1 /********************************************************** 2 * Copyright 2009-2011 VMware, Inc. All rights reserved. 3 * 4 * Permission is hereby granted, free of charge, to any person 5 * obtaining a copy of this software and associated documentation 6 * files (the "Software"), to deal in the Software without 7 * restriction, including without limitation the rights to use, copy, 8 * modify, merge, publish, distribute, sublicense, and/or sell copies 9 * of the Software, and to permit persons to whom the Software is 10 * furnished to do so, subject to the following conditions: 11 * 12 * The above copyright notice and this permission notice shall be 13 * included in all copies or substantial portions of the Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 16 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 17 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 18 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 19 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 20 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 21 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 22 * SOFTWARE. 23 * 24 ********************************************************* 25 * Authors: 26 * Zack Rusin <zackr-at-vmware-dot-com> 27 */ 28 29 #include "xa_context.h" 30 #include "xa_priv.h" 31 #include <math.h> 32 #include "cso_cache/cso_context.h" 33 #include "util/u_inlines.h" 34 #include "util/u_sampler.h" 35 #include "util/u_draw_quad.h" 36 37 #define floatsEqual(x, y) (fabs(x - y) <= 0.00001f * MIN2(fabs(x), fabs(y))) 38 #define floatIsZero(x) (floatsEqual((x) + 1, 1)) 39 40 #define NUM_COMPONENTS 4 41 42 void 43 44 45 renderer_set_constants(struct xa_context *r, 46 int shader_type, const float *params, int param_bytes); 47 48 static inline boolean 49 is_affine(float *matrix) 50 { 51 return floatIsZero(matrix[2]) && floatIsZero(matrix[5]) 52 && floatsEqual(matrix[8], 1); 53 } 54 55 static inline void 56 map_point(float *mat, float x, float y, float *out_x, float *out_y) 57 { 58 if (!mat) { 59 *out_x = x; 60 *out_y = y; 61 return; 62 } 63 64 *out_x = mat[0] * x + mat[3] * y + mat[6]; 65 *out_y = mat[1] * x + mat[4] * y + mat[7]; 66 if (!is_affine(mat)) { 67 float w = 1 / (mat[2] * x + mat[5] * y + mat[8]); 68 69 *out_x *= w; 70 *out_y *= w; 71 } 72 } 73 74 static inline void 75 renderer_draw(struct xa_context *r) 76 { 77 int num_verts = r->buffer_size / (r->attrs_per_vertex * NUM_COMPONENTS); 78 79 if (!r->buffer_size) 80 return; 81 82 if (!r->scissor_valid) { 83 r->scissor.minx = 0; 84 r->scissor.miny = 0; 85 r->scissor.maxx = r->dst->tex->width0; 86 r->scissor.maxy = r->dst->tex->height0; 87 } 88 89 r->pipe->set_scissor_states(r->pipe, 0, 1, &r->scissor); 90 91 cso_set_vertex_elements(r->cso, r->attrs_per_vertex, r->velems); 92 util_draw_user_vertex_buffer(r->cso, r->buffer, PIPE_PRIM_QUADS, 93 num_verts, /* verts */ 94 r->attrs_per_vertex); /* attribs/vert */ 95 r->buffer_size = 0; 96 97 xa_scissor_reset(r); 98 } 99 100 static inline void 101 renderer_draw_conditional(struct xa_context *r, int next_batch) 102 { 103 if (r->buffer_size + next_batch >= XA_VB_SIZE || 104 (next_batch == 0 && r->buffer_size)) { 105 renderer_draw(r); 106 } 107 } 108 109 void 110 renderer_init_state(struct xa_context *r) 111 { 112 struct pipe_depth_stencil_alpha_state dsa; 113 struct pipe_rasterizer_state raster; 114 unsigned i; 115 116 /* set common initial clip state */ 117 memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state)); 118 cso_set_depth_stencil_alpha(r->cso, &dsa); 119 120 /* XXX: move to renderer_init_state? */ 121 memset(&raster, 0, sizeof(struct pipe_rasterizer_state)); 122 raster.half_pixel_center = 1; 123 raster.bottom_edge_rule = 1; 124 raster.depth_clip = 1; 125 raster.scissor = 1; 126 cso_set_rasterizer(r->cso, &raster); 127 128 /* vertex elements state */ 129 memset(&r->velems[0], 0, sizeof(r->velems[0]) * 3); 130 for (i = 0; i < 3; i++) { 131 r->velems[i].src_offset = i * 4 * sizeof(float); 132 r->velems[i].instance_divisor = 0; 133 r->velems[i].vertex_buffer_index = 0; 134 r->velems[i].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; 135 } 136 } 137 138 static inline void 139 add_vertex_color(struct xa_context *r, float x, float y, float color[4]) 140 { 141 float *vertex = r->buffer + r->buffer_size; 142 143 vertex[0] = x; 144 vertex[1] = y; 145 vertex[2] = 0.f; /*z */ 146 vertex[3] = 1.f; /*w */ 147 148 vertex[4] = color[0]; /*r */ 149 vertex[5] = color[1]; /*g */ 150 vertex[6] = color[2]; /*b */ 151 vertex[7] = color[3]; /*a */ 152 153 r->buffer_size += 8; 154 } 155 156 static inline void 157 add_vertex_1tex(struct xa_context *r, float x, float y, float s, float t) 158 { 159 float *vertex = r->buffer + r->buffer_size; 160 161 vertex[0] = x; 162 vertex[1] = y; 163 vertex[2] = 0.f; /*z */ 164 vertex[3] = 1.f; /*w */ 165 166 vertex[4] = s; /*s */ 167 vertex[5] = t; /*t */ 168 vertex[6] = 0.f; /*r */ 169 vertex[7] = 1.f; /*q */ 170 171 r->buffer_size += 8; 172 } 173 174 static inline void 175 add_vertex_2tex(struct xa_context *r, 176 float x, float y, float s0, float t0, float s1, float t1) 177 { 178 float *vertex = r->buffer + r->buffer_size; 179 180 vertex[0] = x; 181 vertex[1] = y; 182 vertex[2] = 0.f; /*z */ 183 vertex[3] = 1.f; /*w */ 184 185 vertex[4] = s0; /*s */ 186 vertex[5] = t0; /*t */ 187 vertex[6] = 0.f; /*r */ 188 vertex[7] = 1.f; /*q */ 189 190 vertex[8] = s1; /*s */ 191 vertex[9] = t1; /*t */ 192 vertex[10] = 0.f; /*r */ 193 vertex[11] = 1.f; /*q */ 194 195 r->buffer_size += 12; 196 } 197 198 static void 199 add_vertex_data1(struct xa_context *r, 200 float srcX, float srcY, float dstX, float dstY, 201 float width, float height, 202 struct pipe_resource *src, const float *src_matrix) 203 { 204 float s0, t0, s1, t1, s2, t2, s3, t3; 205 float pt0[2], pt1[2], pt2[2], pt3[2]; 206 207 pt0[0] = srcX; 208 pt0[1] = srcY; 209 pt1[0] = (srcX + width); 210 pt1[1] = srcY; 211 pt2[0] = (srcX + width); 212 pt2[1] = (srcY + height); 213 pt3[0] = srcX; 214 pt3[1] = (srcY + height); 215 216 if (src_matrix) { 217 map_point((float *)src_matrix, pt0[0], pt0[1], &pt0[0], &pt0[1]); 218 map_point((float *)src_matrix, pt1[0], pt1[1], &pt1[0], &pt1[1]); 219 map_point((float *)src_matrix, pt2[0], pt2[1], &pt2[0], &pt2[1]); 220 map_point((float *)src_matrix, pt3[0], pt3[1], &pt3[0], &pt3[1]); 221 } 222 223 s0 = pt0[0] / src->width0; 224 s1 = pt1[0] / src->width0; 225 s2 = pt2[0] / src->width0; 226 s3 = pt3[0] / src->width0; 227 t0 = pt0[1] / src->height0; 228 t1 = pt1[1] / src->height0; 229 t2 = pt2[1] / src->height0; 230 t3 = pt3[1] / src->height0; 231 232 /* 1st vertex */ 233 add_vertex_1tex(r, dstX, dstY, s0, t0); 234 /* 2nd vertex */ 235 add_vertex_1tex(r, dstX + width, dstY, s1, t1); 236 /* 3rd vertex */ 237 add_vertex_1tex(r, dstX + width, dstY + height, s2, t2); 238 /* 4th vertex */ 239 add_vertex_1tex(r, dstX, dstY + height, s3, t3); 240 } 241 242 static void 243 add_vertex_data2(struct xa_context *r, 244 float srcX, float srcY, float maskX, float maskY, 245 float dstX, float dstY, float width, float height, 246 struct pipe_resource *src, 247 struct pipe_resource *mask, 248 const float *src_matrix, const float *mask_matrix) 249 { 250 float src_s0, src_t0, src_s1, src_t1; 251 float mask_s0, mask_t0, mask_s1, mask_t1; 252 float spt0[2], spt1[2]; 253 float mpt0[2], mpt1[2]; 254 255 spt0[0] = srcX; 256 spt0[1] = srcY; 257 spt1[0] = srcX + width; 258 spt1[1] = srcY + height; 259 260 mpt0[0] = maskX; 261 mpt0[1] = maskY; 262 mpt1[0] = maskX + width; 263 mpt1[1] = maskY + height; 264 265 if (src_matrix) { 266 map_point((float *)src_matrix, spt0[0], spt0[1], &spt0[0], &spt0[1]); 267 map_point((float *)src_matrix, spt1[0], spt1[1], &spt1[0], &spt1[1]); 268 } 269 270 if (mask_matrix) { 271 map_point((float *)mask_matrix, mpt0[0], mpt0[1], &mpt0[0], &mpt0[1]); 272 map_point((float *)mask_matrix, mpt1[0], mpt1[1], &mpt1[0], &mpt1[1]); 273 } 274 275 src_s0 = spt0[0] / src->width0; 276 src_t0 = spt0[1] / src->height0; 277 src_s1 = spt1[0] / src->width0; 278 src_t1 = spt1[1] / src->height0; 279 280 mask_s0 = mpt0[0] / mask->width0; 281 mask_t0 = mpt0[1] / mask->height0; 282 mask_s1 = mpt1[0] / mask->width0; 283 mask_t1 = mpt1[1] / mask->height0; 284 285 /* 1st vertex */ 286 add_vertex_2tex(r, dstX, dstY, 287 src_s0, src_t0, mask_s0, mask_t0); 288 /* 2nd vertex */ 289 add_vertex_2tex(r, dstX + width, dstY, 290 src_s1, src_t0, mask_s1, mask_t0); 291 /* 3rd vertex */ 292 add_vertex_2tex(r, dstX + width, dstY + height, 293 src_s1, src_t1, mask_s1, mask_t1); 294 /* 4th vertex */ 295 add_vertex_2tex(r, dstX, dstY + height, 296 src_s0, src_t1, mask_s0, mask_t1); 297 } 298 299 static void 300 setup_vertex_data_yuv(struct xa_context *r, 301 float srcX, 302 float srcY, 303 float srcW, 304 float srcH, 305 float dstX, 306 float dstY, 307 float dstW, float dstH, struct xa_surface *srf[]) 308 { 309 float s0, t0, s1, t1; 310 float spt0[2], spt1[2]; 311 struct pipe_resource *tex; 312 313 spt0[0] = srcX; 314 spt0[1] = srcY; 315 spt1[0] = srcX + srcW; 316 spt1[1] = srcY + srcH; 317 318 tex = srf[0]->tex; 319 s0 = spt0[0] / tex->width0; 320 t0 = spt0[1] / tex->height0; 321 s1 = spt1[0] / tex->width0; 322 t1 = spt1[1] / tex->height0; 323 324 /* 1st vertex */ 325 add_vertex_1tex(r, dstX, dstY, s0, t0); 326 /* 2nd vertex */ 327 add_vertex_1tex(r, dstX + dstW, dstY, s1, t0); 328 /* 3rd vertex */ 329 add_vertex_1tex(r, dstX + dstW, dstY + dstH, s1, t1); 330 /* 4th vertex */ 331 add_vertex_1tex(r, dstX, dstY + dstH, s0, t1); 332 } 333 334 /* Set up framebuffer, viewport and vertex shader constant buffer 335 * state for a particular destinaton surface. In all our rendering, 336 * these concepts are linked. 337 */ 338 void 339 renderer_bind_destination(struct xa_context *r, 340 struct pipe_surface *surface) 341 { 342 int width = surface->width; 343 int height = surface->height; 344 345 struct pipe_framebuffer_state fb; 346 struct pipe_viewport_state viewport; 347 348 xa_scissor_reset(r); 349 350 /* Framebuffer uses actual surface width/height 351 */ 352 memset(&fb, 0, sizeof fb); 353 fb.width = surface->width; 354 fb.height = surface->height; 355 fb.nr_cbufs = 1; 356 fb.cbufs[0] = surface; 357 fb.zsbuf = 0; 358 359 /* Viewport just touches the bit we're interested in: 360 */ 361 viewport.scale[0] = width / 2.f; 362 viewport.scale[1] = height / 2.f; 363 viewport.scale[2] = 1.0; 364 viewport.translate[0] = width / 2.f; 365 viewport.translate[1] = height / 2.f; 366 viewport.translate[2] = 0.0; 367 368 /* Constant buffer set up to match viewport dimensions: 369 */ 370 if (r->fb_width != width || r->fb_height != height) { 371 float vs_consts[8] = { 372 2.f / width, 2.f / height, 1, 1, 373 -1, -1, 0, 0 374 }; 375 376 r->fb_width = width; 377 r->fb_height = height; 378 379 renderer_set_constants(r, PIPE_SHADER_VERTEX, 380 vs_consts, sizeof vs_consts); 381 } 382 383 cso_set_framebuffer(r->cso, &fb); 384 cso_set_viewport(r->cso, &viewport); 385 } 386 387 void 388 renderer_set_constants(struct xa_context *r, 389 int shader_type, const float *params, int param_bytes) 390 { 391 struct pipe_resource **cbuf = 392 (shader_type == PIPE_SHADER_VERTEX) ? &r->vs_const_buffer : 393 &r->fs_const_buffer; 394 395 pipe_resource_reference(cbuf, NULL); 396 *cbuf = pipe_buffer_create(r->pipe->screen, 397 PIPE_BIND_CONSTANT_BUFFER, PIPE_USAGE_DEFAULT, 398 param_bytes); 399 400 if (*cbuf) { 401 pipe_buffer_write(r->pipe, *cbuf, 0, param_bytes, params); 402 } 403 pipe_set_constant_buffer(r->pipe, shader_type, 0, *cbuf); 404 } 405 406 void 407 renderer_copy_prepare(struct xa_context *r, 408 struct pipe_surface *dst_surface, 409 struct pipe_resource *src_texture, 410 const enum xa_formats src_xa_format, 411 const enum xa_formats dst_xa_format) 412 { 413 struct pipe_context *pipe = r->pipe; 414 struct pipe_screen *screen = pipe->screen; 415 struct xa_shader shader; 416 uint32_t fs_traits = FS_COMPOSITE; 417 418 assert(screen->is_format_supported(screen, dst_surface->format, 419 PIPE_TEXTURE_2D, 0, 420 PIPE_BIND_RENDER_TARGET)); 421 (void)screen; 422 423 renderer_bind_destination(r, dst_surface); 424 425 /* set misc state we care about */ 426 { 427 struct pipe_blend_state blend; 428 429 memset(&blend, 0, sizeof(blend)); 430 blend.rt[0].rgb_src_factor = PIPE_BLENDFACTOR_ONE; 431 blend.rt[0].alpha_src_factor = PIPE_BLENDFACTOR_ONE; 432 blend.rt[0].rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; 433 blend.rt[0].alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; 434 blend.rt[0].colormask = PIPE_MASK_RGBA; 435 cso_set_blend(r->cso, &blend); 436 } 437 438 /* sampler */ 439 { 440 struct pipe_sampler_state sampler; 441 const struct pipe_sampler_state *p_sampler = &sampler; 442 443 memset(&sampler, 0, sizeof(sampler)); 444 sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; 445 sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; 446 sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; 447 sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; 448 sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; 449 sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; 450 sampler.normalized_coords = 1; 451 cso_set_samplers(r->cso, PIPE_SHADER_FRAGMENT, 1, &p_sampler); 452 r->num_bound_samplers = 1; 453 } 454 455 /* texture/sampler view */ 456 { 457 struct pipe_sampler_view templ; 458 struct pipe_sampler_view *src_view; 459 460 u_sampler_view_default_template(&templ, 461 src_texture, src_texture->format); 462 src_view = pipe->create_sampler_view(pipe, src_texture, &templ); 463 cso_set_sampler_views(r->cso, PIPE_SHADER_FRAGMENT, 1, &src_view); 464 pipe_sampler_view_reference(&src_view, NULL); 465 } 466 467 /* shaders */ 468 if (src_texture->format == PIPE_FORMAT_L8_UNORM || 469 src_texture->format == PIPE_FORMAT_R8_UNORM) 470 fs_traits |= FS_SRC_LUMINANCE; 471 if (dst_surface->format == PIPE_FORMAT_L8_UNORM || 472 dst_surface->format == PIPE_FORMAT_R8_UNORM) 473 fs_traits |= FS_DST_LUMINANCE; 474 if (xa_format_a(dst_xa_format) != 0 && 475 xa_format_a(src_xa_format) == 0) 476 fs_traits |= FS_SRC_SET_ALPHA; 477 478 shader = xa_shaders_get(r->shaders, VS_COMPOSITE, fs_traits); 479 cso_set_vertex_shader_handle(r->cso, shader.vs); 480 cso_set_fragment_shader_handle(r->cso, shader.fs); 481 482 r->buffer_size = 0; 483 r->attrs_per_vertex = 2; 484 } 485 486 void 487 renderer_copy(struct xa_context *r, 488 int dx, 489 int dy, 490 int sx, 491 int sy, 492 int width, int height, float src_width, float src_height) 493 { 494 float s0, t0, s1, t1; 495 float x0, y0, x1, y1; 496 497 /* XXX: could put the texcoord scaling calculation into the vertex 498 * shader. 499 */ 500 s0 = sx / src_width; 501 s1 = (sx + width) / src_width; 502 t0 = sy / src_height; 503 t1 = (sy + height) / src_height; 504 505 x0 = dx; 506 x1 = dx + width; 507 y0 = dy; 508 y1 = dy + height; 509 510 /* draw quad */ 511 renderer_draw_conditional(r, 4 * 8); 512 add_vertex_1tex(r, x0, y0, s0, t0); 513 add_vertex_1tex(r, x1, y0, s1, t0); 514 add_vertex_1tex(r, x1, y1, s1, t1); 515 add_vertex_1tex(r, x0, y1, s0, t1); 516 } 517 518 void 519 renderer_draw_yuv(struct xa_context *r, 520 float src_x, 521 float src_y, 522 float src_w, 523 float src_h, 524 int dst_x, 525 int dst_y, int dst_w, int dst_h, struct xa_surface *srf[]) 526 { 527 const int num_attribs = 2; /*pos + tex coord */ 528 529 setup_vertex_data_yuv(r, 530 src_x, src_y, src_w, src_h, 531 dst_x, dst_y, dst_w, dst_h, srf); 532 533 if (!r->scissor_valid) { 534 r->scissor.minx = 0; 535 r->scissor.miny = 0; 536 r->scissor.maxx = r->dst->tex->width0; 537 r->scissor.maxy = r->dst->tex->height0; 538 } 539 540 r->pipe->set_scissor_states(r->pipe, 0, 1, &r->scissor); 541 542 cso_set_vertex_elements(r->cso, num_attribs, r->velems); 543 util_draw_user_vertex_buffer(r->cso, r->buffer, PIPE_PRIM_QUADS, 544 4, /* verts */ 545 num_attribs); /* attribs/vert */ 546 r->buffer_size = 0; 547 548 xa_scissor_reset(r); 549 } 550 551 void 552 renderer_begin_solid(struct xa_context *r) 553 { 554 r->buffer_size = 0; 555 r->attrs_per_vertex = 2; 556 } 557 558 void 559 renderer_solid(struct xa_context *r, 560 int x0, int y0, int x1, int y1, float *color) 561 { 562 /* 563 * debug_printf("solid rect[(%d, %d), (%d, %d)], rgba[%f, %f, %f, %f]\n", 564 * x0, y0, x1, y1, color[0], color[1], color[2], color[3]); */ 565 566 renderer_draw_conditional(r, 4 * 8); 567 568 /* 1st vertex */ 569 add_vertex_color(r, x0, y0, color); 570 /* 2nd vertex */ 571 add_vertex_color(r, x1, y0, color); 572 /* 3rd vertex */ 573 add_vertex_color(r, x1, y1, color); 574 /* 4th vertex */ 575 add_vertex_color(r, x0, y1, color); 576 } 577 578 void 579 renderer_draw_flush(struct xa_context *r) 580 { 581 renderer_draw_conditional(r, 0); 582 } 583 584 void 585 renderer_begin_textures(struct xa_context *r) 586 { 587 r->attrs_per_vertex = 1 + r->num_bound_samplers; 588 r->buffer_size = 0; 589 } 590 591 void 592 renderer_texture(struct xa_context *r, 593 int *pos, 594 int width, int height, 595 const float *src_matrix, 596 const float *mask_matrix) 597 { 598 struct pipe_sampler_view **sampler_view = r->bound_sampler_views; 599 600 #if 0 601 if (src_matrix) { 602 debug_printf("src_matrix = \n"); 603 debug_printf("%f, %f, %f\n", src_matrix[0], src_matrix[1], src_matrix[2]); 604 debug_printf("%f, %f, %f\n", src_matrix[3], src_matrix[4], src_matrix[5]); 605 debug_printf("%f, %f, %f\n", src_matrix[6], src_matrix[7], src_matrix[8]); 606 } 607 if (mask_matrix) { 608 debug_printf("mask_matrix = \n"); 609 debug_printf("%f, %f, %f\n", mask_matrix[0], mask_matrix[1], mask_matrix[2]); 610 debug_printf("%f, %f, %f\n", mask_matrix[3], mask_matrix[4], mask_matrix[5]); 611 debug_printf("%f, %f, %f\n", mask_matrix[6], mask_matrix[7], mask_matrix[8]); 612 } 613 #endif 614 615 switch(r->attrs_per_vertex) { 616 case 2: 617 renderer_draw_conditional(r, 4 * 8); 618 add_vertex_data1(r, 619 pos[0], pos[1], /* src */ 620 pos[4], pos[5], /* dst */ 621 width, height, 622 sampler_view[0]->texture, src_matrix); 623 break; 624 case 3: 625 renderer_draw_conditional(r, 4 * 12); 626 add_vertex_data2(r, 627 pos[0], pos[1], /* src */ 628 pos[2], pos[3], /* mask */ 629 pos[4], pos[5], /* dst */ 630 width, height, 631 sampler_view[0]->texture, sampler_view[1]->texture, 632 src_matrix, mask_matrix); 633 break; 634 default: 635 break; 636 } 637 } 638