src/gallium/drivers/softpipe/sp_quad_fs.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2007 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 /* Vertices are just an array of floats, with all the attributes
  29  * packed.  We currently assume a layout like:
  30  *
  31  * attr[0][0..3] - window position
  32  * attr[1..n][0..3] - remaining attributes.
  33  *
  34  * Attributes are assumed to be 4 floats wide but are packed so that
  35  * all the enabled attributes run contiguously.
  36  */
  37
  38 #include "pipe/p_util.h"
  39 #include "pipe/p_defines.h"
  40 #include "pipe/p_shader_tokens.h"
  41
  42 #include "sp_context.h"
  43 #include "sp_state.h"
  44 #include "sp_headers.h"
  45 #include "sp_quad.h"
  46 #include "sp_texture.h"
  47 #include "sp_tex_sample.h"
  48
  49
  50 struct quad_shade_stage
  51 {
  52    struct quad_stage stage;
  53    struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS];
  54    struct tgsi_exec_machine machine;
  55    struct tgsi_exec_vector *inputs, *outputs;
  56    int colorOutSlot, depthOutSlot;
  57 };
  58
  59
  60 /** cast wrapper */
  61 static INLINE struct quad_shade_stage *
  62 quad_shade_stage(struct quad_stage *qs)
  63 {
  64    return (struct quad_shade_stage *) qs;
  65 }
  66
  67
  68
  69 /**
  70  * Execute fragment shader for the four fragments in the quad.
  71  */
  72 static void
  73 shade_quad(
  74    struct quad_stage *qs,
  75    struct quad_header *quad )
  76 {
  77    struct quad_shade_stage *qss = quad_shade_stage( qs );
  78    struct softpipe_context *softpipe = qs->softpipe;
  79    struct tgsi_exec_machine *machine = &qss->machine;
  80
  81    /* Consts do not require 16 byte alignment. */
  82    machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT];
  83
  84    machine->InterpCoefs = quad->coef;
  85
  86    /* run shader */
  87    quad->mask &= softpipe->fs->run( softpipe->fs,
  88                                     &qss->machine,
  89                                     quad );
  90
  91    /* store result color */
  92    if (qss->colorOutSlot >= 0) {
  93       /* XXX need to handle multiple color outputs someday */
  94       assert(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot]
  95              == TGSI_SEMANTIC_COLOR);
  96       memcpy(
  97              quad->outputs.color,
  98              &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0],
  99              sizeof( quad->outputs.color ) );
 100    }
 101
 102    /*
 103     * XXX the following code for updating quad->outputs.depth
 104     * isn't really needed if we did early z testing.
 105     */
 106
 107    /* store result Z */
 108    if (qss->depthOutSlot >= 0) {
 109       /* output[slot] is new Z */
 110       uint i;
 111       for (i = 0; i < 4; i++) {
 112          quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i];
 113       }
 114    }
 115    else {
 116       /* compute Z values now, as in the quad earlyz stage */
 117       /* XXX we should really only do this if the earlyz stage is not used */
 118       const float fx = (float) quad->x0;
 119       const float fy = (float) quad->y0;
 120       const float dzdx = quad->posCoef->dadx[2];
 121       const float dzdy = quad->posCoef->dady[2];
 122       const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy;
 123
 124       quad->outputs.depth[0] = z0;
 125       quad->outputs.depth[1] = z0 + dzdx;
 126       quad->outputs.depth[2] = z0 + dzdy;
 127       quad->outputs.depth[3] = z0 + dzdx + dzdy;
 128    }
 129
 130    /* shader may cull fragments */
 131    if( quad->mask ) {
 132       qs->next->run( qs->next, quad );
 133    }
 134 }
 135
 136 /**
 137  * Per-primitive (or per-begin?) setup
 138  */
 139 static void shade_begin(struct quad_stage *qs)
 140 {
 141    struct quad_shade_stage *qss = quad_shade_stage(qs);
 142    struct softpipe_context *softpipe = qs->softpipe;
 143    unsigned i;
 144    unsigned num = MAX2(softpipe->num_textures, softpipe->num_samplers);
 145
 146    /* set TGSI sampler state that varies */
 147    for (i = 0; i < num; i++) {
 148       qss->samplers[i].state = softpipe->sampler[i];
 149       qss->samplers[i].texture = softpipe->texture[i];
 150    }
 151
 152    /* find output slots for depth, color */
 153    qss->colorOutSlot = -1;
 154    qss->depthOutSlot = -1;
 155    for (i = 0; i < qss->stage.softpipe->fs->info.num_outputs; i++) {
 156       switch (qss->stage.softpipe->fs->info.output_semantic_name[i]) {
 157       case TGSI_SEMANTIC_POSITION:
 158          qss->depthOutSlot = i;
 159          break;
 160       case TGSI_SEMANTIC_COLOR:
 161          qss->colorOutSlot = i;
 162          break;
 163       }
 164    }
 165
 166    softpipe->fs->prepare( softpipe->fs,
 167                           &qss->machine,
 168                           qss->samplers );
 169
 170    qs->next->begin(qs->next);
 171 }
 172
 173
 174 static void shade_destroy(struct quad_stage *qs)
 175 {
 176    struct quad_shade_stage *qss = (struct quad_shade_stage *) qs;
 177
 178    tgsi_exec_machine_free_data(&qss->machine);
 179    FREE( qss->inputs );
 180    FREE( qss->outputs );
 181    FREE( qs );
 182 }
 183
 184
 185 struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe )
 186 {
 187    struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage);
 188    uint i;
 189
 190    /* allocate storage for program inputs/outputs, aligned to 16 bytes */
 191    qss->inputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->inputs) + 16);
 192    qss->outputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->outputs) + 16);
 193    qss->machine.Inputs = align16(qss->inputs);
 194    qss->machine.Outputs = align16(qss->outputs);
 195
 196    qss->stage.softpipe = softpipe;
 197    qss->stage.begin = shade_begin;
 198    qss->stage.run = shade_quad;
 199    qss->stage.destroy = shade_destroy;
 200
 201    /* set TGSI sampler state that's constant */
 202    for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
 203       assert(softpipe->tex_cache[i]);
 204       qss->samplers[i].get_samples = sp_get_samples;
 205       qss->samplers[i].pipe = &softpipe->pipe;
 206       qss->samplers[i].cache = softpipe->tex_cache[i];
 207    }
 208
 209    tgsi_exec_machine_init( &qss->machine );
 210
 211    return &qss->stage;
 212 }