2 * Copyright © 2015 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
30 #include "anv_private.h"
32 #include "gen8_pack.h"
33 #include "gen9_pack.h"
35 #include "genX_pipeline_util.h"
38 emit_ia_state(struct anv_pipeline
*pipeline
,
39 const VkPipelineInputAssemblyStateCreateInfo
*info
,
40 const struct anv_graphics_pipeline_create_info
*extra
)
42 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_VF_TOPOLOGY
),
43 .PrimitiveTopologyType
= pipeline
->topology
);
47 emit_rs_state(struct anv_pipeline
*pipeline
,
48 const VkPipelineRasterizationStateCreateInfo
*info
,
49 const VkPipelineMultisampleStateCreateInfo
*ms_info
,
50 const struct anv_graphics_pipeline_create_info
*extra
)
55 samples
= ms_info
->rasterizationSamples
;
57 struct GENX(3DSTATE_SF
) sf
= {
58 GENX(3DSTATE_SF_header
),
59 .ViewportTransformEnable
= !(extra
&& extra
->disable_viewport
),
60 .TriangleStripListProvokingVertexSelect
= 0,
61 .LineStripListProvokingVertexSelect
= 0,
62 .TriangleFanProvokingVertexSelect
= 0,
63 .PointWidthSource
= pipeline
->writes_point_size
? Vertex
: State
,
67 /* FINISHME: VkBool32 rasterizerDiscardEnable; */
69 GENX(3DSTATE_SF_pack
)(NULL
, pipeline
->gen8
.sf
, &sf
);
71 struct GENX(3DSTATE_RASTER
) raster
= {
72 GENX(3DSTATE_RASTER_header
),
74 /* For details on 3DSTATE_RASTER multisample state, see the BSpec table
75 * "Multisample Modes State".
77 .DXMultisampleRasterizationEnable
= samples
> 1,
78 .ForcedSampleCount
= FSC_NUMRASTSAMPLES_0
,
79 .ForceMultisampling
= false,
81 .FrontWinding
= vk_to_gen_front_face
[info
->frontFace
],
82 .CullMode
= vk_to_gen_cullmode
[info
->cullMode
],
83 .FrontFaceFillMode
= vk_to_gen_fillmode
[info
->polygonMode
],
84 .BackFaceFillMode
= vk_to_gen_fillmode
[info
->polygonMode
],
85 .ScissorRectangleEnable
= !(extra
&& extra
->disable_scissor
),
87 .ViewportZClipTestEnable
= true,
89 /* GEN9+ splits ViewportZClipTestEnable into near and far enable bits */
90 .ViewportZFarClipTestEnable
= true,
91 .ViewportZNearClipTestEnable
= true,
95 GENX(3DSTATE_RASTER_pack
)(NULL
, pipeline
->gen8
.raster
, &raster
);
99 emit_cb_state(struct anv_pipeline
*pipeline
,
100 const VkPipelineColorBlendStateCreateInfo
*info
,
101 const VkPipelineMultisampleStateCreateInfo
*ms_info
)
103 struct anv_device
*device
= pipeline
->device
;
105 uint32_t num_dwords
= GENX(BLEND_STATE_length
);
106 pipeline
->blend_state
=
107 anv_state_pool_alloc(&device
->dynamic_state_pool
, num_dwords
* 4, 64);
109 struct GENX(BLEND_STATE
) blend_state
= {
110 .AlphaToCoverageEnable
= ms_info
&& ms_info
->alphaToCoverageEnable
,
111 .AlphaToOneEnable
= ms_info
&& ms_info
->alphaToOneEnable
,
114 for (uint32_t i
= 0; i
< info
->attachmentCount
; i
++) {
115 const VkPipelineColorBlendAttachmentState
*a
= &info
->pAttachments
[i
];
117 if (a
->srcColorBlendFactor
!= a
->srcAlphaBlendFactor
||
118 a
->dstColorBlendFactor
!= a
->dstAlphaBlendFactor
||
119 a
->colorBlendOp
!= a
->alphaBlendOp
) {
120 blend_state
.IndependentAlphaBlendEnable
= true;
123 blend_state
.Entry
[i
] = (struct GENX(BLEND_STATE_ENTRY
)) {
124 .LogicOpEnable
= info
->logicOpEnable
,
125 .LogicOpFunction
= vk_to_gen_logic_op
[info
->logicOp
],
126 .ColorBufferBlendEnable
= a
->blendEnable
,
127 .PreBlendSourceOnlyClampEnable
= false,
128 .ColorClampRange
= COLORCLAMP_RTFORMAT
,
129 .PreBlendColorClampEnable
= true,
130 .PostBlendColorClampEnable
= true,
131 .SourceBlendFactor
= vk_to_gen_blend
[a
->srcColorBlendFactor
],
132 .DestinationBlendFactor
= vk_to_gen_blend
[a
->dstColorBlendFactor
],
133 .ColorBlendFunction
= vk_to_gen_blend_op
[a
->colorBlendOp
],
134 .SourceAlphaBlendFactor
= vk_to_gen_blend
[a
->srcAlphaBlendFactor
],
135 .DestinationAlphaBlendFactor
= vk_to_gen_blend
[a
->dstAlphaBlendFactor
],
136 .AlphaBlendFunction
= vk_to_gen_blend_op
[a
->alphaBlendOp
],
137 .WriteDisableAlpha
= !(a
->colorWriteMask
& VK_COLOR_COMPONENT_A_BIT
),
138 .WriteDisableRed
= !(a
->colorWriteMask
& VK_COLOR_COMPONENT_R_BIT
),
139 .WriteDisableGreen
= !(a
->colorWriteMask
& VK_COLOR_COMPONENT_G_BIT
),
140 .WriteDisableBlue
= !(a
->colorWriteMask
& VK_COLOR_COMPONENT_B_BIT
),
143 /* Our hardware applies the blend factor prior to the blend function
144 * regardless of what function is used. Technically, this means the
145 * hardware can do MORE than GL or Vulkan specify. However, it also
146 * means that, for MIN and MAX, we have to stomp the blend factor to
147 * ONE to make it a no-op.
149 if (a
->colorBlendOp
== VK_BLEND_OP_MIN
||
150 a
->colorBlendOp
== VK_BLEND_OP_MAX
) {
151 blend_state
.Entry
[i
].SourceBlendFactor
= BLENDFACTOR_ONE
;
152 blend_state
.Entry
[i
].DestinationBlendFactor
= BLENDFACTOR_ONE
;
154 if (a
->alphaBlendOp
== VK_BLEND_OP_MIN
||
155 a
->alphaBlendOp
== VK_BLEND_OP_MAX
) {
156 blend_state
.Entry
[i
].SourceAlphaBlendFactor
= BLENDFACTOR_ONE
;
157 blend_state
.Entry
[i
].DestinationAlphaBlendFactor
= BLENDFACTOR_ONE
;
161 GENX(BLEND_STATE_pack
)(NULL
, pipeline
->blend_state
.map
, &blend_state
);
162 if (!device
->info
.has_llc
)
163 anv_state_clflush(pipeline
->blend_state
);
165 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_BLEND_STATE_POINTERS
),
166 .BlendStatePointer
= pipeline
->blend_state
.offset
,
167 .BlendStatePointerValid
= true);
171 emit_ds_state(struct anv_pipeline
*pipeline
,
172 const VkPipelineDepthStencilStateCreateInfo
*info
)
174 uint32_t *dw
= ANV_GEN
== 8 ?
175 pipeline
->gen8
.wm_depth_stencil
: pipeline
->gen9
.wm_depth_stencil
;
178 /* We're going to OR this together with the dynamic state. We need
179 * to make sure it's initialized to something useful.
181 memset(pipeline
->gen8
.wm_depth_stencil
, 0,
182 sizeof(pipeline
->gen8
.wm_depth_stencil
));
183 memset(pipeline
->gen9
.wm_depth_stencil
, 0,
184 sizeof(pipeline
->gen9
.wm_depth_stencil
));
188 /* VkBool32 depthBoundsTestEnable; // optional (depth_bounds_test) */
190 struct GENX(3DSTATE_WM_DEPTH_STENCIL
) wm_depth_stencil
= {
191 .DepthTestEnable
= info
->depthTestEnable
,
192 .DepthBufferWriteEnable
= info
->depthWriteEnable
,
193 .DepthTestFunction
= vk_to_gen_compare_op
[info
->depthCompareOp
],
194 .DoubleSidedStencilEnable
= true,
196 .StencilTestEnable
= info
->stencilTestEnable
,
197 .StencilFailOp
= vk_to_gen_stencil_op
[info
->front
.failOp
],
198 .StencilPassDepthPassOp
= vk_to_gen_stencil_op
[info
->front
.passOp
],
199 .StencilPassDepthFailOp
= vk_to_gen_stencil_op
[info
->front
.depthFailOp
],
200 .StencilTestFunction
= vk_to_gen_compare_op
[info
->front
.compareOp
],
201 .BackfaceStencilFailOp
= vk_to_gen_stencil_op
[info
->back
.failOp
],
202 .BackfaceStencilPassDepthPassOp
= vk_to_gen_stencil_op
[info
->back
.passOp
],
203 .BackfaceStencilPassDepthFailOp
=vk_to_gen_stencil_op
[info
->back
.depthFailOp
],
204 .BackfaceStencilTestFunction
= vk_to_gen_compare_op
[info
->back
.compareOp
],
207 GENX(3DSTATE_WM_DEPTH_STENCIL_pack
)(NULL
, dw
, &wm_depth_stencil
);
211 emit_ms_state(struct anv_pipeline
*pipeline
,
212 const VkPipelineMultisampleStateCreateInfo
*info
)
214 uint32_t samples
= 1;
215 uint32_t log2_samples
= 0;
217 /* From the Vulkan 1.0 spec:
218 * If pSampleMask is NULL, it is treated as if the mask has all bits
219 * enabled, i.e. no coverage is removed from fragments.
221 * 3DSTATE_SAMPLE_MASK.SampleMask is 16 bits.
223 uint32_t sample_mask
= 0xffff;
226 samples
= info
->rasterizationSamples
;
227 log2_samples
= __builtin_ffs(samples
) - 1;
230 if (info
&& info
->pSampleMask
)
231 sample_mask
&= info
->pSampleMask
[0];
233 if (info
&& info
->sampleShadingEnable
)
234 anv_finishme("VkPipelineMultisampleStateCreateInfo::sampleShadingEnable");
236 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_MULTISAMPLE
),
238 /* The PRM says that this bit is valid only for DX9:
240 * SW can choose to set this bit only for DX9 API. DX10/OGL API's
241 * should not have any effect by setting or not setting this bit.
243 .PixelPositionOffsetEnable
= false,
245 .PixelLocation
= CENTER
,
246 .NumberofMultisamples
= log2_samples
);
248 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_SAMPLE_MASK
),
249 .SampleMask
= sample_mask
);
253 genX(graphics_pipeline_create
)(
255 struct anv_pipeline_cache
* cache
,
256 const VkGraphicsPipelineCreateInfo
* pCreateInfo
,
257 const struct anv_graphics_pipeline_create_info
*extra
,
258 const VkAllocationCallbacks
* pAllocator
,
259 VkPipeline
* pPipeline
)
261 ANV_FROM_HANDLE(anv_device
, device
, _device
);
262 struct anv_pipeline
*pipeline
;
264 uint32_t offset
, length
;
266 assert(pCreateInfo
->sType
== VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO
);
268 pipeline
= anv_alloc2(&device
->alloc
, pAllocator
, sizeof(*pipeline
), 8,
269 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT
);
270 if (pipeline
== NULL
)
271 return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY
);
273 result
= anv_pipeline_init(pipeline
, device
, cache
,
274 pCreateInfo
, extra
, pAllocator
);
275 if (result
!= VK_SUCCESS
) {
276 anv_free2(&device
->alloc
, pAllocator
, pipeline
);
280 assert(pCreateInfo
->pVertexInputState
);
281 emit_vertex_input(pipeline
, pCreateInfo
->pVertexInputState
, extra
);
282 assert(pCreateInfo
->pInputAssemblyState
);
283 emit_ia_state(pipeline
, pCreateInfo
->pInputAssemblyState
, extra
);
284 assert(pCreateInfo
->pRasterizationState
);
285 emit_rs_state(pipeline
, pCreateInfo
->pRasterizationState
,
286 pCreateInfo
->pMultisampleState
, extra
);
287 emit_ms_state(pipeline
, pCreateInfo
->pMultisampleState
);
288 emit_ds_state(pipeline
, pCreateInfo
->pDepthStencilState
);
289 emit_cb_state(pipeline
, pCreateInfo
->pColorBlendState
,
290 pCreateInfo
->pMultisampleState
);
292 emit_urb_setup(pipeline
);
294 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_CLIP
),
296 .ViewportXYClipTestEnable
= !(extra
&& extra
->disable_viewport
),
297 .MinimumPointWidth
= 0.125,
298 .MaximumPointWidth
= 255.875,
299 .MaximumVPIndex
= pCreateInfo
->pViewportState
->viewportCount
- 1);
301 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_WM
),
302 .StatisticsEnable
= true,
303 .LineEndCapAntialiasingRegionWidth
= _05pixels
,
304 .LineAntialiasingRegionWidth
= _10pixels
,
305 .EarlyDepthStencilControl
= NORMAL
,
306 .ForceThreadDispatchEnable
= NORMAL
,
307 .PointRasterizationRule
= RASTRULE_UPPER_RIGHT
,
308 .BarycentricInterpolationMode
=
309 pipeline
->ps_ksp0
== NO_KERNEL
?
310 0 : pipeline
->wm_prog_data
.barycentric_interp_modes
);
312 const struct brw_gs_prog_data
*gs_prog_data
= &pipeline
->gs_prog_data
;
314 length
= (gs_prog_data
->base
.vue_map
.num_slots
+ 1) / 2 - offset
;
316 if (pipeline
->gs_kernel
== NO_KERNEL
)
317 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_GS
), .Enable
= false);
319 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_GS
),
320 .SingleProgramFlow
= false,
321 .KernelStartPointer
= pipeline
->gs_kernel
,
322 .VectorMaskEnable
= false,
324 .BindingTableEntryCount
= 0,
325 .ExpectedVertexCount
= gs_prog_data
->vertices_in
,
327 .ScratchSpaceBasePointer
= pipeline
->scratch_start
[MESA_SHADER_GEOMETRY
],
328 .PerThreadScratchSpace
= scratch_space(&gs_prog_data
->base
.base
),
330 .OutputVertexSize
= gs_prog_data
->output_vertex_size_hwords
* 2 - 1,
331 .OutputTopology
= gs_prog_data
->output_topology
,
332 .VertexURBEntryReadLength
= gs_prog_data
->base
.urb_read_length
,
333 .IncludeVertexHandles
= gs_prog_data
->base
.include_vue_handles
,
334 .DispatchGRFStartRegisterForURBData
=
335 gs_prog_data
->base
.base
.dispatch_grf_start_reg
,
337 .MaximumNumberofThreads
= device
->info
.max_gs_threads
/ 2 - 1,
338 .ControlDataHeaderSize
= gs_prog_data
->control_data_header_size_hwords
,
339 .DispatchMode
= gs_prog_data
->base
.dispatch_mode
,
340 .StatisticsEnable
= true,
341 .IncludePrimitiveID
= gs_prog_data
->include_primitive_id
,
342 .ReorderMode
= TRAILING
,
345 .ControlDataFormat
= gs_prog_data
->control_data_format
,
347 .StaticOutput
= gs_prog_data
->static_vertex_count
>= 0,
348 .StaticOutputVertexCount
=
349 gs_prog_data
->static_vertex_count
>= 0 ?
350 gs_prog_data
->static_vertex_count
: 0,
352 /* FIXME: mesa sets this based on ctx->Transform.ClipPlanesEnabled:
353 * UserClipDistanceClipTestEnableBitmask_3DSTATE_GS(v)
354 * UserClipDistanceCullTestEnableBitmask(v)
357 .VertexURBEntryOutputReadOffset
= offset
,
358 .VertexURBEntryOutputLength
= length
);
360 const struct brw_vue_prog_data
*vue_prog_data
= &pipeline
->vs_prog_data
.base
;
361 /* Skip the VUE header and position slots */
363 length
= (vue_prog_data
->vue_map
.num_slots
+ 1) / 2 - offset
;
365 uint32_t vs_start
= pipeline
->vs_simd8
!= NO_KERNEL
? pipeline
->vs_simd8
:
368 if (vs_start
== NO_KERNEL
|| (extra
&& extra
->disable_vs
))
369 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_VS
),
370 .FunctionEnable
= false,
371 /* Even if VS is disabled, SBE still gets the amount of
372 * vertex data to read from this field. */
373 .VertexURBEntryOutputReadOffset
= offset
,
374 .VertexURBEntryOutputLength
= length
);
376 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_VS
),
377 .KernelStartPointer
= vs_start
,
378 .SingleVertexDispatch
= false,
379 .VectorMaskEnable
= false,
381 .BindingTableEntryCount
=
382 vue_prog_data
->base
.binding_table
.size_bytes
/ 4,
383 .ThreadDispatchPriority
= false,
384 .FloatingPointMode
= IEEE754
,
385 .IllegalOpcodeExceptionEnable
= false,
386 .AccessesUAV
= false,
387 .SoftwareExceptionEnable
= false,
389 .ScratchSpaceBasePointer
= pipeline
->scratch_start
[MESA_SHADER_VERTEX
],
390 .PerThreadScratchSpace
= scratch_space(&vue_prog_data
->base
),
392 .DispatchGRFStartRegisterForURBData
=
393 vue_prog_data
->base
.dispatch_grf_start_reg
,
394 .VertexURBEntryReadLength
= vue_prog_data
->urb_read_length
,
395 .VertexURBEntryReadOffset
= 0,
397 .MaximumNumberofThreads
= device
->info
.max_vs_threads
- 1,
398 .StatisticsEnable
= false,
399 .SIMD8DispatchEnable
= pipeline
->vs_simd8
!= NO_KERNEL
,
400 .VertexCacheDisable
= false,
401 .FunctionEnable
= true,
403 .VertexURBEntryOutputReadOffset
= offset
,
404 .VertexURBEntryOutputLength
= length
,
405 .UserClipDistanceClipTestEnableBitmask
= 0,
406 .UserClipDistanceCullTestEnableBitmask
= 0);
408 const struct brw_wm_prog_data
*wm_prog_data
= &pipeline
->wm_prog_data
;
410 const int num_thread_bias
= ANV_GEN
== 8 ? 2 : 1;
411 if (pipeline
->ps_ksp0
== NO_KERNEL
) {
412 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_PS
));
413 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_PS_EXTRA
),
414 .PixelShaderValid
= false);
416 /* TODO: We should clean this up. Among other things, this is mostly
417 * shared with other gens.
419 const struct brw_vue_map
*fs_input_map
;
420 if (pipeline
->gs_kernel
== NO_KERNEL
)
421 fs_input_map
= &vue_prog_data
->vue_map
;
423 fs_input_map
= &gs_prog_data
->base
.vue_map
;
425 struct GENX(3DSTATE_SBE_SWIZ
) swiz
= {
426 GENX(3DSTATE_SBE_SWIZ_header
),
429 int max_source_attr
= 0;
430 for (int attr
= 0; attr
< VARYING_SLOT_MAX
; attr
++) {
431 int input_index
= wm_prog_data
->urb_setup
[attr
];
436 int source_attr
= fs_input_map
->varying_to_slot
[attr
];
437 max_source_attr
= MAX2(max_source_attr
, source_attr
);
439 if (input_index
>= 16)
442 if (source_attr
== -1) {
443 /* This attribute does not exist in the VUE--that means that the
444 * vertex shader did not write to it. It could be that it's a
445 * regular varying read by the fragment shader but not written by
446 * the vertex shader or it's gl_PrimitiveID. In the first case the
447 * value is undefined, in the second it needs to be
450 swiz
.Attribute
[input_index
].ConstantSource
= PRIM_ID
;
451 swiz
.Attribute
[input_index
].ComponentOverrideX
= true;
452 swiz
.Attribute
[input_index
].ComponentOverrideY
= true;
453 swiz
.Attribute
[input_index
].ComponentOverrideZ
= true;
454 swiz
.Attribute
[input_index
].ComponentOverrideW
= true;
456 /* We have to subtract two slots to accout for the URB entry output
457 * read offset in the VS and GS stages.
459 swiz
.Attribute
[input_index
].SourceAttribute
= source_attr
- 2;
463 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_SBE
),
464 .AttributeSwizzleEnable
= true,
465 .ForceVertexURBEntryReadLength
= false,
466 .ForceVertexURBEntryReadOffset
= false,
467 .VertexURBEntryReadLength
=
468 DIV_ROUND_UP(max_source_attr
+ 1, 2),
469 .PointSpriteTextureCoordinateOrigin
= UPPERLEFT
,
470 .NumberofSFOutputAttributes
=
471 wm_prog_data
->num_varying_inputs
,
474 .Attribute0ActiveComponentFormat
= ACF_XYZW
,
475 .Attribute1ActiveComponentFormat
= ACF_XYZW
,
476 .Attribute2ActiveComponentFormat
= ACF_XYZW
,
477 .Attribute3ActiveComponentFormat
= ACF_XYZW
,
478 .Attribute4ActiveComponentFormat
= ACF_XYZW
,
479 .Attribute5ActiveComponentFormat
= ACF_XYZW
,
480 .Attribute6ActiveComponentFormat
= ACF_XYZW
,
481 .Attribute7ActiveComponentFormat
= ACF_XYZW
,
482 .Attribute8ActiveComponentFormat
= ACF_XYZW
,
483 .Attribute9ActiveComponentFormat
= ACF_XYZW
,
484 .Attribute10ActiveComponentFormat
= ACF_XYZW
,
485 .Attribute11ActiveComponentFormat
= ACF_XYZW
,
486 .Attribute12ActiveComponentFormat
= ACF_XYZW
,
487 .Attribute13ActiveComponentFormat
= ACF_XYZW
,
488 .Attribute14ActiveComponentFormat
= ACF_XYZW
,
489 .Attribute15ActiveComponentFormat
= ACF_XYZW
,
490 /* wow, much field, very attribute */
491 .Attribute16ActiveComponentFormat
= ACF_XYZW
,
492 .Attribute17ActiveComponentFormat
= ACF_XYZW
,
493 .Attribute18ActiveComponentFormat
= ACF_XYZW
,
494 .Attribute19ActiveComponentFormat
= ACF_XYZW
,
495 .Attribute20ActiveComponentFormat
= ACF_XYZW
,
496 .Attribute21ActiveComponentFormat
= ACF_XYZW
,
497 .Attribute22ActiveComponentFormat
= ACF_XYZW
,
498 .Attribute23ActiveComponentFormat
= ACF_XYZW
,
499 .Attribute24ActiveComponentFormat
= ACF_XYZW
,
500 .Attribute25ActiveComponentFormat
= ACF_XYZW
,
501 .Attribute26ActiveComponentFormat
= ACF_XYZW
,
502 .Attribute27ActiveComponentFormat
= ACF_XYZW
,
503 .Attribute28ActiveComponentFormat
= ACF_XYZW
,
504 .Attribute29ActiveComponentFormat
= ACF_XYZW
,
505 .Attribute28ActiveComponentFormat
= ACF_XYZW
,
506 .Attribute29ActiveComponentFormat
= ACF_XYZW
,
507 .Attribute30ActiveComponentFormat
= ACF_XYZW
,
511 uint32_t *dw
= anv_batch_emit_dwords(&pipeline
->batch
,
512 GENX(3DSTATE_SBE_SWIZ_length
));
513 GENX(3DSTATE_SBE_SWIZ_pack
)(&pipeline
->batch
, dw
, &swiz
);
515 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_PS
),
516 .KernelStartPointer0
= pipeline
->ps_ksp0
,
518 .SingleProgramFlow
= false,
519 .VectorMaskEnable
= true,
522 .ScratchSpaceBasePointer
= pipeline
->scratch_start
[MESA_SHADER_FRAGMENT
],
523 .PerThreadScratchSpace
= scratch_space(&wm_prog_data
->base
),
525 .MaximumNumberofThreadsPerPSD
= 64 - num_thread_bias
,
526 .PositionXYOffsetSelect
= wm_prog_data
->uses_pos_offset
?
527 POSOFFSET_SAMPLE
: POSOFFSET_NONE
,
528 .PushConstantEnable
= wm_prog_data
->base
.nr_params
> 0,
529 ._8PixelDispatchEnable
= pipeline
->ps_simd8
!= NO_KERNEL
,
530 ._16PixelDispatchEnable
= pipeline
->ps_simd16
!= NO_KERNEL
,
531 ._32PixelDispatchEnable
= false,
533 .DispatchGRFStartRegisterForConstantSetupData0
= pipeline
->ps_grf_start0
,
534 .DispatchGRFStartRegisterForConstantSetupData1
= 0,
535 .DispatchGRFStartRegisterForConstantSetupData2
= pipeline
->ps_grf_start2
,
537 .KernelStartPointer1
= 0,
538 .KernelStartPointer2
= pipeline
->ps_ksp2
);
540 bool per_sample_ps
= pCreateInfo
->pMultisampleState
&&
541 pCreateInfo
->pMultisampleState
->sampleShadingEnable
;
543 anv_batch_emit(&pipeline
->batch
, GENX(3DSTATE_PS_EXTRA
),
544 .PixelShaderValid
= true,
545 .PixelShaderKillsPixel
= wm_prog_data
->uses_kill
,
546 .PixelShaderComputedDepthMode
= wm_prog_data
->computed_depth_mode
,
547 .AttributeEnable
= wm_prog_data
->num_varying_inputs
> 0,
548 .oMaskPresenttoRenderTarget
= wm_prog_data
->uses_omask
,
549 .PixelShaderIsPerSample
= per_sample_ps
,
550 .PixelShaderUsesSourceDepth
= wm_prog_data
->uses_src_depth
,
551 .PixelShaderUsesSourceW
= wm_prog_data
->uses_src_w
,
553 .PixelShaderPullsBary
= wm_prog_data
->pulls_bary
,
554 .InputCoverageMaskState
= wm_prog_data
->uses_sample_mask
?
555 ICMS_INNER_CONSERVATIVE
: ICMS_NONE
,
557 .PixelShaderUsesInputCoverageMask
=
558 wm_prog_data
->uses_sample_mask
,
563 *pPipeline
= anv_pipeline_to_handle(pipeline
);