1/************************************************************************** 2 * 3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. 4 * All Rights Reserved. 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a 7 * copy of this software and associated documentation files (the 8 * "Software"), to deal in the Software without restriction, including 9 * without limitation the rights to use, copy, modify, merge, publish, 10 * distribute, sub license, and/or sell copies of the Software, and to 11 * permit persons to whom the Software is furnished to do so, subject to 12 * the following conditions: 13 * 14 * The above copyright notice and this permission notice (including the 15 * next paragraph) shall be included in all copies or substantial portions 16 * of the Software. 17 * 18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS 19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. 21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR 22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, 23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE 24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. 25 * 26 **************************************************************************/ 27 28#ifndef INTELCONTEXT_INC 29#define INTELCONTEXT_INC 30 31 32#include <stdbool.h> 33#include <string.h> 34#include "main/mtypes.h" 35#include "main/mm.h" 36 37#ifdef __cplusplus 38extern "C" { 39 /* Evil hack for using libdrm in a c++ compiler. */ 40 #define virtual virt 41#endif 42 43#include "drm.h" 44#include "intel_bufmgr.h" 45 46#include "intel_screen.h" 47#include "intel_tex_obj.h" 48#include "i915_drm.h" 49 50#ifdef __cplusplus 51 #undef virtual 52#endif 53 54#include "tnl/t_vertex.h" 55 56#define TAG(x) intel##x 57#include "tnl_dd/t_dd_vertex.h" 58#undef TAG 59 60#define DV_PF_555 (1<<8) 61#define DV_PF_565 (2<<8) 62#define DV_PF_8888 (3<<8) 63#define DV_PF_4444 (8<<8) 64#define DV_PF_1555 (9<<8) 65 66struct intel_region; 67struct intel_context; 68 69typedef void (*intel_tri_func) (struct intel_context *, intelVertex *, 70 intelVertex *, intelVertex *); 71typedef void (*intel_line_func) (struct intel_context *, intelVertex *, 72 intelVertex *); 73typedef void (*intel_point_func) (struct intel_context *, intelVertex *); 74 75/** 76 * Bits for intel->Fallback field 77 */ 78/*@{*/ 79#define INTEL_FALLBACK_DRAW_BUFFER 0x1 80#define INTEL_FALLBACK_READ_BUFFER 0x2 81#define INTEL_FALLBACK_DEPTH_BUFFER 0x4 82#define INTEL_FALLBACK_STENCIL_BUFFER 0x8 83#define INTEL_FALLBACK_USER 0x10 84#define INTEL_FALLBACK_RENDERMODE 0x20 85#define INTEL_FALLBACK_TEXTURE 0x40 86#define INTEL_FALLBACK_DRIVER 0x1000 /**< first for drivers */ 87/*@}*/ 88 89extern void intelFallback(struct intel_context *intel, GLbitfield bit, 90 bool mode); 91#define FALLBACK( intel, bit, mode ) intelFallback( intel, bit, mode ) 92 93 94#define INTEL_WRITE_PART 0x1 95#define INTEL_WRITE_FULL 0x2 96#define INTEL_READ 0x4 97 98#define INTEL_MAX_FIXUP 64 99 100#ifndef likely 101#ifdef __GNUC__ 102#define likely(expr) (__builtin_expect(expr, 1)) 103#define unlikely(expr) (__builtin_expect(expr, 0)) 104#else 105#define likely(expr) (expr) 106#define unlikely(expr) (expr) 107#endif 108#endif 109 110struct intel_sync_object { 111 struct gl_sync_object Base; 112 113 /** Batch associated with this sync object */ 114 drm_intel_bo *bo; 115}; 116 117struct brw_context; 118 119struct intel_batchbuffer { 120 /** Current batchbuffer being queued up. */ 121 drm_intel_bo *bo; 122 /** Last BO submitted to the hardware. Used for glFinish(). */ 123 drm_intel_bo *last_bo; 124 /** BO for post-sync nonzero writes for gen6 workaround. */ 125 drm_intel_bo *workaround_bo; 126 bool need_workaround_flush; 127 128 struct cached_batch_item *cached_items; 129 130 uint16_t emit, total; 131 uint16_t used, reserved_space; 132 uint32_t map[8192]; 133#define BATCH_SZ (8192*sizeof(uint32_t)) 134 135 uint32_t state_batch_offset; 136 bool is_blit; 137 bool needs_sol_reset; 138 139 struct { 140 uint16_t used; 141 int reloc_count; 142 } saved; 143}; 144 145/** 146 * intel_context is derived from Mesa's context class: struct gl_context. 147 */ 148struct intel_context 149{ 150 struct gl_context ctx; /**< base class, must be first field */ 151 152 struct 153 { 154 void (*destroy) (struct intel_context * intel); 155 void (*emit_state) (struct intel_context * intel); 156 void (*finish_batch) (struct intel_context * intel); 157 void (*new_batch) (struct intel_context * intel); 158 void (*emit_invarient_state) (struct intel_context * intel); 159 void (*update_texture_state) (struct intel_context * intel); 160 161 void (*render_start) (struct intel_context * intel); 162 void (*render_prevalidate) (struct intel_context * intel); 163 void (*set_draw_region) (struct intel_context * intel, 164 struct intel_region * draw_regions[], 165 struct intel_region * depth_region, 166 GLuint num_regions); 167 void (*update_draw_buffer)(struct intel_context *intel); 168 169 void (*reduced_primitive_state) (struct intel_context * intel, 170 GLenum rprim); 171 172 bool (*check_vertex_size) (struct intel_context * intel, 173 GLuint expected); 174 void (*invalidate_state) (struct intel_context *intel, 175 GLuint new_state); 176 177 void (*assert_not_dirty) (struct intel_context *intel); 178 179 void (*debug_batch)(struct intel_context *intel); 180 void (*annotate_aub)(struct intel_context *intel); 181 bool (*render_target_supported)(struct intel_context *intel, 182 struct gl_renderbuffer *rb); 183 184 /** Can HiZ be enabled on a depthbuffer of the given format? */ 185 bool (*is_hiz_depth_format)(struct intel_context *intel, 186 gl_format format); 187 188 /** 189 * Surface state operations (i965+ only) 190 * \{ 191 */ 192 void (*update_texture_surface)(struct gl_context *ctx, 193 unsigned unit, 194 uint32_t *binding_table, 195 unsigned surf_index); 196 void (*update_renderbuffer_surface)(struct brw_context *brw, 197 struct gl_renderbuffer *rb, 198 unsigned unit); 199 void (*update_null_renderbuffer_surface)(struct brw_context *brw, 200 unsigned unit); 201 void (*create_constant_surface)(struct brw_context *brw, 202 drm_intel_bo *bo, 203 uint32_t offset, 204 int width, 205 uint32_t *out_offset); 206 /** \} */ 207 } vtbl; 208 209 GLbitfield Fallback; /**< mask of INTEL_FALLBACK_x bits */ 210 GLuint NewGLState; 211 212 dri_bufmgr *bufmgr; 213 unsigned int maxBatchSize; 214 215 /** 216 * Generation number of the hardware: 2 is 8xx, 3 is 9xx pre-965, 4 is 965. 217 */ 218 int gen; 219 int gt; 220 bool needs_ff_sync; 221 bool is_haswell; 222 bool is_g4x; 223 bool is_945; 224 bool has_separate_stencil; 225 bool must_use_separate_stencil; 226 bool has_hiz; 227 bool has_llc; 228 bool has_swizzling; 229 230 int urb_size; 231 232 drm_intel_context *hw_ctx; 233 234 struct intel_batchbuffer batch; 235 236 drm_intel_bo *first_post_swapbuffers_batch; 237 bool need_throttle; 238 bool no_batch_wrap; 239 bool tnl_pipeline_running; /**< Set while i915's _tnl_run_pipeline. */ 240 241 struct 242 { 243 GLuint id; 244 uint32_t start_ptr; /**< for i8xx */ 245 uint32_t primitive; /**< Current hardware primitive type */ 246 void (*flush) (struct intel_context *); 247 drm_intel_bo *vb_bo; 248 uint8_t *vb; 249 unsigned int start_offset; /**< Byte offset of primitive sequence */ 250 unsigned int current_offset; /**< Byte offset of next vertex */ 251 unsigned int count; /**< Number of vertices in current primitive */ 252 } prim; 253 254 struct { 255 drm_intel_bo *bo; 256 GLuint offset; 257 uint32_t buffer_len; 258 uint32_t buffer_offset; 259 char buffer[4096]; 260 } upload; 261 262 GLuint stats_wm; 263 264 /* Offsets of fields within the current vertex: 265 */ 266 GLuint coloroffset; 267 GLuint specoffset; 268 GLuint wpos_offset; 269 270 struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; 271 GLuint vertex_attr_count; 272 273 GLfloat polygon_offset_scale; /* dependent on depth_scale, bpp */ 274 275 bool hw_stencil; 276 bool hw_stipple; 277 bool no_rast; 278 bool always_flush_batch; 279 bool always_flush_cache; 280 281 /* State for intelvb.c and inteltris.c. 282 */ 283 GLuint RenderIndex; 284 GLmatrix ViewportMatrix; 285 GLenum render_primitive; 286 GLenum reduced_primitive; /*< Only gen < 6 */ 287 GLuint vertex_size; 288 GLubyte *verts; /* points to tnl->clipspace.vertex_buf */ 289 290 /* Fallback rasterization functions 291 */ 292 intel_point_func draw_point; 293 intel_line_func draw_line; 294 intel_tri_func draw_tri; 295 296 /** 297 * Set if rendering has occured to the drawable's front buffer. 298 * 299 * This is used in the DRI2 case to detect that glFlush should also copy 300 * the contents of the fake front buffer to the real front buffer. 301 */ 302 bool front_buffer_dirty; 303 304 /** 305 * Track whether front-buffer rendering is currently enabled 306 * 307 * A separate flag is used to track this in order to support MRT more 308 * easily. 309 */ 310 bool is_front_buffer_rendering; 311 /** 312 * Track whether front-buffer is the current read target. 313 * 314 * This is closely associated with is_front_buffer_rendering, but may 315 * be set separately. The DRI2 fake front buffer must be referenced 316 * either way. 317 */ 318 bool is_front_buffer_reading; 319 320 /** 321 * Count of intel_regions that are mapped. 322 * 323 * This allows us to assert that no batch buffer is emitted if a 324 * region is mapped. 325 */ 326 int num_mapped_regions; 327 328 bool use_texture_tiling; 329 bool use_early_z; 330 331 int driFd; 332 333 __DRIcontext *driContext; 334 struct intel_screen *intelScreen; 335 void (*saved_viewport)(struct gl_context * ctx, 336 GLint x, GLint y, GLsizei width, GLsizei height); 337 338 /** 339 * Configuration cache 340 */ 341 driOptionCache optionCache; 342}; 343 344extern char *__progname; 345 346 347#define SUBPIXEL_X 0.125 348#define SUBPIXEL_Y 0.125 349 350#define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0])) 351 352/** 353 * Align a value up to an alignment value 354 * 355 * If \c value is not already aligned to the requested alignment value, it 356 * will be rounded up. 357 * 358 * \param value Value to be rounded 359 * \param alignment Alignment value to be used. This must be a power of two. 360 * 361 * \sa ROUND_DOWN_TO() 362 */ 363#define ALIGN(value, alignment) (((value) + alignment - 1) & ~(alignment - 1)) 364 365/** 366 * Align a value down to an alignment value 367 * 368 * If \c value is not already aligned to the requested alignment value, it 369 * will be rounded down. 370 * 371 * \param value Value to be rounded 372 * \param alignment Alignment value to be used. This must be a power of two. 373 * 374 * \sa ALIGN() 375 */ 376#define ROUND_DOWN_TO(value, alignment) ((value) & ~(alignment - 1)) 377 378#define IS_POWER_OF_TWO(val) (((val) & (val - 1)) == 0) 379 380static INLINE uint32_t 381U_FIXED(float value, uint32_t frac_bits) 382{ 383 value *= (1 << frac_bits); 384 return value < 0 ? 0 : value; 385} 386 387static INLINE uint32_t 388S_FIXED(float value, uint32_t frac_bits) 389{ 390 return value * (1 << frac_bits); 391} 392 393#define INTEL_FIREVERTICES(intel) \ 394do { \ 395 if ((intel)->prim.flush) \ 396 (intel)->prim.flush(intel); \ 397} while (0) 398 399/* ================================================================ 400 * From linux kernel i386 header files, copes with odd sizes better 401 * than COPY_DWORDS would: 402 * XXX Put this in src/mesa/main/imports.h ??? 403 */ 404#if defined(i386) || defined(__i386__) 405static INLINE void * __memcpy(void * to, const void * from, size_t n) 406{ 407 int d0, d1, d2; 408 __asm__ __volatile__( 409 "rep ; movsl\n\t" 410 "testb $2,%b4\n\t" 411 "je 1f\n\t" 412 "movsw\n" 413 "1:\ttestb $1,%b4\n\t" 414 "je 2f\n\t" 415 "movsb\n" 416 "2:" 417 : "=&c" (d0), "=&D" (d1), "=&S" (d2) 418 :"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from) 419 : "memory"); 420 return (to); 421} 422#else 423#define __memcpy(a,b,c) memcpy(a,b,c) 424#endif 425 426 427/* ================================================================ 428 * Debugging: 429 */ 430extern int INTEL_DEBUG; 431 432#define DEBUG_TEXTURE 0x1 433#define DEBUG_STATE 0x2 434#define DEBUG_IOCTL 0x4 435#define DEBUG_BLIT 0x8 436#define DEBUG_MIPTREE 0x10 437#define DEBUG_PERF 0x20 438#define DEBUG_VERBOSE 0x40 439#define DEBUG_BATCH 0x80 440#define DEBUG_PIXEL 0x100 441#define DEBUG_BUFMGR 0x200 442#define DEBUG_REGION 0x400 443#define DEBUG_FBO 0x800 444#define DEBUG_GS 0x1000 445#define DEBUG_SYNC 0x2000 446#define DEBUG_PRIMS 0x4000 447#define DEBUG_VERTS 0x8000 448#define DEBUG_DRI 0x10000 449#define DEBUG_SF 0x20000 450#define DEBUG_SANITY 0x40000 451#define DEBUG_SLEEP 0x80000 452#define DEBUG_STATS 0x100000 453#define DEBUG_TILE 0x200000 454#define DEBUG_WM 0x400000 455#define DEBUG_URB 0x800000 456#define DEBUG_VS 0x1000000 457#define DEBUG_CLIP 0x2000000 458#define DEBUG_AUB 0x4000000 459 460#define DBG(...) do { \ 461 if (unlikely(INTEL_DEBUG & FILE_DEBUG_FLAG)) \ 462 printf(__VA_ARGS__); \ 463} while(0) 464 465#define fallback_debug(...) do { \ 466 if (unlikely(INTEL_DEBUG & DEBUG_PERF)) \ 467 printf(__VA_ARGS__); \ 468} while(0) 469 470#define perf_debug(...) do { \ 471 if (unlikely(INTEL_DEBUG & DEBUG_PERF)) \ 472 printf(__VA_ARGS__); \ 473} while(0) 474 475#define PCI_CHIP_845_G 0x2562 476#define PCI_CHIP_I830_M 0x3577 477#define PCI_CHIP_I855_GM 0x3582 478#define PCI_CHIP_I865_G 0x2572 479#define PCI_CHIP_I915_G 0x2582 480#define PCI_CHIP_I915_GM 0x2592 481#define PCI_CHIP_I945_G 0x2772 482#define PCI_CHIP_I945_GM 0x27A2 483#define PCI_CHIP_I945_GME 0x27AE 484#define PCI_CHIP_G33_G 0x29C2 485#define PCI_CHIP_Q35_G 0x29B2 486#define PCI_CHIP_Q33_G 0x29D2 487 488 489/* ================================================================ 490 * intel_context.c: 491 */ 492 493extern bool intelInitContext(struct intel_context *intel, 494 int api, 495 const struct gl_config * mesaVis, 496 __DRIcontext * driContextPriv, 497 void *sharedContextPrivate, 498 struct dd_function_table *functions); 499 500extern void intelFinish(struct gl_context * ctx); 501extern void intel_flush_rendering_to_batch(struct gl_context *ctx); 502extern void _intel_flush(struct gl_context * ctx, const char *file, int line); 503 504#define intel_flush(ctx) _intel_flush(ctx, __FILE__, __LINE__) 505 506extern void intelInitDriverFunctions(struct dd_function_table *functions); 507 508void intel_init_syncobj_functions(struct dd_function_table *functions); 509 510 511/* ================================================================ 512 * intel_state.c: 513 */ 514 515#define COMPAREFUNC_ALWAYS 0 516#define COMPAREFUNC_NEVER 0x1 517#define COMPAREFUNC_LESS 0x2 518#define COMPAREFUNC_EQUAL 0x3 519#define COMPAREFUNC_LEQUAL 0x4 520#define COMPAREFUNC_GREATER 0x5 521#define COMPAREFUNC_NOTEQUAL 0x6 522#define COMPAREFUNC_GEQUAL 0x7 523 524#define STENCILOP_KEEP 0 525#define STENCILOP_ZERO 0x1 526#define STENCILOP_REPLACE 0x2 527#define STENCILOP_INCRSAT 0x3 528#define STENCILOP_DECRSAT 0x4 529#define STENCILOP_INCR 0x5 530#define STENCILOP_DECR 0x6 531#define STENCILOP_INVERT 0x7 532 533#define LOGICOP_CLEAR 0 534#define LOGICOP_NOR 0x1 535#define LOGICOP_AND_INV 0x2 536#define LOGICOP_COPY_INV 0x3 537#define LOGICOP_AND_RVRSE 0x4 538#define LOGICOP_INV 0x5 539#define LOGICOP_XOR 0x6 540#define LOGICOP_NAND 0x7 541#define LOGICOP_AND 0x8 542#define LOGICOP_EQUIV 0x9 543#define LOGICOP_NOOP 0xa 544#define LOGICOP_OR_INV 0xb 545#define LOGICOP_COPY 0xc 546#define LOGICOP_OR_RVRSE 0xd 547#define LOGICOP_OR 0xe 548#define LOGICOP_SET 0xf 549 550#define BLENDFACT_ZERO 0x01 551#define BLENDFACT_ONE 0x02 552#define BLENDFACT_SRC_COLR 0x03 553#define BLENDFACT_INV_SRC_COLR 0x04 554#define BLENDFACT_SRC_ALPHA 0x05 555#define BLENDFACT_INV_SRC_ALPHA 0x06 556#define BLENDFACT_DST_ALPHA 0x07 557#define BLENDFACT_INV_DST_ALPHA 0x08 558#define BLENDFACT_DST_COLR 0x09 559#define BLENDFACT_INV_DST_COLR 0x0a 560#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b 561#define BLENDFACT_CONST_COLOR 0x0c 562#define BLENDFACT_INV_CONST_COLOR 0x0d 563#define BLENDFACT_CONST_ALPHA 0x0e 564#define BLENDFACT_INV_CONST_ALPHA 0x0f 565#define BLENDFACT_MASK 0x0f 566 567enum { 568 DRI_CONF_BO_REUSE_DISABLED, 569 DRI_CONF_BO_REUSE_ALL 570}; 571 572extern int intel_translate_shadow_compare_func(GLenum func); 573extern int intel_translate_compare_func(GLenum func); 574extern int intel_translate_stencil_op(GLenum op); 575extern int intel_translate_blend_factor(GLenum factor); 576extern int intel_translate_logic_op(GLenum opcode); 577 578void intel_update_renderbuffers(__DRIcontext *context, 579 __DRIdrawable *drawable); 580void intel_prepare_render(struct intel_context *intel); 581 582void 583intel_downsample_for_dri2_flush(struct intel_context *intel, 584 __DRIdrawable *drawable); 585 586void i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region, 587 uint32_t buffer_id); 588void intel_init_texture_formats(struct gl_context *ctx); 589 590/*====================================================================== 591 * Inline conversion functions. 592 * These are better-typed than the macros used previously: 593 */ 594static INLINE struct intel_context * 595intel_context(struct gl_context * ctx) 596{ 597 return (struct intel_context *) ctx; 598} 599 600static INLINE bool 601is_power_of_two(uint32_t value) 602{ 603 return (value & (value - 1)) == 0; 604} 605 606#ifdef __cplusplus 607} 608#endif 609 610#endif 611