src/gallium/auxiliary/vl/vl_idct.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 "vl_idct.h"
  29 #include "util/u_draw.h"
  30 #include <assert.h>
  31 #include <pipe/p_context.h>
  32 #include <pipe/p_screen.h>
  33 #include <util/u_inlines.h>
  34 #include <util/u_sampler.h>
  35 #include <util/u_format.h>
  36 #include <tgsi/tgsi_ureg.h>
  37 #include "vl_types.h"
  38
  39 #define BLOCK_WIDTH 8
  40 #define BLOCK_HEIGHT 8
  41 #define SCALE_FACTOR_16_TO_12 (32768.0f / 2048.0f)
  42 #define SCALE_FACTOR_9_TO_16 (256.0f / 32768.0f)
  43
  44 struct vertex_shader_consts
  45 {
  46    struct vertex4f norm;
  47 };
  48
  49 enum VS_INPUT
  50 {
  51    VS_I_RECT,
  52    VS_I_VPOS,
  53
  54    NUM_VS_INPUTS
  55 };
  56
  57 enum VS_OUTPUT
  58 {
  59    VS_O_VPOS,
  60    VS_O_BLOCK,
  61    VS_O_TEX,
  62    VS_O_START,
  63    VS_O_STEP
  64 };
  65
  66
  67 static const float const_matrix[8][8] = {
  68    {  0.3535530f,  0.3535530f,  0.3535530f,  0.3535530f,  0.3535530f,  0.3535530f,  0.353553f,  0.3535530f },
  69    {  0.4903930f,  0.4157350f,  0.2777850f,  0.0975451f, -0.0975452f, -0.2777850f, -0.415735f, -0.4903930f },
  70    {  0.4619400f,  0.1913420f, -0.1913420f, -0.4619400f, -0.4619400f, -0.1913420f,  0.191342f,  0.4619400f },
  71    {  0.4157350f, -0.0975452f, -0.4903930f, -0.2777850f,  0.2777850f,  0.4903930f,  0.097545f, -0.4157350f },
  72    {  0.3535530f, -0.3535530f, -0.3535530f,  0.3535540f,  0.3535530f, -0.3535540f, -0.353553f,  0.3535530f },
  73    {  0.2777850f, -0.4903930f,  0.0975452f,  0.4157350f, -0.4157350f, -0.0975451f,  0.490393f, -0.2777850f },
  74    {  0.1913420f, -0.4619400f,  0.4619400f, -0.1913420f, -0.1913410f,  0.4619400f, -0.461940f,  0.1913420f },
  75    {  0.0975451f, -0.2777850f,  0.4157350f, -0.4903930f,  0.4903930f, -0.4157350f,  0.277786f, -0.0975458f }
  76 };
  77
  78 static const float const_transpose[8][8] = {
  79    {  0.3535530f,  0.4903930f,  0.4619400f,  0.4157350f,  0.3535530f,  0.2777850f,  0.191342f,  0.0975451f },
  80    {  0.3535530f,  0.4157350f,  0.1913420f, -0.0975452f, -0.3535530f, -0.4903930f, -0.461940f, -0.2777850f },
  81    {  0.3535530f,  0.2777850f, -0.1913420f, -0.4903930f, -0.3535530f,  0.0975452f,  0.461940f,  0.4157350f },
  82    {  0.3535530f,  0.0975451f, -0.4619400f, -0.2777850f,  0.3535540f,  0.4157350f, -0.191342f, -0.4903930f },
  83    {  0.3535530f, -0.0975452f, -0.4619400f,  0.2777850f,  0.3535530f, -0.4157350f, -0.191341f,  0.4903930f },
  84    {  0.3535530f, -0.2777850f, -0.1913420f,  0.4903930f, -0.3535540f, -0.0975451f,  0.461940f, -0.4157350f },
  85    {  0.3535530f, -0.4157350f,  0.1913420f,  0.0975450f, -0.3535530f,  0.4903930f, -0.461940f,  0.2777860f },
  86    {  0.3535530f, -0.4903930f,  0.4619400f, -0.4157350f,  0.3535530f, -0.2777850f,  0.191342f, -0.0975458f }
  87 };
  88
  89 /* vertices for a quad covering a block */
  90 static const struct vertex2f const_quad[4] = {
  91    {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f}
  92 };
  93
  94 static void *
  95 create_vert_shader(struct vl_idct *idct)
  96 {
  97    struct ureg_program *shader;
  98    struct ureg_src norm, bs;
  99    struct ureg_src vrect, vpos;
 100    struct ureg_dst scale, t_vpos;
 101    struct ureg_dst o_vpos, o_block, o_tex, o_start, o_step;
 102
 103    shader = ureg_create(TGSI_PROCESSOR_VERTEX);
 104    if (!shader)
 105       return NULL;
 106
 107    norm = ureg_DECL_constant(shader, 0);
 108    bs = ureg_imm2f(shader, BLOCK_WIDTH, BLOCK_HEIGHT);
 109
 110    scale = ureg_DECL_temporary(shader);
 111    t_vpos = ureg_DECL_temporary(shader);
 112
 113    vrect = ureg_DECL_vs_input(shader, VS_I_RECT);
 114    vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
 115
 116    o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
 117    o_block = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK);
 118    o_tex = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX);
 119    o_start = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_START);
 120    o_step = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_STEP);
 121
 122    /*
 123     * scale = norm * mbs;
 124     *
 125     * t_vpos = vpos + vrect
 126     * o_vpos.xy = t_vpos * scale
 127     * o_vpos.zw = vpos
 128     *
 129     * o_block = vrect
 130     * o_tex = t_pos
 131     * o_start = vpos * scale
 132     * o_step = norm
 133     *
 134     */
 135    ureg_MUL(shader, ureg_writemask(scale, TGSI_WRITEMASK_XY), norm, bs);
 136
 137    ureg_ADD(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), vpos, vrect);
 138    ureg_MUL(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos), ureg_src(scale));
 139    ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
 140    ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), vpos);
 141
 142    ureg_MOV(shader, ureg_writemask(o_tex, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
 143    ureg_MOV(shader, ureg_writemask(o_block, TGSI_WRITEMASK_XY), vrect);
 144    ureg_MUL(shader, ureg_writemask(o_start, TGSI_WRITEMASK_XY), vpos, ureg_src(scale));
 145    ureg_MOV(shader, ureg_writemask(o_step, TGSI_WRITEMASK_XY), norm);
 146
 147    ureg_release_temporary(shader, t_vpos);
 148    ureg_release_temporary(shader, scale);
 149
 150    ureg_END(shader);
 151
 152    return ureg_create_shader_and_destroy(shader, idct->pipe);
 153 }
 154
 155 static void
 156 matrix_mul(struct ureg_program *shader, struct ureg_dst dst,
 157            struct ureg_src tc[2], struct ureg_src sampler[2],
 158            struct ureg_src start[2], struct ureg_src step[2],
 159            float scale[2])
 160 {
 161    struct ureg_dst t_tc[2], m[2][2], tmp[2];
 162    unsigned i, j;
 163
 164    for(i = 0; i < 2; ++i) {
 165       t_tc[i] = ureg_DECL_temporary(shader);
 166       for(j = 0; j < 2; ++j)
 167          m[i][j] = ureg_DECL_temporary(shader);
 168       tmp[i] = ureg_DECL_temporary(shader);
 169    }
 170
 171    /*
 172     * m[0..1][0] = ?
 173     * tmp[0..1] = dot4(m[0..1][0], m[0..1][1])
 174     * fragment = tmp[0] + tmp[1]
 175     */
 176    ureg_MOV(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_X), start[0]);
 177    ureg_MOV(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_Y), tc[0]);
 178
 179    ureg_MOV(shader, ureg_writemask(t_tc[1], TGSI_WRITEMASK_X), tc[1]);
 180    ureg_MOV(shader, ureg_writemask(t_tc[1], TGSI_WRITEMASK_Y), start[1]);
 181
 182    for(i = 0; i < 2; ++i) {
 183       for(j = 0; j < 4; ++j) {
 184          /* Nouveau and r600g can't writemask tex dst regs (yet?), do in two steps */
 185          ureg_TEX(shader, tmp[0], TGSI_TEXTURE_2D, ureg_src(t_tc[0]), sampler[0]);
 186          ureg_MOV(shader, ureg_writemask(m[i][0], TGSI_WRITEMASK_X << j), ureg_scalar(ureg_src(tmp[0]), TGSI_SWIZZLE_X));
 187
 188          ureg_TEX(shader, tmp[1], TGSI_TEXTURE_2D, ureg_src(t_tc[1]), sampler[1]);
 189          ureg_MOV(shader, ureg_writemask(m[i][1], TGSI_WRITEMASK_X << j), ureg_scalar(ureg_src(tmp[1]), TGSI_SWIZZLE_X));
 190
 191          ureg_ADD(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_X), ureg_src(t_tc[0]), step[0]);
 192          ureg_ADD(shader, ureg_writemask(t_tc[1], TGSI_WRITEMASK_Y), ureg_src(t_tc[1]), step[1]);
 193       }
 194
 195       if(scale[0] != 1.0f)
 196          ureg_MUL(shader, m[i][0], ureg_src(m[i][0]), ureg_scalar(ureg_imm1f(shader, scale[0]), TGSI_SWIZZLE_X));
 197
 198       if(scale[1] != 1.0f)
 199          ureg_MUL(shader, m[i][1], ureg_src(m[i][1]), ureg_scalar(ureg_imm1f(shader, scale[1]), TGSI_SWIZZLE_X));
 200    }
 201
 202    ureg_DP4(shader, ureg_writemask(tmp[0], TGSI_WRITEMASK_X), ureg_src(m[0][0]), ureg_src(m[0][1]));
 203    ureg_DP4(shader, ureg_writemask(tmp[1], TGSI_WRITEMASK_X), ureg_src(m[1][0]), ureg_src(m[1][1]));
 204    ureg_ADD(shader, ureg_writemask(dst, TGSI_WRITEMASK_X), ureg_src(tmp[0]), ureg_src(tmp[1]));
 205
 206    for(i = 0; i < 2; ++i) {
 207       ureg_release_temporary(shader, t_tc[i]);
 208       for(j = 0; j < 2; ++j)
 209          ureg_release_temporary(shader, m[i][j]);
 210       ureg_release_temporary(shader, tmp[i]);
 211    }
 212 }
 213
 214 static void *
 215 create_transpose_frag_shader(struct vl_idct *idct)
 216 {
 217    struct ureg_program *shader;
 218    struct ureg_src tc[2], sampler[2];
 219    struct ureg_src start[2], step[2];
 220    struct ureg_dst fragment;
 221    float scale[2];
 222
 223    shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
 224    if (!shader)
 225       return NULL;
 226
 227    tc[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK, TGSI_INTERPOLATE_LINEAR);
 228    tc[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX, TGSI_INTERPOLATE_LINEAR);
 229
 230    start[0] = ureg_imm1f(shader, 0.0f);
 231    start[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT);
 232
 233    step[0] = ureg_imm1f(shader, 1.0f / BLOCK_HEIGHT);
 234    step[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_STEP, TGSI_INTERPOLATE_CONSTANT);
 235
 236    sampler[0] = ureg_DECL_sampler(shader, 0);
 237    sampler[1] = ureg_DECL_sampler(shader, 2);
 238
 239    scale[0] = 1.0f;
 240    scale[1] = SCALE_FACTOR_16_TO_12;
 241
 242    fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
 243
 244    //matrix_mul(shader, fragment, tc, sampler, start, step, scale);
 245    //ureg_MOV(shader, fragment, ureg_imm1f(shader, 0.0f));
 246
 247    ureg_END(shader);
 248
 249    return ureg_create_shader_and_destroy(shader, idct->pipe);
 250 }
 251
 252 static void *
 253 create_matrix_frag_shader(struct vl_idct *idct)
 254 {
 255    struct ureg_program *shader;
 256    struct ureg_src tc[2], sampler[2];
 257    struct ureg_src start[2], step[2];
 258    struct ureg_dst tmp, fragment;
 259    float scale[2];
 260
 261    shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
 262    if (!shader)
 263       return NULL;
 264
 265    tmp = ureg_DECL_temporary(shader);
 266
 267    tc[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX, TGSI_INTERPOLATE_LINEAR);
 268    tc[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK, TGSI_INTERPOLATE_LINEAR);
 269
 270    start[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT);
 271    start[1] = ureg_imm1f(shader, 0.0f);
 272
 273    step[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_STEP, TGSI_INTERPOLATE_CONSTANT);
 274    step[1] = ureg_imm1f(shader, 1.0f / BLOCK_WIDTH);
 275
 276    sampler[0] = ureg_DECL_sampler(shader, 3);
 277    sampler[1] = ureg_DECL_sampler(shader, 1);
 278
 279    scale[0] = 1.0f;
 280    scale[1] = 1.0f;
 281
 282    fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
 283
 284 //   matrix_mul(shader, tmp, tc, sampler, start, step, scale);
 285 //   ureg_MUL(shader, fragment, ureg_src(tmp), ureg_scalar(ureg_imm1f(shader, SCALE_FACTOR_9_TO_16), TGSI_SWIZZLE_X));
 286    ureg_TEX(shader, fragment, TGSI_TEXTURE_2D, tc[0], sampler[0]);
 287    //ureg_MUL(shader, , ureg_scalar(ureg_src(tmp), TGSI_SWIZZLE_X), ureg_imm1f(shader, 1.0f));
 288
 289    ureg_END(shader);
 290
 291    return ureg_create_shader_and_destroy(shader, idct->pipe);
 292 }
 293
 294 static void
 295 xfer_buffers_map(struct vl_idct *idct)
 296 {
 297    struct pipe_box rect =
 298    {
 299       0, 0, 0,
 300       idct->destination->width0,
 301       idct->destination->height0,
 302       1
 303    };
 304
 305    idct->tex_transfer = idct->pipe->get_transfer
 306    (
 307 #if 0
 308       idct->pipe, idct->textures.individual.intermediate,
 309 #else
 310       idct->pipe, idct->destination,
 311 #endif
 312       u_subresource(0, 0),
 313       PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
 314       &rect
 315    );
 316
 317    idct->texels = idct->pipe->transfer_map(idct->pipe, idct->tex_transfer);
 318
 319    idct->vectors = pipe_buffer_map
 320    (
 321       idct->pipe,
 322       idct->vertex_bufs.individual.pos.buffer,
 323       PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
 324       &idct->vec_transfer
 325    );
 326 }
 327
 328 static void
 329 xfer_buffers_unmap(struct vl_idct *idct)
 330 {
 331    pipe_buffer_unmap(idct->pipe, idct->vertex_bufs.individual.pos.buffer, idct->vec_transfer);
 332
 333    idct->pipe->transfer_unmap(idct->pipe, idct->tex_transfer);
 334    idct->pipe->transfer_destroy(idct->pipe, idct->tex_transfer);
 335 }
 336
 337 static bool
 338 init_shaders(struct vl_idct *idct)
 339 {
 340    assert(idct);
 341
 342    assert(idct->vs = create_vert_shader(idct));
 343    assert(idct->transpose_fs = create_transpose_frag_shader(idct));
 344    assert(idct->matrix_fs = create_matrix_frag_shader(idct));
 345
 346    return true;
 347 }
 348
 349 static void
 350 cleanup_shaders(struct vl_idct *idct)
 351 {
 352    assert(idct);
 353
 354    idct->pipe->delete_vs_state(idct->pipe, idct->vs);
 355    idct->pipe->delete_fs_state(idct->pipe, idct->transpose_fs);
 356    idct->pipe->delete_fs_state(idct->pipe, idct->matrix_fs);
 357 }
 358
 359 static bool
 360 init_buffers(struct vl_idct *idct)
 361 {
 362    struct pipe_resource template;
 363    struct pipe_sampler_view sampler_view;
 364    struct pipe_vertex_element vertex_elems[2];
 365
 366    const unsigned max_blocks =
 367       align(idct->destination->width0, BLOCK_WIDTH) / BLOCK_WIDTH *
 368       align(idct->destination->height0, BLOCK_HEIGHT) / BLOCK_HEIGHT *
 369       idct->destination->depth0;
 370
 371    unsigned i;
 372
 373    memset(&template, 0, sizeof(struct pipe_resource));
 374    template.target = PIPE_TEXTURE_2D;
 375    template.format = PIPE_FORMAT_R32_FLOAT;
 376    template.last_level = 0;
 377    template.width0 = 8;
 378    template.height0 = 8;
 379    template.depth0 = 1;
 380    template.usage = PIPE_USAGE_IMMUTABLE;
 381    template.bind = PIPE_BIND_SAMPLER_VIEW;
 382    template.flags = 0;
 383
 384    idct->textures.individual.matrix = idct->pipe->screen->resource_create(idct->pipe->screen, &template);
 385    idct->textures.individual.transpose = idct->pipe->screen->resource_create(idct->pipe->screen, &template);
 386
 387    template.format = idct->destination->format;
 388    template.width0 = idct->destination->width0;
 389    template.height0 = idct->destination->height0;
 390    template.depth0 = idct->destination->depth0;
 391    template.usage = PIPE_USAGE_DYNAMIC;
 392    idct->textures.individual.source = idct->pipe->screen->resource_create(idct->pipe->screen, &template);
 393
 394    //template.format = PIPE_FORMAT_R32_FLOAT;
 395    //template.usage = PIPE_USAGE_STATIC;
 396    idct->textures.individual.intermediate = idct->pipe->screen->resource_create(idct->pipe->screen, &template);
 397
 398    for (i = 0; i < 4; ++i) {
 399       u_sampler_view_default_template(&sampler_view, idct->textures.all[i], idct->textures.all[i]->format);
 400       idct->sampler_views.all[i] = idct->pipe->create_sampler_view(idct->pipe, idct->textures.all[i], &sampler_view);
 401    }
 402
 403    idct->vertex_bufs.individual.quad.stride = sizeof(struct vertex2f);
 404    idct->vertex_bufs.individual.quad.max_index = 4 * max_blocks - 1;
 405    idct->vertex_bufs.individual.quad.buffer_offset = 0;
 406    idct->vertex_bufs.individual.quad.buffer = pipe_buffer_create
 407    (
 408       idct->pipe->screen,
 409       PIPE_BIND_VERTEX_BUFFER,
 410       sizeof(struct vertex2f) * 4 * max_blocks
 411    );
 412
 413    idct->vertex_bufs.individual.pos.stride = sizeof(struct vertex2f);
 414    idct->vertex_bufs.individual.pos.max_index = 4 * max_blocks - 1;
 415    idct->vertex_bufs.individual.pos.buffer_offset = 0;
 416    idct->vertex_bufs.individual.pos.buffer = pipe_buffer_create
 417    (
 418       idct->pipe->screen,
 419       PIPE_BIND_VERTEX_BUFFER,
 420       sizeof(struct vertex2f) * 4 * max_blocks
 421    );
 422
 423    /* Rect element */
 424    vertex_elems[0].src_offset = 0;
 425    vertex_elems[0].instance_divisor = 0;
 426    vertex_elems[0].vertex_buffer_index = 0;
 427    vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
 428
 429    /* Pos element */
 430    vertex_elems[1].src_offset = 0;
 431    vertex_elems[1].instance_divisor = 0;
 432    vertex_elems[1].vertex_buffer_index = 1;
 433    vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT;
 434
 435    idct->vertex_elems_state = idct->pipe->create_vertex_elements_state(idct->pipe, 2, vertex_elems);
 436
 437    idct->vs_const_buf = pipe_buffer_create
 438    (
 439       idct->pipe->screen,
 440       PIPE_BIND_CONSTANT_BUFFER,
 441       sizeof(struct vertex_shader_consts)
 442    );
 443
 444    return true;
 445 }
 446
 447 static void
 448 cleanup_buffers(struct vl_idct *idct)
 449 {
 450    unsigned i;
 451
 452    assert(idct);
 453
 454    pipe_resource_reference(&idct->vs_const_buf, NULL);
 455
 456    for (i = 0; i < 4; ++i) {
 457       pipe_sampler_view_reference(&idct->sampler_views.all[i], NULL);
 458       pipe_resource_reference(&idct->textures.all[i], NULL);
 459    }
 460
 461    idct->pipe->delete_vertex_elements_state(idct->pipe, idct->vertex_elems_state);
 462    pipe_resource_reference(&idct->vertex_bufs.individual.quad.buffer, NULL);
 463    pipe_resource_reference(&idct->vertex_bufs.individual.pos.buffer, NULL);
 464 }
 465
 466 static void
 467 init_constants(struct vl_idct *idct)
 468 {
 469    struct pipe_transfer *buf_transfer;
 470    struct vertex_shader_consts *vs_consts;
 471    struct vertex2f *v;
 472
 473    unsigned i;
 474
 475    v = pipe_buffer_map
 476    (
 477       idct->pipe,
 478       idct->vertex_bufs.individual.quad.buffer,
 479       PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
 480       &buf_transfer
 481    );
 482
 483    for ( i = 0; i <= idct->vertex_bufs.individual.quad.max_index; i += 4)
 484      memcpy(v + i, &const_quad, sizeof(const_quad));
 485
 486    pipe_buffer_unmap(idct->pipe, idct->vertex_bufs.individual.quad.buffer, buf_transfer);
 487
 488
 489    v = pipe_buffer_map
 490    (
 491       idct->pipe,
 492       idct->textures.individual.matrix,
 493       PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
 494       &buf_transfer
 495    );
 496
 497    memcpy(v, &const_matrix, sizeof(const_matrix));
 498
 499    pipe_buffer_unmap(idct->pipe, idct->textures.individual.matrix, buf_transfer);
 500
 501    v = pipe_buffer_map
 502    (
 503       idct->pipe,
 504       idct->textures.individual.transpose,
 505       PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
 506       &buf_transfer
 507    );
 508
 509    memcpy(v, &const_transpose, sizeof(const_transpose));
 510
 511    pipe_buffer_unmap(idct->pipe, idct->textures.individual.transpose, buf_transfer);
 512
 513    vs_consts = pipe_buffer_map
 514    (
 515       idct->pipe, idct->vs_const_buf,
 516       PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
 517       &buf_transfer
 518    );
 519
 520    vs_consts->norm.x = 1.0f / idct->destination->width0;
 521    vs_consts->norm.y = 1.0f / idct->destination->height0;
 522
 523    pipe_buffer_unmap(idct->pipe, idct->vs_const_buf, buf_transfer);
 524 }
 525
 526 static void
 527 init_state(struct vl_idct *idct)
 528 {
 529    struct pipe_sampler_state sampler;
 530    unsigned i;
 531
 532    idct->num_blocks = 0;
 533
 534    idct->viewport.scale[0] = idct->destination->width0;
 535    idct->viewport.scale[1] = idct->destination->height0;
 536    idct->viewport.scale[2] = 1;
 537    idct->viewport.scale[3] = 1;
 538    idct->viewport.translate[0] = 0;
 539    idct->viewport.translate[1] = 0;
 540    idct->viewport.translate[2] = 0;
 541    idct->viewport.translate[3] = 0;
 542
 543    idct->fb_state.width = idct->destination->width0;
 544    idct->fb_state.height = idct->destination->height0;
 545    idct->fb_state.nr_cbufs = 1;
 546    idct->fb_state.zsbuf = NULL;
 547
 548    for (i = 0; i < 4; ++i) {
 549       memset(&sampler, 0, sizeof(sampler));
 550       sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
 551       sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
 552       sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
 553       sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
 554       sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
 555       sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
 556       sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
 557       sampler.compare_func = PIPE_FUNC_ALWAYS;
 558       sampler.normalized_coords = 1;
 559       /*sampler.shadow_ambient = ; */
 560       /*sampler.lod_bias = ; */
 561       sampler.min_lod = 0;
 562       /*sampler.max_lod = ; */
 563       /*sampler.border_color[0] = ; */
 564       /*sampler.max_anisotropy = ; */
 565       idct->samplers.all[i] = idct->pipe->create_sampler_state(idct->pipe, &sampler);
 566    }
 567 }
 568
 569 static void
 570 cleanup_state(struct vl_idct *idct)
 571 {
 572    unsigned i;
 573
 574    for (i = 0; i < 4; ++i)
 575       idct->pipe->delete_sampler_state(idct->pipe, idct->samplers.all[i]);
 576 }
 577
 578 bool vl_idct_init(struct vl_idct *idct, struct pipe_context *pipe, struct pipe_resource *dst)
 579 {
 580    assert(idct && pipe && dst);
 581
 582    idct->pipe = pipe;
 583    pipe_resource_reference(&idct->destination, dst);
 584
 585    init_state(idct);
 586
 587    if(!init_shaders(idct))
 588       return false;
 589
 590    if(!init_buffers(idct)) {
 591       cleanup_shaders(idct);
 592       return false;
 593    }
 594
 595    idct->surfaces.intermediate = idct->pipe->screen->get_tex_surface(
 596       idct->pipe->screen, idct->textures.individual.intermediate, 0, 0, 0,
 597       PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET);
 598
 599    idct->surfaces.destination = idct->pipe->screen->get_tex_surface(
 600       idct->pipe->screen, idct->destination, 0, 0, 0,
 601       PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET);
 602
 603    init_constants(idct);
 604    xfer_buffers_map(idct);
 605
 606    return true;
 607 }
 608
 609 void vl_idct_cleanup(struct vl_idct *idct)
 610 {
 611    idct->pipe->screen->tex_surface_destroy(idct->surfaces.destination);
 612    idct->pipe->screen->tex_surface_destroy(idct->surfaces.intermediate);
 613
 614    cleanup_shaders(idct);
 615    cleanup_buffers(idct);
 616
 617    cleanup_state(idct);
 618
 619    pipe_resource_reference(&idct->destination, NULL);
 620 }
 621
 622 void vl_idct_add_block(struct vl_idct *idct, unsigned x, unsigned y, short *block)
 623 {
 624    struct vertex2f v;
 625
 626    unsigned tex_pitch;
 627    short *texels;
 628
 629    unsigned i;
 630
 631    assert(idct);
 632    assert(block);
 633
 634    v.x = x;
 635    v.y = y;
 636
 637    for (i = 0; i < 4; ++i) {
 638       idct->vectors[idct->num_blocks * 4 + i] = v;
 639    }
 640
 641    tex_pitch = idct->tex_transfer->stride / util_format_get_blocksize(idct->tex_transfer->resource->format);
 642    texels = idct->texels + y * tex_pitch * BLOCK_HEIGHT + x * BLOCK_WIDTH;
 643
 644    for (i = 0; i < BLOCK_HEIGHT; ++i)
 645       memcpy(texels + i * tex_pitch, block + i * BLOCK_WIDTH, BLOCK_WIDTH * 2);
 646
 647    idct->num_blocks++;
 648 }
 649
 650 void vl_idct_flush(struct vl_idct *idct)
 651 {
 652    xfer_buffers_unmap(idct);
 653
 654    idct->pipe->set_constant_buffer(idct->pipe, PIPE_SHADER_VERTEX, 0, idct->vs_const_buf);
 655
 656 #if 0
 657    /* first stage */
 658    idct->fb_state.cbufs[0] = idct->surfaces.intermediate;
 659    idct->pipe->set_framebuffer_state(idct->pipe, &idct->fb_state);
 660    idct->pipe->set_viewport_state(idct->pipe, &idct->viewport);
 661
 662    idct->pipe->set_vertex_buffers(idct->pipe, 2, idct->vertex_bufs.all);
 663    idct->pipe->bind_vertex_elements_state(idct->pipe, idct->vertex_elems_state);
 664    idct->pipe->set_fragment_sampler_views(idct->pipe, 4, idct->sampler_views.all);
 665    idct->pipe->bind_fragment_sampler_states(idct->pipe, 4, idct->samplers.all);
 666    idct->pipe->bind_vs_state(idct->pipe, idct->vs);
 667    idct->pipe->bind_fs_state(idct->pipe, idct->transpose_fs);
 668
 669    util_draw_arrays(idct->pipe, PIPE_PRIM_QUADS, 0, idct->num_blocks * 4);
 670
 671    /* second stage */
 672    idct->fb_state.cbufs[0] = idct->surfaces.destination;
 673    idct->pipe->set_framebuffer_state(idct->pipe, &idct->fb_state);
 674    idct->pipe->set_viewport_state(idct->pipe, &idct->viewport);
 675
 676    idct->pipe->set_vertex_buffers(idct->pipe, 2, idct->vertex_bufs.all);
 677    idct->pipe->bind_vertex_elements_state(idct->pipe, idct->vertex_elems_state);
 678    idct->pipe->set_fragment_sampler_views(idct->pipe, 4, idct->sampler_views.all);
 679    idct->pipe->bind_fragment_sampler_states(idct->pipe, 4, idct->samplers.all);
 680    idct->pipe->bind_vs_state(idct->pipe, idct->vs);
 681    idct->pipe->bind_fs_state(idct->pipe, idct->matrix_fs);
 682
 683    util_draw_arrays(idct->pipe, PIPE_PRIM_QUADS, 0, idct->num_blocks * 4);
 684
 685    idct->pipe->flush(idct->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
 686 #endif
 687
 688    idct->num_blocks = 0;
 689    xfer_buffers_map(idct);
 690 }