src/gallium/auxiliary/draw/draw_vs_varient.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 #include "util/u_memory.h"
  34 #include "util/u_math.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_vs.h"
  40 #include "translate/translate.h"
  41 #include "translate/translate_cache.h"
  42
  43 /* A first pass at incorporating vertex fetch/emit functionality into
  44  */
  45 struct draw_vs_varient_generic {
  46    struct draw_vs_varient base;
  47
  48    struct draw_vertex_shader *shader;
  49    struct draw_context *draw;
  50
  51    /* Basic plan is to run these two translate functions before/after
  52     * the vertex shader's existing run_linear() routine to simulate
  53     * the inclusion of this functionality into the shader...
  54     *
  55     * Next will look at actually including it.
  56     */
  57    struct translate *fetch;
  58    struct translate *emit;
  59
  60    unsigned temp_vertex_stride;
  61 };
  62
  63
  64
  65
  66
  67 static void vsvg_set_buffer( struct draw_vs_varient *varient,
  68                              unsigned buffer,
  69                              const void *ptr,
  70                              unsigned stride )
  71 {
  72    struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient;
  73
  74    vsvg->fetch->set_buffer(vsvg->fetch,
  75                            buffer,
  76                            ptr,
  77                            stride);
  78 }
  79
  80
  81 /* Mainly for debug at this stage:
  82  */
  83 static void do_rhw_viewport( struct draw_vs_varient_generic *vsvg,
  84                              unsigned count,
  85                              void *output_buffer )
  86 {
  87    char *ptr = (char *)output_buffer;
  88    const float *scale = vsvg->base.vs->draw->viewport.scale;
  89    const float *trans = vsvg->base.vs->draw->viewport.translate;
  90    unsigned stride = vsvg->temp_vertex_stride;
  91    unsigned j;
  92
  93    ptr += vsvg->base.vs->position_output * 4 * sizeof(float);
  94
  95    for (j = 0; j < count; j++, ptr += stride) {
  96       float *data = (float *)ptr;
  97       float w = 1.0f / data[3];
  98
  99       data[0] = data[0] * w * scale[0] + trans[0];
 100       data[1] = data[1] * w * scale[1] + trans[1];
 101       data[2] = data[2] * w * scale[2] + trans[2];
 102       data[3] = w;
 103    }
 104 }
 105
 106 static void do_viewport( struct draw_vs_varient_generic *vsvg,
 107                          unsigned count,
 108                          void *output_buffer )
 109 {
 110    char *ptr = (char *)output_buffer;
 111    const float *scale = vsvg->base.vs->draw->viewport.scale;
 112    const float *trans = vsvg->base.vs->draw->viewport.translate;
 113    unsigned stride = vsvg->temp_vertex_stride;
 114    unsigned j;
 115
 116    ptr += vsvg->base.vs->position_output * 4 * sizeof(float);
 117
 118    for (j = 0; j < count; j++, ptr += stride) {
 119       float *data = (float *)ptr;
 120
 121       data[0] = data[0] * scale[0] + trans[0];
 122       data[1] = data[1] * scale[1] + trans[1];
 123       data[2] = data[2] * scale[2] + trans[2];
 124    }
 125 }
 126
 127
 128 static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient,
 129                                       const unsigned *elts,
 130                                       unsigned count,
 131                                       void *output_buffer)
 132 {
 133    struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient;
 134    unsigned temp_vertex_stride = vsvg->temp_vertex_stride;
 135    void *temp_buffer = MALLOC( align(count,4) * temp_vertex_stride );
 136
 137    if (0) debug_printf("%s %d \n", __FUNCTION__,  count);
 138
 139    /* Want to do this in small batches for cache locality?
 140     */
 141
 142    vsvg->fetch->run_elts( vsvg->fetch,
 143                           elts,
 144                           count,
 145                           vsvg->draw->instance_id,
 146                           temp_buffer );
 147
 148    vsvg->base.vs->run_linear( vsvg->base.vs,
 149                               temp_buffer,
 150                               temp_buffer,
 151                               (const float (*)[4])vsvg->base.vs->draw->pt.user.vs_constants,
 152                               count,
 153                               temp_vertex_stride,
 154                               temp_vertex_stride);
 155
 156    /* FIXME: geometry shading? */
 157
 158    if (vsvg->base.key.clip) {
 159       /* not really handling clipping, just do the rhw so we can
 160        * see the results...
 161        */
 162       do_rhw_viewport( vsvg,
 163                        count,
 164                        temp_buffer );
 165    }
 166    else if (vsvg->base.key.viewport) {
 167       do_viewport( vsvg,
 168                    count,
 169                    temp_buffer );
 170    }
 171
 172
 173    vsvg->emit->set_buffer( vsvg->emit,
 174                            0,
 175                            temp_buffer,
 176                            temp_vertex_stride );
 177
 178    vsvg->emit->set_buffer( vsvg->emit,
 179                            1,
 180                            &vsvg->draw->rasterizer->point_size,
 181                            0);
 182
 183    vsvg->emit->run( vsvg->emit,
 184                     0, count,
 185                     vsvg->draw->instance_id,
 186                     output_buffer );
 187
 188    FREE(temp_buffer);
 189 }
 190
 191
 192 static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
 193                                         unsigned start,
 194                                         unsigned count,
 195                                         void *output_buffer )
 196 {
 197    struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient;
 198    unsigned temp_vertex_stride = vsvg->temp_vertex_stride;
 199    void *temp_buffer = MALLOC( align(count,4) * temp_vertex_stride );
 200
 201    if (0) debug_printf("%s %d %d (sz %d, %d)\n", __FUNCTION__, start, count,
 202                        vsvg->base.key.output_stride,
 203                        temp_vertex_stride);
 204
 205    vsvg->fetch->run( vsvg->fetch,
 206                      start,
 207                      count,
 208                      vsvg->draw->instance_id,
 209                      temp_buffer );
 210
 211    vsvg->base.vs->run_linear( vsvg->base.vs,
 212                               temp_buffer,
 213                               temp_buffer,
 214                               (const float (*)[4])vsvg->base.vs->draw->pt.user.vs_constants,
 215                               count,
 216                               temp_vertex_stride,
 217                               temp_vertex_stride);
 218
 219    if (vsvg->base.key.clip) {
 220       /* not really handling clipping, just do the rhw so we can
 221        * see the results...
 222        */
 223       do_rhw_viewport( vsvg,
 224                        count,
 225                        temp_buffer );
 226    }
 227    else if (vsvg->base.key.viewport) {
 228       do_viewport( vsvg,
 229                    count,
 230                    temp_buffer );
 231    }
 232
 233    vsvg->emit->set_buffer( vsvg->emit,
 234                            0,
 235                            temp_buffer,
 236                            temp_vertex_stride );
 237
 238    vsvg->emit->set_buffer( vsvg->emit,
 239                            1,
 240                            &vsvg->draw->rasterizer->point_size,
 241                            0);
 242
 243    vsvg->emit->run( vsvg->emit,
 244                     0, count,
 245                     vsvg->draw->instance_id,
 246                     output_buffer );
 247
 248    FREE(temp_buffer);
 249 }
 250
 251
 252
 253
 254
 255 static void vsvg_destroy( struct draw_vs_varient *varient )
 256 {
 257    FREE(varient);
 258 }
 259
 260
 261 struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs,
 262                                                  const struct draw_vs_varient_key *key )
 263 {
 264    unsigned i;
 265    struct translate_key fetch, emit;
 266
 267    struct draw_vs_varient_generic *vsvg = CALLOC_STRUCT( draw_vs_varient_generic );
 268    if (vsvg == NULL)
 269       return NULL;
 270
 271    vsvg->base.key = *key;
 272    vsvg->base.vs = vs;
 273    vsvg->base.set_buffer    = vsvg_set_buffer;
 274    vsvg->base.run_elts      = vsvg_run_elts;
 275    vsvg->base.run_linear    = vsvg_run_linear;
 276    vsvg->base.destroy       = vsvg_destroy;
 277
 278    vsvg->draw = vs->draw;
 279
 280    vsvg->temp_vertex_stride = MAX2(key->nr_inputs,
 281                                    vsvg->base.vs->info.num_outputs) * 4 * sizeof(float);
 282
 283    /* Build free-standing fetch and emit functions:
 284     */
 285    fetch.nr_elements = key->nr_inputs;
 286    fetch.output_stride = vsvg->temp_vertex_stride;
 287    for (i = 0; i < key->nr_inputs; i++) {
 288       fetch.element[i].type = TRANSLATE_ELEMENT_NORMAL;
 289       fetch.element[i].input_format = key->element[i].in.format;
 290       fetch.element[i].input_buffer = key->element[i].in.buffer;
 291       fetch.element[i].input_offset = key->element[i].in.offset;
 292       fetch.element[i].instance_divisor = 0;
 293       fetch.element[i].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
 294       fetch.element[i].output_offset = i * 4 * sizeof(float);
 295       assert(fetch.element[i].output_offset < fetch.output_stride);
 296    }
 297
 298
 299    emit.nr_elements = key->nr_outputs;
 300    emit.output_stride = key->output_stride;
 301    for (i = 0; i < key->nr_outputs; i++) {
 302       if (key->element[i].out.format != EMIT_1F_PSIZE)
 303       {
 304          emit.element[i].type = TRANSLATE_ELEMENT_NORMAL;
 305          emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
 306          emit.element[i].input_buffer = 0;
 307          emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float);
 308          emit.element[i].instance_divisor = 0;
 309          emit.element[i].output_format = draw_translate_vinfo_format(key->element[i].out.format);
 310          emit.element[i].output_offset = key->element[i].out.offset;
 311          assert(emit.element[i].input_offset <= fetch.output_stride);
 312       }
 313       else {
 314          emit.element[i].type = TRANSLATE_ELEMENT_NORMAL;
 315          emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT;
 316          emit.element[i].input_buffer = 1;
 317          emit.element[i].input_offset = 0;
 318          emit.element[i].instance_divisor = 0;
 319          emit.element[i].output_format = PIPE_FORMAT_R32_FLOAT;
 320          emit.element[i].output_offset = key->element[i].out.offset;
 321       }
 322    }
 323
 324    vsvg->fetch = draw_vs_get_fetch( vs->draw, &fetch );
 325    vsvg->emit = draw_vs_get_emit( vs->draw, &emit );
 326
 327    return &vsvg->base;
 328 }
 329
 330
 331
 332
 333