src/mesa/state_tracker/st_atom_shader.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 /**
  29  * State validation for vertex/fragment shaders.
  30  * Note that we have to delay most vertex/fragment shader translation
  31  * until rendering time since the linkage between the vertex outputs and
  32  * fragment inputs can vary depending on the pairing of shaders.
  33  *
  34  * Authors:
  35  *   Brian Paul
  36  */
  37
  38
  39
  40 #include "main/imports.h"
  41 #include "main/mtypes.h"
  42
  43 #include "pipe/p_context.h"
  44 #include "pipe/p_shader_tokens.h"
  45
  46 #include "cso_cache/cso_context.h"
  47
  48 #include "st_context.h"
  49 #include "st_atom.h"
  50 #include "st_program.h"
  51 #include "st_atom_shader.h"
  52 #include "st_mesa_to_tgsi.h"
  53
  54
  55 /**
  56  * This represents a vertex program, especially translated to match
  57  * the inputs of a particular fragment shader.
  58  */
  59 struct translated_vertex_program
  60 {
  61    struct st_vertex_program *master;
  62
  63    /** The fragment shader "signature" this vertex shader is meant for: */
  64    GLbitfield frag_inputs;
  65
  66    /** Compared against master vertex program's serialNo: */
  67    GLuint serialNo;
  68
  69    /** Maps VERT_RESULT_x to slot */
  70    GLuint output_to_slot[VERT_RESULT_MAX];
  71
  72    /** Pointer to the translated vertex program */
  73    struct st_vertex_program *vp;
  74
  75    struct translated_vertex_program *next;  /**< next in linked list */
  76 };
  77
  78
  79
  80 /**
  81  * Given a vertex program output attribute, return the corresponding
  82  * fragment program input attribute.
  83  * \return -1 for vertex outputs that have no corresponding fragment input
  84  */
  85 static GLint
  86 vp_out_to_fp_in(GLuint vertResult)
  87 {
  88    if (vertResult >= VERT_RESULT_TEX0 &&
  89        vertResult < VERT_RESULT_TEX0 + MAX_TEXTURE_COORD_UNITS)
  90       return FRAG_ATTRIB_TEX0 + (vertResult - VERT_RESULT_TEX0);
  91
  92    if (vertResult >= VERT_RESULT_VAR0 &&
  93        vertResult < VERT_RESULT_VAR0 + MAX_VARYING)
  94       return FRAG_ATTRIB_VAR0 + (vertResult - VERT_RESULT_VAR0);
  95
  96    switch (vertResult) {
  97    case VERT_RESULT_HPOS:
  98       return FRAG_ATTRIB_WPOS;
  99    case VERT_RESULT_COL0:
 100       return FRAG_ATTRIB_COL0;
 101    case VERT_RESULT_COL1:
 102       return FRAG_ATTRIB_COL1;
 103    case VERT_RESULT_FOGC:
 104       return FRAG_ATTRIB_FOGC;
 105    default:
 106       /* Back-face colors, edge flags, etc */
 107       return -1;
 108    }
 109 }
 110
 111
 112 /**
 113  * Find a translated vertex program that corresponds to stvp and
 114  * has outputs matched to stfp's inputs.
 115  * This performs vertex and fragment translation (to TGSI) when needed.
 116  */
 117 static struct translated_vertex_program *
 118 find_translated_vp(struct st_context *st,
 119                    struct st_vertex_program *stvp,
 120                    struct st_fragment_program *stfp)
 121 {
 122    static const GLuint UNUSED = ~0;
 123    struct translated_vertex_program *xvp;
 124    const GLbitfield fragInputsRead = stfp->Base.Base.InputsRead;
 125
 126    /*
 127     * Translate fragment program if needed.
 128     */
 129    if (!stfp->state.tokens) {
 130       GLuint inAttr, numIn = 0;
 131
 132       for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) {
 133          if (fragInputsRead & (1 << inAttr)) {
 134             stfp->input_to_slot[inAttr] = numIn;
 135             numIn++;
 136          }
 137          else {
 138             stfp->input_to_slot[inAttr] = UNUSED;
 139          }
 140       }
 141
 142       stfp->num_input_slots = numIn;
 143
 144       assert(stfp->Base.Base.NumInstructions > 1);
 145
 146       st_translate_fragment_program(st, stfp, stfp->input_to_slot);
 147    }
 148
 149
 150    /* See if we've got a translated vertex program whose outputs match
 151     * the fragment program's inputs.
 152     * XXX This could be a hash lookup, using InputsRead as the key.
 153     */
 154    for (xvp = stfp->vertex_programs; xvp; xvp = xvp->next) {
 155       if (xvp->master == stvp && xvp->frag_inputs == fragInputsRead) {
 156          break;
 157       }
 158    }
 159
 160    /* No?  Allocate translated vp object now */
 161    if (!xvp) {
 162       xvp = CALLOC_STRUCT(translated_vertex_program);
 163       xvp->frag_inputs = fragInputsRead;
 164       xvp->master = stvp;
 165
 166       xvp->next = stfp->vertex_programs;
 167       stfp->vertex_programs = xvp;
 168    }
 169
 170    /* See if we need to translate vertex program to TGSI form */
 171    if (xvp->serialNo != stvp->serialNo) {
 172       GLuint outAttr, dummySlot;
 173       const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten;
 174       GLuint numVpOuts = 0;
 175
 176       /* Compute mapping of vertex program outputs to slots, which depends
 177        * on the fragment program's input->slot mapping.
 178        */
 179       for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
 180          /* set default: */
 181          xvp->output_to_slot[outAttr] = UNUSED;
 182
 183          if (outAttr == VERT_RESULT_HPOS) {
 184             /* always put xformed position into slot zero */
 185             xvp->output_to_slot[VERT_RESULT_HPOS] = 0;
 186             numVpOuts++;
 187          }
 188          else if (outputsWritten & (1 << outAttr)) {
 189             /* see if the frag prog wants this vert output */
 190             GLint fpInAttrib = vp_out_to_fp_in(outAttr);
 191             if (fpInAttrib >= 0) {
 192                GLuint fpInSlot = stfp->input_to_slot[fpInAttrib];
 193                if (fpInSlot != ~0) {
 194                   GLuint vpOutSlot = stfp->input_map[fpInSlot];
 195                   xvp->output_to_slot[outAttr] = vpOutSlot;
 196                   numVpOuts++;
 197                }
 198             }
 199             else if (outAttr == VERT_RESULT_PSIZ ||
 200                      outAttr == VERT_RESULT_BFC0 ||
 201                      outAttr == VERT_RESULT_BFC1) {
 202                /* backface colors go into last slots */
 203                xvp->output_to_slot[outAttr] = numVpOuts++;
 204             }
 205          }
 206          /*
 207          printf("output_to_slot[%d] = %d\n", outAttr,
 208                 xvp->output_to_slot[outAttr]);
 209          */
 210       }
 211
 212       /* Unneeded vertex program outputs will go to this slot.
 213        * We could use this info to do dead code elimination in the
 214        * vertex program.
 215        */
 216       dummySlot = numVpOuts;
 217
 218       /* Map vert program outputs that aren't used to the dummy slot */
 219       for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
 220          if (outputsWritten & (1 << outAttr)) {
 221             if (xvp->output_to_slot[outAttr] == UNUSED)
 222                xvp->output_to_slot[outAttr] = dummySlot;
 223          }
 224       }
 225
 226       assert(stvp->Base.Base.NumInstructions > 1);
 227
 228       st_translate_vertex_program(st, stvp, xvp->output_to_slot);
 229
 230       xvp->vp = stvp;
 231
 232       /* translated VP is up to date now */
 233       xvp->serialNo = stvp->serialNo;
 234    }
 235
 236    return xvp;
 237 }
 238
 239
 240 void
 241 st_free_translated_vertex_programs(struct st_context *st,
 242                                    struct translated_vertex_program *xvp)
 243 {
 244    struct translated_vertex_program *next;
 245
 246    while (xvp) {
 247       next = xvp->next;
 248       free(xvp);
 249       xvp = next;
 250    }
 251 }
 252
 253
 254
 255 static void
 256 update_linkage( struct st_context *st )
 257 {
 258    struct st_vertex_program *stvp;
 259    struct st_fragment_program *stfp;
 260    struct translated_vertex_program *xvp;
 261
 262    /* find active shader and params -- Should be covered by
 263     * ST_NEW_VERTEX_PROGRAM
 264     */
 265    assert(st->ctx->VertexProgram._Current);
 266    stvp = st_vertex_program(st->ctx->VertexProgram._Current);
 267
 268    assert(st->ctx->FragmentProgram._Current);
 269    stfp = st_fragment_program(st->ctx->FragmentProgram._Current);
 270
 271    xvp = find_translated_vp(st, stvp, stfp);
 272
 273    st->vp = stvp;
 274    st->fp = stfp;
 275
 276    st->pipe->bind_vs_state(st->pipe, stvp->driver_shader);
 277    st->pipe->bind_fs_state(st->pipe, stfp->driver_shader);
 278
 279    st->vertex_result_to_slot = xvp->output_to_slot;
 280 }
 281
 282
 283 const struct st_tracked_state st_update_shader = {
 284    .name = "st_update_shader",
 285    .dirty = {
 286       .mesa  = 0,
 287       .st   = ST_NEW_VERTEX_PROGRAM | ST_NEW_FRAGMENT_PROGRAM
 288    },
 289    .update = update_linkage
 290 };