vl_idct.c revision 7c4887f5ae642131d7895da5bffda77a6287c6d4
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 "vl_vertex_buffers.h" 30#include "util/u_draw.h" 31#include <assert.h> 32#include <pipe/p_context.h> 33#include <pipe/p_screen.h> 34#include <util/u_inlines.h> 35#include <util/u_sampler.h> 36#include <util/u_format.h> 37#include <tgsi/tgsi_ureg.h> 38#include "vl_types.h" 39 40#define BLOCK_WIDTH 8 41#define BLOCK_HEIGHT 8 42 43#define SCALE_FACTOR_16_TO_9 (32768.0f / 256.0f) 44 45#define STAGE1_SCALE 4.0f 46#define STAGE2_SCALE (SCALE_FACTOR_16_TO_9 / STAGE1_SCALE) 47 48#define NR_RENDER_TARGETS 1 49 50enum VS_INPUT 51{ 52 VS_I_RECT, 53 VS_I_VPOS, 54 55 NUM_VS_INPUTS 56}; 57 58enum VS_OUTPUT 59{ 60 VS_O_VPOS, 61 VS_O_BLOCK, 62 VS_O_TEX, 63 VS_O_START 64}; 65 66static const float const_matrix[8][8] = { 67 { 0.3535530f, 0.3535530f, 0.3535530f, 0.3535530f, 0.3535530f, 0.3535530f, 0.353553f, 0.3535530f }, 68 { 0.4903930f, 0.4157350f, 0.2777850f, 0.0975451f, -0.0975452f, -0.2777850f, -0.415735f, -0.4903930f }, 69 { 0.4619400f, 0.1913420f, -0.1913420f, -0.4619400f, -0.4619400f, -0.1913420f, 0.191342f, 0.4619400f }, 70 { 0.4157350f, -0.0975452f, -0.4903930f, -0.2777850f, 0.2777850f, 0.4903930f, 0.097545f, -0.4157350f }, 71 { 0.3535530f, -0.3535530f, -0.3535530f, 0.3535540f, 0.3535530f, -0.3535540f, -0.353553f, 0.3535530f }, 72 { 0.2777850f, -0.4903930f, 0.0975452f, 0.4157350f, -0.4157350f, -0.0975451f, 0.490393f, -0.2777850f }, 73 { 0.1913420f, -0.4619400f, 0.4619400f, -0.1913420f, -0.1913410f, 0.4619400f, -0.461940f, 0.1913420f }, 74 { 0.0975451f, -0.2777850f, 0.4157350f, -0.4903930f, 0.4903930f, -0.4157350f, 0.277786f, -0.0975458f } 75}; 76 77static void * 78create_vert_shader(struct vl_idct *idct) 79{ 80 struct ureg_program *shader; 81 struct ureg_src scale; 82 struct ureg_src vrect, vpos; 83 struct ureg_dst t_vpos; 84 struct ureg_dst o_vpos, o_block, o_tex, o_start; 85 86 shader = ureg_create(TGSI_PROCESSOR_VERTEX); 87 if (!shader) 88 return NULL; 89 90 t_vpos = ureg_DECL_temporary(shader); 91 92 vrect = ureg_DECL_vs_input(shader, VS_I_RECT); 93 vpos = ureg_DECL_vs_input(shader, VS_I_VPOS); 94 95 o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS); 96 o_block = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK); 97 o_tex = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX); 98 o_start = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_START); 99 100 /* 101 * scale = (BLOCK_WIDTH, BLOCK_HEIGHT) / (dst.width, dst.height) 102 * 103 * t_vpos = vpos + vrect 104 * o_vpos.xy = t_vpos * scale 105 * o_vpos.zw = vpos 106 * 107 * o_block = vrect 108 * o_tex = t_pos 109 * o_start = vpos * scale 110 * 111 */ 112 scale = ureg_imm2f(shader, 113 (float)BLOCK_WIDTH / idct->buffer_width, 114 (float)BLOCK_HEIGHT / idct->buffer_height); 115 116 ureg_ADD(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), vpos, vrect); 117 ureg_MUL(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos), scale); 118 ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos)); 119 ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), vpos); 120 121 ureg_MOV(shader, ureg_writemask(o_block, TGSI_WRITEMASK_XY), vrect); 122 ureg_MOV(shader, ureg_writemask(o_tex, TGSI_WRITEMASK_XY), ureg_src(t_vpos)); 123 ureg_MUL(shader, ureg_writemask(o_start, TGSI_WRITEMASK_XY), vpos, scale); 124 125 ureg_release_temporary(shader, t_vpos); 126 127 ureg_END(shader); 128 129 return ureg_create_shader_and_destroy(shader, idct->pipe); 130} 131 132static void 133fetch_four(struct ureg_program *shader, struct ureg_dst m[2], 134 struct ureg_src tc, struct ureg_src sampler, 135 struct ureg_src start, struct ureg_src block, 136 bool right_side, bool transposed, float size) 137{ 138 struct ureg_dst t_tc; 139 unsigned wm_start = (right_side == transposed) ? TGSI_WRITEMASK_X : TGSI_WRITEMASK_Y; 140 unsigned wm_tc = (right_side == transposed) ? TGSI_WRITEMASK_Y : TGSI_WRITEMASK_X; 141 142 t_tc = ureg_DECL_temporary(shader); 143 m[0] = ureg_DECL_temporary(shader); 144 m[1] = ureg_DECL_temporary(shader); 145 146 /* 147 * t_tc.x = right_side ? start.x : tc.x 148 * t_tc.y = right_side ? tc.y : start.y 149 * m[0..1] = tex(t_tc++, sampler) 150 */ 151 if(!right_side) { 152 ureg_MOV(shader, ureg_writemask(t_tc, wm_start), ureg_scalar(start, TGSI_SWIZZLE_X)); 153 ureg_MOV(shader, ureg_writemask(t_tc, wm_tc), ureg_scalar(tc, TGSI_SWIZZLE_Y)); 154 } else { 155 ureg_MOV(shader, ureg_writemask(t_tc, wm_start), ureg_scalar(start, TGSI_SWIZZLE_Y)); 156 ureg_MOV(shader, ureg_writemask(t_tc, wm_tc), ureg_scalar(tc, TGSI_SWIZZLE_X)); 157 } 158 159#if NR_RENDER_TARGETS == 8 160 ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Z), ureg_scalar(block, TGSI_SWIZZLE_X)); 161#else 162 ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Z), ureg_imm1f(shader, 0.0f)); 163#endif 164 165 ureg_TEX(shader, m[0], TGSI_TEXTURE_3D, ureg_src(t_tc), sampler); 166 ureg_ADD(shader, ureg_writemask(t_tc, wm_start), ureg_src(t_tc), ureg_imm1f(shader, 1.0f / size)); 167 ureg_TEX(shader, m[1], TGSI_TEXTURE_3D, ureg_src(t_tc), sampler); 168 169 ureg_release_temporary(shader, t_tc); 170} 171 172static void 173matrix_mul(struct ureg_program *shader, struct ureg_dst dst, struct ureg_dst l[2], struct ureg_dst r[2]) 174{ 175 struct ureg_dst tmp[2]; 176 unsigned i; 177 178 for(i = 0; i < 2; ++i) { 179 tmp[i] = ureg_DECL_temporary(shader); 180 } 181 182 /* 183 * tmp[0..1] = dot4(m[0][0..1], m[1][0..1]) 184 * dst = tmp[0] + tmp[1] 185 */ 186 ureg_DP4(shader, ureg_writemask(tmp[0], TGSI_WRITEMASK_X), ureg_src(l[0]), ureg_src(r[0])); 187 ureg_DP4(shader, ureg_writemask(tmp[1], TGSI_WRITEMASK_X), ureg_src(l[1]), ureg_src(r[1])); 188 ureg_ADD(shader, dst, 189 ureg_scalar(ureg_src(tmp[0]), TGSI_SWIZZLE_X), 190 ureg_scalar(ureg_src(tmp[1]), TGSI_SWIZZLE_X)); 191 192 for(i = 0; i < 2; ++i) { 193 ureg_release_temporary(shader, tmp[i]); 194 } 195} 196 197static void * 198create_transpose_frag_shader(struct vl_idct *idct) 199{ 200 struct ureg_program *shader; 201 202 struct ureg_src block, tex, sampler[2]; 203 struct ureg_src start[2]; 204 205 struct ureg_dst l[2], r[2]; 206 struct ureg_dst tmp, fragment; 207 208 shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); 209 if (!shader) 210 return NULL; 211 212 block = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK, TGSI_INTERPOLATE_LINEAR); 213 tex = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX, TGSI_INTERPOLATE_CONSTANT); 214 215 sampler[0] = ureg_DECL_sampler(shader, 0); 216 sampler[1] = ureg_DECL_sampler(shader, 1); 217 218 start[0] = ureg_imm1f(shader, 0.0f); 219 start[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT); 220 221 fetch_four(shader, l, block, sampler[0], start[0], block, false, false, BLOCK_WIDTH / 4); 222 fetch_four(shader, r, tex, sampler[1], start[1], block, true, false, idct->buffer_height / 4); 223 224 fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0); 225 226 tmp = ureg_DECL_temporary(shader); 227 matrix_mul(shader, ureg_writemask(tmp, TGSI_WRITEMASK_X), l, r); 228 ureg_MUL(shader, fragment, ureg_src(tmp), ureg_imm1f(shader, STAGE2_SCALE)); 229 230 ureg_release_temporary(shader, tmp); 231 ureg_release_temporary(shader, l[0]); 232 ureg_release_temporary(shader, l[1]); 233 ureg_release_temporary(shader, r[0]); 234 ureg_release_temporary(shader, r[1]); 235 236 ureg_END(shader); 237 238 return ureg_create_shader_and_destroy(shader, idct->pipe); 239} 240 241static void * 242create_matrix_frag_shader(struct vl_idct *idct) 243{ 244 struct ureg_program *shader; 245 246 struct ureg_src tex, block, sampler[2]; 247 struct ureg_src start[2]; 248 249 struct ureg_dst l[4][2], r[2]; 250 struct ureg_dst t_tc, tmp, fragment[NR_RENDER_TARGETS]; 251 252 unsigned i, j; 253 254 shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); 255 if (!shader) 256 return NULL; 257 258 t_tc = ureg_DECL_temporary(shader); 259 tmp = ureg_DECL_temporary(shader); 260 261 tex = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX, TGSI_INTERPOLATE_LINEAR); 262 block = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK, TGSI_INTERPOLATE_LINEAR); 263 264 sampler[0] = ureg_DECL_sampler(shader, 1); 265 sampler[1] = ureg_DECL_sampler(shader, 0); 266 267 start[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT); 268 start[1] = ureg_imm1f(shader, 0.0f); 269 270 for (i = 0; i < NR_RENDER_TARGETS; ++i) 271 fragment[i] = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, i); 272 273 ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Y), tex); 274 for (i = 0; i < 4; ++i) { 275 fetch_four(shader, l[i], ureg_src(t_tc), sampler[0], start[0], block, false, false, idct->buffer_width / 4); 276 ureg_MUL(shader, l[i][0], ureg_src(l[i][0]), ureg_imm1f(shader, STAGE1_SCALE)); 277 ureg_MUL(shader, l[i][1], ureg_src(l[i][1]), ureg_imm1f(shader, STAGE1_SCALE)); 278 if(i != 3) 279 ureg_ADD(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Y), 280 ureg_src(t_tc), ureg_imm1f(shader, 1.0f / idct->buffer_height)); 281 } 282 283 for (i = 0; i < NR_RENDER_TARGETS; ++i) { 284 285#if NR_RENDER_TARGETS == 8 286 ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_X), ureg_imm1f(shader, 1.0f / BLOCK_WIDTH * i)); 287 fetch_four(shader, r, ureg_src(t_tc), sampler[1], start[1], block, true, true, BLOCK_WIDTH / 4); 288#elif NR_RENDER_TARGETS == 1 289 fetch_four(shader, r, block, sampler[1], start[1], block, true, true, BLOCK_WIDTH / 4); 290#else 291#error invalid number of render targets 292#endif 293 294 for (j = 0; j < 4; ++j) { 295 matrix_mul(shader, ureg_writemask(fragment[i], TGSI_WRITEMASK_X << j), l[j], r); 296 } 297 ureg_release_temporary(shader, r[0]); 298 ureg_release_temporary(shader, r[1]); 299 } 300 301 ureg_release_temporary(shader, t_tc); 302 ureg_release_temporary(shader, tmp); 303 304 for (i = 0; i < 4; ++i) { 305 ureg_release_temporary(shader, l[i][0]); 306 ureg_release_temporary(shader, l[i][1]); 307 } 308 309 ureg_END(shader); 310 311 return ureg_create_shader_and_destroy(shader, idct->pipe); 312} 313 314static bool 315init_shaders(struct vl_idct *idct) 316{ 317 idct->vs = create_vert_shader(idct); 318 idct->matrix_fs = create_matrix_frag_shader(idct); 319 idct->transpose_fs = create_transpose_frag_shader(idct); 320 321 return 322 idct->vs != NULL && 323 idct->transpose_fs != NULL && 324 idct->matrix_fs != NULL; 325} 326 327static void 328cleanup_shaders(struct vl_idct *idct) 329{ 330 idct->pipe->delete_vs_state(idct->pipe, idct->vs); 331 idct->pipe->delete_fs_state(idct->pipe, idct->matrix_fs); 332 idct->pipe->delete_fs_state(idct->pipe, idct->transpose_fs); 333} 334 335static bool 336init_state(struct vl_idct *idct) 337{ 338 struct pipe_vertex_element vertex_elems[NUM_VS_INPUTS]; 339 struct pipe_sampler_state sampler; 340 struct pipe_rasterizer_state rs_state; 341 unsigned i; 342 343 assert(idct); 344 345 idct->quad = vl_vb_upload_quads(idct->pipe, idct->max_blocks); 346 347 if(idct->quad.buffer == NULL) 348 return false; 349 350 for (i = 0; i < 4; ++i) { 351 memset(&sampler, 0, sizeof(sampler)); 352 sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; 353 sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; 354 sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; 355 sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; 356 sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; 357 sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; 358 sampler.compare_mode = PIPE_TEX_COMPARE_NONE; 359 sampler.compare_func = PIPE_FUNC_ALWAYS; 360 sampler.normalized_coords = 1; 361 /*sampler.shadow_ambient = ; */ 362 /*sampler.lod_bias = ; */ 363 sampler.min_lod = 0; 364 /*sampler.max_lod = ; */ 365 /*sampler.border_color[0] = ; */ 366 /*sampler.max_anisotropy = ; */ 367 idct->samplers.all[i] = idct->pipe->create_sampler_state(idct->pipe, &sampler); 368 } 369 370 memset(&rs_state, 0, sizeof(rs_state)); 371 /*rs_state.sprite_coord_enable */ 372 rs_state.sprite_coord_mode = PIPE_SPRITE_COORD_UPPER_LEFT; 373 rs_state.point_quad_rasterization = true; 374 rs_state.point_size = BLOCK_WIDTH; 375 rs_state.gl_rasterization_rules = false; 376 idct->rs_state = idct->pipe->create_rasterizer_state(idct->pipe, &rs_state); 377 378 vertex_elems[VS_I_RECT] = vl_vb_get_quad_vertex_element(); 379 380 /* Pos element */ 381 vertex_elems[VS_I_VPOS].src_format = PIPE_FORMAT_R32G32_FLOAT; 382 383 idct->vertex_buffer_stride = vl_vb_element_helper(&vertex_elems[VS_I_VPOS], 1, 1); 384 idct->vertex_elems_state = idct->pipe->create_vertex_elements_state(idct->pipe, 2, vertex_elems); 385 386 return true; 387} 388 389static void 390cleanup_state(struct vl_idct *idct) 391{ 392 unsigned i; 393 394 for (i = 0; i < 4; ++i) 395 idct->pipe->delete_sampler_state(idct->pipe, idct->samplers.all[i]); 396 397 idct->pipe->delete_rasterizer_state(idct->pipe, idct->rs_state); 398 idct->pipe->delete_vertex_elements_state(idct->pipe, idct->vertex_elems_state); 399} 400 401static bool 402init_textures(struct vl_idct *idct, struct vl_idct_buffer *buffer) 403{ 404 struct pipe_resource template; 405 struct pipe_sampler_view sampler_view; 406 unsigned i; 407 408 assert(idct && buffer); 409 410 /* create textures */ 411 memset(&template, 0, sizeof(struct pipe_resource)); 412 template.last_level = 0; 413 template.depth0 = 1; 414 template.bind = PIPE_BIND_SAMPLER_VIEW; 415 template.flags = 0; 416 417 template.target = PIPE_TEXTURE_2D; 418 template.format = PIPE_FORMAT_R16G16B16A16_SNORM; 419 template.width0 = idct->buffer_width / 4; 420 template.height0 = idct->buffer_height; 421 template.depth0 = 1; 422 template.usage = PIPE_USAGE_STREAM; 423 buffer->textures.individual.source = idct->pipe->screen->resource_create(idct->pipe->screen, &template); 424 425 template.target = PIPE_TEXTURE_3D; 426 template.format = PIPE_FORMAT_R16G16B16A16_SNORM; 427 template.width0 = idct->buffer_width / NR_RENDER_TARGETS; 428 template.height0 = idct->buffer_height / 4; 429 template.depth0 = NR_RENDER_TARGETS; 430 template.usage = PIPE_USAGE_STATIC; 431 buffer->textures.individual.intermediate = idct->pipe->screen->resource_create(idct->pipe->screen, &template); 432 433 for (i = 0; i < 4; ++i) { 434 if(buffer->textures.all[i] == NULL) 435 return false; /* a texture failed to allocate */ 436 437 u_sampler_view_default_template(&sampler_view, buffer->textures.all[i], buffer->textures.all[i]->format); 438 buffer->sampler_views.all[i] = idct->pipe->create_sampler_view(idct->pipe, buffer->textures.all[i], &sampler_view); 439 } 440 441 return true; 442} 443 444static void 445cleanup_textures(struct vl_idct *idct, struct vl_idct_buffer *buffer) 446{ 447 unsigned i; 448 449 assert(idct && buffer); 450 451 for (i = 0; i < 4; ++i) { 452 pipe_sampler_view_reference(&buffer->sampler_views.all[i], NULL); 453 pipe_resource_reference(&buffer->textures.all[i], NULL); 454 } 455} 456 457static bool 458init_vertex_buffers(struct vl_idct *idct, struct vl_idct_buffer *buffer) 459{ 460 assert(idct && buffer); 461 462 buffer->vertex_bufs.individual.quad.stride = idct->quad.stride; 463 buffer->vertex_bufs.individual.quad.max_index = idct->quad.max_index; 464 buffer->vertex_bufs.individual.quad.buffer_offset = idct->quad.buffer_offset; 465 pipe_resource_reference(&buffer->vertex_bufs.individual.quad.buffer, idct->quad.buffer); 466 467 buffer->vertex_bufs.individual.pos = vl_vb_init( 468 &buffer->blocks, idct->pipe, idct->max_blocks, 2, 469 idct->vertex_buffer_stride); 470 471 if(buffer->vertex_bufs.individual.pos.buffer == NULL) 472 return false; 473 474 return true; 475} 476 477static void 478cleanup_vertex_buffers(struct vl_idct *idct, struct vl_idct_buffer *buffer) 479{ 480 assert(idct && buffer); 481 482 pipe_resource_reference(&buffer->vertex_bufs.individual.quad.buffer, NULL); 483 pipe_resource_reference(&buffer->vertex_bufs.individual.pos.buffer, NULL); 484 485 vl_vb_cleanup(&buffer->blocks); 486} 487 488struct pipe_resource * 489vl_idct_upload_matrix(struct pipe_context *pipe) 490{ 491 struct pipe_resource template, *matrix; 492 struct pipe_transfer *buf_transfer; 493 unsigned i, j, pitch; 494 float *f; 495 496 struct pipe_box rect = 497 { 498 0, 0, 0, 499 BLOCK_WIDTH / 4, 500 BLOCK_HEIGHT, 501 1 502 }; 503 504 memset(&template, 0, sizeof(struct pipe_resource)); 505 template.target = PIPE_TEXTURE_2D; 506 template.format = PIPE_FORMAT_R32G32B32A32_FLOAT; 507 template.last_level = 0; 508 template.width0 = 2; 509 template.height0 = 8; 510 template.depth0 = 1; 511 template.usage = PIPE_USAGE_IMMUTABLE; 512 template.bind = PIPE_BIND_SAMPLER_VIEW; 513 template.flags = 0; 514 515 matrix = pipe->screen->resource_create(pipe->screen, &template); 516 517 /* matrix */ 518 buf_transfer = pipe->get_transfer 519 ( 520 pipe, matrix, 521 u_subresource(0, 0), 522 PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD, 523 &rect 524 ); 525 pitch = buf_transfer->stride / sizeof(float); 526 527 f = pipe->transfer_map(pipe, buf_transfer); 528 for(i = 0; i < BLOCK_HEIGHT; ++i) 529 for(j = 0; j < BLOCK_WIDTH; ++j) 530 f[i * pitch + j] = const_matrix[j][i]; // transpose 531 532 pipe->transfer_unmap(pipe, buf_transfer); 533 pipe->transfer_destroy(pipe, buf_transfer); 534 535 return matrix; 536} 537 538bool vl_idct_init(struct vl_idct *idct, struct pipe_context *pipe, 539 unsigned buffer_width, unsigned buffer_height, 540 struct pipe_resource *matrix) 541{ 542 assert(idct && pipe && matrix); 543 544 idct->pipe = pipe; 545 idct->buffer_width = buffer_width; 546 idct->buffer_height = buffer_height; 547 pipe_resource_reference(&idct->matrix, matrix); 548 549 idct->max_blocks = 550 align(buffer_width, BLOCK_WIDTH) / BLOCK_WIDTH * 551 align(buffer_height, BLOCK_HEIGHT) / BLOCK_HEIGHT; 552 553 if(!init_shaders(idct)) 554 return false; 555 556 if(!init_state(idct)) { 557 cleanup_shaders(idct); 558 return false; 559 } 560 561 return true; 562} 563 564void 565vl_idct_cleanup(struct vl_idct *idct) 566{ 567 cleanup_shaders(idct); 568 cleanup_state(idct); 569 570 pipe_resource_reference(&idct->matrix, NULL); 571} 572 573bool 574vl_idct_init_buffer(struct vl_idct *idct, struct vl_idct_buffer *buffer, struct pipe_resource *dst) 575{ 576 unsigned i; 577 578 assert(buffer); 579 assert(idct); 580 assert(dst); 581 582 pipe_resource_reference(&buffer->textures.individual.matrix, idct->matrix); 583 pipe_resource_reference(&buffer->textures.individual.transpose, idct->matrix); 584 pipe_resource_reference(&buffer->destination, dst); 585 586 if (!init_textures(idct, buffer)) 587 return false; 588 589 if (!init_vertex_buffers(idct, buffer)) 590 return false; 591 592 /* init state */ 593 buffer->viewport[0].scale[0] = buffer->textures.individual.intermediate->width0; 594 buffer->viewport[0].scale[1] = buffer->textures.individual.intermediate->height0; 595 596 buffer->viewport[1].scale[0] = buffer->destination->width0; 597 buffer->viewport[1].scale[1] = buffer->destination->height0; 598 599 buffer->fb_state[0].width = buffer->textures.individual.intermediate->width0; 600 buffer->fb_state[0].height = buffer->textures.individual.intermediate->height0; 601 602 buffer->fb_state[0].nr_cbufs = NR_RENDER_TARGETS; 603 for(i = 0; i < NR_RENDER_TARGETS; ++i) { 604 buffer->fb_state[0].cbufs[i] = idct->pipe->screen->get_tex_surface( 605 idct->pipe->screen, buffer->textures.individual.intermediate, 0, 0, i, 606 PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET); 607 } 608 609 buffer->fb_state[1].width = buffer->destination->width0; 610 buffer->fb_state[1].height = buffer->destination->height0; 611 612 buffer->fb_state[1].nr_cbufs = 1; 613 buffer->fb_state[1].cbufs[0] = idct->pipe->screen->get_tex_surface( 614 idct->pipe->screen, buffer->destination, 0, 0, 0, 615 PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET); 616 617 for(i = 0; i < 2; ++i) { 618 buffer->viewport[i].scale[2] = 1; 619 buffer->viewport[i].scale[3] = 1; 620 buffer->viewport[i].translate[0] = 0; 621 buffer->viewport[i].translate[1] = 0; 622 buffer->viewport[i].translate[2] = 0; 623 buffer->viewport[i].translate[3] = 0; 624 625 buffer->fb_state[i].zsbuf = NULL; 626 } 627 628 return true; 629} 630 631void 632vl_idct_cleanup_buffer(struct vl_idct *idct, struct vl_idct_buffer *buffer) 633{ 634 unsigned i; 635 636 assert(buffer); 637 638 for(i = 0; i < NR_RENDER_TARGETS; ++i) { 639 idct->pipe->screen->tex_surface_destroy(buffer->fb_state[0].cbufs[i]); 640 } 641 642 idct->pipe->screen->tex_surface_destroy(buffer->fb_state[1].cbufs[0]); 643 644 cleanup_textures(idct, buffer); 645 cleanup_vertex_buffers(idct, buffer); 646} 647 648void 649vl_idct_map_buffers(struct vl_idct *idct, struct vl_idct_buffer *buffer) 650{ 651 assert(idct); 652 653 struct pipe_box rect = 654 { 655 0, 0, 0, 656 buffer->textures.individual.source->width0, 657 buffer->textures.individual.source->height0, 658 1 659 }; 660 661 buffer->tex_transfer = idct->pipe->get_transfer 662 ( 663 idct->pipe, buffer->textures.individual.source, 664 u_subresource(0, 0), 665 PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD, 666 &rect 667 ); 668 669 buffer->texels = idct->pipe->transfer_map(idct->pipe, buffer->tex_transfer); 670 671 vl_vb_map(&buffer->blocks, idct->pipe); 672} 673 674void 675vl_idct_add_block(struct vl_idct_buffer *buffer, unsigned x, unsigned y, short *block) 676{ 677 struct vertex2f v; 678 unsigned tex_pitch; 679 short *texels; 680 681 unsigned i; 682 683 assert(buffer); 684 685 tex_pitch = buffer->tex_transfer->stride / sizeof(short); 686 texels = buffer->texels + y * tex_pitch * BLOCK_HEIGHT + x * BLOCK_WIDTH; 687 688 for (i = 0; i < BLOCK_HEIGHT; ++i) 689 memcpy(texels + i * tex_pitch, block + i * BLOCK_WIDTH, BLOCK_WIDTH * sizeof(short)); 690 691 v.x = x; 692 v.y = y; 693 vl_vb_add_block(&buffer->blocks, (float*)&v); 694} 695 696void 697vl_idct_unmap_buffers(struct vl_idct *idct, struct vl_idct_buffer *buffer) 698{ 699 assert(idct && buffer); 700 701 idct->pipe->transfer_unmap(idct->pipe, buffer->tex_transfer); 702 idct->pipe->transfer_destroy(idct->pipe, buffer->tex_transfer); 703 vl_vb_unmap(&buffer->blocks, idct->pipe); 704} 705 706void 707vl_idct_flush(struct vl_idct *idct, struct vl_idct_buffer *buffer) 708{ 709 unsigned num_verts; 710 711 assert(idct); 712 713 num_verts = vl_vb_restart(&buffer->blocks); 714 715 if(num_verts > 0) { 716 717 idct->pipe->bind_rasterizer_state(idct->pipe, idct->rs_state); 718 idct->pipe->set_vertex_buffers(idct->pipe, 2, buffer->vertex_bufs.all); 719 idct->pipe->bind_vertex_elements_state(idct->pipe, idct->vertex_elems_state); 720 idct->pipe->bind_vs_state(idct->pipe, idct->vs); 721 722 /* first stage */ 723 idct->pipe->set_framebuffer_state(idct->pipe, &buffer->fb_state[0]); 724 idct->pipe->set_viewport_state(idct->pipe, &buffer->viewport[0]); 725 idct->pipe->set_fragment_sampler_views(idct->pipe, 2, buffer->sampler_views.stage[0]); 726 idct->pipe->bind_fragment_sampler_states(idct->pipe, 2, idct->samplers.stage[0]); 727 idct->pipe->bind_fs_state(idct->pipe, idct->matrix_fs); 728 util_draw_arrays(idct->pipe, PIPE_PRIM_QUADS, 0, num_verts); 729 730 /* second stage */ 731 idct->pipe->set_framebuffer_state(idct->pipe, &buffer->fb_state[1]); 732 idct->pipe->set_viewport_state(idct->pipe, &buffer->viewport[1]); 733 idct->pipe->set_fragment_sampler_views(idct->pipe, 2, buffer->sampler_views.stage[1]); 734 idct->pipe->bind_fragment_sampler_states(idct->pipe, 2, idct->samplers.stage[1]); 735 idct->pipe->bind_fs_state(idct->pipe, idct->transpose_fs); 736 util_draw_arrays(idct->pipe, PIPE_PRIM_QUADS, 0, num_verts); 737 } 738} 739