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                           temp_buffer );
 146
 147    vsvg->base.vs->run_linear( vsvg->base.vs,
 148                               temp_buffer,
 149                               temp_buffer,
 150                               (const float (*)[4])vsvg->base.vs->draw->pt.user.constants,
 151                               count,
 152                               temp_vertex_stride,
 153                               temp_vertex_stride);
 154
 155
 156    if (vsvg->base.key.clip) {
 157       /* not really handling clipping, just do the rhw so we can
 158        * see the results...
 159        */
 160       do_rhw_viewport( vsvg,
 161                        count,
 162                        temp_buffer );
 163    }
 164    else if (vsvg->base.key.viewport) {
 165       do_viewport( vsvg,
 166                    count,
 167                    temp_buffer );
 168    }
 169
 170
 171    vsvg->emit->set_buffer( vsvg->emit,
 172                            0,
 173                            temp_buffer,
 174                            temp_vertex_stride );
 175
 176    vsvg->emit->set_buffer( vsvg->emit,
 177                            1,
 178                            &vsvg->draw->rasterizer->point_size,
 179                            0);
 180
 181    vsvg->emit->run( vsvg->emit,
 182                     0, count,
 183                     output_buffer );
 184
 185    FREE(temp_buffer);
 186 }
 187
 188
 189 static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
 190                                         unsigned start,
 191                                         unsigned count,
 192                                         void *output_buffer )
 193 {
 194    struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient;
 195    unsigned temp_vertex_stride = vsvg->temp_vertex_stride;
 196    void *temp_buffer = MALLOC( align(count,4) * temp_vertex_stride );
 197
 198    if (0) debug_printf("%s %d %d (sz %d, %d)\n", __FUNCTION__, start, count,
 199                        vsvg->base.key.output_stride,
 200                        temp_vertex_stride);
 201
 202    vsvg->fetch->run( vsvg->fetch,
 203                      start,
 204                      count,
 205                      temp_buffer );
 206
 207    vsvg->base.vs->run_linear( vsvg->base.vs,
 208                               temp_buffer,
 209                               temp_buffer,
 210                               (const float (*)[4])vsvg->base.vs->draw->pt.user.constants,
 211                               count,
 212                               temp_vertex_stride,
 213                               temp_vertex_stride);
 214
 215    if (vsvg->base.key.clip) {
 216       /* not really handling clipping, just do the rhw so we can
 217        * see the results...
 218        */
 219       do_rhw_viewport( vsvg,
 220                        count,
 221                        temp_buffer );
 222    }
 223    else if (vsvg->base.key.viewport) {
 224       do_viewport( vsvg,
 225                    count,
 226                    temp_buffer );
 227    }
 228
 229    vsvg->emit->set_buffer( vsvg->emit,
 230                            0,
 231                            temp_buffer,
 232                            temp_vertex_stride );
 233
 234    vsvg->emit->set_buffer( vsvg->emit,
 235                            1,
 236                            &vsvg->draw->rasterizer->point_size,
 237                            0);
 238
 239    vsvg->emit->run( vsvg->emit,
 240                     0, count,
 241                     output_buffer );
 242
 243    FREE(temp_buffer);
 244 }
 245
 246
 247
 248
 249
 250 static void vsvg_destroy( struct draw_vs_varient *varient )
 251 {
 252    FREE(varient);
 253 }
 254
 255
 256 struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs,
 257                                                  const struct draw_vs_varient_key *key )
 258 {
 259    unsigned i;
 260    struct translate_key fetch, emit;
 261
 262    struct draw_vs_varient_generic *vsvg = CALLOC_STRUCT( draw_vs_varient_generic );
 263    if (vsvg == NULL)
 264       return NULL;
 265
 266    vsvg->base.key = *key;
 267    vsvg->base.vs = vs;
 268    vsvg->base.set_buffer    = vsvg_set_buffer;
 269    vsvg->base.run_elts      = vsvg_run_elts;
 270    vsvg->base.run_linear    = vsvg_run_linear;
 271    vsvg->base.destroy       = vsvg_destroy;
 272
 273    vsvg->draw = vs->draw;
 274
 275    vsvg->temp_vertex_stride = MAX2(key->nr_inputs,
 276                                    vsvg->base.vs->info.num_outputs) * 4 * sizeof(float);
 277
 278    /* Build free-standing fetch and emit functions:
 279     */
 280    fetch.nr_elements = key->nr_inputs;
 281    fetch.output_stride = vsvg->temp_vertex_stride;
 282    for (i = 0; i < key->nr_inputs; i++) {
 283       fetch.element[i].input_format = key->element[i].in.format;
 284       fetch.element[i].input_buffer = key->element[i].in.buffer;
 285       fetch.element[i].input_offset = key->element[i].in.offset;
 286       fetch.element[i].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
 287       fetch.element[i].output_offset = i * 4 * sizeof(float);
 288       assert(fetch.element[i].output_offset < fetch.output_stride);
 289    }
 290
 291
 292    emit.nr_elements = key->nr_outputs;
 293    emit.output_stride = key->output_stride;
 294    for (i = 0; i < key->nr_outputs; i++) {
 295       if (key->element[i].out.format != EMIT_1F_PSIZE)
 296       {
 297          emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
 298          emit.element[i].input_buffer = 0;
 299          emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float);
 300          emit.element[i].output_format = draw_translate_vinfo_format(key->element[i].out.format);
 301          emit.element[i].output_offset = key->element[i].out.offset;
 302          assert(emit.element[i].input_offset <= fetch.output_stride);
 303       }
 304       else {
 305          emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT;
 306          emit.element[i].input_buffer = 1;
 307          emit.element[i].input_offset = 0;
 308          emit.element[i].output_format = PIPE_FORMAT_R32_FLOAT;
 309          emit.element[i].output_offset = key->element[i].out.offset;
 310       }
 311    }
 312
 313    vsvg->fetch = draw_vs_get_fetch( vs->draw, &fetch );
 314    vsvg->emit = draw_vs_get_emit( vs->draw, &emit );
 315
 316    return &vsvg->base;
 317 }
 318
 319
 320
 321
 322