src/mesa/drivers/dri/i915/intel_tris.c

   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 /** @file intel_tris.c
  29  *
  30  * This file contains functions for managing the vertex buffer and emitting
  31  * primitives into it.
  32  */
  33
  34 #include "main/glheader.h"
  35 #include "main/context.h"
  36 #include "main/macros.h"
  37 #include "main/enums.h"
  38 #include "main/texobj.h"
  39 #include "main/state.h"
  40 #include "main/dd.h"
  41
  42 #include "swrast/swrast.h"
  43 #include "swrast_setup/swrast_setup.h"
  44 #include "tnl/t_context.h"
  45 #include "tnl/t_pipeline.h"
  46 #include "tnl/t_vertex.h"
  47
  48 #include "intel_screen.h"
  49 #include "intel_context.h"
  50 #include "intel_tris.h"
  51 #include "intel_batchbuffer.h"
  52 #include "intel_buffers.h"
  53 #include "intel_reg.h"
  54 #include "intel_span.h"
  55 #include "i830_context.h"
  56 #include "i830_reg.h"
  57
  58 static void intelRenderPrimitive(GLcontext * ctx, GLenum prim);
  59 static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim,
  60                                  GLuint hwprim);
  61
  62 static void
  63 intel_flush_inline_primitive(struct intel_context *intel)
  64 {
  65    GLuint used = intel->batch->ptr - intel->prim.start_ptr;
  66
  67    assert(intel->prim.primitive != ~0);
  68
  69 /*    printf("/\n"); */
  70
  71    if (used < 8)
  72       goto do_discard;
  73
  74    *(int *) intel->prim.start_ptr = (_3DPRIMITIVE |
  75                                      intel->prim.primitive | (used / 4 - 2));
  76
  77    goto finished;
  78
  79  do_discard:
  80    intel->batch->ptr -= used;
  81
  82  finished:
  83    intel->prim.primitive = ~0;
  84    intel->prim.start_ptr = 0;
  85    intel->prim.flush = 0;
  86 }
  87
  88 static void intel_start_inline(struct intel_context *intel, uint32_t prim)
  89 {
  90    BATCH_LOCALS;
  91
  92    intel->vtbl.emit_state(intel);
  93
  94    intel->no_batch_wrap = GL_TRUE;
  95
  96    /*printf("%s *", __progname);*/
  97
  98    /* Emit a slot which will be filled with the inline primitive
  99     * command later.
 100     */
 101    BEGIN_BATCH(1);
 102
 103    assert((intel->batch->dirty_state & (1<<1)) == 0);
 104
 105    intel->prim.start_ptr = intel->batch->ptr;
 106    intel->prim.primitive = prim;
 107    intel->prim.flush = intel_flush_inline_primitive;
 108
 109    OUT_BATCH(0);
 110    ADVANCE_BATCH();
 111
 112    intel->no_batch_wrap = GL_FALSE;
 113 /*    printf(">"); */
 114 }
 115
 116 static void intel_wrap_inline(struct intel_context *intel)
 117 {
 118    GLuint prim = intel->prim.primitive;
 119
 120    intel_flush_inline_primitive(intel);
 121    intel_batchbuffer_flush(intel->batch);
 122    intel_start_inline(intel, prim);  /* ??? */
 123 }
 124
 125 static GLuint *intel_extend_inline(struct intel_context *intel, GLuint dwords)
 126 {
 127    GLuint sz = dwords * sizeof(GLuint);
 128    GLuint *ptr;
 129
 130    assert(intel->prim.flush == intel_flush_inline_primitive);
 131
 132    if (intel_batchbuffer_space(intel->batch) < sz)
 133       intel_wrap_inline(intel);
 134
 135 /*    printf("."); */
 136
 137    intel->vtbl.assert_not_dirty(intel);
 138
 139    ptr = (GLuint *) intel->batch->ptr;
 140    intel->batch->ptr += sz;
 141
 142    return ptr;
 143 }
 144
 145 /** Sets the primitive type for a primitive sequence, flushing as needed. */
 146 void intel_set_prim(struct intel_context *intel, uint32_t prim)
 147 {
 148    /* if we have no VBOs */
 149
 150    if (intel->intelScreen->no_vbo) {
 151       intel_start_inline(intel, prim);
 152       return;
 153    }
 154    if (prim != intel->prim.primitive) {
 155       INTEL_FIREVERTICES(intel);
 156       intel->prim.primitive = prim;
 157    }
 158 }
 159
 160 /** Returns mapped VB space for the given number of vertices */
 161 uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
 162 {
 163    uint32_t *addr;
 164
 165    if (intel->intelScreen->no_vbo) {
 166       return intel_extend_inline(intel, count * intel->vertex_size);
 167    }
 168
 169    /* Check for space in the existing VB */
 170    if (intel->prim.vb_bo == NULL ||
 171        (intel->prim.current_offset +
 172         count * intel->vertex_size * 4) > INTEL_VB_SIZE ||
 173        (intel->prim.count + count) >= (1 << 16)) {
 174       /* Flush existing prim if any */
 175       INTEL_FIREVERTICES(intel);
 176
 177       intel_finish_vb(intel);
 178
 179       /* Start a new VB */
 180       if (intel->prim.vb == NULL)
 181          intel->prim.vb = malloc(INTEL_VB_SIZE);
 182       intel->prim.vb_bo = dri_bo_alloc(intel->bufmgr, "vb",
 183                                        INTEL_VB_SIZE, 4);
 184       intel->prim.start_offset = 0;
 185       intel->prim.current_offset = 0;
 186    }
 187
 188    intel->prim.flush = intel_flush_prim;
 189
 190    addr = (uint32_t *)(intel->prim.vb + intel->prim.current_offset);
 191    intel->prim.current_offset += intel->vertex_size * 4 * count;
 192    intel->prim.count += count;
 193
 194    return addr;
 195 }
 196
 197 /** Dispatches the accumulated primitive to the batchbuffer. */
 198 void intel_flush_prim(struct intel_context *intel)
 199 {
 200    dri_bo *aper_array[2];
 201    dri_bo *vb_bo;
 202    unsigned int offset, count;
 203    BATCH_LOCALS;
 204
 205    /* Must be called after an intel_start_prim. */
 206    assert(intel->prim.primitive != ~0);
 207
 208    if (intel->prim.count == 0)
 209       return;
 210
 211    /* Clear the current prims out of the context state so that a batch flush
 212     * flush triggered by emit_state doesn't loop back to flush_prim again.
 213     */
 214    vb_bo = intel->prim.vb_bo;
 215    dri_bo_reference(vb_bo);
 216    count = intel->prim.count;
 217    intel->prim.count = 0;
 218    offset = intel->prim.start_offset;
 219    intel->prim.start_offset = intel->prim.current_offset;
 220    if (!intel->gen >= 3)
 221       intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128);
 222    intel->prim.flush = NULL;
 223
 224    intel->vtbl.emit_state(intel);
 225
 226    aper_array[0] = intel->batch->buf;
 227    aper_array[1] = vb_bo;
 228    if (dri_bufmgr_check_aperture_space(aper_array, 2)) {
 229       intel_batchbuffer_flush(intel->batch);
 230       intel->vtbl.emit_state(intel);
 231    }
 232
 233    /* Ensure that we don't start a new batch for the following emit, which
 234     * depends on the state just emitted. emit_state should be making sure we
 235     * have the space for this.
 236     */
 237    intel->no_batch_wrap = GL_TRUE;
 238
 239    /* Check that we actually emitted the state into this batch, using the
 240     * UPLOAD_CTX bit as the signal.
 241     */
 242    assert((intel->batch->dirty_state & (1<<1)) == 0);
 243
 244 #if 0
 245    printf("emitting %d..%d=%d vertices size %d\n", offset,
 246           intel->prim.current_offset, count,
 247           intel->vertex_size * 4);
 248 #endif
 249
 250    if (intel->gen >= 3) {
 251       BEGIN_BATCH(5);
 252       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
 253                 I1_LOAD_S(0) | I1_LOAD_S(1) | 1);
 254       assert((offset & !S0_VB_OFFSET_MASK) == 0);
 255       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, offset);
 256       OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) |
 257                 (intel->vertex_size << S1_VERTEX_PITCH_SHIFT));
 258
 259       OUT_BATCH(_3DPRIMITIVE |
 260                 PRIM_INDIRECT |
 261                 PRIM_INDIRECT_SEQUENTIAL |
 262                 intel->prim.primitive |
 263                 count);
 264       OUT_BATCH(0); /* Beginning vertex index */
 265       ADVANCE_BATCH();
 266    } else {
 267       struct i830_context *i830 = i830_context(&intel->ctx);
 268
 269       BEGIN_BATCH(5);
 270       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
 271                 I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
 272       /* S0 */
 273       assert((offset & !S0_VB_OFFSET_MASK_830) == 0);
 274       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
 275                 offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
 276                 S0_VB_ENABLE_830);
 277       /* S2
 278        * This is somewhat unfortunate -- VB width is tied up with
 279        * vertex format data that we've already uploaded through
 280        * _3DSTATE_VFT[01]_CMD.  We may want to replace emits of VFT state with
 281        * STATE_IMMEDIATE_1 like this to avoid duplication.
 282        */
 283       OUT_BATCH((i830->state.Ctx[I830_CTXREG_VF] & VFT0_TEX_COUNT_MASK) >>
 284                 VFT0_TEX_COUNT_SHIFT << S2_TEX_COUNT_SHIFT_830 |
 285                 (i830->state.Ctx[I830_CTXREG_VF2] << 16) |
 286                 intel->vertex_size << S2_VERTEX_0_WIDTH_SHIFT_830);
 287
 288       OUT_BATCH(_3DPRIMITIVE |
 289                 PRIM_INDIRECT |
 290                 PRIM_INDIRECT_SEQUENTIAL |
 291                 intel->prim.primitive |
 292                 count);
 293       OUT_BATCH(0); /* Beginning vertex index */
 294       ADVANCE_BATCH();
 295    }
 296
 297    intel->no_batch_wrap = GL_FALSE;
 298
 299    dri_bo_unreference(vb_bo);
 300 }
 301
 302 /**
 303  * Uploads the locally-accumulated VB into the buffer object.
 304  *
 305  * This avoids us thrashing the cachelines in and out as the buffer gets
 306  * filled, dispatched, then reused as the hardware completes rendering from it,
 307  * and also lets us clflush less if we dispatch with a partially-filled VB.
 308  *
 309  * This is called normally from get_space when we're finishing a BO, but also
 310  * at batch flush time so that we don't try accessing the contents of a
 311  * just-dispatched buffer.
 312  */
 313 void intel_finish_vb(struct intel_context *intel)
 314 {
 315    if (intel->prim.vb_bo == NULL)
 316       return;
 317
 318    dri_bo_subdata(intel->prim.vb_bo, 0, intel->prim.start_offset,
 319                   intel->prim.vb);
 320    dri_bo_unreference(intel->prim.vb_bo);
 321    intel->prim.vb_bo = NULL;
 322 }
 323
 324 /***********************************************************************
 325  *                    Emit primitives as inline vertices               *
 326  ***********************************************************************/
 327
 328 #ifdef __i386__
 329 #define COPY_DWORDS( j, vb, vertsize, v )                       \
 330 do {                                                            \
 331    int __tmp;                                                   \
 332    __asm__ __volatile__( "rep ; movsl"                          \
 333                          : "=%c" (j), "=D" (vb), "=S" (__tmp)   \
 334                          : "0" (vertsize),                      \
 335                          "D" ((long)vb),                        \
 336                          "S" ((long)v) );                       \
 337 } while (0)
 338 #else
 339 #define COPY_DWORDS( j, vb, vertsize, v )       \
 340 do {                                            \
 341    for ( j = 0 ; j < vertsize ; j++ ) {         \
 342       vb[j] = ((GLuint *)v)[j];                 \
 343    }                                            \
 344    vb += vertsize;                              \
 345 } while (0)
 346 #endif
 347
 348 static void
 349 intel_draw_quad(struct intel_context *intel,
 350                 intelVertexPtr v0,
 351                 intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3)
 352 {
 353    GLuint vertsize = intel->vertex_size;
 354    GLuint *vb = intel_get_prim_space(intel, 6);
 355    int j;
 356
 357    COPY_DWORDS(j, vb, vertsize, v0);
 358    COPY_DWORDS(j, vb, vertsize, v1);
 359
 360    /* If smooth shading, draw like a trifan which gives better
 361     * rasterization.  Otherwise draw as two triangles with provoking
 362     * vertex in third position as required for flat shading.
 363     */
 364    if (intel->ctx.Light.ShadeModel == GL_FLAT) {
 365       COPY_DWORDS(j, vb, vertsize, v3);
 366       COPY_DWORDS(j, vb, vertsize, v1);
 367    }
 368    else {
 369       COPY_DWORDS(j, vb, vertsize, v2);
 370       COPY_DWORDS(j, vb, vertsize, v0);
 371    }
 372
 373    COPY_DWORDS(j, vb, vertsize, v2);
 374    COPY_DWORDS(j, vb, vertsize, v3);
 375 }
 376
 377 static void
 378 intel_draw_triangle(struct intel_context *intel,
 379                     intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
 380 {
 381    GLuint vertsize = intel->vertex_size;
 382    GLuint *vb = intel_get_prim_space(intel, 3);
 383    int j;
 384
 385    COPY_DWORDS(j, vb, vertsize, v0);
 386    COPY_DWORDS(j, vb, vertsize, v1);
 387    COPY_DWORDS(j, vb, vertsize, v2);
 388 }
 389
 390
 391 static void
 392 intel_draw_line(struct intel_context *intel,
 393                 intelVertexPtr v0, intelVertexPtr v1)
 394 {
 395    GLuint vertsize = intel->vertex_size;
 396    GLuint *vb = intel_get_prim_space(intel, 2);
 397    int j;
 398
 399    COPY_DWORDS(j, vb, vertsize, v0);
 400    COPY_DWORDS(j, vb, vertsize, v1);
 401 }
 402
 403
 404 static void
 405 intel_draw_point(struct intel_context *intel, intelVertexPtr v0)
 406 {
 407    GLuint vertsize = intel->vertex_size;
 408    GLuint *vb = intel_get_prim_space(intel, 1);
 409    int j;
 410
 411    /* Adjust for sub pixel position -- still required for conform. */
 412    *(float *) &vb[0] = v0->v.x;
 413    *(float *) &vb[1] = v0->v.y;
 414    for (j = 2; j < vertsize; j++)
 415       vb[j] = v0->ui[j];
 416 }
 417
 418
 419
 420 /***********************************************************************
 421  *                Fixup for ARB_point_parameters                       *
 422  ***********************************************************************/
 423
 424 /* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly
 425  * represented in the fragment program InputsRead field.
 426  */
 427 static void
 428 intel_atten_point(struct intel_context *intel, intelVertexPtr v0)
 429 {
 430    GLcontext *ctx = &intel->ctx;
 431    GLfloat psz[4], col[4], restore_psz, restore_alpha;
 432
 433    _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 434    _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 435
 436    restore_psz = psz[0];
 437    restore_alpha = col[3];
 438
 439    if (psz[0] >= ctx->Point.Threshold) {
 440       psz[0] = MIN2(psz[0], ctx->Point.MaxSize);
 441    }
 442    else {
 443       GLfloat dsize = psz[0] / ctx->Point.Threshold;
 444       psz[0] = MAX2(ctx->Point.Threshold, ctx->Point.MinSize);
 445       col[3] *= dsize * dsize;
 446    }
 447
 448    if (psz[0] < 1.0)
 449       psz[0] = 1.0;
 450
 451    if (restore_psz != psz[0] || restore_alpha != col[3]) {
 452       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 453       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 454
 455       intel_draw_point(intel, v0);
 456
 457       psz[0] = restore_psz;
 458       col[3] = restore_alpha;
 459
 460       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 461       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 462    }
 463    else
 464       intel_draw_point(intel, v0);
 465 }
 466
 467
 468
 469
 470
 471 /***********************************************************************
 472  *                Fixup for I915 WPOS texture coordinate                *
 473  ***********************************************************************/
 474
 475
 476
 477 static void
 478 intel_wpos_triangle(struct intel_context *intel,
 479                     intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
 480 {
 481    GLuint offset = intel->wpos_offset;
 482    GLuint size = intel->wpos_size;
 483    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 484    GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
 485    GLfloat *v2_wpos = (GLfloat *)((char *)v2 + offset);
 486
 487    __memcpy(v0_wpos, v0, size);
 488    __memcpy(v1_wpos, v1, size);
 489    __memcpy(v2_wpos, v2, size);
 490
 491    v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
 492    v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
 493    v2_wpos[1] = -v2_wpos[1] + intel->driDrawable->h;
 494
 495
 496    intel_draw_triangle(intel, v0, v1, v2);
 497 }
 498
 499
 500 static void
 501 intel_wpos_line(struct intel_context *intel,
 502                 intelVertexPtr v0, intelVertexPtr v1)
 503 {
 504    GLuint offset = intel->wpos_offset;
 505    GLuint size = intel->wpos_size;
 506    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 507    GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
 508
 509    __memcpy(v0_wpos, v0, size);
 510    __memcpy(v1_wpos, v1, size);
 511
 512    v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
 513    v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
 514
 515    intel_draw_line(intel, v0, v1);
 516 }
 517
 518
 519 static void
 520 intel_wpos_point(struct intel_context *intel, intelVertexPtr v0)
 521 {
 522    GLuint offset = intel->wpos_offset;
 523    GLuint size = intel->wpos_size;
 524    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 525
 526    __memcpy(v0_wpos, v0, size);
 527    v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
 528
 529    intel_draw_point(intel, v0);
 530 }
 531
 532
 533
 534
 535
 536
 537 /***********************************************************************
 538  *          Macros for t_dd_tritmp.h to draw basic primitives          *
 539  ***********************************************************************/
 540
 541 #define TRI( a, b, c )                          \
 542 do {                                            \
 543    if (DO_FALLBACK)                             \
 544       intel->draw_tri( intel, a, b, c );        \
 545    else                                         \
 546       intel_draw_triangle( intel, a, b, c );    \
 547 } while (0)
 548
 549 #define QUAD( a, b, c, d )                      \
 550 do {                                            \
 551    if (DO_FALLBACK) {                           \
 552       intel->draw_tri( intel, a, b, d );        \
 553       intel->draw_tri( intel, b, c, d );        \
 554    } else                                       \
 555       intel_draw_quad( intel, a, b, c, d );     \
 556 } while (0)
 557
 558 #define LINE( v0, v1 )                          \
 559 do {                                            \
 560    if (DO_FALLBACK)                             \
 561       intel->draw_line( intel, v0, v1 );        \
 562    else                                         \
 563       intel_draw_line( intel, v0, v1 );         \
 564 } while (0)
 565
 566 #define POINT( v0 )                             \
 567 do {                                            \
 568    if (DO_FALLBACK)                             \
 569       intel->draw_point( intel, v0 );           \
 570    else                                         \
 571       intel_draw_point( intel, v0 );            \
 572 } while (0)
 573
 574
 575 /***********************************************************************
 576  *              Build render functions from dd templates               *
 577  ***********************************************************************/
 578
 579 #define INTEL_OFFSET_BIT        0x01
 580 #define INTEL_TWOSIDE_BIT       0x02
 581 #define INTEL_UNFILLED_BIT      0x04
 582 #define INTEL_FALLBACK_BIT      0x08
 583 #define INTEL_MAX_TRIFUNC       0x10
 584
 585
 586 static struct
 587 {
 588    tnl_points_func points;
 589    tnl_line_func line;
 590    tnl_triangle_func triangle;
 591    tnl_quad_func quad;
 592 } rast_tab[INTEL_MAX_TRIFUNC];
 593
 594
 595 #define DO_FALLBACK (IND & INTEL_FALLBACK_BIT)
 596 #define DO_OFFSET   (IND & INTEL_OFFSET_BIT)
 597 #define DO_UNFILLED (IND & INTEL_UNFILLED_BIT)
 598 #define DO_TWOSIDE  (IND & INTEL_TWOSIDE_BIT)
 599 #define DO_FLAT      0
 600 #define DO_TRI       1
 601 #define DO_QUAD      1
 602 #define DO_LINE      1
 603 #define DO_POINTS    1
 604 #define DO_FULL_QUAD 1
 605
 606 #define HAVE_RGBA         1
 607 #define HAVE_SPEC         1
 608 #define HAVE_BACK_COLORS  0
 609 #define HAVE_HW_FLATSHADE 1
 610 #define VERTEX            intelVertex
 611 #define TAB               rast_tab
 612
 613 /* Only used to pull back colors into vertices (ie, we know color is
 614  * floating point).
 615  */
 616 #define INTEL_COLOR( dst, src )                         \
 617 do {                                                    \
 618    UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);        \
 619    UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);        \
 620    UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);        \
 621    UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]);        \
 622 } while (0)
 623
 624 #define INTEL_SPEC( dst, src )                          \
 625 do {                                                    \
 626    UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);        \
 627    UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);        \
 628    UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);        \
 629 } while (0)
 630
 631
 632 #define DEPTH_SCALE intel->polygon_offset_scale
 633 #define UNFILLED_TRI unfilled_tri
 634 #define UNFILLED_QUAD unfilled_quad
 635 #define VERT_X(_v) _v->v.x
 636 #define VERT_Y(_v) _v->v.y
 637 #define VERT_Z(_v) _v->v.z
 638 #define AREA_IS_CCW( a ) (a > 0)
 639 #define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint)))
 640
 641 #define VERT_SET_RGBA( v, c )    if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c )
 642 #define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
 643 #define VERT_SAVE_RGBA( idx )    if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
 644 #define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
 645
 646 #define VERT_SET_SPEC( v, c )    if (specoffset) INTEL_SPEC( v->ub4[specoffset], c )
 647 #define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
 648 #define VERT_SAVE_SPEC( idx )    if (specoffset) spec[idx] = v[idx]->ui[specoffset]
 649 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
 650
 651 #define LOCAL_VARS(n)                                                   \
 652    struct intel_context *intel = intel_context(ctx);                    \
 653    GLuint color[n] = { 0, }, spec[n] = { 0, };                          \
 654    GLuint coloroffset = intel->coloroffset;                             \
 655    GLboolean specoffset = intel->specoffset;                            \
 656    (void) color; (void) spec; (void) coloroffset; (void) specoffset;
 657
 658
 659 /***********************************************************************
 660  *                Helpers for rendering unfilled primitives            *
 661  ***********************************************************************/
 662
 663 static const GLuint hw_prim[GL_POLYGON + 1] = {
 664    PRIM3D_POINTLIST,
 665    PRIM3D_LINELIST,
 666    PRIM3D_LINELIST,
 667    PRIM3D_LINELIST,
 668    PRIM3D_TRILIST,
 669    PRIM3D_TRILIST,
 670    PRIM3D_TRILIST,
 671    PRIM3D_TRILIST,
 672    PRIM3D_TRILIST,
 673    PRIM3D_TRILIST
 674 };
 675
 676 #define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] )
 677 #define RENDER_PRIMITIVE intel->render_primitive
 678 #define TAG(x) x
 679 #define IND INTEL_FALLBACK_BIT
 680 #include "tnl_dd/t_dd_unfilled.h"
 681 #undef IND
 682
 683 /***********************************************************************
 684  *                      Generate GL render functions                   *
 685  ***********************************************************************/
 686
 687 #define IND (0)
 688 #define TAG(x) x
 689 #include "tnl_dd/t_dd_tritmp.h"
 690
 691 #define IND (INTEL_OFFSET_BIT)
 692 #define TAG(x) x##_offset
 693 #include "tnl_dd/t_dd_tritmp.h"
 694
 695 #define IND (INTEL_TWOSIDE_BIT)
 696 #define TAG(x) x##_twoside
 697 #include "tnl_dd/t_dd_tritmp.h"
 698
 699 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT)
 700 #define TAG(x) x##_twoside_offset
 701 #include "tnl_dd/t_dd_tritmp.h"
 702
 703 #define IND (INTEL_UNFILLED_BIT)
 704 #define TAG(x) x##_unfilled
 705 #include "tnl_dd/t_dd_tritmp.h"
 706
 707 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
 708 #define TAG(x) x##_offset_unfilled
 709 #include "tnl_dd/t_dd_tritmp.h"
 710
 711 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT)
 712 #define TAG(x) x##_twoside_unfilled
 713 #include "tnl_dd/t_dd_tritmp.h"
 714
 715 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
 716 #define TAG(x) x##_twoside_offset_unfilled
 717 #include "tnl_dd/t_dd_tritmp.h"
 718
 719 #define IND (INTEL_FALLBACK_BIT)
 720 #define TAG(x) x##_fallback
 721 #include "tnl_dd/t_dd_tritmp.h"
 722
 723 #define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
 724 #define TAG(x) x##_offset_fallback
 725 #include "tnl_dd/t_dd_tritmp.h"
 726
 727 #define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT)
 728 #define TAG(x) x##_twoside_fallback
 729 #include "tnl_dd/t_dd_tritmp.h"
 730
 731 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
 732 #define TAG(x) x##_twoside_offset_fallback
 733 #include "tnl_dd/t_dd_tritmp.h"
 734
 735 #define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 736 #define TAG(x) x##_unfilled_fallback
 737 #include "tnl_dd/t_dd_tritmp.h"
 738
 739 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 740 #define TAG(x) x##_offset_unfilled_fallback
 741 #include "tnl_dd/t_dd_tritmp.h"
 742
 743 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 744 #define TAG(x) x##_twoside_unfilled_fallback
 745 #include "tnl_dd/t_dd_tritmp.h"
 746
 747 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \
 748              INTEL_FALLBACK_BIT)
 749 #define TAG(x) x##_twoside_offset_unfilled_fallback
 750 #include "tnl_dd/t_dd_tritmp.h"
 751
 752
 753 static void
 754 init_rast_tab(void)
 755 {
 756    init();
 757    init_offset();
 758    init_twoside();
 759    init_twoside_offset();
 760    init_unfilled();
 761    init_offset_unfilled();
 762    init_twoside_unfilled();
 763    init_twoside_offset_unfilled();
 764    init_fallback();
 765    init_offset_fallback();
 766    init_twoside_fallback();
 767    init_twoside_offset_fallback();
 768    init_unfilled_fallback();
 769    init_offset_unfilled_fallback();
 770    init_twoside_unfilled_fallback();
 771    init_twoside_offset_unfilled_fallback();
 772 }
 773
 774
 775 /***********************************************************************
 776  *                    Rasterization fallback helpers                   *
 777  ***********************************************************************/
 778
 779
 780 /* This code is hit only when a mix of accelerated and unaccelerated
 781  * primitives are being drawn, and only for the unaccelerated
 782  * primitives.
 783  */
 784 static void
 785 intel_fallback_tri(struct intel_context *intel,
 786                    intelVertex * v0, intelVertex * v1, intelVertex * v2)
 787 {
 788    GLcontext *ctx = &intel->ctx;
 789    SWvertex v[3];
 790
 791    if (0)
 792       fprintf(stderr, "\n%s\n", __FUNCTION__);
 793
 794    INTEL_FIREVERTICES(intel);
 795
 796    _swsetup_Translate(ctx, v0, &v[0]);
 797    _swsetup_Translate(ctx, v1, &v[1]);
 798    _swsetup_Translate(ctx, v2, &v[2]);
 799    intelSpanRenderStart(ctx);
 800    _swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
 801    intelSpanRenderFinish(ctx);
 802 }
 803
 804
 805 static void
 806 intel_fallback_line(struct intel_context *intel,
 807                     intelVertex * v0, intelVertex * v1)
 808 {
 809    GLcontext *ctx = &intel->ctx;
 810    SWvertex v[2];
 811
 812    if (0)
 813       fprintf(stderr, "\n%s\n", __FUNCTION__);
 814
 815    INTEL_FIREVERTICES(intel);
 816
 817    _swsetup_Translate(ctx, v0, &v[0]);
 818    _swsetup_Translate(ctx, v1, &v[1]);
 819    intelSpanRenderStart(ctx);
 820    _swrast_Line(ctx, &v[0], &v[1]);
 821    intelSpanRenderFinish(ctx);
 822 }
 823
 824 static void
 825 intel_fallback_point(struct intel_context *intel,
 826                      intelVertex * v0)
 827 {
 828    GLcontext *ctx = &intel->ctx;
 829    SWvertex v[1];
 830
 831    if (0)
 832       fprintf(stderr, "\n%s\n", __FUNCTION__);
 833
 834    INTEL_FIREVERTICES(intel);
 835
 836    _swsetup_Translate(ctx, v0, &v[0]);
 837    intelSpanRenderStart(ctx);
 838    _swrast_Point(ctx, &v[0]);
 839    intelSpanRenderFinish(ctx);
 840 }
 841
 842
 843 /**********************************************************************/
 844 /*               Render unclipped begin/end objects                   */
 845 /**********************************************************************/
 846
 847 #define IND 0
 848 #define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint)))
 849 #define RENDER_POINTS( start, count )   \
 850    for ( ; start < count ; start++) POINT( V(ELT(start)) );
 851 #define RENDER_LINE( v0, v1 )         LINE( V(v0), V(v1) )
 852 #define RENDER_TRI(  v0, v1, v2 )     TRI(  V(v0), V(v1), V(v2) )
 853 #define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) )
 854 #define INIT(x) intelRenderPrimitive( ctx, x )
 855 #undef LOCAL_VARS
 856 #define LOCAL_VARS                                              \
 857     struct intel_context *intel = intel_context(ctx);                   \
 858     GLubyte *vertptr = (GLubyte *)intel->verts;                 \
 859     const GLuint vertsize = intel->vertex_size;         \
 860     const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;       \
 861     (void) elt;
 862 #define RESET_STIPPLE
 863 #define RESET_OCCLUSION
 864 #define PRESERVE_VB_DEFS
 865 #define ELT(x) x
 866 #define TAG(x) intel_##x##_verts
 867 #include "tnl/t_vb_rendertmp.h"
 868 #undef ELT
 869 #undef TAG
 870 #define TAG(x) intel_##x##_elts
 871 #define ELT(x) elt[x]
 872 #include "tnl/t_vb_rendertmp.h"
 873
 874 /**********************************************************************/
 875 /*                   Render clipped primitives                        */
 876 /**********************************************************************/
 877
 878
 879
 880 static void
 881 intelRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
 882 {
 883    struct intel_context *intel = intel_context(ctx);
 884    TNLcontext *tnl = TNL_CONTEXT(ctx);
 885    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
 886    GLuint prim = intel->render_primitive;
 887
 888    /* Render the new vertices as an unclipped polygon.
 889     */
 890    {
 891       GLuint *tmp = VB->Elts;
 892       VB->Elts = (GLuint *) elts;
 893       tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n,
 894                                                   PRIM_BEGIN | PRIM_END);
 895       VB->Elts = tmp;
 896    }
 897
 898    /* Restore the render primitive
 899     */
 900    if (prim != GL_POLYGON)
 901       tnl->Driver.Render.PrimitiveNotify(ctx, prim);
 902 }
 903
 904 static void
 905 intelRenderClippedLine(GLcontext * ctx, GLuint ii, GLuint jj)
 906 {
 907    TNLcontext *tnl = TNL_CONTEXT(ctx);
 908
 909    tnl->Driver.Render.Line(ctx, ii, jj);
 910 }
 911
 912 static void
 913 intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
 914 {
 915    struct intel_context *intel = intel_context(ctx);
 916    const GLuint vertsize = intel->vertex_size;
 917    GLuint *vb = intel_get_prim_space(intel, (n - 2) * 3);
 918    GLubyte *vertptr = (GLubyte *) intel->verts;
 919    const GLuint *start = (const GLuint *) V(elts[0]);
 920    int i, j;
 921
 922    for (i = 2; i < n; i++) {
 923       COPY_DWORDS(j, vb, vertsize, V(elts[i - 1]));
 924       COPY_DWORDS(j, vb, vertsize, V(elts[i]));
 925       COPY_DWORDS(j, vb, vertsize, start);
 926    }
 927 }
 928
 929 /**********************************************************************/
 930 /*                    Choose render functions                         */
 931 /**********************************************************************/
 932
 933
 934
 935
 936 #define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN | DD_POINT_SMOOTH | DD_TRI_SMOOTH)
 937 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED)
 938
 939 void
 940 intelChooseRenderState(GLcontext * ctx)
 941 {
 942    TNLcontext *tnl = TNL_CONTEXT(ctx);
 943    struct intel_context *intel = intel_context(ctx);
 944    GLuint flags = ctx->_TriangleCaps;
 945    const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current;
 946    GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS));
 947    GLuint index = 0;
 948
 949    if (INTEL_DEBUG & DEBUG_STATE)
 950       fprintf(stderr, "\n%s\n", __FUNCTION__);
 951
 952    if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) {
 953
 954       if (flags & ANY_RASTER_FLAGS) {
 955          if (flags & DD_TRI_LIGHT_TWOSIDE)
 956             index |= INTEL_TWOSIDE_BIT;
 957          if (flags & DD_TRI_OFFSET)
 958             index |= INTEL_OFFSET_BIT;
 959          if (flags & DD_TRI_UNFILLED)
 960             index |= INTEL_UNFILLED_BIT;
 961       }
 962
 963       if (have_wpos) {
 964          intel->draw_point = intel_wpos_point;
 965          intel->draw_line = intel_wpos_line;
 966          intel->draw_tri = intel_wpos_triangle;
 967
 968          /* Make sure these get called:
 969           */
 970          index |= INTEL_FALLBACK_BIT;
 971       }
 972       else {
 973          intel->draw_point = intel_draw_point;
 974          intel->draw_line = intel_draw_line;
 975          intel->draw_tri = intel_draw_triangle;
 976       }
 977
 978       /* Hook in fallbacks for specific primitives.
 979        */
 980       if (flags & ANY_FALLBACK_FLAGS) {
 981          if (flags & DD_LINE_STIPPLE)
 982             intel->draw_line = intel_fallback_line;
 983
 984          if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple)
 985             intel->draw_tri = intel_fallback_tri;
 986
 987          if (flags & DD_TRI_SMOOTH) {
 988             if (intel->conformance_mode > 0)
 989                intel->draw_tri = intel_fallback_tri;
 990          }
 991
 992          if (flags & DD_POINT_ATTEN) {
 993             if (0)
 994                intel->draw_point = intel_atten_point;
 995             else
 996                intel->draw_point = intel_fallback_point;
 997          }
 998
 999          if (flags & DD_POINT_SMOOTH) {
1000             if (intel->conformance_mode > 0)
1001                intel->draw_point = intel_fallback_point;
1002          }
1003
1004          index |= INTEL_FALLBACK_BIT;
1005       }
1006    }
1007
1008    if (intel->RenderIndex != index) {
1009       intel->RenderIndex = index;
1010
1011       tnl->Driver.Render.Points = rast_tab[index].points;
1012       tnl->Driver.Render.Line = rast_tab[index].line;
1013       tnl->Driver.Render.Triangle = rast_tab[index].triangle;
1014       tnl->Driver.Render.Quad = rast_tab[index].quad;
1015
1016       if (index == 0) {
1017          tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts;
1018          tnl->Driver.Render.PrimTabElts = intel_render_tab_elts;
1019          tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */
1020          tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly;
1021       }
1022       else {
1023          tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
1024          tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
1025          tnl->Driver.Render.ClippedLine = intelRenderClippedLine;
1026          tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly;
1027       }
1028    }
1029 }
1030
1031 static const GLenum reduced_prim[GL_POLYGON + 1] = {
1032    GL_POINTS,
1033    GL_LINES,
1034    GL_LINES,
1035    GL_LINES,
1036    GL_TRIANGLES,
1037    GL_TRIANGLES,
1038    GL_TRIANGLES,
1039    GL_TRIANGLES,
1040    GL_TRIANGLES,
1041    GL_TRIANGLES
1042 };
1043
1044
1045 /**********************************************************************/
1046 /*                 High level hooks for t_vb_render.c                 */
1047 /**********************************************************************/
1048
1049
1050
1051
1052 static void
1053 intelRunPipeline(GLcontext * ctx)
1054 {
1055    struct intel_context *intel = intel_context(ctx);
1056
1057    _mesa_lock_context_textures(ctx);
1058
1059    if (ctx->NewState)
1060       _mesa_update_state_locked(ctx);
1061
1062    if (intel->NewGLState) {
1063       if (intel->NewGLState & _NEW_TEXTURE) {
1064          intel->vtbl.update_texture_state(intel);
1065       }
1066
1067       if (!intel->Fallback) {
1068          if (intel->NewGLState & _INTEL_NEW_RENDERSTATE)
1069             intelChooseRenderState(ctx);
1070       }
1071
1072       intel->NewGLState = 0;
1073    }
1074
1075    intel_map_vertex_shader_textures(ctx);
1076    _tnl_run_pipeline(ctx);
1077    intel_unmap_vertex_shader_textures(ctx);
1078
1079    _mesa_unlock_context_textures(ctx);
1080 }
1081
1082 static void
1083 intelRenderStart(GLcontext * ctx)
1084 {
1085    struct intel_context *intel = intel_context(ctx);
1086
1087    intel_check_front_buffer_rendering(intel);
1088    intel->vtbl.render_start(intel_context(ctx));
1089    intel->vtbl.emit_state(intel);
1090 }
1091
1092 static void
1093 intelRenderFinish(GLcontext * ctx)
1094 {
1095    struct intel_context *intel = intel_context(ctx);
1096
1097    if (intel->RenderIndex & INTEL_FALLBACK_BIT)
1098       _swrast_flush(ctx);
1099
1100    INTEL_FIREVERTICES(intel);
1101 }
1102
1103
1104
1105
1106  /* System to flush dma and emit state changes based on the rasterized
1107   * primitive.
1108   */
1109 static void
1110 intelRasterPrimitive(GLcontext * ctx, GLenum rprim, GLuint hwprim)
1111 {
1112    struct intel_context *intel = intel_context(ctx);
1113
1114    if (0)
1115       fprintf(stderr, "%s %s %x\n", __FUNCTION__,
1116               _mesa_lookup_enum_by_nr(rprim), hwprim);
1117
1118    intel->vtbl.reduced_primitive_state(intel, rprim);
1119
1120    /* Start a new primitive.  Arrange to have it flushed later on.
1121     */
1122    if (hwprim != intel->prim.primitive) {
1123       INTEL_FIREVERTICES(intel);
1124
1125       intel_set_prim(intel, hwprim);
1126    }
1127 }
1128
1129
1130  /*
1131   */
1132 static void
1133 intelRenderPrimitive(GLcontext * ctx, GLenum prim)
1134 {
1135    struct intel_context *intel = intel_context(ctx);
1136
1137    if (0)
1138       fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
1139
1140    /* Let some clipping routines know which primitive they're dealing
1141     * with.
1142     */
1143    intel->render_primitive = prim;
1144
1145    /* Shortcircuit this when called from t_dd_rendertmp.h for unfilled
1146     * triangles.  The rasterized primitive will always be reset by
1147     * lower level functions in that case, potentially pingponging the
1148     * state:
1149     */
1150    if (reduced_prim[prim] == GL_TRIANGLES &&
1151        (ctx->_TriangleCaps & DD_TRI_UNFILLED))
1152       return;
1153
1154    /* Set some primitive-dependent state and Start? a new primitive.
1155     */
1156    intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]);
1157 }
1158
1159
1160  /**********************************************************************/
1161  /*           Transition to/from hardware rasterization.               */
1162  /**********************************************************************/
1163
1164 static char *fallbackStrings[] = {
1165    [0] = "Draw buffer",
1166    [1] = "Read buffer",
1167    [2] = "Depth buffer",
1168    [3] = "Stencil buffer",
1169    [4] = "User disable",
1170    [5] = "Render mode",
1171
1172    [12] = "Texture",
1173    [13] = "Color mask",
1174    [14] = "Stencil",
1175    [15] = "Stipple",
1176    [16] = "Program",
1177    [17] = "Logic op",
1178    [18] = "Smooth polygon",
1179    [19] = "Smooth point",
1180 };
1181
1182
1183 static char *
1184 getFallbackString(GLuint bit)
1185 {
1186    int i = 0;
1187    while (bit > 1) {
1188       i++;
1189       bit >>= 1;
1190    }
1191    return fallbackStrings[i];
1192 }
1193
1194
1195
1196 /**
1197  * Enable/disable a fallback flag.
1198  * \param bit  one of INTEL_FALLBACK_x flags.
1199  */
1200 void
1201 intelFallback(struct intel_context *intel, GLbitfield bit, GLboolean mode)
1202 {
1203    GLcontext *ctx = &intel->ctx;
1204    TNLcontext *tnl = TNL_CONTEXT(ctx);
1205    const GLbitfield oldfallback = intel->Fallback;
1206
1207    if (mode) {
1208       intel->Fallback |= bit;
1209       if (oldfallback == 0) {
1210          intelFlush(ctx);
1211          if (INTEL_DEBUG & DEBUG_FALLBACKS)
1212             fprintf(stderr, "ENTER FALLBACK %x: %s\n",
1213                     bit, getFallbackString(bit));
1214          _swsetup_Wakeup(ctx);
1215          intel->RenderIndex = ~0;
1216       }
1217    }
1218    else {
1219       intel->Fallback &= ~bit;
1220       if (oldfallback == bit) {
1221          _swrast_flush(ctx);
1222          if (INTEL_DEBUG & DEBUG_FALLBACKS)
1223             fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit));
1224          tnl->Driver.Render.Start = intelRenderStart;
1225          tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1226          tnl->Driver.Render.Finish = intelRenderFinish;
1227          tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1228          tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1229          tnl->Driver.Render.Interp = _tnl_interp;
1230
1231          _tnl_invalidate_vertex_state(ctx, ~0);
1232          _tnl_invalidate_vertices(ctx, ~0);
1233          _tnl_install_attrs(ctx,
1234                             intel->vertex_attrs,
1235                             intel->vertex_attr_count,
1236                             intel->ViewportMatrix.m, 0);
1237
1238          intel->NewGLState |= _INTEL_NEW_RENDERSTATE;
1239       }
1240    }
1241 }
1242
1243 union fi
1244 {
1245    GLfloat f;
1246    GLint i;
1247 };
1248
1249 /**********************************************************************/
1250 /*                            Initialization.                         */
1251 /**********************************************************************/
1252
1253
1254 void
1255 intelInitTriFuncs(GLcontext * ctx)
1256 {
1257    TNLcontext *tnl = TNL_CONTEXT(ctx);
1258    static int firsttime = 1;
1259
1260    if (firsttime) {
1261       init_rast_tab();
1262       firsttime = 0;
1263    }
1264
1265    tnl->Driver.RunPipeline = intelRunPipeline;
1266    tnl->Driver.Render.Start = intelRenderStart;
1267    tnl->Driver.Render.Finish = intelRenderFinish;
1268    tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1269    tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple;
1270    tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1271    tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1272    tnl->Driver.Render.Interp = _tnl_interp;
1273 }