src/mesa/pipe/softpipe/sp_quad_fs.c

   1 /*
   2  * Mesa 3-D graphics library
   3  * Version:  6.5
   4  *
   5  * Copyright (C) 1999-2005  Brian Paul   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  * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
  21  * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  22  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  23  */
  24
  25 /* Vertices are just an array of floats, with all the attributes
  26  * packed.  We currently assume a layout like:
  27  *
  28  * attr[0][0..3] - window position
  29  * attr[1..n][0..3] - remaining attributes.
  30  *
  31  * Attributes are assumed to be 4 floats wide but are packed so that
  32  * all the enabled attributes run contiguously.
  33  */
  34
  35 #include "glheader.h"
  36 #include "imports.h"
  37 #include "sp_context.h"
  38 #include "sp_headers.h"
  39 #include "sp_quad.h"
  40 #include "core/tgsi_core.h"
  41
  42 struct exec_machine {
  43    const struct setup_coefficient *coef; /**< will point to quad->coef */
  44
  45    GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE];
  46 };
  47
  48
  49 /**
  50  * Compute quad's attributes values, as constants (GL_FLAT shading).
  51  */
  52 static INLINE void cinterp( struct exec_machine *exec,
  53                             GLuint attrib,
  54                             GLuint i )
  55 {
  56    GLuint j;
  57
  58    for (j = 0; j < QUAD_SIZE; j++) {
  59       exec->attr[attrib][i][j] = exec->coef[attrib].a0[i];
  60    }
  61 }
  62
  63
  64 /**
  65  * Compute quad's attribute values by linear interpolation.
  66  *
  67  * Push into the fp:
  68  *
  69  *   INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx
  70  *   INPUT[attr] = MAD INPUT[attr],   COEF_DADY[attr], INPUT_WPOS.yyyy
  71  */
  72 static INLINE void linterp( struct exec_machine *exec,
  73                             GLuint attrib,
  74                             GLuint i )
  75 {
  76    GLuint j;
  77
  78    for (j = 0; j < QUAD_SIZE; j++) {
  79       const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j];
  80       const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j];
  81       exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] +
  82                                   exec->coef[attrib].dadx[i] * x +
  83                                   exec->coef[attrib].dady[i] * y);
  84    }
  85 }
  86
  87
  88 /**
  89  * Compute quad's attribute values by linear interpolation with
  90  * perspective correction.
  91  *
  92  * Push into the fp:
  93  *
  94  *   INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx
  95  *   INPUT[attr] = MAD INPUT[attr],   COEF_DADY[attr], INPUT_WPOS.yyyy
  96  *   INPUT[attr] = MUL INPUT[attr],   INPUT_WPOS.wwww
  97  *
  98  * (Or should that be 1/w ???)
  99  */
 100 static INLINE void pinterp( struct exec_machine *exec,
 101                             GLuint attrib,
 102                             GLuint i )
 103 {
 104    GLuint j;
 105
 106    for (j = 0; j < QUAD_SIZE; j++) {
 107       const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j];
 108       const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j];
 109       const GLfloat invW = exec->attr[FRAG_ATTRIB_WPOS][3][j];
 110       exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] +
 111                                    exec->coef[attrib].dadx[i] * x +
 112                                    exec->coef[attrib].dady[i] * y) * invW);
 113    }
 114 }
 115
 116
 117
 118 /* This should be done by the fragment shader execution unit (code
 119  * generated from the decl instructions).  Do it here for now.
 120  */
 121 static void
 122 shade_quad( struct quad_stage *qs, struct quad_header *quad )
 123 {
 124    const struct softpipe_context *softpipe = qs->softpipe;
 125    struct exec_machine exec;
 126    const GLfloat fx = quad->x0;
 127    const GLfloat fy = quad->y0;
 128    GLuint i, j;
 129
 130    exec.coef = quad->coef;
 131
 132    /* Position:
 133     */
 134    exec.attr[FRAG_ATTRIB_WPOS][0][0] = fx;
 135    exec.attr[FRAG_ATTRIB_WPOS][0][1] = fx + 1.0;
 136    exec.attr[FRAG_ATTRIB_WPOS][0][2] = fx;
 137    exec.attr[FRAG_ATTRIB_WPOS][0][3] = fx + 1.0;
 138
 139    exec.attr[FRAG_ATTRIB_WPOS][1][0] = fy;
 140    exec.attr[FRAG_ATTRIB_WPOS][1][1] = fy;
 141    exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0;
 142    exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0;
 143
 144    /* Z and W are done by linear interpolation */
 145    if (softpipe->need_z) {
 146       linterp(&exec, 0, 2);   /* attr[0].z */
 147    }
 148
 149    if (softpipe->need_w) {
 150       linterp(&exec, 0, 3);  /* attr[0].w */
 151       /*invert(&exec, 0, 3);*/
 152    }
 153
 154    /* Interpolate all the remaining attributes.  This will get pushed
 155     * into the fragment program's responsibilities at some point.
 156     * Start at 1 to skip fragment position attribute (computed above).
 157     */
 158    for (i = 1; i < quad->nr_attrs; i++) {
 159       switch (softpipe->interp[i]) {
 160       case INTERP_CONSTANT:
 161          for (j = 0; j < NUM_CHANNELS; j++)
 162             cinterp(&exec, i, j);
 163          break;
 164
 165       case INTERP_LINEAR:
 166          for (j = 0; j < NUM_CHANNELS; j++)
 167             linterp(&exec, i, j);
 168          break;
 169
 170       case INTERP_PERSPECTIVE:
 171          for (j = 0; j < NUM_CHANNELS; j++)
 172             pinterp(&exec, i, j);
 173          break;
 174       }
 175    }
 176
 177 #if 1
 178    /*softpipe->run_fs( tri->fp, quad, &tri->outputs );*/
 179
 180    {
 181       struct tgsi_exec_machine machine;
 182       struct tgsi_exec_vector inputs[FRAG_ATTRIB_MAX + 1];
 183       struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1];
 184       struct tgsi_exec_vector *ainputs;
 185       struct tgsi_exec_vector *aoutputs;
 186       GLuint i /*, total*/;
 187
 188 #ifdef DEBUG
 189       memset(&machine, 0, sizeof(machine));
 190 #endif
 191
 192       ainputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs );
 193       aoutputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs );
 194
 195 #if 0
 196       for( i = total = 0; i < PIPE_ATTRIB_MAX; i++ ) {
 197          GLuint attr;
 198
 199          attr = softpipe->fp_attr_to_slot[i];
 200          if( attr || total == 0) {
 201             assert( total < FRAG_ATTRIB_MAX );
 202             assert( attr < FRAG_ATTRIB_MAX );
 203             assert( sizeof( ainputs[0] ) == sizeof( exec.attr[0] ) );
 204
 205             memcpy(
 206                &ainputs[total],
 207                exec.attr[attr],
 208                sizeof( ainputs[0] ) );
 209             total++;
 210          }
 211       }
 212 #else
 213       /* load input registers */
 214       /* XXX simpler than above, but might not be right... */
 215       for (i = 0; i < softpipe->nr_attrs; i++) {
 216          memcpy(
 217                 &ainputs[i],
 218                 exec.attr[i],
 219                 sizeof( ainputs[0] ) );
 220       }
 221 #endif
 222
 223       /* init machine state */
 224       tgsi_exec_machine_init(
 225          &machine,
 226          softpipe->fs.tokens );
 227
 228       machine.Inputs = ainputs;
 229       machine.Outputs = aoutputs;
 230       machine.Consts = softpipe->fs.constants->constant; /* XXX alignment? */
 231
 232       /* run shader */
 233       tgsi_exec_machine_run( &machine );
 234
 235       /* store result color */
 236       memcpy(quad->outputs.color,
 237              &aoutputs[FRAG_ATTRIB_COL0].xyzw[0].f[0],
 238              sizeof(quad->outputs.color));
 239       if (softpipe->need_z) {
 240          /* XXX temporary */
 241          quad->outputs.depth[0] = exec.attr[0][2][0];
 242          quad->outputs.depth[1] = exec.attr[0][2][1];
 243          quad->outputs.depth[2] = exec.attr[0][2][2];
 244          quad->outputs.depth[3] = exec.attr[0][2][3];
 245       }
 246    }
 247 #else
 248    {
 249       GLuint attr = softpipe->fp_attr_to_slot[FRAG_ATTRIB_COL0];
 250       assert(attr);
 251
 252       memcpy(quad->outputs.color,
 253              exec.attr[attr],
 254              sizeof(quad->outputs.color));
 255
 256       if (softpipe->need_z) {
 257          quad->outputs.depth[0] = exec.attr[0][2][0];
 258          quad->outputs.depth[1] = exec.attr[0][2][1];
 259          quad->outputs.depth[2] = exec.attr[0][2][2];
 260          quad->outputs.depth[3] = exec.attr[0][2][3];
 261       }
 262    }
 263 #endif
 264
 265    /* shader may cull fragments */
 266    if (quad->mask)
 267       qs->next->run(qs->next, quad);
 268 }
 269
 270
 271
 272 struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe )
 273 {
 274    struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
 275
 276    stage->softpipe = softpipe;
 277    stage->run = shade_quad;
 278
 279    return stage;
 280 }