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    uint32_t batch_flags = LOOP_CLIPRECTS;
  93    BATCH_LOCALS;
  94
  95    intel->vtbl.emit_state(intel);
  96
  97    intel->no_batch_wrap = GL_TRUE;
  98
  99    /*_mesa_printf("%s *", __progname);*/
 100
 101    /* Emit a slot which will be filled with the inline primitive
 102     * command later.
 103     */
 104    BEGIN_BATCH(2, batch_flags);
 105    OUT_BATCH(0);
 106
 107    assert((intel->batch->dirty_state & (1<<1)) == 0);
 108
 109    intel->prim.start_ptr = intel->batch->ptr;
 110    intel->prim.primitive = prim;
 111    intel->prim.flush = intel_flush_inline_primitive;
 112
 113    OUT_BATCH(0);
 114    ADVANCE_BATCH();
 115
 116    intel->no_batch_wrap = GL_FALSE;
 117 /*    _mesa_printf(">"); */
 118 }
 119
 120 static void intel_wrap_inline(struct intel_context *intel)
 121 {
 122    GLuint prim = intel->prim.primitive;
 123
 124    intel_flush_inline_primitive(intel);
 125    intel_batchbuffer_flush(intel->batch);
 126    intel_start_inline(intel, prim);  /* ??? */
 127 }
 128
 129 static GLuint *intel_extend_inline(struct intel_context *intel, GLuint dwords)
 130 {
 131    GLuint sz = dwords * sizeof(GLuint);
 132    GLuint *ptr;
 133
 134    assert(intel->prim.flush == intel_flush_inline_primitive);
 135
 136    if (intel_batchbuffer_space(intel->batch) < sz)
 137       intel_wrap_inline(intel);
 138
 139 /*    _mesa_printf("."); */
 140
 141    intel->vtbl.assert_not_dirty(intel);
 142
 143    ptr = (GLuint *) intel->batch->ptr;
 144    intel->batch->ptr += sz;
 145
 146    return ptr;
 147 }
 148
 149 /** Sets the primitive type for a primitive sequence, flushing as needed. */
 150 void intel_set_prim(struct intel_context *intel, uint32_t prim)
 151 {
 152    /* if we have no VBOs */
 153
 154    if (intel->intelScreen->no_vbo) {
 155       intel_start_inline(intel, prim);
 156       return;
 157    }
 158    if (prim != intel->prim.primitive) {
 159       INTEL_FIREVERTICES(intel);
 160       intel->prim.primitive = prim;
 161    }
 162 }
 163
 164 /** Returns mapped VB space for the given number of vertices */
 165 uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
 166 {
 167    uint32_t *addr;
 168
 169    if (intel->intelScreen->no_vbo) {
 170       return intel_extend_inline(intel, count * intel->vertex_size);
 171    }
 172
 173    /* Check for space in the existing VB */
 174    if (intel->prim.vb_bo == NULL ||
 175        (intel->prim.current_offset +
 176         count * intel->vertex_size * 4) > INTEL_VB_SIZE ||
 177        (intel->prim.count + count) >= (1 << 16)) {
 178       /* Flush existing prim if any */
 179       INTEL_FIREVERTICES(intel);
 180
 181       intel_finish_vb(intel);
 182
 183       /* Start a new VB */
 184       if (intel->prim.vb == NULL)
 185          intel->prim.vb = malloc(INTEL_VB_SIZE);
 186       intel->prim.vb_bo = dri_bo_alloc(intel->bufmgr, "vb",
 187                                        INTEL_VB_SIZE, 4);
 188       intel->prim.start_offset = 0;
 189       intel->prim.current_offset = 0;
 190    }
 191
 192    intel->prim.flush = intel_flush_prim;
 193
 194    addr = (uint32_t *)(intel->prim.vb + intel->prim.current_offset);
 195    intel->prim.current_offset += intel->vertex_size * 4 * count;
 196    intel->prim.count += count;
 197
 198    return addr;
 199 }
 200
 201 /** Dispatches the accumulated primitive to the batchbuffer. */
 202 void intel_flush_prim(struct intel_context *intel)
 203 {
 204    dri_bo *aper_array[2];
 205    dri_bo *vb_bo;
 206    unsigned int offset, count;
 207    BATCH_LOCALS;
 208
 209    /* Must be called after an intel_start_prim. */
 210    assert(intel->prim.primitive != ~0);
 211
 212    if (intel->prim.count == 0)
 213       return;
 214
 215    /* Clear the current prims out of the context state so that a batch flush
 216     * flush triggered by emit_state doesn't loop back to flush_prim again.
 217     */
 218    vb_bo = intel->prim.vb_bo;
 219    dri_bo_reference(vb_bo);
 220    count = intel->prim.count;
 221    intel->prim.count = 0;
 222    offset = intel->prim.start_offset;
 223    intel->prim.start_offset = intel->prim.current_offset;
 224    if (!IS_9XX(intel->intelScreen->deviceID))
 225       intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128);
 226    intel->prim.flush = NULL;
 227
 228    intel->vtbl.emit_state(intel);
 229
 230    aper_array[0] = intel->batch->buf;
 231    aper_array[1] = vb_bo;
 232    if (dri_bufmgr_check_aperture_space(aper_array, 2)) {
 233       intel_batchbuffer_flush(intel->batch);
 234       intel->vtbl.emit_state(intel);
 235    }
 236
 237    /* Ensure that we don't start a new batch for the following emit, which
 238     * depends on the state just emitted. emit_state should be making sure we
 239     * have the space for this.
 240     */
 241    intel->no_batch_wrap = GL_TRUE;
 242
 243    /* Check that we actually emitted the state into this batch, using the
 244     * UPLOAD_CTX bit as the signal.
 245     */
 246    assert((intel->batch->dirty_state & (1<<1)) == 0);
 247
 248 #if 0
 249    printf("emitting %d..%d=%d vertices size %d\n", offset,
 250           intel->prim.current_offset, count,
 251           intel->vertex_size * 4);
 252 #endif
 253
 254    if (IS_9XX(intel->intelScreen->deviceID)) {
 255       BEGIN_BATCH(5, LOOP_CLIPRECTS);
 256       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
 257                 I1_LOAD_S(0) | I1_LOAD_S(1) | 1);
 258       assert((offset & !S0_VB_OFFSET_MASK) == 0);
 259       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, offset);
 260       OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) |
 261                 (intel->vertex_size << S1_VERTEX_PITCH_SHIFT));
 262
 263       OUT_BATCH(_3DPRIMITIVE |
 264                 PRIM_INDIRECT |
 265                 PRIM_INDIRECT_SEQUENTIAL |
 266                 intel->prim.primitive |
 267                 count);
 268       OUT_BATCH(0); /* Beginning vertex index */
 269       ADVANCE_BATCH();
 270    } else {
 271       struct i830_context *i830 = i830_context(&intel->ctx);
 272
 273       BEGIN_BATCH(5, LOOP_CLIPRECTS);
 274       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
 275                 I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
 276       /* S0 */
 277       assert((offset & !S0_VB_OFFSET_MASK_830) == 0);
 278       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
 279                 offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
 280                 S0_VB_ENABLE_830);
 281       /* S2
 282        * This is somewhat unfortunate -- VB width is tied up with
 283        * vertex format data that we've already uploaded through
 284        * _3DSTATE_VFT[01]_CMD.  We may want to replace emits of VFT state with
 285        * STATE_IMMEDIATE_1 like this to avoid duplication.
 286        */
 287       OUT_BATCH((i830->state.Ctx[I830_CTXREG_VF] & VFT0_TEX_COUNT_MASK) >>
 288                 VFT0_TEX_COUNT_SHIFT << S2_TEX_COUNT_SHIFT_830 |
 289                 (i830->state.Ctx[I830_CTXREG_VF2] << 16) |
 290                 intel->vertex_size << S2_VERTEX_0_WIDTH_SHIFT_830);
 291
 292       OUT_BATCH(_3DPRIMITIVE |
 293                 PRIM_INDIRECT |
 294                 PRIM_INDIRECT_SEQUENTIAL |
 295                 intel->prim.primitive |
 296                 count);
 297       OUT_BATCH(0); /* Beginning vertex index */
 298       ADVANCE_BATCH();
 299    }
 300
 301    intel->no_batch_wrap = GL_FALSE;
 302
 303    dri_bo_unreference(vb_bo);
 304 }
 305
 306 /**
 307  * Uploads the locally-accumulated VB into the buffer object.
 308  *
 309  * This avoids us thrashing the cachelines in and out as the buffer gets
 310  * filled, dispatched, then reused as the hardware completes rendering from it,
 311  * and also lets us clflush less if we dispatch with a partially-filled VB.
 312  *
 313  * This is called normally from get_space when we're finishing a BO, but also
 314  * at batch flush time so that we don't try accessing the contents of a
 315  * just-dispatched buffer.
 316  */
 317 void intel_finish_vb(struct intel_context *intel)
 318 {
 319    if (intel->prim.vb_bo == NULL)
 320       return;
 321
 322    dri_bo_subdata(intel->prim.vb_bo, 0, intel->prim.start_offset,
 323                   intel->prim.vb);
 324    dri_bo_unreference(intel->prim.vb_bo);
 325    intel->prim.vb_bo = NULL;
 326 }
 327
 328 /***********************************************************************
 329  *                    Emit primitives as inline vertices               *
 330  ***********************************************************************/
 331
 332 #ifdef __i386__
 333 #define COPY_DWORDS( j, vb, vertsize, v )                       \
 334 do {                                                            \
 335    int __tmp;                                                   \
 336    __asm__ __volatile__( "rep ; movsl"                          \
 337                          : "=%c" (j), "=D" (vb), "=S" (__tmp)   \
 338                          : "0" (vertsize),                      \
 339                          "D" ((long)vb),                        \
 340                          "S" ((long)v) );                       \
 341 } while (0)
 342 #else
 343 #define COPY_DWORDS( j, vb, vertsize, v )       \
 344 do {                                            \
 345    for ( j = 0 ; j < vertsize ; j++ ) {         \
 346       vb[j] = ((GLuint *)v)[j];                 \
 347    }                                            \
 348    vb += vertsize;                              \
 349 } while (0)
 350 #endif
 351
 352 static void
 353 intel_draw_quad(struct intel_context *intel,
 354                 intelVertexPtr v0,
 355                 intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3)
 356 {
 357    GLuint vertsize = intel->vertex_size;
 358    GLuint *vb = intel_get_prim_space(intel, 6);
 359    int j;
 360
 361    COPY_DWORDS(j, vb, vertsize, v0);
 362    COPY_DWORDS(j, vb, vertsize, v1);
 363
 364    /* If smooth shading, draw like a trifan which gives better
 365     * rasterization.  Otherwise draw as two triangles with provoking
 366     * vertex in third position as required for flat shading.
 367     */
 368    if (intel->ctx.Light.ShadeModel == GL_FLAT) {
 369       COPY_DWORDS(j, vb, vertsize, v3);
 370       COPY_DWORDS(j, vb, vertsize, v1);
 371    }
 372    else {
 373       COPY_DWORDS(j, vb, vertsize, v2);
 374       COPY_DWORDS(j, vb, vertsize, v0);
 375    }
 376
 377    COPY_DWORDS(j, vb, vertsize, v2);
 378    COPY_DWORDS(j, vb, vertsize, v3);
 379 }
 380
 381 static void
 382 intel_draw_triangle(struct intel_context *intel,
 383                     intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
 384 {
 385    GLuint vertsize = intel->vertex_size;
 386    GLuint *vb = intel_get_prim_space(intel, 3);
 387    int j;
 388
 389    COPY_DWORDS(j, vb, vertsize, v0);
 390    COPY_DWORDS(j, vb, vertsize, v1);
 391    COPY_DWORDS(j, vb, vertsize, v2);
 392 }
 393
 394
 395 static void
 396 intel_draw_line(struct intel_context *intel,
 397                 intelVertexPtr v0, intelVertexPtr v1)
 398 {
 399    GLuint vertsize = intel->vertex_size;
 400    GLuint *vb = intel_get_prim_space(intel, 2);
 401    int j;
 402
 403    COPY_DWORDS(j, vb, vertsize, v0);
 404    COPY_DWORDS(j, vb, vertsize, v1);
 405 }
 406
 407
 408 static void
 409 intel_draw_point(struct intel_context *intel, intelVertexPtr v0)
 410 {
 411    GLuint vertsize = intel->vertex_size;
 412    GLuint *vb = intel_get_prim_space(intel, 1);
 413    int j;
 414
 415    /* Adjust for sub pixel position -- still required for conform. */
 416    *(float *) &vb[0] = v0->v.x;
 417    *(float *) &vb[1] = v0->v.y;
 418    for (j = 2; j < vertsize; j++)
 419       vb[j] = v0->ui[j];
 420 }
 421
 422
 423
 424 /***********************************************************************
 425  *                Fixup for ARB_point_parameters                       *
 426  ***********************************************************************/
 427
 428 /* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly
 429  * represented in the fragment program InputsRead field.
 430  */
 431 static void
 432 intel_atten_point(struct intel_context *intel, intelVertexPtr v0)
 433 {
 434    GLcontext *ctx = &intel->ctx;
 435    GLfloat psz[4], col[4], restore_psz, restore_alpha;
 436
 437    _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 438    _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 439
 440    restore_psz = psz[0];
 441    restore_alpha = col[3];
 442
 443    if (psz[0] >= ctx->Point.Threshold) {
 444       psz[0] = MIN2(psz[0], ctx->Point.MaxSize);
 445    }
 446    else {
 447       GLfloat dsize = psz[0] / ctx->Point.Threshold;
 448       psz[0] = MAX2(ctx->Point.Threshold, ctx->Point.MinSize);
 449       col[3] *= dsize * dsize;
 450    }
 451
 452    if (psz[0] < 1.0)
 453       psz[0] = 1.0;
 454
 455    if (restore_psz != psz[0] || restore_alpha != col[3]) {
 456       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 457       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 458
 459       intel_draw_point(intel, v0);
 460
 461       psz[0] = restore_psz;
 462       col[3] = restore_alpha;
 463
 464       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz);
 465       _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col);
 466    }
 467    else
 468       intel_draw_point(intel, v0);
 469 }
 470
 471
 472
 473
 474
 475 /***********************************************************************
 476  *                Fixup for I915 WPOS texture coordinate                *
 477  ***********************************************************************/
 478
 479
 480
 481 static void
 482 intel_wpos_triangle(struct intel_context *intel,
 483                     intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2)
 484 {
 485    GLuint offset = intel->wpos_offset;
 486    GLuint size = intel->wpos_size;
 487    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 488    GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
 489    GLfloat *v2_wpos = (GLfloat *)((char *)v2 + offset);
 490
 491    __memcpy(v0_wpos, v0, size);
 492    __memcpy(v1_wpos, v1, size);
 493    __memcpy(v2_wpos, v2, size);
 494
 495    v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
 496    v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
 497    v2_wpos[1] = -v2_wpos[1] + intel->driDrawable->h;
 498
 499
 500    intel_draw_triangle(intel, v0, v1, v2);
 501 }
 502
 503
 504 static void
 505 intel_wpos_line(struct intel_context *intel,
 506                 intelVertexPtr v0, intelVertexPtr v1)
 507 {
 508    GLuint offset = intel->wpos_offset;
 509    GLuint size = intel->wpos_size;
 510    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 511    GLfloat *v1_wpos = (GLfloat *)((char *)v1 + offset);
 512
 513    __memcpy(v0_wpos, v0, size);
 514    __memcpy(v1_wpos, v1, size);
 515
 516    v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
 517    v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
 518
 519    intel_draw_line(intel, v0, v1);
 520 }
 521
 522
 523 static void
 524 intel_wpos_point(struct intel_context *intel, intelVertexPtr v0)
 525 {
 526    GLuint offset = intel->wpos_offset;
 527    GLuint size = intel->wpos_size;
 528    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 529
 530    __memcpy(v0_wpos, v0, size);
 531    v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
 532
 533    intel_draw_point(intel, v0);
 534 }
 535
 536
 537
 538
 539
 540
 541 /***********************************************************************
 542  *          Macros for t_dd_tritmp.h to draw basic primitives          *
 543  ***********************************************************************/
 544
 545 #define TRI( a, b, c )                          \
 546 do {                                            \
 547    if (DO_FALLBACK)                             \
 548       intel->draw_tri( intel, a, b, c );        \
 549    else                                         \
 550       intel_draw_triangle( intel, a, b, c );    \
 551 } while (0)
 552
 553 #define QUAD( a, b, c, d )                      \
 554 do {                                            \
 555    if (DO_FALLBACK) {                           \
 556       intel->draw_tri( intel, a, b, d );        \
 557       intel->draw_tri( intel, b, c, d );        \
 558    } else                                       \
 559       intel_draw_quad( intel, a, b, c, d );     \
 560 } while (0)
 561
 562 #define LINE( v0, v1 )                          \
 563 do {                                            \
 564    if (DO_FALLBACK)                             \
 565       intel->draw_line( intel, v0, v1 );        \
 566    else                                         \
 567       intel_draw_line( intel, v0, v1 );         \
 568 } while (0)
 569
 570 #define POINT( v0 )                             \
 571 do {                                            \
 572    if (DO_FALLBACK)                             \
 573       intel->draw_point( intel, v0 );           \
 574    else                                         \
 575       intel_draw_point( intel, v0 );            \
 576 } while (0)
 577
 578
 579 /***********************************************************************
 580  *              Build render functions from dd templates               *
 581  ***********************************************************************/
 582
 583 #define INTEL_OFFSET_BIT        0x01
 584 #define INTEL_TWOSIDE_BIT       0x02
 585 #define INTEL_UNFILLED_BIT      0x04
 586 #define INTEL_FALLBACK_BIT      0x08
 587 #define INTEL_MAX_TRIFUNC       0x10
 588
 589
 590 static struct
 591 {
 592    tnl_points_func points;
 593    tnl_line_func line;
 594    tnl_triangle_func triangle;
 595    tnl_quad_func quad;
 596 } rast_tab[INTEL_MAX_TRIFUNC];
 597
 598
 599 #define DO_FALLBACK (IND & INTEL_FALLBACK_BIT)
 600 #define DO_OFFSET   (IND & INTEL_OFFSET_BIT)
 601 #define DO_UNFILLED (IND & INTEL_UNFILLED_BIT)
 602 #define DO_TWOSIDE  (IND & INTEL_TWOSIDE_BIT)
 603 #define DO_FLAT      0
 604 #define DO_TRI       1
 605 #define DO_QUAD      1
 606 #define DO_LINE      1
 607 #define DO_POINTS    1
 608 #define DO_FULL_QUAD 1
 609
 610 #define HAVE_RGBA         1
 611 #define HAVE_SPEC         1
 612 #define HAVE_BACK_COLORS  0
 613 #define HAVE_HW_FLATSHADE 1
 614 #define VERTEX            intelVertex
 615 #define TAB               rast_tab
 616
 617 /* Only used to pull back colors into vertices (ie, we know color is
 618  * floating point).
 619  */
 620 #define INTEL_COLOR( dst, src )                         \
 621 do {                                                    \
 622    UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);        \
 623    UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);        \
 624    UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);        \
 625    UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]);        \
 626 } while (0)
 627
 628 #define INTEL_SPEC( dst, src )                          \
 629 do {                                                    \
 630    UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]);        \
 631    UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]);        \
 632    UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]);        \
 633 } while (0)
 634
 635
 636 #define DEPTH_SCALE intel->polygon_offset_scale
 637 #define UNFILLED_TRI unfilled_tri
 638 #define UNFILLED_QUAD unfilled_quad
 639 #define VERT_X(_v) _v->v.x
 640 #define VERT_Y(_v) _v->v.y
 641 #define VERT_Z(_v) _v->v.z
 642 #define AREA_IS_CCW( a ) (a > 0)
 643 #define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint)))
 644
 645 #define VERT_SET_RGBA( v, c )    if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c )
 646 #define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset]
 647 #define VERT_SAVE_RGBA( idx )    if (coloroffset) color[idx] = v[idx]->ui[coloroffset]
 648 #define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx]
 649
 650 #define VERT_SET_SPEC( v, c )    if (specoffset) INTEL_SPEC( v->ub4[specoffset], c )
 651 #define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset])
 652 #define VERT_SAVE_SPEC( idx )    if (specoffset) spec[idx] = v[idx]->ui[specoffset]
 653 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
 654
 655 #define LOCAL_VARS(n)                                                   \
 656    struct intel_context *intel = intel_context(ctx);                    \
 657    GLuint color[n] = { 0, }, spec[n] = { 0, };                          \
 658    GLuint coloroffset = intel->coloroffset;                             \
 659    GLboolean specoffset = intel->specoffset;                            \
 660    (void) color; (void) spec; (void) coloroffset; (void) specoffset;
 661
 662
 663 /***********************************************************************
 664  *                Helpers for rendering unfilled primitives            *
 665  ***********************************************************************/
 666
 667 static const GLuint hw_prim[GL_POLYGON + 1] = {
 668    PRIM3D_POINTLIST,
 669    PRIM3D_LINELIST,
 670    PRIM3D_LINELIST,
 671    PRIM3D_LINELIST,
 672    PRIM3D_TRILIST,
 673    PRIM3D_TRILIST,
 674    PRIM3D_TRILIST,
 675    PRIM3D_TRILIST,
 676    PRIM3D_TRILIST,
 677    PRIM3D_TRILIST
 678 };
 679
 680 #define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] )
 681 #define RENDER_PRIMITIVE intel->render_primitive
 682 #define TAG(x) x
 683 #define IND INTEL_FALLBACK_BIT
 684 #include "tnl_dd/t_dd_unfilled.h"
 685 #undef IND
 686
 687 /***********************************************************************
 688  *                      Generate GL render functions                   *
 689  ***********************************************************************/
 690
 691 #define IND (0)
 692 #define TAG(x) x
 693 #include "tnl_dd/t_dd_tritmp.h"
 694
 695 #define IND (INTEL_OFFSET_BIT)
 696 #define TAG(x) x##_offset
 697 #include "tnl_dd/t_dd_tritmp.h"
 698
 699 #define IND (INTEL_TWOSIDE_BIT)
 700 #define TAG(x) x##_twoside
 701 #include "tnl_dd/t_dd_tritmp.h"
 702
 703 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT)
 704 #define TAG(x) x##_twoside_offset
 705 #include "tnl_dd/t_dd_tritmp.h"
 706
 707 #define IND (INTEL_UNFILLED_BIT)
 708 #define TAG(x) x##_unfilled
 709 #include "tnl_dd/t_dd_tritmp.h"
 710
 711 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
 712 #define TAG(x) x##_offset_unfilled
 713 #include "tnl_dd/t_dd_tritmp.h"
 714
 715 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT)
 716 #define TAG(x) x##_twoside_unfilled
 717 #include "tnl_dd/t_dd_tritmp.h"
 718
 719 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT)
 720 #define TAG(x) x##_twoside_offset_unfilled
 721 #include "tnl_dd/t_dd_tritmp.h"
 722
 723 #define IND (INTEL_FALLBACK_BIT)
 724 #define TAG(x) x##_fallback
 725 #include "tnl_dd/t_dd_tritmp.h"
 726
 727 #define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
 728 #define TAG(x) x##_offset_fallback
 729 #include "tnl_dd/t_dd_tritmp.h"
 730
 731 #define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT)
 732 #define TAG(x) x##_twoside_fallback
 733 #include "tnl_dd/t_dd_tritmp.h"
 734
 735 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT)
 736 #define TAG(x) x##_twoside_offset_fallback
 737 #include "tnl_dd/t_dd_tritmp.h"
 738
 739 #define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 740 #define TAG(x) x##_unfilled_fallback
 741 #include "tnl_dd/t_dd_tritmp.h"
 742
 743 #define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 744 #define TAG(x) x##_offset_unfilled_fallback
 745 #include "tnl_dd/t_dd_tritmp.h"
 746
 747 #define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT)
 748 #define TAG(x) x##_twoside_unfilled_fallback
 749 #include "tnl_dd/t_dd_tritmp.h"
 750
 751 #define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \
 752              INTEL_FALLBACK_BIT)
 753 #define TAG(x) x##_twoside_offset_unfilled_fallback
 754 #include "tnl_dd/t_dd_tritmp.h"
 755
 756
 757 static void
 758 init_rast_tab(void)
 759 {
 760    init();
 761    init_offset();
 762    init_twoside();
 763    init_twoside_offset();
 764    init_unfilled();
 765    init_offset_unfilled();
 766    init_twoside_unfilled();
 767    init_twoside_offset_unfilled();
 768    init_fallback();
 769    init_offset_fallback();
 770    init_twoside_fallback();
 771    init_twoside_offset_fallback();
 772    init_unfilled_fallback();
 773    init_offset_unfilled_fallback();
 774    init_twoside_unfilled_fallback();
 775    init_twoside_offset_unfilled_fallback();
 776 }
 777
 778
 779 /***********************************************************************
 780  *                    Rasterization fallback helpers                   *
 781  ***********************************************************************/
 782
 783
 784 /* This code is hit only when a mix of accelerated and unaccelerated
 785  * primitives are being drawn, and only for the unaccelerated
 786  * primitives.
 787  */
 788 static void
 789 intel_fallback_tri(struct intel_context *intel,
 790                    intelVertex * v0, intelVertex * v1, intelVertex * v2)
 791 {
 792    GLcontext *ctx = &intel->ctx;
 793    SWvertex v[3];
 794
 795    if (0)
 796       fprintf(stderr, "\n%s\n", __FUNCTION__);
 797
 798    INTEL_FIREVERTICES(intel);
 799
 800    _swsetup_Translate(ctx, v0, &v[0]);
 801    _swsetup_Translate(ctx, v1, &v[1]);
 802    _swsetup_Translate(ctx, v2, &v[2]);
 803    intelSpanRenderStart(ctx);
 804    _swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
 805    intelSpanRenderFinish(ctx);
 806 }
 807
 808
 809 static void
 810 intel_fallback_line(struct intel_context *intel,
 811                     intelVertex * v0, intelVertex * v1)
 812 {
 813    GLcontext *ctx = &intel->ctx;
 814    SWvertex v[2];
 815
 816    if (0)
 817       fprintf(stderr, "\n%s\n", __FUNCTION__);
 818
 819    INTEL_FIREVERTICES(intel);
 820
 821    _swsetup_Translate(ctx, v0, &v[0]);
 822    _swsetup_Translate(ctx, v1, &v[1]);
 823    intelSpanRenderStart(ctx);
 824    _swrast_Line(ctx, &v[0], &v[1]);
 825    intelSpanRenderFinish(ctx);
 826 }
 827
 828 static void
 829 intel_fallback_point(struct intel_context *intel,
 830                      intelVertex * v0)
 831 {
 832    GLcontext *ctx = &intel->ctx;
 833    SWvertex v[1];
 834
 835    if (0)
 836       fprintf(stderr, "\n%s\n", __FUNCTION__);
 837
 838    INTEL_FIREVERTICES(intel);
 839
 840    _swsetup_Translate(ctx, v0, &v[0]);
 841    intelSpanRenderStart(ctx);
 842    _swrast_Point(ctx, &v[0]);
 843    intelSpanRenderFinish(ctx);
 844 }
 845
 846
 847 /**********************************************************************/
 848 /*               Render unclipped begin/end objects                   */
 849 /**********************************************************************/
 850
 851 #define IND 0
 852 #define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint)))
 853 #define RENDER_POINTS( start, count )   \
 854    for ( ; start < count ; start++) POINT( V(ELT(start)) );
 855 #define RENDER_LINE( v0, v1 )         LINE( V(v0), V(v1) )
 856 #define RENDER_TRI(  v0, v1, v2 )     TRI(  V(v0), V(v1), V(v2) )
 857 #define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) )
 858 #define INIT(x) intelRenderPrimitive( ctx, x )
 859 #undef LOCAL_VARS
 860 #define LOCAL_VARS                                              \
 861     struct intel_context *intel = intel_context(ctx);                   \
 862     GLubyte *vertptr = (GLubyte *)intel->verts;                 \
 863     const GLuint vertsize = intel->vertex_size;         \
 864     const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;       \
 865     (void) elt;
 866 #define RESET_STIPPLE
 867 #define RESET_OCCLUSION
 868 #define PRESERVE_VB_DEFS
 869 #define ELT(x) x
 870 #define TAG(x) intel_##x##_verts
 871 #include "tnl/t_vb_rendertmp.h"
 872 #undef ELT
 873 #undef TAG
 874 #define TAG(x) intel_##x##_elts
 875 #define ELT(x) elt[x]
 876 #include "tnl/t_vb_rendertmp.h"
 877
 878 /**********************************************************************/
 879 /*                   Render clipped primitives                        */
 880 /**********************************************************************/
 881
 882
 883
 884 static void
 885 intelRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
 886 {
 887    struct intel_context *intel = intel_context(ctx);
 888    TNLcontext *tnl = TNL_CONTEXT(ctx);
 889    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
 890    GLuint prim = intel->render_primitive;
 891
 892    /* Render the new vertices as an unclipped polygon.
 893     */
 894    {
 895       GLuint *tmp = VB->Elts;
 896       VB->Elts = (GLuint *) elts;
 897       tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n,
 898                                                   PRIM_BEGIN | PRIM_END);
 899       VB->Elts = tmp;
 900    }
 901
 902    /* Restore the render primitive
 903     */
 904    if (prim != GL_POLYGON)
 905       tnl->Driver.Render.PrimitiveNotify(ctx, prim);
 906 }
 907
 908 static void
 909 intelRenderClippedLine(GLcontext * ctx, GLuint ii, GLuint jj)
 910 {
 911    TNLcontext *tnl = TNL_CONTEXT(ctx);
 912
 913    tnl->Driver.Render.Line(ctx, ii, jj);
 914 }
 915
 916 static void
 917 intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n)
 918 {
 919    struct intel_context *intel = intel_context(ctx);
 920    const GLuint vertsize = intel->vertex_size;
 921    GLuint *vb = intel_get_prim_space(intel, (n - 2) * 3);
 922    GLubyte *vertptr = (GLubyte *) intel->verts;
 923    const GLuint *start = (const GLuint *) V(elts[0]);
 924    int i, j;
 925
 926    for (i = 2; i < n; i++) {
 927       COPY_DWORDS(j, vb, vertsize, V(elts[i - 1]));
 928       COPY_DWORDS(j, vb, vertsize, V(elts[i]));
 929       COPY_DWORDS(j, vb, vertsize, start);
 930    }
 931 }
 932
 933 /**********************************************************************/
 934 /*                    Choose render functions                         */
 935 /**********************************************************************/
 936
 937
 938
 939
 940 #define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN | DD_POINT_SMOOTH | DD_TRI_SMOOTH)
 941 #define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED)
 942
 943 void
 944 intelChooseRenderState(GLcontext * ctx)
 945 {
 946    TNLcontext *tnl = TNL_CONTEXT(ctx);
 947    struct intel_context *intel = intel_context(ctx);
 948    GLuint flags = ctx->_TriangleCaps;
 949    const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current;
 950    GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS));
 951    GLuint index = 0;
 952
 953    if (INTEL_DEBUG & DEBUG_STATE)
 954       fprintf(stderr, "\n%s\n", __FUNCTION__);
 955
 956    if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) {
 957
 958       if (flags & ANY_RASTER_FLAGS) {
 959          if (flags & DD_TRI_LIGHT_TWOSIDE)
 960             index |= INTEL_TWOSIDE_BIT;
 961          if (flags & DD_TRI_OFFSET)
 962             index |= INTEL_OFFSET_BIT;
 963          if (flags & DD_TRI_UNFILLED)
 964             index |= INTEL_UNFILLED_BIT;
 965       }
 966
 967       if (have_wpos) {
 968          intel->draw_point = intel_wpos_point;
 969          intel->draw_line = intel_wpos_line;
 970          intel->draw_tri = intel_wpos_triangle;
 971
 972          /* Make sure these get called:
 973           */
 974          index |= INTEL_FALLBACK_BIT;
 975       }
 976       else {
 977          intel->draw_point = intel_draw_point;
 978          intel->draw_line = intel_draw_line;
 979          intel->draw_tri = intel_draw_triangle;
 980       }
 981
 982       /* Hook in fallbacks for specific primitives.
 983        */
 984       if (flags & ANY_FALLBACK_FLAGS) {
 985          if (flags & DD_LINE_STIPPLE)
 986             intel->draw_line = intel_fallback_line;
 987
 988          if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple)
 989             intel->draw_tri = intel_fallback_tri;
 990
 991          if (flags & DD_TRI_SMOOTH) {
 992             if (intel->conformance_mode > 0)
 993                intel->draw_tri = intel_fallback_tri;
 994          }
 995
 996          if (flags & DD_POINT_ATTEN) {
 997             if (0)
 998                intel->draw_point = intel_atten_point;
 999             else
1000                intel->draw_point = intel_fallback_point;
1001          }
1002
1003          if (flags & DD_POINT_SMOOTH) {
1004             if (intel->conformance_mode > 0)
1005                intel->draw_point = intel_fallback_point;
1006          }
1007
1008          index |= INTEL_FALLBACK_BIT;
1009       }
1010    }
1011
1012    if (intel->RenderIndex != index) {
1013       intel->RenderIndex = index;
1014
1015       tnl->Driver.Render.Points = rast_tab[index].points;
1016       tnl->Driver.Render.Line = rast_tab[index].line;
1017       tnl->Driver.Render.Triangle = rast_tab[index].triangle;
1018       tnl->Driver.Render.Quad = rast_tab[index].quad;
1019
1020       if (index == 0) {
1021          tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts;
1022          tnl->Driver.Render.PrimTabElts = intel_render_tab_elts;
1023          tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */
1024          tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly;
1025       }
1026       else {
1027          tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
1028          tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
1029          tnl->Driver.Render.ClippedLine = intelRenderClippedLine;
1030          tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly;
1031       }
1032    }
1033 }
1034
1035 static const GLenum reduced_prim[GL_POLYGON + 1] = {
1036    GL_POINTS,
1037    GL_LINES,
1038    GL_LINES,
1039    GL_LINES,
1040    GL_TRIANGLES,
1041    GL_TRIANGLES,
1042    GL_TRIANGLES,
1043    GL_TRIANGLES,
1044    GL_TRIANGLES,
1045    GL_TRIANGLES
1046 };
1047
1048
1049 /**********************************************************************/
1050 /*                 High level hooks for t_vb_render.c                 */
1051 /**********************************************************************/
1052
1053
1054
1055
1056 static void
1057 intelRunPipeline(GLcontext * ctx)
1058 {
1059    struct intel_context *intel = intel_context(ctx);
1060
1061    _mesa_lock_context_textures(ctx);
1062
1063    if (ctx->NewState)
1064       _mesa_update_state_locked(ctx);
1065
1066    if (intel->NewGLState) {
1067       if (intel->NewGLState & _NEW_TEXTURE) {
1068          intel->vtbl.update_texture_state(intel);
1069       }
1070
1071       if (!intel->Fallback) {
1072          if (intel->NewGLState & _INTEL_NEW_RENDERSTATE)
1073             intelChooseRenderState(ctx);
1074       }
1075
1076       intel->NewGLState = 0;
1077    }
1078
1079    _tnl_run_pipeline(ctx);
1080
1081    _mesa_unlock_context_textures(ctx);
1082 }
1083
1084 static void
1085 intelRenderStart(GLcontext * ctx)
1086 {
1087    struct intel_context *intel = intel_context(ctx);
1088
1089    intel->vtbl.render_start(intel_context(ctx));
1090    intel->vtbl.emit_state(intel);
1091 }
1092
1093 static void
1094 intelRenderFinish(GLcontext * ctx)
1095 {
1096    struct intel_context *intel = intel_context(ctx);
1097
1098    if (intel->RenderIndex & INTEL_FALLBACK_BIT)
1099       _swrast_flush(ctx);
1100
1101    INTEL_FIREVERTICES(intel);
1102 }
1103
1104
1105
1106
1107  /* System to flush dma and emit state changes based on the rasterized
1108   * primitive.
1109   */
1110 static void
1111 intelRasterPrimitive(GLcontext * ctx, GLenum rprim, GLuint hwprim)
1112 {
1113    struct intel_context *intel = intel_context(ctx);
1114
1115    if (0)
1116       fprintf(stderr, "%s %s %x\n", __FUNCTION__,
1117               _mesa_lookup_enum_by_nr(rprim), hwprim);
1118
1119    intel->vtbl.reduced_primitive_state(intel, rprim);
1120
1121    /* Start a new primitive.  Arrange to have it flushed later on.
1122     */
1123    if (hwprim != intel->prim.primitive) {
1124       INTEL_FIREVERTICES(intel);
1125
1126       intel_set_prim(intel, hwprim);
1127    }
1128 }
1129
1130
1131  /*
1132   */
1133 static void
1134 intelRenderPrimitive(GLcontext * ctx, GLenum prim)
1135 {
1136    struct intel_context *intel = intel_context(ctx);
1137
1138    if (0)
1139       fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
1140
1141    /* Let some clipping routines know which primitive they're dealing
1142     * with.
1143     */
1144    intel->render_primitive = prim;
1145
1146    /* Shortcircuit this when called from t_dd_rendertmp.h for unfilled
1147     * triangles.  The rasterized primitive will always be reset by
1148     * lower level functions in that case, potentially pingponging the
1149     * state:
1150     */
1151    if (reduced_prim[prim] == GL_TRIANGLES &&
1152        (ctx->_TriangleCaps & DD_TRI_UNFILLED))
1153       return;
1154
1155    /* Set some primitive-dependent state and Start? a new primitive.
1156     */
1157    intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]);
1158 }
1159
1160
1161  /**********************************************************************/
1162  /*           Transition to/from hardware rasterization.               */
1163  /**********************************************************************/
1164
1165 static char *fallbackStrings[] = {
1166    [0] = "Draw buffer",
1167    [1] = "Read buffer",
1168    [2] = "Depth buffer",
1169    [3] = "Stencil buffer",
1170    [4] = "User disable",
1171    [5] = "Render mode",
1172
1173    [12] = "Texture",
1174    [13] = "Color mask",
1175    [14] = "Stencil",
1176    [15] = "Stipple",
1177    [16] = "Program",
1178    [17] = "Logic op",
1179    [18] = "Smooth polygon",
1180    [19] = "Smooth point",
1181 };
1182
1183
1184 static char *
1185 getFallbackString(GLuint bit)
1186 {
1187    int i = 0;
1188    while (bit > 1) {
1189       i++;
1190       bit >>= 1;
1191    }
1192    return fallbackStrings[i];
1193 }
1194
1195
1196
1197 void
1198 intelFallback(struct intel_context *intel, GLuint bit, GLboolean mode)
1199 {
1200    GLcontext *ctx = &intel->ctx;
1201    TNLcontext *tnl = TNL_CONTEXT(ctx);
1202    GLuint oldfallback = intel->Fallback;
1203
1204    if (mode) {
1205       intel->Fallback |= bit;
1206       if (oldfallback == 0) {
1207          intelFlush(ctx);
1208          if (INTEL_DEBUG & DEBUG_FALLBACKS)
1209             fprintf(stderr, "ENTER FALLBACK %x: %s\n",
1210                     bit, getFallbackString(bit));
1211          _swsetup_Wakeup(ctx);
1212          intel->RenderIndex = ~0;
1213       }
1214    }
1215    else {
1216       intel->Fallback &= ~bit;
1217       if (oldfallback == bit) {
1218          _swrast_flush(ctx);
1219          if (INTEL_DEBUG & DEBUG_FALLBACKS)
1220             fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit));
1221          tnl->Driver.Render.Start = intelRenderStart;
1222          tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1223          tnl->Driver.Render.Finish = intelRenderFinish;
1224          tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1225          tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1226          tnl->Driver.Render.Interp = _tnl_interp;
1227
1228          _tnl_invalidate_vertex_state(ctx, ~0);
1229          _tnl_invalidate_vertices(ctx, ~0);
1230          _tnl_install_attrs(ctx,
1231                             intel->vertex_attrs,
1232                             intel->vertex_attr_count,
1233                             intel->ViewportMatrix.m, 0);
1234
1235          intel->NewGLState |= _INTEL_NEW_RENDERSTATE;
1236       }
1237    }
1238 }
1239
1240 union fi
1241 {
1242    GLfloat f;
1243    GLint i;
1244 };
1245
1246
1247 /**********************************************************************/
1248 /*             Used only with the metaops callbacks.                  */
1249 /**********************************************************************/
1250 static void
1251 intel_meta_draw_poly(struct intel_context *intel,
1252                      GLuint n,
1253                      GLfloat xy[][2],
1254                      GLfloat z, GLuint color, GLfloat tex[][2])
1255 {
1256    union fi *vb;
1257    GLint i;
1258    unsigned int saved_vertex_size = intel->vertex_size;
1259
1260    LOCK_HARDWARE(intel);
1261
1262    intel->vertex_size = 6;
1263
1264    /* All 3d primitives should be emitted with LOOP_CLIPRECTS,
1265     * otherwise the drawing origin (DR4) might not be set correctly.
1266     */
1267    intel_set_prim(intel, PRIM3D_TRIFAN);
1268    vb = (union fi *) intel_get_prim_space(intel, n);
1269
1270    for (i = 0; i < n; i++) {
1271       vb[0].f = xy[i][0];
1272       vb[1].f = xy[i][1];
1273       vb[2].f = z;
1274       vb[3].i = color;
1275       vb[4].f = tex[i][0];
1276       vb[5].f = tex[i][1];
1277       vb += 6;
1278    }
1279
1280    INTEL_FIREVERTICES(intel);
1281
1282    intel->vertex_size = saved_vertex_size;
1283
1284    UNLOCK_HARDWARE(intel);
1285 }
1286
1287 static void
1288 intel_meta_draw_quad(struct intel_context *intel,
1289                      GLfloat x0, GLfloat x1,
1290                      GLfloat y0, GLfloat y1,
1291                      GLfloat z,
1292                      GLuint color,
1293                      GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1)
1294 {
1295    GLfloat xy[4][2];
1296    GLfloat tex[4][2];
1297
1298    xy[0][0] = x0;
1299    xy[0][1] = y0;
1300    xy[1][0] = x1;
1301    xy[1][1] = y0;
1302    xy[2][0] = x1;
1303    xy[2][1] = y1;
1304    xy[3][0] = x0;
1305    xy[3][1] = y1;
1306
1307    tex[0][0] = s0;
1308    tex[0][1] = t0;
1309    tex[1][0] = s1;
1310    tex[1][1] = t0;
1311    tex[2][0] = s1;
1312    tex[2][1] = t1;
1313    tex[3][0] = s0;
1314    tex[3][1] = t1;
1315
1316    intel_meta_draw_poly(intel, 4, xy, z, color, tex);
1317 }
1318
1319
1320
1321 /**********************************************************************/
1322 /*                            Initialization.                         */
1323 /**********************************************************************/
1324
1325
1326 void
1327 intelInitTriFuncs(GLcontext * ctx)
1328 {
1329    struct intel_context *intel = intel_context(ctx);
1330    TNLcontext *tnl = TNL_CONTEXT(ctx);
1331    static int firsttime = 1;
1332
1333    if (firsttime) {
1334       init_rast_tab();
1335       firsttime = 0;
1336    }
1337
1338    tnl->Driver.RunPipeline = intelRunPipeline;
1339    tnl->Driver.Render.Start = intelRenderStart;
1340    tnl->Driver.Render.Finish = intelRenderFinish;
1341    tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive;
1342    tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple;
1343    tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
1344    tnl->Driver.Render.CopyPV = _tnl_copy_pv;
1345    tnl->Driver.Render.Interp = _tnl_interp;
1346
1347    intel->vtbl.meta_draw_quad = intel_meta_draw_quad;
1348 }