src/mesa/drivers/dri/i965/brw_context.h

   1 /*
   2  Copyright (C) Intel Corp.  2006.  All Rights Reserved.
   3  Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
   4  develop this 3D driver.
   5
   6  Permission is hereby granted, free of charge, to any person obtaining
   7  a 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, sublicense, 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
  16  portions of the Software.
  17
  18  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  19  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  20  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
  21  IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
  22  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
  23  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
  24  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  25
  26  **********************************************************************/
  27  /*
  28   * Authors:
  29   *   Keith Whitwell <keith@tungstengraphics.com>
  30   */
  31
  32
  33 #ifndef BRWCONTEXT_INC
  34 #define BRWCONTEXT_INC
  35
  36 #include "intel_context.h"
  37 #include "brw_structs.h"
  38 #include "main/imports.h"
  39
  40
  41 /* Glossary:
  42  *
  43  * URB - uniform resource buffer.  A mid-sized buffer which is
  44  * partitioned between the fixed function units and used for passing
  45  * values (vertices, primitives, constants) between them.
  46  *
  47  * CURBE - constant URB entry.  An urb region (entry) used to hold
  48  * constant values which the fixed function units can be instructed to
  49  * preload into the GRF when spawning a thread.
  50  *
  51  * VUE - vertex URB entry.  An urb entry holding a vertex and usually
  52  * a vertex header.  The header contains control information and
  53  * things like primitive type, Begin/end flags and clip codes.
  54  *
  55  * PUE - primitive URB entry.  An urb entry produced by the setup (SF)
  56  * unit holding rasterization and interpolation parameters.
  57  *
  58  * GRF - general register file.  One of several register files
  59  * addressable by programmed threads.  The inputs (r0, payload, curbe,
  60  * urb) of the thread are preloaded to this area before the thread is
  61  * spawned.  The registers are individually 8 dwords wide and suitable
  62  * for general usage.  Registers holding thread input values are not
  63  * special and may be overwritten.
  64  *
  65  * MRF - message register file.  Threads communicate (and terminate)
  66  * by sending messages.  Message parameters are placed in contiguous
  67  * MRF registers.  All program output is via these messages.  URB
  68  * entries are populated by sending a message to the shared URB
  69  * function containing the new data, together with a control word,
  70  * often an unmodified copy of R0.
  71  *
  72  * R0 - GRF register 0.  Typically holds control information used when
  73  * sending messages to other threads.
  74  *
  75  * EU or GEN4 EU: The name of the programmable subsystem of the
  76  * i965 hardware.  Threads are executed by the EU, the registers
  77  * described above are part of the EU architecture.
  78  *
  79  * Fixed function units:
  80  *
  81  * CS - Command streamer.  Notional first unit, little software
  82  * interaction.  Holds the URB entries used for constant data, ie the
  83  * CURBEs.
  84  *
  85  * VF/VS - Vertex Fetch / Vertex Shader.  The fixed function part of
  86  * this unit is responsible for pulling vertices out of vertex buffers
  87  * in vram and injecting them into the processing pipe as VUEs.  If
  88  * enabled, it first passes them to a VS thread which is a good place
  89  * for the driver to implement any active vertex shader.
  90  *
  91  * GS - Geometry Shader.  This corresponds to a new DX10 concept.  If
  92  * enabled, incoming strips etc are passed to GS threads in individual
  93  * line/triangle/point units.  The GS thread may perform arbitary
  94  * computation and emit whatever primtives with whatever vertices it
  95  * chooses.  This makes GS an excellent place to implement GL's
  96  * unfilled polygon modes, though of course it is capable of much
  97  * more.  Additionally, GS is used to translate away primitives not
  98  * handled by latter units, including Quads and Lineloops.
  99  *
 100  * CS - Clipper.  Mesa's clipping algorithms are imported to run on
 101  * this unit.  The fixed function part performs cliptesting against
 102  * the 6 fixed clipplanes and makes descisions on whether or not the
 103  * incoming primitive needs to be passed to a thread for clipping.
 104  * User clip planes are handled via cooperation with the VS thread.
 105  *
 106  * SF - Strips Fans or Setup: Triangles are prepared for
 107  * rasterization.  Interpolation coefficients are calculated.
 108  * Flatshading and two-side lighting usually performed here.
 109  *
 110  * WM - Windower.  Interpolation of vertex attributes performed here.
 111  * Fragment shader implemented here.  SIMD aspects of EU taken full
 112  * advantage of, as pixels are processed in blocks of 16.
 113  *
 114  * CC - Color Calculator.  No EU threads associated with this unit.
 115  * Handles blending and (presumably) depth and stencil testing.
 116  */
 117
 118
 119 #define BRW_MAX_CURBE                    (32*16)
 120
 121 struct brw_context;
 122
 123 #define BRW_NEW_URB_FENCE               0x1
 124 #define BRW_NEW_FRAGMENT_PROGRAM        0x2
 125 #define BRW_NEW_VERTEX_PROGRAM          0x4
 126 #define BRW_NEW_INPUT_DIMENSIONS        0x8
 127 #define BRW_NEW_CURBE_OFFSETS           0x10
 128 #define BRW_NEW_REDUCED_PRIMITIVE       0x20
 129 #define BRW_NEW_PRIMITIVE               0x40
 130 #define BRW_NEW_CONTEXT                 0x80
 131 #define BRW_NEW_WM_INPUT_DIMENSIONS     0x100
 132 #define BRW_NEW_PSP                     0x800
 133 #define BRW_NEW_WM_SURFACES             0x1000
 134 #define BRW_NEW_BINDING_TABLE           0x2000
 135 #define BRW_NEW_INDICES                 0x4000
 136 #define BRW_NEW_VERTICES                0x8000
 137 /**
 138  * Used for any batch entry with a relocated pointer that will be used
 139  * by any 3D rendering.
 140  */
 141 #define BRW_NEW_BATCH                   0x10000
 142 /** brw->depth_region updated */
 143 #define BRW_NEW_DEPTH_BUFFER            0x20000
 144 #define BRW_NEW_NR_WM_SURFACES          0x40000
 145 #define BRW_NEW_NR_VS_SURFACES          0x80000
 146 #define BRW_NEW_INDEX_BUFFER            0x100000
 147 #define BRW_NEW_VS_CONSTBUF             0x200000
 148 #define BRW_NEW_WM_CONSTBUF             0x200000
 149
 150 struct brw_state_flags {
 151    /** State update flags signalled by mesa internals */
 152    GLuint mesa;
 153    /**
 154     * State update flags signalled as the result of brw_tracked_state updates
 155     */
 156    GLuint brw;
 157    /** State update flags signalled by brw_state_cache.c searches */
 158    GLuint cache;
 159 };
 160
 161
 162 /** Subclass of Mesa vertex program */
 163 struct brw_vertex_program {
 164    struct gl_vertex_program program;
 165    GLuint id;
 166    GLboolean use_const_buffer;
 167 };
 168
 169
 170 /** Subclass of Mesa fragment program */
 171 struct brw_fragment_program {
 172    struct gl_fragment_program program;
 173    GLuint id;  /**< serial no. to identify frag progs, never re-used */
 174    GLboolean isGLSL;  /**< really, any IF/LOOP/CONT/BREAK instructions */
 175
 176    GLboolean use_const_buffer;
 177
 178    /** for debugging, which texture units are referenced */
 179    GLbitfield tex_units_used;
 180 };
 181
 182
 183 /* Data about a particular attempt to compile a program.  Note that
 184  * there can be many of these, each in a different GL state
 185  * corresponding to a different brw_wm_prog_key struct, with different
 186  * compiled programs:
 187  */
 188 struct brw_wm_prog_data {
 189    GLuint curb_read_length;
 190    GLuint urb_read_length;
 191
 192    GLuint first_curbe_grf;
 193    GLuint total_grf;
 194    GLuint total_scratch;
 195
 196    GLuint nr_params;       /**< number of float params/constants */
 197    GLboolean error;
 198
 199    /* Pointer to tracked values (only valid once
 200     * _mesa_load_state_parameters has been called at runtime).
 201     */
 202    const GLfloat *param[BRW_MAX_CURBE];
 203 };
 204
 205 struct brw_sf_prog_data {
 206    GLuint urb_read_length;
 207    GLuint total_grf;
 208
 209    /* Each vertex may have upto 12 attributes, 4 components each,
 210     * except WPOS which requires only 2.  (11*4 + 2) == 44 ==> 11
 211     * rows.
 212     *
 213     * Actually we use 4 for each, so call it 12 rows.
 214     */
 215    GLuint urb_entry_size;
 216 };
 217
 218 struct brw_clip_prog_data {
 219    GLuint curb_read_length;     /* user planes? */
 220    GLuint clip_mode;
 221    GLuint urb_read_length;
 222    GLuint total_grf;
 223 };
 224
 225 struct brw_gs_prog_data {
 226    GLuint urb_read_length;
 227    GLuint total_grf;
 228 };
 229
 230 struct brw_vs_prog_data {
 231    GLuint curb_read_length;
 232    GLuint urb_read_length;
 233    GLuint total_grf;
 234    GLbitfield64 outputs_written;
 235    GLuint nr_params;       /**< number of float params/constants */
 236
 237    GLuint inputs_read;
 238
 239    /* Used for calculating urb partitions:
 240     */
 241    GLuint urb_entry_size;
 242 };
 243
 244
 245 /* Size == 0 if output either not written, or always [0,0,0,1]
 246  */
 247 struct brw_vs_ouput_sizes {
 248    GLubyte output_size[VERT_RESULT_MAX];
 249 };
 250
 251
 252 /** Number of texture sampler units */
 253 #define BRW_MAX_TEX_UNIT 16
 254
 255 /** Max number of render targets in a shader */
 256 #define BRW_MAX_DRAW_BUFFERS 8
 257
 258 /**
 259  * Size of our surface binding table for the WM.
 260  * This contains pointers to the drawing surfaces and current texture
 261  * objects and shader constant buffers (+2).
 262  */
 263 #define BRW_WM_MAX_SURF (BRW_MAX_DRAW_BUFFERS + BRW_MAX_TEX_UNIT + 1)
 264
 265 /**
 266  * Helpers to convert drawing buffers, textures and constant buffers
 267  * to surface binding table indexes, for WM.
 268  */
 269 #define SURF_INDEX_DRAW(d)           (d)
 270 #define SURF_INDEX_FRAG_CONST_BUFFER (BRW_MAX_DRAW_BUFFERS)
 271 #define SURF_INDEX_TEXTURE(t)        (BRW_MAX_DRAW_BUFFERS + 1 + (t))
 272
 273 /**
 274  * Size of surface binding table for the VS.
 275  * Only one constant buffer for now.
 276  */
 277 #define BRW_VS_MAX_SURF 1
 278
 279 /**
 280  * Only a VS constant buffer
 281  */
 282 #define SURF_INDEX_VERT_CONST_BUFFER 0
 283
 284
 285 enum brw_cache_id {
 286    BRW_BLEND_STATE,
 287    BRW_DEPTH_STENCIL_STATE,
 288    BRW_COLOR_CALC_STATE,
 289    BRW_CC_VP,
 290    BRW_CC_UNIT,
 291    BRW_WM_PROG,
 292    BRW_SAMPLER_DEFAULT_COLOR,
 293    BRW_SAMPLER,
 294    BRW_WM_UNIT,
 295    BRW_SF_PROG,
 296    BRW_SF_VP,
 297    BRW_SF_UNIT, /* scissor state on gen6 */
 298    BRW_VS_UNIT,
 299    BRW_VS_PROG,
 300    BRW_GS_UNIT,
 301    BRW_GS_PROG,
 302    BRW_CLIP_VP,
 303    BRW_CLIP_UNIT,
 304    BRW_CLIP_PROG,
 305
 306    BRW_MAX_CACHE
 307 };
 308
 309 struct brw_cache_item {
 310    /**
 311     * Effectively part of the key, cache_id identifies what kind of state
 312     * buffer is involved, and also which brw->state.dirty.cache flag should
 313     * be set when this cache item is chosen.
 314     */
 315    enum brw_cache_id cache_id;
 316    /** 32-bit hash of the key data */
 317    GLuint hash;
 318    GLuint key_size;             /* for variable-sized keys */
 319    const void *key;
 320    drm_intel_bo **reloc_bufs;
 321    GLuint nr_reloc_bufs;
 322
 323    drm_intel_bo *bo;
 324
 325    struct brw_cache_item *next;
 326 };
 327
 328
 329
 330 struct brw_cache {
 331    struct brw_context *brw;
 332
 333    struct brw_cache_item **items;
 334    GLuint size, n_items;
 335
 336    char *name[BRW_MAX_CACHE];
 337
 338    /* Record of the last BOs chosen for each cache_id.  Used to set
 339     * brw->state.dirty.cache when a new cache item is chosen.
 340     */
 341    drm_intel_bo *last_bo[BRW_MAX_CACHE];
 342 };
 343
 344
 345 /* Considered adding a member to this struct to document which flags
 346  * an update might raise so that ordering of the state atoms can be
 347  * checked or derived at runtime.  Dropped the idea in favor of having
 348  * a debug mode where the state is monitored for flags which are
 349  * raised that have already been tested against.
 350  */
 351 struct brw_tracked_state {
 352    struct brw_state_flags dirty;
 353    void (*prepare)( struct brw_context *brw );
 354    void (*emit)( struct brw_context *brw );
 355 };
 356
 357 /* Flags for brw->state.cache.
 358  */
 359 #define CACHE_NEW_BLEND_STATE            (1<<BRW_BLEND_STATE)
 360 #define CACHE_NEW_DEPTH_STENCIL_STATE    (1<<BRW_DEPTH_STENCIL_STATE)
 361 #define CACHE_NEW_COLOR_CALC_STATE       (1<<BRW_COLOR_CALC_STATE)
 362 #define CACHE_NEW_CC_VP                  (1<<BRW_CC_VP)
 363 #define CACHE_NEW_CC_UNIT                (1<<BRW_CC_UNIT)
 364 #define CACHE_NEW_WM_PROG                (1<<BRW_WM_PROG)
 365 #define CACHE_NEW_SAMPLER_DEFAULT_COLOR  (1<<BRW_SAMPLER_DEFAULT_COLOR)
 366 #define CACHE_NEW_SAMPLER                (1<<BRW_SAMPLER)
 367 #define CACHE_NEW_WM_UNIT                (1<<BRW_WM_UNIT)
 368 #define CACHE_NEW_SF_PROG                (1<<BRW_SF_PROG)
 369 #define CACHE_NEW_SF_VP                  (1<<BRW_SF_VP)
 370 #define CACHE_NEW_SF_UNIT                (1<<BRW_SF_UNIT)
 371 #define CACHE_NEW_VS_UNIT                (1<<BRW_VS_UNIT)
 372 #define CACHE_NEW_VS_PROG                (1<<BRW_VS_PROG)
 373 #define CACHE_NEW_GS_UNIT                (1<<BRW_GS_UNIT)
 374 #define CACHE_NEW_GS_PROG                (1<<BRW_GS_PROG)
 375 #define CACHE_NEW_CLIP_VP                (1<<BRW_CLIP_VP)
 376 #define CACHE_NEW_CLIP_UNIT              (1<<BRW_CLIP_UNIT)
 377 #define CACHE_NEW_CLIP_PROG              (1<<BRW_CLIP_PROG)
 378
 379 struct brw_cached_batch_item {
 380    struct header *header;
 381    GLuint sz;
 382    struct brw_cached_batch_item *next;
 383 };
 384
 385
 386
 387 /* Protect against a future where VERT_ATTRIB_MAX > 32.  Wouldn't life
 388  * be easier if C allowed arrays of packed elements?
 389  */
 390 #define ATTRIB_BIT_DWORDS  ((VERT_ATTRIB_MAX+31)/32)
 391
 392 struct brw_vertex_element {
 393    const struct gl_client_array *glarray;
 394
 395    /** The corresponding Mesa vertex attribute */
 396    gl_vert_attrib attrib;
 397    /** Size of a complete element */
 398    GLuint element_size;
 399    /** Number of uploaded elements for this input. */
 400    GLuint count;
 401    /** Byte stride between elements in the uploaded array */
 402    GLuint stride;
 403    /** Offset of the first element within the buffer object */
 404    unsigned int offset;
 405    /** Buffer object containing the uploaded vertex data */
 406    drm_intel_bo *bo;
 407 };
 408
 409
 410
 411 struct brw_vertex_info {
 412    GLuint sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[VERT_ATTRIB_MAX] */
 413 };
 414
 415 struct brw_query_object {
 416    struct gl_query_object Base;
 417
 418    /** Last query BO associated with this query. */
 419    drm_intel_bo *bo;
 420    /** First index in bo with query data for this object. */
 421    int first_index;
 422    /** Last index in bo with query data for this object. */
 423    int last_index;
 424 };
 425
 426
 427 /**
 428  * brw_context is derived from intel_context.
 429  */
 430 struct brw_context
 431 {
 432    struct intel_context intel;  /**< base class, must be first field */
 433    GLuint primitive;
 434
 435    GLboolean emit_state_always;
 436    GLboolean has_surface_tile_offset;
 437    GLboolean has_compr4;
 438    GLboolean has_negative_rhw_bug;
 439    GLboolean has_aa_line_parameters;
 440    GLboolean has_pln;
 441 ;
 442    struct {
 443       struct brw_state_flags dirty;
 444
 445       GLuint nr_color_regions;
 446       struct intel_region *color_regions[MAX_DRAW_BUFFERS];
 447       struct intel_region *depth_region;
 448
 449       /**
 450        * List of buffers accumulated in brw_validate_state to receive
 451        * drm_intel_bo_check_aperture treatment before exec, so we can
 452        * know if we should flush the batch and try again before
 453        * emitting primitives.
 454        *
 455        * This can be a fixed number as we only have a limited number of
 456        * objects referenced from the batchbuffer in a primitive emit,
 457        * consisting of the vertex buffers, pipelined state pointers,
 458        * the CURBE, the depth buffer, and a query BO.
 459        */
 460       drm_intel_bo *validated_bos[VERT_ATTRIB_MAX + BRW_WM_MAX_SURF + 16];
 461       int validated_bo_count;
 462    } state;
 463
 464    struct brw_cache cache;
 465    struct brw_cached_batch_item *cached_batch_items;
 466
 467    struct {
 468       struct brw_vertex_element inputs[VERT_ATTRIB_MAX];
 469
 470       struct brw_vertex_element *enabled[VERT_ATTRIB_MAX];
 471       GLuint nr_enabled;
 472
 473 #define BRW_NR_UPLOAD_BUFS 17
 474 #define BRW_UPLOAD_INIT_SIZE (128*1024)
 475
 476       struct {
 477          drm_intel_bo *bo;
 478          GLuint offset;
 479       } upload;
 480
 481       /* Summary of size and varying of active arrays, so we can check
 482        * for changes to this state:
 483        */
 484       struct brw_vertex_info info;
 485       unsigned int min_index, max_index;
 486    } vb;
 487
 488    struct {
 489       /**
 490        * Index buffer for this draw_prims call.
 491        *
 492        * Updates are signaled by BRW_NEW_INDICES.
 493        */
 494       const struct _mesa_index_buffer *ib;
 495
 496       /* Updates to these fields are signaled by BRW_NEW_INDEX_BUFFER. */
 497       drm_intel_bo *bo;
 498       unsigned int offset;
 499       unsigned int size;
 500       /* Offset to index buffer index to use in CMD_3D_PRIM so that we can
 501        * avoid re-uploading the IB packet over and over if we're actually
 502        * referencing the same index buffer.
 503        */
 504       unsigned int start_vertex_offset;
 505    } ib;
 506
 507    /* Active vertex program:
 508     */
 509    const struct gl_vertex_program *vertex_program;
 510    const struct gl_fragment_program *fragment_program;
 511
 512
 513    /* For populating the gtt:
 514     */
 515    GLuint next_free_page;
 516
 517    /* hw-dependent 3DSTATE_VF_STATISTICS opcode */
 518    uint32_t CMD_VF_STATISTICS;
 519    /* hw-dependent 3DSTATE_PIPELINE_SELECT opcode */
 520    uint32_t CMD_PIPELINE_SELECT;
 521    int vs_max_threads;
 522    int wm_max_threads;
 523
 524    /* BRW_NEW_URB_ALLOCATIONS:
 525     */
 526    struct {
 527       GLuint vsize;             /* vertex size plus header in urb registers */
 528       GLuint csize;             /* constant buffer size in urb registers */
 529       GLuint sfsize;            /* setup data size in urb registers */
 530
 531       GLboolean constrained;
 532
 533       GLuint nr_vs_entries;
 534       GLuint nr_gs_entries;
 535       GLuint nr_clip_entries;
 536       GLuint nr_sf_entries;
 537       GLuint nr_cs_entries;
 538
 539       /* gen6 */
 540       GLuint vs_size;
 541 /*       GLuint gs_size; */
 542 /*       GLuint clip_size; */
 543 /*       GLuint sf_size; */
 544 /*       GLuint cs_size; */
 545
 546       GLuint vs_start;
 547       GLuint gs_start;
 548       GLuint clip_start;
 549       GLuint sf_start;
 550       GLuint cs_start;
 551       GLuint size; /* Hardware URB size, in KB. */
 552    } urb;
 553
 554
 555    /* BRW_NEW_CURBE_OFFSETS:
 556     */
 557    struct {
 558       GLuint wm_start;  /**< pos of first wm const in CURBE buffer */
 559       GLuint wm_size;   /**< number of float[4] consts, multiple of 16 */
 560       GLuint clip_start;
 561       GLuint clip_size;
 562       GLuint vs_start;
 563       GLuint vs_size;
 564       GLuint total_size;
 565
 566       drm_intel_bo *curbe_bo;
 567       /** Offset within curbe_bo of space for current curbe entry */
 568       GLuint curbe_offset;
 569       /** Offset within curbe_bo of space for next curbe entry */
 570       GLuint curbe_next_offset;
 571
 572       /**
 573        * Copy of the last set of CURBEs uploaded.  Frequently we'll end up
 574        * in brw_curbe.c with the same set of constant data to be uploaded,
 575        * so we'd rather not upload new constants in that case (it can cause
 576        * a pipeline bubble since only up to 4 can be pipelined at a time).
 577        */
 578       GLfloat *last_buf;
 579       /**
 580        * Allocation for where to calculate the next set of CURBEs.
 581        * It's a hot enough path that malloc/free of that data matters.
 582        */
 583       GLfloat *next_buf;
 584       GLuint last_bufsz;
 585    } curbe;
 586
 587    struct {
 588       struct brw_vs_prog_data *prog_data;
 589       int8_t *constant_map; /* variable array following prog_data */
 590
 591       drm_intel_bo *prog_bo;
 592       drm_intel_bo *state_bo;
 593       drm_intel_bo *const_bo;
 594
 595       /** Binding table of pointers to surf_bo entries */
 596       drm_intel_bo *bind_bo;
 597       uint32_t bind_bo_offset;
 598       drm_intel_bo *surf_bo[BRW_VS_MAX_SURF];
 599       uint32_t surf_offset[BRW_VS_MAX_SURF];
 600       GLuint nr_surfaces;
 601    } vs;
 602
 603    struct {
 604       struct brw_gs_prog_data *prog_data;
 605
 606       GLboolean prog_active;
 607       drm_intel_bo *prog_bo;
 608       drm_intel_bo *state_bo;
 609    } gs;
 610
 611    struct {
 612       struct brw_clip_prog_data *prog_data;
 613
 614       drm_intel_bo *prog_bo;
 615       drm_intel_bo *state_bo;
 616       drm_intel_bo *vp_bo;
 617    } clip;
 618
 619
 620    struct {
 621       struct brw_sf_prog_data *prog_data;
 622
 623       drm_intel_bo *prog_bo;
 624       drm_intel_bo *state_bo;
 625       drm_intel_bo *vp_bo;
 626    } sf;
 627
 628    struct {
 629       struct brw_wm_prog_data *prog_data;
 630       struct brw_wm_compile *compile_data;
 631
 632       /** Input sizes, calculated from active vertex program.
 633        * One bit per fragment program input attribute.
 634        */
 635       GLbitfield input_size_masks[4];
 636
 637       /** Array of surface default colors (texture border color) */
 638       drm_intel_bo *sdc_bo[BRW_MAX_TEX_UNIT];
 639
 640       GLuint render_surf;
 641       GLuint nr_surfaces;
 642
 643       GLuint max_threads;
 644       drm_intel_bo *scratch_bo;
 645
 646       GLuint sampler_count;
 647       drm_intel_bo *sampler_bo;
 648
 649       /** Binding table of pointers to surf_bo entries */
 650       drm_intel_bo *bind_bo;
 651       uint32_t bind_bo_offset;
 652       drm_intel_bo *surf_bo[BRW_WM_MAX_SURF];
 653       uint32_t surf_offset[BRW_WM_MAX_SURF];
 654
 655       drm_intel_bo *prog_bo;
 656       drm_intel_bo *state_bo;
 657       drm_intel_bo *const_bo;
 658    } wm;
 659
 660
 661    struct {
 662       /* gen4 */
 663       drm_intel_bo *prog_bo;
 664       drm_intel_bo *vp_bo;
 665
 666       /* gen6 */
 667       drm_intel_bo *blend_state_bo;
 668       drm_intel_bo *depth_stencil_state_bo;
 669       drm_intel_bo *color_calc_state_bo;
 670
 671       drm_intel_bo *state_bo;
 672       uint32_t state_offset;
 673    } cc;
 674
 675    struct {
 676       struct brw_query_object *obj;
 677       drm_intel_bo *bo;
 678       int index;
 679       GLboolean active;
 680    } query;
 681    /* Used to give every program string a unique id
 682     */
 683    GLuint program_id;
 684 };
 685
 686
 687 #define BRW_PACKCOLOR8888(r,g,b,a)  ((r<<24) | (g<<16) | (b<<8) | a)
 688
 689
 690
 691 /*======================================================================
 692  * brw_vtbl.c
 693  */
 694 void brwInitVtbl( struct brw_context *brw );
 695
 696 /*======================================================================
 697  * brw_context.c
 698  */
 699 GLboolean brwCreateContext( int api,
 700                             const __GLcontextModes *mesaVis,
 701                             __DRIcontext *driContextPriv,
 702                             void *sharedContextPrivate);
 703
 704 /*======================================================================
 705  * brw_queryobj.c
 706  */
 707 void brw_init_queryobj_functions(struct dd_function_table *functions);
 708 void brw_prepare_query_begin(struct brw_context *brw);
 709 void brw_emit_query_begin(struct brw_context *brw);
 710 void brw_emit_query_end(struct brw_context *brw);
 711
 712 /*======================================================================
 713  * brw_state_dump.c
 714  */
 715 void brw_debug_batch(struct intel_context *intel);
 716
 717 /*======================================================================
 718  * brw_tex.c
 719  */
 720 void brw_validate_textures( struct brw_context *brw );
 721
 722
 723 /*======================================================================
 724  * brw_program.c
 725  */
 726 void brwInitFragProgFuncs( struct dd_function_table *functions );
 727
 728
 729 /* brw_urb.c
 730  */
 731 void brw_upload_urb_fence(struct brw_context *brw);
 732
 733 /* brw_cc.c */
 734 void brw_update_cc_vp(struct brw_context *brw);
 735
 736 /* brw_curbe.c
 737  */
 738 void brw_upload_cs_urb_state(struct brw_context *brw);
 739
 740 /* brw_disasm.c */
 741 int brw_disasm (FILE *file, struct brw_instruction *inst, int gen);
 742
 743 /* brw_state.c */
 744 void brw_enable(GLcontext * ctx, GLenum cap, GLboolean state);
 745 void brw_depth_range(GLcontext *ctx, GLclampd nearval, GLclampd farval);
 746
 747 /*======================================================================
 748  * Inline conversion functions.  These are better-typed than the
 749  * macros used previously:
 750  */
 751 static INLINE struct brw_context *
 752 brw_context( GLcontext *ctx )
 753 {
 754    return (struct brw_context *)ctx;
 755 }
 756
 757 static INLINE struct brw_vertex_program *
 758 brw_vertex_program(struct gl_vertex_program *p)
 759 {
 760    return (struct brw_vertex_program *) p;
 761 }
 762
 763 static INLINE const struct brw_vertex_program *
 764 brw_vertex_program_const(const struct gl_vertex_program *p)
 765 {
 766    return (const struct brw_vertex_program *) p;
 767 }
 768
 769 static INLINE struct brw_fragment_program *
 770 brw_fragment_program(struct gl_fragment_program *p)
 771 {
 772    return (struct brw_fragment_program *) p;
 773 }
 774
 775 static INLINE const struct brw_fragment_program *
 776 brw_fragment_program_const(const struct gl_fragment_program *p)
 777 {
 778    return (const struct brw_fragment_program *) p;
 779 }
 780
 781 #endif
 782