src/mesa/tnl/t_vb_program.c

   1 /*
   2  * Mesa 3-D graphics library
   3  *
   4  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
   5  * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
   6  *
   7  * Permission is hereby granted, free of charge, to any person obtaining a
   8  * copy of this software and associated documentation files (the "Software"),
   9  * to deal in the Software without restriction, including without limitation
  10  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
  11  * and/or sell copies of the Software, and to permit persons to whom the
  12  * Software is furnished to do so, subject to the following conditions:
  13  *
  14  * The above copyright notice and this permission notice shall be included
  15  * in all copies or substantial portions of the Software.
  16  *
  17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  18  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  20  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
  21  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
  22  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
  23  * OTHER DEALINGS IN THE SOFTWARE.
  24  */
  25
  26
  27 /**
  28  * \file tnl/t_vb_program.c
  29  * \brief Pipeline stage for executing vertex programs.
  30  * \author Brian Paul,  Keith Whitwell
  31  */
  32
  33
  34 #include "main/glheader.h"
  35 #include "main/macros.h"
  36 #include "main/imports.h"
  37 #include "main/samplerobj.h"
  38 #include "math/m_xform.h"
  39 #include "program/prog_instruction.h"
  40 #include "program/prog_statevars.h"
  41 #include "program/prog_execute.h"
  42 #include "swrast/s_context.h"
  43
  44 #include "tnl/tnl.h"
  45 #include "tnl/t_context.h"
  46 #include "tnl/t_pipeline.h"
  47
  48
  49 #ifdef NAN_CHECK
  50 /** Check for NaNs and very large values */
  51 static inline void
  52 check_float(float x)
  53 {
  54    assert(!IS_INF_OR_NAN(x));
  55    assert(1.0e-15 <= x && x <= 1.0e15);
  56 }
  57 #endif
  58
  59
  60 /*!
  61  * Private storage for the vertex program pipeline stage.
  62  */
  63 struct vp_stage_data {
  64    /** The results of running the vertex program go into these arrays. */
  65    GLvector4f results[VARYING_SLOT_MAX];
  66
  67    GLvector4f ndcCoords;              /**< normalized device coords */
  68    GLubyte *clipmask;                 /**< clip flags */
  69    GLubyte ormask, andmask;           /**< for clipping */
  70
  71    GLboolean vertex_textures;
  72
  73    struct gl_program_machine machine;
  74 };
  75
  76
  77 #define VP_STAGE_DATA(stage) ((struct vp_stage_data *)(stage->privatePtr))
  78
  79
  80 static void
  81 userclip( struct gl_context *ctx,
  82           GLvector4f *clip,
  83           GLubyte *clipmask,
  84           GLubyte *clipormask,
  85           GLubyte *clipandmask )
  86 {
  87    GLuint p;
  88
  89    for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
  90       if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
  91          GLuint nr, i;
  92          const GLfloat a = ctx->Transform._ClipUserPlane[p][0];
  93          const GLfloat b = ctx->Transform._ClipUserPlane[p][1];
  94          const GLfloat c = ctx->Transform._ClipUserPlane[p][2];
  95          const GLfloat d = ctx->Transform._ClipUserPlane[p][3];
  96          GLfloat *coord = (GLfloat *)clip->data;
  97          GLuint stride = clip->stride;
  98          GLuint count = clip->count;
  99
 100          for (nr = 0, i = 0 ; i < count ; i++) {
 101             GLfloat dp = (coord[0] * a +
 102                           coord[1] * b +
 103                           coord[2] * c +
 104                           coord[3] * d);
 105
 106             if (dp < 0) {
 107                nr++;
 108                clipmask[i] |= CLIP_USER_BIT;
 109             }
 110
 111             STRIDE_F(coord, stride);
 112          }
 113
 114          if (nr > 0) {
 115             *clipormask |= CLIP_USER_BIT;
 116             if (nr == count) {
 117                *clipandmask |= CLIP_USER_BIT;
 118                return;
 119             }
 120          }
 121       }
 122    }
 123 }
 124
 125
 126 static GLboolean
 127 do_ndc_cliptest(struct gl_context *ctx, struct vp_stage_data *store)
 128 {
 129    TNLcontext *tnl = TNL_CONTEXT(ctx);
 130    struct vertex_buffer *VB = &tnl->vb;
 131    /* Cliptest and perspective divide.  Clip functions must clear
 132     * the clipmask.
 133     */
 134    store->ormask = 0;
 135    store->andmask = CLIP_FRUSTUM_BITS;
 136
 137    tnl_clip_prepare(ctx);
 138
 139    if (tnl->NeedNdcCoords) {
 140       VB->NdcPtr =
 141          _mesa_clip_tab[VB->ClipPtr->size]( VB->ClipPtr,
 142                                             &store->ndcCoords,
 143                                             store->clipmask,
 144                                             &store->ormask,
 145                                             &store->andmask,
 146                                             !ctx->Transform.DepthClamp );
 147    }
 148    else {
 149       VB->NdcPtr = NULL;
 150       _mesa_clip_np_tab[VB->ClipPtr->size]( VB->ClipPtr,
 151                                             NULL,
 152                                             store->clipmask,
 153                                             &store->ormask,
 154                                             &store->andmask,
 155                                             !ctx->Transform.DepthClamp );
 156    }
 157
 158    if (store->andmask) {
 159       /* All vertices are outside the frustum */
 160       return GL_FALSE;
 161    }
 162
 163    /* Test userclip planes.  This contributes to VB->ClipMask.
 164     */
 165    /** XXX NEW_SLANG _Enabled ??? */
 166    if (ctx->Transform.ClipPlanesEnabled && (!ctx->VertexProgram._Enabled ||
 167       ctx->VertexProgram.Current->IsPositionInvariant)) {
 168       userclip( ctx,
 169                 VB->ClipPtr,
 170                 store->clipmask,
 171                 &store->ormask,
 172                 &store->andmask );
 173
 174       if (store->andmask) {
 175          return GL_FALSE;
 176       }
 177    }
 178
 179    VB->ClipAndMask = store->andmask;
 180    VB->ClipOrMask = store->ormask;
 181    VB->ClipMask = store->clipmask;
 182
 183    return GL_TRUE;
 184 }
 185
 186
 187 /**
 188  * XXX the texture sampling code in this module is a bit of a hack.
 189  * The texture sampling code is in swrast, though it doesn't have any
 190  * real dependencies on the rest of swrast.  It should probably be
 191  * moved into main/ someday.
 192  */
 193 static void
 194 vp_fetch_texel(struct gl_context *ctx, const GLfloat texcoord[4], GLfloat lambda,
 195                GLuint unit, GLfloat color[4])
 196 {
 197    SWcontext *swrast = SWRAST_CONTEXT(ctx);
 198
 199    /* XXX use a float-valued TextureSample routine here!!! */
 200    swrast->TextureSample[unit](ctx, _mesa_get_samplerobj(ctx, unit),
 201                                ctx->Texture.Unit[unit]._Current,
 202                                1, (const GLfloat (*)[4]) texcoord,
 203                                &lambda,  (GLfloat (*)[4]) color);
 204 }
 205
 206
 207 /**
 208  * Called via ctx->Driver.ProgramStringNotify() after a new vertex program
 209  * string has been parsed.
 210  */
 211 GLboolean
 212 _tnl_program_string(struct gl_context *ctx, GLenum target, struct gl_program *program)
 213 {
 214    /* No-op.
 215     * If we had derived anything from the program that was private to this
 216     * stage we'd recompute/validate it here.
 217     */
 218    return GL_TRUE;
 219 }
 220
 221
 222 /**
 223  * Initialize virtual machine state prior to executing vertex program.
 224  */
 225 static void
 226 init_machine(struct gl_context *ctx, struct gl_program_machine *machine,
 227              GLuint instID)
 228 {
 229    /* Input registers get initialized from the current vertex attribs */
 230    memcpy(machine->VertAttribs, ctx->Current.Attrib,
 231           MAX_VERTEX_GENERIC_ATTRIBS * 4 * sizeof(GLfloat));
 232
 233    machine->NumDeriv = 0;
 234
 235    /* init call stack */
 236    machine->StackDepth = 0;
 237
 238    machine->FetchTexelLod = vp_fetch_texel;
 239    machine->FetchTexelDeriv = NULL; /* not used by vertex programs */
 240
 241    machine->Samplers = ctx->VertexProgram._Current->Base.SamplerUnits;
 242
 243    machine->SystemValues[SYSTEM_VALUE_INSTANCE_ID][0] = (GLfloat) instID;
 244 }
 245
 246
 247 /**
 248  * Map the texture images which the vertex program will access (if any).
 249  */
 250 static void
 251 map_textures(struct gl_context *ctx, const struct gl_vertex_program *vp)
 252 {
 253    GLuint u;
 254
 255    for (u = 0; u < ctx->Const.Program[MESA_SHADER_VERTEX].MaxTextureImageUnits; u++) {
 256       if (vp->Base.TexturesUsed[u]) {
 257          /* Note: _Current *should* correspond to the target indicated
 258           * in TexturesUsed[u].
 259           */
 260          _swrast_map_texture(ctx, ctx->Texture.Unit[u]._Current);
 261       }
 262    }
 263 }
 264
 265
 266 /**
 267  * Unmap the texture images which were used by the vertex program (if any).
 268  */
 269 static void
 270 unmap_textures(struct gl_context *ctx, const struct gl_vertex_program *vp)
 271 {
 272    GLuint u;
 273
 274    for (u = 0; u < ctx->Const.Program[MESA_SHADER_VERTEX].MaxTextureImageUnits; u++) {
 275       if (vp->Base.TexturesUsed[u]) {
 276          /* Note: _Current *should* correspond to the target indicated
 277           * in TexturesUsed[u].
 278           */
 279          _swrast_unmap_texture(ctx, ctx->Texture.Unit[u]._Current);
 280       }
 281    }
 282 }
 283
 284
 285 /**
 286  * This function executes vertex programs
 287  */
 288 static GLboolean
 289 run_vp( struct gl_context *ctx, struct tnl_pipeline_stage *stage )
 290 {
 291    TNLcontext *tnl = TNL_CONTEXT(ctx);
 292    struct vp_stage_data *store = VP_STAGE_DATA(stage);
 293    struct vertex_buffer *VB = &tnl->vb;
 294    struct gl_vertex_program *program = ctx->VertexProgram._Current;
 295    struct gl_program_machine *machine = &store->machine;
 296    GLuint outputs[VARYING_SLOT_MAX], numOutputs;
 297    GLuint i, j;
 298
 299    if (!program)
 300       return GL_TRUE;
 301
 302    /* ARB program or vertex shader */
 303    _mesa_load_state_parameters(ctx, program->Base.Parameters);
 304
 305    /* make list of outputs to save some time below */
 306    numOutputs = 0;
 307    for (i = 0; i < VARYING_SLOT_MAX; i++) {
 308       if (program->Base.OutputsWritten & BITFIELD64_BIT(i)) {
 309          outputs[numOutputs++] = i;
 310       }
 311    }
 312
 313    /* Allocate result vectors.  We delay this until now to avoid allocating
 314     * memory that would never be used if we don't run the software tnl pipeline.
 315     */
 316    if (!store->results[0].storage) {
 317       for (i = 0; i < VARYING_SLOT_MAX; i++) {
 318          assert(!store->results[i].storage);
 319          _mesa_vector4f_alloc( &store->results[i], 0, VB->Size, 32 );
 320          store->results[i].size = 4;
 321       }
 322    }
 323
 324    map_textures(ctx, program);
 325
 326    for (i = 0; i < VB->Count; i++) {
 327       GLuint attr;
 328
 329       init_machine(ctx, machine, tnl->CurInstance);
 330
 331 #if 0
 332       printf("Input  %d: %f, %f, %f, %f\n", i,
 333              VB->AttribPtr[0]->data[i][0],
 334              VB->AttribPtr[0]->data[i][1],
 335              VB->AttribPtr[0]->data[i][2],
 336              VB->AttribPtr[0]->data[i][3]);
 337       printf("   color: %f, %f, %f, %f\n",
 338              VB->AttribPtr[3]->data[i][0],
 339              VB->AttribPtr[3]->data[i][1],
 340              VB->AttribPtr[3]->data[i][2],
 341              VB->AttribPtr[3]->data[i][3]);
 342       printf("  normal: %f, %f, %f, %f\n",
 343              VB->AttribPtr[2]->data[i][0],
 344              VB->AttribPtr[2]->data[i][1],
 345              VB->AttribPtr[2]->data[i][2],
 346              VB->AttribPtr[2]->data[i][3]);
 347 #endif
 348
 349       /* the vertex array case */
 350       for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
 351          if (program->Base.InputsRead & BITFIELD64_BIT(attr)) {
 352             const GLubyte *ptr = (const GLubyte*) VB->AttribPtr[attr]->data;
 353             const GLuint size = VB->AttribPtr[attr]->size;
 354             const GLuint stride = VB->AttribPtr[attr]->stride;
 355             const GLfloat *data = (GLfloat *) (ptr + stride * i);
 356 #ifdef NAN_CHECK
 357             check_float(data[0]);
 358             check_float(data[1]);
 359             check_float(data[2]);
 360             check_float(data[3]);
 361 #endif
 362             COPY_CLEAN_4V(machine->VertAttribs[attr], size, data);
 363          }
 364       }
 365
 366       /* execute the program */
 367       _mesa_execute_program(ctx, &program->Base, machine);
 368
 369       /* copy the output registers into the VB->attribs arrays */
 370       for (j = 0; j < numOutputs; j++) {
 371          const GLuint attr = outputs[j];
 372 #ifdef NAN_CHECK
 373          check_float(machine->Outputs[attr][0]);
 374          check_float(machine->Outputs[attr][1]);
 375          check_float(machine->Outputs[attr][2]);
 376          check_float(machine->Outputs[attr][3]);
 377 #endif
 378          COPY_4V(store->results[attr].data[i], machine->Outputs[attr]);
 379       }
 380
 381       /* FOGC is a special case.  Fragment shader expects (f,0,0,1) */
 382       if (program->Base.OutputsWritten & BITFIELD64_BIT(VARYING_SLOT_FOGC)) {
 383          store->results[VARYING_SLOT_FOGC].data[i][1] = 0.0;
 384          store->results[VARYING_SLOT_FOGC].data[i][2] = 0.0;
 385          store->results[VARYING_SLOT_FOGC].data[i][3] = 1.0;
 386       }
 387 #ifdef NAN_CHECK
 388       assert(machine->Outputs[0][3] != 0.0F);
 389 #endif
 390 #if 0
 391       printf("HPOS: %f %f %f %f\n",
 392              machine->Outputs[0][0],
 393              machine->Outputs[0][1],
 394              machine->Outputs[0][2],
 395              machine->Outputs[0][3]);
 396 #endif
 397    }
 398
 399    unmap_textures(ctx, program);
 400
 401    if (program->IsPositionInvariant) {
 402       /* We need the exact same transform as in the fixed function path here
 403        * to guarantee invariance, depending on compiler optimization flags
 404        * results could be different otherwise.
 405        */
 406       VB->ClipPtr = TransformRaw( &store->results[0],
 407                                   &ctx->_ModelProjectMatrix,
 408                                   VB->AttribPtr[0] );
 409
 410       /* Drivers expect this to be clean to element 4...
 411        */
 412       switch (VB->ClipPtr->size) {
 413       case 1:
 414          /* impossible */
 415       case 2:
 416          _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 2 );
 417          /* fall-through */
 418       case 3:
 419          _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 3 );
 420          /* fall-through */
 421       case 4:
 422          break;
 423       }
 424    }
 425    else {
 426       /* Setup the VB pointers so that the next pipeline stages get
 427        * their data from the right place (the program output arrays).
 428        */
 429       VB->ClipPtr = &store->results[VARYING_SLOT_POS];
 430       VB->ClipPtr->size = 4;
 431       VB->ClipPtr->count = VB->Count;
 432    }
 433
 434    VB->AttribPtr[VERT_ATTRIB_COLOR0] = &store->results[VARYING_SLOT_COL0];
 435    VB->AttribPtr[VERT_ATTRIB_COLOR1] = &store->results[VARYING_SLOT_COL1];
 436    VB->AttribPtr[VERT_ATTRIB_FOG] = &store->results[VARYING_SLOT_FOGC];
 437    VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->results[VARYING_SLOT_PSIZ];
 438    VB->BackfaceColorPtr = &store->results[VARYING_SLOT_BFC0];
 439    VB->BackfaceSecondaryColorPtr = &store->results[VARYING_SLOT_BFC1];
 440
 441    for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
 442       VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]
 443          = &store->results[VARYING_SLOT_TEX0 + i];
 444    }
 445
 446    for (i = 0; i < ctx->Const.MaxVarying; i++) {
 447       if (program->Base.OutputsWritten & BITFIELD64_BIT(VARYING_SLOT_VAR0 + i)) {
 448          /* Note: varying results get put into the generic attributes */
 449          VB->AttribPtr[VERT_ATTRIB_GENERIC0+i]
 450             = &store->results[VARYING_SLOT_VAR0 + i];
 451       }
 452    }
 453
 454
 455    /* Perform NDC and cliptest operations:
 456     */
 457    return do_ndc_cliptest(ctx, store);
 458 }
 459
 460
 461 /**
 462  * Called the first time stage->run is called.  In effect, don't
 463  * allocate data until the first time the stage is run.
 464  */
 465 static GLboolean
 466 init_vp(struct gl_context *ctx, struct tnl_pipeline_stage *stage)
 467 {
 468    TNLcontext *tnl = TNL_CONTEXT(ctx);
 469    struct vertex_buffer *VB = &(tnl->vb);
 470    struct vp_stage_data *store;
 471    const GLuint size = VB->Size;
 472
 473    stage->privatePtr = calloc(1, sizeof(*store));
 474    store = VP_STAGE_DATA(stage);
 475    if (!store)
 476       return GL_FALSE;
 477
 478    /* a few other misc allocations */
 479    _mesa_vector4f_alloc( &store->ndcCoords, 0, size, 32 );
 480    store->clipmask = _mesa_align_malloc(sizeof(GLubyte)*size, 32 );
 481
 482    return GL_TRUE;
 483 }
 484
 485
 486 /**
 487  * Destructor for this pipeline stage.
 488  */
 489 static void
 490 dtr(struct tnl_pipeline_stage *stage)
 491 {
 492    struct vp_stage_data *store = VP_STAGE_DATA(stage);
 493
 494    if (store) {
 495       GLuint i;
 496
 497       /* free the vertex program result arrays */
 498       for (i = 0; i < VARYING_SLOT_MAX; i++)
 499          _mesa_vector4f_free( &store->results[i] );
 500
 501       /* free misc arrays */
 502       _mesa_vector4f_free( &store->ndcCoords );
 503       _mesa_align_free( store->clipmask );
 504
 505       free( store );
 506       stage->privatePtr = NULL;
 507    }
 508 }
 509
 510
 511 static void
 512 validate_vp_stage(struct gl_context *ctx, struct tnl_pipeline_stage *stage)
 513 {
 514    if (ctx->VertexProgram._Current) {
 515       _swrast_update_texture_samplers(ctx);
 516    }
 517 }
 518
 519
 520
 521 /**
 522  * Public description of this pipeline stage.
 523  */
 524 const struct tnl_pipeline_stage _tnl_vertex_program_stage =
 525 {
 526    "vertex-program",
 527    NULL,                        /* private_data */
 528    init_vp,                     /* create */
 529    dtr,                         /* destroy */
 530    validate_vp_stage,           /* validate */
 531    run_vp                       /* run -- initially set to ctr */
 532 };