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             0x400000
 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
 175    /** for debugging, which texture units are referenced */
 176    GLbitfield tex_units_used;
 177 };
 178
 179 struct brw_shader {
 180    struct gl_shader base;
 181
 182    /** Shader IR transformed for native compile, at link time. */
 183    struct exec_list *ir;
 184 };
 185
 186 struct brw_shader_program {
 187    struct gl_shader_program base;
 188 };
 189
 190 enum param_conversion {
 191    PARAM_NO_CONVERT,
 192    PARAM_CONVERT_F2I,
 193    PARAM_CONVERT_F2U,
 194    PARAM_CONVERT_F2B,
 195 };
 196
 197 /* Data about a particular attempt to compile a program.  Note that
 198  * there can be many of these, each in a different GL state
 199  * corresponding to a different brw_wm_prog_key struct, with different
 200  * compiled programs:
 201  */
 202 struct brw_wm_prog_data {
 203    GLuint curb_read_length;
 204    GLuint urb_read_length;
 205
 206    GLuint first_curbe_grf;
 207    GLuint total_grf;
 208    GLuint total_scratch;
 209
 210    GLuint nr_params;       /**< number of float params/constants */
 211    GLuint nr_pull_params;
 212    GLboolean error;
 213    int dispatch_width;
 214
 215    /* Pointer to tracked values (only valid once
 216     * _mesa_load_state_parameters has been called at runtime).
 217     */
 218    const float *param[MAX_UNIFORMS * 4]; /* should be: BRW_MAX_CURBE */
 219    enum param_conversion param_convert[MAX_UNIFORMS * 4];
 220    const float *pull_param[MAX_UNIFORMS * 4];
 221    enum param_conversion pull_param_convert[MAX_UNIFORMS * 4];
 222 };
 223
 224 struct brw_sf_prog_data {
 225    GLuint urb_read_length;
 226    GLuint total_grf;
 227
 228    /* Each vertex may have upto 12 attributes, 4 components each,
 229     * except WPOS which requires only 2.  (11*4 + 2) == 44 ==> 11
 230     * rows.
 231     *
 232     * Actually we use 4 for each, so call it 12 rows.
 233     */
 234    GLuint urb_entry_size;
 235 };
 236
 237 struct brw_clip_prog_data {
 238    GLuint curb_read_length;     /* user planes? */
 239    GLuint clip_mode;
 240    GLuint urb_read_length;
 241    GLuint total_grf;
 242 };
 243
 244 struct brw_gs_prog_data {
 245    GLuint urb_read_length;
 246    GLuint total_grf;
 247 };
 248
 249 struct brw_vs_prog_data {
 250    GLuint curb_read_length;
 251    GLuint urb_read_length;
 252    GLuint total_grf;
 253    GLbitfield64 outputs_written;
 254    GLuint nr_params;       /**< number of float params/constants */
 255
 256    GLuint inputs_read;
 257
 258    /* Used for calculating urb partitions:
 259     */
 260    GLuint urb_entry_size;
 261 };
 262
 263
 264 /* Size == 0 if output either not written, or always [0,0,0,1]
 265  */
 266 struct brw_vs_ouput_sizes {
 267    GLubyte output_size[VERT_RESULT_MAX];
 268 };
 269
 270
 271 /** Number of texture sampler units */
 272 #define BRW_MAX_TEX_UNIT 16
 273
 274 /** Max number of render targets in a shader */
 275 #define BRW_MAX_DRAW_BUFFERS 8
 276
 277 /**
 278  * Size of our surface binding table for the WM.
 279  * This contains pointers to the drawing surfaces and current texture
 280  * objects and shader constant buffers (+2).
 281  */
 282 #define BRW_WM_MAX_SURF (BRW_MAX_DRAW_BUFFERS + BRW_MAX_TEX_UNIT + 1)
 283
 284 /**
 285  * Helpers to convert drawing buffers, textures and constant buffers
 286  * to surface binding table indexes, for WM.
 287  */
 288 #define SURF_INDEX_DRAW(d)           (d)
 289 #define SURF_INDEX_FRAG_CONST_BUFFER (BRW_MAX_DRAW_BUFFERS)
 290 #define SURF_INDEX_TEXTURE(t)        (BRW_MAX_DRAW_BUFFERS + 1 + (t))
 291
 292 /**
 293  * Size of surface binding table for the VS.
 294  * Only one constant buffer for now.
 295  */
 296 #define BRW_VS_MAX_SURF 1
 297
 298 /**
 299  * Only a VS constant buffer
 300  */
 301 #define SURF_INDEX_VERT_CONST_BUFFER 0
 302
 303
 304 enum brw_cache_id {
 305    BRW_BLEND_STATE,
 306    BRW_DEPTH_STENCIL_STATE,
 307    BRW_COLOR_CALC_STATE,
 308    BRW_CC_VP,
 309    BRW_CC_UNIT,
 310    BRW_WM_PROG,
 311    BRW_SAMPLER_DEFAULT_COLOR,
 312    BRW_SAMPLER,
 313    BRW_WM_UNIT,
 314    BRW_SF_PROG,
 315    BRW_SF_VP,
 316    BRW_SF_UNIT, /* scissor state on gen6 */
 317    BRW_VS_UNIT,
 318    BRW_VS_PROG,
 319    BRW_GS_UNIT,
 320    BRW_GS_PROG,
 321    BRW_CLIP_VP,
 322    BRW_CLIP_UNIT,
 323    BRW_CLIP_PROG,
 324
 325    BRW_MAX_CACHE
 326 };
 327
 328 struct brw_cache_item {
 329    /**
 330     * Effectively part of the key, cache_id identifies what kind of state
 331     * buffer is involved, and also which brw->state.dirty.cache flag should
 332     * be set when this cache item is chosen.
 333     */
 334    enum brw_cache_id cache_id;
 335    /** 32-bit hash of the key data */
 336    GLuint hash;
 337    GLuint key_size;             /* for variable-sized keys */
 338    const void *key;
 339    drm_intel_bo **reloc_bufs;
 340    GLuint nr_reloc_bufs;
 341
 342    drm_intel_bo *bo;
 343
 344    struct brw_cache_item *next;
 345 };
 346
 347
 348
 349 struct brw_cache {
 350    struct brw_context *brw;
 351
 352    struct brw_cache_item **items;
 353    GLuint size, n_items;
 354
 355    char *name[BRW_MAX_CACHE];
 356
 357    /* Record of the last BOs chosen for each cache_id.  Used to set
 358     * brw->state.dirty.cache when a new cache item is chosen.
 359     */
 360    drm_intel_bo *last_bo[BRW_MAX_CACHE];
 361 };
 362
 363
 364 /* Considered adding a member to this struct to document which flags
 365  * an update might raise so that ordering of the state atoms can be
 366  * checked or derived at runtime.  Dropped the idea in favor of having
 367  * a debug mode where the state is monitored for flags which are
 368  * raised that have already been tested against.
 369  */
 370 struct brw_tracked_state {
 371    struct brw_state_flags dirty;
 372    void (*prepare)( struct brw_context *brw );
 373    void (*emit)( struct brw_context *brw );
 374 };
 375
 376 /* Flags for brw->state.cache.
 377  */
 378 #define CACHE_NEW_BLEND_STATE            (1<<BRW_BLEND_STATE)
 379 #define CACHE_NEW_DEPTH_STENCIL_STATE    (1<<BRW_DEPTH_STENCIL_STATE)
 380 #define CACHE_NEW_COLOR_CALC_STATE       (1<<BRW_COLOR_CALC_STATE)
 381 #define CACHE_NEW_CC_VP                  (1<<BRW_CC_VP)
 382 #define CACHE_NEW_CC_UNIT                (1<<BRW_CC_UNIT)
 383 #define CACHE_NEW_WM_PROG                (1<<BRW_WM_PROG)
 384 #define CACHE_NEW_SAMPLER_DEFAULT_COLOR  (1<<BRW_SAMPLER_DEFAULT_COLOR)
 385 #define CACHE_NEW_SAMPLER                (1<<BRW_SAMPLER)
 386 #define CACHE_NEW_WM_UNIT                (1<<BRW_WM_UNIT)
 387 #define CACHE_NEW_SF_PROG                (1<<BRW_SF_PROG)
 388 #define CACHE_NEW_SF_VP                  (1<<BRW_SF_VP)
 389 #define CACHE_NEW_SF_UNIT                (1<<BRW_SF_UNIT)
 390 #define CACHE_NEW_VS_UNIT                (1<<BRW_VS_UNIT)
 391 #define CACHE_NEW_VS_PROG                (1<<BRW_VS_PROG)
 392 #define CACHE_NEW_GS_UNIT                (1<<BRW_GS_UNIT)
 393 #define CACHE_NEW_GS_PROG                (1<<BRW_GS_PROG)
 394 #define CACHE_NEW_CLIP_VP                (1<<BRW_CLIP_VP)
 395 #define CACHE_NEW_CLIP_UNIT              (1<<BRW_CLIP_UNIT)
 396 #define CACHE_NEW_CLIP_PROG              (1<<BRW_CLIP_PROG)
 397
 398 struct brw_cached_batch_item {
 399    struct header *header;
 400    GLuint sz;
 401    struct brw_cached_batch_item *next;
 402 };
 403
 404
 405
 406 /* Protect against a future where VERT_ATTRIB_MAX > 32.  Wouldn't life
 407  * be easier if C allowed arrays of packed elements?
 408  */
 409 #define ATTRIB_BIT_DWORDS  ((VERT_ATTRIB_MAX+31)/32)
 410
 411 struct brw_vertex_element {
 412    const struct gl_client_array *glarray;
 413
 414    /** The corresponding Mesa vertex attribute */
 415    gl_vert_attrib attrib;
 416    /** Size of a complete element */
 417    GLuint element_size;
 418    /** Number of uploaded elements for this input. */
 419    GLuint count;
 420    /** Byte stride between elements in the uploaded array */
 421    GLuint stride;
 422    /** Offset of the first element within the buffer object */
 423    unsigned int offset;
 424    /** Buffer object containing the uploaded vertex data */
 425    drm_intel_bo *bo;
 426 };
 427
 428
 429
 430 struct brw_vertex_info {
 431    GLuint sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[VERT_ATTRIB_MAX] */
 432 };
 433
 434 struct brw_query_object {
 435    struct gl_query_object Base;
 436
 437    /** Last query BO associated with this query. */
 438    drm_intel_bo *bo;
 439    /** First index in bo with query data for this object. */
 440    int first_index;
 441    /** Last index in bo with query data for this object. */
 442    int last_index;
 443 };
 444
 445
 446 /**
 447  * brw_context is derived from intel_context.
 448  */
 449 struct brw_context
 450 {
 451    struct intel_context intel;  /**< base class, must be first field */
 452    GLuint primitive;
 453
 454    GLboolean emit_state_always;
 455    GLboolean has_surface_tile_offset;
 456    GLboolean has_compr4;
 457    GLboolean has_negative_rhw_bug;
 458    GLboolean has_aa_line_parameters;
 459    GLboolean has_pln;
 460 ;
 461    struct {
 462       struct brw_state_flags dirty;
 463
 464       struct intel_region *depth_region;
 465
 466       /**
 467        * List of buffers accumulated in brw_validate_state to receive
 468        * drm_intel_bo_check_aperture treatment before exec, so we can
 469        * know if we should flush the batch and try again before
 470        * emitting primitives.
 471        *
 472        * This can be a fixed number as we only have a limited number of
 473        * objects referenced from the batchbuffer in a primitive emit,
 474        * consisting of the vertex buffers, pipelined state pointers,
 475        * the CURBE, the depth buffer, and a query BO.
 476        */
 477       drm_intel_bo *validated_bos[VERT_ATTRIB_MAX + BRW_WM_MAX_SURF + 16];
 478       int validated_bo_count;
 479    } state;
 480
 481    struct brw_cache cache;
 482    struct brw_cached_batch_item *cached_batch_items;
 483
 484    struct {
 485       struct brw_vertex_element inputs[VERT_ATTRIB_MAX];
 486
 487       struct brw_vertex_element *enabled[VERT_ATTRIB_MAX];
 488       GLuint nr_enabled;
 489
 490       /* Summary of size and varying of active arrays, so we can check
 491        * for changes to this state:
 492        */
 493       struct brw_vertex_info info;
 494       unsigned int min_index, max_index;
 495    } vb;
 496
 497    struct {
 498       /**
 499        * Index buffer for this draw_prims call.
 500        *
 501        * Updates are signaled by BRW_NEW_INDICES.
 502        */
 503       const struct _mesa_index_buffer *ib;
 504
 505       /* Updates to these fields are signaled by BRW_NEW_INDEX_BUFFER. */
 506       drm_intel_bo *bo;
 507       unsigned int offset;
 508       unsigned int size;
 509       /* Offset to index buffer index to use in CMD_3D_PRIM so that we can
 510        * avoid re-uploading the IB packet over and over if we're actually
 511        * referencing the same index buffer.
 512        */
 513       unsigned int start_vertex_offset;
 514    } ib;
 515
 516    /* Active vertex program:
 517     */
 518    const struct gl_vertex_program *vertex_program;
 519    const struct gl_fragment_program *fragment_program;
 520
 521
 522    /* For populating the gtt:
 523     */
 524    GLuint next_free_page;
 525
 526    /* hw-dependent 3DSTATE_VF_STATISTICS opcode */
 527    uint32_t CMD_VF_STATISTICS;
 528    /* hw-dependent 3DSTATE_PIPELINE_SELECT opcode */
 529    uint32_t CMD_PIPELINE_SELECT;
 530    int vs_max_threads;
 531    int wm_max_threads;
 532
 533    /* BRW_NEW_URB_ALLOCATIONS:
 534     */
 535    struct {
 536       GLuint vsize;             /* vertex size plus header in urb registers */
 537       GLuint csize;             /* constant buffer size in urb registers */
 538       GLuint sfsize;            /* setup data size in urb registers */
 539
 540       GLboolean constrained;
 541
 542       GLuint nr_vs_entries;
 543       GLuint nr_gs_entries;
 544       GLuint nr_clip_entries;
 545       GLuint nr_sf_entries;
 546       GLuint nr_cs_entries;
 547
 548       /* gen6 */
 549       GLuint vs_size;
 550 /*       GLuint gs_size; */
 551 /*       GLuint clip_size; */
 552 /*       GLuint sf_size; */
 553 /*       GLuint cs_size; */
 554
 555       GLuint vs_start;
 556       GLuint gs_start;
 557       GLuint clip_start;
 558       GLuint sf_start;
 559       GLuint cs_start;
 560       GLuint size; /* Hardware URB size, in KB. */
 561    } urb;
 562
 563
 564    /* BRW_NEW_CURBE_OFFSETS:
 565     */
 566    struct {
 567       GLuint wm_start;  /**< pos of first wm const in CURBE buffer */
 568       GLuint wm_size;   /**< number of float[4] consts, multiple of 16 */
 569       GLuint clip_start;
 570       GLuint clip_size;
 571       GLuint vs_start;
 572       GLuint vs_size;
 573       GLuint total_size;
 574
 575       drm_intel_bo *curbe_bo;
 576       /** Offset within curbe_bo of space for current curbe entry */
 577       GLuint curbe_offset;
 578       /** Offset within curbe_bo of space for next curbe entry */
 579       GLuint curbe_next_offset;
 580
 581       /**
 582        * Copy of the last set of CURBEs uploaded.  Frequently we'll end up
 583        * in brw_curbe.c with the same set of constant data to be uploaded,
 584        * so we'd rather not upload new constants in that case (it can cause
 585        * a pipeline bubble since only up to 4 can be pipelined at a time).
 586        */
 587       GLfloat *last_buf;
 588       /**
 589        * Allocation for where to calculate the next set of CURBEs.
 590        * It's a hot enough path that malloc/free of that data matters.
 591        */
 592       GLfloat *next_buf;
 593       GLuint last_bufsz;
 594    } curbe;
 595
 596    struct {
 597       struct brw_vs_prog_data *prog_data;
 598       int8_t *constant_map; /* variable array following prog_data */
 599
 600       drm_intel_bo *prog_bo;
 601       drm_intel_bo *state_bo;
 602       drm_intel_bo *const_bo;
 603
 604       /** Binding table of pointers to surf_bo entries */
 605       uint32_t bind_bo_offset;
 606       uint32_t surf_offset[BRW_VS_MAX_SURF];
 607       GLuint nr_surfaces;
 608    } vs;
 609
 610    struct {
 611       struct brw_gs_prog_data *prog_data;
 612
 613       GLboolean prog_active;
 614       drm_intel_bo *prog_bo;
 615       drm_intel_bo *state_bo;
 616    } gs;
 617
 618    struct {
 619       struct brw_clip_prog_data *prog_data;
 620
 621       drm_intel_bo *prog_bo;
 622       drm_intel_bo *state_bo;
 623       drm_intel_bo *vp_bo;
 624    } clip;
 625
 626
 627    struct {
 628       struct brw_sf_prog_data *prog_data;
 629
 630       drm_intel_bo *prog_bo;
 631       drm_intel_bo *state_bo;
 632       drm_intel_bo *vp_bo;
 633    } sf;
 634
 635    struct {
 636       struct brw_wm_prog_data *prog_data;
 637       struct brw_wm_compile *compile_data;
 638
 639       /** Input sizes, calculated from active vertex program.
 640        * One bit per fragment program input attribute.
 641        */
 642       GLbitfield input_size_masks[4];
 643
 644       /** Array of surface default colors (texture border color) */
 645       drm_intel_bo *sdc_bo[BRW_MAX_TEX_UNIT];
 646
 647       GLuint render_surf;
 648       GLuint nr_surfaces;
 649
 650       GLuint max_threads;
 651       drm_intel_bo *scratch_bo;
 652
 653       GLuint sampler_count;
 654       drm_intel_bo *sampler_bo;
 655
 656       /** Binding table of pointers to surf_bo entries */
 657       uint32_t bind_bo_offset;
 658       uint32_t surf_offset[BRW_WM_MAX_SURF];
 659
 660       drm_intel_bo *prog_bo;
 661       drm_intel_bo *state_bo;
 662       drm_intel_bo *const_bo; /* pull constant buffer. */
 663       /**
 664        *  This is the push constant BO on gen6.
 665        *
 666        * Pre-gen6, push constants live in the CURBE.
 667        */
 668       drm_intel_bo *push_const_bo;
 669    } wm;
 670
 671
 672    struct {
 673       /* gen4 */
 674       drm_intel_bo *prog_bo;
 675       drm_intel_bo *vp_bo;
 676
 677       /* gen6 */
 678       drm_intel_bo *blend_state_bo;
 679       drm_intel_bo *depth_stencil_state_bo;
 680       drm_intel_bo *color_calc_state_bo;
 681
 682       uint32_t state_offset;
 683    } cc;
 684
 685    struct {
 686       struct brw_query_object *obj;
 687       drm_intel_bo *bo;
 688       int index;
 689       GLboolean active;
 690    } query;
 691    /* Used to give every program string a unique id
 692     */
 693    GLuint program_id;
 694 };
 695
 696
 697 #define BRW_PACKCOLOR8888(r,g,b,a)  ((r<<24) | (g<<16) | (b<<8) | a)
 698
 699 struct brw_instruction_info {
 700     char    *name;
 701     int     nsrc;
 702     int     ndst;
 703     GLboolean is_arith;
 704 };
 705 extern const struct brw_instruction_info brw_opcodes[128];
 706
 707 /*======================================================================
 708  * brw_vtbl.c
 709  */
 710 void brwInitVtbl( struct brw_context *brw );
 711
 712 /*======================================================================
 713  * brw_context.c
 714  */
 715 GLboolean brwCreateContext( int api,
 716                             const struct gl_config *mesaVis,
 717                             __DRIcontext *driContextPriv,
 718                             void *sharedContextPrivate);
 719
 720 /*======================================================================
 721  * brw_queryobj.c
 722  */
 723 void brw_init_queryobj_functions(struct dd_function_table *functions);
 724 void brw_prepare_query_begin(struct brw_context *brw);
 725 void brw_emit_query_begin(struct brw_context *brw);
 726 void brw_emit_query_end(struct brw_context *brw);
 727
 728 /*======================================================================
 729  * brw_state_dump.c
 730  */
 731 void brw_debug_batch(struct intel_context *intel);
 732
 733 /*======================================================================
 734  * brw_tex.c
 735  */
 736 void brw_validate_textures( struct brw_context *brw );
 737
 738
 739 /*======================================================================
 740  * brw_program.c
 741  */
 742 void brwInitFragProgFuncs( struct dd_function_table *functions );
 743
 744
 745 /* brw_urb.c
 746  */
 747 void brw_upload_urb_fence(struct brw_context *brw);
 748
 749 /* brw_cc.c */
 750 void brw_update_cc_vp(struct brw_context *brw);
 751
 752 /* brw_curbe.c
 753  */
 754 void brw_upload_cs_urb_state(struct brw_context *brw);
 755
 756 /* brw_disasm.c */
 757 int brw_disasm (FILE *file, struct brw_instruction *inst, int gen);
 758
 759 /* brw_state.c */
 760 void brw_enable(struct gl_context * ctx, GLenum cap, GLboolean state);
 761 void brw_depth_range(struct gl_context *ctx, GLclampd nearval, GLclampd farval);
 762
 763 /*======================================================================
 764  * Inline conversion functions.  These are better-typed than the
 765  * macros used previously:
 766  */
 767 static INLINE struct brw_context *
 768 brw_context( struct gl_context *ctx )
 769 {
 770    return (struct brw_context *)ctx;
 771 }
 772
 773 static INLINE struct brw_vertex_program *
 774 brw_vertex_program(struct gl_vertex_program *p)
 775 {
 776    return (struct brw_vertex_program *) p;
 777 }
 778
 779 static INLINE const struct brw_vertex_program *
 780 brw_vertex_program_const(const struct gl_vertex_program *p)
 781 {
 782    return (const struct brw_vertex_program *) p;
 783 }
 784
 785 static INLINE struct brw_fragment_program *
 786 brw_fragment_program(struct gl_fragment_program *p)
 787 {
 788    return (struct brw_fragment_program *) p;
 789 }
 790
 791 static INLINE const struct brw_fragment_program *
 792 brw_fragment_program_const(const struct gl_fragment_program *p)
 793 {
 794    return (const struct brw_fragment_program *) p;
 795 }
 796
 797 static inline
 798 float convert_param(enum param_conversion conversion, float param)
 799 {
 800    union {
 801       float f;
 802       uint32_t u;
 803       int32_t i;
 804    } fi;
 805
 806    switch (conversion) {
 807    case PARAM_NO_CONVERT:
 808       return param;
 809    case PARAM_CONVERT_F2I:
 810       fi.i = param;
 811       return fi.f;
 812    case PARAM_CONVERT_F2U:
 813       fi.u = param;
 814       return fi.f;
 815    case PARAM_CONVERT_F2B:
 816       if (param != 0.0)
 817          fi.i = 1;
 818       else
 819          fi.i = 0;
 820       return fi.f;
 821    default:
 822       return param;
 823    }
 824 }
 825
 826 GLboolean brw_do_cubemap_normalize(struct exec_list *instructions);
 827
 828 #endif
 829