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 "intel_tex.h"
  56 #include "intel_chipset.h"
  57 #include "i830_context.h"
  58 #include "i830_reg.h"
  59
  60 static void intelRenderPrimitive(GLcontext * ctx, GLenum prim);
  61 static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim,
  62                                  GLuint hwprim);
  63
  64 static void
  65 intel_flush_inline_primitive(struct intel_context *intel)
  66 {
  67    GLuint used = intel->batch->ptr - intel->prim.start_ptr;
  68
  69    assert(intel->prim.primitive != ~0);
  70
  71 /*    _mesa_printf("/\n"); */
  72
  73    if (used < 8)
  74       goto do_discard;
  75
  76    *(int *) intel->prim.start_ptr = (_3DPRIMITIVE |
  77                                      intel->prim.primitive | (used / 4 - 2));
  78
  79    goto finished;
  80
  81  do_discard:
  82    intel->batch->ptr -= used;
  83
  84  finished:
  85    intel->prim.primitive = ~0;
  86    intel->prim.start_ptr = 0;
  87    intel->prim.flush = 0;
  88 }
  89
  90 static void intel_start_inline(struct intel_context *intel, uint32_t prim)
  91 {
  92    BATCH_LOCALS;
  93    uint32_t batch_flags = LOOP_CLIPRECTS;
  94
  95    intel_wait_flips(intel);
  96    intel->vtbl.emit_state(intel);
  97
  98    intel->no_batch_wrap = GL_TRUE;
  99
 100    /*_mesa_printf("%s *", __progname);*/
 101
 102    /* Emit a slot which will be filled with the inline primitive
 103     * command later.
 104     */
 105    BEGIN_BATCH(2, batch_flags);
 106    OUT_BATCH(0);
 107
 108    assert((intel->batch->dirty_state & (1<<1)) == 0);
 109
 110    intel->prim.start_ptr = intel->batch->ptr;
 111    intel->prim.primitive = prim;
 112    intel->prim.flush = intel_flush_inline_primitive;
 113
 114    OUT_BATCH(0);
 115    ADVANCE_BATCH();
 116
 117    intel->no_batch_wrap = GL_FALSE;
 118 /*    _mesa_printf(">"); */
 119 }
 120
 121 static void intel_wrap_inline(struct intel_context *intel)
 122 {
 123    GLuint prim = intel->prim.primitive;
 124
 125    intel_flush_inline_primitive(intel);
 126    intel_batchbuffer_flush(intel->batch);
 127    intel_start_inline(intel, prim);  /* ??? */
 128 }
 129
 130 static GLuint *intel_extend_inline(struct intel_context *intel, GLuint dwords)
 131 {
 132    GLuint sz = dwords * sizeof(GLuint);
 133    GLuint *ptr;
 134
 135    assert(intel->prim.flush == intel_flush_inline_primitive);
 136
 137    if (intel_batchbuffer_space(intel->batch) < sz)
 138       intel_wrap_inline(intel);
 139
 140 /*    _mesa_printf("."); */
 141
 142    intel->vtbl.assert_not_dirty(intel);
 143
 144    ptr = (GLuint *) intel->batch->ptr;
 145    intel->batch->ptr += sz;
 146
 147    return ptr;
 148 }
 149
 150 /** Sets the primitive type for a primitive sequence, flushing as needed. */
 151 void intel_set_prim(struct intel_context *intel, uint32_t prim)
 152 {
 153    /* if we have no VBOs */
 154
 155    if (intel->intelScreen->no_vbo) {
 156       intel_start_inline(intel, prim);
 157       return;
 158    }
 159    if (prim != intel->prim.primitive) {
 160       INTEL_FIREVERTICES(intel);
 161       intel->prim.primitive = prim;
 162    }
 163 }
 164
 165 /** Returns mapped VB space for the given number of vertices */
 166 uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
 167 {
 168    uint32_t *addr;
 169
 170    if (intel->intelScreen->no_vbo) {
 171       return intel_extend_inline(intel, count * intel->vertex_size);
 172    }
 173
 174    /* Check for space in the existing VB */
 175    if (intel->prim.vb_bo == NULL ||
 176        (intel->prim.current_offset +
 177         count * intel->vertex_size * 4) > INTEL_VB_SIZE ||
 178        (intel->prim.count + count) >= (1 << 16)) {
 179       /* Flush existing prim if any */
 180       INTEL_FIREVERTICES(intel);
 181
 182       intel_finish_vb(intel);
 183
 184       /* Start a new VB */
 185       if (intel->prim.vb == NULL)
 186          intel->prim.vb = malloc(INTEL_VB_SIZE);
 187       intel->prim.vb_bo = dri_bo_alloc(intel->bufmgr, "vb",
 188                                        INTEL_VB_SIZE, 4);
 189       intel->prim.start_offset = 0;
 190       intel->prim.current_offset = 0;
 191    }
 192
 193    intel->prim.flush = intel_flush_prim;
 194
 195    addr = (uint32_t *)(intel->prim.vb + intel->prim.current_offset);
 196    intel->prim.current_offset += intel->vertex_size * 4 * count;
 197    intel->prim.count += count;
 198
 199    return addr;
 200 }
 201
 202 /** Dispatches the accumulated primitive to the batchbuffer. */
 203 void intel_flush_prim(struct intel_context *intel)
 204 {
 205    BATCH_LOCALS;
 206    dri_bo *aper_array[2];
 207    dri_bo *vb_bo;
 208    unsigned int offset, count;
 209
 210    /* Must be called after an intel_start_prim. */
 211    assert(intel->prim.primitive != ~0);
 212
 213    if (intel->prim.count == 0)
 214       return;
 215
 216    /* Clear the current prims out of the context state so that a batch flush
 217     * flush triggered by wait_flips or emit_state doesn't loop back to
 218     * flush_prim again.
 219     */
 220    vb_bo = intel->prim.vb_bo;
 221    dri_bo_reference(vb_bo);
 222    count = intel->prim.count;
 223    intel->prim.count = 0;
 224    offset = intel->prim.start_offset;
 225    intel->prim.start_offset = intel->prim.current_offset;
 226    if (!IS_9XX(intel->intelScreen->deviceID))
 227       intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128);
 228    intel->prim.flush = NULL;
 229
 230    intel_wait_flips(intel);
 231
 232    intel->vtbl.emit_state(intel);
 233
 234    aper_array[0] = intel->batch->buf;
 235    aper_array[1] = vb_bo;
 236    if (dri_bufmgr_check_aperture_space(aper_array, 2)) {
 237       intel_batchbuffer_flush(intel->batch);
 238       intel->vtbl.emit_state(intel);
 239    }
 240
 241    /* Ensure that we don't start a new batch for the following emit, which
 242     * depends on the state just emitted. emit_state should be making sure we
 243     * have the space for this.
 244     */
 245    intel->no_batch_wrap = GL_TRUE;
 246
 247    /* Check that we actually emitted the state into this batch, using the
 248     * UPLOAD_CTX bit as the signal.
 249     */
 250    assert((intel->batch->dirty_state & (1<<1)) == 0);
 251
 252 #if 0
 253    printf("emitting %d..%d=%d vertices size %d\n", offset,
 254           intel->prim.current_offset, count,
 255           intel->vertex_size * 4);
 256 #endif
 257
 258    if (IS_9XX(intel->intelScreen->deviceID)) {
 259       BEGIN_BATCH(5, LOOP_CLIPRECTS);
 260       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
 261                 I1_LOAD_S(0) | I1_LOAD_S(1) | 1);
 262       assert((offset & !S0_VB_OFFSET_MASK) == 0);
 263       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, offset);
 264       OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) |
 265                 (intel->vertex_size << S1_VERTEX_PITCH_SHIFT));
 266
 267       OUT_BATCH(_3DPRIMITIVE |
 268                 PRIM_INDIRECT |
 269                 PRIM_INDIRECT_SEQUENTIAL |
 270                 intel->prim.primitive |
 271                 count);
 272       OUT_BATCH(0); /* Beginning vertex index */
 273       ADVANCE_BATCH();
 274    } else {
 275       struct i830_context *i830 = i830_context(&intel->ctx);
 276
 277       BEGIN_BATCH(5, LOOP_CLIPRECTS);
 278       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
 279                 I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
 280       /* S0 */
 281       assert((offset & !S0_VB_OFFSET_MASK_830) == 0);
 282       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
 283                 offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
 284                 S0_VB_ENABLE_830);
 285       /* S2
 286        * This is somewhat unfortunate -- VB width is tied up with
 287        * vertex format data that we've already uploaded through
 288        * _3DSTATE_VFT[01]_CMD.  We may want to replace emits of VFT state with
 289        * STATE_IMMEDIATE_1 like this to avoid duplication.
 290        */
 291       OUT_BATCH((i830->state.Ctx[I830_CTXREG_VF] & VFT0_TEX_COUNT_MASK) >>
 292                 VFT0_TEX_COUNT_SHIFT << S2_TEX_COUNT_SHIFT_830 |
 293                 (i830->state.Ctx[I830_CTXREG_VF2] << 16) |
 294                 intel->vertex_size << S2_VERTEX_0_WIDTH_SHIFT_830);
 295
 296       OUT_BATCH(_3DPRIMITIVE |
 297                 PRIM_INDIRECT |
 298                 PRIM_INDIRECT_SEQUENTIAL |
 299                 intel->prim.primitive |
 300                 count);
 301       OUT_BATCH(0); /* Beginning vertex index */
 302       ADVANCE_BATCH();
 303    }
 304
 305    intel->no_batch_wrap = GL_FALSE;
 306
 307    dri_bo_unreference(vb_bo);
 308 }
 309
 310 /**
 311  * Uploads the locally-accumulated VB into the buffer object.
 312  *
 313  * This avoids us thrashing the cachelines in and out as the buffer gets
 314  * filled, dispatched, then reused as the hardware completes rendering from it,
 315  * and also lets us clflush less if we dispatch with a partially-filled VB.
 316  *
 317  * This is called normally from get_space when we're finishing a BO, but also
 318  * at batch flush time so that we don't try accessing the contents of a
 319  * just-dispatched buffer.
 320  */
 321 void intel_finish_vb(struct intel_context *intel)
 322 {
 323    if (intel->prim.vb_bo == NULL)
 324       return;
 325
 326    dri_bo_subdata(intel->prim.vb_bo, 0, intel->prim.start_offset,
 327                   intel->prim.vb);
 328    dri_bo_unreference(intel->prim.vb_bo);
 329    intel->prim.vb_bo = NULL;
 330 }
 331
 332 /***********************************************************************
 333  *                    Emit primitives as inline vertices               *
 334  ***********************************************************************/
 335
 336 #ifdef __i386__
 337 #define COPY_DWORDS( j, vb, vertsize, v )                       \
 338 do {                                                            \
 339    int __tmp;                                                   \
 340    __asm__ __volatile__( "rep ; movsl"                          \
 341                          : "=%c" (j), "=D" (vb), "=S" (__tmp)   \
 342                          : "0" (vertsize),                      \
 343                          "D" ((long)vb),                        \
 344                          "S" ((long)v) );                       \
 345 } while (0)
 346 #else
 347 #define COPY_DWORDS( j, vb, vertsize, v )       \
 348 do {                                            \
 349    for ( j = 0 ; j < vertsize ; j++ ) {         \
 350       vb[j] = ((GLuint *)v)[j];                 \
 351    }                                            \
 352    vb += vertsize;                              \
 353 } while (0)
 354 #endif
 355
 356 static void
 357 intel_draw_quad(struct intel_context *intel,
 358                 intelVertexPtr v0,
 359                 intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3)
 360 {
 361    GLuint vertsize = intel->vertex_size;
 362    GLuint *vb = intel_get_prim_space(intel, 6);
 363    int j;
 364
 365    COPY_DWORDS(j, vb, vertsize, v0);
 366    COPY_DWORDS(j, vb, vertsize, v1);
 367
 368    /* If smooth shading, draw like a trifan which gives better
 369     * rasterization.  Otherwise draw as two triangles with provoking
 370     * vertex in third position as required for flat shading.
 371     */
 372    if (intel->ctx.Light.ShadeModel == GL_FLAT) {
 373       COPY_DWORDS(j, vb, vertsize, v3);
 374       COPY_DWORDS(j, vb, vertsize, v1);
 375    }
 376    else {
 377       COPY_DWORDS(j, vb, vertsize, v2);
 378       COPY_DWORDS(j, vb, vertsize, v0);
 379    }
 380
 381    COPY_DWORDS(j, vb, vertsize, v2);
 382    COPY_DWORDS(j, vb, vertsize, v3);
 383 }
 384
 385 static void
 386 intel_draw_triangle(struct intel_context *intel,
 387                     intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
 388 {
 389    GLuint vertsize = intel->vertex_size;
 390    GLuint *vb = intel_get_prim_space(intel, 3);
 391    int j;
 392
 393    COPY_DWORDS(j, vb, vertsize, v0);
 394    COPY_DWORDS(j, vb, vertsize, v1);
 395    COPY_DWORDS(j, vb, vertsize, v2);
 396 }
 397
 398
 399 static void
 400 intel_draw_line(struct intel_context *intel,
 401                 intelVertexPtr v0, intelVertexPtr v1)
 402 {
 403    GLuint vertsize = intel->vertex_size;
 404    GLuint *vb = intel_get_prim_space(intel, 2);
 405    int j;
 406
 407    COPY_DWORDS(j, vb, vertsize, v0);
 408    COPY_DWORDS(j, vb, vertsize, v1);
 409 }
 410
 411
 412 static void
 413 intel_draw_point(struct intel_context *intel, intelVertexPtr v0)
 414 {
 415    GLuint vertsize = intel->vertex_size;
 416    GLuint *vb = intel_get_prim_space(intel, 1);
 417    int j;
 418
 419    /* Adjust for sub pixel position -- still required for conform. */
 420    *(float *) &vb[0] = v0->v.x;
 421    *(float *) &vb[1] = v0->v.y;
 422    for (j = 2; j < vertsize; j++)
 423       vb[j] = v0->ui[j];
 424 }
 425
 426
 427
 428 /***********************************************************************
 429  *                Fixup for ARB_point_parameters                       *
 430  ***********************************************************************/
 431
 432 /* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly
 433  * represented in the fragment program InputsRead field.
 434  */
 435 static void
 436 intel_atten_point(struct intel_context *intel, intelVertexPtr v0)
 437 {
 438    GLcontext *ctx = &intel->ctx;
 439    GLfloat psz[4], col[4], restore_psz, restore_alpha;
 440
 441    _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 442    _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 443
 444    restore_psz = psz[0];
 445    restore_alpha = col[3];
 446
 447    if (psz[0] >= ctx->Point.Threshold) {
 448       psz[0] = MIN2(psz[0], ctx->Point.MaxSize);
 449    }
 450    else {
 451       GLfloat dsize = psz[0] / ctx->Point.Threshold;
 452       psz[0] = MAX2(ctx->Point.Threshold, ctx->Point.MinSize);
 453       col[3] *= dsize * dsize;
 454    }
 455
 456    if (psz[0] < 1.0)
 457       psz[0] = 1.0;
 458
 459    if (restore_psz != psz[0] || restore_alpha != col[3]) {
 460       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 461       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 462
 463       intel_draw_point(intel, v0);
 464
 465       psz[0] = restore_psz;
 466       col[3] = restore_alpha;
 467
 468       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 469       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 470    }
 471    else
 472       intel_draw_point(intel, v0);
 473 }
 474
 475
 476
 477
 478
 479 /***********************************************************************
 480  *                Fixup for I915 WPOS texture coordinate                *
 481  ***********************************************************************/
 482
 483
 484
 485 static void
 486 intel_wpos_triangle(struct intel_context *intel,
 487                     intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
 488 {
 489    GLuint offset = intel->wpos_offset;
 490    GLuint size = intel->wpos_size;
 491    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 492    GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
 493    GLfloat *v2_wpos = (GLfloat *)((char *)v2 + offset);
 494
 495    __memcpy(v0_wpos, v0, size);
 496    __memcpy(v1_wpos, v1, size);
 497    __memcpy(v2_wpos, v2, size);
 498
 499    v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
 500    v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
 501    v2_wpos[1] = -v2_wpos[1] + intel->driDrawable->h;
 502
 503
 504    intel_draw_triangle(intel, v0, v1, v2);
 505 }
 506
 507
 508 static void
 509 intel_wpos_line(struct intel_context *intel,
 510                 intelVertexPtr v0, intelVertexPtr v1)
 511 {
 512    GLuint offset = intel->wpos_offset;
 513    GLuint size = intel->wpos_size;
 514    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 515    GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
 516
 517    __memcpy(v0_wpos, v0, size);
 518    __memcpy(v1_wpos, v1, size);
 519
 520    v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
 521    v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
 522
 523    intel_draw_line(intel, v0, v1);
 524 }
 525
 526
 527 static void
 528 intel_wpos_point(struct intel_context *intel, intelVertexPtr v0)
 529 {
 530    GLuint offset = intel->wpos_offset;
 531    GLuint size = intel->wpos_size;
 532    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 533
 534    __memcpy(v0_wpos, v0, size);
 535    v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
 536
 537    intel_draw_point(intel, v0);
 538 }
 539
 540
 541
 542
 543
 544
 545 /***********************************************************************
 546  *          Macros for t_dd_tritmp.h to draw basic primitives          *
 547  ***********************************************************************/
 548
 549 #define TRI( a, b, c )                          \
 550 do {                                            \
 551    if (DO_FALLBACK)                             \
 552       intel->draw_tri( intel, a, b, c );        \
 553    else                                         \
 554       intel_draw_triangle( intel, a, b, c );    \
 555 } while (0)
 556
 557 #define QUAD( a, b, c, d )                      \
 558 do {                                            \
 559    if (DO_FALLBACK) {                           \
 560       intel->draw_tri( intel, a, b, d );        \
 561       intel->draw_tri( intel, b, c, d );        \
 562    } else                                       \
 563       intel_draw_quad( intel, a, b, c, d );     \
 564 } while (0)
 565
 566 #define LINE( v0, v1 )                          \
 567 do {                                            \
 568    if (DO_FALLBACK)                             \
 569       intel->draw_line( intel, v0, v1 );        \
 570    else                                         \
 571       intel_draw_line( intel, v0, v1 );         \
 572 } while (0)
 573
 574 #define POINT( v0 )                             \
 575 do {                                            \
 576    if (DO_FALLBACK)                             \
 577       intel->draw_point( intel, v0 );           \
 578    else                                         \
 579       intel_draw_point( intel, v0 );            \
 580 } while (0)
 581
 582
 583 /***********************************************************************
 584  *              Build render functions from dd templates               *
 585  ***********************************************************************/
 586
 587 #define INTEL_OFFSET_BIT        0x01
 588 #define INTEL_TWOSIDE_BIT       0x02
 589 #define INTEL_UNFILLED_BIT      0x04
 590 #define INTEL_FALLBACK_BIT      0x08
 591 #define INTEL_MAX_TRIFUNC       0x10
 592
 593
 594 static struct
 595 {
 596    tnl_points_func points;
 597    tnl_line_func line;
 598    tnl_triangle_func triangle;
 599    tnl_quad_func quad;
 600 } rast_tab[INTEL_MAX_TRIFUNC];
 601
 602
 603 #define DO_FALLBACK (IND & INTEL_FALLBACK_BIT)
 604 #define DO_OFFSET   (IND & INTEL_OFFSET_BIT)
 605 #define DO_UNFILLED (IND & INTEL_UNFILLED_BIT)
 606 #define DO_TWOSIDE  (IND & INTEL_TWOSIDE_BIT)
 607 #define DO_FLAT      0
 608 #define DO_TRI       1
 609 #define DO_QUAD      1
 610 #define DO_LINE      1
 611 #define DO_POINTS    1
 612 #define DO_FULL_QUAD 1
 613
 614 #define HAVE_RGBA         1
 615 #define HAVE_SPEC         1
 616 #define HAVE_BACK_COLORS  0
 617 #define HAVE_HW_FLATSHADE 1
 618 #define VERTEX            intelVertex
 619 #define TAB               rast_tab
 620
 621 /* Only used to pull back colors into vertices (ie, we know color is
 622  * floating point).
 623  */
 624 #define INTEL_COLOR( 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    UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]);        \
 630 } while (0)
 631
 632 #define INTEL_SPEC( dst, src )                          \
 633 do {                                                    \
 634    UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);        \
 635    UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);        \
 636    UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);        \
 637 } while (0)
 638
 639
 640 #define DEPTH_SCALE intel->polygon_offset_scale
 641 #define UNFILLED_TRI unfilled_tri
 642 #define UNFILLED_QUAD unfilled_quad
 643 #define VERT_X(_v) _v->v.x
 644 #define VERT_Y(_v) _v->v.y
 645 #define VERT_Z(_v) _v->v.z
 646 #define AREA_IS_CCW( a ) (a > 0)
 647 #define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint)))
 648
 649 #define VERT_SET_RGBA( v, c )    if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c )
 650 #define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
 651 #define VERT_SAVE_RGBA( idx )    if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
 652 #define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
 653
 654 #define VERT_SET_SPEC( v, c )    if (specoffset) INTEL_SPEC( v->ub4[specoffset], c )
 655 #define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
 656 #define VERT_SAVE_SPEC( idx )    if (specoffset) spec[idx] = v[idx]->ui[specoffset]
 657 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
 658
 659 #define LOCAL_VARS(n)                                                   \
 660    struct intel_context *intel = intel_context(ctx);                    \
 661    GLuint color[n] = { 0, }, spec[n] = { 0, };                          \
 662    GLuint coloroffset = intel->coloroffset;                             \
 663    GLboolean specoffset = intel->specoffset;                            \
 664    (void) color; (void) spec; (void) coloroffset; (void) specoffset;
 665
 666
 667 /***********************************************************************
 668  *                Helpers for rendering unfilled primitives            *
 669  ***********************************************************************/
 670
 671 static const GLuint hw_prim[GL_POLYGON + 1] = {
 672    PRIM3D_POINTLIST,
 673    PRIM3D_LINELIST,
 674    PRIM3D_LINELIST,
 675    PRIM3D_LINELIST,
 676    PRIM3D_TRILIST,
 677    PRIM3D_TRILIST,
 678    PRIM3D_TRILIST,
 679    PRIM3D_TRILIST,
 680    PRIM3D_TRILIST,
 681    PRIM3D_TRILIST
 682 };
 683
 684 #define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] )
 685 #define RENDER_PRIMITIVE intel->render_primitive
 686 #define TAG(x) x
 687 #define IND INTEL_FALLBACK_BIT
 688 #include "tnl_dd/t_dd_unfilled.h"
 689 #undef IND
 690
 691 /***********************************************************************
 692  *                      Generate GL render functions                   *
 693  ***********************************************************************/
 694
 695 #define IND (0)
 696 #define TAG(x) x
 697 #include "tnl_dd/t_dd_tritmp.h"
 698
 699 #define IND (INTEL_OFFSET_BIT)
 700 #define TAG(x) x##_offset
 701 #include "tnl_dd/t_dd_tritmp.h"
 702
 703 #define IND (INTEL_TWOSIDE_BIT)
 704 #define TAG(x) x##_twoside
 705 #include "tnl_dd/t_dd_tritmp.h"
 706
 707 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT)
 708 #define TAG(x) x##_twoside_offset
 709 #include "tnl_dd/t_dd_tritmp.h"
 710
 711 #define IND (INTEL_UNFILLED_BIT)
 712 #define TAG(x) x##_unfilled
 713 #include "tnl_dd/t_dd_tritmp.h"
 714
 715 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
 716 #define TAG(x) x##_offset_unfilled
 717 #include "tnl_dd/t_dd_tritmp.h"
 718
 719 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT)
 720 #define TAG(x) x##_twoside_unfilled
 721 #include "tnl_dd/t_dd_tritmp.h"
 722
 723 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
 724 #define TAG(x) x##_twoside_offset_unfilled
 725 #include "tnl_dd/t_dd_tritmp.h"
 726
 727 #define IND (INTEL_FALLBACK_BIT)
 728 #define TAG(x) x##_fallback
 729 #include "tnl_dd/t_dd_tritmp.h"
 730
 731 #define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
 732 #define TAG(x) x##_offset_fallback
 733 #include "tnl_dd/t_dd_tritmp.h"
 734
 735 #define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT)
 736 #define TAG(x) x##_twoside_fallback
 737 #include "tnl_dd/t_dd_tritmp.h"
 738
 739 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
 740 #define TAG(x) x##_twoside_offset_fallback
 741 #include "tnl_dd/t_dd_tritmp.h"
 742
 743 #define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 744 #define TAG(x) x##_unfilled_fallback
 745 #include "tnl_dd/t_dd_tritmp.h"
 746
 747 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 748 #define TAG(x) x##_offset_unfilled_fallback
 749 #include "tnl_dd/t_dd_tritmp.h"
 750
 751 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 752 #define TAG(x) x##_twoside_unfilled_fallback
 753 #include "tnl_dd/t_dd_tritmp.h"
 754
 755 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \
 756              INTEL_FALLBACK_BIT)
 757 #define TAG(x) x##_twoside_offset_unfilled_fallback
 758 #include "tnl_dd/t_dd_tritmp.h"
 759
 760
 761 static void
 762 init_rast_tab(void)
 763 {
 764    init();
 765    init_offset();
 766    init_twoside();
 767    init_twoside_offset();
 768    init_unfilled();
 769    init_offset_unfilled();
 770    init_twoside_unfilled();
 771    init_twoside_offset_unfilled();
 772    init_fallback();
 773    init_offset_fallback();
 774    init_twoside_fallback();
 775    init_twoside_offset_fallback();
 776    init_unfilled_fallback();
 777    init_offset_unfilled_fallback();
 778    init_twoside_unfilled_fallback();
 779    init_twoside_offset_unfilled_fallback();
 780 }
 781
 782
 783 /***********************************************************************
 784  *                    Rasterization fallback helpers                   *
 785  ***********************************************************************/
 786
 787
 788 /* This code is hit only when a mix of accelerated and unaccelerated
 789  * primitives are being drawn, and only for the unaccelerated
 790  * primitives.
 791  */
 792 static void
 793 intel_fallback_tri(struct intel_context *intel,
 794                    intelVertex * v0, intelVertex * v1, intelVertex * v2)
 795 {
 796    GLcontext *ctx = &intel->ctx;
 797    SWvertex v[3];
 798
 799    if (0)
 800       fprintf(stderr, "\n%s\n", __FUNCTION__);
 801
 802    INTEL_FIREVERTICES(intel);
 803
 804    _swsetup_Translate(ctx, v0, &v[0]);
 805    _swsetup_Translate(ctx, v1, &v[1]);
 806    _swsetup_Translate(ctx, v2, &v[2]);
 807    intelSpanRenderStart(ctx);
 808    _swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
 809    intelSpanRenderFinish(ctx);
 810 }
 811
 812
 813 static void
 814 intel_fallback_line(struct intel_context *intel,
 815                     intelVertex * v0, intelVertex * v1)
 816 {
 817    GLcontext *ctx = &intel->ctx;
 818    SWvertex v[2];
 819
 820    if (0)
 821       fprintf(stderr, "\n%s\n", __FUNCTION__);
 822
 823    INTEL_FIREVERTICES(intel);
 824
 825    _swsetup_Translate(ctx, v0, &v[0]);
 826    _swsetup_Translate(ctx, v1, &v[1]);
 827    intelSpanRenderStart(ctx);
 828    _swrast_Line(ctx, &v[0], &v[1]);
 829    intelSpanRenderFinish(ctx);
 830 }
 831
 832 static void
 833 intel_fallback_point(struct intel_context *intel,
 834                      intelVertex * v0)
 835 {
 836    GLcontext *ctx = &intel->ctx;
 837    SWvertex v[1];
 838
 839    if (0)
 840       fprintf(stderr, "\n%s\n", __FUNCTION__);
 841
 842    INTEL_FIREVERTICES(intel);
 843
 844    _swsetup_Translate(ctx, v0, &v[0]);
 845    intelSpanRenderStart(ctx);
 846    _swrast_Point(ctx, &v[0]);
 847    intelSpanRenderFinish(ctx);
 848 }
 849
 850
 851 /**********************************************************************/
 852 /*               Render unclipped begin/end objects                   */
 853 /**********************************************************************/
 854
 855 #define IND 0
 856 #define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint)))
 857 #define RENDER_POINTS( start, count )   \
 858    for ( ; start < count ; start++) POINT( V(ELT(start)) );
 859 #define RENDER_LINE( v0, v1 )         LINE( V(v0), V(v1) )
 860 #define RENDER_TRI(  v0, v1, v2 )     TRI(  V(v0), V(v1), V(v2) )
 861 #define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) )
 862 #define INIT(x) intelRenderPrimitive( ctx, x )
 863 #undef LOCAL_VARS
 864 #define LOCAL_VARS                                              \
 865     struct intel_context *intel = intel_context(ctx);                   \
 866     GLubyte *vertptr = (GLubyte *)intel->verts;                 \
 867     const GLuint vertsize = intel->vertex_size;         \
 868     const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;       \
 869     (void) elt;
 870 #define RESET_STIPPLE
 871 #define RESET_OCCLUSION
 872 #define PRESERVE_VB_DEFS
 873 #define ELT(x) x
 874 #define TAG(x) intel_##x##_verts
 875 #include "tnl/t_vb_rendertmp.h"
 876 #undef ELT
 877 #undef TAG
 878 #define TAG(x) intel_##x##_elts
 879 #define ELT(x) elt[x]
 880 #include "tnl/t_vb_rendertmp.h"
 881
 882 /**********************************************************************/
 883 /*                   Render clipped primitives                        */
 884 /**********************************************************************/
 885
 886
 887
 888 static void
 889 intelRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
 890 {
 891    struct intel_context *intel = intel_context(ctx);
 892    TNLcontext *tnl = TNL_CONTEXT(ctx);
 893    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
 894    GLuint prim = intel->render_primitive;
 895
 896    /* Render the new vertices as an unclipped polygon.
 897     */
 898    {
 899       GLuint *tmp = VB->Elts;
 900       VB->Elts = (GLuint *) elts;
 901       tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n,
 902                                                   PRIM_BEGIN | PRIM_END);
 903       VB->Elts = tmp;
 904    }
 905
 906    /* Restore the render primitive
 907     */
 908    if (prim != GL_POLYGON)
 909       tnl->Driver.Render.PrimitiveNotify(ctx, prim);
 910 }
 911
 912 static void
 913 intelRenderClippedLine(GLcontext * ctx, GLuint ii, GLuint jj)
 914 {
 915    TNLcontext *tnl = TNL_CONTEXT(ctx);
 916
 917    tnl->Driver.Render.Line(ctx, ii, jj);
 918 }
 919
 920 static void
 921 intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
 922 {
 923    struct intel_context *intel = intel_context(ctx);
 924    const GLuint vertsize = intel->vertex_size;
 925    GLuint *vb = intel_get_prim_space(intel, (n - 2) * 3);
 926    GLubyte *vertptr = (GLubyte *) intel->verts;
 927    const GLuint *start = (const GLuint *) V(elts[0]);
 928    int i, j;
 929
 930    for (i = 2; i < n; i++) {
 931       COPY_DWORDS(j, vb, vertsize, V(elts[i - 1]));
 932       COPY_DWORDS(j, vb, vertsize, V(elts[i]));
 933       COPY_DWORDS(j, vb, vertsize, start);
 934    }
 935 }
 936
 937 /**********************************************************************/
 938 /*                    Choose render functions                         */
 939 /**********************************************************************/
 940
 941
 942
 943
 944 #define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN | DD_POINT_SMOOTH | DD_TRI_SMOOTH)
 945 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED)
 946
 947 void
 948 intelChooseRenderState(GLcontext * ctx)
 949 {
 950    TNLcontext *tnl = TNL_CONTEXT(ctx);
 951    struct intel_context *intel = intel_context(ctx);
 952    GLuint flags = ctx->_TriangleCaps;
 953    const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current;
 954    GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS));
 955    GLuint index = 0;
 956
 957    if (INTEL_DEBUG & DEBUG_STATE)
 958       fprintf(stderr, "\n%s\n", __FUNCTION__);
 959
 960    if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) {
 961
 962       if (flags & ANY_RASTER_FLAGS) {
 963          if (flags & DD_TRI_LIGHT_TWOSIDE)
 964             index |= INTEL_TWOSIDE_BIT;
 965          if (flags & DD_TRI_OFFSET)
 966             index |= INTEL_OFFSET_BIT;
 967          if (flags & DD_TRI_UNFILLED)
 968             index |= INTEL_UNFILLED_BIT;
 969       }
 970
 971       if (have_wpos) {
 972          intel->draw_point = intel_wpos_point;
 973          intel->draw_line = intel_wpos_line;
 974          intel->draw_tri = intel_wpos_triangle;
 975
 976          /* Make sure these get called:
 977           */
 978          index |= INTEL_FALLBACK_BIT;
 979       }
 980       else {
 981          intel->draw_point = intel_draw_point;
 982          intel->draw_line = intel_draw_line;
 983          intel->draw_tri = intel_draw_triangle;
 984       }
 985
 986       /* Hook in fallbacks for specific primitives.
 987        */
 988       if (flags & ANY_FALLBACK_FLAGS) {
 989          if (flags & DD_LINE_STIPPLE)
 990             intel->draw_line = intel_fallback_line;
 991
 992          if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple)
 993             intel->draw_tri = intel_fallback_tri;
 994
 995          if (flags & DD_TRI_SMOOTH) {
 996             if (intel->strict_conformance)
 997                intel->draw_tri = intel_fallback_tri;
 998          }
 999
1000          if (flags & DD_POINT_ATTEN) {
1001             if (0)
1002                intel->draw_point = intel_atten_point;
1003             else
1004                intel->draw_point = intel_fallback_point;
1005          }
1006
1007          if (flags & DD_POINT_SMOOTH) {
1008             if (intel->strict_conformance)
1009                intel->draw_point = intel_fallback_point;
1010          }
1011
1012          index |= INTEL_FALLBACK_BIT;
1013       }
1014    }
1015
1016    if (intel->RenderIndex != index) {
1017       intel->RenderIndex = index;
1018
1019       tnl->Driver.Render.Points = rast_tab[index].points;
1020       tnl->Driver.Render.Line = rast_tab[index].line;
1021       tnl->Driver.Render.Triangle = rast_tab[index].triangle;
1022       tnl->Driver.Render.Quad = rast_tab[index].quad;
1023
1024       if (index == 0) {
1025          tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts;
1026          tnl->Driver.Render.PrimTabElts = intel_render_tab_elts;
1027          tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */
1028          tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly;
1029       }
1030       else {
1031          tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
1032          tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
1033          tnl->Driver.Render.ClippedLine = intelRenderClippedLine;
1034          tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly;
1035       }
1036    }
1037 }
1038
1039 static const GLenum reduced_prim[GL_POLYGON + 1] = {
1040    GL_POINTS,
1041    GL_LINES,
1042    GL_LINES,
1043    GL_LINES,
1044    GL_TRIANGLES,
1045    GL_TRIANGLES,
1046    GL_TRIANGLES,
1047    GL_TRIANGLES,
1048    GL_TRIANGLES,
1049    GL_TRIANGLES
1050 };
1051
1052
1053 /**********************************************************************/
1054 /*                 High level hooks for t_vb_render.c                 */
1055 /**********************************************************************/
1056
1057
1058
1059
1060 static void
1061 intelRunPipeline(GLcontext * ctx)
1062 {
1063    struct intel_context *intel = intel_context(ctx);
1064
1065    _mesa_lock_context_textures(ctx);
1066
1067    if (ctx->NewState)
1068       _mesa_update_state_locked(ctx);
1069
1070    if (intel->NewGLState) {
1071       if (intel->NewGLState & _NEW_TEXTURE) {
1072          intel->vtbl.update_texture_state(intel);
1073       }
1074
1075       if (!intel->Fallback) {
1076          if (intel->NewGLState & _INTEL_NEW_RENDERSTATE)
1077             intelChooseRenderState(ctx);
1078       }
1079
1080       intel->NewGLState = 0;
1081    }
1082
1083    _tnl_run_pipeline(ctx);
1084
1085    _mesa_unlock_context_textures(ctx);
1086 }
1087
1088 static void
1089 intelRenderStart(GLcontext * ctx)
1090 {
1091    struct intel_context *intel = intel_context(ctx);
1092
1093    intel->vtbl.render_start(intel_context(ctx));
1094    intel->vtbl.emit_state(intel);
1095 }
1096
1097 static void
1098 intelRenderFinish(GLcontext * ctx)
1099 {
1100    struct intel_context *intel = intel_context(ctx);
1101
1102    if (intel->RenderIndex & INTEL_FALLBACK_BIT)
1103       _swrast_flush(ctx);
1104
1105    INTEL_FIREVERTICES(intel);
1106 }
1107
1108
1109
1110
1111  /* System to flush dma and emit state changes based on the rasterized
1112   * primitive.
1113   */
1114 static void
1115 intelRasterPrimitive(GLcontext * ctx, GLenum rprim, GLuint hwprim)
1116 {
1117    struct intel_context *intel = intel_context(ctx);
1118
1119    if (0)
1120       fprintf(stderr, "%s %s %x\n", __FUNCTION__,
1121               _mesa_lookup_enum_by_nr(rprim), hwprim);
1122
1123    intel->vtbl.reduced_primitive_state(intel, rprim);
1124
1125    /* Start a new primitive.  Arrange to have it flushed later on.
1126     */
1127    if (hwprim != intel->prim.primitive) {
1128       INTEL_FIREVERTICES(intel);
1129
1130       intel_set_prim(intel, hwprim);
1131    }
1132 }
1133
1134
1135  /*
1136   */
1137 static void
1138 intelRenderPrimitive(GLcontext * ctx, GLenum prim)
1139 {
1140    struct intel_context *intel = intel_context(ctx);
1141
1142    if (0)
1143       fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
1144
1145    /* Let some clipping routines know which primitive they're dealing
1146     * with.
1147     */
1148    intel->render_primitive = prim;
1149
1150    /* Shortcircuit this when called from t_dd_rendertmp.h for unfilled
1151     * triangles.  The rasterized primitive will always be reset by
1152     * lower level functions in that case, potentially pingponging the
1153     * state:
1154     */
1155    if (reduced_prim[prim] == GL_TRIANGLES &&
1156        (ctx->_TriangleCaps & DD_TRI_UNFILLED))
1157       return;
1158
1159    /* Set some primitive-dependent state and Start? a new primitive.
1160     */
1161    intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]);
1162 }
1163
1164
1165  /**********************************************************************/
1166  /*           Transition to/from hardware rasterization.               */
1167  /**********************************************************************/
1168
1169 static char *fallbackStrings[] = {
1170    [0] = "Draw buffer",
1171    [1] = "Read buffer",
1172    [2] = "Depth buffer",
1173    [3] = "Stencil buffer",
1174    [4] = "User disable",
1175    [5] = "Render mode",
1176
1177    [12] = "Texture",
1178    [13] = "Color mask",
1179    [14] = "Stencil",
1180    [15] = "Stipple",
1181    [16] = "Program",
1182    [17] = "Logic op",
1183    [18] = "Smooth polygon",
1184    [19] = "Smooth point",
1185 };
1186
1187
1188 static char *
1189 getFallbackString(GLuint bit)
1190 {
1191    int i = 0;
1192    while (bit > 1) {
1193       i++;
1194       bit >>= 1;
1195    }
1196    return fallbackStrings[i];
1197 }
1198
1199
1200
1201 void
1202 intelFallback(struct intel_context *intel, GLuint bit, GLboolean mode)
1203 {
1204    GLcontext *ctx = &intel->ctx;
1205    TNLcontext *tnl = TNL_CONTEXT(ctx);
1206    GLuint oldfallback = intel->Fallback;
1207
1208    if (mode) {
1209       intel->Fallback |= bit;
1210       if (oldfallback == 0) {
1211          intelFlush(ctx);
1212          if (INTEL_DEBUG & DEBUG_FALLBACKS)
1213             fprintf(stderr, "ENTER FALLBACK %x: %s\n",
1214                     bit, getFallbackString(bit));
1215          _swsetup_Wakeup(ctx);
1216          intel->RenderIndex = ~0;
1217       }
1218    }
1219    else {
1220       intel->Fallback &= ~bit;
1221       if (oldfallback == bit) {
1222          _swrast_flush(ctx);
1223          if (INTEL_DEBUG & DEBUG_FALLBACKS)
1224             fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit));
1225          tnl->Driver.Render.Start = intelRenderStart;
1226          tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1227          tnl->Driver.Render.Finish = intelRenderFinish;
1228          tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1229          tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1230          tnl->Driver.Render.Interp = _tnl_interp;
1231
1232          _tnl_invalidate_vertex_state(ctx, ~0);
1233          _tnl_invalidate_vertices(ctx, ~0);
1234          _tnl_install_attrs(ctx,
1235                             intel->vertex_attrs,
1236                             intel->vertex_attr_count,
1237                             intel->ViewportMatrix.m, 0);
1238
1239          intel->NewGLState |= _INTEL_NEW_RENDERSTATE;
1240       }
1241    }
1242 }
1243
1244 union fi
1245 {
1246    GLfloat f;
1247    GLint i;
1248 };
1249
1250
1251 /**********************************************************************/
1252 /*             Used only with the metaops callbacks.                  */
1253 /**********************************************************************/
1254 static void
1255 intel_meta_draw_poly(struct intel_context *intel,
1256                      GLuint n,
1257                      GLfloat xy[][2],
1258                      GLfloat z, GLuint color, GLfloat tex[][2])
1259 {
1260    union fi *vb;
1261    GLint i;
1262    GLboolean was_locked = intel->locked;
1263    unsigned int saved_vertex_size = intel->vertex_size;
1264
1265    if (!was_locked)
1266        LOCK_HARDWARE(intel);
1267
1268    intel->vertex_size = 6;
1269
1270    /* All 3d primitives should be emitted with LOOP_CLIPRECTS,
1271     * otherwise the drawing origin (DR4) might not be set correctly.
1272     */
1273    intel_set_prim(intel, PRIM3D_TRIFAN);
1274    vb = (union fi *) intel_get_prim_space(intel, n);
1275
1276    for (i = 0; i < n; i++) {
1277       vb[0].f = xy[i][0];
1278       vb[1].f = xy[i][1];
1279       vb[2].f = z;
1280       vb[3].i = color;
1281       vb[4].f = tex[i][0];
1282       vb[5].f = tex[i][1];
1283       vb += 6;
1284    }
1285
1286    INTEL_FIREVERTICES(intel);
1287
1288    intel->vertex_size = saved_vertex_size;
1289
1290    if (!was_locked)
1291        UNLOCK_HARDWARE(intel);
1292 }
1293
1294 static void
1295 intel_meta_draw_quad(struct intel_context *intel,
1296                      GLfloat x0, GLfloat x1,
1297                      GLfloat y0, GLfloat y1,
1298                      GLfloat z,
1299                      GLuint color,
1300                      GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1)
1301 {
1302    GLfloat xy[4][2];
1303    GLfloat tex[4][2];
1304
1305    xy[0][0] = x0;
1306    xy[0][1] = y0;
1307    xy[1][0] = x1;
1308    xy[1][1] = y0;
1309    xy[2][0] = x1;
1310    xy[2][1] = y1;
1311    xy[3][0] = x0;
1312    xy[3][1] = y1;
1313
1314    tex[0][0] = s0;
1315    tex[0][1] = t0;
1316    tex[1][0] = s1;
1317    tex[1][1] = t0;
1318    tex[2][0] = s1;
1319    tex[2][1] = t1;
1320    tex[3][0] = s0;
1321    tex[3][1] = t1;
1322
1323    intel_meta_draw_poly(intel, 4, xy, z, color, tex);
1324 }
1325
1326
1327
1328 /**********************************************************************/
1329 /*                            Initialization.                         */
1330 /**********************************************************************/
1331
1332
1333 void
1334 intelInitTriFuncs(GLcontext * ctx)
1335 {
1336    struct intel_context *intel = intel_context(ctx);
1337    TNLcontext *tnl = TNL_CONTEXT(ctx);
1338    static int firsttime = 1;
1339
1340    if (firsttime) {
1341       init_rast_tab();
1342       firsttime = 0;
1343    }
1344
1345    tnl->Driver.RunPipeline = intelRunPipeline;
1346    tnl->Driver.Render.Start = intelRenderStart;
1347    tnl->Driver.Render.Finish = intelRenderFinish;
1348    tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1349    tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple;
1350    tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1351    tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1352    tnl->Driver.Render.Interp = _tnl_interp;
1353
1354    intel->vtbl.meta_draw_quad = intel_meta_draw_quad;
1355 }