intel_context.h revision 27eedca3e0b9ed47545b8cea8229e3d0a732a103
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; 133 uint32_t *cpu_map; 134#define BATCH_SZ (8192*sizeof(uint32_t)) 135 136 uint32_t state_batch_offset; 137 bool is_blit; 138 bool needs_sol_reset; 139 140 struct { 141 uint16_t used; 142 int reloc_count; 143 } saved; 144}; 145 146/** 147 * intel_context is derived from Mesa's context class: struct gl_context. 148 */ 149struct intel_context 150{ 151 struct gl_context ctx; /**< base class, must be first field */ 152 153 struct 154 { 155 void (*destroy) (struct intel_context * intel); 156 void (*emit_state) (struct intel_context * intel); 157 void (*finish_batch) (struct intel_context * intel); 158 void (*new_batch) (struct intel_context * intel); 159 void (*emit_invarient_state) (struct intel_context * intel); 160 void (*update_texture_state) (struct intel_context * intel); 161 162 void (*render_start) (struct intel_context * intel); 163 void (*render_prevalidate) (struct intel_context * intel); 164 void (*set_draw_region) (struct intel_context * intel, 165 struct intel_region * draw_regions[], 166 struct intel_region * depth_region, 167 GLuint num_regions); 168 void (*update_draw_buffer)(struct intel_context *intel); 169 170 void (*reduced_primitive_state) (struct intel_context * intel, 171 GLenum rprim); 172 173 bool (*check_vertex_size) (struct intel_context * intel, 174 GLuint expected); 175 void (*invalidate_state) (struct intel_context *intel, 176 GLuint new_state); 177 178 void (*assert_not_dirty) (struct intel_context *intel); 179 180 void (*debug_batch)(struct intel_context *intel); 181 void (*annotate_aub)(struct intel_context *intel); 182 bool (*render_target_supported)(struct intel_context *intel, 183 struct gl_renderbuffer *rb); 184 185 /** 186 * Surface state operations (i965+ only) 187 * \{ 188 */ 189 void (*update_texture_surface)(struct gl_context *ctx, 190 unsigned unit, 191 uint32_t *binding_table, 192 unsigned surf_index); 193 void (*update_renderbuffer_surface)(struct brw_context *brw, 194 struct gl_renderbuffer *rb, 195 bool layered, 196 unsigned unit); 197 void (*update_null_renderbuffer_surface)(struct brw_context *brw, 198 unsigned unit); 199 void (*create_constant_surface)(struct brw_context *brw, 200 drm_intel_bo *bo, 201 uint32_t offset, 202 uint32_t size, 203 uint32_t *out_offset, 204 bool dword_pitch); 205 /** \} */ 206 } vtbl; 207 208 GLbitfield Fallback; /**< mask of INTEL_FALLBACK_x bits */ 209 GLuint NewGLState; 210 211 dri_bufmgr *bufmgr; 212 unsigned int maxBatchSize; 213 214 /** 215 * Generation number of the hardware: 2 is 8xx, 3 is 9xx pre-965, 4 is 965. 216 */ 217 int gen; 218 int gt; 219 bool needs_ff_sync; 220 bool is_haswell; 221 bool is_baytrail; 222 bool is_g4x; 223 bool is_945; 224 bool has_llc; 225 bool has_swizzling; 226 227 int urb_size; 228 229 drm_intel_context *hw_ctx; 230 231 struct intel_batchbuffer batch; 232 233 drm_intel_bo *first_post_swapbuffers_batch; 234 bool need_throttle; 235 bool no_batch_wrap; 236 bool tnl_pipeline_running; /**< Set while i915's _tnl_run_pipeline. */ 237 238 /** 239 * Set if we're either a debug context or the INTEL_DEBUG=perf environment 240 * variable is set, this is the flag indicating to do expensive work that 241 * might lead to a perf_debug() call. 242 */ 243 bool perf_debug; 244 245 struct 246 { 247 GLuint id; 248 uint32_t start_ptr; /**< for i8xx */ 249 uint32_t primitive; /**< Current hardware primitive type */ 250 void (*flush) (struct intel_context *); 251 drm_intel_bo *vb_bo; 252 uint8_t *vb; 253 unsigned int start_offset; /**< Byte offset of primitive sequence */ 254 unsigned int current_offset; /**< Byte offset of next vertex */ 255 unsigned int count; /**< Number of vertices in current primitive */ 256 } prim; 257 258 struct { 259 drm_intel_bo *bo; 260 GLuint offset; 261 uint32_t buffer_len; 262 uint32_t buffer_offset; 263 char buffer[4096]; 264 } upload; 265 266 uint32_t max_gtt_map_object_size; 267 268 GLuint stats_wm; 269 270 /* Offsets of fields within the current vertex: 271 */ 272 GLuint coloroffset; 273 GLuint specoffset; 274 GLuint wpos_offset; 275 276 struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; 277 GLuint vertex_attr_count; 278 279 GLfloat polygon_offset_scale; /* dependent on depth_scale, bpp */ 280 281 bool hw_stencil; 282 bool hw_stipple; 283 bool no_rast; 284 bool always_flush_batch; 285 bool always_flush_cache; 286 bool disable_throttling; 287 288 /* State for intelvb.c and inteltris.c. 289 */ 290 GLuint RenderIndex; 291 GLmatrix ViewportMatrix; 292 GLenum render_primitive; 293 GLenum reduced_primitive; /*< Only gen < 6 */ 294 GLuint vertex_size; 295 GLubyte *verts; /* points to tnl->clipspace.vertex_buf */ 296 297 /* Fallback rasterization functions 298 */ 299 intel_point_func draw_point; 300 intel_line_func draw_line; 301 intel_tri_func draw_tri; 302 303 /** 304 * Set if rendering has occured to the drawable's front buffer. 305 * 306 * This is used in the DRI2 case to detect that glFlush should also copy 307 * the contents of the fake front buffer to the real front buffer. 308 */ 309 bool front_buffer_dirty; 310 311 /** 312 * Track whether front-buffer rendering is currently enabled 313 * 314 * A separate flag is used to track this in order to support MRT more 315 * easily. 316 */ 317 bool is_front_buffer_rendering; 318 /** 319 * Track whether front-buffer is the current read target. 320 * 321 * This is closely associated with is_front_buffer_rendering, but may 322 * be set separately. The DRI2 fake front buffer must be referenced 323 * either way. 324 */ 325 bool is_front_buffer_reading; 326 327 bool use_early_z; 328 329 int driFd; 330 331 __DRIcontext *driContext; 332 struct intel_screen *intelScreen; 333 void (*saved_viewport)(struct gl_context * ctx, 334 GLint x, GLint y, GLsizei width, GLsizei height); 335 336 /** 337 * Configuration cache 338 */ 339 driOptionCache optionCache; 340}; 341 342extern char *__progname; 343 344 345#define SUBPIXEL_X 0.125 346#define SUBPIXEL_Y 0.125 347 348/** 349 * Align a value down to an alignment value 350 * 351 * If \c value is not already aligned to the requested alignment value, it 352 * will be rounded down. 353 * 354 * \param value Value to be rounded 355 * \param alignment Alignment value to be used. This must be a power of two. 356 * 357 * \sa ALIGN() 358 */ 359#define ROUND_DOWN_TO(value, alignment) ((value) & ~(alignment - 1)) 360 361static INLINE uint32_t 362U_FIXED(float value, uint32_t frac_bits) 363{ 364 value *= (1 << frac_bits); 365 return value < 0 ? 0 : value; 366} 367 368static INLINE uint32_t 369S_FIXED(float value, uint32_t frac_bits) 370{ 371 return value * (1 << frac_bits); 372} 373 374#define INTEL_FIREVERTICES(intel) \ 375do { \ 376 if ((intel)->prim.flush) \ 377 (intel)->prim.flush(intel); \ 378} while (0) 379 380/* ================================================================ 381 * From linux kernel i386 header files, copes with odd sizes better 382 * than COPY_DWORDS would: 383 * XXX Put this in src/mesa/main/imports.h ??? 384 */ 385#if defined(i386) || defined(__i386__) 386static INLINE void * __memcpy(void * to, const void * from, size_t n) 387{ 388 int d0, d1, d2; 389 __asm__ __volatile__( 390 "rep ; movsl\n\t" 391 "testb $2,%b4\n\t" 392 "je 1f\n\t" 393 "movsw\n" 394 "1:\ttestb $1,%b4\n\t" 395 "je 2f\n\t" 396 "movsb\n" 397 "2:" 398 : "=&c" (d0), "=&D" (d1), "=&S" (d2) 399 :"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from) 400 : "memory"); 401 return (to); 402} 403#else 404#define __memcpy(a,b,c) memcpy(a,b,c) 405#endif 406 407 408/* ================================================================ 409 * Debugging: 410 */ 411extern int INTEL_DEBUG; 412 413#define DEBUG_TEXTURE 0x1 414#define DEBUG_STATE 0x2 415#define DEBUG_IOCTL 0x4 416#define DEBUG_BLIT 0x8 417#define DEBUG_MIPTREE 0x10 418#define DEBUG_PERF 0x20 419#define DEBUG_BATCH 0x80 420#define DEBUG_PIXEL 0x100 421#define DEBUG_BUFMGR 0x200 422#define DEBUG_REGION 0x400 423#define DEBUG_FBO 0x800 424#define DEBUG_GS 0x1000 425#define DEBUG_SYNC 0x2000 426#define DEBUG_PRIMS 0x4000 427#define DEBUG_VERTS 0x8000 428#define DEBUG_DRI 0x10000 429#define DEBUG_SF 0x20000 430#define DEBUG_STATS 0x100000 431#define DEBUG_WM 0x400000 432#define DEBUG_URB 0x800000 433#define DEBUG_VS 0x1000000 434#define DEBUG_CLIP 0x2000000 435#define DEBUG_AUB 0x4000000 436#define DEBUG_SHADER_TIME 0x8000000 437#define DEBUG_BLORP 0x10000000 438#define DEBUG_NO16 0x20000000 439 440#ifdef HAVE_ANDROID_PLATFORM 441#define LOG_TAG "INTEL-MESA" 442#include <cutils/log.h> 443#ifndef ALOGW 444#define ALOGW LOGW 445#endif 446#define dbg_printf(...) ALOGW(__VA_ARGS__) 447#else 448#define dbg_printf(...) printf(__VA_ARGS__) 449#endif /* HAVE_ANDROID_PLATFORM */ 450 451#define DBG(...) do { \ 452 if (unlikely(INTEL_DEBUG & FILE_DEBUG_FLAG)) \ 453 dbg_printf(__VA_ARGS__); \ 454} while(0) 455 456#define perf_debug(...) do { \ 457 static GLuint msg_id = 0; \ 458 if (unlikely(INTEL_DEBUG & DEBUG_PERF)) \ 459 dbg_printf(__VA_ARGS__); \ 460 if (intel->perf_debug) \ 461 _mesa_gl_debug(&intel->ctx, &msg_id, \ 462 MESA_DEBUG_TYPE_PERFORMANCE, \ 463 MESA_DEBUG_SEVERITY_MEDIUM, \ 464 __VA_ARGS__); \ 465} while(0) 466 467#define WARN_ONCE(cond, fmt...) do { \ 468 if (unlikely(cond)) { \ 469 static bool _warned = false; \ 470 static GLuint msg_id = 0; \ 471 if (!_warned) { \ 472 fprintf(stderr, "WARNING: "); \ 473 fprintf(stderr, fmt); \ 474 _warned = true; \ 475 \ 476 _mesa_gl_debug(ctx, &msg_id, \ 477 MESA_DEBUG_TYPE_OTHER, \ 478 MESA_DEBUG_SEVERITY_HIGH, fmt); \ 479 } \ 480 } \ 481} while (0) 482 483#define PCI_CHIP_845_G 0x2562 484#define PCI_CHIP_I830_M 0x3577 485#define PCI_CHIP_I855_GM 0x3582 486#define PCI_CHIP_I865_G 0x2572 487#define PCI_CHIP_I915_G 0x2582 488#define PCI_CHIP_I915_GM 0x2592 489#define PCI_CHIP_I945_G 0x2772 490#define PCI_CHIP_I945_GM 0x27A2 491#define PCI_CHIP_I945_GME 0x27AE 492#define PCI_CHIP_G33_G 0x29C2 493#define PCI_CHIP_Q35_G 0x29B2 494#define PCI_CHIP_Q33_G 0x29D2 495 496 497/* ================================================================ 498 * intel_context.c: 499 */ 500 501extern bool intelInitContext(struct intel_context *intel, 502 int api, 503 unsigned major_version, 504 unsigned minor_version, 505 const struct gl_config * mesaVis, 506 __DRIcontext * driContextPriv, 507 void *sharedContextPrivate, 508 struct dd_function_table *functions, 509 unsigned *dri_ctx_error); 510 511extern void intelFinish(struct gl_context * ctx); 512extern void intel_flush_rendering_to_batch(struct gl_context *ctx); 513extern void _intel_flush(struct gl_context * ctx, const char *file, int line); 514 515#define intel_flush(ctx) _intel_flush(ctx, __FILE__, __LINE__) 516 517extern void intelInitDriverFunctions(struct dd_function_table *functions); 518 519void intel_init_syncobj_functions(struct dd_function_table *functions); 520 521 522/* ================================================================ 523 * intel_state.c: 524 */ 525 526#define COMPAREFUNC_ALWAYS 0 527#define COMPAREFUNC_NEVER 0x1 528#define COMPAREFUNC_LESS 0x2 529#define COMPAREFUNC_EQUAL 0x3 530#define COMPAREFUNC_LEQUAL 0x4 531#define COMPAREFUNC_GREATER 0x5 532#define COMPAREFUNC_NOTEQUAL 0x6 533#define COMPAREFUNC_GEQUAL 0x7 534 535#define STENCILOP_KEEP 0 536#define STENCILOP_ZERO 0x1 537#define STENCILOP_REPLACE 0x2 538#define STENCILOP_INCRSAT 0x3 539#define STENCILOP_DECRSAT 0x4 540#define STENCILOP_INCR 0x5 541#define STENCILOP_DECR 0x6 542#define STENCILOP_INVERT 0x7 543 544#define LOGICOP_CLEAR 0 545#define LOGICOP_NOR 0x1 546#define LOGICOP_AND_INV 0x2 547#define LOGICOP_COPY_INV 0x3 548#define LOGICOP_AND_RVRSE 0x4 549#define LOGICOP_INV 0x5 550#define LOGICOP_XOR 0x6 551#define LOGICOP_NAND 0x7 552#define LOGICOP_AND 0x8 553#define LOGICOP_EQUIV 0x9 554#define LOGICOP_NOOP 0xa 555#define LOGICOP_OR_INV 0xb 556#define LOGICOP_COPY 0xc 557#define LOGICOP_OR_RVRSE 0xd 558#define LOGICOP_OR 0xe 559#define LOGICOP_SET 0xf 560 561#define BLENDFACT_ZERO 0x01 562#define BLENDFACT_ONE 0x02 563#define BLENDFACT_SRC_COLR 0x03 564#define BLENDFACT_INV_SRC_COLR 0x04 565#define BLENDFACT_SRC_ALPHA 0x05 566#define BLENDFACT_INV_SRC_ALPHA 0x06 567#define BLENDFACT_DST_ALPHA 0x07 568#define BLENDFACT_INV_DST_ALPHA 0x08 569#define BLENDFACT_DST_COLR 0x09 570#define BLENDFACT_INV_DST_COLR 0x0a 571#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b 572#define BLENDFACT_CONST_COLOR 0x0c 573#define BLENDFACT_INV_CONST_COLOR 0x0d 574#define BLENDFACT_CONST_ALPHA 0x0e 575#define BLENDFACT_INV_CONST_ALPHA 0x0f 576#define BLENDFACT_MASK 0x0f 577 578enum { 579 DRI_CONF_BO_REUSE_DISABLED, 580 DRI_CONF_BO_REUSE_ALL 581}; 582 583extern int intel_translate_shadow_compare_func(GLenum func); 584extern int intel_translate_compare_func(GLenum func); 585extern int intel_translate_stencil_op(GLenum op); 586extern int intel_translate_blend_factor(GLenum factor); 587extern int intel_translate_logic_op(GLenum opcode); 588 589void intel_update_renderbuffers(__DRIcontext *context, 590 __DRIdrawable *drawable); 591void intel_prepare_render(struct intel_context *intel); 592 593void i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region, 594 uint32_t buffer_id); 595void intel_init_texture_formats(struct gl_context *ctx); 596 597/*====================================================================== 598 * Inline conversion functions. 599 * These are better-typed than the macros used previously: 600 */ 601static INLINE struct intel_context * 602intel_context(struct gl_context * ctx) 603{ 604 return (struct intel_context *) ctx; 605} 606 607static INLINE bool 608is_power_of_two(uint32_t value) 609{ 610 return (value & (value - 1)) == 0; 611} 612 613#ifdef __cplusplus 614} 615#endif 616 617#endif 618