src/mesa/tnl/t_vertex.c

   1 /*
   2  * Copyright 2003 Tungsten Graphics, inc.
   3  * All Rights Reserved.
   4  *
   5  * Permission is hereby granted, free of charge, to any person obtaining a
   6  * copy of this software and associated documentation files (the "Software"),
   7  * to deal in the Software without restriction, including without limitation
   8  * on the rights to use, copy, modify, merge, publish, distribute, sub
   9  * license, and/or sell copies of the Software, and to permit persons to whom
  10  * the Software is furnished to do so, subject to the following conditions:
  11  *
  12  * The above copyright notice and this permission notice (including the next
  13  * paragraph) shall be included in all copies or substantial portions of the
  14  * Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
  19  * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
  20  * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
  21  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  22  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  23  *
  24  * Authors:
  25  *    Keith Whitwell <keithw@tungstengraphics.com>
  26  */
  27
  28 #include "glheader.h"
  29 #include "context.h"
  30 #include "colormac.h"
  31
  32 #include "t_context.h"
  33 #include "t_vertex.h"
  34
  35 #define DBG 0
  36
  37 /* Build and manage clipspace/ndc/window vertices.
  38  */
  39
  40 static GLboolean match_fastpath( struct tnl_clipspace *vtx,
  41                                  const struct tnl_clipspace_fastpath *fp)
  42 {
  43    GLuint j;
  44
  45    if (vtx->attr_count != fp->attr_count)
  46       return GL_FALSE;
  47
  48    for (j = 0; j < vtx->attr_count; j++)
  49       if (vtx->attr[j].format != fp->attr[j].format)
  50          return GL_FALSE;
  51
  52    if (fp->match_strides) {
  53       if (vtx->vertex_size != fp->vertex_size)
  54          return GL_FALSE;
  55
  56       for (j = 0; j < vtx->attr_count; j++)
  57          if (vtx->attr[j].inputstride != fp->attr[j].stride)
  58             return GL_FALSE;
  59    }
  60
  61    return GL_TRUE;
  62 }
  63
  64 static GLboolean search_fastpath_emit( struct tnl_clipspace *vtx )
  65 {
  66    struct tnl_clipspace_fastpath *fp = vtx->fastpath;
  67
  68    for ( ; fp ; fp = fp->next) {
  69       if (match_fastpath(vtx, fp)) {
  70          vtx->emit = fp->func;
  71          return GL_TRUE;
  72       }
  73    }
  74
  75    return GL_FALSE;
  76 }
  77
  78 void _tnl_register_fastpath( struct tnl_clipspace *vtx,
  79                              GLboolean match_strides )
  80 {
  81    struct tnl_clipspace_fastpath *fastpath = CALLOC_STRUCT(tnl_clipspace_fastpath);
  82    GLuint i;
  83
  84    fastpath->vertex_size = vtx->vertex_size;
  85    fastpath->attr_count = vtx->attr_count;
  86    fastpath->match_strides = match_strides;
  87    fastpath->func = vtx->emit;
  88    fastpath->attr = MALLOC(vtx->attr_count * sizeof(fastpath->attr[0]));
  89
  90    for (i = 0; i < vtx->attr_count; i++) {
  91       fastpath->attr[i].format = vtx->attr[i].format;
  92       fastpath->attr[i].stride = vtx->attr[i].inputstride;
  93    }
  94
  95    fastpath->next = vtx->fastpath;
  96    vtx->fastpath = fastpath;
  97 }
  98
  99
 100
 101 /***********************************************************************
 102  * Build codegen functions or return generic ones:
 103  */
 104 static void choose_emit_func( GLcontext *ctx, GLuint count, GLubyte *dest)
 105 {
 106    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
 107    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 108    struct tnl_clipspace_attr *a = vtx->attr;
 109    const GLuint attr_count = vtx->attr_count;
 110    GLuint j;
 111
 112    for (j = 0; j < attr_count; j++) {
 113       GLvector4f *vptr = VB->AttribPtr[a[j].attrib];
 114       a[j].inputstride = vptr->stride;
 115       a[j].inputsize = vptr->size;
 116       a[j].emit = a[j].insert[vptr->size - 1]; /* not always used */
 117    }
 118
 119    vtx->emit = NULL;
 120
 121    /* Does this match an existing (hardwired, codegen or known-bad)
 122     * fastpath?
 123     */
 124    if (search_fastpath_emit(vtx)) {
 125       /* Use this result.  If it is null, then it is already known
 126        * that the current state will fail for codegen and there is no
 127        * point trying again.
 128        */
 129    }
 130    else if (vtx->codegen_emit) {
 131       vtx->codegen_emit(ctx);
 132    }
 133
 134    if (!vtx->emit) {
 135       _tnl_generate_hardwired_emit(ctx);
 136    }
 137
 138    /* Otherwise use the generic version:
 139     */
 140    if (!vtx->emit)
 141       vtx->emit = _tnl_generic_emit;
 142
 143    vtx->emit( ctx, count, dest );
 144 }
 145
 146
 147
 148 static void choose_interp_func( GLcontext *ctx,
 149                                 GLfloat t,
 150                                 GLuint edst, GLuint eout, GLuint ein,
 151                                 GLboolean force_boundary )
 152 {
 153    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 154
 155    if (vtx->need_extras &&
 156        (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
 157       vtx->interp = _tnl_generic_interp_extras;
 158    } else {
 159       vtx->interp = _tnl_generic_interp;
 160    }
 161
 162    vtx->interp( ctx, t, edst, eout, ein, force_boundary );
 163 }
 164
 165
 166 static void choose_copy_pv_func(  GLcontext *ctx, GLuint edst, GLuint esrc )
 167 {
 168    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 169
 170    if (vtx->need_extras &&
 171        (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
 172       vtx->copy_pv = _tnl_generic_copy_pv_extras;
 173    } else {
 174       vtx->copy_pv = _tnl_generic_copy_pv;
 175    }
 176
 177    vtx->copy_pv( ctx, edst, esrc );
 178 }
 179
 180
 181 /***********************************************************************
 182  * Public entrypoints, mostly dispatch to the above:
 183  */
 184
 185
 186 /* Interpolate between two vertices to produce a third:
 187  */
 188 void _tnl_interp( GLcontext *ctx,
 189                   GLfloat t,
 190                   GLuint edst, GLuint eout, GLuint ein,
 191                   GLboolean force_boundary )
 192 {
 193    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 194    vtx->interp( ctx, t, edst, eout, ein, force_boundary );
 195 }
 196
 197 /* Copy colors from one vertex to another:
 198  */
 199 void _tnl_copy_pv(  GLcontext *ctx, GLuint edst, GLuint esrc )
 200 {
 201    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 202    vtx->copy_pv( ctx, edst, esrc );
 203 }
 204
 205
 206 /* Extract a named attribute from a hardware vertex.  Will have to
 207  * reverse any viewport transformation, swizzling or other conversions
 208  * which may have been applied:
 209  */
 210 void _tnl_get_attr( GLcontext *ctx, const void *vin,
 211                               GLenum attr, GLfloat *dest )
 212 {
 213    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 214    const struct tnl_clipspace_attr *a = vtx->attr;
 215    const GLuint attr_count = vtx->attr_count;
 216    GLuint j;
 217
 218    for (j = 0; j < attr_count; j++) {
 219       if (a[j].attrib == attr) {
 220          a[j].extract( &a[j], dest, (GLubyte *)vin + a[j].vertoffset );
 221          return;
 222       }
 223    }
 224
 225    /* Else return the value from ctx->Current.
 226     */
 227    _mesa_memcpy( dest, ctx->Current.Attrib[attr], 4*sizeof(GLfloat));
 228 }
 229
 230
 231 /* Complementary operation to the above.
 232  */
 233 void _tnl_set_attr( GLcontext *ctx, void *vout,
 234                     GLenum attr, const GLfloat *src )
 235 {
 236    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 237    const struct tnl_clipspace_attr *a = vtx->attr;
 238    const GLuint attr_count = vtx->attr_count;
 239    GLuint j;
 240
 241    for (j = 0; j < attr_count; j++) {
 242       if (a[j].attrib == attr) {
 243          a[j].insert[4-1]( &a[j], (GLubyte *)vout + a[j].vertoffset, src );
 244          return;
 245       }
 246    }
 247 }
 248
 249
 250 void *_tnl_get_vertex( GLcontext *ctx, GLuint nr )
 251 {
 252    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 253
 254    return vtx->vertex_buf + nr * vtx->vertex_size;
 255 }
 256
 257 void _tnl_invalidate_vertex_state( GLcontext *ctx, GLuint new_state )
 258 {
 259    if (new_state & (_DD_NEW_TRI_LIGHT_TWOSIDE|_DD_NEW_TRI_UNFILLED) ) {
 260       struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 261       vtx->new_inputs = ~0;
 262       vtx->interp = choose_interp_func;
 263       vtx->copy_pv = choose_copy_pv_func;
 264    }
 265 }
 266
 267 static void invalidate_funcs( struct tnl_clipspace *vtx )
 268 {
 269    vtx->emit = choose_emit_func;
 270    vtx->interp = choose_interp_func;
 271    vtx->copy_pv = choose_copy_pv_func;
 272    vtx->new_inputs = ~0;
 273 }
 274
 275 GLuint _tnl_install_attrs( GLcontext *ctx, const struct tnl_attr_map *map,
 276                            GLuint nr, const GLfloat *vp,
 277                            GLuint unpacked_size )
 278 {
 279    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 280    GLuint offset = 0;
 281    GLuint i, j;
 282
 283    assert(nr < _TNL_ATTRIB_MAX);
 284    assert(nr == 0 || map[0].attrib == VERT_ATTRIB_POS);
 285
 286    vtx->new_inputs = ~0;
 287    vtx->need_viewport = GL_FALSE;
 288
 289    if (vp) {
 290       vtx->need_viewport = GL_TRUE;
 291       vtx->vp_scale[0] = vp[MAT_SX];
 292       vtx->vp_scale[1] = vp[MAT_SY];
 293       vtx->vp_scale[2] = vp[MAT_SZ];
 294       vtx->vp_scale[3] = 1.0;
 295       vtx->vp_xlate[0] = vp[MAT_TX];
 296       vtx->vp_xlate[1] = vp[MAT_TY];
 297       vtx->vp_xlate[2] = vp[MAT_TZ];
 298       vtx->vp_xlate[3] = 0.0;
 299    }
 300
 301    for (j = 0, i = 0; i < nr; i++) {
 302       const GLuint format = map[i].format;
 303       if (format == EMIT_PAD) {
 304          if (DBG)
 305             _mesa_printf("%d: pad %d, offset %d\n", i,
 306                          map[i].offset, offset);
 307
 308          offset += map[i].offset;
 309
 310       }
 311       else {
 312          GLuint tmpoffset;
 313
 314          if (unpacked_size)
 315             tmpoffset = map[i].offset;
 316          else
 317             tmpoffset = offset;
 318
 319          if (vtx->attr_count != j ||
 320              vtx->attr[j].attrib != map[i].attrib ||
 321              vtx->attr[j].format != format ||
 322              vtx->attr[j].vertoffset != tmpoffset) {
 323             invalidate_funcs(vtx);
 324
 325             vtx->attr[j].attrib = map[i].attrib;
 326             vtx->attr[j].format = format;
 327             vtx->attr[j].vp = vp;
 328             vtx->attr[j].insert = _tnl_format_info[format].insert;
 329             vtx->attr[j].extract = _tnl_format_info[format].extract;
 330             vtx->attr[j].vertattrsize = _tnl_format_info[format].attrsize;
 331             vtx->attr[j].vertoffset = tmpoffset;
 332          }
 333
 334
 335          if (DBG)
 336             _mesa_printf("%d: %s, vp %p, offset %d\n", i,
 337                          _tnl_format_info[format].name, (void *)vp,
 338                          vtx->attr[j].vertoffset);
 339
 340          offset += _tnl_format_info[format].attrsize;
 341          j++;
 342       }
 343    }
 344
 345    vtx->attr_count = j;
 346
 347    if (unpacked_size)
 348       vtx->vertex_size = unpacked_size;
 349    else
 350       vtx->vertex_size = offset;
 351
 352    assert(vtx->vertex_size <= vtx->max_vertex_size);
 353    return vtx->vertex_size;
 354 }
 355
 356
 357
 358 void _tnl_invalidate_vertices( GLcontext *ctx, GLuint newinputs )
 359 {
 360    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 361    vtx->new_inputs |= newinputs;
 362 }
 363
 364
 365 /* This event has broader use beyond this file - will move elsewhere
 366  * and probably invoke a driver callback.
 367  */
 368 void _tnl_notify_pipeline_output_change( GLcontext *ctx )
 369 {
 370    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 371    invalidate_funcs(vtx);
 372 }
 373
 374 static void update_input_ptrs( GLcontext *ctx, GLuint start )
 375 {
 376    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
 377    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 378    struct tnl_clipspace_attr *a = vtx->attr;
 379    const GLuint count = vtx->attr_count;
 380    GLuint j;
 381
 382    for (j = 0; j < count; j++) {
 383       GLvector4f *vptr = VB->AttribPtr[a[j].attrib];
 384
 385       if (vtx->emit != choose_emit_func) {
 386          assert(a[j].inputstride == vptr->stride);
 387          assert(a[j].inputsize == vptr->size);
 388       }
 389
 390       a[j].inputptr = ((GLubyte *)vptr->data) + start * vptr->stride;
 391    }
 392 }
 393
 394
 395 void _tnl_build_vertices( GLcontext *ctx,
 396                           GLuint start,
 397                           GLuint end,
 398                           GLuint newinputs )
 399 {
 400    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 401    update_input_ptrs( ctx, start );
 402    vtx->emit( ctx, end - start,
 403               (GLubyte *)(vtx->vertex_buf +
 404                           start * vtx->vertex_size));
 405 }
 406
 407 /* Emit VB vertices start..end to dest.  Note that VB vertex at
 408  * postion start will be emitted to dest at position zero.
 409  */
 410 void *_tnl_emit_vertices_to_buffer( GLcontext *ctx,
 411                                     GLuint start,
 412                                     GLuint end,
 413                                     void *dest )
 414 {
 415    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 416
 417    update_input_ptrs(ctx, start);
 418
 419    /* Note: dest should not be adjusted for non-zero 'start' values:
 420     */
 421    vtx->emit( ctx, end - start, dest );
 422    return (void *)((GLubyte *)dest + vtx->vertex_size * (end - start));
 423 }
 424
 425
 426 void _tnl_init_vertices( GLcontext *ctx,
 427                         GLuint vb_size,
 428                         GLuint max_vertex_size )
 429 {
 430    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 431
 432    _tnl_install_attrs( ctx, NULL, 0, NULL, 0 );
 433
 434    vtx->need_extras = GL_TRUE;
 435    if (max_vertex_size > vtx->max_vertex_size) {
 436       _tnl_free_vertices( ctx );
 437       vtx->max_vertex_size = max_vertex_size;
 438       vtx->vertex_buf = (GLubyte *)ALIGN_CALLOC(vb_size * max_vertex_size, 32 );
 439       invalidate_funcs(vtx);
 440    }
 441
 442    switch(CHAN_TYPE) {
 443    case GL_UNSIGNED_BYTE:
 444       vtx->chan_scale[0] = 255.0;
 445       vtx->chan_scale[1] = 255.0;
 446       vtx->chan_scale[2] = 255.0;
 447       vtx->chan_scale[3] = 255.0;
 448       break;
 449    case GL_UNSIGNED_SHORT:
 450       vtx->chan_scale[0] = 65535.0;
 451       vtx->chan_scale[1] = 65535.0;
 452       vtx->chan_scale[2] = 65535.0;
 453       vtx->chan_scale[3] = 65535.0;
 454       break;
 455    default:
 456       vtx->chan_scale[0] = 1.0;
 457       vtx->chan_scale[1] = 1.0;
 458       vtx->chan_scale[2] = 1.0;
 459       vtx->chan_scale[3] = 1.0;
 460       break;
 461    }
 462
 463    vtx->identity[0] = 0.0;
 464    vtx->identity[1] = 0.0;
 465    vtx->identity[2] = 0.0;
 466    vtx->identity[3] = 1.0;
 467
 468    vtx->codegen_emit = NULL;
 469
 470 #ifdef __i386__
 471    if (getenv("MESA_EXPERIMENTAL"))
 472       vtx->codegen_emit = _tnl_generate_sse_emit;
 473 #endif
 474 }
 475
 476
 477 void _tnl_free_vertices( GLcontext *ctx )
 478 {
 479    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 480    struct tnl_clipspace_fastpath *fp, *tmp;
 481
 482    if (vtx->vertex_buf) {
 483       ALIGN_FREE(vtx->vertex_buf);
 484       vtx->vertex_buf = NULL;
 485    }
 486
 487    for (fp = vtx->fastpath ; fp ; fp = tmp) {
 488       tmp = fp->next;
 489       FREE(fp->attr);
 490       FREE((void *)fp->func);
 491       FREE(fp);
 492    }
 493
 494    vtx->fastpath = NULL;
 495 }