src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.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  /*
  29   * Authors:
  30   *   Keith Whitwell <keith@tungstengraphics.com>
  31   */
  32
  33
  34 #include "pipe/p_util.h"
  35 #include "draw/draw_context.h"
  36 #include "draw/draw_private.h"
  37 #include "draw/draw_vbuf.h"
  38 #include "draw/draw_vertex.h"
  39 #include "draw/draw_pt.h"
  40 #include "draw/draw_vs.h"
  41
  42 #include "translate/translate.h"
  43
  44 struct fetch_shade_emit;
  45
  46
  47 /* Prototype fetch, shade, emit-hw-verts all in one go.
  48  */
  49 struct fetch_shade_emit {
  50    struct draw_pt_middle_end base;
  51    struct draw_context *draw;
  52
  53
  54    /* Temporaries:
  55     */
  56    const float *constants;
  57    unsigned pitch[PIPE_MAX_ATTRIBS];
  58    const ubyte *src[PIPE_MAX_ATTRIBS];
  59    unsigned prim;
  60
  61    struct draw_vs_varient_key key;
  62    struct draw_vs_varient *active;
  63
  64
  65    const struct vertex_info *vinfo;
  66 };
  67
  68
  69
  70
  71 static void fse_prepare( struct draw_pt_middle_end *middle,
  72                          unsigned prim,
  73                          unsigned opt )
  74 {
  75    struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
  76    struct draw_context *draw = fse->draw;
  77    unsigned num_vs_inputs = draw->vs.vertex_shader->info.num_inputs;
  78    const struct vertex_info *vinfo;
  79    unsigned i;
  80
  81
  82    if (!draw->render->set_primitive( draw->render,
  83                                      prim )) {
  84       assert(0);
  85       return;
  86    }
  87
  88    /* Must do this after set_primitive() above:
  89     */
  90    fse->vinfo = vinfo = draw->render->get_vertex_info(draw->render);
  91
  92
  93
  94    fse->key.output_stride = vinfo->size * 4;
  95    fse->key.nr_outputs = vinfo->num_attribs;
  96    fse->key.nr_inputs = num_vs_inputs;
  97
  98    fse->key.nr_elements = MAX2(fse->key.nr_outputs,     /* outputs - translate to hw format */
  99                                fse->key.nr_inputs);     /* inputs - fetch from api format */
 100
 101    fse->key.viewport = !draw->identity_viewport;
 102    fse->key.clip = !draw->bypass_clipping;
 103    fse->key.pad = 0;
 104
 105    memset(fse->key.element, 0,
 106           fse->key.nr_elements * sizeof(fse->key.element[0]));
 107
 108    for (i = 0; i < num_vs_inputs; i++) {
 109       const struct pipe_vertex_element *src = &draw->pt.vertex_element[i];
 110       fse->key.element[i].in.format = src->src_format;
 111
 112       /* Consider ignoring these, ie make generated programs
 113        * independent of this state:
 114        */
 115       fse->key.element[i].in.buffer = src->vertex_buffer_index;
 116       fse->key.element[i].in.offset = src->src_offset;
 117    }
 118
 119
 120    {
 121       unsigned dst_offset = 0;
 122
 123       for (i = 0; i < vinfo->num_attribs; i++) {
 124          unsigned emit_sz = 0;
 125
 126          switch (vinfo->emit[i]) {
 127          case EMIT_4F:
 128             emit_sz = 4 * sizeof(float);
 129             break;
 130          case EMIT_3F:
 131             emit_sz = 3 * sizeof(float);
 132             break;
 133          case EMIT_2F:
 134             emit_sz = 2 * sizeof(float);
 135             break;
 136          case EMIT_1F:
 137             emit_sz = 1 * sizeof(float);
 138             break;
 139          case EMIT_1F_PSIZE:
 140             emit_sz = 1 * sizeof(float);
 141             break;
 142          case EMIT_4UB:
 143             emit_sz = 4 * sizeof(ubyte);
 144             break;
 145          default:
 146             assert(0);
 147             break;
 148          }
 149
 150          /* The elements in the key correspond to vertex shader output
 151           * numbers, not to positions in the hw vertex description --
 152           * that's handled by the output_offset field.
 153           */
 154          fse->key.element[i].out.format = vinfo->emit[i];
 155          fse->key.element[i].out.vs_output = vinfo->src_index[i];
 156          fse->key.element[i].out.offset = dst_offset;
 157
 158          dst_offset += emit_sz;
 159          assert(fse->key.output_stride >= dst_offset);
 160       }
 161    }
 162
 163
 164    /* Would normally look up a vertex shader and peruse its list of
 165     * varients somehow.  We omitted that step and put all the
 166     * hardcoded "shaders" into an array.  We're just making the
 167     * assumption that this happens to be a matching shader...  ie
 168     * you're running isosurf, aren't you?
 169     */
 170    fse->active = draw_vs_lookup_varient( draw->vs.vertex_shader,
 171                                          &fse->key );
 172
 173    if (!fse->active) {
 174       assert(0);
 175       return ;
 176    }
 177
 178    /* Now set buffer pointers:
 179     */
 180    for (i = 0; i < num_vs_inputs; i++) {
 181       unsigned buf = draw->pt.vertex_element[i].vertex_buffer_index;
 182
 183       fse->active->set_input( fse->active,
 184                               i,
 185
 186                               ((const ubyte *) draw->pt.user.vbuffer[buf] +
 187                                draw->pt.vertex_buffer[buf].buffer_offset),
 188
 189                               draw->pt.vertex_buffer[buf].pitch );
 190    }
 191
 192    //return TRUE;
 193 }
 194
 195
 196
 197
 198
 199
 200
 201 static void fse_run_linear( struct draw_pt_middle_end *middle,
 202                             unsigned start,
 203                             unsigned count )
 204 {
 205    struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
 206    struct draw_context *draw = fse->draw;
 207    unsigned alloc_count = align(count, 4);
 208    char *hw_verts;
 209
 210    /* XXX: need to flush to get prim_vbuf.c to release its allocation??
 211     */
 212    draw_do_flush( draw, DRAW_FLUSH_BACKEND );
 213
 214    hw_verts = draw->render->allocate_vertices( draw->render,
 215                                                (ushort)fse->key.output_stride,
 216                                                (ushort)alloc_count );
 217
 218    if (!hw_verts) {
 219       assert(0);
 220       return;
 221    }
 222
 223    /* Single routine to fetch vertices, run shader and emit HW verts.
 224     * Clipping is done elsewhere -- either by the API or on hardware,
 225     * or for some other reason not required...
 226     */
 227    fse->active->run_linear( fse->active,
 228                             start, count,
 229                             hw_verts );
 230
 231    /* Draw arrays path to avoid re-emitting index list again and
 232     * again.
 233     */
 234    draw->render->draw_arrays( draw->render,
 235                               0,
 236                               count );
 237
 238    if (0) {
 239       unsigned i;
 240       for (i = 0; i < count; i++) {
 241          debug_printf("\n\n%s vertex %d: (stride %d, offset %d)\n", __FUNCTION__, i,
 242                       fse->key.output_stride,
 243                       fse->key.output_stride * i);
 244
 245          draw_dump_emitted_vertex( fse->vinfo,
 246                                    (const uint8_t *)hw_verts + fse->key.output_stride * i );
 247       }
 248    }
 249
 250
 251    draw->render->release_vertices( draw->render,
 252                                    hw_verts,
 253                                    fse->key.output_stride,
 254                                    count );
 255 }
 256
 257
 258 static void
 259 fse_run(struct draw_pt_middle_end *middle,
 260         const unsigned *fetch_elts,
 261         unsigned fetch_count,
 262         const ushort *draw_elts,
 263         unsigned draw_count )
 264 {
 265    struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
 266    struct draw_context *draw = fse->draw;
 267    unsigned alloc_count = align(fetch_count, 4);
 268    void *hw_verts;
 269
 270    /* XXX: need to flush to get prim_vbuf.c to release its allocation??
 271     */
 272    draw_do_flush( draw, DRAW_FLUSH_BACKEND );
 273
 274    hw_verts = draw->render->allocate_vertices( draw->render,
 275                                                (ushort)fse->key.output_stride,
 276                                                (ushort)alloc_count );
 277    if (!hw_verts) {
 278       assert(0);
 279       return;
 280    }
 281
 282
 283    /* Single routine to fetch vertices, run shader and emit HW verts.
 284     */
 285    fse->active->run_elts( fse->active,
 286                           fetch_elts,
 287                           fetch_count,
 288                           hw_verts );
 289
 290    draw->render->draw( draw->render,
 291                        draw_elts,
 292                        draw_count );
 293
 294    if (0) {
 295       unsigned i;
 296       for (i = 0; i < fetch_count; i++) {
 297          debug_printf("\n\n%s vertex %d:\n", __FUNCTION__, i);
 298          draw_dump_emitted_vertex( fse->vinfo,
 299                                    (const uint8_t *)hw_verts +
 300                                    fse->key.output_stride * i );
 301       }
 302    }
 303
 304
 305    draw->render->release_vertices( draw->render,
 306                                    hw_verts,
 307                                    fse->key.output_stride,
 308                                    fetch_count );
 309
 310 }
 311
 312
 313 static void fse_finish( struct draw_pt_middle_end *middle )
 314 {
 315 }
 316
 317
 318 static void
 319 fse_destroy( struct draw_pt_middle_end *middle )
 320 {
 321    FREE(middle);
 322 }
 323
 324 struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw )
 325 {
 326    struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit);
 327    if (!fse)
 328       return NULL;
 329
 330    fse->base.prepare = fse_prepare;
 331    fse->base.run = fse_run;
 332    fse->base.run_linear = fse_run_linear;
 333    fse->base.finish = fse_finish;
 334    fse->base.destroy = fse_destroy;
 335    fse->draw = draw;
 336
 337    return &fse->base;
 338 }