBACK_OVERFLOW_ARB ||
133 query->Base.Target == GL_TRANSFORM_FEEDBACK_STREAM_OVERFLOW_ARB) &&
134 !can_do_mi_math_and_lrr(brw->screen))
135 return;
136
137 switch (mode) {
138 case GL_QUERY_WAIT:
139 case GL_QUERY_NO_WAIT:
140 case GL_QUERY_BY_REGION_WAIT:
141 case GL_QUERY_BY_REGION_NO_WAIT:
142 inverted = false;
143 break;
144 case GL_QUERY_WAIT_INVERTED:
145 case GL_QUERY_NO_WAIT_INVERTED:
146 case GL_QUERY_BY_REGION_WAIT_INVERTED:
147 case GL_QUERY_BY_REGION_NO_WAIT_INVERTED:
148 inverted = true;
149 break;
150 default:
151 unreachable("Unexpected conditional render mode");
152 }
153
154 /* If there are already samples from a BLT operation or if the query object
155 * is ready then we can avoid looking at the values in the buffer and just
156 * decide whether to draw using the CPU without stalling.
157 */
158 if (query->Base.Result || query->Base.Ready)
159 set_predicate_enable(brw, (query->Base.Result != 0) ^ inverted);
160 else
161 set_predicate_for_result(brw, query, inverted);
162 }
163
164 static void
165 brw_end_conditional_render(struct gl_context *ctx,
166 struct gl_query_object *q)
167 {
168 struct brw_context *brw = brw_context(ctx);
169
170 /* When there is no longer a conditional render in progress it should
171 * always render.
172 */
173 brw->predicate.state = BRW_PREDICATE_STATE_RENDER;
174 }
175
176 void
177 brw_init_conditional_render_functions(struct dd_function_table *functions)
178 {
179 functions->BeginConditionalRender = brw_begin_conditional_render;
180 functions->EndConditionalRender = brw_end_conditional_render;
181 }
182
183 bool
184 brw_check_conditional_render(struct brw_context *brw)
185 {
186 const struct gl_query_object *query = brw->ctx.Query.CondRenderQuery;
187
188 const bool query_is_xfb = query &&
189 (query->Target == GL_TRANSFORM_FEEDBACK_OVERFLOW_ARB ||
190 query->Target == GL_TRANSFORM_FEEDBACK_STREAM_OVERFLOW_ARB);
191
192 if (brw->predicate.supported &&
193 (can_do_mi_math_and_lrr(brw->screen) || !query_is_xfb)) {
194 /* In some cases it is possible to determine that the primitives should
195 * be skipped without needing the predicate enable bit and still without
196 * stalling.
197 */
198 return brw->predicate.state != BRW_PREDICATE_STATE_DONT_RENDER;
199 } else if (query) {
200 perf_debug("Conditional rendering is implemented in software and may "
201 "stall.\n");
202 return _mesa_check_conditional_render(&brw->ctx);
203 } else {
204 return true;
205 }
206 }