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 "st_context.h"
  47 #include "st_cache.h"
  48 #include "st_atom.h"
  49 #include "st_program.h"
  50 #include "st_atom_shader.h"
  51 #include "st_mesa_to_tgsi.h"
  52
  53
  54 /**
  55  * This represents a vertex program, especially translated to match
  56  * the inputs of a particular fragment shader.
  57  */
  58 struct translated_vertex_program
  59 {
  60    struct st_vertex_program *master;
  61
  62    /** The fragment shader "signature" this vertex shader is meant for: */
  63    GLbitfield frag_inputs;
  64
  65    /** Compared against master vertex program's serialNo: */
  66    GLuint serialNo;
  67
  68    /** Maps VERT_RESULT_x to slot */
  69    GLuint output_to_slot[VERT_RESULT_MAX];
  70
  71    /** The program in TGSI format */
  72    struct tgsi_token tokens[ST_MAX_SHADER_TOKENS];
  73
  74    /** Pointer to the translated, cached vertex shader */
  75    const struct cso_vertex_shader *vs;
  76
  77    struct translated_vertex_program *next;  /**< next in linked list */
  78 };
  79
  80
  81
  82 /**
  83  * Free data hanging off the st vert prog.
  84  */
  85 void
  86 st_remove_vertex_program(struct st_context *st, struct st_vertex_program *stvp)
  87 {
  88    /* no-op, for now? */
  89 }
  90
  91
  92 /**
  93  * Free data hanging off the st frag prog.
  94  */
  95 void
  96 st_remove_fragment_program(struct st_context *st,
  97                           struct st_fragment_program *stfp)
  98 {
  99    struct translated_vertex_program *xvp, *next;
 100
 101    for (xvp = stfp->vertex_programs; xvp; xvp = next) {
 102       next = xvp->next;
 103       /* XXX free xvp->vs */
 104       free(xvp);
 105    }
 106 }
 107
 108
 109
 110 /**
 111  * Given a vertex program output attribute, return the corresponding
 112  * fragment program input attribute.
 113  * \return -1 for vertex outputs that have no corresponding fragment input
 114  */
 115 static GLint
 116 vp_out_to_fp_in(GLuint vertResult)
 117 {
 118    if (vertResult >= VERT_RESULT_TEX0 &&
 119        vertResult < VERT_RESULT_TEX0 + MAX_TEXTURE_COORD_UNITS)
 120       return FRAG_ATTRIB_TEX0 + (vertResult - VERT_RESULT_TEX0);
 121
 122    if (vertResult >= VERT_RESULT_VAR0 &&
 123        vertResult < VERT_RESULT_VAR0 + MAX_VARYING)
 124       return FRAG_ATTRIB_VAR0 + (vertResult - VERT_RESULT_VAR0);
 125
 126    switch (vertResult) {
 127    case VERT_RESULT_HPOS:
 128       return FRAG_ATTRIB_WPOS;
 129    case VERT_RESULT_COL0:
 130       return FRAG_ATTRIB_COL0;
 131    case VERT_RESULT_COL1:
 132       return FRAG_ATTRIB_COL1;
 133    case VERT_RESULT_FOGC:
 134       return FRAG_ATTRIB_FOGC;
 135    default:
 136       /* Back-face colors, edge flags, etc */
 137       return -1;
 138    }
 139 }
 140
 141
 142 /**
 143  * Find a translated vertex program that corresponds to stvp and
 144  * has outputs matched to stfp's inputs.
 145  * This performs vertex and fragment translation (to TGSI) when needed.
 146  */
 147 static struct translated_vertex_program *
 148 find_translated_vp(struct st_context *st,
 149                    struct st_vertex_program *stvp,
 150                    struct st_fragment_program *stfp)
 151 {
 152    static const GLuint UNUSED = ~0;
 153    struct translated_vertex_program *xvp;
 154    const GLbitfield fragInputsRead
 155       = stfp->Base.Base.InputsRead | FRAG_BIT_WPOS;
 156
 157    /*
 158     * Translate fragment program if needed.
 159     */
 160    if (!stfp->fs) {
 161       GLuint inAttr, numIn = 0;
 162
 163       for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) {
 164          if (fragInputsRead & (1 << inAttr)) {
 165             stfp->input_to_slot[inAttr] = numIn;
 166             numIn++;
 167          }
 168          else {
 169             stfp->input_to_slot[inAttr] = UNUSED;
 170          }
 171       }
 172
 173       stfp->num_input_slots = numIn;
 174
 175       assert(stfp->Base.Base.NumInstructions > 1);
 176
 177       (void) st_translate_fragment_program(st, stfp,
 178                                            stfp->input_to_slot,
 179                                            stfp->tokens,
 180                                            ST_MAX_SHADER_TOKENS);
 181       assert(stfp->fs);
 182    }
 183
 184
 185    /* See if we've got a translated vertex program whose outputs match
 186     * the fragment program's inputs.
 187     * XXX This could be a hash lookup, using InputsRead as the key.
 188     */
 189    for (xvp = stfp->vertex_programs; xvp; xvp = xvp->next) {
 190       if (xvp->master == stvp && xvp->frag_inputs == fragInputsRead) {
 191          break;
 192       }
 193    }
 194
 195    /* No?  Allocate translated vp object now */
 196    if (!xvp) {
 197       xvp = CALLOC_STRUCT(translated_vertex_program);
 198       xvp->frag_inputs = fragInputsRead;
 199       xvp->master = stvp;
 200
 201       xvp->next = stfp->vertex_programs;
 202       stfp->vertex_programs = xvp;
 203    }
 204
 205    /* See if we need to translate vertex program to TGSI form */
 206    if (xvp->serialNo != stvp->serialNo) {
 207       GLuint outAttr, dummySlot;
 208       const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten;
 209
 210       /* Compute mapping of vertex program outputs to slots, which depends
 211        * on the fragment program's input->slot mapping.
 212        */
 213       for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
 214          /* set default: */
 215          xvp->output_to_slot[outAttr] = UNUSED;
 216
 217          if (outputsWritten & (1 << outAttr)) {
 218             /* see if the frag prog wants this vert output */
 219             GLint fpIn = vp_out_to_fp_in(outAttr);
 220             if (fpIn >= 0) {
 221                xvp->output_to_slot[outAttr] = stfp->input_to_slot[fpIn];
 222             }
 223          }
 224       }
 225
 226       /* Unneeded vertex program outputs will go to this slot.
 227        * We could use this info to do dead code elimination in the
 228        * vertex program.
 229        */
 230       dummySlot = stfp->num_input_slots;
 231
 232       /* Map vert program outputs that aren't used to the dummy slot */
 233       for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
 234          if (outputsWritten & (1 << outAttr)) {
 235             if (xvp->output_to_slot[outAttr] == UNUSED)
 236                xvp->output_to_slot[outAttr] = dummySlot;
 237          }
 238       }
 239
 240       assert(stvp->Base.Base.NumInstructions > 1);
 241
 242       xvp->vs = st_translate_vertex_program(st, stvp,
 243                                             xvp->output_to_slot,
 244                                             xvp->tokens,
 245                                             ST_MAX_SHADER_TOKENS);
 246       assert(xvp->vs);
 247       stvp->vs = NULL; /* don't want to use this */
 248
 249       /* translated VP is up to date now */
 250       xvp->serialNo = stvp->serialNo;
 251    }
 252
 253    return xvp;
 254 }
 255
 256
 257 static void
 258 update_linkage( struct st_context *st )
 259 {
 260    struct st_vertex_program *stvp;
 261    struct st_fragment_program *stfp;
 262    struct translated_vertex_program *xvp;
 263
 264    /* find active shader and params -- Should be covered by
 265     * ST_NEW_VERTEX_PROGRAM
 266     */
 267    assert(st->ctx->VertexProgram._Current);
 268    stvp = st_vertex_program(st->ctx->VertexProgram._Current);
 269
 270    assert(st->ctx->FragmentProgram._Current);
 271    stfp = st_fragment_program(st->ctx->FragmentProgram._Current);
 272
 273    xvp = find_translated_vp(st, stvp, stfp);
 274
 275    st->vp = stvp;
 276    st->state.vs = xvp->vs;
 277    st->pipe->bind_vs_state(st->pipe, st->state.vs->data);
 278
 279    st->fp = stfp;
 280    st->state.fs = stfp->fs;
 281    st->pipe->bind_fs_state(st->pipe, st->state.fs->data);
 282
 283    st->vertex_result_to_slot = xvp->output_to_slot;
 284 }
 285
 286
 287 const struct st_tracked_state st_update_shader = {
 288    .name = "st_update_shader",
 289    .dirty = {
 290       .mesa  = 0,
 291       .st   = ST_NEW_VERTEX_PROGRAM | ST_NEW_FRAGMENT_PROGRAM
 292    },
 293    .update = update_linkage
 294 };