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
25 #include "intel_batchbuffer.h"
26 #include "intel_fbo.h"
28 #include "blorp_priv.h"
29 #include "brw_compiler.h"
31 #include "brw_state.h"
34 blorp_init(struct blorp_context
*blorp
, void *driver_ctx
,
35 struct isl_device
*isl_dev
)
37 blorp
->driver_ctx
= driver_ctx
;
38 blorp
->isl_dev
= isl_dev
;
42 blorp_finish(struct blorp_context
*blorp
)
44 blorp
->driver_ctx
= NULL
;
48 blorp_batch_init(struct blorp_context
*blorp
,
49 struct blorp_batch
*batch
, void *driver_batch
)
52 batch
->driver_batch
= driver_batch
;
56 blorp_batch_finish(struct blorp_batch
*batch
)
62 brw_blorp_surface_info_init(struct blorp_context
*blorp
,
63 struct brw_blorp_surface_info
*info
,
64 const struct brw_blorp_surf
*surf
,
65 unsigned int level
, unsigned int layer
,
66 enum isl_format format
, bool is_render_target
)
68 /* Layer is a physical layer, so if this is a 2D multisample array texture
69 * using INTEL_MSAA_LAYOUT_UMS or INTEL_MSAA_LAYOUT_CMS, then it had better
70 * be a multiple of num_samples.
72 unsigned layer_multiplier
= 1;
73 if (surf
->surf
->msaa_layout
== ISL_MSAA_LAYOUT_ARRAY
) {
74 assert(layer
% surf
->surf
->samples
== 0);
75 layer_multiplier
= surf
->surf
->samples
;
78 if (format
== ISL_FORMAT_UNSUPPORTED
)
79 format
= surf
->surf
->format
;
81 if (format
== ISL_FORMAT_R24_UNORM_X8_TYPELESS
) {
82 /* Unfortunately, ISL_FORMAT_R24_UNORM_X8_TYPELESS it isn't supported as
83 * a render target, which would prevent us from blitting to 24-bit
84 * depth. The miptree consists of 32 bits per pixel, arranged as 24-bit
85 * depth values interleaved with 8 "don't care" bits. Since depth
86 * values don't require any blending, it doesn't matter how we interpret
87 * the bit pattern as long as we copy the right amount of data, so just
88 * map it as 8-bit BGRA.
90 format
= ISL_FORMAT_B8G8R8A8_UNORM
;
91 } else if (surf
->surf
->usage
& ISL_SURF_USAGE_STENCIL_BIT
) {
92 assert(surf
->surf
->format
== ISL_FORMAT_R8_UINT
);
93 /* Prior to Broadwell, we can't render to R8_UINT */
94 if (blorp
->isl_dev
->info
->gen
< 8)
95 format
= ISL_FORMAT_R8_UNORM
;
98 info
->surf
= *surf
->surf
;
99 info
->addr
= surf
->addr
;
101 info
->aux_usage
= surf
->aux_usage
;
102 if (info
->aux_usage
!= ISL_AUX_USAGE_NONE
) {
103 info
->aux_surf
= *surf
->aux_surf
;
104 info
->aux_addr
= surf
->aux_addr
;
107 info
->clear_color
= surf
->clear_color
;
109 info
->view
= (struct isl_view
) {
110 .usage
= is_render_target
? ISL_SURF_USAGE_RENDER_TARGET_BIT
:
111 ISL_SURF_USAGE_TEXTURE_BIT
,
116 ISL_CHANNEL_SELECT_RED
,
117 ISL_CHANNEL_SELECT_GREEN
,
118 ISL_CHANNEL_SELECT_BLUE
,
119 ISL_CHANNEL_SELECT_ALPHA
,
123 if (!is_render_target
&&
124 (info
->surf
.dim
== ISL_SURF_DIM_3D
||
125 info
->surf
.msaa_layout
== ISL_MSAA_LAYOUT_ARRAY
)) {
126 /* 3-D textures don't support base_array layer and neither do 2-D
127 * multisampled textures on IVB so we need to pass it through the
128 * sampler in those cases. These are also two cases where we are
129 * guaranteed that we won't be doing any funny surface hacks.
131 info
->view
.base_array_layer
= 0;
132 info
->view
.array_len
= MAX2(info
->surf
.logical_level0_px
.depth
,
133 info
->surf
.logical_level0_px
.array_len
);
134 info
->z_offset
= layer
/ layer_multiplier
;
136 info
->view
.base_array_layer
= layer
/ layer_multiplier
;
137 info
->view
.array_len
= 1;
144 brw_blorp_params_init(struct brw_blorp_params
*params
)
146 memset(params
, 0, sizeof(*params
));
147 params
->hiz_op
= GEN6_HIZ_OP_NONE
;
148 params
->fast_clear_op
= 0;
149 params
->num_draw_buffers
= 1;
150 params
->num_layers
= 1;
154 brw_blorp_init_wm_prog_key(struct brw_wm_prog_key
*wm_key
)
156 memset(wm_key
, 0, sizeof(*wm_key
));
157 wm_key
->nr_color_regions
= 1;
158 for (int i
= 0; i
< MAX_SAMPLERS
; i
++)
159 wm_key
->tex
.swizzles
[i
] = SWIZZLE_XYZW
;
163 nir_uniform_type_size(const struct glsl_type
*type
)
165 /* Only very basic types are allowed */
166 assert(glsl_type_is_vector_or_scalar(type
));
167 assert(glsl_get_bit_size(type
) == 32);
169 return glsl_get_vector_elements(type
) * 4;
173 brw_blorp_compile_nir_shader(struct blorp_context
*blorp
, struct nir_shader
*nir
,
174 const struct brw_wm_prog_key
*wm_key
,
176 struct brw_blorp_prog_data
*prog_data
,
177 unsigned *program_size
)
179 const struct brw_compiler
*compiler
= blorp
->compiler
;
181 void *mem_ctx
= ralloc_context(NULL
);
183 /* Calling brw_preprocess_nir and friends is destructive and, if cloning is
184 * enabled, may end up completely replacing the nir_shader. Therefore, we
185 * own it and might as well put it in our context for easy cleanup.
187 ralloc_steal(mem_ctx
, nir
);
189 compiler
->glsl_compiler_options
[MESA_SHADER_FRAGMENT
].NirOptions
;
191 struct brw_wm_prog_data wm_prog_data
;
192 memset(&wm_prog_data
, 0, sizeof(wm_prog_data
));
194 wm_prog_data
.base
.nr_params
= 0;
195 wm_prog_data
.base
.param
= NULL
;
197 /* BLORP always just uses the first two binding table entries */
198 wm_prog_data
.binding_table
.render_target_start
= BLORP_RENDERBUFFER_BT_INDEX
;
199 wm_prog_data
.base
.binding_table
.texture_start
= BLORP_TEXTURE_BT_INDEX
;
201 nir
= brw_preprocess_nir(compiler
, nir
);
202 nir_remove_dead_variables(nir
, nir_var_shader_in
);
203 nir_shader_gather_info(nir
, nir_shader_get_entrypoint(nir
));
205 /* Uniforms are required to be lowered before going into compile_fs. For
206 * BLORP, we'll assume that whoever builds the shader sets the location
207 * they want so we just need to lower them and figure out how many we have
210 nir
->num_uniforms
= 0;
211 nir_foreach_variable(var
, &nir
->uniforms
) {
212 var
->data
.driver_location
= var
->data
.location
;
213 unsigned end
= var
->data
.location
+ nir_uniform_type_size(var
->type
);
214 nir
->num_uniforms
= MAX2(nir
->num_uniforms
, end
);
216 nir_lower_io(nir
, nir_var_uniform
, nir_uniform_type_size
);
218 const unsigned *program
=
219 brw_compile_fs(compiler
, blorp
->driver_ctx
, mem_ctx
,
220 wm_key
, &wm_prog_data
, nir
,
221 NULL
, -1, -1, false, use_repclear
, program_size
, NULL
);
223 /* Copy the relavent bits of wm_prog_data over into the blorp prog data */
224 prog_data
->dispatch_8
= wm_prog_data
.dispatch_8
;
225 prog_data
->dispatch_16
= wm_prog_data
.dispatch_16
;
226 prog_data
->first_curbe_grf_0
= wm_prog_data
.base
.dispatch_grf_start_reg
;
227 prog_data
->first_curbe_grf_2
= wm_prog_data
.dispatch_grf_start_reg_2
;
228 prog_data
->ksp_offset_2
= wm_prog_data
.prog_offset_2
;
229 prog_data
->persample_msaa_dispatch
= wm_prog_data
.persample_dispatch
;
230 prog_data
->flat_inputs
= wm_prog_data
.flat_inputs
;
231 prog_data
->num_varying_inputs
= wm_prog_data
.num_varying_inputs
;
232 prog_data
->inputs_read
= nir
->info
.inputs_read
;
234 assert(wm_prog_data
.base
.nr_params
== 0);
240 blorp_gen6_hiz_op(struct blorp_batch
*batch
,
241 struct brw_blorp_surf
*surf
, unsigned level
, unsigned layer
,
244 struct brw_blorp_params params
;
245 brw_blorp_params_init(¶ms
);
249 brw_blorp_surface_info_init(batch
->blorp
, ¶ms
.depth
, surf
, level
, layer
,
250 surf
->surf
->format
, true);
252 /* Align the rectangle primitive to 8x4 pixels.
254 * During fast depth clears, the emitted rectangle primitive must be
255 * aligned to 8x4 pixels. From the Ivybridge PRM, Vol 2 Part 1 Section
256 * 11.5.3.1 Depth Buffer Clear (and the matching section in the Sandybridge
258 * If Number of Multisamples is NUMSAMPLES_1, the rectangle must be
259 * aligned to an 8x4 pixel block relative to the upper left corner
260 * of the depth buffer [...]
262 * For hiz resolves, the rectangle must also be 8x4 aligned. Item
263 * WaHizAmbiguate8x4Aligned from the Haswell workarounds page and the
264 * Ivybridge simulator require the alignment.
266 * To be safe, let's just align the rect for all hiz operations and all
267 * hardware generations.
269 * However, for some miptree slices of a Z24 texture, emitting an 8x4
270 * aligned rectangle that covers the slice may clobber adjacent slices if
271 * we strictly adhered to the texture alignments specified in the PRM. The
272 * Ivybridge PRM, Section "Alignment Unit Size", states that
273 * SURFACE_STATE.Surface_Horizontal_Alignment should be 4 for Z24 surfaces,
274 * not 8. But commit 1f112cc increased the alignment from 4 to 8, which
275 * prevents the clobbering.
277 params
.x1
= minify(params
.depth
.surf
.logical_level0_px
.width
,
278 params
.depth
.view
.base_level
);
279 params
.y1
= minify(params
.depth
.surf
.logical_level0_px
.height
,
280 params
.depth
.view
.base_level
);
281 params
.x1
= ALIGN(params
.x1
, 8);
282 params
.y1
= ALIGN(params
.y1
, 4);
284 if (params
.depth
.view
.base_level
== 0) {
285 /* TODO: What about MSAA? */
286 params
.depth
.surf
.logical_level0_px
.width
= params
.x1
;
287 params
.depth
.surf
.logical_level0_px
.height
= params
.y1
;
290 params
.dst
.surf
.samples
= params
.depth
.surf
.samples
;
291 params
.dst
.surf
.logical_level0_px
= params
.depth
.surf
.logical_level0_px
;
293 switch (surf
->surf
->format
) {
294 case ISL_FORMAT_R16_UNORM
:
295 params
.depth_format
= BRW_DEPTHFORMAT_D16_UNORM
;
297 case ISL_FORMAT_R32_FLOAT
:
298 params
.depth_format
= BRW_DEPTHFORMAT_D32_FLOAT
;
300 case ISL_FORMAT_R24_UNORM_X8_TYPELESS
:
301 params
.depth_format
= BRW_DEPTHFORMAT_D24_UNORM_X8_UINT
;
304 unreachable("not reached");
307 batch
->blorp
->exec(batch
, ¶ms
);