src/gallium/drivers/llvmpipe/lp_quad_fs.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2008-2009 VMware, Inc.
   4  * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
   5  * 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
   9  * "Software"), to deal in the Software without restriction, including
  10  * without limitation the rights to use, copy, modify, merge, publish,
  11  * distribute, sub license, and/or sell copies of the Software, and to
  12  * permit persons to whom the Software is furnished to do so, subject to
  13  * the following conditions:
  14  *
  15  * The above copyright notice and this permission notice (including the
  16  * next paragraph) shall be included in all copies or substantial portions
  17  * of the Software.
  18  *
  19  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  20  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  21  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
  22  * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
  23  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  24  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  25  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  26  *
  27  **************************************************************************/
  28
  29 /* Vertices are just an array of floats, with all the attributes
  30  * packed.  We currently assume a layout like:
  31  *
  32  * attr[0][0..3] - window position
  33  * attr[1..n][0..3] - remaining attributes.
  34  *
  35  * Attributes are assumed to be 4 floats wide but are packed so that
  36  * all the enabled attributes run contiguously.
  37  */
  38
  39 #include "util/u_math.h"
  40 #include "util/u_memory.h"
  41 #include "pipe/p_defines.h"
  42 #include "pipe/p_shader_tokens.h"
  43
  44 #include "lp_context.h"
  45 #include "lp_state.h"
  46 #include "lp_quad.h"
  47 #include "lp_quad_pipe.h"
  48 #include "lp_texture.h"
  49 #include "lp_tex_sample.h"
  50
  51
  52 struct quad_shade_stage
  53 {
  54    struct quad_stage stage;  /**< base class */
  55
  56    union tgsi_exec_channel ALIGN16_ATTRIB pos[NUM_CHANNELS];
  57    float ALIGN16_ATTRIB a0[PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
  58    float ALIGN16_ATTRIB dadx[PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
  59    float ALIGN16_ATTRIB dady[PIPE_MAX_SHADER_INPUTS][NUM_CHANNELS];
  60
  61    struct tgsi_exec_vector ALIGN16_ATTRIB outputs[PIPE_MAX_ATTRIBS];
  62 };
  63
  64
  65 /** cast wrapper */
  66 static INLINE struct quad_shade_stage *
  67 quad_shade_stage(struct quad_stage *qs)
  68 {
  69    return (struct quad_shade_stage *) qs;
  70 }
  71
  72
  73 static void
  74 setup_pos_vector(struct quad_shade_stage *qss,
  75                  const struct tgsi_interp_coef *coef,
  76                  float x, float y)
  77 {
  78    uint chan;
  79
  80    /* do X */
  81    qss->pos[0].f[0] = x;
  82    qss->pos[0].f[1] = x + 1;
  83    qss->pos[0].f[2] = x;
  84    qss->pos[0].f[3] = x + 1;
  85
  86    /* do Y */
  87    qss->pos[1].f[0] = y;
  88    qss->pos[1].f[1] = y;
  89    qss->pos[1].f[2] = y + 1;
  90    qss->pos[1].f[3] = y + 1;
  91
  92    /* do Z and W for all fragments in the quad */
  93    for (chan = 2; chan < 4; chan++) {
  94       const float dadx = coef->dadx[chan];
  95       const float dady = coef->dady[chan];
  96       const float a0 = coef->a0[chan] + dadx * x + dady * y;
  97       qss->pos[chan].f[0] = a0;
  98       qss->pos[chan].f[1] = a0 + dadx;
  99       qss->pos[chan].f[2] = a0 + dady;
 100       qss->pos[chan].f[3] = a0 + dadx + dady;
 101    }
 102 }
 103
 104
 105 static void
 106 setup_coef_vector(struct quad_shade_stage *qss,
 107                   const struct tgsi_interp_coef *coef)
 108 {
 109    unsigned num_inputs = qss->stage.llvmpipe->fs->info.num_inputs;
 110    unsigned attrib, chan, i;
 111
 112    for (attrib = 0; attrib < num_inputs; ++attrib) {
 113       for (chan = 0; chan < NUM_CHANNELS; ++chan) {
 114          qss->a0[attrib][chan] = coef[attrib].a0[chan];
 115          qss->dadx[attrib][chan] = coef[attrib].dadx[chan];
 116          qss->dady[attrib][chan] = coef[attrib].dady[chan];
 117       }
 118    }
 119 }
 120
 121
 122 /**
 123  * Execute fragment shader for the four fragments in the quad.
 124  */
 125 static boolean
 126 shade_quad(struct quad_stage *qs, struct quad_header *quad)
 127 {
 128    struct quad_shade_stage *qss = quad_shade_stage( qs );
 129    struct llvmpipe_context *llvmpipe = qs->llvmpipe;
 130    void *constants;
 131    struct tgsi_sampler **samplers;
 132    boolean z_written;
 133
 134    /* Compute X, Y, Z, W vals for this quad */
 135    setup_pos_vector(qss,
 136                     quad->posCoef,
 137                     (float)quad->input.x0, (float)quad->input.y0);
 138
 139
 140    constants = llvmpipe->mapped_constants[PIPE_SHADER_FRAGMENT];
 141    samplers = (struct tgsi_sampler **)llvmpipe->tgsi.frag_samplers_list;
 142
 143    /* run shader */
 144    llvmpipe->fs->jit_function( qss->pos,
 145                                qss->a0, qss->dadx, qss->dady,
 146                                constants,
 147                                qss->outputs,
 148                                samplers);
 149
 150    /* FIXME */
 151 #if 0
 152    quad->inout.mask &= ... ;
 153    if (quad->inout.mask == 0)
 154       return FALSE;
 155 #endif
 156
 157    /* store outputs */
 158    z_written = FALSE;
 159    {
 160       const ubyte *sem_name = llvmpipe->fs->info.output_semantic_name;
 161       const ubyte *sem_index = llvmpipe->fs->info.output_semantic_index;
 162       const uint n = qss->stage.llvmpipe->fs->info.num_outputs;
 163       uint i;
 164       for (i = 0; i < n; i++) {
 165          switch (sem_name[i]) {
 166          case TGSI_SEMANTIC_COLOR:
 167             {
 168                uint cbuf = sem_index[i];
 169                memcpy(quad->output.color[cbuf],
 170                       &qss->outputs[i].xyzw[0].f[0],
 171                       sizeof(quad->output.color[0]) );
 172             }
 173             break;
 174          case TGSI_SEMANTIC_POSITION:
 175             {
 176                uint j;
 177                for (j = 0; j < 4; j++) {
 178                   quad->output.depth[j] = qss->outputs[0].xyzw[2].f[j];
 179                }
 180                z_written = TRUE;
 181             }
 182             break;
 183          }
 184       }
 185    }
 186
 187    return TRUE;
 188 }
 189
 190
 191
 192 static void
 193 coverage_quad(struct quad_stage *qs, struct quad_header *quad)
 194 {
 195    struct llvmpipe_context *llvmpipe = qs->llvmpipe;
 196    uint cbuf;
 197
 198    /* loop over colorbuffer outputs */
 199    for (cbuf = 0; cbuf < llvmpipe->framebuffer.nr_cbufs; cbuf++) {
 200       float (*quadColor)[4] = quad->output.color[cbuf];
 201       unsigned j;
 202       for (j = 0; j < QUAD_SIZE; j++) {
 203          assert(quad->input.coverage[j] >= 0.0);
 204          assert(quad->input.coverage[j] <= 1.0);
 205          quadColor[3][j] *= quad->input.coverage[j];
 206       }
 207    }
 208 }
 209
 210
 211
 212 static void
 213 shade_quads(struct quad_stage *qs,
 214                  struct quad_header *quads[],
 215                  unsigned nr)
 216 {
 217    struct quad_shade_stage *qss = quad_shade_stage( qs );
 218    unsigned i, pass = 0;
 219
 220    setup_coef_vector(qss,
 221                      quads[0]->coef);
 222
 223    for (i = 0; i < nr; i++) {
 224       if (!shade_quad(qs, quads[i]))
 225          continue;
 226
 227       if (/*do_coverage*/ 0)
 228          coverage_quad( qs, quads[i] );
 229
 230       quads[pass++] = quads[i];
 231    }
 232
 233    if (pass)
 234       qs->next->run(qs->next, quads, pass);
 235 }
 236
 237
 238
 239
 240
 241 /**
 242  * Per-primitive (or per-begin?) setup
 243  */
 244 static void
 245 shade_begin(struct quad_stage *qs)
 246 {
 247    qs->next->begin(qs->next);
 248 }
 249
 250
 251 static void
 252 shade_destroy(struct quad_stage *qs)
 253 {
 254    FREE( qs );
 255 }
 256
 257
 258 struct quad_stage *
 259 lp_quad_shade_stage( struct llvmpipe_context *llvmpipe )
 260 {
 261    struct quad_shade_stage *qss;
 262
 263    qss = CALLOC_STRUCT(quad_shade_stage);
 264    if (!qss)
 265       return NULL;
 266
 267    qss->stage.llvmpipe = llvmpipe;
 268    qss->stage.begin = shade_begin;
 269    qss->stage.run = shade_quads;
 270    qss->stage.destroy = shade_destroy;
 271
 272    return &qss->stage;
 273 }