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 #include "i915_context.h"
  58
  59 static void intelRenderPrimitive(struct gl_context * ctx, GLenum prim);
  60 static void intelRasterPrimitive(struct gl_context * ctx, GLenum rprim,
  61                                  GLuint hwprim);
  62
  63 static void
  64 intel_flush_inline_primitive(struct intel_context *intel)
  65 {
  66    GLuint used = intel->batch.used - intel->prim.start_ptr;
  67
  68    assert(intel->prim.primitive != ~0);
  69
  70 /*    printf("/\n"); */
  71
  72    if (used < 2)
  73       goto do_discard;
  74
  75    intel->batch.map[intel->prim.start_ptr] =
  76       _3DPRIMITIVE | intel->prim.primitive | (used - 2);
  77
  78    goto finished;
  79
  80  do_discard:
  81    intel->batch.used = intel->prim.start_ptr;
  82
  83  finished:
  84    intel->prim.primitive = ~0;
  85    intel->prim.start_ptr = 0;
  86    intel->prim.flush = 0;
  87 }
  88
  89 static void intel_start_inline(struct intel_context *intel, uint32_t prim)
  90 {
  91    BATCH_LOCALS;
  92
  93    intel->vtbl.emit_state(intel);
  94
  95    intel->no_batch_wrap = GL_TRUE;
  96
  97    /*printf("%s *", __progname);*/
  98
  99    /* Emit a slot which will be filled with the inline primitive
 100     * command later.
 101     */
 102    BEGIN_BATCH(1);
 103
 104    intel->prim.start_ptr = intel->batch.used;
 105    intel->prim.primitive = prim;
 106    intel->prim.flush = intel_flush_inline_primitive;
 107
 108    OUT_BATCH(0);
 109    ADVANCE_BATCH();
 110
 111    intel->no_batch_wrap = GL_FALSE;
 112 /*    printf(">"); */
 113 }
 114
 115 static void intel_wrap_inline(struct intel_context *intel)
 116 {
 117    GLuint prim = intel->prim.primitive;
 118
 119    intel_flush_inline_primitive(intel);
 120    intel_batchbuffer_flush(intel);
 121    intel_start_inline(intel, prim);  /* ??? */
 122 }
 123
 124 static GLuint *intel_extend_inline(struct intel_context *intel, GLuint dwords)
 125 {
 126    GLuint *ptr;
 127
 128    assert(intel->prim.flush == intel_flush_inline_primitive);
 129
 130    if (intel_batchbuffer_space(intel) < dwords * sizeof(GLuint))
 131       intel_wrap_inline(intel);
 132
 133 /*    printf("."); */
 134
 135    intel->vtbl.assert_not_dirty(intel);
 136
 137    ptr = intel->batch.map + intel->batch.used;
 138    intel->batch.used += dwords;
 139
 140    return ptr;
 141 }
 142
 143 /** Sets the primitive type for a primitive sequence, flushing as needed. */
 144 void intel_set_prim(struct intel_context *intel, uint32_t prim)
 145 {
 146    /* if we have no VBOs */
 147
 148    if (intel->intelScreen->no_vbo) {
 149       intel_start_inline(intel, prim);
 150       return;
 151    }
 152    if (prim != intel->prim.primitive) {
 153       INTEL_FIREVERTICES(intel);
 154       intel->prim.primitive = prim;
 155    }
 156 }
 157
 158 /** Returns mapped VB space for the given number of vertices */
 159 uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
 160 {
 161    uint32_t *addr;
 162
 163    if (intel->intelScreen->no_vbo) {
 164       return intel_extend_inline(intel, count * intel->vertex_size);
 165    }
 166
 167    /* Check for space in the existing VB */
 168    if (intel->prim.vb_bo == NULL ||
 169        (intel->prim.current_offset +
 170         count * intel->vertex_size * 4) > INTEL_VB_SIZE ||
 171        (intel->prim.count + count) >= (1 << 16)) {
 172       /* Flush existing prim if any */
 173       INTEL_FIREVERTICES(intel);
 174
 175       intel_finish_vb(intel);
 176
 177       /* Start a new VB */
 178       if (intel->prim.vb == NULL)
 179          intel->prim.vb = malloc(INTEL_VB_SIZE);
 180       intel->prim.vb_bo = drm_intel_bo_alloc(intel->bufmgr, "vb",
 181                                              INTEL_VB_SIZE, 4);
 182       intel->prim.start_offset = 0;
 183       intel->prim.current_offset = 0;
 184    }
 185
 186    intel->prim.flush = intel_flush_prim;
 187
 188    addr = (uint32_t *)(intel->prim.vb + intel->prim.current_offset);
 189    intel->prim.current_offset += intel->vertex_size * 4 * count;
 190    intel->prim.count += count;
 191
 192    return addr;
 193 }
 194
 195 /** Dispatches the accumulated primitive to the batchbuffer. */
 196 void intel_flush_prim(struct intel_context *intel)
 197 {
 198    drm_intel_bo *aper_array[2];
 199    drm_intel_bo *vb_bo;
 200    unsigned int offset, count;
 201    BATCH_LOCALS;
 202
 203    /* Must be called after an intel_start_prim. */
 204    assert(intel->prim.primitive != ~0);
 205
 206    if (intel->prim.count == 0)
 207       return;
 208
 209    /* Clear the current prims out of the context state so that a batch flush
 210     * flush triggered by emit_state doesn't loop back to flush_prim again.
 211     */
 212    vb_bo = intel->prim.vb_bo;
 213    drm_intel_bo_reference(vb_bo);
 214    count = intel->prim.count;
 215    intel->prim.count = 0;
 216    offset = intel->prim.start_offset;
 217    intel->prim.start_offset = intel->prim.current_offset;
 218    if (intel->gen < 3)
 219       intel->prim.current_offset = intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128);
 220    intel->prim.flush = NULL;
 221
 222    intel->vtbl.emit_state(intel);
 223
 224    aper_array[0] = intel->batch.bo;
 225    aper_array[1] = vb_bo;
 226    if (dri_bufmgr_check_aperture_space(aper_array, 2)) {
 227       intel_batchbuffer_flush(intel);
 228       intel->vtbl.emit_state(intel);
 229    }
 230
 231    /* Ensure that we don't start a new batch for the following emit, which
 232     * depends on the state just emitted. emit_state should be making sure we
 233     * have the space for this.
 234     */
 235    intel->no_batch_wrap = GL_TRUE;
 236
 237 #if 0
 238    printf("emitting %d..%d=%d vertices size %d\n", offset,
 239           intel->prim.current_offset, count,
 240           intel->vertex_size * 4);
 241 #endif
 242
 243    if (intel->gen >= 3) {
 244       struct i915_context *i915 = i915_context(&intel->ctx);
 245       unsigned int cmd = 0, len = 0;
 246
 247       if (vb_bo != i915->current_vb_bo) {
 248          cmd |= I1_LOAD_S(0);
 249          len++;
 250       }
 251
 252       if (intel->vertex_size != i915->current_vertex_size) {
 253          cmd |= I1_LOAD_S(1);
 254          len++;
 255       }
 256       if (len)
 257          len++;
 258
 259       BEGIN_BATCH(2+len);
 260       if (cmd)
 261          OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | cmd | (len - 2));
 262       if (vb_bo != i915->current_vb_bo) {
 263          OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, 0);
 264          i915->current_vb_bo = vb_bo;
 265       }
 266       if (intel->vertex_size != i915->current_vertex_size) {
 267          OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) |
 268                    (intel->vertex_size << S1_VERTEX_PITCH_SHIFT));
 269          i915->current_vertex_size = intel->vertex_size;
 270       }
 271       OUT_BATCH(_3DPRIMITIVE |
 272                 PRIM_INDIRECT |
 273                 PRIM_INDIRECT_SEQUENTIAL |
 274                 intel->prim.primitive |
 275                 count);
 276       OUT_BATCH(offset / (intel->vertex_size * 4));
 277       ADVANCE_BATCH();
 278    } else {
 279       struct i830_context *i830 = i830_context(&intel->ctx);
 280
 281       BEGIN_BATCH(5);
 282       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
 283                 I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
 284       /* S0 */
 285       assert((offset & ~S0_VB_OFFSET_MASK_830) == 0);
 286       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
 287                 offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
 288                 S0_VB_ENABLE_830);
 289       /* S2
 290        * This is somewhat unfortunate -- VB width is tied up with
 291        * vertex format data that we've already uploaded through
 292        * _3DSTATE_VFT[01]_CMD.  We may want to replace emits of VFT state with
 293        * STATE_IMMEDIATE_1 like this to avoid duplication.
 294        */
 295       OUT_BATCH((i830->state.Ctx[I830_CTXREG_VF] & VFT0_TEX_COUNT_MASK) >>
 296                 VFT0_TEX_COUNT_SHIFT << S2_TEX_COUNT_SHIFT_830 |
 297                 (i830->state.Ctx[I830_CTXREG_VF2] << 16) |
 298                 intel->vertex_size << S2_VERTEX_0_WIDTH_SHIFT_830);
 299
 300       OUT_BATCH(_3DPRIMITIVE |
 301                 PRIM_INDIRECT |
 302                 PRIM_INDIRECT_SEQUENTIAL |
 303                 intel->prim.primitive |
 304                 count);
 305       OUT_BATCH(0); /* Beginning vertex index */
 306       ADVANCE_BATCH();
 307    }
 308
 309    intel->no_batch_wrap = GL_FALSE;
 310
 311    drm_intel_bo_unreference(vb_bo);
 312 }
 313
 314 /**
 315  * Uploads the locally-accumulated VB into the buffer object.
 316  *
 317  * This avoids us thrashing the cachelines in and out as the buffer gets
 318  * filled, dispatched, then reused as the hardware completes rendering from it,
 319  * and also lets us clflush less if we dispatch with a partially-filled VB.
 320  *
 321  * This is called normally from get_space when we're finishing a BO, but also
 322  * at batch flush time so that we don't try accessing the contents of a
 323  * just-dispatched buffer.
 324  */
 325 void intel_finish_vb(struct intel_context *intel)
 326 {
 327    if (intel->prim.vb_bo == NULL)
 328       return;
 329
 330    drm_intel_bo_subdata(intel->prim.vb_bo, 0, intel->prim.start_offset,
 331                         intel->prim.vb);
 332    drm_intel_bo_unreference(intel->prim.vb_bo);
 333    intel->prim.vb_bo = NULL;
 334 }
 335
 336 /***********************************************************************
 337  *                    Emit primitives as inline vertices               *
 338  ***********************************************************************/
 339
 340 #ifdef __i386__
 341 #define COPY_DWORDS( j, vb, vertsize, v )                       \
 342 do {                                                            \
 343    int __tmp;                                                   \
 344    __asm__ __volatile__( "rep ; movsl"                          \
 345                          : "=%c" (j), "=D" (vb), "=S" (__tmp)   \
 346                          : "0" (vertsize),                      \
 347                          "D" ((long)vb),                        \
 348                          "S" ((long)v) );                       \
 349 } while (0)
 350 #else
 351 #define COPY_DWORDS( j, vb, vertsize, v )       \
 352 do {                                            \
 353    for ( j = 0 ; j < vertsize ; j++ ) {         \
 354       vb[j] = ((GLuint *)v)[j];                 \
 355    }                                            \
 356    vb += vertsize;                              \
 357 } while (0)
 358 #endif
 359
 360 static void
 361 intel_draw_quad(struct intel_context *intel,
 362                 intelVertexPtr v0,
 363                 intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3)
 364 {
 365    GLuint vertsize = intel->vertex_size;
 366    GLuint *vb = intel_get_prim_space(intel, 6);
 367    int j;
 368
 369    COPY_DWORDS(j, vb, vertsize, v0);
 370    COPY_DWORDS(j, vb, vertsize, v1);
 371
 372    /* If smooth shading, draw like a trifan which gives better
 373     * rasterization.  Otherwise draw as two triangles with provoking
 374     * vertex in third position as required for flat shading.
 375     */
 376    if (intel->ctx.Light.ShadeModel == GL_FLAT) {
 377       COPY_DWORDS(j, vb, vertsize, v3);
 378       COPY_DWORDS(j, vb, vertsize, v1);
 379    }
 380    else {
 381       COPY_DWORDS(j, vb, vertsize, v2);
 382       COPY_DWORDS(j, vb, vertsize, v0);
 383    }
 384
 385    COPY_DWORDS(j, vb, vertsize, v2);
 386    COPY_DWORDS(j, vb, vertsize, v3);
 387 }
 388
 389 static void
 390 intel_draw_triangle(struct intel_context *intel,
 391                     intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
 392 {
 393    GLuint vertsize = intel->vertex_size;
 394    GLuint *vb = intel_get_prim_space(intel, 3);
 395    int j;
 396
 397    COPY_DWORDS(j, vb, vertsize, v0);
 398    COPY_DWORDS(j, vb, vertsize, v1);
 399    COPY_DWORDS(j, vb, vertsize, v2);
 400 }
 401
 402
 403 static void
 404 intel_draw_line(struct intel_context *intel,
 405                 intelVertexPtr v0, intelVertexPtr v1)
 406 {
 407    GLuint vertsize = intel->vertex_size;
 408    GLuint *vb = intel_get_prim_space(intel, 2);
 409    int j;
 410
 411    COPY_DWORDS(j, vb, vertsize, v0);
 412    COPY_DWORDS(j, vb, vertsize, v1);
 413 }
 414
 415
 416 static void
 417 intel_draw_point(struct intel_context *intel, intelVertexPtr v0)
 418 {
 419    GLuint vertsize = intel->vertex_size;
 420    GLuint *vb = intel_get_prim_space(intel, 1);
 421    int j;
 422
 423    /* Adjust for sub pixel position -- still required for conform. */
 424    *(float *) &vb[0] = v0->v.x;
 425    *(float *) &vb[1] = v0->v.y;
 426    for (j = 2; j < vertsize; j++)
 427       vb[j] = v0->ui[j];
 428 }
 429
 430
 431
 432 /***********************************************************************
 433  *                Fixup for ARB_point_parameters                       *
 434  ***********************************************************************/
 435
 436 /* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly
 437  * represented in the fragment program InputsRead field.
 438  */
 439 static void
 440 intel_atten_point(struct intel_context *intel, intelVertexPtr v0)
 441 {
 442    struct gl_context *ctx = &intel->ctx;
 443    GLfloat psz[4], col[4], restore_psz, restore_alpha;
 444
 445    _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 446    _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 447
 448    restore_psz = psz[0];
 449    restore_alpha = col[3];
 450
 451    if (psz[0] >= ctx->Point.Threshold) {
 452       psz[0] = MIN2(psz[0], ctx->Point.MaxSize);
 453    }
 454    else {
 455       GLfloat dsize = psz[0] / ctx->Point.Threshold;
 456       psz[0] = MAX2(ctx->Point.Threshold, ctx->Point.MinSize);
 457       col[3] *= dsize * dsize;
 458    }
 459
 460    if (psz[0] < 1.0)
 461       psz[0] = 1.0;
 462
 463    if (restore_psz != psz[0] || restore_alpha != col[3]) {
 464       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 465       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 466
 467       intel_draw_point(intel, v0);
 468
 469       psz[0] = restore_psz;
 470       col[3] = restore_alpha;
 471
 472       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 473       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 474    }
 475    else
 476       intel_draw_point(intel, v0);
 477 }
 478
 479
 480
 481
 482
 483 /***********************************************************************
 484  *                Fixup for I915 WPOS texture coordinate                *
 485  ***********************************************************************/
 486
 487
 488
 489 static void
 490 intel_wpos_triangle(struct intel_context *intel,
 491                     intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
 492 {
 493    const struct gl_framebuffer *fb = intel->ctx.DrawBuffer;
 494    GLuint offset = intel->wpos_offset;
 495    GLuint size = intel->wpos_size;
 496    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 497    GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
 498    GLfloat *v2_wpos = (GLfloat *)((char *)v2 + offset);
 499
 500    __memcpy(v0_wpos, v0, size);
 501    __memcpy(v1_wpos, v1, size);
 502    __memcpy(v2_wpos, v2, size);
 503
 504    if (!fb->Name) {
 505       v0_wpos[1] = -v0_wpos[1] + fb->Height;
 506       v1_wpos[1] = -v1_wpos[1] + fb->Height;
 507       v2_wpos[1] = -v2_wpos[1] + fb->Height;
 508    }
 509
 510    intel_draw_triangle(intel, v0, v1, v2);
 511 }
 512
 513
 514 static void
 515 intel_wpos_line(struct intel_context *intel,
 516                 intelVertexPtr v0, intelVertexPtr v1)
 517 {
 518    const struct gl_framebuffer *fb = intel->ctx.DrawBuffer;
 519    GLuint offset = intel->wpos_offset;
 520    GLuint size = intel->wpos_size;
 521    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 522    GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
 523
 524    __memcpy(v0_wpos, v0, size);
 525    __memcpy(v1_wpos, v1, size);
 526
 527    if (!fb->Name) {
 528       v0_wpos[1] = -v0_wpos[1] + fb->Height;
 529       v1_wpos[1] = -v1_wpos[1] + fb->Height;
 530    }
 531
 532    intel_draw_line(intel, v0, v1);
 533 }
 534
 535
 536 static void
 537 intel_wpos_point(struct intel_context *intel, intelVertexPtr v0)
 538 {
 539    const struct gl_framebuffer *fb = intel->ctx.DrawBuffer;
 540    GLuint offset = intel->wpos_offset;
 541    GLuint size = intel->wpos_size;
 542    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 543
 544    __memcpy(v0_wpos, v0, size);
 545
 546    if (!fb->Name)
 547       v0_wpos[1] = -v0_wpos[1] + fb->Height;
 548
 549    intel_draw_point(intel, v0);
 550 }
 551
 552
 553
 554
 555
 556
 557 /***********************************************************************
 558  *          Macros for t_dd_tritmp.h to draw basic primitives          *
 559  ***********************************************************************/
 560
 561 #define TRI( a, b, c )                          \
 562 do {                                            \
 563    if (DO_FALLBACK)                             \
 564       intel->draw_tri( intel, a, b, c );        \
 565    else                                         \
 566       intel_draw_triangle( intel, a, b, c );    \
 567 } while (0)
 568
 569 #define QUAD( a, b, c, d )                      \
 570 do {                                            \
 571    if (DO_FALLBACK) {                           \
 572       intel->draw_tri( intel, a, b, d );        \
 573       intel->draw_tri( intel, b, c, d );        \
 574    } else                                       \
 575       intel_draw_quad( intel, a, b, c, d );     \
 576 } while (0)
 577
 578 #define LINE( v0, v1 )                          \
 579 do {                                            \
 580    if (DO_FALLBACK)                             \
 581       intel->draw_line( intel, v0, v1 );        \
 582    else                                         \
 583       intel_draw_line( intel, v0, v1 );         \
 584 } while (0)
 585
 586 #define POINT( v0 )                             \
 587 do {                                            \
 588    if (DO_FALLBACK)                             \
 589       intel->draw_point( intel, v0 );           \
 590    else                                         \
 591       intel_draw_point( intel, v0 );            \
 592 } while (0)
 593
 594
 595 /***********************************************************************
 596  *              Build render functions from dd templates               *
 597  ***********************************************************************/
 598
 599 #define INTEL_OFFSET_BIT        0x01
 600 #define INTEL_TWOSIDE_BIT       0x02
 601 #define INTEL_UNFILLED_BIT      0x04
 602 #define INTEL_FALLBACK_BIT      0x08
 603 #define INTEL_MAX_TRIFUNC       0x10
 604
 605
 606 static struct
 607 {
 608    tnl_points_func points;
 609    tnl_line_func line;
 610    tnl_triangle_func triangle;
 611    tnl_quad_func quad;
 612 } rast_tab[INTEL_MAX_TRIFUNC];
 613
 614
 615 #define DO_FALLBACK (IND & INTEL_FALLBACK_BIT)
 616 #define DO_OFFSET   (IND & INTEL_OFFSET_BIT)
 617 #define DO_UNFILLED (IND & INTEL_UNFILLED_BIT)
 618 #define DO_TWOSIDE  (IND & INTEL_TWOSIDE_BIT)
 619 #define DO_FLAT      0
 620 #define DO_TRI       1
 621 #define DO_QUAD      1
 622 #define DO_LINE      1
 623 #define DO_POINTS    1
 624 #define DO_FULL_QUAD 1
 625
 626 #define HAVE_SPEC         1
 627 #define HAVE_BACK_COLORS  0
 628 #define HAVE_HW_FLATSHADE 1
 629 #define VERTEX            intelVertex
 630 #define TAB               rast_tab
 631
 632 /* Only used to pull back colors into vertices (ie, we know color is
 633  * floating point).
 634  */
 635 #define INTEL_COLOR( dst, src )                         \
 636 do {                                                    \
 637    UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);        \
 638    UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);        \
 639    UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);        \
 640    UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]);        \
 641 } while (0)
 642
 643 #define INTEL_SPEC( dst, src )                          \
 644 do {                                                    \
 645    UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);        \
 646    UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);        \
 647    UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);        \
 648 } while (0)
 649
 650
 651 #define DEPTH_SCALE intel->polygon_offset_scale
 652 #define UNFILLED_TRI unfilled_tri
 653 #define UNFILLED_QUAD unfilled_quad
 654 #define VERT_X(_v) _v->v.x
 655 #define VERT_Y(_v) _v->v.y
 656 #define VERT_Z(_v) _v->v.z
 657 #define AREA_IS_CCW( a ) (a > 0)
 658 #define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint)))
 659
 660 #define VERT_SET_RGBA( v, c )    if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c )
 661 #define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
 662 #define VERT_SAVE_RGBA( idx )    if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
 663 #define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
 664
 665 #define VERT_SET_SPEC( v, c )    if (specoffset) INTEL_SPEC( v->ub4[specoffset], c )
 666 #define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
 667 #define VERT_SAVE_SPEC( idx )    if (specoffset) spec[idx] = v[idx]->ui[specoffset]
 668 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
 669
 670 #define LOCAL_VARS(n)                                                   \
 671    struct intel_context *intel = intel_context(ctx);                    \
 672    GLuint color[n] = { 0, }, spec[n] = { 0, };                          \
 673    GLuint coloroffset = intel->coloroffset;                             \
 674    GLboolean specoffset = intel->specoffset;                            \
 675    (void) color; (void) spec; (void) coloroffset; (void) specoffset;
 676
 677
 678 /***********************************************************************
 679  *                Helpers for rendering unfilled primitives            *
 680  ***********************************************************************/
 681
 682 static const GLuint hw_prim[GL_POLYGON + 1] = {
 683    PRIM3D_POINTLIST,
 684    PRIM3D_LINELIST,
 685    PRIM3D_LINELIST,
 686    PRIM3D_LINELIST,
 687    PRIM3D_TRILIST,
 688    PRIM3D_TRILIST,
 689    PRIM3D_TRILIST,
 690    PRIM3D_TRILIST,
 691    PRIM3D_TRILIST,
 692    PRIM3D_TRILIST
 693 };
 694
 695 #define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] )
 696 #define RENDER_PRIMITIVE intel->render_primitive
 697 #define TAG(x) x
 698 #define IND INTEL_FALLBACK_BIT
 699 #include "tnl_dd/t_dd_unfilled.h"
 700 #undef IND
 701
 702 /***********************************************************************
 703  *                      Generate GL render functions                   *
 704  ***********************************************************************/
 705
 706 #define IND (0)
 707 #define TAG(x) x
 708 #include "tnl_dd/t_dd_tritmp.h"
 709
 710 #define IND (INTEL_OFFSET_BIT)
 711 #define TAG(x) x##_offset
 712 #include "tnl_dd/t_dd_tritmp.h"
 713
 714 #define IND (INTEL_TWOSIDE_BIT)
 715 #define TAG(x) x##_twoside
 716 #include "tnl_dd/t_dd_tritmp.h"
 717
 718 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT)
 719 #define TAG(x) x##_twoside_offset
 720 #include "tnl_dd/t_dd_tritmp.h"
 721
 722 #define IND (INTEL_UNFILLED_BIT)
 723 #define TAG(x) x##_unfilled
 724 #include "tnl_dd/t_dd_tritmp.h"
 725
 726 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
 727 #define TAG(x) x##_offset_unfilled
 728 #include "tnl_dd/t_dd_tritmp.h"
 729
 730 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT)
 731 #define TAG(x) x##_twoside_unfilled
 732 #include "tnl_dd/t_dd_tritmp.h"
 733
 734 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
 735 #define TAG(x) x##_twoside_offset_unfilled
 736 #include "tnl_dd/t_dd_tritmp.h"
 737
 738 #define IND (INTEL_FALLBACK_BIT)
 739 #define TAG(x) x##_fallback
 740 #include "tnl_dd/t_dd_tritmp.h"
 741
 742 #define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
 743 #define TAG(x) x##_offset_fallback
 744 #include "tnl_dd/t_dd_tritmp.h"
 745
 746 #define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT)
 747 #define TAG(x) x##_twoside_fallback
 748 #include "tnl_dd/t_dd_tritmp.h"
 749
 750 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
 751 #define TAG(x) x##_twoside_offset_fallback
 752 #include "tnl_dd/t_dd_tritmp.h"
 753
 754 #define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 755 #define TAG(x) x##_unfilled_fallback
 756 #include "tnl_dd/t_dd_tritmp.h"
 757
 758 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 759 #define TAG(x) x##_offset_unfilled_fallback
 760 #include "tnl_dd/t_dd_tritmp.h"
 761
 762 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 763 #define TAG(x) x##_twoside_unfilled_fallback
 764 #include "tnl_dd/t_dd_tritmp.h"
 765
 766 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \
 767              INTEL_FALLBACK_BIT)
 768 #define TAG(x) x##_twoside_offset_unfilled_fallback
 769 #include "tnl_dd/t_dd_tritmp.h"
 770
 771
 772 static void
 773 init_rast_tab(void)
 774 {
 775    init();
 776    init_offset();
 777    init_twoside();
 778    init_twoside_offset();
 779    init_unfilled();
 780    init_offset_unfilled();
 781    init_twoside_unfilled();
 782    init_twoside_offset_unfilled();
 783    init_fallback();
 784    init_offset_fallback();
 785    init_twoside_fallback();
 786    init_twoside_offset_fallback();
 787    init_unfilled_fallback();
 788    init_offset_unfilled_fallback();
 789    init_twoside_unfilled_fallback();
 790    init_twoside_offset_unfilled_fallback();
 791 }
 792
 793
 794 /***********************************************************************
 795  *                    Rasterization fallback helpers                   *
 796  ***********************************************************************/
 797
 798
 799 /* This code is hit only when a mix of accelerated and unaccelerated
 800  * primitives are being drawn, and only for the unaccelerated
 801  * primitives.
 802  */
 803 static void
 804 intel_fallback_tri(struct intel_context *intel,
 805                    intelVertex * v0, intelVertex * v1, intelVertex * v2)
 806 {
 807    struct gl_context *ctx = &intel->ctx;
 808    SWvertex v[3];
 809
 810    if (0)
 811       fprintf(stderr, "\n%s\n", __FUNCTION__);
 812
 813    INTEL_FIREVERTICES(intel);
 814
 815    _swsetup_Translate(ctx, v0, &v[0]);
 816    _swsetup_Translate(ctx, v1, &v[1]);
 817    _swsetup_Translate(ctx, v2, &v[2]);
 818    intelSpanRenderStart(ctx);
 819    _swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
 820    intelSpanRenderFinish(ctx);
 821 }
 822
 823
 824 static void
 825 intel_fallback_line(struct intel_context *intel,
 826                     intelVertex * v0, intelVertex * v1)
 827 {
 828    struct gl_context *ctx = &intel->ctx;
 829    SWvertex v[2];
 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    _swsetup_Translate(ctx, v1, &v[1]);
 838    intelSpanRenderStart(ctx);
 839    _swrast_Line(ctx, &v[0], &v[1]);
 840    intelSpanRenderFinish(ctx);
 841 }
 842
 843 static void
 844 intel_fallback_point(struct intel_context *intel,
 845                      intelVertex * v0)
 846 {
 847    struct gl_context *ctx = &intel->ctx;
 848    SWvertex v[1];
 849
 850    if (0)
 851       fprintf(stderr, "\n%s\n", __FUNCTION__);
 852
 853    INTEL_FIREVERTICES(intel);
 854
 855    _swsetup_Translate(ctx, v0, &v[0]);
 856    intelSpanRenderStart(ctx);
 857    _swrast_Point(ctx, &v[0]);
 858    intelSpanRenderFinish(ctx);
 859 }
 860
 861
 862 /**********************************************************************/
 863 /*               Render unclipped begin/end objects                   */
 864 /**********************************************************************/
 865
 866 #define IND 0
 867 #define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint)))
 868 #define RENDER_POINTS( start, count )   \
 869    for ( ; start < count ; start++) POINT( V(ELT(start)) );
 870 #define RENDER_LINE( v0, v1 )         LINE( V(v0), V(v1) )
 871 #define RENDER_TRI(  v0, v1, v2 )     TRI(  V(v0), V(v1), V(v2) )
 872 #define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) )
 873 #define INIT(x) intelRenderPrimitive( ctx, x )
 874 #undef LOCAL_VARS
 875 #define LOCAL_VARS                                              \
 876     struct intel_context *intel = intel_context(ctx);                   \
 877     GLubyte *vertptr = (GLubyte *)intel->verts;                 \
 878     const GLuint vertsize = intel->vertex_size;         \
 879     const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;       \
 880     (void) elt;
 881 #define RESET_STIPPLE
 882 #define RESET_OCCLUSION
 883 #define PRESERVE_VB_DEFS
 884 #define ELT(x) x
 885 #define TAG(x) intel_##x##_verts
 886 #include "tnl/t_vb_rendertmp.h"
 887 #undef ELT
 888 #undef TAG
 889 #define TAG(x) intel_##x##_elts
 890 #define ELT(x) elt[x]
 891 #include "tnl/t_vb_rendertmp.h"
 892
 893 /**********************************************************************/
 894 /*                   Render clipped primitives                        */
 895 /**********************************************************************/
 896
 897
 898
 899 static void
 900 intelRenderClippedPoly(struct gl_context * ctx, const GLuint * elts, GLuint n)
 901 {
 902    struct intel_context *intel = intel_context(ctx);
 903    TNLcontext *tnl = TNL_CONTEXT(ctx);
 904    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
 905    GLuint prim = intel->render_primitive;
 906
 907    /* Render the new vertices as an unclipped polygon.
 908     */
 909    {
 910       GLuint *tmp = VB->Elts;
 911       VB->Elts = (GLuint *) elts;
 912       tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n,
 913                                                   PRIM_BEGIN | PRIM_END);
 914       VB->Elts = tmp;
 915    }
 916
 917    /* Restore the render primitive
 918     */
 919    if (prim != GL_POLYGON)
 920       tnl->Driver.Render.PrimitiveNotify(ctx, prim);
 921 }
 922
 923 static void
 924 intelRenderClippedLine(struct gl_context * ctx, GLuint ii, GLuint jj)
 925 {
 926    TNLcontext *tnl = TNL_CONTEXT(ctx);
 927
 928    tnl->Driver.Render.Line(ctx, ii, jj);
 929 }
 930
 931 static void
 932 intelFastRenderClippedPoly(struct gl_context * ctx, const GLuint * elts, GLuint n)
 933 {
 934    struct intel_context *intel = intel_context(ctx);
 935    const GLuint vertsize = intel->vertex_size;
 936    GLuint *vb = intel_get_prim_space(intel, (n - 2) * 3);
 937    GLubyte *vertptr = (GLubyte *) intel->verts;
 938    const GLuint *start = (const GLuint *) V(elts[0]);
 939    int i, j;
 940
 941    for (i = 2; i < n; i++) {
 942       COPY_DWORDS(j, vb, vertsize, V(elts[i - 1]));
 943       COPY_DWORDS(j, vb, vertsize, V(elts[i]));
 944       COPY_DWORDS(j, vb, vertsize, start);
 945    }
 946 }
 947
 948 /**********************************************************************/
 949 /*                    Choose render functions                         */
 950 /**********************************************************************/
 951
 952
 953
 954
 955 #define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN | DD_POINT_SMOOTH | DD_TRI_SMOOTH)
 956 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED)
 957
 958 void
 959 intelChooseRenderState(struct gl_context * ctx)
 960 {
 961    TNLcontext *tnl = TNL_CONTEXT(ctx);
 962    struct intel_context *intel = intel_context(ctx);
 963    GLuint flags = ctx->_TriangleCaps;
 964    const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current;
 965    GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS));
 966    GLuint index = 0;
 967
 968    if (INTEL_DEBUG & DEBUG_STATE)
 969       fprintf(stderr, "\n%s\n", __FUNCTION__);
 970
 971    if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) {
 972
 973       if (flags & ANY_RASTER_FLAGS) {
 974          if (flags & DD_TRI_LIGHT_TWOSIDE)
 975             index |= INTEL_TWOSIDE_BIT;
 976          if (flags & DD_TRI_OFFSET)
 977             index |= INTEL_OFFSET_BIT;
 978          if (flags & DD_TRI_UNFILLED)
 979             index |= INTEL_UNFILLED_BIT;
 980       }
 981
 982       if (have_wpos) {
 983          intel->draw_point = intel_wpos_point;
 984          intel->draw_line = intel_wpos_line;
 985          intel->draw_tri = intel_wpos_triangle;
 986
 987          /* Make sure these get called:
 988           */
 989          index |= INTEL_FALLBACK_BIT;
 990       }
 991       else {
 992          intel->draw_point = intel_draw_point;
 993          intel->draw_line = intel_draw_line;
 994          intel->draw_tri = intel_draw_triangle;
 995       }
 996
 997       /* Hook in fallbacks for specific primitives.
 998        */
 999       if (flags & ANY_FALLBACK_FLAGS) {
1000          if (flags & DD_LINE_STIPPLE)
1001             intel->draw_line = intel_fallback_line;
1002
1003          if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple)
1004             intel->draw_tri = intel_fallback_tri;
1005
1006          if (flags & DD_TRI_SMOOTH) {
1007             if (intel->conformance_mode > 0)
1008                intel->draw_tri = intel_fallback_tri;
1009          }
1010
1011          if (flags & DD_POINT_ATTEN) {
1012             if (0)
1013                intel->draw_point = intel_atten_point;
1014             else
1015                intel->draw_point = intel_fallback_point;
1016          }
1017
1018          if (flags & DD_POINT_SMOOTH) {
1019             if (intel->conformance_mode > 0)
1020                intel->draw_point = intel_fallback_point;
1021          }
1022
1023          index |= INTEL_FALLBACK_BIT;
1024       }
1025    }
1026
1027    if (intel->RenderIndex != index) {
1028       intel->RenderIndex = index;
1029
1030       tnl->Driver.Render.Points = rast_tab[index].points;
1031       tnl->Driver.Render.Line = rast_tab[index].line;
1032       tnl->Driver.Render.Triangle = rast_tab[index].triangle;
1033       tnl->Driver.Render.Quad = rast_tab[index].quad;
1034
1035       if (index == 0) {
1036          tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts;
1037          tnl->Driver.Render.PrimTabElts = intel_render_tab_elts;
1038          tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */
1039          tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly;
1040       }
1041       else {
1042          tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
1043          tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
1044          tnl->Driver.Render.ClippedLine = intelRenderClippedLine;
1045          tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly;
1046       }
1047    }
1048 }
1049
1050 static const GLenum reduced_prim[GL_POLYGON + 1] = {
1051    GL_POINTS,
1052    GL_LINES,
1053    GL_LINES,
1054    GL_LINES,
1055    GL_TRIANGLES,
1056    GL_TRIANGLES,
1057    GL_TRIANGLES,
1058    GL_TRIANGLES,
1059    GL_TRIANGLES,
1060    GL_TRIANGLES
1061 };
1062
1063
1064 /**********************************************************************/
1065 /*                 High level hooks for t_vb_render.c                 */
1066 /**********************************************************************/
1067
1068
1069
1070
1071 static void
1072 intelRunPipeline(struct gl_context * ctx)
1073 {
1074    struct intel_context *intel = intel_context(ctx);
1075
1076    _mesa_lock_context_textures(ctx);
1077
1078    if (ctx->NewState)
1079       _mesa_update_state_locked(ctx);
1080
1081    if (intel->NewGLState) {
1082       if (intel->NewGLState & _NEW_TEXTURE) {
1083          intel->vtbl.update_texture_state(intel);
1084       }
1085
1086       if (!intel->Fallback) {
1087          if (intel->NewGLState & _INTEL_NEW_RENDERSTATE)
1088             intelChooseRenderState(ctx);
1089       }
1090
1091       intel->NewGLState = 0;
1092    }
1093
1094    intel_map_vertex_shader_textures(ctx);
1095    _tnl_run_pipeline(ctx);
1096    intel_unmap_vertex_shader_textures(ctx);
1097
1098    _mesa_unlock_context_textures(ctx);
1099 }
1100
1101 static void
1102 intelRenderStart(struct gl_context * ctx)
1103 {
1104    struct intel_context *intel = intel_context(ctx);
1105
1106    intel_check_front_buffer_rendering(intel);
1107    intel->vtbl.render_start(intel_context(ctx));
1108    intel->vtbl.emit_state(intel);
1109 }
1110
1111 static void
1112 intelRenderFinish(struct gl_context * ctx)
1113 {
1114    struct intel_context *intel = intel_context(ctx);
1115
1116    if (intel->RenderIndex & INTEL_FALLBACK_BIT)
1117       _swrast_flush(ctx);
1118
1119    INTEL_FIREVERTICES(intel);
1120 }
1121
1122
1123
1124
1125  /* System to flush dma and emit state changes based on the rasterized
1126   * primitive.
1127   */
1128 static void
1129 intelRasterPrimitive(struct gl_context * ctx, GLenum rprim, GLuint hwprim)
1130 {
1131    struct intel_context *intel = intel_context(ctx);
1132
1133    if (0)
1134       fprintf(stderr, "%s %s %x\n", __FUNCTION__,
1135               _mesa_lookup_enum_by_nr(rprim), hwprim);
1136
1137    intel->vtbl.reduced_primitive_state(intel, rprim);
1138
1139    /* Start a new primitive.  Arrange to have it flushed later on.
1140     */
1141    if (hwprim != intel->prim.primitive) {
1142       INTEL_FIREVERTICES(intel);
1143
1144       intel_set_prim(intel, hwprim);
1145    }
1146 }
1147
1148
1149  /*
1150   */
1151 static void
1152 intelRenderPrimitive(struct gl_context * ctx, GLenum prim)
1153 {
1154    struct intel_context *intel = intel_context(ctx);
1155
1156    if (0)
1157       fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
1158
1159    /* Let some clipping routines know which primitive they're dealing
1160     * with.
1161     */
1162    intel->render_primitive = prim;
1163
1164    /* Shortcircuit this when called from t_dd_rendertmp.h for unfilled
1165     * triangles.  The rasterized primitive will always be reset by
1166     * lower level functions in that case, potentially pingponging the
1167     * state:
1168     */
1169    if (reduced_prim[prim] == GL_TRIANGLES &&
1170        (ctx->_TriangleCaps & DD_TRI_UNFILLED))
1171       return;
1172
1173    /* Set some primitive-dependent state and Start? a new primitive.
1174     */
1175    intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]);
1176 }
1177
1178
1179  /**********************************************************************/
1180  /*           Transition to/from hardware rasterization.               */
1181  /**********************************************************************/
1182
1183 static char *fallbackStrings[] = {
1184    [0] = "Draw buffer",
1185    [1] = "Read buffer",
1186    [2] = "Depth buffer",
1187    [3] = "Stencil buffer",
1188    [4] = "User disable",
1189    [5] = "Render mode",
1190
1191    [12] = "Texture",
1192    [13] = "Color mask",
1193    [14] = "Stencil",
1194    [15] = "Stipple",
1195    [16] = "Program",
1196    [17] = "Logic op",
1197    [18] = "Smooth polygon",
1198    [19] = "Smooth point",
1199    [20] = "point sprite coord origin",
1200    [21] = "depth/color drawing offset",
1201 };
1202
1203
1204 static char *
1205 getFallbackString(GLuint bit)
1206 {
1207    int i = 0;
1208    while (bit > 1) {
1209       i++;
1210       bit >>= 1;
1211    }
1212    return fallbackStrings[i];
1213 }
1214
1215
1216
1217 /**
1218  * Enable/disable a fallback flag.
1219  * \param bit  one of INTEL_FALLBACK_x flags.
1220  */
1221 void
1222 intelFallback(struct intel_context *intel, GLbitfield bit, GLboolean mode)
1223 {
1224    struct gl_context *ctx = &intel->ctx;
1225    TNLcontext *tnl = TNL_CONTEXT(ctx);
1226    const GLbitfield oldfallback = intel->Fallback;
1227
1228    if (mode) {
1229       intel->Fallback |= bit;
1230       if (oldfallback == 0) {
1231          intel_flush(ctx);
1232          if (INTEL_DEBUG & DEBUG_FALLBACKS)
1233             fprintf(stderr, "ENTER FALLBACK %x: %s\n",
1234                     bit, getFallbackString(bit));
1235          _swsetup_Wakeup(ctx);
1236          intel->RenderIndex = ~0;
1237       }
1238    }
1239    else {
1240       intel->Fallback &= ~bit;
1241       if (oldfallback == bit) {
1242          _swrast_flush(ctx);
1243          if (INTEL_DEBUG & DEBUG_FALLBACKS)
1244             fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit));
1245          tnl->Driver.Render.Start = intelRenderStart;
1246          tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1247          tnl->Driver.Render.Finish = intelRenderFinish;
1248          tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1249          tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1250          tnl->Driver.Render.Interp = _tnl_interp;
1251
1252          _tnl_invalidate_vertex_state(ctx, ~0);
1253          _tnl_invalidate_vertices(ctx, ~0);
1254          _tnl_install_attrs(ctx,
1255                             intel->vertex_attrs,
1256                             intel->vertex_attr_count,
1257                             intel->ViewportMatrix.m, 0);
1258
1259          intel->NewGLState |= _INTEL_NEW_RENDERSTATE;
1260       }
1261    }
1262 }
1263
1264 union fi
1265 {
1266    GLfloat f;
1267    GLint i;
1268 };
1269
1270 /**********************************************************************/
1271 /*                            Initialization.                         */
1272 /**********************************************************************/
1273
1274
1275 void
1276 intelInitTriFuncs(struct gl_context * ctx)
1277 {
1278    TNLcontext *tnl = TNL_CONTEXT(ctx);
1279    static int firsttime = 1;
1280
1281    if (firsttime) {
1282       init_rast_tab();
1283       firsttime = 0;
1284    }
1285
1286    tnl->Driver.RunPipeline = intelRunPipeline;
1287    tnl->Driver.Render.Start = intelRenderStart;
1288    tnl->Driver.Render.Finish = intelRenderFinish;
1289    tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1290    tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple;
1291    tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1292    tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1293    tnl->Driver.Render.Interp = _tnl_interp;
1294 }