src/gallium/auxiliary/vl/vl_vertex_buffers.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2010 Christian König
   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 #include <assert.h>
  29 #include <pipe/p_context.h>
  30 #include <pipe/p_screen.h>
  31 #include <util/u_memory.h>
  32 #include <util/u_inlines.h>
  33 #include <util/u_format.h>
  34 #include "vl_vertex_buffers.h"
  35 #include "vl_types.h"
  36
  37 struct vl_vertex_stream
  38 {
  39    struct vertex2s pos;
  40    struct {
  41       int8_t y;
  42       int8_t cr;
  43       int8_t cb;
  44       int8_t flag;
  45    } eb[2][2];
  46    struct vertex2s mv[4];
  47 };
  48
  49 /* vertices for a quad covering a block */
  50 static const struct vertex2f block_quad[4] = {
  51    {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f}
  52 };
  53
  54 struct pipe_vertex_buffer
  55 vl_vb_upload_quads(struct pipe_context *pipe, unsigned blocks_x, unsigned blocks_y)
  56 {
  57    struct pipe_vertex_buffer quad;
  58    struct pipe_transfer *buf_transfer;
  59    struct vertex4f *v;
  60
  61    unsigned x, y, i;
  62
  63    assert(pipe);
  64
  65    /* create buffer */
  66    quad.stride = sizeof(struct vertex4f);
  67    quad.buffer_offset = 0;
  68    quad.buffer = pipe_buffer_create
  69    (
  70       pipe->screen,
  71       PIPE_BIND_VERTEX_BUFFER,
  72       PIPE_USAGE_STATIC,
  73       sizeof(struct vertex4f) * 4 * blocks_x * blocks_y
  74    );
  75
  76    if(!quad.buffer)
  77       return quad;
  78
  79    /* and fill it */
  80    v = pipe_buffer_map
  81    (
  82       pipe,
  83       quad.buffer,
  84       PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
  85       &buf_transfer
  86    );
  87
  88    for ( y = 0; y < blocks_y; ++y) {
  89       for ( x = 0; x < blocks_x; ++x) {
  90          for (i = 0; i < 4; ++i, ++v) {
  91             v->x = block_quad[i].x;
  92             v->y = block_quad[i].y;
  93
  94             v->z = x;
  95             v->w = y;
  96          }
  97       }
  98    }
  99
 100    pipe_buffer_unmap(pipe, buf_transfer);
 101
 102    return quad;
 103 }
 104
 105 static struct pipe_vertex_element
 106 vl_vb_get_quad_vertex_element(void)
 107 {
 108    struct pipe_vertex_element element;
 109
 110    /* setup rectangle element */
 111    element.src_offset = 0;
 112    element.instance_divisor = 0;
 113    element.vertex_buffer_index = 0;
 114    element.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
 115
 116    return element;
 117 }
 118
 119 static void
 120 vl_vb_element_helper(struct pipe_vertex_element* elements, unsigned num_elements,
 121                               unsigned vertex_buffer_index)
 122 {
 123    unsigned i, offset = 0;
 124
 125    assert(elements && num_elements);
 126
 127    for ( i = 0; i < num_elements; ++i ) {
 128       elements[i].src_offset = offset;
 129       elements[i].instance_divisor = 1;
 130       elements[i].vertex_buffer_index = vertex_buffer_index;
 131       offset += util_format_get_blocksize(elements[i].src_format);
 132    }
 133 }
 134
 135 void *
 136 vl_vb_get_elems_state(struct pipe_context *pipe, bool include_mvs)
 137 {
 138    struct pipe_vertex_element vertex_elems[NUM_VS_INPUTS];
 139
 140    unsigned i;
 141
 142    memset(&vertex_elems, 0, sizeof(vertex_elems));
 143    vertex_elems[VS_I_RECT] = vl_vb_get_quad_vertex_element();
 144
 145    /* Position element */
 146    vertex_elems[VS_I_VPOS].src_format = PIPE_FORMAT_R16G16_SSCALED;
 147
 148    /* y, cr, cb empty block element top left block */
 149    vertex_elems[VS_I_EB_0_0].src_format = PIPE_FORMAT_R8G8B8A8_SSCALED;
 150
 151    /* y, cr, cb empty block element top right block */
 152    vertex_elems[VS_I_EB_0_1].src_format = PIPE_FORMAT_R8G8B8A8_SSCALED;
 153
 154    /* y, cr, cb empty block element bottom left block */
 155    vertex_elems[VS_I_EB_1_0].src_format = PIPE_FORMAT_R8G8B8A8_SSCALED;
 156
 157    /* y, cr, cb empty block element bottom right block */
 158    vertex_elems[VS_I_EB_1_1].src_format = PIPE_FORMAT_R8G8B8A8_SSCALED;
 159
 160    for (i = 0; i < 4; ++i)
 161       /* motion vector 0..4 element */
 162       vertex_elems[VS_I_MV0 + i].src_format = PIPE_FORMAT_R16G16_SSCALED;
 163
 164    vl_vb_element_helper(&vertex_elems[VS_I_VPOS], NUM_VS_INPUTS - (include_mvs ? 1 : 5), 1);
 165
 166    return pipe->create_vertex_elements_state(pipe, NUM_VS_INPUTS - (include_mvs ? 0 : 4), vertex_elems);
 167 }
 168
 169 struct pipe_vertex_buffer
 170 vl_vb_init(struct vl_vertex_buffer *buffer, struct pipe_context *pipe, unsigned size)
 171 {
 172    struct pipe_vertex_buffer buf;
 173
 174    assert(buffer);
 175
 176    buffer->size = size;
 177    buffer->num_not_empty = 0;
 178    buffer->num_empty = 0;
 179
 180    buf.stride = sizeof(struct vl_vertex_stream);
 181    buf.buffer_offset = 0;
 182    buf.buffer = pipe_buffer_create
 183    (
 184       pipe->screen,
 185       PIPE_BIND_VERTEX_BUFFER,
 186       PIPE_USAGE_STREAM,
 187       sizeof(struct vl_vertex_stream) * size
 188    );
 189
 190    pipe_resource_reference(&buffer->resource, buf.buffer);
 191
 192    vl_vb_map(buffer, pipe);
 193
 194    return buf;
 195 }
 196
 197 void
 198 vl_vb_map(struct vl_vertex_buffer *buffer, struct pipe_context *pipe)
 199 {
 200    assert(buffer && pipe);
 201
 202    buffer->start = pipe_buffer_map
 203    (
 204       pipe,
 205       buffer->resource,
 206       PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
 207       &buffer->transfer
 208    );
 209    buffer->end = buffer->start + buffer->resource->width0 / sizeof(struct vl_vertex_stream);
 210 }
 211
 212 static void
 213 get_motion_vectors(struct pipe_mpeg12_macroblock *mb, struct vertex2s mv[4])
 214 {
 215    switch (mb->mb_type) {
 216       case PIPE_MPEG12_MACROBLOCK_TYPE_BI:
 217       {
 218          if (mb->mo_type == PIPE_MPEG12_MOTION_TYPE_FRAME) {
 219             mv[2].x = mb->pmv[0][1][0];
 220             mv[2].y = mb->pmv[0][1][1];
 221
 222          } else {
 223             mv[2].x = mb->pmv[0][1][0];
 224             mv[2].y = mb->pmv[0][1][1] - (mb->pmv[0][1][1] % 4);
 225
 226             mv[3].x = mb->pmv[1][1][0];
 227             mv[3].y = mb->pmv[1][1][1] - (mb->pmv[1][1][1] % 4);
 228
 229             if(mb->mvfs[0][1]) mv[2].y += 2;
 230             if(!mb->mvfs[1][1]) mv[3].y -= 2;
 231          }
 232
 233          /* fall-through */
 234       }
 235       case PIPE_MPEG12_MACROBLOCK_TYPE_FWD:
 236       case PIPE_MPEG12_MACROBLOCK_TYPE_BKWD:
 237       {
 238          if (mb->mb_type == PIPE_MPEG12_MACROBLOCK_TYPE_BKWD) {
 239
 240             if (mb->mo_type == PIPE_MPEG12_MOTION_TYPE_FRAME) {
 241                mv[0].x = mb->pmv[0][1][0];
 242                mv[0].y = mb->pmv[0][1][1];
 243
 244             } else {
 245                mv[0].x = mb->pmv[0][1][0];
 246                mv[0].y = mb->pmv[0][1][1] - (mb->pmv[0][1][1] % 4);
 247
 248                mv[1].x = mb->pmv[1][1][0];
 249                mv[1].y = mb->pmv[1][1][1] - (mb->pmv[1][1][1] % 4);
 250
 251                if(mb->mvfs[0][1]) mv[0].y += 2;
 252                if(!mb->mvfs[1][1]) mv[1].y -= 2;
 253             }
 254
 255          } else {
 256
 257             if (mb->mo_type == PIPE_MPEG12_MOTION_TYPE_FRAME) {
 258                mv[0].x = mb->pmv[0][0][0];
 259                mv[0].y = mb->pmv[0][0][1];
 260
 261             } else {
 262                mv[0].x = mb->pmv[0][0][0];
 263                mv[0].y = mb->pmv[0][0][1] - (mb->pmv[0][0][1] % 4);
 264
 265                mv[1].x = mb->pmv[1][0][0];
 266                mv[1].y = mb->pmv[1][0][1] - (mb->pmv[1][0][1] % 4);
 267
 268                if(mb->mvfs[0][0]) mv[0].y += 2;
 269                if(!mb->mvfs[1][0]) mv[1].y -= 2;
 270             }
 271          }
 272       }
 273       default:
 274          break;
 275    }
 276 }
 277
 278 void
 279 vl_vb_add_block(struct vl_vertex_buffer *buffer, struct pipe_mpeg12_macroblock *mb,
 280                 const unsigned (*empty_block_mask)[3][2][2])
 281 {
 282    struct vl_vertex_stream *stream;
 283    unsigned i, j;
 284
 285    assert(buffer);
 286    assert(mb);
 287
 288    if(mb->cbp)
 289       stream = buffer->start + buffer->num_not_empty++;
 290    else
 291       stream = buffer->end - ++buffer->num_empty;
 292
 293    stream->pos.x = mb->mbx;
 294    stream->pos.y = mb->mby;
 295
 296    for ( i = 0; i < 2; ++i) {
 297       for ( j = 0; j < 2; ++j) {
 298          stream->eb[i][j].y = !(mb->cbp & (*empty_block_mask)[0][i][j]);
 299          stream->eb[i][j].cr = !(mb->cbp & (*empty_block_mask)[1][i][j]);
 300          stream->eb[i][j].cb = !(mb->cbp & (*empty_block_mask)[2][i][j]);
 301       }
 302    }
 303    stream->eb[0][0].flag = mb->dct_type == PIPE_MPEG12_DCT_TYPE_FIELD;
 304    stream->eb[0][1].flag = mb->mo_type == PIPE_MPEG12_MOTION_TYPE_FRAME;
 305    stream->eb[1][0].flag = mb->mb_type == PIPE_MPEG12_MACROBLOCK_TYPE_BKWD;
 306    switch (mb->mb_type) {
 307       case PIPE_MPEG12_MACROBLOCK_TYPE_INTRA:
 308          stream->eb[1][1].flag = -1;
 309          break;
 310
 311       case PIPE_MPEG12_MACROBLOCK_TYPE_FWD:
 312       case PIPE_MPEG12_MACROBLOCK_TYPE_BKWD:
 313          stream->eb[1][1].flag = 1;
 314          break;
 315
 316       case PIPE_MPEG12_MACROBLOCK_TYPE_BI:
 317          stream->eb[1][1].flag = 0;
 318          break;
 319
 320       default:
 321          assert(0);
 322    }
 323
 324    get_motion_vectors(mb, stream->mv);
 325 }
 326
 327 void
 328 vl_vb_unmap(struct vl_vertex_buffer *buffer, struct pipe_context *pipe)
 329 {
 330    assert(buffer && pipe);
 331
 332    pipe_buffer_unmap(pipe, buffer->transfer);
 333 }
 334
 335 void
 336 vl_vb_restart(struct vl_vertex_buffer *buffer,
 337               unsigned *not_empty_start_instance, unsigned *not_empty_num_instances,
 338               unsigned *empty_start_instance, unsigned *empty_num_instances)
 339 {
 340    assert(buffer);
 341
 342    *not_empty_start_instance = 0;
 343    *not_empty_num_instances = buffer->num_not_empty;
 344    *empty_start_instance = buffer->size - buffer->num_empty;
 345    *empty_num_instances = buffer->num_empty;
 346
 347    buffer->num_not_empty = 0;
 348    buffer->num_empty = 0;
 349 }
 350
 351 void
 352 vl_vb_cleanup(struct vl_vertex_buffer *buffer)
 353 {
 354    assert(buffer);
 355
 356    pipe_resource_reference(&buffer->resource, NULL);
 357 }