d3d11_context.h revision a359eb80c5e141f625cfe42b4d97bf78cf25d128
1/************************************************************************** 2 * 3 * Copyright 2010 Luca Barbieri 4 * 5 * Permission is hereby granted, free of charge, to any person obtaining 6 * a copy of this software and associated documentation files (the 7 * "Software"), to deal in the Software without restriction, including 8 * without limitation the rights to use, copy, modify, merge, publish, 9 * distribute, sublicense, and/or sell copies of the Software, and to 10 * permit persons to whom the Software is furnished to do so, subject to 11 * the following conditions: 12 * 13 * The above copyright notice and this permission notice (including the 14 * next paragraph) shall be included in all copies or substantial 15 * portions of the Software. 16 * 17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 18 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 19 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. 20 * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE 21 * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION 22 * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION 23 * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. 24 * 25 **************************************************************************/ 26 27/* used to unbind things, we need 128 due to resources */ 28static const void* zero_data[128]; 29 30#define UPDATE_VIEWS_SHIFT (D3D11_STAGES * 0) 31#define UPDATE_SAMPLERS_SHIFT (D3D11_STAGES * 1) 32#define UPDATE_VERTEX_BUFFERS (1 << (D3D11_STAGES * 2)) 33 34#if API >= 11 35template<typename PtrTraits> 36struct GalliumD3D11DeviceContext : 37 public GalliumD3D11DeviceChild<ID3D11DeviceContext> 38{ 39#else 40template<bool threadsafe> 41struct GalliumD3D10Device : public GalliumD3D10ScreenImpl<threadsafe> 42{ 43 typedef simple_ptr_traits PtrTraits; 44 typedef GalliumD3D10Device GalliumD3D10DeviceContext; 45#endif 46 47 refcnt_ptr<GalliumD3D11Shader<>, PtrTraits> shaders[D3D11_STAGES]; 48 refcnt_ptr<GalliumD3D11InputLayout, PtrTraits> input_layout; 49 refcnt_ptr<GalliumD3D11Buffer, PtrTraits> index_buffer; 50 refcnt_ptr<GalliumD3D11RasterizerState, PtrTraits> rasterizer_state; 51 refcnt_ptr<GalliumD3D11DepthStencilState, PtrTraits> depth_stencil_state; 52 refcnt_ptr<GalliumD3D11BlendState, PtrTraits> blend_state; 53 refcnt_ptr<GalliumD3D11DepthStencilView, PtrTraits> depth_stencil_view; 54 refcnt_ptr<GalliumD3D11Predicate, PtrTraits> render_predicate; 55 56 refcnt_ptr<GalliumD3D11Buffer, PtrTraits> constant_buffers[D3D11_STAGES][D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT]; 57 refcnt_ptr<GalliumD3D11ShaderResourceView, PtrTraits> shader_resource_views[D3D11_STAGES][D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT]; 58 refcnt_ptr<GalliumD3D11SamplerState, PtrTraits> samplers[D3D11_STAGES][D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT]; 59 refcnt_ptr<GalliumD3D11Buffer, PtrTraits> input_buffers[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; 60 refcnt_ptr<GalliumD3D11RenderTargetView, PtrTraits> render_target_views[D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT]; 61 refcnt_ptr<GalliumD3D11Buffer, PtrTraits> so_targets[D3D11_SO_BUFFER_SLOT_COUNT]; 62 63#if API >= 11 64 refcnt_ptr<ID3D11UnorderedAccessView, PtrTraits> cs_unordered_access_views[D3D11_PS_CS_UAV_REGISTER_COUNT]; 65 refcnt_ptr<ID3D11UnorderedAccessView, PtrTraits> om_unordered_access_views[D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT]; 66#endif 67 68 D3D11_VIEWPORT viewports[D3D11_VIEWPORT_AND_SCISSORRECT_OBJECT_COUNT_PER_PIPELINE]; 69 D3D11_RECT scissor_rects[D3D11_VIEWPORT_AND_SCISSORRECT_OBJECT_COUNT_PER_PIPELINE]; 70 unsigned so_offsets[D3D11_SO_BUFFER_SLOT_COUNT]; 71 D3D11_PRIMITIVE_TOPOLOGY primitive_topology; 72 DXGI_FORMAT index_format; 73 unsigned index_offset; 74 BOOL render_predicate_value; 75 float blend_color[4]; 76 unsigned sample_mask; 77 unsigned stencil_ref; 78 bool depth_clamp; 79 80 void* default_input_layout; 81 void* default_rasterizer; 82 void* default_depth_stencil; 83 void* default_blend; 84 void* default_sampler; 85 void* ld_sampler; 86 void * default_shaders[D3D11_STAGES]; 87 88 // derived state 89 int primitive_mode; 90 struct pipe_vertex_buffer vertex_buffers[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; 91 struct pipe_resource* so_buffers[D3D11_SO_BUFFER_SLOT_COUNT]; 92 struct pipe_sampler_view* sampler_views[D3D11_STAGES][D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT]; 93 struct 94 { 95 void* ld; // accessed with a -1 index from v 96 void* v[D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT]; 97 } sampler_csos[D3D11_STAGES]; 98 struct pipe_resource * buffers[D3D11_SO_BUFFER_SLOT_COUNT]; 99 unsigned num_shader_resource_views[D3D11_STAGES]; 100 unsigned num_samplers[D3D11_STAGES]; 101 unsigned num_vertex_buffers; 102 unsigned num_render_target_views; 103 unsigned num_viewports; 104 unsigned num_scissor_rects; 105 unsigned num_so_targets; 106 107 struct pipe_context* pipe; 108 unsigned update_flags; 109 110 bool owns_pipe; 111 unsigned context_flags; 112 113 GalliumD3D11Caps caps; 114 115 cso_context* cso_ctx; 116 gen_mipmap_state* gen_mipmap; 117 118#if API >= 11 119#define SYNCHRONIZED do {} while(0) 120 121 GalliumD3D11DeviceContext(GalliumD3D11Screen* device, pipe_context* pipe, bool owns_pipe, unsigned context_flags = 0) 122 : GalliumD3D11DeviceChild<ID3D11DeviceContext>(device), pipe(pipe), owns_pipe(owns_pipe), context_flags(context_flags) 123 { 124 caps = device->screen_caps; 125 init_context(); 126 } 127 128 ~GalliumD3D11DeviceContext() 129 { 130 destroy_context(); 131 } 132#else 133#define SYNCHRONIZED lock_t<maybe_mutex_t<threadsafe> > lock_(this->mutex) 134 135 GalliumD3D10Device(pipe_screen* screen, pipe_context* pipe, bool owns_pipe, unsigned creation_flags, IDXGIAdapter* adapter) 136 : GalliumD3D10ScreenImpl<threadsafe>(screen, pipe, owns_pipe, creation_flags, adapter), pipe(pipe), owns_pipe(owns_pipe), context_flags(0) 137 { 138 caps = this->screen_caps; 139 init_context(); 140 } 141 142 ~GalliumD3D10Device() 143 { 144 destroy_context(); 145 } 146#endif 147 148 void init_context() 149 { 150 if(!pipe->begin_query) 151 caps.queries = false; 152 if(!pipe->render_condition) 153 caps.render_condition = false; 154 if(!pipe->bind_gs_state) 155 { 156 caps.gs = false; 157 caps.stages = 2; 158 } 159 if(!pipe->set_stream_output_buffers) 160 caps.so = false; 161 if(!pipe->set_geometry_sampler_views) 162 caps.stages_with_sampling &=~ (1 << PIPE_SHADER_GEOMETRY); 163 if(!pipe->set_fragment_sampler_views) 164 caps.stages_with_sampling &=~ (1 << PIPE_SHADER_FRAGMENT); 165 if(!pipe->set_vertex_sampler_views) 166 caps.stages_with_sampling &=~ (1 << PIPE_SHADER_VERTEX); 167 168 update_flags = 0; 169 170 // pipeline state 171 memset(viewports, 0, sizeof(viewports)); 172 memset(scissor_rects, 0, sizeof(scissor_rects)); 173 memset(so_offsets, 0, sizeof(so_offsets)); 174 primitive_topology = D3D11_PRIMITIVE_TOPOLOGY_UNDEFINED; 175 index_format = DXGI_FORMAT_UNKNOWN; 176 index_offset = 0; 177 render_predicate_value = 0; 178 memset(blend_color, 0, sizeof(blend_color)); 179 sample_mask = ~0; 180 stencil_ref = 0; 181 depth_clamp = 0; 182 183 // derived state 184 primitive_mode = 0; 185 memset(vertex_buffers, 0, sizeof(vertex_buffers)); 186 memset(so_buffers, 0, sizeof(so_buffers)); 187 memset(sampler_views, 0, sizeof(sampler_views)); 188 memset(sampler_csos, 0, sizeof(sampler_csos)); 189 memset(num_shader_resource_views, 0, sizeof(num_shader_resource_views)); 190 memset(num_samplers, 0, sizeof(num_samplers)); 191 num_vertex_buffers = 0; 192 num_render_target_views = 0; 193 num_viewports = 0; 194 num_scissor_rects = 0; 195 num_so_targets = 0; 196 197 default_input_layout = pipe->create_vertex_elements_state(pipe, 0, 0); 198 199 struct pipe_rasterizer_state rasterizerd; 200 memset(&rasterizerd, 0, sizeof(rasterizerd)); 201 rasterizerd.gl_rasterization_rules = 1; 202 rasterizerd.cull_face = PIPE_FACE_BACK; 203 default_rasterizer = pipe->create_rasterizer_state(pipe, &rasterizerd); 204 205 struct pipe_depth_stencil_alpha_state depth_stencild; 206 memset(&depth_stencild, 0, sizeof(depth_stencild)); 207 depth_stencild.depth.enabled = TRUE; 208 depth_stencild.depth.writemask = 1; 209 depth_stencild.depth.func = PIPE_FUNC_LESS; 210 default_depth_stencil = pipe->create_depth_stencil_alpha_state(pipe, &depth_stencild); 211 212 struct pipe_blend_state blendd; 213 memset(&blendd, 0, sizeof(blendd)); 214 blendd.rt[0].colormask = 0xf; 215 default_blend = pipe->create_blend_state(pipe, &blendd); 216 217 struct pipe_sampler_state samplerd; 218 memset(&samplerd, 0, sizeof(samplerd)); 219 samplerd.normalized_coords = 1; 220 samplerd.min_img_filter = PIPE_TEX_FILTER_LINEAR; 221 samplerd.mag_img_filter = PIPE_TEX_FILTER_LINEAR; 222 samplerd.min_mip_filter = PIPE_TEX_MIPFILTER_LINEAR; 223 samplerd.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; 224 samplerd.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; 225 samplerd.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; 226 samplerd.border_color[0] = 1.0f; 227 samplerd.border_color[1] = 1.0f; 228 samplerd.border_color[2] = 1.0f; 229 samplerd.border_color[3] = 1.0f; 230 samplerd.min_lod = -FLT_MAX; 231 samplerd.max_lod = FLT_MAX; 232 samplerd.max_anisotropy = 1; 233 default_sampler = pipe->create_sampler_state(pipe, &samplerd); 234 235 memset(&samplerd, 0, sizeof(samplerd)); 236 samplerd.normalized_coords = 0; 237 samplerd.min_img_filter = PIPE_TEX_FILTER_NEAREST; 238 samplerd.mag_img_filter = PIPE_TEX_FILTER_NEAREST; 239 samplerd.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; 240 samplerd.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_BORDER; 241 samplerd.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_BORDER; 242 samplerd.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_BORDER; 243 samplerd.min_lod = -FLT_MAX; 244 samplerd.max_lod = FLT_MAX; 245 samplerd.max_anisotropy = 1; 246 ld_sampler = pipe->create_sampler_state(pipe, &samplerd); 247 248 for(unsigned s = 0; s < D3D11_STAGES; ++s) 249 { 250 sampler_csos[s].ld = ld_sampler; 251 for(unsigned i = 0; i < D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT; ++i) 252 sampler_csos[s].v[i] = default_sampler; 253 } 254 255 // TODO: should this really be empty shaders, or should they be all-passthrough? 256 memset(default_shaders, 0, sizeof(default_shaders)); 257 struct ureg_program *ureg; 258 ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT); 259 ureg_END(ureg); 260 default_shaders[PIPE_SHADER_FRAGMENT] = ureg_create_shader_and_destroy(ureg, pipe); 261 262 ureg = ureg_create(TGSI_PROCESSOR_VERTEX); 263 ureg_END(ureg); 264 default_shaders[PIPE_SHADER_VERTEX] = ureg_create_shader_and_destroy(ureg, pipe); 265 266 cso_ctx = cso_create_context(pipe); 267 gen_mipmap = util_create_gen_mipmap(pipe, cso_ctx); 268 269 RestoreGalliumState(); 270 } 271 272 void destroy_context() 273 { 274 util_destroy_gen_mipmap(gen_mipmap); 275 cso_destroy_context(cso_ctx); 276 277 pipe->bind_vertex_elements_state(pipe, 0); 278 pipe->delete_vertex_elements_state(pipe, default_input_layout); 279 280 pipe->bind_rasterizer_state(pipe, 0); 281 pipe->delete_rasterizer_state(pipe, default_rasterizer); 282 283 pipe->bind_depth_stencil_alpha_state(pipe, 0); 284 pipe->delete_depth_stencil_alpha_state(pipe, default_depth_stencil); 285 286 pipe->bind_blend_state(pipe, 0); 287 pipe->delete_blend_state(pipe, default_blend); 288 289 pipe->bind_fragment_sampler_states(pipe, 0, 0); 290 pipe->bind_vertex_sampler_states(pipe, 0, 0); 291 if(pipe->bind_geometry_sampler_states) 292 pipe->bind_geometry_sampler_states(pipe, 0, 0); 293 pipe->delete_sampler_state(pipe, default_sampler); 294 pipe->delete_sampler_state(pipe, ld_sampler); 295 296 pipe->bind_fs_state(pipe, 0); 297 pipe->delete_fs_state(pipe, default_shaders[PIPE_SHADER_FRAGMENT]); 298 299 pipe->bind_vs_state(pipe, 0); 300 pipe->delete_vs_state(pipe, default_shaders[PIPE_SHADER_VERTEX]); 301 302 if(owns_pipe) 303 pipe->destroy(pipe); 304 } 305 306 virtual unsigned STDMETHODCALLTYPE GetContextFlags(void) 307 { 308 return context_flags; 309 } 310#if API >= 11 311#define SET_SHADER_EXTRA_ARGS , \ 312 ID3D11ClassInstance *const *ppClassInstances, \ 313 unsigned count 314#define GET_SHADER_EXTRA_ARGS , \ 315 ID3D11ClassInstance **ppClassInstances, \ 316 unsigned *out_count 317#else 318#define SET_SHADER_EXTRA_ARGS 319#define GET_SHADER_EXTRA_ARGS 320#endif 321 322/* On Windows D3D11, SetConstantBuffers and SetShaderResources crash if passed a null pointer. 323 * Instead, you have to pass a pointer to nulls to unbind things. 324 * We do the same. 325 * TODO: is D3D10 the same? 326 */ 327 template<unsigned s> 328 void xs_set_shader(GalliumD3D11Shader<>* shader) 329 { 330 if(shader != shaders[s].p) 331 { 332 shaders[s] = shader; 333 void* shader_cso = shader ? shader->object : default_shaders[s]; 334 switch(s) 335 { 336 case PIPE_SHADER_VERTEX: 337 pipe->bind_vs_state(pipe, shader_cso); 338 break; 339 case PIPE_SHADER_FRAGMENT: 340 pipe->bind_fs_state(pipe, shader_cso); 341 break; 342 case PIPE_SHADER_GEOMETRY: 343 pipe->bind_gs_state(pipe, shader_cso); 344 break; 345 } 346 update_flags |= (1 << (UPDATE_SAMPLERS_SHIFT + s)) | (1 << (UPDATE_VIEWS_SHIFT + s)); 347 } 348 } 349 350 template<unsigned s> 351 void xs_set_constant_buffers(unsigned start, unsigned count, GalliumD3D11Buffer *const *constbufs) 352 { 353 for(unsigned i = 0; i < count; ++i) 354 { 355 if(constbufs[i] != constant_buffers[s][i].p) 356 { 357 constant_buffers[s][i] = constbufs[i]; 358 if(s < caps.stages && start + i < caps.constant_buffers[s]) 359 pipe->set_constant_buffer(pipe, s, start + i, constbufs[i] ? constbufs[i]->resource : NULL); 360 } 361 } 362 } 363 364 template<unsigned s> 365 void xs_set_shader_resources(unsigned start, unsigned count, GalliumD3D11ShaderResourceView *const *srvs) 366 { 367 int last_different = -1; 368 for(unsigned i = 0; i < count; ++i) 369 { 370 if(shader_resource_views[s][start + i].p != srvs[i]) 371 { 372 shader_resource_views[s][start + i] = srvs[i]; 373 sampler_views[s][start + i] = srvs[i] ? srvs[i]->object : 0; 374 last_different = i; 375 } 376 } 377 if(last_different >= 0) 378 { 379 num_shader_resource_views[s] = std::max(num_shader_resource_views[s], start + last_different + 1); 380 update_flags |= 1 << (UPDATE_VIEWS_SHIFT + s); 381 } 382 } 383 384 template<unsigned s> 385 void xs_set_samplers(unsigned start, unsigned count, GalliumD3D11SamplerState *const *samps) 386 { 387 int last_different = -1; 388 for(unsigned i = 0; i < count; ++i) 389 { 390 if(samplers[s][start + i].p != samps[i]) 391 { 392 samplers[s][start + i] = samps[i]; 393 sampler_csos[s].v[start + i] = samps[i] ? samps[i]->object : default_sampler; 394 } 395 if(last_different >= 0) 396 { 397 num_samplers[s] = std::max(num_samplers[s], start + last_different + 1); 398 update_flags |= (UPDATE_SAMPLERS_SHIFT + s); 399 } 400 } 401 } 402 403#define IMPLEMENT_SHADER_STAGE(XS, Stage) \ 404 virtual void STDMETHODCALLTYPE XS##SetShader( \ 405 ID3D11##Stage##Shader *pShader \ 406 SET_SHADER_EXTRA_ARGS) \ 407 { \ 408 SYNCHRONIZED; \ 409 xs_set_shader<D3D11_STAGE_##XS>((GalliumD3D11Shader<>*)pShader); \ 410 } \ 411 virtual void STDMETHODCALLTYPE XS##GetShader(\ 412 ID3D11##Stage##Shader **ppShader \ 413 GET_SHADER_EXTRA_ARGS) \ 414 { \ 415 SYNCHRONIZED; \ 416 *ppShader = (ID3D11##Stage##Shader*)shaders[D3D11_STAGE_##XS].ref(); \ 417 } \ 418 virtual void STDMETHODCALLTYPE XS##SetConstantBuffers(\ 419 unsigned start, \ 420 unsigned count, \ 421 ID3D11Buffer *const* constant_buffers) \ 422 { \ 423 SYNCHRONIZED; \ 424 xs_set_constant_buffers<D3D11_STAGE_##XS>(start, count, (GalliumD3D11Buffer *const *)constant_buffers); \ 425 } \ 426 virtual void STDMETHODCALLTYPE XS##GetConstantBuffers(\ 427 unsigned start, \ 428 unsigned count, \ 429 ID3D11Buffer **out_constant_buffers) \ 430 { \ 431 SYNCHRONIZED; \ 432 for(unsigned i = 0; i < count; ++i) \ 433 out_constant_buffers[i] = constant_buffers[D3D11_STAGE_##XS][start + i].ref(); \ 434 } \ 435 virtual void STDMETHODCALLTYPE XS##SetShaderResources(\ 436 unsigned start, \ 437 unsigned count, \ 438 ID3D11ShaderResourceView *const *new_shader_resource_views) \ 439 { \ 440 SYNCHRONIZED; \ 441 xs_set_shader_resources<D3D11_STAGE_##XS>(start, count, (GalliumD3D11ShaderResourceView *const *)new_shader_resource_views); \ 442 } \ 443 virtual void STDMETHODCALLTYPE XS##GetShaderResources(\ 444 unsigned start, \ 445 unsigned count, \ 446 ID3D11ShaderResourceView **out_shader_resource_views) \ 447 { \ 448 SYNCHRONIZED; \ 449 for(unsigned i = 0; i < count; ++i) \ 450 out_shader_resource_views[i] = shader_resource_views[D3D11_STAGE_##XS][start + i].ref(); \ 451 } \ 452 virtual void STDMETHODCALLTYPE XS##SetSamplers(\ 453 unsigned start, \ 454 unsigned count, \ 455 ID3D11SamplerState *const *new_samplers) \ 456 { \ 457 SYNCHRONIZED; \ 458 xs_set_samplers<D3D11_STAGE_##XS>(start, count, (GalliumD3D11SamplerState *const *)new_samplers); \ 459 } \ 460 virtual void STDMETHODCALLTYPE XS##GetSamplers( \ 461 unsigned start, \ 462 unsigned count, \ 463 ID3D11SamplerState **out_samplers) \ 464 { \ 465 SYNCHRONIZED; \ 466 for(unsigned i = 0; i < count; ++i) \ 467 out_samplers[i] = samplers[D3D11_STAGE_##XS][start + i].ref(); \ 468 } 469 470#define DO_VS(x) x 471#define DO_GS(x) do {if(caps.gs) {x;}} while(0) 472#define DO_PS(x) x 473#define DO_HS(x) 474#define DO_DS(x) 475#define DO_CS(x) 476 IMPLEMENT_SHADER_STAGE(VS, Vertex) 477 IMPLEMENT_SHADER_STAGE(GS, Geometry) 478 IMPLEMENT_SHADER_STAGE(PS, Pixel) 479 480#if API >= 11 481 IMPLEMENT_SHADER_STAGE(HS, Hull) 482 IMPLEMENT_SHADER_STAGE(DS, Domain) 483 IMPLEMENT_SHADER_STAGE(CS, Compute) 484 485 virtual void STDMETHODCALLTYPE CSSetUnorderedAccessViews( 486 unsigned start, 487 unsigned count, 488 ID3D11UnorderedAccessView *const *new_unordered_access_views, 489 const unsigned *new_uav_initial_counts) 490 { 491 SYNCHRONIZED; 492 for(unsigned i = 0; i < count; ++i) 493 cs_unordered_access_views[start + i] = new_unordered_access_views[i]; 494 } 495 496 virtual void STDMETHODCALLTYPE CSGetUnorderedAccessViews( 497 unsigned start, 498 unsigned count, 499 ID3D11UnorderedAccessView **out_unordered_access_views) 500 { 501 SYNCHRONIZED; 502 for(unsigned i = 0; i < count; ++i) 503 out_unordered_access_views[i] = cs_unordered_access_views[start + i].ref(); 504 } 505#endif 506 507 template<unsigned s> 508 void update_stage() 509 { 510 if(update_flags & (1 << (UPDATE_VIEWS_SHIFT + s))) 511 { 512 while(num_shader_resource_views[s] && !sampler_views[s][num_shader_resource_views[s] - 1]) \ 513 --num_shader_resource_views[s]; 514 if((1 << s) & caps.stages_with_sampling) 515 { 516 struct pipe_sampler_view* views_to_bind[PIPE_MAX_SAMPLERS]; 517 unsigned num_views_to_bind = shaders[s] ? shaders[s]->slot_to_resource.size() : 0; 518 for(unsigned i = 0; i < num_views_to_bind; ++i) 519 { 520 views_to_bind[i] = sampler_views[s][shaders[s]->slot_to_resource[i]]; 521 } 522 switch(s) 523 { 524 case PIPE_SHADER_VERTEX: 525 pipe->set_vertex_sampler_views(pipe, num_views_to_bind, views_to_bind); 526 break; 527 case PIPE_SHADER_FRAGMENT: 528 pipe->set_fragment_sampler_views(pipe, num_views_to_bind, views_to_bind); 529 break; 530 case PIPE_SHADER_GEOMETRY: 531 pipe->set_geometry_sampler_views(pipe, num_views_to_bind, views_to_bind); 532 break; 533 } 534 } 535 } 536 537 if(update_flags & (1 << (UPDATE_SAMPLERS_SHIFT + s))) 538 { 539 while(num_samplers[s] && !sampler_csos[s].v[num_samplers[s] - 1]) 540 --num_samplers[s]; 541 if((1 << s) & caps.stages_with_sampling) 542 { 543 void* samplers_to_bind[PIPE_MAX_SAMPLERS]; 544 unsigned num_samplers_to_bind = shaders[s] ? shaders[s]->slot_to_sampler.size() : 0; 545 for(unsigned i = 0; i < num_samplers_to_bind; ++i) 546 { 547 // index can be -1 to access sampler_csos[s].ld 548 samplers_to_bind[i] = *(sampler_csos[s].v + shaders[s]->slot_to_sampler[i]); 549 } 550 switch(s) 551 { 552 case PIPE_SHADER_VERTEX: 553 pipe->bind_vertex_sampler_states(pipe, num_samplers_to_bind, samplers_to_bind); 554 break; 555 case PIPE_SHADER_FRAGMENT: 556 pipe->bind_fragment_sampler_states(pipe, num_samplers_to_bind, samplers_to_bind); 557 break; 558 case PIPE_SHADER_GEOMETRY: 559 pipe->bind_geometry_sampler_states(pipe, num_samplers_to_bind, samplers_to_bind); 560 break; 561 } 562 } 563 } 564 } 565 566 void update_state() 567 { 568 update_stage<D3D11_STAGE_PS>(); 569 update_stage<D3D11_STAGE_VS>(); 570 update_stage<D3D11_STAGE_GS>(); 571#if API >= 11 572 update_stage<D3D11_STAGE_HS>(); 573 update_stage<D3D11_STAGE_DS>(); 574 update_stage<D3D11_STAGE_CS>(); 575#endif 576 577 if(update_flags & UPDATE_VERTEX_BUFFERS) 578 { 579 while(num_vertex_buffers && !vertex_buffers[num_vertex_buffers - 1].buffer) 580 --num_vertex_buffers; 581 pipe->set_vertex_buffers(pipe, num_vertex_buffers, vertex_buffers); 582 } 583 584 update_flags = 0; 585 } 586 587 virtual void STDMETHODCALLTYPE IASetInputLayout( 588 ID3D11InputLayout *new_input_layout) 589 { 590 SYNCHRONIZED; 591 if(new_input_layout != input_layout.p) 592 { 593 input_layout = new_input_layout; 594 pipe->bind_vertex_elements_state(pipe, new_input_layout ? ((GalliumD3D11InputLayout*)new_input_layout)->object : default_input_layout); 595 } 596 } 597 598 virtual void STDMETHODCALLTYPE IAGetInputLayout( 599 ID3D11InputLayout **out_input_layout) 600 { 601 SYNCHRONIZED; 602 *out_input_layout = input_layout.ref(); 603 } 604 605 virtual void STDMETHODCALLTYPE IASetVertexBuffers( 606 unsigned start, 607 unsigned count, 608 ID3D11Buffer *const *new_vertex_buffers, 609 const unsigned *new_strides, 610 const unsigned *new_offsets) 611 { 612 SYNCHRONIZED; 613 int last_different = -1; 614 for(unsigned i = 0; i < count; ++i) 615 { 616 ID3D11Buffer* buffer = new_vertex_buffers[i]; 617 if(buffer != input_buffers[start + i].p 618 || vertex_buffers[start + i].buffer_offset != new_offsets[i] 619 || vertex_buffers[start + i].stride != new_offsets[i] 620 ) 621 { 622 input_buffers[start + i] = buffer; 623 vertex_buffers[start + i].buffer = buffer ? ((GalliumD3D11Buffer*)buffer)->resource : 0; 624 vertex_buffers[start + i].buffer_offset = new_offsets[i]; 625 vertex_buffers[start + i].stride = new_strides[i]; 626 vertex_buffers[start + i].max_index = ~0; 627 last_different = i; 628 } 629 } 630 if(last_different >= 0) 631 { 632 num_vertex_buffers = std::max(num_vertex_buffers, start + count); 633 update_flags |= UPDATE_VERTEX_BUFFERS; 634 } 635 } 636 637 virtual void STDMETHODCALLTYPE IAGetVertexBuffers( 638 unsigned start, 639 unsigned count, 640 ID3D11Buffer **out_vertex_buffers, 641 unsigned *out_strides, 642 unsigned *out_offsets) 643 { 644 SYNCHRONIZED; 645 if(out_vertex_buffers) 646 { 647 for(unsigned i = 0; i < count; ++i) 648 out_vertex_buffers[i] = input_buffers[start + i].ref(); 649 } 650 651 if(out_offsets) 652 { 653 for(unsigned i = 0; i < count; ++i) 654 out_offsets[i] = vertex_buffers[start + i].buffer_offset; 655 } 656 657 if(out_strides) 658 { 659 for(unsigned i = 0; i < count; ++i) 660 out_strides[i] = vertex_buffers[start + i].stride; 661 } 662 } 663 664 void set_index_buffer() 665 { 666 pipe_index_buffer ib; 667 if(!index_buffer) 668 { 669 memset(&ib, 0, sizeof(ib)); 670 } 671 else 672 { 673 if(index_format == DXGI_FORMAT_R32_UINT) 674 ib.index_size = 4; 675 else if(index_format == DXGI_FORMAT_R16_UINT) 676 ib.index_size = 2; 677 else 678 ib.index_size = 1; 679 ib.offset = index_offset; 680 ib.buffer = index_buffer ? ((GalliumD3D11Buffer*)index_buffer.p)->resource : 0; 681 } 682 pipe->set_index_buffer(pipe, &ib); 683 } 684 685 virtual void STDMETHODCALLTYPE IASetIndexBuffer( 686 ID3D11Buffer *new_index_buffer, 687 DXGI_FORMAT new_index_format, 688 unsigned new_index_offset) 689 { 690 SYNCHRONIZED; 691 if(index_buffer.p != new_index_buffer || index_format != new_index_format || index_offset != new_index_offset) 692 { 693 index_buffer = new_index_buffer; 694 index_format = new_index_format; 695 index_offset = new_index_offset; 696 697 set_index_buffer(); 698 } 699 } 700 701 virtual void STDMETHODCALLTYPE IAGetIndexBuffer( 702 ID3D11Buffer **out_index_buffer, 703 DXGI_FORMAT *out_index_format, 704 unsigned *out_index_offset) 705 { 706 SYNCHRONIZED; 707 if(out_index_buffer) 708 *out_index_buffer = index_buffer.ref(); 709 if(out_index_format) 710 *out_index_format = index_format; 711 if(out_index_offset) 712 *out_index_offset = index_offset; 713 } 714 715 virtual void STDMETHODCALLTYPE IASetPrimitiveTopology( 716 D3D11_PRIMITIVE_TOPOLOGY new_primitive_topology) 717 { 718 SYNCHRONIZED; 719 if(primitive_topology != new_primitive_topology) 720 { 721 if(new_primitive_topology < D3D_PRIMITIVE_TOPOLOGY_COUNT) 722 primitive_mode = d3d_to_pipe_prim[new_primitive_topology]; 723 else 724 primitive_mode = 0; 725 primitive_topology = new_primitive_topology; 726 } 727 } 728 729 virtual void STDMETHODCALLTYPE IAGetPrimitiveTopology( 730 D3D11_PRIMITIVE_TOPOLOGY *out_primitive_topology) 731 { 732 SYNCHRONIZED; 733 *out_primitive_topology = primitive_topology; 734 } 735 736 virtual void STDMETHODCALLTYPE DrawIndexed( 737 unsigned index_count, 738 unsigned start_index_location, 739 int base_vertex_location) 740 { 741 SYNCHRONIZED; 742 if(update_flags) 743 update_state(); 744 745 pipe_draw_info info; 746 info.mode = primitive_mode; 747 info.indexed = TRUE; 748 info.count = index_count; 749 info.start = start_index_location; 750 info.index_bias = base_vertex_location; 751 info.min_index = 0; 752 info.max_index = ~0; 753 info.start_instance = 0; 754 info.instance_count = 1; 755 756 pipe->draw_vbo(pipe, &info); 757 } 758 759 virtual void STDMETHODCALLTYPE Draw( 760 unsigned vertex_count, 761 unsigned start_vertex_location) 762 { 763 SYNCHRONIZED; 764 if(update_flags) 765 update_state(); 766 767 pipe_draw_info info; 768 info.mode = primitive_mode; 769 info.indexed = FALSE; 770 info.count = vertex_count; 771 info.start = start_vertex_location; 772 info.index_bias = 0; 773 info.min_index = 0; 774 info.max_index = ~0; 775 info.start_instance = 0; 776 info.instance_count = 1; 777 778 pipe->draw_vbo(pipe, &info); 779 } 780 781 virtual void STDMETHODCALLTYPE DrawIndexedInstanced( 782 unsigned index_countPerInstance, 783 unsigned instance_count, 784 unsigned start_index_location, 785 int base_vertex_location, 786 unsigned start_instance_location) 787 { 788 SYNCHRONIZED; 789 if(update_flags) 790 update_state(); 791 792 pipe_draw_info info; 793 info.mode = primitive_mode; 794 info.indexed = TRUE; 795 info.count = index_countPerInstance; 796 info.start = start_index_location; 797 info.index_bias = base_vertex_location; 798 info.min_index = 0; 799 info.max_index = ~0; 800 info.start_instance = start_instance_location; 801 info.instance_count = instance_count; 802 803 pipe->draw_vbo(pipe, &info); 804 } 805 806 virtual void STDMETHODCALLTYPE DrawInstanced( 807 unsigned vertex_countPerInstance, 808 unsigned instance_count, 809 unsigned start_vertex_location, 810 unsigned start_instance_location) 811 { 812 SYNCHRONIZED; 813 if(update_flags) 814 update_state(); 815 816 pipe_draw_info info; 817 info.mode = primitive_mode; 818 info.indexed = FALSE; 819 info.count = vertex_countPerInstance; 820 info.start = start_vertex_location; 821 info.index_bias = 0; 822 info.min_index = 0; 823 info.max_index = ~0; 824 info.start_instance = start_instance_location; 825 info.instance_count = instance_count; 826 827 pipe->draw_vbo(pipe, &info); 828 } 829 830 virtual void STDMETHODCALLTYPE DrawAuto(void) 831 { 832 if(!caps.so) 833 return; 834 835 SYNCHRONIZED; 836 if(update_flags) 837 update_state(); 838 839 pipe->draw_stream_output(pipe, primitive_mode); 840 } 841 842 virtual void STDMETHODCALLTYPE DrawIndexedInstancedIndirect( 843 ID3D11Buffer *buffer, 844 unsigned aligned_byte_offset) 845 { 846 SYNCHRONIZED; 847 if(update_flags) 848 update_state(); 849 850 struct { 851 unsigned count; 852 unsigned instance_count; 853 unsigned start; 854 unsigned index_bias; 855 } data; 856 857 pipe_buffer_read(pipe, ((GalliumD3D11Buffer*)buffer)->resource, aligned_byte_offset, sizeof(data), &data); 858 859 pipe_draw_info info; 860 info.mode = primitive_mode; 861 info.indexed = TRUE; 862 info.start = data.start; 863 info.count = data.count; 864 info.index_bias = data.index_bias; 865 info.min_index = 0; 866 info.max_index = ~0; 867 info.start_instance = 0; 868 info.instance_count = data.instance_count; 869 870 pipe->draw_vbo(pipe, &info); 871 } 872 873 virtual void STDMETHODCALLTYPE DrawInstancedIndirect( 874 ID3D11Buffer *buffer, 875 unsigned aligned_byte_offset) 876 { 877 SYNCHRONIZED; 878 if(update_flags) 879 update_state(); 880 881 struct { 882 unsigned count; 883 unsigned instance_count; 884 unsigned start; 885 } data; 886 887 pipe_buffer_read(pipe, ((GalliumD3D11Buffer*)buffer)->resource, aligned_byte_offset, sizeof(data), &data); 888 889 pipe_draw_info info; 890 info.mode = primitive_mode; 891 info.indexed = FALSE; 892 info.start = data.start; 893 info.count = data.count; 894 info.index_bias = 0; 895 info.min_index = 0; 896 info.max_index = ~0; 897 info.start_instance = 0; 898 info.instance_count = data.instance_count; 899 900 pipe->draw_vbo(pipe, &info); 901 } 902 903#if API >= 11 904 virtual void STDMETHODCALLTYPE Dispatch( 905 unsigned thread_group_count_x, 906 unsigned thread_group_count_y, 907 unsigned thread_group_count_z) 908 { 909// uncomment this when this is implemented 910// SYNCHRONIZED; 911// if(update_flags) 912// update_state(); 913 } 914 915 virtual void STDMETHODCALLTYPE DispatchIndirect( 916 ID3D11Buffer *buffer, 917 unsigned aligned_byte_offset) 918 { 919// uncomment this when this is implemented 920// SYNCHRONIZED; 921// if(update_flags) 922// update_state(); 923 } 924#endif 925 926 void set_clip() 927 { 928 pipe_clip_state clip; 929 clip.nr = 0; 930 clip.depth_clamp = depth_clamp; 931 pipe->set_clip_state(pipe, &clip); 932 } 933 934 virtual void STDMETHODCALLTYPE RSSetState( 935 ID3D11RasterizerState *new_rasterizer_state) 936 { 937 SYNCHRONIZED; 938 if(new_rasterizer_state != rasterizer_state.p) 939 { 940 rasterizer_state = new_rasterizer_state; 941 pipe->bind_rasterizer_state(pipe, new_rasterizer_state ? ((GalliumD3D11RasterizerState*)new_rasterizer_state)->object : default_rasterizer); 942 bool new_depth_clamp = new_rasterizer_state ? ((GalliumD3D11RasterizerState*)new_rasterizer_state)->depth_clamp : false; 943 if(depth_clamp != new_depth_clamp) 944 { 945 depth_clamp = new_depth_clamp; 946 set_clip(); 947 } 948 } 949 } 950 951 virtual void STDMETHODCALLTYPE RSGetState( 952 ID3D11RasterizerState **out_rasterizer_state) 953 { 954 SYNCHRONIZED; 955 *out_rasterizer_state = rasterizer_state.ref(); 956 } 957 958 void set_viewport() 959 { 960 // TODO: is depth correct? it seems D3D10/11 uses a [-1,1]x[-1,1]x[0,1] cube 961 pipe_viewport_state viewport; 962 float half_width = viewports[0].Width * 0.5f; 963 float half_height = viewports[0].Height * 0.5f; 964 965 viewport.scale[0] = half_width; 966 viewport.scale[1] = -half_height; 967 viewport.scale[2] = (viewports[0].MaxDepth - viewports[0].MinDepth); 968 viewport.scale[3] = 1.0f; 969 viewport.translate[0] = half_width + viewports[0].TopLeftX; 970 viewport.translate[1] = half_height + viewports[0].TopLeftY; 971 viewport.translate[2] = viewports[0].MinDepth; 972 viewport.translate[3] = 1.0f; 973 pipe->set_viewport_state(pipe, &viewport); 974 } 975 976 virtual void STDMETHODCALLTYPE RSSetViewports( 977 unsigned count, 978 const D3D11_VIEWPORT *new_viewports) 979 { 980 SYNCHRONIZED; 981 if(count) 982 { 983 if(memcmp(&viewports[0], &new_viewports[0], sizeof(viewports[0]))) 984 { 985 viewports[0] = new_viewports[0]; 986 set_viewport(); 987 } 988 for(unsigned i = 1; i < count; ++i) 989 viewports[i] = new_viewports[i]; 990 } 991 else if(num_viewports) 992 { 993 // TODO: what should we do here? 994 memset(&viewports[0], 0, sizeof(viewports[0])); 995 set_viewport(); 996 } 997 num_viewports = count; 998 } 999 1000 virtual void STDMETHODCALLTYPE RSGetViewports( 1001 unsigned *out_count, 1002 D3D11_VIEWPORT *out_viewports) 1003 { 1004 SYNCHRONIZED; 1005 if(out_viewports) 1006 { 1007 unsigned i; 1008 for(i = 0; i < std::min(*out_count, num_viewports); ++i) 1009 out_viewports[i] = viewports[i]; 1010 1011 memset(out_viewports + i, 0, (*out_count - i) * sizeof(D3D11_VIEWPORT)); 1012 } 1013 1014 *out_count = num_viewports; 1015 } 1016 1017 void set_scissor() 1018 { 1019 pipe_scissor_state scissor; 1020 scissor.minx = scissor_rects[0].left; 1021 scissor.miny = scissor_rects[0].top; 1022 scissor.maxx = scissor_rects[0].right; 1023 scissor.maxy = scissor_rects[0].bottom; 1024 pipe->set_scissor_state(pipe, &scissor); 1025 } 1026 1027 virtual void STDMETHODCALLTYPE RSSetScissorRects( 1028 unsigned count, 1029 const D3D11_RECT *new_rects) 1030 { 1031 SYNCHRONIZED; 1032 if(count) 1033 { 1034 if(memcmp(&scissor_rects[0], &new_rects[0], sizeof(scissor_rects[0]))) 1035 { 1036 scissor_rects[0] = new_rects[0]; 1037 set_scissor(); 1038 } 1039 for(unsigned i = 1; i < count; ++i) 1040 scissor_rects[i] = new_rects[i]; 1041 } 1042 else if(num_scissor_rects) 1043 { 1044 // TODO: what should we do here? 1045 memset(&scissor_rects[0], 0, sizeof(scissor_rects[0])); 1046 set_scissor(); 1047 } 1048 1049 num_scissor_rects = count; 1050 } 1051 1052 virtual void STDMETHODCALLTYPE RSGetScissorRects( 1053 unsigned *out_count, 1054 D3D11_RECT *out_rects) 1055 { 1056 SYNCHRONIZED; 1057 if(out_rects) 1058 { 1059 unsigned i; 1060 for(i = 0; i < std::min(*out_count, num_scissor_rects); ++i) 1061 out_rects[i] = scissor_rects[i]; 1062 1063 memset(out_rects + i, 0, (*out_count - i) * sizeof(D3D11_RECT)); 1064 } 1065 1066 *out_count = num_scissor_rects; 1067 } 1068 1069 virtual void STDMETHODCALLTYPE OMSetBlendState( 1070 ID3D11BlendState *new_blend_state, 1071 const float new_blend_factor[4], 1072 unsigned new_sample_mask) 1073 { 1074 SYNCHRONIZED; 1075 float white[4] = {1.0f, 1.0f, 1.0f, 1.0f}; 1076 1077 if(blend_state.p != new_blend_state) 1078 { 1079 pipe->bind_blend_state(pipe, new_blend_state ? ((GalliumD3D11BlendState*)new_blend_state)->object : default_blend); 1080 blend_state = new_blend_state; 1081 } 1082 1083 // Windows D3D11 does this, even though it's apparently undocumented 1084 if(!new_blend_factor) 1085 new_blend_factor = white; 1086 1087 if(memcmp(blend_color, new_blend_factor, sizeof(blend_color))) 1088 { 1089 pipe->set_blend_color(pipe, (struct pipe_blend_color*)new_blend_factor); 1090 memcpy(blend_color, new_blend_factor, sizeof(blend_color)); 1091 } 1092 1093 if(sample_mask != new_sample_mask) 1094 { 1095 pipe->set_sample_mask(pipe, new_sample_mask); 1096 sample_mask = new_sample_mask; 1097 } 1098 } 1099 1100 virtual void STDMETHODCALLTYPE OMGetBlendState( 1101 ID3D11BlendState **out_blend_state, 1102 float out_blend_factor[4], 1103 unsigned *out_sample_mask) 1104 { 1105 SYNCHRONIZED; 1106 if(out_blend_state) 1107 *out_blend_state = blend_state.ref(); 1108 if(out_blend_factor) 1109 memcpy(out_blend_factor, blend_color, sizeof(blend_color)); 1110 if(out_sample_mask) 1111 *out_sample_mask = sample_mask; 1112 } 1113 1114 void set_stencil_ref() 1115 { 1116 struct pipe_stencil_ref sref; 1117 sref.ref_value[0] = stencil_ref; 1118 sref.ref_value[1] = stencil_ref; 1119 pipe->set_stencil_ref(pipe, &sref); 1120 } 1121 1122 virtual void STDMETHODCALLTYPE OMSetDepthStencilState( 1123 ID3D11DepthStencilState *new_depth_stencil_state, 1124 unsigned new_stencil_ref) 1125 { 1126 SYNCHRONIZED; 1127 if(new_depth_stencil_state != depth_stencil_state.p) 1128 { 1129 pipe->bind_depth_stencil_alpha_state(pipe, new_depth_stencil_state ? ((GalliumD3D11DepthStencilState*)new_depth_stencil_state)->object : default_depth_stencil); 1130 depth_stencil_state = new_depth_stencil_state; 1131 } 1132 1133 if(new_stencil_ref != stencil_ref) 1134 { 1135 stencil_ref = new_stencil_ref; 1136 set_stencil_ref(); 1137 } 1138 } 1139 1140 virtual void STDMETHODCALLTYPE OMGetDepthStencilState( 1141 ID3D11DepthStencilState **out_depth_stencil_state, 1142 unsigned *out_stencil_ref) 1143 { 1144 SYNCHRONIZED; 1145 if(*out_depth_stencil_state) 1146 *out_depth_stencil_state = depth_stencil_state.ref(); 1147 if(out_stencil_ref) 1148 *out_stencil_ref = stencil_ref; 1149 } 1150 1151 void set_framebuffer() 1152 { 1153 struct pipe_framebuffer_state fb; 1154 memset(&fb, 0, sizeof(fb)); 1155 if(depth_stencil_view) 1156 { 1157 struct pipe_surface* surf = ((GalliumD3D11DepthStencilView*)depth_stencil_view.p)->object; 1158 fb.zsbuf = surf; 1159 if(surf->width > fb.width) 1160 fb.width = surf->width; 1161 if(surf->height > fb.height) 1162 fb.height = surf->height; 1163 } 1164 fb.nr_cbufs = num_render_target_views; 1165 unsigned i; 1166 for(i = 0; i < num_render_target_views; ++i) 1167 { 1168 if(render_target_views[i]) 1169 { 1170 struct pipe_surface* surf = ((GalliumD3D11RenderTargetView*)render_target_views[i].p)->object; 1171 fb.cbufs[i] = surf; 1172 if(surf->width > fb.width) 1173 fb.width = surf->width; 1174 if(surf->height > fb.height) 1175 fb.height = surf->height; 1176 } 1177 } 1178 1179 pipe->set_framebuffer_state(pipe, &fb); 1180 } 1181 1182 /* TODO: the docs say that we should unbind conflicting resources (e.g. those bound for read while we are binding them for write too), but we aren't. 1183 * Hopefully nobody relies on this happening 1184 */ 1185 1186 virtual void STDMETHODCALLTYPE OMSetRenderTargets( 1187 unsigned count, 1188 ID3D11RenderTargetView *const *new_render_target_views, 1189 ID3D11DepthStencilView *new_depth_stencil_view) 1190 { 1191 SYNCHRONIZED; 1192 if(!new_render_target_views) 1193 count = 0; 1194 if(count == num_render_target_views) 1195 { 1196 for(unsigned i = 0; i < count; ++i) 1197 { 1198 if(new_render_target_views[i] != render_target_views[i].p) 1199 goto changed; 1200 } 1201 return; 1202 } 1203changed: 1204 depth_stencil_view = new_depth_stencil_view; 1205 unsigned i; 1206 for(i = 0; i < count; ++i) 1207 { 1208 render_target_views[i] = new_render_target_views[i]; 1209#if API >= 11 1210 om_unordered_access_views[i] = (ID3D11UnorderedAccessView*)NULL; 1211#endif 1212 } 1213 for(; i < num_render_target_views; ++i) 1214 render_target_views[i] = (ID3D11RenderTargetView*)NULL; 1215 num_render_target_views = count; 1216 set_framebuffer(); 1217 } 1218 1219 virtual void STDMETHODCALLTYPE OMGetRenderTargets( 1220 unsigned count, 1221 ID3D11RenderTargetView **out_render_target_views, 1222 ID3D11DepthStencilView **out_depth_stencil_view) 1223 { 1224 SYNCHRONIZED; 1225 if(out_render_target_views) 1226 { 1227 unsigned i; 1228 for(i = 0; i < std::min(num_render_target_views, count); ++i) 1229 out_render_target_views[i] = render_target_views[i].ref(); 1230 1231 for(; i < count; ++i) 1232 out_render_target_views[i] = 0; 1233 } 1234 1235 if(out_depth_stencil_view) 1236 *out_depth_stencil_view = depth_stencil_view.ref(); 1237 } 1238 1239#if API >= 11 1240 /* TODO: what is this supposed to do _exactly_? are we doing the right thing? */ 1241 virtual void STDMETHODCALLTYPE OMSetRenderTargetsAndUnorderedAccessViews( 1242 unsigned rtv_count, 1243 ID3D11RenderTargetView *const *new_render_target_views, 1244 ID3D11DepthStencilView *new_depth_stencil_view, 1245 unsigned uav_start, 1246 unsigned uav_count, 1247 ID3D11UnorderedAccessView *const *new_unordered_access_views, 1248 const unsigned *new_uav_initial_counts) 1249 { 1250 SYNCHRONIZED; 1251 if(rtv_count != D3D11_KEEP_RENDER_TARGETS_AND_DEPTH_STENCIL) 1252 OMSetRenderTargets(rtv_count, new_render_target_views, new_depth_stencil_view); 1253 1254 if(uav_count != D3D11_KEEP_UNORDERED_ACCESS_VIEWS) 1255 { 1256 for(unsigned i = 0; i < uav_count; ++i) 1257 { 1258 om_unordered_access_views[uav_start + i] = new_unordered_access_views[i]; 1259 render_target_views[uav_start + i] = (ID3D11RenderTargetView*)0; 1260 } 1261 } 1262 } 1263 1264 virtual void STDMETHODCALLTYPE OMGetRenderTargetsAndUnorderedAccessViews( 1265 unsigned rtv_count, 1266 ID3D11RenderTargetView **out_render_target_views, 1267 ID3D11DepthStencilView **out_depth_stencil_view, 1268 unsigned uav_start, 1269 unsigned uav_count, 1270 ID3D11UnorderedAccessView **out_unordered_access_views) 1271 { 1272 SYNCHRONIZED; 1273 if(out_render_target_views) 1274 OMGetRenderTargets(rtv_count, out_render_target_views, out_depth_stencil_view); 1275 1276 if(out_unordered_access_views) 1277 { 1278 for(unsigned i = 0; i < uav_count; ++i) 1279 out_unordered_access_views[i] = om_unordered_access_views[uav_start + i].ref(); 1280 } 1281 } 1282#endif 1283 1284 virtual void STDMETHODCALLTYPE SOSetTargets( 1285 unsigned count, 1286 ID3D11Buffer *const *new_so_targets, 1287 const unsigned *new_offsets) 1288 { 1289 SYNCHRONIZED; 1290 unsigned i; 1291 if(!new_so_targets) 1292 count = 0; 1293 bool changed = false; 1294 for(i = 0; i < count; ++i) 1295 { 1296 ID3D11Buffer* buffer = new_so_targets[i]; 1297 if(buffer != so_targets[i].p || new_offsets[i] != so_offsets[i]) 1298 { 1299 so_buffers[i] = buffer ? ((GalliumD3D11Buffer*)buffer)->resource : 0; 1300 so_targets[i] = buffer; 1301 so_offsets[i] = new_offsets[i]; 1302 changed = true; 1303 } 1304 } 1305 for(; i < D3D11_SO_BUFFER_SLOT_COUNT; ++i) 1306 { 1307 if(so_targets[i].p || so_offsets[i]) 1308 { 1309 changed = true; 1310 so_targets[i] = (ID3D11Buffer*)0; 1311 so_offsets[i] = 0; 1312 } 1313 } 1314 num_so_targets = count; 1315 1316 if(changed && caps.so) 1317 pipe->set_stream_output_buffers(pipe, so_buffers, (int*)so_offsets, num_so_targets); 1318 } 1319 1320 virtual void STDMETHODCALLTYPE SOGetTargets( 1321 unsigned count, 1322 ID3D11Buffer **out_so_targets 1323#if API < 11 1324 , UINT *out_offsets 1325#endif 1326 ) 1327 { 1328 SYNCHRONIZED; 1329 for(unsigned i = 0; i < count; ++i) 1330 { 1331 out_so_targets[i] = so_targets[i].ref(); 1332#if API < 11 1333 out_offsets[i] = so_offsets[i]; 1334#endif 1335 } 1336 } 1337 1338 virtual void STDMETHODCALLTYPE Begin( 1339 ID3D11Asynchronous *async) 1340 { 1341 SYNCHRONIZED; 1342 if(caps.queries) 1343 pipe->begin_query(pipe, ((GalliumD3D11Asynchronous<>*)async)->query); 1344 } 1345 1346 virtual void STDMETHODCALLTYPE End( 1347 ID3D11Asynchronous *async) 1348 { 1349 SYNCHRONIZED; 1350 if(caps.queries) 1351 pipe->end_query(pipe, ((GalliumD3D11Asynchronous<>*)async)->query); 1352 } 1353 1354 virtual HRESULT STDMETHODCALLTYPE GetData( 1355 ID3D11Asynchronous *iasync, 1356 void *out_data, 1357 unsigned data_size, 1358 unsigned get_data_flags) 1359 { 1360 SYNCHRONIZED; 1361 if(!caps.queries) 1362 return E_NOTIMPL; 1363 1364 GalliumD3D11Asynchronous<>* async = (GalliumD3D11Asynchronous<>*)iasync; 1365 void* tmp_data = alloca(async->data_size); 1366 boolean ret = pipe->get_query_result(pipe, async->query, !(get_data_flags & D3D11_ASYNC_GETDATA_DONOTFLUSH), tmp_data); 1367 if(out_data) 1368 memcpy(out_data, tmp_data, std::min(async->data_size, data_size)); 1369 return ret ? S_OK : S_FALSE; 1370 } 1371 1372 void set_render_condition() 1373 { 1374 if(caps.render_condition) 1375 { 1376 if(!render_predicate) 1377 pipe->render_condition(pipe, 0, 0); 1378 else 1379 { 1380 GalliumD3D11Predicate* predicate = (GalliumD3D11Predicate*)render_predicate.p; 1381 if(!render_predicate_value && predicate->desc.Query == D3D11_QUERY_OCCLUSION_PREDICATE) 1382 { 1383 unsigned mode = (predicate->desc.MiscFlags & D3D11_QUERY_MISC_PREDICATEHINT) ? PIPE_RENDER_COND_NO_WAIT : PIPE_RENDER_COND_WAIT; 1384 pipe->render_condition(pipe, predicate->query, mode); 1385 } 1386 else 1387 { 1388 /* TODO: add inverted predication to Gallium*/ 1389 pipe->render_condition(pipe, 0, 0); 1390 } 1391 } 1392 } 1393 } 1394 1395 virtual void STDMETHODCALLTYPE SetPredication( 1396 ID3D11Predicate *new_predicate, 1397 BOOL new_predicate_value) 1398 { 1399 SYNCHRONIZED; 1400 if(render_predicate.p != new_predicate || render_predicate_value != new_predicate_value) 1401 { 1402 render_predicate = new_predicate; 1403 render_predicate_value = new_predicate_value; 1404 set_render_condition(); 1405 } 1406 } 1407 1408 virtual void STDMETHODCALLTYPE GetPredication( 1409 ID3D11Predicate **out_predicate, 1410 BOOL *out_predicate_value) 1411 { 1412 SYNCHRONIZED; 1413 if(out_predicate) 1414 *out_predicate = render_predicate.ref(); 1415 if(out_predicate_value) 1416 *out_predicate_value = render_predicate_value; 1417 } 1418 1419 static pipe_subresource d3d11_to_pipe_subresource(struct pipe_resource* resource, unsigned subresource) 1420 { 1421 pipe_subresource sr; 1422 if(subresource <= resource->last_level) 1423 { 1424 sr.level = subresource; 1425 sr.face = 0; 1426 } 1427 else 1428 { 1429 unsigned levels = resource->last_level + 1; 1430 sr.level = subresource % levels; 1431 sr.face = subresource / levels; 1432 } 1433 return sr; 1434 } 1435 1436 /* TODO: deferred contexts will need a different implementation of this, 1437 * because we can't put the transfer info into the resource itself. 1438 * Also, there are very different restrictions, for obvious reasons. 1439 */ 1440 virtual HRESULT STDMETHODCALLTYPE Map( 1441 ID3D11Resource *iresource, 1442 unsigned subresource, 1443 D3D11_MAP map_type, 1444 unsigned map_flags, 1445 D3D11_MAPPED_SUBRESOURCE *mapped_resource) 1446 { 1447 SYNCHRONIZED; 1448 GalliumD3D11Resource<>* resource = (GalliumD3D11Resource<>*)iresource; 1449 if(resource->transfers.count(subresource)) 1450 return E_FAIL; 1451 pipe_subresource sr = d3d11_to_pipe_subresource(resource->resource, subresource); 1452 pipe_box box = d3d11_to_pipe_box(resource->resource, sr.level, 0); 1453 unsigned usage = 0; 1454 if(map_type == D3D11_MAP_READ) 1455 usage = PIPE_TRANSFER_READ; 1456 else if(map_type == D3D11_MAP_WRITE) 1457 usage = PIPE_TRANSFER_WRITE; 1458 else if(map_type == D3D11_MAP_READ_WRITE) 1459 usage = PIPE_TRANSFER_READ_WRITE; 1460 else if(map_type == D3D11_MAP_WRITE_DISCARD) 1461 usage = PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD; 1462 else if(map_type == D3D11_MAP_WRITE_NO_OVERWRITE) 1463 usage = PIPE_TRANSFER_WRITE | PIPE_TRANSFER_NOOVERWRITE; 1464 else 1465 return E_INVALIDARG; 1466 if(map_type & D3D10_MAP_FLAG_DO_NOT_WAIT) 1467 usage |= PIPE_TRANSFER_DONTBLOCK; 1468 struct pipe_transfer* transfer = pipe->get_transfer(pipe, resource->resource, sr, usage, &box); 1469 if(!transfer) { 1470 if(map_type & D3D10_MAP_FLAG_DO_NOT_WAIT) 1471 return DXGI_ERROR_WAS_STILL_DRAWING; 1472 else 1473 return E_FAIL; 1474 } 1475 resource->transfers[subresource] = transfer; 1476 mapped_resource->pData = pipe->transfer_map(pipe, transfer); 1477 mapped_resource->RowPitch = transfer->stride; 1478 mapped_resource->DepthPitch = transfer->slice_stride; 1479 return S_OK; 1480 } 1481 1482 virtual void STDMETHODCALLTYPE Unmap( 1483 ID3D11Resource *iresource, 1484 unsigned subresource) 1485 { 1486 SYNCHRONIZED; 1487 GalliumD3D11Resource<>* resource = (GalliumD3D11Resource<>*)iresource; 1488 std::unordered_map<unsigned, pipe_transfer*>::iterator i = resource->transfers.find(subresource); 1489 if(i != resource->transfers.end()) 1490 { 1491 pipe->transfer_unmap(pipe, i->second); 1492 pipe->transfer_destroy(pipe, i->second); 1493 resource->transfers.erase(i); 1494 } 1495 } 1496 1497 virtual void STDMETHODCALLTYPE CopySubresourceRegion( 1498 ID3D11Resource *dst_resource, 1499 unsigned dst_subresource, 1500 unsigned dst_x, 1501 unsigned dst_y, 1502 unsigned dst_z, 1503 ID3D11Resource *src_resource, 1504 unsigned src_subresource, 1505 const D3D11_BOX *src_box) 1506 { 1507 SYNCHRONIZED; 1508 GalliumD3D11Resource<>* dst = (GalliumD3D11Resource<>*)dst_resource; 1509 GalliumD3D11Resource<>* src = (GalliumD3D11Resource<>*)src_resource; 1510 pipe_subresource subdst = d3d11_to_pipe_subresource(dst->resource, dst_subresource); 1511 pipe_subresource subsrc = d3d11_to_pipe_subresource(src->resource, src_subresource); 1512 pipe_box box = d3d11_to_pipe_box(src->resource, subsrc.level, src_box); 1513 for(unsigned i = 0; i < box.depth; ++i) 1514 { 1515 pipe->resource_copy_region(pipe, 1516 dst->resource, subdst, dst_x, dst_y, dst_z + i, 1517 src->resource, subsrc, box.x, box.y, box.z + i, 1518 box.width, box.height); 1519 } 1520 } 1521 1522 virtual void STDMETHODCALLTYPE CopyResource( 1523 ID3D11Resource *dst_resource, 1524 ID3D11Resource *src_resource) 1525 { 1526 SYNCHRONIZED; 1527 GalliumD3D11Resource<>* dst = (GalliumD3D11Resource<>*)dst_resource; 1528 GalliumD3D11Resource<>* src = (GalliumD3D11Resource<>*)src_resource; 1529 pipe_subresource sr; 1530 unsigned faces = dst->resource->target == PIPE_TEXTURE_CUBE ? 6 : 1; 1531 1532 for(sr.face = 0; sr.face < faces; ++sr.face) 1533 { 1534 for(sr.level = 0; sr.level <= dst->resource->last_level; ++sr.level) 1535 { 1536 unsigned w = u_minify(dst->resource->width0, sr.level); 1537 unsigned h = u_minify(dst->resource->height0, sr.level); 1538 unsigned d = u_minify(dst->resource->depth0, sr.level); 1539 for(unsigned i = 0; i < d; ++i) 1540 { 1541 pipe->resource_copy_region(pipe, 1542 dst->resource, sr, 0, 0, i, 1543 src->resource, sr, 0, 0, i, 1544 w, h); 1545 } 1546 } 1547 } 1548 } 1549 1550 virtual void STDMETHODCALLTYPE UpdateSubresource( 1551 ID3D11Resource *dst_resource, 1552 unsigned dst_subresource, 1553 const D3D11_BOX *pDstBox, 1554 const void *pSrcData, 1555 unsigned src_row_pitch, 1556 unsigned src_depth_pitch) 1557 { 1558 SYNCHRONIZED; 1559 GalliumD3D11Resource<>* dst = (GalliumD3D11Resource<>*)dst_resource; 1560 pipe_subresource subdst = d3d11_to_pipe_subresource(dst->resource, dst_subresource); 1561 pipe_box box = d3d11_to_pipe_box(dst->resource, subdst.level, pDstBox); 1562 pipe->transfer_inline_write(pipe, dst->resource, subdst, PIPE_TRANSFER_WRITE, &box, pSrcData, src_row_pitch, src_depth_pitch); 1563 } 1564 1565#if API >= 11 1566 virtual void STDMETHODCALLTYPE CopyStructureCount( 1567 ID3D11Buffer *dst_buffer, 1568 unsigned dst_aligned_byte_offset, 1569 ID3D11UnorderedAccessView *src_view) 1570 { 1571 SYNCHRONIZED; 1572 } 1573#endif 1574 1575 virtual void STDMETHODCALLTYPE ClearRenderTargetView( 1576 ID3D11RenderTargetView *render_target_view, 1577 const float color[4]) 1578 { 1579 SYNCHRONIZED; 1580 GalliumD3D11RenderTargetView* view = ((GalliumD3D11RenderTargetView*)render_target_view); 1581 pipe->clear_render_target(pipe, view->object, color, 0, 0, view->object->width, view->object->height); 1582 } 1583 1584 virtual void STDMETHODCALLTYPE ClearDepthStencilView( 1585 ID3D11DepthStencilView *depth_stencil_view, 1586 unsigned clear_flags, 1587 float depth, 1588 UINT8 stencil) 1589 { 1590 SYNCHRONIZED; 1591 GalliumD3D11DepthStencilView* view = ((GalliumD3D11DepthStencilView*)depth_stencil_view); 1592 unsigned flags = 0; 1593 if(clear_flags & D3D11_CLEAR_DEPTH) 1594 flags |= PIPE_CLEAR_DEPTH; 1595 if(clear_flags & D3D11_CLEAR_STENCIL) 1596 flags |= PIPE_CLEAR_STENCIL; 1597 pipe->clear_depth_stencil(pipe, view->object, flags, depth, stencil, 0, 0, view->object->width, view->object->height); 1598 } 1599 1600#if API >= 11 1601 virtual void STDMETHODCALLTYPE ClearUnorderedAccessViewUint( 1602 ID3D11UnorderedAccessView *unordered_access_view, 1603 const unsigned values[4]) 1604 { 1605 SYNCHRONIZED; 1606 } 1607 1608 virtual void STDMETHODCALLTYPE ClearUnorderedAccessViewFloat( 1609 ID3D11UnorderedAccessView *unordered_access_view, 1610 const float values[4]) 1611 { 1612 SYNCHRONIZED; 1613 } 1614#endif 1615 1616 void restore_gallium_state_blit_only() 1617 { 1618 pipe->bind_blend_state(pipe, blend_state.p ? blend_state.p->object : default_blend); 1619 pipe->bind_depth_stencil_alpha_state(pipe, depth_stencil_state.p ? depth_stencil_state.p->object : default_depth_stencil); 1620 pipe->bind_rasterizer_state(pipe, rasterizer_state.p ? rasterizer_state.p->object : default_rasterizer); 1621 pipe->bind_vertex_elements_state(pipe, input_layout.p ? input_layout.p->object : default_input_layout); 1622 pipe->bind_fs_state(pipe, shaders[D3D11_STAGE_PS].p ? shaders[D3D11_STAGE_PS].p->object : default_shaders[PIPE_SHADER_FRAGMENT]); 1623 pipe->bind_vs_state(pipe, shaders[D3D11_STAGE_VS].p ? shaders[D3D11_STAGE_VS].p->object : default_shaders[PIPE_SHADER_VERTEX]); 1624 if(caps.gs) 1625 pipe->bind_gs_state(pipe, shaders[D3D11_STAGE_GS].p ? shaders[D3D11_STAGE_GS].p->object : default_shaders[PIPE_SHADER_GEOMETRY]); 1626 set_framebuffer(); 1627 set_viewport(); 1628 set_clip(); 1629 set_render_condition(); 1630 // TODO: restore stream output 1631 1632 update_flags |= UPDATE_VERTEX_BUFFERS | (1 << (UPDATE_SAMPLERS_SHIFT + D3D11_STAGE_PS)) | (1 << (UPDATE_VIEWS_SHIFT + D3D11_STAGE_PS)); 1633 } 1634 1635 virtual void STDMETHODCALLTYPE RestoreGalliumStateBlitOnly() 1636 { 1637 SYNCHRONIZED; 1638 restore_gallium_state_blit_only(); 1639 } 1640 1641 virtual void STDMETHODCALLTYPE GenerateMips( 1642 ID3D11ShaderResourceView *shader_resource_view) 1643 { 1644 SYNCHRONIZED; 1645 1646 GalliumD3D11ShaderResourceView* view = (GalliumD3D11ShaderResourceView*)shader_resource_view; 1647 if(caps.gs) 1648 pipe->bind_gs_state(pipe, 0); 1649 if(caps.so) 1650 pipe->bind_stream_output_state(pipe, 0); 1651 if(pipe->render_condition) 1652 pipe->render_condition(pipe, 0, 0); 1653 util_gen_mipmap(gen_mipmap, view->object, 0, 0, view->object->texture->last_level, PIPE_TEX_FILTER_LINEAR); 1654 restore_gallium_state_blit_only(); 1655 } 1656 1657 virtual void STDMETHODCALLTYPE RestoreGalliumState() 1658 { 1659 SYNCHRONIZED; 1660 restore_gallium_state_blit_only(); 1661 1662 set_index_buffer(); 1663 set_stencil_ref(); 1664 pipe->set_blend_color(pipe, (struct pipe_blend_color*)blend_color); 1665 pipe->set_sample_mask(pipe, sample_mask); 1666 1667 for(unsigned s = 0; s < 3; ++s) 1668 { 1669 unsigned num = std::min(caps.constant_buffers[s], (unsigned)D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT); 1670 for(unsigned i = 0; i < num; ++i) 1671 pipe->set_constant_buffer(pipe, s, i, constant_buffers[s][i].p ? constant_buffers[s][i].p->resource : 0); 1672 } 1673 1674 if(caps.so) 1675 pipe->set_stream_output_buffers(pipe, so_buffers, (int*)so_offsets, num_so_targets); 1676 1677 update_flags |= (1 << (UPDATE_SAMPLERS_SHIFT + D3D11_STAGE_VS)) | (1 << (UPDATE_VIEWS_SHIFT + D3D11_STAGE_VS)); 1678 update_flags |= (1 << (UPDATE_SAMPLERS_SHIFT + D3D11_STAGE_GS)) | (1 << (UPDATE_VIEWS_SHIFT + D3D11_STAGE_GS)); 1679 1680 set_scissor(); 1681 } 1682 1683#if API >= 11 1684 /* TODO: hack SRVs or sampler states to handle this, or add to Gallium */ 1685 virtual void STDMETHODCALLTYPE SetResourceMinLOD( 1686 ID3D11Resource *iresource, 1687 float min_lod) 1688 { 1689 SYNCHRONIZED; 1690 GalliumD3D11Resource<>* resource = (GalliumD3D11Resource<>*)iresource; 1691 if(resource->min_lod != min_lod) 1692 { 1693 // TODO: actually do anything? 1694 resource->min_lod = min_lod; 1695 } 1696 } 1697 1698 virtual float STDMETHODCALLTYPE GetResourceMinLOD( 1699 ID3D11Resource *iresource) 1700 { 1701 SYNCHRONIZED; 1702 GalliumD3D11Resource<>* resource = (GalliumD3D11Resource<>*)iresource; 1703 return resource->min_lod; 1704 } 1705#endif 1706 1707 virtual void STDMETHODCALLTYPE ResolveSubresource( 1708 ID3D11Resource *dst_resource, 1709 unsigned dst_subresource, 1710 ID3D11Resource *src_resource, 1711 unsigned src_subresource, 1712 DXGI_FORMAT format) 1713 { 1714 SYNCHRONIZED; 1715 GalliumD3D11Resource<>* dst = (GalliumD3D11Resource<>*)dst_resource; 1716 GalliumD3D11Resource<>* src = (GalliumD3D11Resource<>*)src_resource; 1717 pipe_subresource subdst = d3d11_to_pipe_subresource(dst->resource, dst_subresource); 1718 pipe_subresource subsrc = d3d11_to_pipe_subresource(src->resource, src_subresource); 1719 pipe->resource_resolve(pipe, dst->resource, subdst, src->resource, subsrc); 1720 } 1721 1722#if API >= 11 1723 virtual void STDMETHODCALLTYPE ExecuteCommandList( 1724 ID3D11CommandList *command_list, 1725 BOOL restore_context_state) 1726 { 1727 SYNCHRONIZED; 1728 } 1729 1730 virtual HRESULT STDMETHODCALLTYPE FinishCommandList( 1731 BOOL restore_deferred_context_state, 1732 ID3D11CommandList **out_command_list) 1733 { 1734 SYNCHRONIZED; 1735 return E_NOTIMPL; 1736 } 1737#endif 1738 1739 virtual void STDMETHODCALLTYPE ClearState(void) 1740 { 1741 /* we don't take a lock here because we would deadlock otherwise 1742 * TODO: this is probably incorrect, because ClearState should likely be atomic. 1743 * However, I can't think of any correct usage that would be affected by this 1744 * being non-atomic, and making this atomic is quite expensive and complicates 1745 * the code 1746 */ 1747 1748 // we qualify all calls so that we avoid virtual dispatch and might get them inlined 1749 // TODO: make sure all this gets inlined, which might require more compiler flags 1750 // TODO: optimize this 1751#if API >= 11 1752 GalliumD3D11DeviceContext::PSSetShader(0, 0, 0); 1753 GalliumD3D11DeviceContext::GSSetShader(0, 0, 0); 1754 GalliumD3D11DeviceContext::VSSetShader(0, 0, 0); 1755 GalliumD3D11DeviceContext::HSSetShader(0, 0, 0); 1756 GalliumD3D11DeviceContext::DSSetShader(0, 0, 0); 1757 GalliumD3D11DeviceContext::CSSetShader(0, 0, 0); 1758#else 1759 GalliumD3D11DeviceContext::PSSetShader(0); 1760 GalliumD3D11DeviceContext::GSSetShader(0); 1761 GalliumD3D11DeviceContext::VSSetShader(0); 1762#endif 1763 1764 GalliumD3D11DeviceContext::IASetInputLayout(0); 1765 GalliumD3D11DeviceContext::IASetIndexBuffer(0, DXGI_FORMAT_UNKNOWN, 0); 1766 GalliumD3D11DeviceContext::RSSetState(0); 1767 GalliumD3D11DeviceContext::OMSetDepthStencilState(0, 0); 1768 GalliumD3D11DeviceContext::OMSetBlendState(0, (float*)zero_data, ~0); 1769 GalliumD3D11DeviceContext::SetPredication(0, 0); 1770 GalliumD3D11DeviceContext::IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_UNDEFINED); 1771 1772 GalliumD3D11DeviceContext::PSSetConstantBuffers(0, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, (ID3D11Buffer**)zero_data); 1773 GalliumD3D11DeviceContext::GSSetConstantBuffers(0, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, (ID3D11Buffer**)zero_data); 1774 GalliumD3D11DeviceContext::VSSetConstantBuffers(0, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, (ID3D11Buffer**)zero_data); 1775#if API >= 11 1776 GalliumD3D11DeviceContext::HSSetConstantBuffers(0, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, (ID3D11Buffer**)zero_data); 1777 GalliumD3D11DeviceContext::DSSetConstantBuffers(0, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, (ID3D11Buffer**)zero_data); 1778 GalliumD3D11DeviceContext::CSSetConstantBuffers(0, D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, (ID3D11Buffer**)zero_data); 1779#endif 1780 1781 GalliumD3D11DeviceContext::IASetVertexBuffers(0, num_vertex_buffers, (ID3D11Buffer**)zero_data, (unsigned*)zero_data, (unsigned*)zero_data); 1782#if API >= 11 1783 GalliumD3D11DeviceContext::OMSetRenderTargetsAndUnorderedAccessViews(0, 0, 0 , 0, 0, 0, 0); 1784#else 1785 GalliumD3D11DeviceContext::OMSetRenderTargets(0, 0, 0 ); 1786#endif 1787 GalliumD3D11DeviceContext::SOSetTargets(0, 0, 0); 1788 1789 GalliumD3D11DeviceContext::PSSetShaderResources(0, num_shader_resource_views[D3D11_STAGE_PS], (ID3D11ShaderResourceView**)zero_data); 1790 GalliumD3D11DeviceContext::GSSetShaderResources(0, num_shader_resource_views[D3D11_STAGE_GS], (ID3D11ShaderResourceView**)zero_data); 1791 GalliumD3D11DeviceContext::VSSetShaderResources(0, num_shader_resource_views[D3D11_STAGE_VS], (ID3D11ShaderResourceView**)zero_data); 1792#if API >= 11 1793 GalliumD3D11DeviceContext::HSSetShaderResources(0, num_shader_resource_views[D3D11_STAGE_HS], (ID3D11ShaderResourceView**)zero_data); 1794 GalliumD3D11DeviceContext::DSSetShaderResources(0, num_shader_resource_views[D3D11_STAGE_DS], (ID3D11ShaderResourceView**)zero_data); 1795 GalliumD3D11DeviceContext::CSSetShaderResources(0, num_shader_resource_views[D3D11_STAGE_CS], (ID3D11ShaderResourceView**)zero_data); 1796#endif 1797 1798 GalliumD3D11DeviceContext::PSSetSamplers(0, num_shader_resource_views[D3D11_STAGE_PS], (ID3D11SamplerState**)zero_data); 1799 GalliumD3D11DeviceContext::GSSetSamplers(0, num_shader_resource_views[D3D11_STAGE_GS], (ID3D11SamplerState**)zero_data); 1800 GalliumD3D11DeviceContext::VSSetSamplers(0, num_shader_resource_views[D3D11_STAGE_VS], (ID3D11SamplerState**)zero_data); 1801#if API >= 11 1802 GalliumD3D11DeviceContext::HSSetSamplers(0, num_shader_resource_views[D3D11_STAGE_HS], (ID3D11SamplerState**)zero_data); 1803 GalliumD3D11DeviceContext::DSSetSamplers(0, num_shader_resource_views[D3D11_STAGE_DS], (ID3D11SamplerState**)zero_data); 1804 GalliumD3D11DeviceContext::CSSetSamplers(0, num_shader_resource_views[D3D11_STAGE_CS], (ID3D11SamplerState**)zero_data); 1805#endif 1806 1807 GalliumD3D11DeviceContext::RSSetViewports(0, 0); 1808 GalliumD3D11DeviceContext::RSSetScissorRects(0, 0); 1809 } 1810 1811 virtual void STDMETHODCALLTYPE Flush(void) 1812 { 1813 SYNCHRONIZED; 1814 pipe->flush(pipe, PIPE_FLUSH_FRAME, 0); 1815 } 1816 1817 /* In Direct3D 10, if the reference count of an object drops to 0, it is automatically 1818 * cleanly unbound from the pipeline. 1819 * In Direct3D 11, the pipeline holds a reference. 1820 * 1821 * Note that instead of always scanning the pipeline on destruction, we could 1822 * maintain the internal reference count on DirectX 10 and use it to check if an 1823 * object is still bound. 1824 * Presumably, on average, scanning is faster if the application is well written. 1825 */ 1826#if API < 11 1827#define IMPLEMENT_SIMPLE_UNBIND(name, member, gallium, def) \ 1828 void Unbind##name(ID3D11##name* state) \ 1829 { \ 1830 SYNCHRONIZED; \ 1831 if((void*)state == (void*)member.p) \ 1832 { \ 1833 member.p = 0; \ 1834 pipe->bind_##gallium##_state(pipe, default_##def); \ 1835 } \ 1836 } 1837 IMPLEMENT_SIMPLE_UNBIND(BlendState, blend_state, blend, blend) 1838 IMPLEMENT_SIMPLE_UNBIND(RasterizerState, rasterizer_state, rasterizer, rasterizer) 1839 IMPLEMENT_SIMPLE_UNBIND(DepthStencilState, depth_stencil_state, depth_stencil_alpha, depth_stencil) 1840 IMPLEMENT_SIMPLE_UNBIND(InputLayout, input_layout, vertex_elements, input_layout) 1841 IMPLEMENT_SIMPLE_UNBIND(PixelShader, shaders[D3D11_STAGE_PS], fs, shaders[D3D11_STAGE_PS]) 1842 IMPLEMENT_SIMPLE_UNBIND(VertexShader, shaders[D3D11_STAGE_VS], vs, shaders[D3D11_STAGE_VS]) 1843 IMPLEMENT_SIMPLE_UNBIND(GeometryShader, shaders[D3D11_STAGE_GS], gs, shaders[D3D11_STAGE_GS]) 1844 1845 void UnbindPredicate(ID3D11Predicate* predicate) 1846 { 1847 SYNCHRONIZED; 1848 if(predicate == render_predicate) 1849 { 1850 render_predicate.p = NULL; 1851 render_predicate_value = 0; 1852 pipe->render_condition(pipe, 0, 0); 1853 } 1854 } 1855 1856 void UnbindSamplerState(ID3D11SamplerState* state) 1857 { 1858 SYNCHRONIZED; 1859 for(unsigned s = 0; s < D3D11_STAGES; ++s) 1860 { 1861 for(unsigned i = 0; i < num_samplers[s]; ++i) 1862 { 1863 if(samplers[s][i] == state) 1864 { 1865 samplers[s][i].p = NULL; 1866 sampler_csos[s].v[i] = NULL; 1867 update_flags |= (1 << (UPDATE_SAMPLERS_SHIFT + s)); 1868 } 1869 } 1870 } 1871 } 1872 1873 void UnbindBuffer(ID3D11Buffer* buffer) 1874 { 1875 SYNCHRONIZED; 1876 if(buffer == index_buffer) 1877 { 1878 index_buffer.p = 0; 1879 index_format = DXGI_FORMAT_UNKNOWN; 1880 index_offset = 0; 1881 struct pipe_index_buffer ib; 1882 memset(&ib, 0, sizeof(ib)); 1883 pipe->set_index_buffer(pipe, &ib); 1884 } 1885 1886 for(unsigned i = 0; i < num_vertex_buffers; ++i) 1887 { 1888 if(buffer == input_buffers[i]) 1889 { 1890 input_buffers[i].p = 0; 1891 memset(&vertex_buffers[num_vertex_buffers], 0, sizeof(vertex_buffers[num_vertex_buffers])); 1892 update_flags |= UPDATE_VERTEX_BUFFERS; 1893 } 1894 } 1895 1896 for(unsigned s = 0; s < D3D11_STAGES; ++s) 1897 { 1898 for(unsigned i = 0; i < sizeof(constant_buffers) / sizeof(constant_buffers[0]); ++i) 1899 { 1900 if(constant_buffers[s][i] == buffer) 1901 { 1902 constant_buffers[s][i] = (ID3D10Buffer*)NULL; 1903 pipe->set_constant_buffer(pipe, s, i, NULL); 1904 } 1905 } 1906 } 1907 } 1908 1909 void UnbindDepthStencilView(ID3D11DepthStencilView * view) 1910 { 1911 SYNCHRONIZED; 1912 if(view == depth_stencil_view) 1913 { 1914 depth_stencil_view.p = NULL; 1915 set_framebuffer(); 1916 } 1917 } 1918 1919 void UnbindRenderTargetView(ID3D11RenderTargetView* view) 1920 { 1921 SYNCHRONIZED; 1922 bool any_bound = false; 1923 for(unsigned i = 0; i < num_render_target_views; ++i) 1924 { 1925 if(render_target_views[i] == view) 1926 { 1927 render_target_views[i].p = NULL; 1928 any_bound = true; 1929 } 1930 } 1931 if(any_bound) 1932 set_framebuffer(); 1933 } 1934 1935 void UnbindShaderResourceView(ID3D11ShaderResourceView* view) 1936 { 1937 SYNCHRONIZED; 1938 for(unsigned s = 0; s < D3D11_STAGES; ++s) 1939 { 1940 for(unsigned i = 0; i < num_shader_resource_views[s]; ++i) 1941 { 1942 if(shader_resource_views[s][i] == view) 1943 { 1944 shader_resource_views[s][i].p = NULL; 1945 sampler_views[s][i] = NULL; 1946 update_flags |= (1 << (UPDATE_VIEWS_SHIFT + s)); 1947 } 1948 } 1949 } 1950 } 1951#endif 1952 1953#undef SYNCHRONIZED 1954}; 1955 1956#if API >= 11 1957/* This approach serves two purposes. 1958 * First, we don't want to do an atomic operation to manipulate the reference 1959 * count every time something is bound/unbound to the pipeline, since they are 1960 * expensive. 1961 * Fortunately, the immediate context can only be used by a single thread, so 1962 * we don't have to use them, as long as a separate reference count is used 1963 * (see dual_refcnt_t). 1964 * 1965 * Second, we want to avoid the Device -> DeviceContext -> bound DeviceChild -> Device 1966 * garbage cycle. 1967 * To avoid it, DeviceChild doesn't hold a reference to Device as usual, but adds 1968 * one for each external reference count, while internal nonatomic_add_ref doesn't 1969 * add any. 1970 * 1971 * Note that ideally we would to eliminate the non-atomic op too, but this is more 1972 * complicated, since we would either need to use garbage collection and give up 1973 * deterministic destruction (especially bad for large textures), or scan the whole 1974 * pipeline state every time the reference count of object drops to 0, which risks 1975 * pathological slowdowns. 1976 * 1977 * Since this microoptimization should matter relatively little, let's avoid it for now. 1978 * 1979 * Note that deferred contexts don't use this, since as a whole, they must thread-safe. 1980 * Eliminating the atomic ops for deferred contexts seems substantially harder. 1981 * This might be a problem if they are used in a one-shot multithreaded rendering 1982 * fashion, where SMP cacheline bouncing on the reference count may be visible. 1983 * 1984 * The idea would be to attach a structure of reference counts indexed by deferred 1985 * context id to each object. Ideally, this should be organized like ext2 block pointers. 1986 * 1987 * Every deferred context would get a reference count in its own cacheline. 1988 * The external count is protected by a lock bit, and there is also a "lock bit" in each 1989 * internal count. 1990 * 1991 * When the external count has to be dropped to 0, the lock bit is taken and all internal 1992 * reference counts are scanned, taking a count of them. A flag would also be set on them. 1993 * Deferred context manipulation would notice the flag, and update the count. 1994 * Once the count goes to zero, the object is freed. 1995 * 1996 * The problem of this is that if the external reference count ping-pongs between 1997 * zero and non-zero, the scans will take a lot of time. 1998 * 1999 * The idea to solve this is to compute the scans in a binary-tree like fashion, where 2000 * each binary tree node would have a "determined bit", which would be invalidated 2001 * by manipulations. 2002 * 2003 * However, all this complexity might actually be a loss in most cases, so let's just 2004 * stick to a single atomic refcnt for now. 2005 * 2006 * Also, we don't even support deferred contexts yet, so this can wait. 2007 */ 2008struct nonatomic_device_child_ptr_traits 2009{ 2010 static void add_ref(void* p) 2011 { 2012 if(p) 2013 ((GalliumD3D11DeviceChild<>*)p)->nonatomic_add_ref(); 2014 } 2015 2016 static void release(void* p) 2017 { 2018 if(p) 2019 ((GalliumD3D11DeviceChild<>*)p)->nonatomic_release(); 2020 } 2021}; 2022 2023struct GalliumD3D11ImmediateDeviceContext 2024 : public GalliumD3D11DeviceContext<nonatomic_device_child_ptr_traits> 2025{ 2026 GalliumD3D11ImmediateDeviceContext(GalliumD3D11Screen* device, pipe_context* pipe, unsigned context_flags = 0) 2027 : GalliumD3D11DeviceContext<nonatomic_device_child_ptr_traits>(device, pipe, context_flags) 2028 { 2029 // not necessary, but tests that the API at least basically works 2030 ClearState(); 2031 } 2032 2033 /* we do this since otherwise we would have a garbage cycle between this and the device */ 2034 virtual ULONG STDMETHODCALLTYPE AddRef() 2035 { 2036 return this->device->AddRef(); 2037 } 2038 2039 virtual ULONG STDMETHODCALLTYPE Release() 2040 { 2041 return this->device->Release(); 2042 } 2043 2044 virtual D3D11_DEVICE_CONTEXT_TYPE STDMETHODCALLTYPE GetType() 2045 { 2046 return D3D11_DEVICE_CONTEXT_IMMEDIATE; 2047 } 2048}; 2049 2050static ID3D11DeviceContext* GalliumD3D11ImmediateDeviceContext_Create(GalliumD3D11Screen* device, struct pipe_context* pipe, bool owns_pipe) 2051{ 2052 return new GalliumD3D11ImmediateDeviceContext(device, pipe, owns_pipe); 2053} 2054 2055static void GalliumD3D11ImmediateDeviceContext_RestoreGalliumState(ID3D11DeviceContext* context) 2056{ 2057 ((GalliumD3D11ImmediateDeviceContext*)context)->RestoreGalliumState(); 2058} 2059 2060static void GalliumD3D11ImmediateDeviceContext_RestoreGalliumStateBlitOnly(ID3D11DeviceContext* context) 2061{ 2062 ((GalliumD3D11ImmediateDeviceContext*)context)->RestoreGalliumStateBlitOnly(); 2063} 2064 2065static void GalliumD3D11ImmediateDeviceContext_Destroy(ID3D11DeviceContext* context) 2066{ 2067 delete (GalliumD3D11ImmediateDeviceContext*)context; 2068} 2069#endif 2070