vk/pipeline: Zero out the depth-stencil state when not in use
[mesa.git] / src / vulkan / pipeline.c
1 /*
2 * Copyright © 2015 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 #include <assert.h>
25 #include <stdbool.h>
26 #include <string.h>
27 #include <unistd.h>
28 #include <fcntl.h>
29
30 #include "private.h"
31
32 // Shader functions
33
34 VkResult anv_CreateShader(
35 VkDevice _device,
36 const VkShaderCreateInfo* pCreateInfo,
37 VkShader* pShader)
38 {
39 struct anv_device *device = (struct anv_device *) _device;
40 struct anv_shader *shader;
41
42 assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_SHADER_CREATE_INFO);
43
44 shader = anv_device_alloc(device, sizeof(*shader) + pCreateInfo->codeSize, 8,
45 VK_SYSTEM_ALLOC_TYPE_API_OBJECT);
46 if (shader == NULL)
47 return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
48
49 shader->size = pCreateInfo->codeSize;
50 memcpy(shader->data, pCreateInfo->pCode, shader->size);
51
52 *pShader = (VkShader) shader;
53
54 return VK_SUCCESS;
55 }
56
57 // Pipeline functions
58
59 static void
60 emit_vertex_input(struct anv_pipeline *pipeline, VkPipelineVertexInputCreateInfo *info)
61 {
62 const uint32_t num_dwords = 1 + info->attributeCount * 2;
63 uint32_t *p;
64 bool instancing_enable[32];
65
66 pipeline->vb_used = 0;
67 for (uint32_t i = 0; i < info->bindingCount; i++) {
68 const VkVertexInputBindingDescription *desc =
69 &info->pVertexBindingDescriptions[i];
70
71 pipeline->vb_used |= 1 << desc->binding;
72 pipeline->binding_stride[desc->binding] = desc->strideInBytes;
73
74 /* Step rate is programmed per vertex element (attribute), not
75 * binding. Set up a map of which bindings step per instance, for
76 * reference by vertex element setup. */
77 switch (desc->stepRate) {
78 default:
79 case VK_VERTEX_INPUT_STEP_RATE_VERTEX:
80 instancing_enable[desc->binding] = false;
81 break;
82 case VK_VERTEX_INPUT_STEP_RATE_INSTANCE:
83 instancing_enable[desc->binding] = true;
84 break;
85 }
86 }
87
88 p = anv_batch_emitn(&pipeline->batch, num_dwords,
89 GEN8_3DSTATE_VERTEX_ELEMENTS);
90
91 for (uint32_t i = 0; i < info->attributeCount; i++) {
92 const VkVertexInputAttributeDescription *desc =
93 &info->pVertexAttributeDescriptions[i];
94 const struct anv_format *format = anv_format_for_vk_format(desc->format);
95
96 struct GEN8_VERTEX_ELEMENT_STATE element = {
97 .VertexBufferIndex = desc->binding,
98 .Valid = true,
99 .SourceElementFormat = format->format,
100 .EdgeFlagEnable = false,
101 .SourceElementOffset = desc->offsetInBytes,
102 .Component0Control = VFCOMP_STORE_SRC,
103 .Component1Control = format->channels >= 2 ? VFCOMP_STORE_SRC : VFCOMP_STORE_0,
104 .Component2Control = format->channels >= 3 ? VFCOMP_STORE_SRC : VFCOMP_STORE_0,
105 .Component3Control = format->channels >= 4 ? VFCOMP_STORE_SRC : VFCOMP_STORE_1_FP
106 };
107 GEN8_VERTEX_ELEMENT_STATE_pack(NULL, &p[1 + i * 2], &element);
108
109 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VF_INSTANCING,
110 .InstancingEnable = instancing_enable[desc->binding],
111 .VertexElementIndex = i,
112 /* Vulkan so far doesn't have an instance divisor, so
113 * this is always 1 (ignored if not instancing). */
114 .InstanceDataStepRate = 1);
115 }
116
117 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VF_SGVS,
118 .VertexIDEnable = pipeline->vs_prog_data.uses_vertexid,
119 .VertexIDComponentNumber = 2,
120 .VertexIDElementOffset = info->bindingCount,
121 .InstanceIDEnable = pipeline->vs_prog_data.uses_instanceid,
122 .InstanceIDComponentNumber = 3,
123 .InstanceIDElementOffset = info->bindingCount);
124 }
125
126 static void
127 emit_ia_state(struct anv_pipeline *pipeline,
128 VkPipelineIaStateCreateInfo *info,
129 const struct anv_pipeline_create_info *extra)
130 {
131 static const uint32_t vk_to_gen_primitive_type[] = {
132 [VK_PRIMITIVE_TOPOLOGY_POINT_LIST] = _3DPRIM_POINTLIST,
133 [VK_PRIMITIVE_TOPOLOGY_LINE_LIST] = _3DPRIM_LINELIST,
134 [VK_PRIMITIVE_TOPOLOGY_LINE_STRIP] = _3DPRIM_LINESTRIP,
135 [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST] = _3DPRIM_TRILIST,
136 [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP] = _3DPRIM_TRISTRIP,
137 [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_FAN] = _3DPRIM_TRIFAN,
138 [VK_PRIMITIVE_TOPOLOGY_LINE_LIST_ADJ] = _3DPRIM_LINELIST_ADJ,
139 [VK_PRIMITIVE_TOPOLOGY_LINE_STRIP_ADJ] = _3DPRIM_LISTSTRIP_ADJ,
140 [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST_ADJ] = _3DPRIM_TRILIST_ADJ,
141 [VK_PRIMITIVE_TOPOLOGY_TRIANGLE_STRIP_ADJ] = _3DPRIM_TRISTRIP_ADJ,
142 [VK_PRIMITIVE_TOPOLOGY_PATCH] = _3DPRIM_PATCHLIST_1
143 };
144 uint32_t topology = vk_to_gen_primitive_type[info->topology];
145
146 if (extra && extra->use_rectlist)
147 topology = _3DPRIM_RECTLIST;
148
149 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VF,
150 .IndexedDrawCutIndexEnable = info->primitiveRestartEnable,
151 .CutIndex = info->primitiveRestartIndex);
152 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VF_TOPOLOGY,
153 .PrimitiveTopologyType = topology);
154 }
155
156 static void
157 emit_rs_state(struct anv_pipeline *pipeline, VkPipelineRsStateCreateInfo *info,
158 const struct anv_pipeline_create_info *extra)
159 {
160 static const uint32_t vk_to_gen_cullmode[] = {
161 [VK_CULL_MODE_NONE] = CULLMODE_NONE,
162 [VK_CULL_MODE_FRONT] = CULLMODE_FRONT,
163 [VK_CULL_MODE_BACK] = CULLMODE_BACK,
164 [VK_CULL_MODE_FRONT_AND_BACK] = CULLMODE_BOTH
165 };
166
167 static const uint32_t vk_to_gen_fillmode[] = {
168 [VK_FILL_MODE_POINTS] = RASTER_POINT,
169 [VK_FILL_MODE_WIREFRAME] = RASTER_WIREFRAME,
170 [VK_FILL_MODE_SOLID] = RASTER_SOLID
171 };
172
173 static const uint32_t vk_to_gen_front_face[] = {
174 [VK_FRONT_FACE_CCW] = CounterClockwise,
175 [VK_FRONT_FACE_CW] = Clockwise
176 };
177
178 static const uint32_t vk_to_gen_coordinate_origin[] = {
179 [VK_COORDINATE_ORIGIN_UPPER_LEFT] = UPPERLEFT,
180 [VK_COORDINATE_ORIGIN_LOWER_LEFT] = LOWERLEFT
181 };
182
183 struct GEN8_3DSTATE_SF sf = {
184 GEN8_3DSTATE_SF_header,
185 .ViewportTransformEnable = !(extra && extra->disable_viewport),
186 .TriangleStripListProvokingVertexSelect =
187 info->provokingVertex == VK_PROVOKING_VERTEX_FIRST ? 0 : 2,
188 .LineStripListProvokingVertexSelect =
189 info->provokingVertex == VK_PROVOKING_VERTEX_FIRST ? 0 : 1,
190 .TriangleFanProvokingVertexSelect =
191 info->provokingVertex == VK_PROVOKING_VERTEX_FIRST ? 0 : 2,
192 .PointWidthSource = info->programPointSize ? Vertex : State,
193 };
194
195 /* FINISHME: bool32_t rasterizerDiscardEnable; */
196
197 GEN8_3DSTATE_SF_pack(NULL, pipeline->state_sf, &sf);
198
199 struct GEN8_3DSTATE_RASTER raster = {
200 GEN8_3DSTATE_RASTER_header,
201 .FrontWinding = vk_to_gen_front_face[info->frontFace],
202 .CullMode = vk_to_gen_cullmode[info->cullMode],
203 .FrontFaceFillMode = vk_to_gen_fillmode[info->fillMode],
204 .BackFaceFillMode = vk_to_gen_fillmode[info->fillMode],
205 .ScissorRectangleEnable = !(extra && extra->disable_scissor),
206 .ViewportZClipTestEnable = info->depthClipEnable
207 };
208
209 GEN8_3DSTATE_RASTER_pack(NULL, pipeline->state_raster, &raster);
210
211 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_SBE,
212 .ForceVertexURBEntryReadLength = false,
213 .ForceVertexURBEntryReadOffset = false,
214 .PointSpriteTextureCoordinateOrigin =
215 vk_to_gen_coordinate_origin[info->pointOrigin],
216 .NumberofSFOutputAttributes =
217 pipeline->wm_prog_data.num_varying_inputs);
218
219 }
220
221 static void
222 emit_cb_state(struct anv_pipeline *pipeline, VkPipelineCbStateCreateInfo *info)
223 {
224 struct anv_device *device = pipeline->device;
225
226 static const uint32_t vk_to_gen_logic_op[] = {
227 [VK_LOGIC_OP_COPY] = LOGICOP_COPY,
228 [VK_LOGIC_OP_CLEAR] = LOGICOP_CLEAR,
229 [VK_LOGIC_OP_AND] = LOGICOP_AND,
230 [VK_LOGIC_OP_AND_REVERSE] = LOGICOP_AND_REVERSE,
231 [VK_LOGIC_OP_AND_INVERTED] = LOGICOP_AND_INVERTED,
232 [VK_LOGIC_OP_NOOP] = LOGICOP_NOOP,
233 [VK_LOGIC_OP_XOR] = LOGICOP_XOR,
234 [VK_LOGIC_OP_OR] = LOGICOP_OR,
235 [VK_LOGIC_OP_NOR] = LOGICOP_NOR,
236 [VK_LOGIC_OP_EQUIV] = LOGICOP_EQUIV,
237 [VK_LOGIC_OP_INVERT] = LOGICOP_INVERT,
238 [VK_LOGIC_OP_OR_REVERSE] = LOGICOP_OR_REVERSE,
239 [VK_LOGIC_OP_COPY_INVERTED] = LOGICOP_COPY_INVERTED,
240 [VK_LOGIC_OP_OR_INVERTED] = LOGICOP_OR_INVERTED,
241 [VK_LOGIC_OP_NAND] = LOGICOP_NAND,
242 [VK_LOGIC_OP_SET] = LOGICOP_SET,
243 };
244
245 static const uint32_t vk_to_gen_blend[] = {
246 [VK_BLEND_ZERO] = BLENDFACTOR_ZERO,
247 [VK_BLEND_ONE] = BLENDFACTOR_ONE,
248 [VK_BLEND_SRC_COLOR] = BLENDFACTOR_SRC_COLOR,
249 [VK_BLEND_ONE_MINUS_SRC_COLOR] = BLENDFACTOR_INV_SRC_COLOR,
250 [VK_BLEND_DEST_COLOR] = BLENDFACTOR_DST_COLOR,
251 [VK_BLEND_ONE_MINUS_DEST_COLOR] = BLENDFACTOR_INV_DST_COLOR,
252 [VK_BLEND_SRC_ALPHA] = BLENDFACTOR_SRC_ALPHA,
253 [VK_BLEND_ONE_MINUS_SRC_ALPHA] = BLENDFACTOR_INV_SRC_ALPHA,
254 [VK_BLEND_DEST_ALPHA] = BLENDFACTOR_DST_ALPHA,
255 [VK_BLEND_ONE_MINUS_DEST_ALPHA] = BLENDFACTOR_INV_DST_ALPHA,
256 [VK_BLEND_CONSTANT_COLOR] = BLENDFACTOR_CONST_COLOR,
257 [VK_BLEND_ONE_MINUS_CONSTANT_COLOR] = BLENDFACTOR_INV_CONST_COLOR,
258 [VK_BLEND_CONSTANT_ALPHA] = BLENDFACTOR_CONST_ALPHA,
259 [VK_BLEND_ONE_MINUS_CONSTANT_ALPHA] = BLENDFACTOR_INV_CONST_ALPHA,
260 [VK_BLEND_SRC_ALPHA_SATURATE] = BLENDFACTOR_SRC_ALPHA_SATURATE,
261 [VK_BLEND_SRC1_COLOR] = BLENDFACTOR_SRC1_COLOR,
262 [VK_BLEND_ONE_MINUS_SRC1_COLOR] = BLENDFACTOR_INV_SRC1_COLOR,
263 [VK_BLEND_SRC1_ALPHA] = BLENDFACTOR_SRC1_ALPHA,
264 [VK_BLEND_ONE_MINUS_SRC1_ALPHA] = BLENDFACTOR_INV_SRC1_ALPHA,
265 };
266
267 static const uint32_t vk_to_gen_blend_op[] = {
268 [VK_BLEND_OP_ADD] = BLENDFUNCTION_ADD,
269 [VK_BLEND_OP_SUBTRACT] = BLENDFUNCTION_SUBTRACT,
270 [VK_BLEND_OP_REVERSE_SUBTRACT] = BLENDFUNCTION_REVERSE_SUBTRACT,
271 [VK_BLEND_OP_MIN] = BLENDFUNCTION_MIN,
272 [VK_BLEND_OP_MAX] = BLENDFUNCTION_MAX,
273 };
274
275 uint32_t num_dwords = 1 + info->attachmentCount * 2;
276 pipeline->blend_state =
277 anv_state_pool_alloc(&device->dynamic_state_pool, num_dwords * 4, 64);
278
279 struct GEN8_BLEND_STATE blend_state = {
280 .AlphaToCoverageEnable = info->alphaToCoverageEnable,
281 };
282
283 uint32_t *state = pipeline->blend_state.map;
284 GEN8_BLEND_STATE_pack(NULL, state, &blend_state);
285
286 for (uint32_t i = 0; i < info->attachmentCount; i++) {
287 const VkPipelineCbAttachmentState *a = &info->pAttachments[i];
288
289 struct GEN8_BLEND_STATE_ENTRY entry = {
290 .LogicOpEnable = info->logicOpEnable,
291 .LogicOpFunction = vk_to_gen_logic_op[info->logicOp],
292 .ColorBufferBlendEnable = a->blendEnable,
293 .PreBlendSourceOnlyClampEnable = false,
294 .PreBlendColorClampEnable = false,
295 .PostBlendColorClampEnable = false,
296 .SourceBlendFactor = vk_to_gen_blend[a->srcBlendColor],
297 .DestinationBlendFactor = vk_to_gen_blend[a->destBlendColor],
298 .ColorBlendFunction = vk_to_gen_blend_op[a->blendOpColor],
299 .SourceAlphaBlendFactor = vk_to_gen_blend[a->srcBlendAlpha],
300 .DestinationAlphaBlendFactor = vk_to_gen_blend[a->destBlendAlpha],
301 .AlphaBlendFunction = vk_to_gen_blend_op[a->blendOpAlpha],
302 .WriteDisableAlpha = !(a->channelWriteMask & VK_CHANNEL_A_BIT),
303 .WriteDisableRed = !(a->channelWriteMask & VK_CHANNEL_R_BIT),
304 .WriteDisableGreen = !(a->channelWriteMask & VK_CHANNEL_G_BIT),
305 .WriteDisableBlue = !(a->channelWriteMask & VK_CHANNEL_B_BIT),
306 };
307
308 GEN8_BLEND_STATE_ENTRY_pack(NULL, state + i * 2 + 1, &entry);
309 }
310
311 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_BLEND_STATE_POINTERS,
312 .BlendStatePointer = pipeline->blend_state.offset,
313 .BlendStatePointerValid = true);
314 }
315
316 static const uint32_t vk_to_gen_compare_op[] = {
317 [VK_COMPARE_OP_NEVER] = COMPAREFUNCTION_NEVER,
318 [VK_COMPARE_OP_LESS] = COMPAREFUNCTION_LESS,
319 [VK_COMPARE_OP_EQUAL] = COMPAREFUNCTION_EQUAL,
320 [VK_COMPARE_OP_LESS_EQUAL] = COMPAREFUNCTION_LEQUAL,
321 [VK_COMPARE_OP_GREATER] = COMPAREFUNCTION_GREATER,
322 [VK_COMPARE_OP_NOT_EQUAL] = COMPAREFUNCTION_NOTEQUAL,
323 [VK_COMPARE_OP_GREATER_EQUAL] = COMPAREFUNCTION_GEQUAL,
324 [VK_COMPARE_OP_ALWAYS] = COMPAREFUNCTION_ALWAYS,
325 };
326
327 static const uint32_t vk_to_gen_stencil_op[] = {
328 [VK_STENCIL_OP_KEEP] = 0,
329 [VK_STENCIL_OP_ZERO] = 0,
330 [VK_STENCIL_OP_REPLACE] = 0,
331 [VK_STENCIL_OP_INC_CLAMP] = 0,
332 [VK_STENCIL_OP_DEC_CLAMP] = 0,
333 [VK_STENCIL_OP_INVERT] = 0,
334 [VK_STENCIL_OP_INC_WRAP] = 0,
335 [VK_STENCIL_OP_DEC_WRAP] = 0
336 };
337
338 static void
339 emit_ds_state(struct anv_pipeline *pipeline, VkPipelineDsStateCreateInfo *info)
340 {
341 if (info == NULL) {
342 /* We're going to OR this together with the dynamic state. We need
343 * to make sure it's initialized to something useful.
344 */
345 memset(pipeline->state_wm_depth_stencil, 0,
346 sizeof(pipeline->state_wm_depth_stencil));
347 return;
348 }
349
350 /* bool32_t depthBoundsEnable; // optional (depth_bounds_test) */
351
352 struct GEN8_3DSTATE_WM_DEPTH_STENCIL wm_depth_stencil = {
353 .DepthTestEnable = info->depthTestEnable,
354 .DepthBufferWriteEnable = info->depthWriteEnable,
355 .DepthTestFunction = vk_to_gen_compare_op[info->depthCompareOp],
356 .DoubleSidedStencilEnable = true,
357
358 .StencilTestEnable = info->stencilTestEnable,
359 .StencilFailOp = vk_to_gen_stencil_op[info->front.stencilFailOp],
360 .StencilPassDepthPassOp = vk_to_gen_stencil_op[info->front.stencilPassOp],
361 .StencilPassDepthFailOp = vk_to_gen_stencil_op[info->front.stencilDepthFailOp],
362 .StencilTestFunction = vk_to_gen_compare_op[info->front.stencilCompareOp],
363 .BackfaceStencilFailOp = vk_to_gen_stencil_op[info->back.stencilFailOp],
364 .BackfaceStencilPassDepthPassOp = vk_to_gen_stencil_op[info->back.stencilPassOp],
365 .BackfaceStencilPassDepthFailOp =vk_to_gen_stencil_op[info->back.stencilDepthFailOp],
366 .BackfaceStencilTestFunction = vk_to_gen_compare_op[info->back.stencilCompareOp],
367 };
368
369 GEN8_3DSTATE_WM_DEPTH_STENCIL_pack(NULL, pipeline->state_wm_depth_stencil, &wm_depth_stencil);
370 }
371
372 VkResult anv_CreateGraphicsPipeline(
373 VkDevice device,
374 const VkGraphicsPipelineCreateInfo* pCreateInfo,
375 VkPipeline* pPipeline)
376 {
377 return anv_pipeline_create(device, pCreateInfo, NULL, pPipeline);
378 }
379
380 static void
381 anv_pipeline_destroy(struct anv_device *device,
382 struct anv_object *object,
383 VkObjectType obj_type)
384 {
385 struct anv_pipeline *pipeline = (struct anv_pipeline*) object;
386
387 assert(obj_type == VK_OBJECT_TYPE_PIPELINE);
388
389 anv_compiler_free(pipeline);
390 anv_reloc_list_finish(&pipeline->batch.relocs, pipeline->device);
391 anv_state_stream_finish(&pipeline->program_stream);
392 anv_state_pool_free(&device->dynamic_state_pool, pipeline->blend_state);
393 anv_device_free(pipeline->device, pipeline);
394 }
395
396 VkResult
397 anv_pipeline_create(
398 VkDevice _device,
399 const VkGraphicsPipelineCreateInfo* pCreateInfo,
400 const struct anv_pipeline_create_info * extra,
401 VkPipeline* pPipeline)
402 {
403 struct anv_device *device = (struct anv_device *) _device;
404 struct anv_pipeline *pipeline;
405 const struct anv_common *common;
406 VkPipelineShaderStageCreateInfo *shader_create_info;
407 VkPipelineIaStateCreateInfo *ia_info = NULL;
408 VkPipelineRsStateCreateInfo *rs_info = NULL;
409 VkPipelineDsStateCreateInfo *ds_info = NULL;
410 VkPipelineCbStateCreateInfo *cb_info = NULL;
411 VkPipelineVertexInputCreateInfo *vi_info = NULL;
412 VkResult result;
413 uint32_t offset, length;
414
415 assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO);
416
417 pipeline = anv_device_alloc(device, sizeof(*pipeline), 8,
418 VK_SYSTEM_ALLOC_TYPE_API_OBJECT);
419 if (pipeline == NULL)
420 return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
421
422 pipeline->base.destructor = anv_pipeline_destroy;
423 pipeline->device = device;
424 pipeline->layout = (struct anv_pipeline_layout *) pCreateInfo->layout;
425 memset(pipeline->shaders, 0, sizeof(pipeline->shaders));
426
427 result = anv_reloc_list_init(&pipeline->batch.relocs, device);
428 if (result != VK_SUCCESS) {
429 anv_device_free(device, pipeline);
430 return result;
431 }
432 pipeline->batch.next = pipeline->batch.start = pipeline->batch_data;
433 pipeline->batch.end = pipeline->batch.start + sizeof(pipeline->batch_data);
434
435 anv_state_stream_init(&pipeline->program_stream,
436 &device->instruction_block_pool);
437
438 for (common = pCreateInfo->pNext; common; common = common->pNext) {
439 switch (common->sType) {
440 case VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_CREATE_INFO:
441 vi_info = (VkPipelineVertexInputCreateInfo *) common;
442 break;
443 case VK_STRUCTURE_TYPE_PIPELINE_IA_STATE_CREATE_INFO:
444 ia_info = (VkPipelineIaStateCreateInfo *) common;
445 break;
446 case VK_STRUCTURE_TYPE_PIPELINE_TESS_STATE_CREATE_INFO:
447 anv_finishme("VK_STRUCTURE_TYPE_PIPELINE_TESS_STATE_CREATE_INFO");
448 break;
449 case VK_STRUCTURE_TYPE_PIPELINE_VP_STATE_CREATE_INFO:
450 anv_finishme("VK_STRUCTURE_TYPE_PIPELINE_VP_STATE_CREATE_INFO");
451 break;
452 case VK_STRUCTURE_TYPE_PIPELINE_RS_STATE_CREATE_INFO:
453 rs_info = (VkPipelineRsStateCreateInfo *) common;
454 break;
455 case VK_STRUCTURE_TYPE_PIPELINE_MS_STATE_CREATE_INFO:
456 anv_finishme("VK_STRUCTURE_TYPE_PIPELINE_MS_STATE_CREATE_INFO");
457 break;
458 case VK_STRUCTURE_TYPE_PIPELINE_CB_STATE_CREATE_INFO:
459 cb_info = (VkPipelineCbStateCreateInfo *) common;
460 break;
461 case VK_STRUCTURE_TYPE_PIPELINE_DS_STATE_CREATE_INFO:
462 ds_info = (VkPipelineDsStateCreateInfo *) common;
463 break;
464 case VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO:
465 shader_create_info = (VkPipelineShaderStageCreateInfo *) common;
466 pipeline->shaders[shader_create_info->shader.stage] =
467 (struct anv_shader *) shader_create_info->shader.shader;
468 break;
469 default:
470 break;
471 }
472 }
473
474 pipeline->use_repclear = extra && extra->use_repclear;
475
476 anv_compiler_run(device->compiler, pipeline);
477
478 /* FIXME: The compiler dead-codes FS inputs when we don't have a VS, so we
479 * hard code this to num_attributes - 2. This is because the attributes
480 * include VUE header and position, which aren't counted as varying
481 * inputs. */
482 if (pipeline->vs_simd8 == NO_KERNEL)
483 pipeline->wm_prog_data.num_varying_inputs = vi_info->attributeCount - 2;
484
485 assert(vi_info);
486 emit_vertex_input(pipeline, vi_info);
487 assert(ia_info);
488 emit_ia_state(pipeline, ia_info, extra);
489 assert(rs_info);
490 emit_rs_state(pipeline, rs_info, extra);
491 emit_ds_state(pipeline, ds_info);
492 emit_cb_state(pipeline, cb_info);
493
494 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_CLIP,
495 .ClipEnable = true,
496 .ViewportXYClipTestEnable = !(extra && extra->disable_viewport));
497
498 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_WM,
499 .StatisticsEnable = true,
500 .LineEndCapAntialiasingRegionWidth = _05pixels,
501 .LineAntialiasingRegionWidth = _10pixels,
502 .EarlyDepthStencilControl = NORMAL,
503 .ForceThreadDispatchEnable = NORMAL,
504 .PointRasterizationRule = RASTRULE_UPPER_RIGHT,
505 .BarycentricInterpolationMode =
506 pipeline->wm_prog_data.barycentric_interp_modes);
507
508 uint32_t samples = 1;
509 uint32_t log2_samples = __builtin_ffs(samples) - 1;
510 bool enable_sampling = samples > 1 ? true : false;
511
512 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_MULTISAMPLE,
513 .PixelPositionOffsetEnable = enable_sampling,
514 .PixelLocation = CENTER,
515 .NumberofMultisamples = log2_samples);
516
517 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_URB_VS,
518 .VSURBStartingAddress = pipeline->urb.vs_start,
519 .VSURBEntryAllocationSize = pipeline->urb.vs_size - 1,
520 .VSNumberofURBEntries = pipeline->urb.nr_vs_entries);
521
522 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_URB_GS,
523 .GSURBStartingAddress = pipeline->urb.gs_start,
524 .GSURBEntryAllocationSize = pipeline->urb.gs_size - 1,
525 .GSNumberofURBEntries = pipeline->urb.nr_gs_entries);
526
527 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_URB_HS,
528 .HSURBStartingAddress = pipeline->urb.vs_start,
529 .HSURBEntryAllocationSize = 0,
530 .HSNumberofURBEntries = 0);
531
532 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_URB_DS,
533 .DSURBStartingAddress = pipeline->urb.vs_start,
534 .DSURBEntryAllocationSize = 0,
535 .DSNumberofURBEntries = 0);
536
537 const struct brw_gs_prog_data *gs_prog_data = &pipeline->gs_prog_data;
538 offset = 1;
539 length = (gs_prog_data->base.vue_map.num_slots + 1) / 2 - offset;
540
541 if (pipeline->gs_vec4 == NO_KERNEL)
542 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_GS, .Enable = false);
543 else
544 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_GS,
545 .SingleProgramFlow = false,
546 .KernelStartPointer = pipeline->gs_vec4,
547 .VectorMaskEnable = Vmask,
548 .SamplerCount = 0,
549 .BindingTableEntryCount = 0,
550 .ExpectedVertexCount = pipeline->gs_vertex_count,
551
552 .PerThreadScratchSpace = 0,
553 .ScratchSpaceBasePointer = 0,
554
555 .OutputVertexSize = gs_prog_data->output_vertex_size_hwords * 2 - 1,
556 .OutputTopology = gs_prog_data->output_topology,
557 .VertexURBEntryReadLength = gs_prog_data->base.urb_read_length,
558 .DispatchGRFStartRegisterForURBData =
559 gs_prog_data->base.base.dispatch_grf_start_reg,
560
561 .MaximumNumberofThreads = device->info.max_gs_threads,
562 .ControlDataHeaderSize = gs_prog_data->control_data_header_size_hwords,
563 //pipeline->gs_prog_data.dispatch_mode |
564 .StatisticsEnable = true,
565 .IncludePrimitiveID = gs_prog_data->include_primitive_id,
566 .ReorderMode = TRAILING,
567 .Enable = true,
568
569 .ControlDataFormat = gs_prog_data->control_data_format,
570
571 /* FIXME: mesa sets this based on ctx->Transform.ClipPlanesEnabled:
572 * UserClipDistanceClipTestEnableBitmask_3DSTATE_GS(v)
573 * UserClipDistanceCullTestEnableBitmask(v)
574 */
575
576 .VertexURBEntryOutputReadOffset = offset,
577 .VertexURBEntryOutputLength = length);
578
579 //trp_generate_blend_hw_cmds(batch, pipeline);
580
581 const struct brw_vue_prog_data *vue_prog_data = &pipeline->vs_prog_data.base;
582 /* Skip the VUE header and position slots */
583 offset = 1;
584 length = (vue_prog_data->vue_map.num_slots + 1) / 2 - offset;
585
586 if (pipeline->vs_simd8 == NO_KERNEL || (extra && extra->disable_vs))
587 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VS,
588 .FunctionEnable = false,
589 .VertexURBEntryOutputReadOffset = 1,
590 /* Even if VS is disabled, SBE still gets the amount of
591 * vertex data to read from this field. We use attribute
592 * count - 1, as we don't count the VUE header here. */
593 .VertexURBEntryOutputLength =
594 DIV_ROUND_UP(vi_info->attributeCount - 1, 2));
595 else
596 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VS,
597 .KernelStartPointer = pipeline->vs_simd8,
598 .SingleVertexDispatch = Multiple,
599 .VectorMaskEnable = Dmask,
600 .SamplerCount = 0,
601 .BindingTableEntryCount =
602 vue_prog_data->base.binding_table.size_bytes / 4,
603 .ThreadDispatchPriority = Normal,
604 .FloatingPointMode = IEEE754,
605 .IllegalOpcodeExceptionEnable = false,
606 .AccessesUAV = false,
607 .SoftwareExceptionEnable = false,
608
609 /* FIXME: pointer needs to be assigned outside as it aliases
610 * PerThreadScratchSpace.
611 */
612 .ScratchSpaceBasePointer = 0,
613 .PerThreadScratchSpace = 0,
614
615 .DispatchGRFStartRegisterForURBData =
616 vue_prog_data->base.dispatch_grf_start_reg,
617 .VertexURBEntryReadLength = vue_prog_data->urb_read_length,
618 .VertexURBEntryReadOffset = 0,
619
620 .MaximumNumberofThreads = device->info.max_vs_threads - 1,
621 .StatisticsEnable = false,
622 .SIMD8DispatchEnable = true,
623 .VertexCacheDisable = ia_info->disableVertexReuse,
624 .FunctionEnable = true,
625
626 .VertexURBEntryOutputReadOffset = offset,
627 .VertexURBEntryOutputLength = length,
628 .UserClipDistanceClipTestEnableBitmask = 0,
629 .UserClipDistanceCullTestEnableBitmask = 0);
630
631 const struct brw_wm_prog_data *wm_prog_data = &pipeline->wm_prog_data;
632 uint32_t ksp0, ksp2, grf_start0, grf_start2;
633
634 ksp2 = 0;
635 grf_start2 = 0;
636 if (pipeline->ps_simd8 != NO_KERNEL) {
637 ksp0 = pipeline->ps_simd8;
638 grf_start0 = wm_prog_data->base.dispatch_grf_start_reg;
639 if (pipeline->ps_simd16 != NO_KERNEL) {
640 ksp2 = pipeline->ps_simd16;
641 grf_start2 = wm_prog_data->dispatch_grf_start_reg_16;
642 }
643 } else if (pipeline->ps_simd16 != NO_KERNEL) {
644 ksp0 = pipeline->ps_simd16;
645 grf_start0 = wm_prog_data->dispatch_grf_start_reg_16;
646 } else {
647 unreachable("no ps shader");
648 }
649
650 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_PS,
651 .KernelStartPointer0 = ksp0,
652
653 .SingleProgramFlow = false,
654 .VectorMaskEnable = true,
655 .SamplerCount = 1,
656
657 .ScratchSpaceBasePointer = 0,
658 .PerThreadScratchSpace = 0,
659
660 .MaximumNumberofThreadsPerPSD = 64 - 2,
661 .PositionXYOffsetSelect = wm_prog_data->uses_pos_offset ?
662 POSOFFSET_SAMPLE: POSOFFSET_NONE,
663 .PushConstantEnable = wm_prog_data->base.nr_params > 0,
664 ._8PixelDispatchEnable = pipeline->ps_simd8 != NO_KERNEL,
665 ._16PixelDispatchEnable = pipeline->ps_simd16 != NO_KERNEL,
666 ._32PixelDispatchEnable = false,
667
668 .DispatchGRFStartRegisterForConstantSetupData0 = grf_start0,
669 .DispatchGRFStartRegisterForConstantSetupData1 = 0,
670 .DispatchGRFStartRegisterForConstantSetupData2 = grf_start2,
671
672 .KernelStartPointer1 = 0,
673 .KernelStartPointer2 = ksp2);
674
675 bool per_sample_ps = false;
676 anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_PS_EXTRA,
677 .PixelShaderValid = true,
678 .PixelShaderKillsPixel = wm_prog_data->uses_kill,
679 .PixelShaderComputedDepthMode = wm_prog_data->computed_depth_mode,
680 .AttributeEnable = wm_prog_data->num_varying_inputs > 0,
681 .oMaskPresenttoRenderTarget = wm_prog_data->uses_omask,
682 .PixelShaderIsPerSample = per_sample_ps);
683
684 *pPipeline = (VkPipeline) pipeline;
685
686 return VK_SUCCESS;
687 }
688
689 VkResult anv_CreateGraphicsPipelineDerivative(
690 VkDevice device,
691 const VkGraphicsPipelineCreateInfo* pCreateInfo,
692 VkPipeline basePipeline,
693 VkPipeline* pPipeline)
694 {
695 stub_return(VK_UNSUPPORTED);
696 }
697
698 VkResult anv_CreateComputePipeline(
699 VkDevice device,
700 const VkComputePipelineCreateInfo* pCreateInfo,
701 VkPipeline* pPipeline)
702 {
703 stub_return(VK_UNSUPPORTED);
704 }
705
706 VkResult anv_StorePipeline(
707 VkDevice device,
708 VkPipeline pipeline,
709 size_t* pDataSize,
710 void* pData)
711 {
712 stub_return(VK_UNSUPPORTED);
713 }
714
715 VkResult anv_LoadPipeline(
716 VkDevice device,
717 size_t dataSize,
718 const void* pData,
719 VkPipeline* pPipeline)
720 {
721 stub_return(VK_UNSUPPORTED);
722 }
723
724 VkResult anv_LoadPipelineDerivative(
725 VkDevice device,
726 size_t dataSize,
727 const void* pData,
728 VkPipeline basePipeline,
729 VkPipeline* pPipeline)
730 {
731 stub_return(VK_UNSUPPORTED);
732 }
733
734 // Pipeline layout functions
735
736 VkResult anv_CreatePipelineLayout(
737 VkDevice _device,
738 const VkPipelineLayoutCreateInfo* pCreateInfo,
739 VkPipelineLayout* pPipelineLayout)
740 {
741 struct anv_device *device = (struct anv_device *) _device;
742 struct anv_pipeline_layout *layout;
743
744 assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO);
745
746 layout = anv_device_alloc(device, sizeof(*layout), 8,
747 VK_SYSTEM_ALLOC_TYPE_API_OBJECT);
748 if (layout == NULL)
749 return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
750
751 layout->num_sets = pCreateInfo->descriptorSetCount;
752
753 uint32_t surface_start[VK_NUM_SHADER_STAGE] = { 0, };
754 uint32_t sampler_start[VK_NUM_SHADER_STAGE] = { 0, };
755
756 for (uint32_t s = 0; s < VK_NUM_SHADER_STAGE; s++) {
757 layout->stage[s].surface_count = 0;
758 layout->stage[s].sampler_count = 0;
759 }
760
761 for (uint32_t i = 0; i < pCreateInfo->descriptorSetCount; i++) {
762 struct anv_descriptor_set_layout *set_layout =
763 (struct anv_descriptor_set_layout *) pCreateInfo->pSetLayouts[i];
764
765 layout->set[i].layout = set_layout;
766 for (uint32_t s = 0; s < VK_NUM_SHADER_STAGE; s++) {
767 layout->set[i].surface_start[s] = surface_start[s];
768 surface_start[s] += set_layout->stage[s].surface_count;
769 layout->set[i].sampler_start[s] = sampler_start[s];
770 sampler_start[s] += set_layout->stage[s].sampler_count;
771
772 layout->stage[s].surface_count += set_layout->stage[s].surface_count;
773 layout->stage[s].sampler_count += set_layout->stage[s].sampler_count;
774 }
775 }
776
777 *pPipelineLayout = (VkPipelineLayout) layout;
778
779 return VK_SUCCESS;
780 }