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
  58    struct tgsi_exec_vector ALIGN16_ATTRIB outputs[PIPE_MAX_ATTRIBS];
  59 };
  60
  61
  62 /** cast wrapper */
  63 static INLINE struct quad_shade_stage *
  64 quad_shade_stage(struct quad_stage *qs)
  65 {
  66    return (struct quad_shade_stage *) qs;
  67 }
  68
  69
  70 static void
  71 setup_pos_vector(struct quad_shade_stage *qss,
  72                  const struct tgsi_interp_coef *coef,
  73                  float x, float y)
  74 {
  75    uint chan;
  76
  77    /* do X */
  78    qss->pos[0].f[0] = x;
  79    qss->pos[0].f[1] = x + 1;
  80    qss->pos[0].f[2] = x;
  81    qss->pos[0].f[3] = x + 1;
  82
  83    /* do Y */
  84    qss->pos[1].f[0] = y;
  85    qss->pos[1].f[1] = y;
  86    qss->pos[1].f[2] = y + 1;
  87    qss->pos[1].f[3] = y + 1;
  88
  89    /* do Z and W for all fragments in the quad */
  90    for (chan = 2; chan < 4; chan++) {
  91       const float dadx = coef->dadx[chan];
  92       const float dady = coef->dady[chan];
  93       const float a0 = coef->a0[chan] + dadx * x + dady * y;
  94       qss->pos[chan].f[0] = a0;
  95       qss->pos[chan].f[1] = a0 + dadx;
  96       qss->pos[chan].f[2] = a0 + dady;
  97       qss->pos[chan].f[3] = a0 + dadx + dady;
  98    }
  99 }
 100
 101
 102 /**
 103  * Execute fragment shader for the four fragments in the quad.
 104  */
 105 static boolean
 106 shade_quad(struct quad_stage *qs, struct quad_header *quad)
 107 {
 108    struct quad_shade_stage *qss = quad_shade_stage( qs );
 109    struct llvmpipe_context *llvmpipe = qs->llvmpipe;
 110    void *constants;
 111    struct tgsi_sampler **samplers;
 112    boolean z_written;
 113
 114    /* Compute X, Y, Z, W vals for this quad */
 115    setup_pos_vector(qss,
 116                     quad->posCoef,
 117                     (float)quad->input.x0, (float)quad->input.y0);
 118
 119
 120    constants = llvmpipe->mapped_constants[PIPE_SHADER_FRAGMENT];
 121    samplers = (struct tgsi_sampler **)llvmpipe->tgsi.frag_samplers_list;
 122
 123    /* run shader */
 124    llvmpipe->fs->jit_function( qss->pos,
 125                                quad->coef->a0,
 126                                quad->coef->dadx,
 127                                quad->coef->dady,
 128                                constants,
 129                                qss->outputs,
 130                                samplers);
 131
 132    /* FIXME */
 133 #if 0
 134    quad->inout.mask &= ... ;
 135    if (quad->inout.mask == 0)
 136       return FALSE;
 137 #endif
 138
 139    /* store outputs */
 140    z_written = FALSE;
 141    {
 142       const ubyte *sem_name = llvmpipe->fs->info.output_semantic_name;
 143       const ubyte *sem_index = llvmpipe->fs->info.output_semantic_index;
 144       const uint n = qss->stage.llvmpipe->fs->info.num_outputs;
 145       uint i;
 146       for (i = 0; i < n; i++) {
 147          switch (sem_name[i]) {
 148          case TGSI_SEMANTIC_COLOR:
 149             {
 150                uint cbuf = sem_index[i];
 151                memcpy(quad->output.color[cbuf],
 152                       &qss->outputs[i].xyzw[0].f[0],
 153                       sizeof(quad->output.color[0]) );
 154             }
 155             break;
 156          case TGSI_SEMANTIC_POSITION:
 157             {
 158                uint j;
 159                for (j = 0; j < 4; j++) {
 160                   quad->output.depth[j] = qss->outputs[0].xyzw[2].f[j];
 161                }
 162                z_written = TRUE;
 163             }
 164             break;
 165          }
 166       }
 167    }
 168
 169    return TRUE;
 170 }
 171
 172
 173
 174 static void
 175 coverage_quad(struct quad_stage *qs, struct quad_header *quad)
 176 {
 177    struct llvmpipe_context *llvmpipe = qs->llvmpipe;
 178    uint cbuf;
 179
 180    /* loop over colorbuffer outputs */
 181    for (cbuf = 0; cbuf < llvmpipe->framebuffer.nr_cbufs; cbuf++) {
 182       float (*quadColor)[4] = quad->output.color[cbuf];
 183       unsigned j;
 184       for (j = 0; j < QUAD_SIZE; j++) {
 185          assert(quad->input.coverage[j] >= 0.0);
 186          assert(quad->input.coverage[j] <= 1.0);
 187          quadColor[3][j] *= quad->input.coverage[j];
 188       }
 189    }
 190 }
 191
 192
 193
 194 static void
 195 shade_quads(struct quad_stage *qs,
 196                  struct quad_header *quads[],
 197                  unsigned nr)
 198 {
 199    struct quad_shade_stage *qss = quad_shade_stage( qs );
 200    unsigned i, pass = 0;
 201
 202    for (i = 0; i < nr; i++) {
 203       if (!shade_quad(qs, quads[i]))
 204          continue;
 205
 206       if (/*do_coverage*/ 0)
 207          coverage_quad( qs, quads[i] );
 208
 209       quads[pass++] = quads[i];
 210    }
 211
 212    if (pass)
 213       qs->next->run(qs->next, quads, pass);
 214 }
 215
 216
 217
 218
 219
 220 /**
 221  * Per-primitive (or per-begin?) setup
 222  */
 223 static void
 224 shade_begin(struct quad_stage *qs)
 225 {
 226    qs->next->begin(qs->next);
 227 }
 228
 229
 230 static void
 231 shade_destroy(struct quad_stage *qs)
 232 {
 233    FREE( qs );
 234 }
 235
 236
 237 struct quad_stage *
 238 lp_quad_shade_stage( struct llvmpipe_context *llvmpipe )
 239 {
 240    struct quad_shade_stage *qss;
 241
 242    qss = CALLOC_STRUCT(quad_shade_stage);
 243    if (!qss)
 244       return NULL;
 245
 246    qss->stage.llvmpipe = llvmpipe;
 247    qss->stage.begin = shade_begin;
 248    qss->stage.run = shade_quads;
 249    qss->stage.destroy = shade_destroy;
 250
 251    return &qss->stage;
 252 }