2 * Copyright © 2012 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 #include "main/blend.h"
25 #include "main/mtypes.h"
26 #include "main/samplerobj.h"
27 #include "main/texformat.h"
28 #include "program/prog_parameter.h"
30 #include "intel_mipmap_tree.h"
31 #include "intel_batchbuffer.h"
32 #include "intel_tex.h"
33 #include "intel_fbo.h"
34 #include "intel_buffer_objects.h"
36 #include "brw_context.h"
37 #include "brw_state.h"
38 #include "brw_defines.h"
42 * Convert an swizzle enumeration (i.e. SWIZZLE_X) to one of the Gen7.5+
43 * "Shader Channel Select" enumerations (i.e. HSW_SCS_RED). The mappings are
45 * SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W, SWIZZLE_ZERO, SWIZZLE_ONE
48 * SCS_RED, SCS_GREEN, SCS_BLUE, SCS_ALPHA, SCS_ZERO, SCS_ONE
50 * which is simply adding 4 then modding by 8 (or anding with 7).
53 swizzle_to_scs(unsigned swizzle
)
55 return (swizzle
+ 4) & 7;
59 surface_tiling_mode(uint32_t tiling
)
63 return GEN8_SURFACE_TILING_X
;
65 return GEN8_SURFACE_TILING_Y
;
67 return GEN8_SURFACE_TILING_NONE
;
72 vertical_alignment(struct intel_mipmap_tree
*mt
)
74 switch (mt
->align_h
) {
76 return GEN8_SURFACE_VALIGN_4
;
78 return GEN8_SURFACE_VALIGN_8
;
80 return GEN8_SURFACE_VALIGN_16
;
82 unreachable("Unsupported vertical surface alignment.");
87 horizontal_alignment(struct intel_mipmap_tree
*mt
)
89 switch (mt
->align_w
) {
91 return GEN8_SURFACE_HALIGN_4
;
93 return GEN8_SURFACE_HALIGN_8
;
95 return GEN8_SURFACE_HALIGN_16
;
97 unreachable("Unsupported horizontal surface alignment.");
102 allocate_surface_state(struct brw_context
*brw
, uint32_t *out_offset
)
104 int dwords
= brw
->gen
>= 9 ? 16 : 13;
105 uint32_t *surf
= brw_state_batch(brw
, AUB_TRACE_SURFACE_STATE
,
106 dwords
* 4, 64, out_offset
);
107 memset(surf
, 0, dwords
* 4);
112 gen8_emit_buffer_surface_state(struct brw_context
*brw
,
113 uint32_t *out_offset
,
115 unsigned buffer_offset
,
116 unsigned surface_format
,
117 unsigned buffer_size
,
121 const unsigned mocs
= brw
->gen
>= 9 ? SKL_MOCS_WB
: BDW_MOCS_WB
;
122 uint32_t *surf
= allocate_surface_state(brw
, out_offset
);
124 surf
[0] = BRW_SURFACE_BUFFER
<< BRW_SURFACE_TYPE_SHIFT
|
125 surface_format
<< BRW_SURFACE_FORMAT_SHIFT
|
126 BRW_SURFACE_RC_READ_WRITE
;
127 surf
[1] = SET_FIELD(mocs
, GEN8_SURFACE_MOCS
);
129 surf
[2] = SET_FIELD((buffer_size
- 1) & 0x7f, GEN7_SURFACE_WIDTH
) |
130 SET_FIELD(((buffer_size
- 1) >> 7) & 0x3fff, GEN7_SURFACE_HEIGHT
);
131 surf
[3] = SET_FIELD(((buffer_size
- 1) >> 21) & 0x3f, BRW_SURFACE_DEPTH
) |
133 surf
[7] = SET_FIELD(HSW_SCS_RED
, GEN7_SURFACE_SCS_R
) |
134 SET_FIELD(HSW_SCS_GREEN
, GEN7_SURFACE_SCS_G
) |
135 SET_FIELD(HSW_SCS_BLUE
, GEN7_SURFACE_SCS_B
) |
136 SET_FIELD(HSW_SCS_ALPHA
, GEN7_SURFACE_SCS_A
);
138 *((uint64_t *) &surf
[8]) = (bo
? bo
->offset64
: 0) + buffer_offset
;
140 /* Emit relocation to surface contents. */
142 drm_intel_bo_emit_reloc(brw
->batch
.bo
, *out_offset
+ 8 * 4,
143 bo
, buffer_offset
, I915_GEM_DOMAIN_SAMPLER
,
144 rw
? I915_GEM_DOMAIN_SAMPLER
: 0);
149 gen8_update_texture_surface(struct gl_context
*ctx
,
151 uint32_t *surf_offset
,
154 struct brw_context
*brw
= brw_context(ctx
);
155 struct gl_texture_object
*tObj
= ctx
->Texture
.Unit
[unit
]._Current
;
156 struct intel_texture_object
*intelObj
= intel_texture_object(tObj
);
157 struct intel_mipmap_tree
*mt
= intelObj
->mt
;
158 struct gl_texture_image
*firstImage
= tObj
->Image
[0][tObj
->BaseLevel
];
159 struct gl_sampler_object
*sampler
= _mesa_get_samplerobj(ctx
, unit
);
160 struct intel_mipmap_tree
*aux_mt
= NULL
;
161 uint32_t aux_mode
= 0;
162 mesa_format format
= intelObj
->_Format
;
163 uint32_t mocs_wb
= brw
->gen
>= 9 ? SKL_MOCS_WB
: BDW_MOCS_WB
;
165 if (tObj
->Target
== GL_TEXTURE_BUFFER
) {
166 brw_update_buffer_texture_surface(ctx
, unit
, surf_offset
);
170 if (tObj
->StencilSampling
&& firstImage
->_BaseFormat
== GL_DEPTH_STENCIL
) {
172 format
= MESA_FORMAT_S_UINT8
;
175 unsigned tiling_mode
, pitch
;
176 if (format
== MESA_FORMAT_S_UINT8
) {
177 tiling_mode
= GEN8_SURFACE_TILING_W
;
178 pitch
= 2 * mt
->pitch
;
180 tiling_mode
= surface_tiling_mode(mt
->tiling
);
186 aux_mode
= GEN8_SURFACE_AUX_MODE_MCS
;
189 /* If this is a view with restricted NumLayers, then our effective depth
190 * is not just the miptree depth.
192 uint32_t effective_depth
=
193 (tObj
->Immutable
&& tObj
->Target
!= GL_TEXTURE_3D
) ? tObj
->NumLayers
194 : mt
->logical_depth0
;
196 uint32_t tex_format
= translate_tex_format(brw
, format
, sampler
->sRGBDecode
);
198 uint32_t *surf
= allocate_surface_state(brw
, surf_offset
);
200 surf
[0] = translate_tex_target(tObj
->Target
) << BRW_SURFACE_TYPE_SHIFT
|
201 tex_format
<< BRW_SURFACE_FORMAT_SHIFT
|
202 vertical_alignment(mt
) |
203 horizontal_alignment(mt
) |
206 if (tObj
->Target
== GL_TEXTURE_CUBE_MAP
||
207 tObj
->Target
== GL_TEXTURE_CUBE_MAP_ARRAY
) {
208 surf
[0] |= BRW_SURFACE_CUBEFACE_ENABLES
;
211 if (mt
->logical_depth0
> 1 && tObj
->Target
!= GL_TEXTURE_3D
)
212 surf
[0] |= GEN8_SURFACE_IS_ARRAY
;
214 surf
[1] = SET_FIELD(mocs_wb
, GEN8_SURFACE_MOCS
) | mt
->qpitch
>> 2;
216 surf
[2] = SET_FIELD(mt
->logical_width0
- 1, GEN7_SURFACE_WIDTH
) |
217 SET_FIELD(mt
->logical_height0
- 1, GEN7_SURFACE_HEIGHT
);
219 surf
[3] = SET_FIELD(effective_depth
- 1, BRW_SURFACE_DEPTH
) | (pitch
- 1);
221 surf
[4] = gen7_surface_msaa_bits(mt
->num_samples
, mt
->msaa_layout
) |
222 SET_FIELD(tObj
->MinLayer
, GEN7_SURFACE_MIN_ARRAY_ELEMENT
) |
223 SET_FIELD(effective_depth
- 1,
224 GEN7_SURFACE_RENDER_TARGET_VIEW_EXTENT
);
226 surf
[5] = SET_FIELD(tObj
->MinLevel
+ tObj
->BaseLevel
- mt
->first_level
,
227 GEN7_SURFACE_MIN_LOD
) |
228 (intelObj
->_MaxLevel
- tObj
->BaseLevel
); /* mip count */
231 surf
[6] = SET_FIELD(mt
->qpitch
/ 4, GEN8_SURFACE_AUX_QPITCH
) |
232 SET_FIELD((aux_mt
->pitch
/ 128) - 1, GEN8_SURFACE_AUX_PITCH
) |
238 /* Handling GL_ALPHA as a surface format override breaks 1.30+ style
239 * texturing functions that return a float, as our code generation always
240 * selects the .x channel (which would always be 0).
242 const bool alpha_depth
= tObj
->DepthMode
== GL_ALPHA
&&
243 (firstImage
->_BaseFormat
== GL_DEPTH_COMPONENT
||
244 firstImage
->_BaseFormat
== GL_DEPTH_STENCIL
);
246 surf
[7] = mt
->fast_clear_color_value
;
249 unlikely(alpha_depth
) ? SWIZZLE_XYZW
: brw_get_texture_swizzle(ctx
, tObj
);
251 SET_FIELD(swizzle_to_scs(GET_SWZ(swizzle
, 0)), GEN7_SURFACE_SCS_R
) |
252 SET_FIELD(swizzle_to_scs(GET_SWZ(swizzle
, 1)), GEN7_SURFACE_SCS_G
) |
253 SET_FIELD(swizzle_to_scs(GET_SWZ(swizzle
, 2)), GEN7_SURFACE_SCS_B
) |
254 SET_FIELD(swizzle_to_scs(GET_SWZ(swizzle
, 3)), GEN7_SURFACE_SCS_A
);
256 *((uint64_t *) &surf
[8]) = mt
->bo
->offset64
+ mt
->offset
; /* reloc */
259 *((uint64_t *) &surf
[10]) = aux_mt
->bo
->offset64
;
260 drm_intel_bo_emit_reloc(brw
->batch
.bo
, *surf_offset
+ 10 * 4,
262 I915_GEM_DOMAIN_SAMPLER
, 0);
269 /* Emit relocation to surface contents */
270 drm_intel_bo_emit_reloc(brw
->batch
.bo
,
271 *surf_offset
+ 8 * 4,
274 I915_GEM_DOMAIN_SAMPLER
, 0);
278 * Creates a null surface.
280 * This is used when the shader doesn't write to any color output. An FB
281 * write to target 0 will still be emitted, because that's how the thread is
282 * terminated (and computed depth is returned), so we need to have the
283 * hardware discard the target 0 color output..
286 gen8_emit_null_surface_state(struct brw_context
*brw
,
290 uint32_t *out_offset
)
292 uint32_t *surf
= allocate_surface_state(brw
, out_offset
);
294 surf
[0] = BRW_SURFACE_NULL
<< BRW_SURFACE_TYPE_SHIFT
|
295 BRW_SURFACEFORMAT_B8G8R8A8_UNORM
<< BRW_SURFACE_FORMAT_SHIFT
|
296 GEN8_SURFACE_TILING_Y
;
297 surf
[2] = SET_FIELD(width
- 1, GEN7_SURFACE_WIDTH
) |
298 SET_FIELD(height
- 1, GEN7_SURFACE_HEIGHT
);
302 * Sets up a surface state structure to point at the given region.
303 * While it is only used for the front/back buffer currently, it should be
304 * usable for further buffers when doing ARB_draw_buffer support.
307 gen8_update_renderbuffer_surface(struct brw_context
*brw
,
308 struct gl_renderbuffer
*rb
,
312 struct gl_context
*ctx
= &brw
->ctx
;
313 struct intel_renderbuffer
*irb
= intel_renderbuffer(rb
);
314 struct intel_mipmap_tree
*mt
= irb
->mt
;
315 struct intel_mipmap_tree
*aux_mt
= NULL
;
316 uint32_t aux_mode
= 0;
317 unsigned width
= mt
->logical_width0
;
318 unsigned height
= mt
->logical_height0
;
319 unsigned pitch
= mt
->pitch
;
320 uint32_t tiling
= mt
->tiling
;
323 bool is_array
= false;
324 int depth
= MAX2(irb
->layer_count
, 1);
325 const int min_array_element
= (mt
->format
== MESA_FORMAT_S_UINT8
) ?
326 irb
->mt_layer
: (irb
->mt_layer
/ MAX2(mt
->num_samples
, 1));
328 rb
->TexImage
? rb
->TexImage
->TexObject
->Target
: GL_TEXTURE_2D
;
329 uint32_t surf_index
=
330 brw
->wm
.prog_data
->binding_table
.render_target_start
+ unit
;
331 /* FINISHME: Use PTE MOCS on Skylake. */
332 uint32_t mocs
= brw
->gen
>= 9 ? SKL_MOCS_WT
: BDW_MOCS_PTE
;
334 intel_miptree_used_for_rendering(mt
);
337 case GL_TEXTURE_CUBE_MAP_ARRAY
:
338 case GL_TEXTURE_CUBE_MAP
:
339 surf_type
= BRW_SURFACE_2D
;
344 depth
= MAX2(irb
->mt
->logical_depth0
, 1);
347 surf_type
= translate_tex_target(gl_target
);
348 is_array
= _mesa_tex_target_is_array(gl_target
);
353 /* Render targets can't use IMS layout. Stencil in turn gets configured as
354 * single sampled and indexed manually by the program.
356 if (mt
->format
== MESA_FORMAT_S_UINT8
) {
357 brw_configure_w_tiled(mt
, true, &width
, &height
, &pitch
,
360 assert(mt
->msaa_layout
!= INTEL_MSAA_LAYOUT_IMS
);
361 assert(brw_render_target_supported(brw
, rb
));
362 mesa_format rb_format
= _mesa_get_render_format(ctx
,
363 intel_rb_format(irb
));
364 format
= brw
->render_target_format
[rb_format
];
365 if (unlikely(!brw
->format_supported_as_render_target
[rb_format
]))
366 _mesa_problem(ctx
, "%s: renderbuffer format %s unsupported\n",
367 __FUNCTION__
, _mesa_get_format_name(rb_format
));
372 aux_mode
= GEN8_SURFACE_AUX_MODE_MCS
;
376 allocate_surface_state(brw
, &brw
->wm
.base
.surf_offset
[surf_index
]);
378 surf
[0] = (surf_type
<< BRW_SURFACE_TYPE_SHIFT
) |
379 (is_array
? GEN7_SURFACE_IS_ARRAY
: 0) |
380 (format
<< BRW_SURFACE_FORMAT_SHIFT
) |
381 vertical_alignment(mt
) |
382 horizontal_alignment(mt
) |
383 surface_tiling_mode(tiling
);
385 surf
[1] = SET_FIELD(mocs
, GEN8_SURFACE_MOCS
) | mt
->qpitch
>> 2;
387 surf
[2] = SET_FIELD(width
- 1, GEN7_SURFACE_WIDTH
) |
388 SET_FIELD(height
- 1, GEN7_SURFACE_HEIGHT
);
390 surf
[3] = (depth
- 1) << BRW_SURFACE_DEPTH_SHIFT
|
391 (pitch
- 1); /* Surface Pitch */
393 surf
[4] = min_array_element
<< GEN7_SURFACE_MIN_ARRAY_ELEMENT_SHIFT
|
394 (depth
- 1) << GEN7_SURFACE_RENDER_TARGET_VIEW_EXTENT_SHIFT
;
396 if (mt
->format
!= MESA_FORMAT_S_UINT8
)
397 surf
[4] |= gen7_surface_msaa_bits(mt
->num_samples
, mt
->msaa_layout
);
399 surf
[5] = irb
->mt_level
- irb
->mt
->first_level
;
402 surf
[6] = SET_FIELD(mt
->qpitch
/ 4, GEN8_SURFACE_AUX_QPITCH
) |
403 SET_FIELD((aux_mt
->pitch
/ 128) - 1, GEN8_SURFACE_AUX_PITCH
) |
409 surf
[7] = mt
->fast_clear_color_value
|
410 SET_FIELD(HSW_SCS_RED
, GEN7_SURFACE_SCS_R
) |
411 SET_FIELD(HSW_SCS_GREEN
, GEN7_SURFACE_SCS_G
) |
412 SET_FIELD(HSW_SCS_BLUE
, GEN7_SURFACE_SCS_B
) |
413 SET_FIELD(HSW_SCS_ALPHA
, GEN7_SURFACE_SCS_A
);
415 assert(mt
->offset
% mt
->cpp
== 0);
416 *((uint64_t *) &surf
[8]) = mt
->bo
->offset64
+ mt
->offset
; /* reloc */
419 *((uint64_t *) &surf
[10]) = aux_mt
->bo
->offset64
;
420 drm_intel_bo_emit_reloc(brw
->batch
.bo
,
421 brw
->wm
.base
.surf_offset
[surf_index
] + 10 * 4,
423 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
);
430 drm_intel_bo_emit_reloc(brw
->batch
.bo
,
431 brw
->wm
.base
.surf_offset
[surf_index
] + 8 * 4,
434 I915_GEM_DOMAIN_RENDER
,
435 I915_GEM_DOMAIN_RENDER
);
439 gen8_init_vtable_surface_functions(struct brw_context
*brw
)
441 brw
->vtbl
.update_texture_surface
= gen8_update_texture_surface
;
442 brw
->vtbl
.update_renderbuffer_surface
= gen8_update_renderbuffer_surface
;
443 brw
->vtbl
.emit_null_surface_state
= gen8_emit_null_surface_state
;
444 brw
->vtbl
.emit_buffer_surface_state
= gen8_emit_buffer_surface_state
;