i965: Fix PBO cache coherency issue after _mesa_meta_pbo_GetTexSubImage().
[mesa.git] / src / mesa / drivers / dri / i965 / gen6_depth_state.c
1 /*
2 * Copyright (c) 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
24
25 #include "intel_batchbuffer.h"
26 #include "intel_fbo.h"
27 #include "intel_mipmap_tree.h"
28
29 #include "brw_context.h"
30 #include "brw_state.h"
31 #include "brw_defines.h"
32
33 #include "main/mtypes.h"
34 #include "main/fbobject.h"
35 #include "main/glformats.h"
36
37 void
38 gen6_emit_depth_stencil_hiz(struct brw_context *brw,
39 struct intel_mipmap_tree *depth_mt,
40 uint32_t depth_offset, uint32_t depthbuffer_format,
41 uint32_t depth_surface_type,
42 struct intel_mipmap_tree *stencil_mt,
43 bool hiz, bool separate_stencil,
44 uint32_t width, uint32_t height,
45 uint32_t tile_x, uint32_t tile_y)
46 {
47 struct gl_context *ctx = &brw->ctx;
48 struct gl_framebuffer *fb = ctx->DrawBuffer;
49 uint32_t surftype;
50 unsigned int depth = 1;
51 GLenum gl_target = GL_TEXTURE_2D;
52 unsigned int lod;
53 const struct intel_mipmap_tree *mt = depth_mt ? depth_mt : stencil_mt;
54 const struct intel_renderbuffer *irb = NULL;
55 const struct gl_renderbuffer *rb = NULL;
56
57 /* Enable the hiz bit if we're doing separate stencil, because it and the
58 * separate stencil bit must have the same value. From Section 2.11.5.6.1.1
59 * 3DSTATE_DEPTH_BUFFER, Bit 1.21 "Separate Stencil Enable":
60 * [DevIL]: If this field is enabled, Hierarchical Depth Buffer
61 * Enable must also be enabled.
62 *
63 * [DevGT]: This field must be set to the same value (enabled or
64 * disabled) as Hierarchical Depth Buffer Enable
65 */
66 bool enable_hiz_ss = hiz || separate_stencil;
67
68 intel_emit_depth_stall_flushes(brw);
69
70 irb = intel_get_renderbuffer(fb, BUFFER_DEPTH);
71 if (!irb)
72 irb = intel_get_renderbuffer(fb, BUFFER_STENCIL);
73 rb = (struct gl_renderbuffer*) irb;
74
75 if (rb) {
76 depth = MAX2(rb->Depth, 1);
77 if (rb->TexImage)
78 gl_target = rb->TexImage->TexObject->Target;
79 }
80
81 switch (gl_target) {
82 case GL_TEXTURE_CUBE_MAP_ARRAY:
83 case GL_TEXTURE_CUBE_MAP:
84 /* The PRM claims that we should use BRW_SURFACE_CUBE for this
85 * situation, but experiments show that gl_Layer doesn't work when we do
86 * this. So we use BRW_SURFACE_2D, since for rendering purposes this is
87 * equivalent.
88 */
89 surftype = BRW_SURFACE_2D;
90 depth *= 6;
91 break;
92 default:
93 surftype = translate_tex_target(gl_target);
94 break;
95 }
96
97 const unsigned min_array_element = irb ? irb->mt_layer : 0;
98
99 lod = irb ? irb->mt_level - irb->mt->first_level : 0;
100
101 if (mt) {
102 width = mt->logical_width0;
103 height = mt->logical_height0;
104 }
105
106 BEGIN_BATCH(7);
107 /* 3DSTATE_DEPTH_BUFFER dw0 */
108 OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (7 - 2));
109
110 /* 3DSTATE_DEPTH_BUFFER dw1 */
111 OUT_BATCH((depth_mt ? depth_mt->pitch - 1 : 0) |
112 (depthbuffer_format << 18) |
113 ((enable_hiz_ss ? 1 : 0) << 21) | /* separate stencil enable */
114 ((enable_hiz_ss ? 1 : 0) << 22) | /* hiz enable */
115 (BRW_TILEWALK_YMAJOR << 26) |
116 ((depth_mt ? depth_mt->tiling != I915_TILING_NONE : 1)
117 << 27) |
118 (surftype << 29));
119
120 /* 3DSTATE_DEPTH_BUFFER dw2 */
121 if (depth_mt) {
122 OUT_RELOC(depth_mt->bo,
123 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
124 0);
125 } else {
126 OUT_BATCH(0);
127 }
128
129 /* 3DSTATE_DEPTH_BUFFER dw3 */
130 OUT_BATCH(((width - 1) << 6) |
131 ((height - 1) << 19) |
132 lod << 2);
133
134 /* 3DSTATE_DEPTH_BUFFER dw4 */
135 OUT_BATCH((depth - 1) << 21 |
136 min_array_element << 10 |
137 (depth - 1) << 1);
138
139 /* 3DSTATE_DEPTH_BUFFER dw5 */
140 OUT_BATCH(0);
141 assert(tile_x == 0 && tile_y == 0);
142
143 /* 3DSTATE_DEPTH_BUFFER dw6 */
144 OUT_BATCH(0);
145
146 ADVANCE_BATCH();
147
148 if (hiz || separate_stencil) {
149 /*
150 * In the 3DSTATE_DEPTH_BUFFER batch emitted above, the 'separate
151 * stencil enable' and 'hiz enable' bits were set. Therefore we must
152 * emit 3DSTATE_HIER_DEPTH_BUFFER and 3DSTATE_STENCIL_BUFFER. Even if
153 * there is no stencil buffer, 3DSTATE_STENCIL_BUFFER must be emitted;
154 * failure to do so causes hangs on gen5 and a stall on gen6.
155 */
156
157 /* Emit hiz buffer. */
158 if (hiz) {
159 struct intel_mipmap_tree *hiz_mt = depth_mt->hiz_buf->mt;
160 uint32_t offset = 0;
161
162 if (hiz_mt->array_layout == ALL_SLICES_AT_EACH_LOD) {
163 offset = intel_miptree_get_aligned_offset(
164 hiz_mt,
165 hiz_mt->level[lod].level_x,
166 hiz_mt->level[lod].level_y,
167 false);
168 }
169
170 BEGIN_BATCH(3);
171 OUT_BATCH((_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
172 OUT_BATCH(hiz_mt->pitch - 1);
173 OUT_RELOC(hiz_mt->bo,
174 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
175 offset);
176 ADVANCE_BATCH();
177 } else {
178 BEGIN_BATCH(3);
179 OUT_BATCH((_3DSTATE_HIER_DEPTH_BUFFER << 16) | (3 - 2));
180 OUT_BATCH(0);
181 OUT_BATCH(0);
182 ADVANCE_BATCH();
183 }
184
185 /* Emit stencil buffer. */
186 if (separate_stencil) {
187 uint32_t offset = 0;
188
189 if (stencil_mt->array_layout == ALL_SLICES_AT_EACH_LOD) {
190 if (stencil_mt->format == MESA_FORMAT_S_UINT8) {
191 /* Note: we can't compute the stencil offset using
192 * intel_region_get_aligned_offset(), because stencil_region
193 * claims that the region is untiled even though it's W tiled.
194 */
195 offset =
196 stencil_mt->level[lod].level_y * stencil_mt->pitch +
197 stencil_mt->level[lod].level_x * 64;
198 } else {
199 offset = intel_miptree_get_aligned_offset(
200 stencil_mt,
201 stencil_mt->level[lod].level_x,
202 stencil_mt->level[lod].level_y,
203 false);
204 }
205 }
206
207 BEGIN_BATCH(3);
208 OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
209 /* The stencil buffer has quirky pitch requirements. From Vol 2a,
210 * 11.5.6.2.1 3DSTATE_STENCIL_BUFFER, field "Surface Pitch":
211 * The pitch must be set to 2x the value computed based on width, as
212 * the stencil buffer is stored with two rows interleaved.
213 */
214 OUT_BATCH(2 * stencil_mt->pitch - 1);
215 OUT_RELOC(stencil_mt->bo,
216 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
217 offset);
218 ADVANCE_BATCH();
219 } else {
220 BEGIN_BATCH(3);
221 OUT_BATCH((_3DSTATE_STENCIL_BUFFER << 16) | (3 - 2));
222 OUT_BATCH(0);
223 OUT_BATCH(0);
224 ADVANCE_BATCH();
225 }
226 }
227
228 /*
229 * On Gen >= 6, emit clear params for safety. If using hiz, then clear
230 * params must be emitted.
231 *
232 * From Section 2.11.5.6.4.1 3DSTATE_CLEAR_PARAMS:
233 * 3DSTATE_CLEAR_PARAMS packet must follow the DEPTH_BUFFER_STATE packet
234 * when HiZ is enabled and the DEPTH_BUFFER_STATE changes.
235 */
236 BEGIN_BATCH(2);
237 OUT_BATCH(_3DSTATE_CLEAR_PARAMS << 16 |
238 GEN5_DEPTH_CLEAR_VALID |
239 (2 - 2));
240 OUT_BATCH(depth_mt ? depth_mt->depth_clear_value : 0);
241 ADVANCE_BATCH();
242 }