1 // SPDX-License-Identifier: MIT 2 // 3 // Copyright 2024 Advanced Micro Devices, Inc. 4 5 #include "dc_spl.h" 6 #include "dc_spl_scl_filters.h" 7 #include "dc_spl_scl_easf_filters.h" 8 #include "dc_spl_isharp_filters.h" 9 #include "spl_debug.h" 10 11 #define IDENTITY_RATIO(ratio) (spl_fixpt_u3d19(ratio) == (1 << 19)) 12 #define MIN_VIEWPORT_SIZE 12 13 14 static bool spl_is_yuv420(enum spl_pixel_format format) 15 { 16 if ((format >= SPL_PIXEL_FORMAT_420BPP8) && 17 (format <= SPL_PIXEL_FORMAT_420BPP10)) 18 return true; 19 20 return false; 21 } 22 23 static bool spl_is_rgb8(enum spl_pixel_format format) 24 { 25 if (format == SPL_PIXEL_FORMAT_ARGB8888) 26 return true; 27 28 return false; 29 } 30 31 static bool spl_is_video_format(enum spl_pixel_format format) 32 { 33 if (format >= SPL_PIXEL_FORMAT_VIDEO_BEGIN 34 && format <= SPL_PIXEL_FORMAT_VIDEO_END) 35 return true; 36 else 37 return false; 38 } 39 40 static bool spl_is_subsampled_format(enum spl_pixel_format format) 41 { 42 if (format >= SPL_PIXEL_FORMAT_SUBSAMPLED_BEGIN 43 && format <= SPL_PIXEL_FORMAT_SUBSAMPLED_END) 44 return true; 45 else 46 return false; 47 } 48 49 static struct spl_rect intersect_rec(const struct spl_rect *r0, const struct spl_rect *r1) 50 { 51 struct spl_rect rec; 52 int r0_x_end = r0->x + r0->width; 53 int r1_x_end = r1->x + r1->width; 54 int r0_y_end = r0->y + r0->height; 55 int r1_y_end = r1->y + r1->height; 56 57 rec.x = r0->x > r1->x ? r0->x : r1->x; 58 rec.width = r0_x_end > r1_x_end ? r1_x_end - rec.x : r0_x_end - rec.x; 59 rec.y = r0->y > r1->y ? r0->y : r1->y; 60 rec.height = r0_y_end > r1_y_end ? r1_y_end - rec.y : r0_y_end - rec.y; 61 62 /* in case that there is no intersection */ 63 if (rec.width < 0 || rec.height < 0) 64 memset(&rec, 0, sizeof(rec)); 65 66 return rec; 67 } 68 69 static struct spl_rect shift_rec(const struct spl_rect *rec_in, int x, int y) 70 { 71 struct spl_rect rec_out = *rec_in; 72 73 rec_out.x += x; 74 rec_out.y += y; 75 76 return rec_out; 77 } 78 79 static void spl_opp_adjust_rect(struct spl_rect *rec, const struct spl_opp_adjust *adjust) 80 { 81 if ((rec->x + adjust->x) >= 0) 82 rec->x += adjust->x; 83 84 if ((rec->y + adjust->y) >= 0) 85 rec->y += adjust->y; 86 87 if ((rec->width + adjust->width) >= 1) 88 rec->width += adjust->width; 89 90 if ((rec->height + adjust->height) >= 1) 91 rec->height += adjust->height; 92 } 93 94 static struct spl_rect calculate_plane_rec_in_timing_active( 95 struct spl_in *spl_in, 96 const struct spl_rect *rec_in) 97 { 98 /* 99 * The following diagram shows an example where we map a 1920x1200 100 * desktop to a 2560x1440 timing with a plane rect in the middle 101 * of the screen. To map a plane rect from Stream Source to Timing 102 * Active space, we first multiply stream scaling ratios (i.e 2304/1920 103 * horizontal and 1440/1200 vertical) to the plane's x and y, then 104 * we add stream destination offsets (i.e 128 horizontal, 0 vertical). 105 * This will give us a plane rect's position in Timing Active. However 106 * we have to remove the fractional. The rule is that we find left/right 107 * and top/bottom positions and round the value to the adjacent integer. 108 * 109 * Stream Source Space 110 * ------------ 111 * __________________________________________________ 112 * |Stream Source (1920 x 1200) ^ | 113 * | y | 114 * | <------- w --------|> | 115 * | __________________V | 116 * |<-- x -->|Plane//////////////| ^ | 117 * | |(pre scale)////////| | | 118 * | |///////////////////| | | 119 * | |///////////////////| h | 120 * | |///////////////////| | | 121 * | |///////////////////| | | 122 * | |///////////////////| V | 123 * | | 124 * | | 125 * |__________________________________________________| 126 * 127 * 128 * Timing Active Space 129 * --------------------------------- 130 * 131 * Timing Active (2560 x 1440) 132 * __________________________________________________ 133 * |*****| Stteam Destination (2304 x 1440) |*****| 134 * |*****| |*****| 135 * |<128>| |*****| 136 * |*****| __________________ |*****| 137 * |*****| |Plane/////////////| |*****| 138 * |*****| |(post scale)//////| |*****| 139 * |*****| |//////////////////| |*****| 140 * |*****| |//////////////////| |*****| 141 * |*****| |//////////////////| |*****| 142 * |*****| |//////////////////| |*****| 143 * |*****| |*****| 144 * |*****| |*****| 145 * |*****| |*****| 146 * |*****|______________________________________|*****| 147 * 148 * So the resulting formulas are shown below: 149 * 150 * recout_x = 128 + round(plane_x * 2304 / 1920) 151 * recout_w = 128 + round((plane_x + plane_w) * 2304 / 1920) - recout_x 152 * recout_y = 0 + round(plane_y * 1440 / 1200) 153 * recout_h = 0 + round((plane_y + plane_h) * 1440 / 1200) - recout_y 154 * 155 * NOTE: fixed point division is not error free. To reduce errors 156 * introduced by fixed point division, we divide only after 157 * multiplication is complete. 158 */ 159 const struct spl_rect *stream_src = &spl_in->basic_out.src_rect; 160 const struct spl_rect *stream_dst = &spl_in->basic_out.dst_rect; 161 struct spl_rect rec_out = {0}; 162 struct spl_fixed31_32 temp; 163 164 165 temp = spl_fixpt_from_fraction(rec_in->x * (long long)stream_dst->width, 166 stream_src->width); 167 rec_out.x = stream_dst->x + spl_fixpt_round(temp); 168 169 temp = spl_fixpt_from_fraction( 170 (rec_in->x + rec_in->width) * (long long)stream_dst->width, 171 stream_src->width); 172 rec_out.width = stream_dst->x + spl_fixpt_round(temp) - rec_out.x; 173 174 temp = spl_fixpt_from_fraction(rec_in->y * (long long)stream_dst->height, 175 stream_src->height); 176 rec_out.y = stream_dst->y + spl_fixpt_round(temp); 177 178 temp = spl_fixpt_from_fraction( 179 (rec_in->y + rec_in->height) * (long long)stream_dst->height, 180 stream_src->height); 181 rec_out.height = stream_dst->y + spl_fixpt_round(temp) - rec_out.y; 182 183 return rec_out; 184 } 185 186 static struct spl_rect calculate_mpc_slice_in_timing_active( 187 struct spl_in *spl_in, 188 struct spl_rect *plane_clip_rec) 189 { 190 bool use_recout_width_aligned = 191 spl_in->basic_in.num_h_slices_recout_width_align.use_recout_width_aligned; 192 int mpc_slice_count = 193 spl_in->basic_in.num_h_slices_recout_width_align.num_slices_recout_width.mpc_num_h_slices; 194 int recout_width_align = 195 spl_in->basic_in.num_h_slices_recout_width_align.num_slices_recout_width.mpc_recout_width_align; 196 int mpc_slice_idx = spl_in->basic_in.mpc_h_slice_index; 197 int epimo = mpc_slice_count - plane_clip_rec->width % mpc_slice_count - 1; 198 struct spl_rect mpc_rec; 199 200 if (use_recout_width_aligned) { 201 mpc_rec.width = recout_width_align; 202 if ((mpc_rec.width * (mpc_slice_idx + 1)) > plane_clip_rec->width) { 203 mpc_rec.width = plane_clip_rec->width % recout_width_align; 204 mpc_rec.x = plane_clip_rec->x + recout_width_align * mpc_slice_idx; 205 } else 206 mpc_rec.x = plane_clip_rec->x + mpc_rec.width * mpc_slice_idx; 207 mpc_rec.height = plane_clip_rec->height; 208 mpc_rec.y = plane_clip_rec->y; 209 210 } else { 211 mpc_rec.width = plane_clip_rec->width / mpc_slice_count; 212 mpc_rec.x = plane_clip_rec->x + mpc_rec.width * mpc_slice_idx; 213 mpc_rec.height = plane_clip_rec->height; 214 mpc_rec.y = plane_clip_rec->y; 215 } 216 SPL_ASSERT(mpc_slice_count == 1 || 217 spl_in->basic_out.view_format != SPL_VIEW_3D_SIDE_BY_SIDE || 218 mpc_rec.width % 2 == 0); 219 220 /* extra pixels in the division remainder need to go to pipes after 221 * the extra pixel index minus one(epimo) defined here as: 222 */ 223 if (mpc_slice_idx > epimo) { 224 mpc_rec.x += mpc_slice_idx - epimo - 1; 225 mpc_rec.width += 1; 226 } 227 228 if (spl_in->basic_out.view_format == SPL_VIEW_3D_TOP_AND_BOTTOM) { 229 SPL_ASSERT(mpc_rec.height % 2 == 0); 230 mpc_rec.height /= 2; 231 } 232 return mpc_rec; 233 } 234 235 static struct spl_rect calculate_odm_slice_in_timing_active(struct spl_in *spl_in) 236 { 237 int odm_slice_count = spl_in->basic_out.odm_combine_factor; 238 int odm_slice_idx = spl_in->odm_slice_index; 239 bool is_last_odm_slice = (odm_slice_idx + 1) == odm_slice_count; 240 int h_active = spl_in->basic_out.output_size.width; 241 int v_active = spl_in->basic_out.output_size.height; 242 int odm_slice_width; 243 struct spl_rect odm_rec; 244 245 if (spl_in->basic_out.odm_combine_factor > 0) { 246 odm_slice_width = h_active / odm_slice_count; 247 /* 248 * deprecated, caller must pass in odm slice rect i.e OPP input 249 * rect in timing active for the new interface. 250 */ 251 if (spl_in->basic_out.use_two_pixels_per_container && (odm_slice_width % 2)) 252 odm_slice_width++; 253 254 odm_rec.x = odm_slice_width * odm_slice_idx; 255 odm_rec.width = is_last_odm_slice ? 256 /* last slice width is the reminder of h_active */ 257 h_active - odm_slice_width * (odm_slice_count - 1) : 258 /* odm slice width is the floor of h_active / count */ 259 odm_slice_width; 260 odm_rec.y = 0; 261 odm_rec.height = v_active; 262 263 return odm_rec; 264 } 265 266 return spl_in->basic_out.odm_slice_rect; 267 } 268 269 static void spl_calculate_recout(struct spl_in *spl_in, struct spl_scratch *spl_scratch, struct spl_out *spl_out) 270 { 271 /* 272 * A plane clip represents the desired plane size and position in Stream 273 * Source Space. Stream Source is the destination where all planes are 274 * blended (i.e. positioned, scaled and overlaid). It is a canvas where 275 * all planes associated with the current stream are drawn together. 276 * After Stream Source is completed, we will further scale and 277 * reposition the entire canvas of the stream source to Stream 278 * Destination in Timing Active Space. This could be due to display 279 * overscan adjustment where we will need to rescale and reposition all 280 * the planes so they can fit into a TV with overscan or downscale 281 * upscale features such as GPU scaling or VSR. 282 * 283 * This two step blending is a virtual procedure in software. In 284 * hardware there is no such thing as Stream Source. all planes are 285 * blended once in Timing Active Space. Software virtualizes a Stream 286 * Source space to decouple the math complicity so scaling param 287 * calculation focuses on one step at a time. 288 * 289 * In the following two diagrams, user applied 10% overscan adjustment 290 * so the Stream Source needs to be scaled down a little before mapping 291 * to Timing Active Space. As a result the Plane Clip is also scaled 292 * down by the same ratio, Plane Clip position (i.e. x and y) with 293 * respect to Stream Source is also scaled down. To map it in Timing 294 * Active Space additional x and y offsets from Stream Destination are 295 * added to Plane Clip as well. 296 * 297 * Stream Source Space 298 * ------------ 299 * __________________________________________________ 300 * |Stream Source (3840 x 2160) ^ | 301 * | y | 302 * | | | 303 * | __________________V | 304 * |<-- x -->|Plane Clip/////////| | 305 * | |(pre scale)////////| | 306 * | |///////////////////| | 307 * | |///////////////////| | 308 * | |///////////////////| | 309 * | |///////////////////| | 310 * | |///////////////////| | 311 * | | 312 * | | 313 * |__________________________________________________| 314 * 315 * 316 * Timing Active Space (3840 x 2160) 317 * --------------------------------- 318 * 319 * Timing Active 320 * __________________________________________________ 321 * | y_____________________________________________ | 322 * |x |Stream Destination (3456 x 1944) | | 323 * | | | | 324 * | | __________________ | | 325 * | | |Plane Clip////////| | | 326 * | | |(post scale)//////| | | 327 * | | |//////////////////| | | 328 * | | |//////////////////| | | 329 * | | |//////////////////| | | 330 * | | |//////////////////| | | 331 * | | | | 332 * | | | | 333 * | |____________________________________________| | 334 * |__________________________________________________| 335 * 336 * 337 * In Timing Active Space a plane clip could be further sliced into 338 * pieces called MPC slices. Each Pipe Context is responsible for 339 * processing only one MPC slice so the plane processing workload can be 340 * distributed to multiple DPP Pipes. MPC slices could be blended 341 * together to a single ODM slice. Each ODM slice is responsible for 342 * processing a portion of Timing Active divided horizontally so the 343 * output pixel processing workload can be distributed to multiple OPP 344 * pipes. All ODM slices are mapped together in ODM block so all MPC 345 * slices belong to different ODM slices could be pieced together to 346 * form a single image in Timing Active. MPC slices must belong to 347 * single ODM slice. If an MPC slice goes across ODM slice boundary, it 348 * needs to be divided into two MPC slices one for each ODM slice. 349 * 350 * In the following diagram the output pixel processing workload is 351 * divided horizontally into two ODM slices one for each OPP blend tree. 352 * OPP0 blend tree is responsible for processing left half of Timing 353 * Active, while OPP2 blend tree is responsible for processing right 354 * half. 355 * 356 * The plane has two MPC slices. However since the right MPC slice goes 357 * across ODM boundary, two DPP pipes are needed one for each OPP blend 358 * tree. (i.e. DPP1 for OPP0 blend tree and DPP2 for OPP2 blend tree). 359 * 360 * Assuming that we have a Pipe Context associated with OPP0 and DPP1 361 * working on processing the plane in the diagram. We want to know the 362 * width and height of the shaded rectangle and its relative position 363 * with respect to the ODM slice0. This is called the recout of the pipe 364 * context. 365 * 366 * Planes can be at arbitrary size and position and there could be an 367 * arbitrary number of MPC and ODM slices. The algorithm needs to take 368 * all scenarios into account. 369 * 370 * Timing Active Space (3840 x 2160) 371 * --------------------------------- 372 * 373 * Timing Active 374 * __________________________________________________ 375 * |OPP0(ODM slice0)^ |OPP2(ODM slice1) | 376 * | y | | 377 * | | <- w -> | 378 * | _____V________|____ | 379 * | |DPP0 ^ |DPP1 |DPP2| | 380 * |<------ x |-----|->|/////| | | 381 * | | | |/////| | | 382 * | | h |/////| | | 383 * | | | |/////| | | 384 * | |_____V__|/////|____| | 385 * | | | 386 * | | | 387 * | | | 388 * |_________________________|________________________| 389 * 390 * 391 */ 392 struct spl_rect plane_clip; 393 struct spl_rect mpc_slice_of_plane_clip; 394 struct spl_rect odm_slice; 395 struct spl_rect overlapping_area; 396 397 plane_clip = calculate_plane_rec_in_timing_active(spl_in, 398 &spl_in->basic_in.clip_rect); 399 /* guard plane clip from drawing beyond stream dst here */ 400 plane_clip = intersect_rec(&plane_clip, 401 &spl_in->basic_out.dst_rect); 402 mpc_slice_of_plane_clip = calculate_mpc_slice_in_timing_active( 403 spl_in, &plane_clip); 404 odm_slice = calculate_odm_slice_in_timing_active(spl_in); 405 overlapping_area = intersect_rec(&mpc_slice_of_plane_clip, &odm_slice); 406 407 if (overlapping_area.height > 0 && 408 overlapping_area.width > 0) { 409 /* shift the overlapping area so it is with respect to current 410 * ODM slice's position 411 */ 412 spl_scratch->scl_data.recout = shift_rec( 413 &overlapping_area, 414 -odm_slice.x, -odm_slice.y); 415 spl_scratch->scl_data.recout.height -= 416 spl_in->debug.visual_confirm_base_offset; 417 spl_scratch->scl_data.recout.height -= 418 spl_in->debug.visual_confirm_dpp_offset; 419 } else 420 /* if there is no overlap, zero recout */ 421 memset(&spl_scratch->scl_data.recout, 0, 422 sizeof(struct spl_rect)); 423 } 424 425 /* Calculate scaling ratios */ 426 static void spl_calculate_scaling_ratios(struct spl_in *spl_in, 427 struct spl_scratch *spl_scratch, 428 struct spl_out *spl_out) 429 { 430 const int in_w = spl_in->basic_out.src_rect.width; 431 const int in_h = spl_in->basic_out.src_rect.height; 432 const int out_w = spl_in->basic_out.dst_rect.width; 433 const int out_h = spl_in->basic_out.dst_rect.height; 434 struct spl_rect surf_src = spl_in->basic_in.src_rect; 435 436 /*Swap surf_src height and width since scaling ratios are in recout rotation*/ 437 if (spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_90 || 438 spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_270) 439 spl_swap(surf_src.height, surf_src.width); 440 441 spl_scratch->scl_data.ratios.horz = spl_fixpt_from_fraction( 442 surf_src.width, 443 spl_in->basic_in.dst_rect.width); 444 spl_scratch->scl_data.ratios.vert = spl_fixpt_from_fraction( 445 surf_src.height, 446 spl_in->basic_in.dst_rect.height); 447 448 if (spl_in->basic_out.view_format == SPL_VIEW_3D_SIDE_BY_SIDE) 449 spl_scratch->scl_data.ratios.horz.value *= 2; 450 else if (spl_in->basic_out.view_format == SPL_VIEW_3D_TOP_AND_BOTTOM) 451 spl_scratch->scl_data.ratios.vert.value *= 2; 452 453 spl_scratch->scl_data.ratios.vert.value = spl_div64_s64( 454 spl_scratch->scl_data.ratios.vert.value * in_h, out_h); 455 spl_scratch->scl_data.ratios.horz.value = spl_div64_s64( 456 spl_scratch->scl_data.ratios.horz.value * in_w, out_w); 457 458 spl_scratch->scl_data.ratios.horz_c = spl_scratch->scl_data.ratios.horz; 459 spl_scratch->scl_data.ratios.vert_c = spl_scratch->scl_data.ratios.vert; 460 461 if (spl_is_yuv420(spl_in->basic_in.format)) { 462 spl_scratch->scl_data.ratios.horz_c.value /= 2; 463 spl_scratch->scl_data.ratios.vert_c.value /= 2; 464 } 465 spl_scratch->scl_data.ratios.horz = spl_fixpt_truncate( 466 spl_scratch->scl_data.ratios.horz, 19); 467 spl_scratch->scl_data.ratios.vert = spl_fixpt_truncate( 468 spl_scratch->scl_data.ratios.vert, 19); 469 spl_scratch->scl_data.ratios.horz_c = spl_fixpt_truncate( 470 spl_scratch->scl_data.ratios.horz_c, 19); 471 spl_scratch->scl_data.ratios.vert_c = spl_fixpt_truncate( 472 spl_scratch->scl_data.ratios.vert_c, 19); 473 474 /* 475 * Coefficient table and some registers are different based on ratio 476 * that is output/input. Currently we calculate input/output 477 * Store 1/ratio in recip_ratio for those lookups 478 */ 479 spl_scratch->scl_data.recip_ratios.horz = spl_fixpt_recip( 480 spl_scratch->scl_data.ratios.horz); 481 spl_scratch->scl_data.recip_ratios.vert = spl_fixpt_recip( 482 spl_scratch->scl_data.ratios.vert); 483 spl_scratch->scl_data.recip_ratios.horz_c = spl_fixpt_recip( 484 spl_scratch->scl_data.ratios.horz_c); 485 spl_scratch->scl_data.recip_ratios.vert_c = spl_fixpt_recip( 486 spl_scratch->scl_data.ratios.vert_c); 487 } 488 489 /* Calculate Viewport size */ 490 static void spl_calculate_viewport_size(struct spl_in *spl_in, struct spl_scratch *spl_scratch) 491 { 492 spl_scratch->scl_data.viewport.width = spl_fixpt_ceil(spl_fixpt_mul_int(spl_scratch->scl_data.ratios.horz, 493 spl_scratch->scl_data.recout.width)); 494 spl_scratch->scl_data.viewport.height = spl_fixpt_ceil(spl_fixpt_mul_int(spl_scratch->scl_data.ratios.vert, 495 spl_scratch->scl_data.recout.height)); 496 spl_scratch->scl_data.viewport_c.width = spl_fixpt_ceil(spl_fixpt_mul_int(spl_scratch->scl_data.ratios.horz_c, 497 spl_scratch->scl_data.recout.width)); 498 spl_scratch->scl_data.viewport_c.height = spl_fixpt_ceil(spl_fixpt_mul_int(spl_scratch->scl_data.ratios.vert_c, 499 spl_scratch->scl_data.recout.height)); 500 if (spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_90 || 501 spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_270) { 502 spl_swap(spl_scratch->scl_data.viewport.width, spl_scratch->scl_data.viewport.height); 503 spl_swap(spl_scratch->scl_data.viewport_c.width, spl_scratch->scl_data.viewport_c.height); 504 } 505 } 506 507 static void spl_get_vp_scan_direction(enum spl_rotation_angle rotation, 508 bool horizontal_mirror, 509 bool *orthogonal_rotation, 510 bool *flip_vert_scan_dir, 511 bool *flip_horz_scan_dir) 512 { 513 *orthogonal_rotation = false; 514 *flip_vert_scan_dir = false; 515 *flip_horz_scan_dir = false; 516 if (rotation == SPL_ROTATION_ANGLE_180) { 517 *flip_vert_scan_dir = true; 518 *flip_horz_scan_dir = true; 519 } else if (rotation == SPL_ROTATION_ANGLE_90) { 520 *orthogonal_rotation = true; 521 *flip_horz_scan_dir = true; 522 } else if (rotation == SPL_ROTATION_ANGLE_270) { 523 *orthogonal_rotation = true; 524 *flip_vert_scan_dir = true; 525 } 526 527 if (horizontal_mirror) 528 *flip_horz_scan_dir = !*flip_horz_scan_dir; 529 } 530 531 /* 532 * We completely calculate vp offset, size and inits here based entirely on scaling 533 * ratios and recout for pixel perfect pipe combine. 534 */ 535 static void spl_calculate_init_and_vp(bool flip_scan_dir, 536 int recout_offset_within_recout_full, 537 int recout_size, 538 int src_size, 539 int taps, 540 struct spl_fixed31_32 ratio, 541 struct spl_fixed31_32 init_adj, 542 struct spl_fixed31_32 *init, 543 int *vp_offset, 544 int *vp_size) 545 { 546 struct spl_fixed31_32 temp; 547 int int_part; 548 549 /* 550 * First of the taps starts sampling pixel number <init_int_part> corresponding to recout 551 * pixel 1. Next recout pixel samples int part of <init + scaling ratio> and so on. 552 * All following calculations are based on this logic. 553 * 554 * Init calculated according to formula: 555 * init = (scaling_ratio + number_of_taps + 1) / 2 556 * init_bot = init + scaling_ratio 557 * to get pixel perfect combine add the fraction from calculating vp offset 558 */ 559 temp = spl_fixpt_mul_int(ratio, recout_offset_within_recout_full); 560 *vp_offset = spl_fixpt_floor(temp); 561 temp.value &= 0xffffffff; 562 *init = spl_fixpt_add(spl_fixpt_div_int(spl_fixpt_add_int(ratio, taps + 1), 2), temp); 563 *init = spl_fixpt_add(*init, init_adj); 564 *init = spl_fixpt_truncate(*init, 19); 565 566 /* 567 * If viewport has non 0 offset and there are more taps than covered by init then 568 * we should decrease the offset and increase init so we are never sampling 569 * outside of viewport. 570 */ 571 int_part = spl_fixpt_floor(*init); 572 if (int_part < taps) { 573 int_part = taps - int_part; 574 if (int_part > *vp_offset) 575 int_part = *vp_offset; 576 *vp_offset -= int_part; 577 *init = spl_fixpt_add_int(*init, int_part); 578 } 579 /* 580 * If taps are sampling outside of viewport at end of recout and there are more pixels 581 * available in the surface we should increase the viewport size, regardless set vp to 582 * only what is used. 583 */ 584 temp = spl_fixpt_add(*init, spl_fixpt_mul_int(ratio, recout_size - 1)); 585 *vp_size = spl_fixpt_floor(temp); 586 if (*vp_size + *vp_offset > src_size) 587 *vp_size = src_size - *vp_offset; 588 589 /* We did all the math assuming we are scanning same direction as display does, 590 * however mirror/rotation changes how vp scans vs how it is offset. If scan direction 591 * is flipped we simply need to calculate offset from the other side of plane. 592 * Note that outside of viewport all scaling hardware works in recout space. 593 */ 594 if (flip_scan_dir) 595 *vp_offset = src_size - *vp_offset - *vp_size; 596 } 597 598 /*Calculate inits and viewport */ 599 static void spl_calculate_inits_and_viewports(struct spl_in *spl_in, 600 struct spl_scratch *spl_scratch) 601 { 602 struct spl_rect src = spl_in->basic_in.src_rect; 603 struct spl_rect recout_dst_in_active_timing; 604 struct spl_rect recout_clip_in_active_timing; 605 struct spl_rect recout_clip_in_recout_dst; 606 struct spl_rect overlap_in_active_timing; 607 struct spl_rect odm_slice = calculate_odm_slice_in_timing_active(spl_in); 608 int vpc_div = spl_is_subsampled_format(spl_in->basic_in.format) ? 2 : 1; 609 bool orthogonal_rotation, flip_vert_scan_dir, flip_horz_scan_dir; 610 struct spl_fixed31_32 init_adj_h = spl_fixpt_zero; 611 struct spl_fixed31_32 init_adj_v = spl_fixpt_zero; 612 613 recout_clip_in_active_timing = shift_rec( 614 &spl_scratch->scl_data.recout, odm_slice.x, odm_slice.y); 615 recout_dst_in_active_timing = calculate_plane_rec_in_timing_active( 616 spl_in, &spl_in->basic_in.dst_rect); 617 overlap_in_active_timing = intersect_rec(&recout_clip_in_active_timing, 618 &recout_dst_in_active_timing); 619 if (overlap_in_active_timing.width > 0 && 620 overlap_in_active_timing.height > 0) 621 recout_clip_in_recout_dst = shift_rec(&overlap_in_active_timing, 622 -recout_dst_in_active_timing.x, 623 -recout_dst_in_active_timing.y); 624 else 625 memset(&recout_clip_in_recout_dst, 0, sizeof(struct spl_rect)); 626 /* 627 * Work in recout rotation since that requires less transformations 628 */ 629 spl_get_vp_scan_direction( 630 spl_in->basic_in.rotation, 631 spl_in->basic_in.horizontal_mirror, 632 &orthogonal_rotation, 633 &flip_vert_scan_dir, 634 &flip_horz_scan_dir); 635 636 if (spl_is_subsampled_format(spl_in->basic_in.format)) { 637 /* this gives the direction of the cositing (negative will move 638 * left, right otherwise) 639 */ 640 int sign = 1; 641 642 switch (spl_in->basic_in.cositing) { 643 644 case CHROMA_COSITING_TOPLEFT: 645 init_adj_h = spl_fixpt_from_fraction(sign, 4); 646 init_adj_v = spl_fixpt_from_fraction(sign, 4); 647 break; 648 case CHROMA_COSITING_LEFT: 649 init_adj_h = spl_fixpt_from_fraction(sign, 4); 650 init_adj_v = spl_fixpt_zero; 651 break; 652 case CHROMA_COSITING_NONE: 653 default: 654 init_adj_h = spl_fixpt_zero; 655 init_adj_v = spl_fixpt_zero; 656 break; 657 } 658 } 659 660 if (orthogonal_rotation) { 661 spl_swap(src.width, src.height); 662 spl_swap(flip_vert_scan_dir, flip_horz_scan_dir); 663 spl_swap(init_adj_h, init_adj_v); 664 } 665 666 spl_calculate_init_and_vp( 667 flip_horz_scan_dir, 668 recout_clip_in_recout_dst.x, 669 spl_scratch->scl_data.recout.width, 670 src.width, 671 spl_scratch->scl_data.taps.h_taps, 672 spl_scratch->scl_data.ratios.horz, 673 spl_fixpt_zero, 674 &spl_scratch->scl_data.inits.h, 675 &spl_scratch->scl_data.viewport.x, 676 &spl_scratch->scl_data.viewport.width); 677 spl_calculate_init_and_vp( 678 flip_horz_scan_dir, 679 recout_clip_in_recout_dst.x, 680 spl_scratch->scl_data.recout.width, 681 src.width / vpc_div, 682 spl_scratch->scl_data.taps.h_taps_c, 683 spl_scratch->scl_data.ratios.horz_c, 684 init_adj_h, 685 &spl_scratch->scl_data.inits.h_c, 686 &spl_scratch->scl_data.viewport_c.x, 687 &spl_scratch->scl_data.viewport_c.width); 688 spl_calculate_init_and_vp( 689 flip_vert_scan_dir, 690 recout_clip_in_recout_dst.y, 691 spl_scratch->scl_data.recout.height, 692 src.height, 693 spl_scratch->scl_data.taps.v_taps, 694 spl_scratch->scl_data.ratios.vert, 695 spl_fixpt_zero, 696 &spl_scratch->scl_data.inits.v, 697 &spl_scratch->scl_data.viewport.y, 698 &spl_scratch->scl_data.viewport.height); 699 spl_calculate_init_and_vp( 700 flip_vert_scan_dir, 701 recout_clip_in_recout_dst.y, 702 spl_scratch->scl_data.recout.height, 703 src.height / vpc_div, 704 spl_scratch->scl_data.taps.v_taps_c, 705 spl_scratch->scl_data.ratios.vert_c, 706 init_adj_v, 707 &spl_scratch->scl_data.inits.v_c, 708 &spl_scratch->scl_data.viewport_c.y, 709 &spl_scratch->scl_data.viewport_c.height); 710 if (orthogonal_rotation) { 711 spl_swap(spl_scratch->scl_data.viewport.x, spl_scratch->scl_data.viewport.y); 712 spl_swap(spl_scratch->scl_data.viewport.width, spl_scratch->scl_data.viewport.height); 713 spl_swap(spl_scratch->scl_data.viewport_c.x, spl_scratch->scl_data.viewport_c.y); 714 spl_swap(spl_scratch->scl_data.viewport_c.width, spl_scratch->scl_data.viewport_c.height); 715 } 716 spl_scratch->scl_data.viewport.x += src.x; 717 spl_scratch->scl_data.viewport.y += src.y; 718 SPL_ASSERT(src.x % vpc_div == 0 && src.y % vpc_div == 0); 719 spl_scratch->scl_data.viewport_c.x += src.x / vpc_div; 720 spl_scratch->scl_data.viewport_c.y += src.y / vpc_div; 721 } 722 723 static void spl_handle_3d_recout(struct spl_in *spl_in, struct spl_rect *recout) 724 { 725 /* 726 * Handle side by side and top bottom 3d recout offsets after vp calculation 727 * since 3d is special and needs to calculate vp as if there is no recout offset 728 * This may break with rotation, good thing we aren't mixing hw rotation and 3d 729 */ 730 if (spl_in->basic_in.mpc_h_slice_index) { 731 SPL_ASSERT(spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_0 || 732 (spl_in->basic_out.view_format != SPL_VIEW_3D_TOP_AND_BOTTOM && 733 spl_in->basic_out.view_format != SPL_VIEW_3D_SIDE_BY_SIDE)); 734 if (spl_in->basic_out.view_format == SPL_VIEW_3D_TOP_AND_BOTTOM) 735 recout->y += recout->height; 736 else if (spl_in->basic_out.view_format == SPL_VIEW_3D_SIDE_BY_SIDE) 737 recout->x += recout->width; 738 } 739 } 740 741 static void spl_clamp_viewport(struct spl_rect *viewport, int min_viewport_size) 742 { 743 if (min_viewport_size == 0) 744 min_viewport_size = MIN_VIEWPORT_SIZE; 745 /* Clamp minimum viewport size */ 746 if (viewport->height < min_viewport_size) 747 viewport->height = min_viewport_size; 748 if (viewport->width < min_viewport_size) 749 viewport->width = min_viewport_size; 750 } 751 752 static enum scl_mode spl_get_dscl_mode(const struct spl_in *spl_in, 753 const struct spl_scaler_data *data, 754 bool enable_isharp, bool enable_easf) 755 { 756 const long long one = spl_fixpt_one.value; 757 enum spl_pixel_format pixel_format = spl_in->basic_in.format; 758 759 /* Bypass if ratio is 1:1 with no ISHARP or force scale on */ 760 if (data->ratios.horz.value == one 761 && data->ratios.vert.value == one 762 && data->ratios.horz_c.value == one 763 && data->ratios.vert_c.value == one 764 && !spl_in->basic_out.always_scale 765 && !enable_isharp) 766 return SCL_MODE_SCALING_444_BYPASS; 767 768 if (!spl_is_subsampled_format(pixel_format)) { 769 if (spl_is_video_format(pixel_format)) 770 return SCL_MODE_SCALING_444_YCBCR_ENABLE; 771 else 772 return SCL_MODE_SCALING_444_RGB_ENABLE; 773 } 774 775 /* 776 * Bypass YUV if Y is 1:1 with no ISHARP 777 * Do not bypass UV at 1:1 for cositing to be applied 778 */ 779 if (!enable_isharp) { 780 if (data->ratios.horz.value == one && data->ratios.vert.value == one) 781 return SCL_MODE_SCALING_420_LUMA_BYPASS; 782 } 783 784 return SCL_MODE_SCALING_420_YCBCR_ENABLE; 785 } 786 787 static void spl_choose_lls_policy(enum spl_pixel_format format, 788 enum linear_light_scaling *lls_pref) 789 { 790 if (spl_is_subsampled_format(format)) 791 *lls_pref = LLS_PREF_NO; 792 else /* RGB or YUV444 */ 793 *lls_pref = LLS_PREF_YES; 794 } 795 796 /* Enable EASF ?*/ 797 static bool enable_easf(struct spl_in *spl_in, struct spl_scratch *spl_scratch) 798 { 799 int vratio = 0; 800 int hratio = 0; 801 bool skip_easf = false; 802 803 if (spl_in->disable_easf) 804 skip_easf = true; 805 806 vratio = spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert); 807 hratio = spl_fixpt_ceil(spl_scratch->scl_data.ratios.horz); 808 809 /* 810 * No EASF support for downscaling > 2:1 811 * EASF support for upscaling or downscaling up to 2:1 812 */ 813 if ((vratio > 2) || (hratio > 2)) 814 skip_easf = true; 815 816 /* 817 * If lls_pref is LLS_PREF_DONT_CARE, then use pixel format 818 * to determine whether to use LINEAR or NONLINEAR scaling 819 */ 820 if (spl_in->lls_pref == LLS_PREF_DONT_CARE) 821 spl_choose_lls_policy(spl_in->basic_in.format, 822 &spl_in->lls_pref); 823 824 /* Check for linear scaling or EASF preferred */ 825 if (spl_in->lls_pref != LLS_PREF_YES && !spl_in->prefer_easf) 826 skip_easf = true; 827 828 return skip_easf; 829 } 830 831 /* Check if video is in fullscreen mode */ 832 static bool spl_is_video_fullscreen(struct spl_in *spl_in) 833 { 834 if (spl_is_video_format(spl_in->basic_in.format) && spl_in->is_fullscreen) 835 return true; 836 return false; 837 } 838 839 static bool spl_get_isharp_en(struct spl_in *spl_in, 840 struct spl_scratch *spl_scratch) 841 { 842 bool enable_isharp = false; 843 int vratio = 0; 844 int hratio = 0; 845 struct spl_taps taps = spl_scratch->scl_data.taps; 846 bool fullscreen = spl_is_video_fullscreen(spl_in); 847 848 /* Return if adaptive sharpness is disabled */ 849 if (spl_in->adaptive_sharpness.enable == false) 850 return enable_isharp; 851 852 vratio = spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert); 853 hratio = spl_fixpt_ceil(spl_scratch->scl_data.ratios.horz); 854 855 /* No iSHARP support for downscaling */ 856 if (vratio > 1 || hratio > 1) 857 return enable_isharp; 858 859 // Scaling is up to 1:1 (no scaling) or upscaling 860 861 /* 862 * Apply sharpness to RGB and YUV (NV12/P010) 863 * surfaces based on policy setting 864 */ 865 if (!spl_is_video_format(spl_in->basic_in.format) && 866 (spl_in->sharpen_policy == SHARPEN_YUV)) 867 return enable_isharp; 868 else if ((spl_is_video_format(spl_in->basic_in.format) && !fullscreen) && 869 (spl_in->sharpen_policy == SHARPEN_RGB_FULLSCREEN_YUV)) 870 return enable_isharp; 871 else if (!spl_in->is_fullscreen && 872 spl_in->sharpen_policy == SHARPEN_FULLSCREEN_ALL) 873 return enable_isharp; 874 875 /* 876 * Apply sharpness if supports horizontal taps 4,6 AND 877 * vertical taps 3, 4, 6 878 */ 879 if ((taps.h_taps == 4 || taps.h_taps == 6) && 880 (taps.v_taps == 3 || taps.v_taps == 4 || taps.v_taps == 6)) 881 enable_isharp = true; 882 883 return enable_isharp; 884 } 885 886 /* Calculate number of tap with adaptive scaling off */ 887 static void spl_get_taps_non_adaptive_scaler( 888 struct spl_scratch *spl_scratch, const struct spl_taps *in_taps) 889 { 890 bool check_max_downscale = false; 891 892 if (in_taps->h_taps == 0) { 893 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.horz) > 1) 894 spl_scratch->scl_data.taps.h_taps = spl_min(2 * spl_fixpt_ceil( 895 spl_scratch->scl_data.ratios.horz), 8); 896 else 897 spl_scratch->scl_data.taps.h_taps = 4; 898 } else 899 spl_scratch->scl_data.taps.h_taps = in_taps->h_taps; 900 901 if (in_taps->v_taps == 0) { 902 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert) > 1) 903 spl_scratch->scl_data.taps.v_taps = spl_min(2 * spl_fixpt_ceil( 904 spl_scratch->scl_data.ratios.vert), 8); 905 else 906 spl_scratch->scl_data.taps.v_taps = 4; 907 } else 908 spl_scratch->scl_data.taps.v_taps = in_taps->v_taps; 909 910 if (in_taps->v_taps_c == 0) { 911 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert_c) > 1) 912 spl_scratch->scl_data.taps.v_taps_c = spl_min(2 * spl_fixpt_ceil( 913 spl_scratch->scl_data.ratios.vert_c), 8); 914 else 915 spl_scratch->scl_data.taps.v_taps_c = 4; 916 } else 917 spl_scratch->scl_data.taps.v_taps_c = in_taps->v_taps_c; 918 919 if (in_taps->h_taps_c == 0) { 920 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.horz_c) > 1) 921 spl_scratch->scl_data.taps.h_taps_c = spl_min(2 * spl_fixpt_ceil( 922 spl_scratch->scl_data.ratios.horz_c), 8); 923 else 924 spl_scratch->scl_data.taps.h_taps_c = 4; 925 } else if ((in_taps->h_taps_c % 2) != 0 && in_taps->h_taps_c != 1) 926 /* Only 1 and even h_taps_c are supported by hw */ 927 spl_scratch->scl_data.taps.h_taps_c = in_taps->h_taps_c - 1; 928 else 929 spl_scratch->scl_data.taps.h_taps_c = in_taps->h_taps_c; 930 931 932 /* 933 * Max downscale supported is 6.0x. Add ASSERT to catch if go beyond that 934 */ 935 check_max_downscale = spl_fixpt_le(spl_scratch->scl_data.ratios.horz, 936 spl_fixpt_from_fraction(6, 1)); 937 SPL_ASSERT(check_max_downscale); 938 check_max_downscale = spl_fixpt_le(spl_scratch->scl_data.ratios.vert, 939 spl_fixpt_from_fraction(6, 1)); 940 SPL_ASSERT(check_max_downscale); 941 check_max_downscale = spl_fixpt_le(spl_scratch->scl_data.ratios.horz_c, 942 spl_fixpt_from_fraction(6, 1)); 943 SPL_ASSERT(check_max_downscale); 944 check_max_downscale = spl_fixpt_le(spl_scratch->scl_data.ratios.vert_c, 945 spl_fixpt_from_fraction(6, 1)); 946 SPL_ASSERT(check_max_downscale); 947 948 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz)) 949 spl_scratch->scl_data.taps.h_taps = 1; 950 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert)) 951 spl_scratch->scl_data.taps.v_taps = 1; 952 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz_c)) 953 spl_scratch->scl_data.taps.h_taps_c = 1; 954 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert_c)) 955 spl_scratch->scl_data.taps.v_taps_c = 1; 956 957 } 958 959 /* Calculate optimal number of taps */ 960 static bool spl_get_optimal_number_of_taps( 961 int max_downscale_src_width, struct spl_in *spl_in, struct spl_scratch *spl_scratch, 962 const struct spl_taps *in_taps, bool *enable_easf_v, bool *enable_easf_h, 963 bool *enable_isharp) 964 { 965 int num_part_y, num_part_c; 966 unsigned int max_taps_y, max_taps_c; 967 unsigned int min_taps_y, min_taps_c; 968 enum lb_memory_config lb_config; 969 bool skip_easf = false; 970 bool is_subsampled = spl_is_subsampled_format(spl_in->basic_in.format); 971 972 if (spl_scratch->scl_data.viewport.width > spl_scratch->scl_data.h_active && 973 max_downscale_src_width != 0 && 974 spl_scratch->scl_data.viewport.width > max_downscale_src_width) { 975 spl_get_taps_non_adaptive_scaler(spl_scratch, in_taps); 976 *enable_easf_v = false; 977 *enable_easf_h = false; 978 *enable_isharp = false; 979 return false; 980 } 981 982 /* Disable adaptive scaler and sharpener when integer scaling is enabled */ 983 if (spl_in->scaling_quality.integer_scaling) { 984 spl_get_taps_non_adaptive_scaler(spl_scratch, in_taps); 985 *enable_easf_v = false; 986 *enable_easf_h = false; 987 *enable_isharp = false; 988 return true; 989 } 990 991 /* Check if we are using EASF or not */ 992 skip_easf = enable_easf(spl_in, spl_scratch); 993 994 /* 995 * Set default taps if none are provided 996 * From programming guide: taps = min{ ceil(2*H_RATIO,1), 8} for downscaling 997 * taps = 4 for upscaling 998 */ 999 if (skip_easf) 1000 spl_get_taps_non_adaptive_scaler(spl_scratch, in_taps); 1001 else { 1002 if (spl_is_video_format(spl_in->basic_in.format)) { 1003 spl_scratch->scl_data.taps.h_taps = 6; 1004 spl_scratch->scl_data.taps.v_taps = 6; 1005 spl_scratch->scl_data.taps.h_taps_c = 4; 1006 spl_scratch->scl_data.taps.v_taps_c = 4; 1007 } else { /* RGB */ 1008 spl_scratch->scl_data.taps.h_taps = 6; 1009 spl_scratch->scl_data.taps.v_taps = 6; 1010 spl_scratch->scl_data.taps.h_taps_c = 6; 1011 spl_scratch->scl_data.taps.v_taps_c = 6; 1012 } 1013 } 1014 1015 /*Ensure we can support the requested number of vtaps*/ 1016 min_taps_y = spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert); 1017 min_taps_c = spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert_c); 1018 1019 /* Use LB_MEMORY_CONFIG_3 for 4:2:0 */ 1020 if (spl_is_yuv420(spl_in->basic_in.format)) 1021 lb_config = LB_MEMORY_CONFIG_3; 1022 else 1023 lb_config = LB_MEMORY_CONFIG_0; 1024 // Determine max vtap support by calculating how much line buffer can fit 1025 spl_in->callbacks.spl_calc_lb_num_partitions(spl_in->basic_out.alpha_en, &spl_scratch->scl_data, 1026 lb_config, &num_part_y, &num_part_c); 1027 /* MAX_V_TAPS = MIN (NUM_LINES - MAX(CEILING(V_RATIO,1)-2, 0), 8) */ 1028 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert) > 2) 1029 if ((spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert) - 2) > num_part_y) 1030 max_taps_y = 0; 1031 else 1032 max_taps_y = num_part_y - (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert) - 2); 1033 else 1034 max_taps_y = num_part_y; 1035 1036 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert_c) > 2) 1037 if ((spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert_c) - 2) > num_part_c) 1038 max_taps_c = 0; 1039 else 1040 max_taps_c = num_part_c - (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert_c) - 2); 1041 else 1042 max_taps_c = num_part_c; 1043 1044 if (max_taps_y < min_taps_y) 1045 return false; 1046 else if (max_taps_c < min_taps_c) 1047 return false; 1048 1049 if (spl_scratch->scl_data.taps.v_taps > max_taps_y) 1050 spl_scratch->scl_data.taps.v_taps = max_taps_y; 1051 1052 if (spl_scratch->scl_data.taps.v_taps_c > max_taps_c) 1053 spl_scratch->scl_data.taps.v_taps_c = max_taps_c; 1054 1055 if (!skip_easf) { 1056 /* 1057 * RGB ( L + NL ) and Linear HDR support 6x6, 6x4, 6x3, 4x4, 4x3 1058 * NL YUV420 only supports 6x6, 6x4 for Y and 4x4 for UV 1059 * 1060 * If LB does not support 3, 4, or 6 taps, then disable EASF_V 1061 * and only enable EASF_H. So for RGB, support 6x2, 4x2 1062 * and for NL YUV420, support 6x2 for Y and 4x2 for UV 1063 * 1064 * All other cases, have to disable EASF_V and EASF_H 1065 * 1066 * If optimal no of taps is 5, then set it to 4 1067 * If optimal no of taps is 7 or 8, then fine since max tap is 6 1068 * 1069 */ 1070 if (spl_scratch->scl_data.taps.v_taps == 5) 1071 spl_scratch->scl_data.taps.v_taps = 4; 1072 1073 if (spl_scratch->scl_data.taps.v_taps_c == 5) 1074 spl_scratch->scl_data.taps.v_taps_c = 4; 1075 1076 if (spl_scratch->scl_data.taps.h_taps == 5) 1077 spl_scratch->scl_data.taps.h_taps = 4; 1078 1079 if (spl_scratch->scl_data.taps.h_taps_c == 5) 1080 spl_scratch->scl_data.taps.h_taps_c = 4; 1081 1082 if (spl_is_video_format(spl_in->basic_in.format)) { 1083 if (spl_scratch->scl_data.taps.h_taps <= 4) { 1084 *enable_easf_v = false; 1085 *enable_easf_h = false; 1086 } else if (spl_scratch->scl_data.taps.v_taps <= 3) { 1087 *enable_easf_v = false; 1088 *enable_easf_h = true; 1089 } else { 1090 *enable_easf_v = true; 1091 *enable_easf_h = true; 1092 } 1093 SPL_ASSERT((spl_scratch->scl_data.taps.v_taps > 1) && 1094 (spl_scratch->scl_data.taps.v_taps_c > 1)); 1095 } else { /* RGB */ 1096 if (spl_scratch->scl_data.taps.h_taps <= 3) { 1097 *enable_easf_v = false; 1098 *enable_easf_h = false; 1099 } else if (spl_scratch->scl_data.taps.v_taps < 3) { 1100 *enable_easf_v = false; 1101 *enable_easf_h = true; 1102 } else { 1103 *enable_easf_v = true; 1104 *enable_easf_h = true; 1105 } 1106 SPL_ASSERT(spl_scratch->scl_data.taps.v_taps > 1); 1107 } 1108 } else { 1109 *enable_easf_v = false; 1110 *enable_easf_h = false; 1111 } // end of if prefer_easf 1112 1113 /* Sharpener requires scaler to be enabled, including for 1:1 1114 * Check if ISHARP can be enabled 1115 * If ISHARP is not enabled, set taps to 1 if ratio is 1:1 1116 * except for chroma taps. Keep previous taps so it can 1117 * handle cositing 1118 */ 1119 1120 *enable_isharp = spl_get_isharp_en(spl_in, spl_scratch); 1121 if (!*enable_isharp && !spl_in->basic_out.always_scale) { 1122 if ((IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz)) && 1123 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert))) { 1124 spl_scratch->scl_data.taps.h_taps = 1; 1125 spl_scratch->scl_data.taps.v_taps = 1; 1126 1127 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz_c) && !is_subsampled) 1128 spl_scratch->scl_data.taps.h_taps_c = 1; 1129 1130 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert_c) && !is_subsampled) 1131 spl_scratch->scl_data.taps.v_taps_c = 1; 1132 1133 *enable_easf_v = false; 1134 *enable_easf_h = false; 1135 } else { 1136 if ((!*enable_easf_h) && 1137 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz))) 1138 spl_scratch->scl_data.taps.h_taps = 1; 1139 1140 if ((!*enable_easf_v) && 1141 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert))) 1142 spl_scratch->scl_data.taps.v_taps = 1; 1143 1144 if ((!*enable_easf_h) && !is_subsampled && 1145 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz_c))) 1146 spl_scratch->scl_data.taps.h_taps_c = 1; 1147 1148 if ((!*enable_easf_v) && !is_subsampled && 1149 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert_c))) 1150 spl_scratch->scl_data.taps.v_taps_c = 1; 1151 } 1152 } 1153 return true; 1154 } 1155 1156 static void spl_set_black_color_data(enum spl_pixel_format format, 1157 struct scl_black_color *scl_black_color) 1158 { 1159 bool ycbcr = spl_is_video_format(format); 1160 if (ycbcr) { 1161 scl_black_color->offset_rgb_y = BLACK_OFFSET_RGB_Y; 1162 scl_black_color->offset_rgb_cbcr = BLACK_OFFSET_CBCR; 1163 } else { 1164 scl_black_color->offset_rgb_y = 0x0; 1165 scl_black_color->offset_rgb_cbcr = 0x0; 1166 } 1167 } 1168 1169 static void spl_set_manual_ratio_init_data(struct dscl_prog_data *dscl_prog_data, 1170 const struct spl_scaler_data *scl_data) 1171 { 1172 struct spl_fixed31_32 bot; 1173 1174 dscl_prog_data->ratios.h_scale_ratio = spl_fixpt_u3d19(scl_data->ratios.horz) << 5; 1175 dscl_prog_data->ratios.v_scale_ratio = spl_fixpt_u3d19(scl_data->ratios.vert) << 5; 1176 dscl_prog_data->ratios.h_scale_ratio_c = spl_fixpt_u3d19(scl_data->ratios.horz_c) << 5; 1177 dscl_prog_data->ratios.v_scale_ratio_c = spl_fixpt_u3d19(scl_data->ratios.vert_c) << 5; 1178 /* 1179 * 0.24 format for fraction, first five bits zeroed 1180 */ 1181 dscl_prog_data->init.h_filter_init_frac = 1182 spl_fixpt_u0d19(scl_data->inits.h) << 5; 1183 dscl_prog_data->init.h_filter_init_int = 1184 spl_fixpt_floor(scl_data->inits.h); 1185 dscl_prog_data->init.h_filter_init_frac_c = 1186 spl_fixpt_u0d19(scl_data->inits.h_c) << 5; 1187 dscl_prog_data->init.h_filter_init_int_c = 1188 spl_fixpt_floor(scl_data->inits.h_c); 1189 dscl_prog_data->init.v_filter_init_frac = 1190 spl_fixpt_u0d19(scl_data->inits.v) << 5; 1191 dscl_prog_data->init.v_filter_init_int = 1192 spl_fixpt_floor(scl_data->inits.v); 1193 dscl_prog_data->init.v_filter_init_frac_c = 1194 spl_fixpt_u0d19(scl_data->inits.v_c) << 5; 1195 dscl_prog_data->init.v_filter_init_int_c = 1196 spl_fixpt_floor(scl_data->inits.v_c); 1197 1198 bot = spl_fixpt_add(scl_data->inits.v, scl_data->ratios.vert); 1199 dscl_prog_data->init.v_filter_init_bot_frac = spl_fixpt_u0d19(bot) << 5; 1200 dscl_prog_data->init.v_filter_init_bot_int = spl_fixpt_floor(bot); 1201 bot = spl_fixpt_add(scl_data->inits.v_c, scl_data->ratios.vert_c); 1202 dscl_prog_data->init.v_filter_init_bot_frac_c = spl_fixpt_u0d19(bot) << 5; 1203 dscl_prog_data->init.v_filter_init_bot_int_c = spl_fixpt_floor(bot); 1204 } 1205 1206 static void spl_set_taps_data(struct dscl_prog_data *dscl_prog_data, 1207 const struct spl_scaler_data *scl_data) 1208 { 1209 dscl_prog_data->taps.v_taps = scl_data->taps.v_taps - 1; 1210 dscl_prog_data->taps.h_taps = scl_data->taps.h_taps - 1; 1211 dscl_prog_data->taps.v_taps_c = scl_data->taps.v_taps_c - 1; 1212 dscl_prog_data->taps.h_taps_c = scl_data->taps.h_taps_c - 1; 1213 } 1214 1215 /* Populate dscl prog data structure from scaler data calculated by SPL */ 1216 static void spl_set_dscl_prog_data(struct spl_in *spl_in, struct spl_scratch *spl_scratch, 1217 struct spl_out *spl_out, bool enable_easf_v, bool enable_easf_h, bool enable_isharp) 1218 { 1219 struct dscl_prog_data *dscl_prog_data = spl_out->dscl_prog_data; 1220 1221 const struct spl_scaler_data *data = &spl_scratch->scl_data; 1222 1223 struct scl_black_color *scl_black_color = &dscl_prog_data->scl_black_color; 1224 1225 bool enable_easf = enable_easf_v || enable_easf_h; 1226 1227 // Set values for recout 1228 dscl_prog_data->recout = spl_scratch->scl_data.recout; 1229 // Set values for MPC Size 1230 dscl_prog_data->mpc_size.width = spl_scratch->scl_data.h_active; 1231 dscl_prog_data->mpc_size.height = spl_scratch->scl_data.v_active; 1232 1233 // SCL_MODE - Set SCL_MODE data 1234 dscl_prog_data->dscl_mode = spl_get_dscl_mode(spl_in, data, enable_isharp, 1235 enable_easf); 1236 1237 // SCL_BLACK_COLOR 1238 spl_set_black_color_data(spl_in->basic_in.format, scl_black_color); 1239 1240 /* Manually calculate scale ratio and init values */ 1241 spl_set_manual_ratio_init_data(dscl_prog_data, data); 1242 1243 // Set HTaps/VTaps 1244 spl_set_taps_data(dscl_prog_data, data); 1245 // Set viewport 1246 dscl_prog_data->viewport = spl_scratch->scl_data.viewport; 1247 // Set viewport_c 1248 dscl_prog_data->viewport_c = spl_scratch->scl_data.viewport_c; 1249 // Set filters data 1250 spl_set_filters_data(dscl_prog_data, data, enable_easf_v, enable_easf_h); 1251 } 1252 1253 /* Calculate C0-C3 coefficients based on HDR_mult */ 1254 static void spl_calculate_c0_c3_hdr(struct dscl_prog_data *dscl_prog_data, uint32_t sdr_white_level_nits) 1255 { 1256 struct spl_fixed31_32 hdr_mult, c0_mult, c1_mult, c2_mult; 1257 struct spl_fixed31_32 c0_calc, c1_calc, c2_calc; 1258 struct spl_custom_float_format fmt; 1259 uint32_t hdr_multx100_int; 1260 1261 if ((sdr_white_level_nits >= 80) && (sdr_white_level_nits <= 480)) 1262 hdr_multx100_int = sdr_white_level_nits * 100 / 80; 1263 else 1264 hdr_multx100_int = 100; /* default for 80 nits otherwise */ 1265 1266 hdr_mult = spl_fixpt_from_fraction((long long)hdr_multx100_int, 100LL); 1267 c0_mult = spl_fixpt_from_fraction(2126LL, 10000LL); 1268 c1_mult = spl_fixpt_from_fraction(7152LL, 10000LL); 1269 c2_mult = spl_fixpt_from_fraction(722LL, 10000LL); 1270 1271 c0_calc = spl_fixpt_mul(hdr_mult, spl_fixpt_mul(c0_mult, spl_fixpt_from_fraction( 1272 16384LL, 125LL))); 1273 c1_calc = spl_fixpt_mul(hdr_mult, spl_fixpt_mul(c1_mult, spl_fixpt_from_fraction( 1274 16384LL, 125LL))); 1275 c2_calc = spl_fixpt_mul(hdr_mult, spl_fixpt_mul(c2_mult, spl_fixpt_from_fraction( 1276 16384LL, 125LL))); 1277 1278 fmt.exponenta_bits = 5; 1279 fmt.mantissa_bits = 10; 1280 fmt.sign = true; 1281 1282 // fp1.5.10, C0 coefficient (LN_rec709: HDR_MULT * 0.212600 * 2^14/125) 1283 spl_convert_to_custom_float_format(c0_calc, &fmt, &dscl_prog_data->easf_matrix_c0); 1284 // fp1.5.10, C1 coefficient (LN_rec709: HDR_MULT * 0.715200 * 2^14/125) 1285 spl_convert_to_custom_float_format(c1_calc, &fmt, &dscl_prog_data->easf_matrix_c1); 1286 // fp1.5.10, C2 coefficient (LN_rec709: HDR_MULT * 0.072200 * 2^14/125) 1287 spl_convert_to_custom_float_format(c2_calc, &fmt, &dscl_prog_data->easf_matrix_c2); 1288 dscl_prog_data->easf_matrix_c3 = 0x0; // fp1.5.10, C3 coefficient 1289 } 1290 1291 /* Set EASF data */ 1292 static void spl_set_easf_data(struct spl_scratch *spl_scratch, struct spl_out *spl_out, bool enable_easf_v, 1293 bool enable_easf_h, enum linear_light_scaling lls_pref, 1294 enum spl_pixel_format format, enum system_setup setup, 1295 uint32_t sdr_white_level_nits) 1296 { 1297 struct dscl_prog_data *dscl_prog_data = spl_out->dscl_prog_data; 1298 if (enable_easf_v) { 1299 dscl_prog_data->easf_v_en = true; 1300 dscl_prog_data->easf_v_ring = 0; 1301 dscl_prog_data->easf_v_sharp_factor = 0; 1302 dscl_prog_data->easf_v_bf1_en = 1; // 1-bit, BF1 calculation enable, 0=disable, 1=enable 1303 dscl_prog_data->easf_v_bf2_mode = 0xF; // 4-bit, BF2 calculation mode 1304 /* 2-bit, BF3 chroma mode correction calculation mode */ 1305 dscl_prog_data->easf_v_bf3_mode = spl_get_v_bf3_mode( 1306 spl_scratch->scl_data.recip_ratios.vert); 1307 /* FP1.5.10 [ minCoef ]*/ 1308 dscl_prog_data->easf_v_ringest_3tap_dntilt_uptilt = 1309 spl_get_3tap_dntilt_uptilt_offset(spl_scratch->scl_data.taps.v_taps, 1310 spl_scratch->scl_data.recip_ratios.vert); 1311 /* FP1.5.10 [ upTiltMaxVal ]*/ 1312 dscl_prog_data->easf_v_ringest_3tap_uptilt_max = 1313 spl_get_3tap_uptilt_maxval(spl_scratch->scl_data.taps.v_taps, 1314 spl_scratch->scl_data.recip_ratios.vert); 1315 /* FP1.5.10 [ dnTiltSlope ]*/ 1316 dscl_prog_data->easf_v_ringest_3tap_dntilt_slope = 1317 spl_get_3tap_dntilt_slope(spl_scratch->scl_data.taps.v_taps, 1318 spl_scratch->scl_data.recip_ratios.vert); 1319 /* FP1.5.10 [ upTilt1Slope ]*/ 1320 dscl_prog_data->easf_v_ringest_3tap_uptilt1_slope = 1321 spl_get_3tap_uptilt1_slope(spl_scratch->scl_data.taps.v_taps, 1322 spl_scratch->scl_data.recip_ratios.vert); 1323 /* FP1.5.10 [ upTilt2Slope ]*/ 1324 dscl_prog_data->easf_v_ringest_3tap_uptilt2_slope = 1325 spl_get_3tap_uptilt2_slope(spl_scratch->scl_data.taps.v_taps, 1326 spl_scratch->scl_data.recip_ratios.vert); 1327 /* FP1.5.10 [ upTilt2Offset ]*/ 1328 dscl_prog_data->easf_v_ringest_3tap_uptilt2_offset = 1329 spl_get_3tap_uptilt2_offset(spl_scratch->scl_data.taps.v_taps, 1330 spl_scratch->scl_data.recip_ratios.vert); 1331 /* FP1.5.10; (2.0) Ring reducer gain for 4 or 6-tap mode [H_REDUCER_GAIN4] */ 1332 dscl_prog_data->easf_v_ringest_eventap_reduceg1 = 1333 spl_get_reducer_gain4(spl_scratch->scl_data.taps.v_taps, 1334 spl_scratch->scl_data.recip_ratios.vert); 1335 /* FP1.5.10; (2.5) Ring reducer gain for 6-tap mode [V_REDUCER_GAIN6] */ 1336 dscl_prog_data->easf_v_ringest_eventap_reduceg2 = 1337 spl_get_reducer_gain6(spl_scratch->scl_data.taps.v_taps, 1338 spl_scratch->scl_data.recip_ratios.vert); 1339 /* FP1.5.10; (-0.135742) Ring gain for 6-tap set to -139/1024 */ 1340 dscl_prog_data->easf_v_ringest_eventap_gain1 = 1341 spl_get_gainRing4(spl_scratch->scl_data.taps.v_taps, 1342 spl_scratch->scl_data.recip_ratios.vert); 1343 /* FP1.5.10; (-0.024414) Ring gain for 6-tap set to -25/1024 */ 1344 dscl_prog_data->easf_v_ringest_eventap_gain2 = 1345 spl_get_gainRing6(spl_scratch->scl_data.taps.v_taps, 1346 spl_scratch->scl_data.recip_ratios.vert); 1347 dscl_prog_data->easf_v_bf_maxa = 63; //Vertical Max BF value A in U0.6 format.Selected if V_FCNTL == 0 1348 dscl_prog_data->easf_v_bf_maxb = 63; //Vertical Max BF value A in U0.6 format.Selected if V_FCNTL == 1 1349 dscl_prog_data->easf_v_bf_mina = 0; //Vertical Min BF value A in U0.6 format.Selected if V_FCNTL == 0 1350 dscl_prog_data->easf_v_bf_minb = 0; //Vertical Min BF value A in U0.6 format.Selected if V_FCNTL == 1 1351 if (lls_pref == LLS_PREF_YES) { 1352 dscl_prog_data->easf_v_bf2_flat1_gain = 4; // U1.3, BF2 Flat1 Gain control 1353 dscl_prog_data->easf_v_bf2_flat2_gain = 8; // U4.0, BF2 Flat2 Gain control 1354 dscl_prog_data->easf_v_bf2_roc_gain = 4; // U2.2, Rate Of Change control 1355 1356 dscl_prog_data->easf_v_bf1_pwl_in_seg0 = 0x600; // S0.10, BF1 PWL Segment 0 = -512 1357 dscl_prog_data->easf_v_bf1_pwl_base_seg0 = 0; // U0.6, BF1 Base PWL Segment 0 1358 dscl_prog_data->easf_v_bf1_pwl_slope_seg0 = 3; // S7.3, BF1 Slope PWL Segment 0 1359 dscl_prog_data->easf_v_bf1_pwl_in_seg1 = 0x7EC; // S0.10, BF1 PWL Segment 1 = -20 1360 dscl_prog_data->easf_v_bf1_pwl_base_seg1 = 12; // U0.6, BF1 Base PWL Segment 1 1361 dscl_prog_data->easf_v_bf1_pwl_slope_seg1 = 326; // S7.3, BF1 Slope PWL Segment 1 1362 dscl_prog_data->easf_v_bf1_pwl_in_seg2 = 0; // S0.10, BF1 PWL Segment 2 1363 dscl_prog_data->easf_v_bf1_pwl_base_seg2 = 63; // U0.6, BF1 Base PWL Segment 2 1364 dscl_prog_data->easf_v_bf1_pwl_slope_seg2 = 0; // S7.3, BF1 Slope PWL Segment 2 1365 dscl_prog_data->easf_v_bf1_pwl_in_seg3 = 16; // S0.10, BF1 PWL Segment 3 1366 dscl_prog_data->easf_v_bf1_pwl_base_seg3 = 63; // U0.6, BF1 Base PWL Segment 3 1367 dscl_prog_data->easf_v_bf1_pwl_slope_seg3 = 0x7C8; // S7.3, BF1 Slope PWL Segment 3 = -56 1368 dscl_prog_data->easf_v_bf1_pwl_in_seg4 = 32; // S0.10, BF1 PWL Segment 4 1369 dscl_prog_data->easf_v_bf1_pwl_base_seg4 = 56; // U0.6, BF1 Base PWL Segment 4 1370 dscl_prog_data->easf_v_bf1_pwl_slope_seg4 = 0x7D0; // S7.3, BF1 Slope PWL Segment 4 = -48 1371 dscl_prog_data->easf_v_bf1_pwl_in_seg5 = 48; // S0.10, BF1 PWL Segment 5 1372 dscl_prog_data->easf_v_bf1_pwl_base_seg5 = 50; // U0.6, BF1 Base PWL Segment 5 1373 dscl_prog_data->easf_v_bf1_pwl_slope_seg5 = 0x710; // S7.3, BF1 Slope PWL Segment 5 = -240 1374 dscl_prog_data->easf_v_bf1_pwl_in_seg6 = 64; // S0.10, BF1 PWL Segment 6 1375 dscl_prog_data->easf_v_bf1_pwl_base_seg6 = 20; // U0.6, BF1 Base PWL Segment 6 1376 dscl_prog_data->easf_v_bf1_pwl_slope_seg6 = 0x760; // S7.3, BF1 Slope PWL Segment 6 = -160 1377 dscl_prog_data->easf_v_bf1_pwl_in_seg7 = 80; // S0.10, BF1 PWL Segment 7 1378 dscl_prog_data->easf_v_bf1_pwl_base_seg7 = 0; // U0.6, BF1 Base PWL Segment 7 1379 1380 dscl_prog_data->easf_v_bf3_pwl_in_set0 = 0x000; // FP0.6.6, BF3 Input value PWL Segment 0 1381 dscl_prog_data->easf_v_bf3_pwl_base_set0 = 63; // S0.6, BF3 Base PWL Segment 0 1382 dscl_prog_data->easf_v_bf3_pwl_slope_set0 = 0x12C5; // FP1.6.6, BF3 Slope PWL Segment 0 1383 dscl_prog_data->easf_v_bf3_pwl_in_set1 = 1384 0x0B37; // FP0.6.6, BF3 Input value PWL Segment 1 (0.0078125 * 125^3) 1385 dscl_prog_data->easf_v_bf3_pwl_base_set1 = 62; // S0.6, BF3 Base PWL Segment 1 1386 dscl_prog_data->easf_v_bf3_pwl_slope_set1 = 1387 0x13B8; // FP1.6.6, BF3 Slope PWL Segment 1 1388 dscl_prog_data->easf_v_bf3_pwl_in_set2 = 1389 0x0BB7; // FP0.6.6, BF3 Input value PWL Segment 2 (0.03125 * 125^3) 1390 dscl_prog_data->easf_v_bf3_pwl_base_set2 = 20; // S0.6, BF3 Base PWL Segment 2 1391 dscl_prog_data->easf_v_bf3_pwl_slope_set2 = 1392 0x1356; // FP1.6.6, BF3 Slope PWL Segment 2 1393 dscl_prog_data->easf_v_bf3_pwl_in_set3 = 1394 0x0BF7; // FP0.6.6, BF3 Input value PWL Segment 3 (0.0625 * 125^3) 1395 dscl_prog_data->easf_v_bf3_pwl_base_set3 = 0; // S0.6, BF3 Base PWL Segment 3 1396 dscl_prog_data->easf_v_bf3_pwl_slope_set3 = 1397 0x136B; // FP1.6.6, BF3 Slope PWL Segment 3 1398 dscl_prog_data->easf_v_bf3_pwl_in_set4 = 1399 0x0C37; // FP0.6.6, BF3 Input value PWL Segment 4 (0.125 * 125^3) 1400 dscl_prog_data->easf_v_bf3_pwl_base_set4 = 0x4E; // S0.6, BF3 Base PWL Segment 4 = -50 1401 dscl_prog_data->easf_v_bf3_pwl_slope_set4 = 1402 0x1200; // FP1.6.6, BF3 Slope PWL Segment 4 1403 dscl_prog_data->easf_v_bf3_pwl_in_set5 = 1404 0x0CF7; // FP0.6.6, BF3 Input value PWL Segment 5 (1.0 * 125^3) 1405 dscl_prog_data->easf_v_bf3_pwl_base_set5 = 0x41; // S0.6, BF3 Base PWL Segment 5 = -63 1406 } else { 1407 dscl_prog_data->easf_v_bf2_flat1_gain = 13; // U1.3, BF2 Flat1 Gain control 1408 dscl_prog_data->easf_v_bf2_flat2_gain = 15; // U4.0, BF2 Flat2 Gain control 1409 dscl_prog_data->easf_v_bf2_roc_gain = 14; // U2.2, Rate Of Change control 1410 1411 dscl_prog_data->easf_v_bf1_pwl_in_seg0 = 0x440; // S0.10, BF1 PWL Segment 0 = -960 1412 dscl_prog_data->easf_v_bf1_pwl_base_seg0 = 0; // U0.6, BF1 Base PWL Segment 0 1413 dscl_prog_data->easf_v_bf1_pwl_slope_seg0 = 2; // S7.3, BF1 Slope PWL Segment 0 1414 dscl_prog_data->easf_v_bf1_pwl_in_seg1 = 0x7C4; // S0.10, BF1 PWL Segment 1 = -60 1415 dscl_prog_data->easf_v_bf1_pwl_base_seg1 = 12; // U0.6, BF1 Base PWL Segment 1 1416 dscl_prog_data->easf_v_bf1_pwl_slope_seg1 = 109; // S7.3, BF1 Slope PWL Segment 1 1417 dscl_prog_data->easf_v_bf1_pwl_in_seg2 = 0; // S0.10, BF1 PWL Segment 2 1418 dscl_prog_data->easf_v_bf1_pwl_base_seg2 = 63; // U0.6, BF1 Base PWL Segment 2 1419 dscl_prog_data->easf_v_bf1_pwl_slope_seg2 = 0; // S7.3, BF1 Slope PWL Segment 2 1420 dscl_prog_data->easf_v_bf1_pwl_in_seg3 = 48; // S0.10, BF1 PWL Segment 3 1421 dscl_prog_data->easf_v_bf1_pwl_base_seg3 = 63; // U0.6, BF1 Base PWL Segment 3 1422 dscl_prog_data->easf_v_bf1_pwl_slope_seg3 = 0x7ED; // S7.3, BF1 Slope PWL Segment 3 = -19 1423 dscl_prog_data->easf_v_bf1_pwl_in_seg4 = 96; // S0.10, BF1 PWL Segment 4 1424 dscl_prog_data->easf_v_bf1_pwl_base_seg4 = 56; // U0.6, BF1 Base PWL Segment 4 1425 dscl_prog_data->easf_v_bf1_pwl_slope_seg4 = 0x7F0; // S7.3, BF1 Slope PWL Segment 4 = -16 1426 dscl_prog_data->easf_v_bf1_pwl_in_seg5 = 144; // S0.10, BF1 PWL Segment 5 1427 dscl_prog_data->easf_v_bf1_pwl_base_seg5 = 50; // U0.6, BF1 Base PWL Segment 5 1428 dscl_prog_data->easf_v_bf1_pwl_slope_seg5 = 0x7B0; // S7.3, BF1 Slope PWL Segment 5 = -80 1429 dscl_prog_data->easf_v_bf1_pwl_in_seg6 = 192; // S0.10, BF1 PWL Segment 6 1430 dscl_prog_data->easf_v_bf1_pwl_base_seg6 = 20; // U0.6, BF1 Base PWL Segment 6 1431 dscl_prog_data->easf_v_bf1_pwl_slope_seg6 = 0x7CB; // S7.3, BF1 Slope PWL Segment 6 = -53 1432 dscl_prog_data->easf_v_bf1_pwl_in_seg7 = 240; // S0.10, BF1 PWL Segment 7 1433 dscl_prog_data->easf_v_bf1_pwl_base_seg7 = 0; // U0.6, BF1 Base PWL Segment 7 1434 1435 dscl_prog_data->easf_v_bf3_pwl_in_set0 = 0x000; // FP0.6.6, BF3 Input value PWL Segment 0 1436 dscl_prog_data->easf_v_bf3_pwl_base_set0 = 63; // S0.6, BF3 Base PWL Segment 0 1437 dscl_prog_data->easf_v_bf3_pwl_slope_set0 = 0x0000; // FP1.6.6, BF3 Slope PWL Segment 0 1438 dscl_prog_data->easf_v_bf3_pwl_in_set1 = 1439 0x06C0; // FP0.6.6, BF3 Input value PWL Segment 1 (0.0625) 1440 dscl_prog_data->easf_v_bf3_pwl_base_set1 = 63; // S0.6, BF3 Base PWL Segment 1 1441 dscl_prog_data->easf_v_bf3_pwl_slope_set1 = 0x1896; // FP1.6.6, BF3 Slope PWL Segment 1 1442 dscl_prog_data->easf_v_bf3_pwl_in_set2 = 1443 0x0700; // FP0.6.6, BF3 Input value PWL Segment 2 (0.125) 1444 dscl_prog_data->easf_v_bf3_pwl_base_set2 = 20; // S0.6, BF3 Base PWL Segment 2 1445 dscl_prog_data->easf_v_bf3_pwl_slope_set2 = 0x1810; // FP1.6.6, BF3 Slope PWL Segment 2 1446 dscl_prog_data->easf_v_bf3_pwl_in_set3 = 1447 0x0740; // FP0.6.6, BF3 Input value PWL Segment 3 (0.25) 1448 dscl_prog_data->easf_v_bf3_pwl_base_set3 = 0; // S0.6, BF3 Base PWL Segment 3 1449 dscl_prog_data->easf_v_bf3_pwl_slope_set3 = 1450 0x1878; // FP1.6.6, BF3 Slope PWL Segment 3 1451 dscl_prog_data->easf_v_bf3_pwl_in_set4 = 1452 0x0761; // FP0.6.6, BF3 Input value PWL Segment 4 (0.375) 1453 dscl_prog_data->easf_v_bf3_pwl_base_set4 = 0x44; // S0.6, BF3 Base PWL Segment 4 = -60 1454 dscl_prog_data->easf_v_bf3_pwl_slope_set4 = 0x1760; // FP1.6.6, BF3 Slope PWL Segment 4 1455 dscl_prog_data->easf_v_bf3_pwl_in_set5 = 1456 0x0780; // FP0.6.6, BF3 Input value PWL Segment 5 (0.5) 1457 dscl_prog_data->easf_v_bf3_pwl_base_set5 = 0x41; // S0.6, BF3 Base PWL Segment 5 = -63 1458 } 1459 } else 1460 dscl_prog_data->easf_v_en = false; 1461 1462 if (enable_easf_h) { 1463 dscl_prog_data->easf_h_en = true; 1464 dscl_prog_data->easf_h_ring = 0; 1465 dscl_prog_data->easf_h_sharp_factor = 0; 1466 dscl_prog_data->easf_h_bf1_en = 1467 1; // 1-bit, BF1 calculation enable, 0=disable, 1=enable 1468 dscl_prog_data->easf_h_bf2_mode = 1469 0xF; // 4-bit, BF2 calculation mode 1470 /* 2-bit, BF3 chroma mode correction calculation mode */ 1471 dscl_prog_data->easf_h_bf3_mode = spl_get_h_bf3_mode( 1472 spl_scratch->scl_data.recip_ratios.horz); 1473 /* FP1.5.10; (2.0) Ring reducer gain for 4 or 6-tap mode [H_REDUCER_GAIN4] */ 1474 dscl_prog_data->easf_h_ringest_eventap_reduceg1 = 1475 spl_get_reducer_gain4(spl_scratch->scl_data.taps.h_taps, 1476 spl_scratch->scl_data.recip_ratios.horz); 1477 /* FP1.5.10; (2.5) Ring reducer gain for 6-tap mode [V_REDUCER_GAIN6] */ 1478 dscl_prog_data->easf_h_ringest_eventap_reduceg2 = 1479 spl_get_reducer_gain6(spl_scratch->scl_data.taps.h_taps, 1480 spl_scratch->scl_data.recip_ratios.horz); 1481 /* FP1.5.10; (-0.135742) Ring gain for 6-tap set to -139/1024 */ 1482 dscl_prog_data->easf_h_ringest_eventap_gain1 = 1483 spl_get_gainRing4(spl_scratch->scl_data.taps.h_taps, 1484 spl_scratch->scl_data.recip_ratios.horz); 1485 /* FP1.5.10; (-0.024414) Ring gain for 6-tap set to -25/1024 */ 1486 dscl_prog_data->easf_h_ringest_eventap_gain2 = 1487 spl_get_gainRing6(spl_scratch->scl_data.taps.h_taps, 1488 spl_scratch->scl_data.recip_ratios.horz); 1489 dscl_prog_data->easf_h_bf_maxa = 63; //Horz Max BF value A in U0.6 format.Selected if H_FCNTL==0 1490 dscl_prog_data->easf_h_bf_maxb = 63; //Horz Max BF value B in U0.6 format.Selected if H_FCNTL==1 1491 dscl_prog_data->easf_h_bf_mina = 0; //Horz Min BF value B in U0.6 format.Selected if H_FCNTL==0 1492 dscl_prog_data->easf_h_bf_minb = 0; //Horz Min BF value B in U0.6 format.Selected if H_FCNTL==1 1493 if (lls_pref == LLS_PREF_YES) { 1494 dscl_prog_data->easf_h_bf2_flat1_gain = 4; // U1.3, BF2 Flat1 Gain control 1495 dscl_prog_data->easf_h_bf2_flat2_gain = 8; // U4.0, BF2 Flat2 Gain control 1496 dscl_prog_data->easf_h_bf2_roc_gain = 4; // U2.2, Rate Of Change control 1497 1498 dscl_prog_data->easf_h_bf1_pwl_in_seg0 = 0x600; // S0.10, BF1 PWL Segment 0 = -512 1499 dscl_prog_data->easf_h_bf1_pwl_base_seg0 = 0; // U0.6, BF1 Base PWL Segment 0 1500 dscl_prog_data->easf_h_bf1_pwl_slope_seg0 = 3; // S7.3, BF1 Slope PWL Segment 0 1501 dscl_prog_data->easf_h_bf1_pwl_in_seg1 = 0x7EC; // S0.10, BF1 PWL Segment 1 = -20 1502 dscl_prog_data->easf_h_bf1_pwl_base_seg1 = 12; // U0.6, BF1 Base PWL Segment 1 1503 dscl_prog_data->easf_h_bf1_pwl_slope_seg1 = 326; // S7.3, BF1 Slope PWL Segment 1 1504 dscl_prog_data->easf_h_bf1_pwl_in_seg2 = 0; // S0.10, BF1 PWL Segment 2 1505 dscl_prog_data->easf_h_bf1_pwl_base_seg2 = 63; // U0.6, BF1 Base PWL Segment 2 1506 dscl_prog_data->easf_h_bf1_pwl_slope_seg2 = 0; // S7.3, BF1 Slope PWL Segment 2 1507 dscl_prog_data->easf_h_bf1_pwl_in_seg3 = 16; // S0.10, BF1 PWL Segment 3 1508 dscl_prog_data->easf_h_bf1_pwl_base_seg3 = 63; // U0.6, BF1 Base PWL Segment 3 1509 dscl_prog_data->easf_h_bf1_pwl_slope_seg3 = 0x7C8; // S7.3, BF1 Slope PWL Segment 3 = -56 1510 dscl_prog_data->easf_h_bf1_pwl_in_seg4 = 32; // S0.10, BF1 PWL Segment 4 1511 dscl_prog_data->easf_h_bf1_pwl_base_seg4 = 56; // U0.6, BF1 Base PWL Segment 4 1512 dscl_prog_data->easf_h_bf1_pwl_slope_seg4 = 0x7D0; // S7.3, BF1 Slope PWL Segment 4 = -48 1513 dscl_prog_data->easf_h_bf1_pwl_in_seg5 = 48; // S0.10, BF1 PWL Segment 5 1514 dscl_prog_data->easf_h_bf1_pwl_base_seg5 = 50; // U0.6, BF1 Base PWL Segment 5 1515 dscl_prog_data->easf_h_bf1_pwl_slope_seg5 = 0x710; // S7.3, BF1 Slope PWL Segment 5 = -240 1516 dscl_prog_data->easf_h_bf1_pwl_in_seg6 = 64; // S0.10, BF1 PWL Segment 6 1517 dscl_prog_data->easf_h_bf1_pwl_base_seg6 = 20; // U0.6, BF1 Base PWL Segment 6 1518 dscl_prog_data->easf_h_bf1_pwl_slope_seg6 = 0x760; // S7.3, BF1 Slope PWL Segment 6 = -160 1519 dscl_prog_data->easf_h_bf1_pwl_in_seg7 = 80; // S0.10, BF1 PWL Segment 7 1520 dscl_prog_data->easf_h_bf1_pwl_base_seg7 = 0; // U0.6, BF1 Base PWL Segment 7 1521 1522 dscl_prog_data->easf_h_bf3_pwl_in_set0 = 0x000; // FP0.6.6, BF3 Input value PWL Segment 0 1523 dscl_prog_data->easf_h_bf3_pwl_base_set0 = 63; // S0.6, BF3 Base PWL Segment 0 1524 dscl_prog_data->easf_h_bf3_pwl_slope_set0 = 0x12C5; // FP1.6.6, BF3 Slope PWL Segment 0 1525 dscl_prog_data->easf_h_bf3_pwl_in_set1 = 1526 0x0B37; // FP0.6.6, BF3 Input value PWL Segment 1 (0.0078125 * 125^3) 1527 dscl_prog_data->easf_h_bf3_pwl_base_set1 = 62; // S0.6, BF3 Base PWL Segment 1 1528 dscl_prog_data->easf_h_bf3_pwl_slope_set1 = 0x13B8; // FP1.6.6, BF3 Slope PWL Segment 1 1529 dscl_prog_data->easf_h_bf3_pwl_in_set2 = 1530 0x0BB7; // FP0.6.6, BF3 Input value PWL Segment 2 (0.03125 * 125^3) 1531 dscl_prog_data->easf_h_bf3_pwl_base_set2 = 20; // S0.6, BF3 Base PWL Segment 2 1532 dscl_prog_data->easf_h_bf3_pwl_slope_set2 = 0x1356; // FP1.6.6, BF3 Slope PWL Segment 2 1533 dscl_prog_data->easf_h_bf3_pwl_in_set3 = 1534 0x0BF7; // FP0.6.6, BF3 Input value PWL Segment 3 (0.0625 * 125^3) 1535 dscl_prog_data->easf_h_bf3_pwl_base_set3 = 0; // S0.6, BF3 Base PWL Segment 3 1536 dscl_prog_data->easf_h_bf3_pwl_slope_set3 = 0x136B; // FP1.6.6, BF3 Slope PWL Segment 3 1537 dscl_prog_data->easf_h_bf3_pwl_in_set4 = 1538 0x0C37; // FP0.6.6, BF3 Input value PWL Segment 4 (0.125 * 125^3) 1539 dscl_prog_data->easf_h_bf3_pwl_base_set4 = 0x4E; // S0.6, BF3 Base PWL Segment 4 = -50 1540 dscl_prog_data->easf_h_bf3_pwl_slope_set4 = 0x1200; // FP1.6.6, BF3 Slope PWL Segment 4 1541 dscl_prog_data->easf_h_bf3_pwl_in_set5 = 1542 0x0CF7; // FP0.6.6, BF3 Input value PWL Segment 5 (1.0 * 125^3) 1543 dscl_prog_data->easf_h_bf3_pwl_base_set5 = 0x41; // S0.6, BF3 Base PWL Segment 5 = -63 1544 } else { 1545 dscl_prog_data->easf_h_bf2_flat1_gain = 13; // U1.3, BF2 Flat1 Gain control 1546 dscl_prog_data->easf_h_bf2_flat2_gain = 15; // U4.0, BF2 Flat2 Gain control 1547 dscl_prog_data->easf_h_bf2_roc_gain = 14; // U2.2, Rate Of Change control 1548 1549 dscl_prog_data->easf_h_bf1_pwl_in_seg0 = 0x440; // S0.10, BF1 PWL Segment 0 = -960 1550 dscl_prog_data->easf_h_bf1_pwl_base_seg0 = 0; // U0.6, BF1 Base PWL Segment 0 1551 dscl_prog_data->easf_h_bf1_pwl_slope_seg0 = 2; // S7.3, BF1 Slope PWL Segment 0 1552 dscl_prog_data->easf_h_bf1_pwl_in_seg1 = 0x7C4; // S0.10, BF1 PWL Segment 1 = -60 1553 dscl_prog_data->easf_h_bf1_pwl_base_seg1 = 12; // U0.6, BF1 Base PWL Segment 1 1554 dscl_prog_data->easf_h_bf1_pwl_slope_seg1 = 109; // S7.3, BF1 Slope PWL Segment 1 1555 dscl_prog_data->easf_h_bf1_pwl_in_seg2 = 0; // S0.10, BF1 PWL Segment 2 1556 dscl_prog_data->easf_h_bf1_pwl_base_seg2 = 63; // U0.6, BF1 Base PWL Segment 2 1557 dscl_prog_data->easf_h_bf1_pwl_slope_seg2 = 0; // S7.3, BF1 Slope PWL Segment 2 1558 dscl_prog_data->easf_h_bf1_pwl_in_seg3 = 48; // S0.10, BF1 PWL Segment 3 1559 dscl_prog_data->easf_h_bf1_pwl_base_seg3 = 63; // U0.6, BF1 Base PWL Segment 3 1560 dscl_prog_data->easf_h_bf1_pwl_slope_seg3 = 0x7ED; // S7.3, BF1 Slope PWL Segment 3 = -19 1561 dscl_prog_data->easf_h_bf1_pwl_in_seg4 = 96; // S0.10, BF1 PWL Segment 4 1562 dscl_prog_data->easf_h_bf1_pwl_base_seg4 = 56; // U0.6, BF1 Base PWL Segment 4 1563 dscl_prog_data->easf_h_bf1_pwl_slope_seg4 = 0x7F0; // S7.3, BF1 Slope PWL Segment 4 = -16 1564 dscl_prog_data->easf_h_bf1_pwl_in_seg5 = 144; // S0.10, BF1 PWL Segment 5 1565 dscl_prog_data->easf_h_bf1_pwl_base_seg5 = 50; // U0.6, BF1 Base PWL Segment 5 1566 dscl_prog_data->easf_h_bf1_pwl_slope_seg5 = 0x7B0; // S7.3, BF1 Slope PWL Segment 5 = -80 1567 dscl_prog_data->easf_h_bf1_pwl_in_seg6 = 192; // S0.10, BF1 PWL Segment 6 1568 dscl_prog_data->easf_h_bf1_pwl_base_seg6 = 20; // U0.6, BF1 Base PWL Segment 6 1569 dscl_prog_data->easf_h_bf1_pwl_slope_seg6 = 0x7CB; // S7.3, BF1 Slope PWL Segment 6 = -53 1570 dscl_prog_data->easf_h_bf1_pwl_in_seg7 = 240; // S0.10, BF1 PWL Segment 7 1571 dscl_prog_data->easf_h_bf1_pwl_base_seg7 = 0; // U0.6, BF1 Base PWL Segment 7 1572 1573 dscl_prog_data->easf_h_bf3_pwl_in_set0 = 0x000; // FP0.6.6, BF3 Input value PWL Segment 0 1574 dscl_prog_data->easf_h_bf3_pwl_base_set0 = 63; // S0.6, BF3 Base PWL Segment 0 1575 dscl_prog_data->easf_h_bf3_pwl_slope_set0 = 0x0000; // FP1.6.6, BF3 Slope PWL Segment 0 1576 dscl_prog_data->easf_h_bf3_pwl_in_set1 = 1577 0x06C0; // FP0.6.6, BF3 Input value PWL Segment 1 (0.0625) 1578 dscl_prog_data->easf_h_bf3_pwl_base_set1 = 63; // S0.6, BF3 Base PWL Segment 1 1579 dscl_prog_data->easf_h_bf3_pwl_slope_set1 = 0x1896; // FP1.6.6, BF3 Slope PWL Segment 1 1580 dscl_prog_data->easf_h_bf3_pwl_in_set2 = 1581 0x0700; // FP0.6.6, BF3 Input value PWL Segment 2 (0.125) 1582 dscl_prog_data->easf_h_bf3_pwl_base_set2 = 20; // S0.6, BF3 Base PWL Segment 2 1583 dscl_prog_data->easf_h_bf3_pwl_slope_set2 = 0x1810; // FP1.6.6, BF3 Slope PWL Segment 2 1584 dscl_prog_data->easf_h_bf3_pwl_in_set3 = 1585 0x0740; // FP0.6.6, BF3 Input value PWL Segment 3 (0.25) 1586 dscl_prog_data->easf_h_bf3_pwl_base_set3 = 0; // S0.6, BF3 Base PWL Segment 3 1587 dscl_prog_data->easf_h_bf3_pwl_slope_set3 = 0x1878; // FP1.6.6, BF3 Slope PWL Segment 3 1588 dscl_prog_data->easf_h_bf3_pwl_in_set4 = 1589 0x0761; // FP0.6.6, BF3 Input value PWL Segment 4 (0.375) 1590 dscl_prog_data->easf_h_bf3_pwl_base_set4 = 0x44; // S0.6, BF3 Base PWL Segment 4 = -60 1591 dscl_prog_data->easf_h_bf3_pwl_slope_set4 = 0x1760; // FP1.6.6, BF3 Slope PWL Segment 4 1592 dscl_prog_data->easf_h_bf3_pwl_in_set5 = 1593 0x0780; // FP0.6.6, BF3 Input value PWL Segment 5 (0.5) 1594 dscl_prog_data->easf_h_bf3_pwl_base_set5 = 0x41; // S0.6, BF3 Base PWL Segment 5 = -63 1595 } // if (lls_pref == LLS_PREF_YES) 1596 } else 1597 dscl_prog_data->easf_h_en = false; 1598 1599 if (lls_pref == LLS_PREF_YES) { 1600 dscl_prog_data->easf_ltonl_en = 1; // Linear input 1601 if ((setup == HDR_L) && (spl_is_rgb8(format))) { 1602 /* Calculate C0-C3 coefficients based on HDR multiplier */ 1603 spl_calculate_c0_c3_hdr(dscl_prog_data, sdr_white_level_nits); 1604 } else { // HDR_L ( DWM ) and SDR_L 1605 dscl_prog_data->easf_matrix_c0 = 1606 0x4EF7; // fp1.5.10, C0 coefficient (LN_rec709: 0.2126 * (2^14)/125 = 27.86590720) 1607 dscl_prog_data->easf_matrix_c1 = 1608 0x55DC; // fp1.5.10, C1 coefficient (LN_rec709: 0.7152 * (2^14)/125 = 93.74269440) 1609 dscl_prog_data->easf_matrix_c2 = 1610 0x48BB; // fp1.5.10, C2 coefficient (LN_rec709: 0.0722 * (2^14)/125 = 9.46339840) 1611 dscl_prog_data->easf_matrix_c3 = 1612 0x0; // fp1.5.10, C3 coefficient 1613 } 1614 } else { 1615 dscl_prog_data->easf_ltonl_en = 0; // Non-Linear input 1616 dscl_prog_data->easf_matrix_c0 = 1617 0x3434; // fp1.5.10, C0 coefficient (LN_BT2020: 0.262695312500000) 1618 dscl_prog_data->easf_matrix_c1 = 1619 0x396D; // fp1.5.10, C1 coefficient (LN_BT2020: 0.678222656250000) 1620 dscl_prog_data->easf_matrix_c2 = 1621 0x2B97; // fp1.5.10, C2 coefficient (LN_BT2020: 0.059295654296875) 1622 dscl_prog_data->easf_matrix_c3 = 1623 0x0; // fp1.5.10, C3 coefficient 1624 } 1625 1626 if (spl_is_subsampled_format(format)) { /* TODO: 0 = RGB, 1 = YUV */ 1627 dscl_prog_data->easf_matrix_mode = 1; 1628 /* 1629 * 2-bit, BF3 chroma mode correction calculation mode 1630 * Needs to be disabled for YUV420 mode 1631 * Override lookup value 1632 */ 1633 dscl_prog_data->easf_v_bf3_mode = 0; 1634 dscl_prog_data->easf_h_bf3_mode = 0; 1635 } else 1636 dscl_prog_data->easf_matrix_mode = 0; 1637 1638 } 1639 1640 /*Set isharp noise detection */ 1641 static void spl_set_isharp_noise_det_mode(struct dscl_prog_data *dscl_prog_data, 1642 const struct spl_scaler_data *data) 1643 { 1644 // ISHARP_NOISEDET_MODE 1645 // 0: 3x5 as VxH 1646 // 1: 4x5 as VxH 1647 // 2: 1648 // 3: 5x5 as VxH 1649 if (data->taps.v_taps == 6) 1650 dscl_prog_data->isharp_noise_det.mode = 3; 1651 else if (data->taps.v_taps == 4) 1652 dscl_prog_data->isharp_noise_det.mode = 1; 1653 else if (data->taps.v_taps == 3) 1654 dscl_prog_data->isharp_noise_det.mode = 0; 1655 }; 1656 /* Set Sharpener data */ 1657 static void spl_set_isharp_data(struct dscl_prog_data *dscl_prog_data, 1658 struct adaptive_sharpness adp_sharpness, bool enable_isharp, 1659 enum linear_light_scaling lls_pref, enum spl_pixel_format format, 1660 const struct spl_scaler_data *data, struct spl_fixed31_32 ratio, 1661 enum system_setup setup, enum scale_to_sharpness_policy scale_to_sharpness_policy) 1662 { 1663 /* Turn off sharpener if not required */ 1664 if (!enable_isharp) { 1665 dscl_prog_data->isharp_en = 0; 1666 return; 1667 } 1668 1669 spl_build_isharp_1dlut_from_reference_curve(ratio, setup, adp_sharpness, 1670 scale_to_sharpness_policy); 1671 memcpy(dscl_prog_data->isharp_delta, spl_get_pregen_filter_isharp_1D_lut(setup), 1672 sizeof(uint32_t) * ISHARP_LUT_TABLE_SIZE); 1673 dscl_prog_data->sharpness_level = adp_sharpness.sharpness_level; 1674 1675 dscl_prog_data->isharp_en = 1; // ISHARP_EN 1676 // Set ISHARP_NOISEDET_MODE if htaps = 6-tap 1677 if (data->taps.h_taps == 6) { 1678 dscl_prog_data->isharp_noise_det.enable = 1; /* ISHARP_NOISEDET_EN */ 1679 spl_set_isharp_noise_det_mode(dscl_prog_data, data); /* ISHARP_NOISEDET_MODE */ 1680 } else 1681 dscl_prog_data->isharp_noise_det.enable = 0; // ISHARP_NOISEDET_EN 1682 // Program noise detection threshold 1683 dscl_prog_data->isharp_noise_det.uthreshold = 24; // ISHARP_NOISEDET_UTHRE 1684 dscl_prog_data->isharp_noise_det.dthreshold = 4; // ISHARP_NOISEDET_DTHRE 1685 // Program noise detection gain 1686 dscl_prog_data->isharp_noise_det.pwl_start_in = 3; // ISHARP_NOISEDET_PWL_START_IN 1687 dscl_prog_data->isharp_noise_det.pwl_end_in = 13; // ISHARP_NOISEDET_PWL_END_IN 1688 dscl_prog_data->isharp_noise_det.pwl_slope = 1623; // ISHARP_NOISEDET_PWL_SLOPE 1689 1690 if (lls_pref == LLS_PREF_NO) /* ISHARP_FMT_MODE */ 1691 dscl_prog_data->isharp_fmt.mode = 1; 1692 else 1693 dscl_prog_data->isharp_fmt.mode = 0; 1694 1695 dscl_prog_data->isharp_fmt.norm = 0x3C00; // ISHARP_FMT_NORM 1696 dscl_prog_data->isharp_lba.mode = 0; // ISHARP_LBA_MODE 1697 1698 if (setup == SDR_L) { 1699 // ISHARP_LBA_PWL_SEG0: ISHARP Local Brightness Adjustment PWL Segment 0 1700 dscl_prog_data->isharp_lba.in_seg[0] = 0; // ISHARP LBA PWL for Seg 0. INPUT value in U0.10 format 1701 dscl_prog_data->isharp_lba.base_seg[0] = 0; // ISHARP LBA PWL for Seg 0. BASE value in U0.6 format 1702 dscl_prog_data->isharp_lba.slope_seg[0] = 62; // ISHARP LBA for Seg 0. SLOPE value in S5.3 format 1703 // ISHARP_LBA_PWL_SEG1: ISHARP LBA PWL Segment 1 1704 dscl_prog_data->isharp_lba.in_seg[1] = 130; // ISHARP LBA PWL for Seg 1. INPUT value in U0.10 format 1705 dscl_prog_data->isharp_lba.base_seg[1] = 63; // ISHARP LBA PWL for Seg 1. BASE value in U0.6 format 1706 dscl_prog_data->isharp_lba.slope_seg[1] = 0; // ISHARP LBA for Seg 1. SLOPE value in S5.3 format 1707 // ISHARP_LBA_PWL_SEG2: ISHARP LBA PWL Segment 2 1708 dscl_prog_data->isharp_lba.in_seg[2] = 450; // ISHARP LBA PWL for Seg 2. INPUT value in U0.10 format 1709 dscl_prog_data->isharp_lba.base_seg[2] = 63; // ISHARP LBA PWL for Seg 2. BASE value in U0.6 format 1710 dscl_prog_data->isharp_lba.slope_seg[2] = 0x18D; // ISHARP LBA for Seg 2. SLOPE value in S5.3 format = -115 1711 // ISHARP_LBA_PWL_SEG3: ISHARP LBA PWL Segment 3 1712 dscl_prog_data->isharp_lba.in_seg[3] = 520; // ISHARP LBA PWL for Seg 3.INPUT value in U0.10 format 1713 dscl_prog_data->isharp_lba.base_seg[3] = 0; // ISHARP LBA PWL for Seg 3. BASE value in U0.6 format 1714 dscl_prog_data->isharp_lba.slope_seg[3] = 0; // ISHARP LBA for Seg 3. SLOPE value in S5.3 format 1715 // ISHARP_LBA_PWL_SEG4: ISHARP LBA PWL Segment 4 1716 dscl_prog_data->isharp_lba.in_seg[4] = 520; // ISHARP LBA PWL for Seg 4.INPUT value in U0.10 format 1717 dscl_prog_data->isharp_lba.base_seg[4] = 0; // ISHARP LBA PWL for Seg 4. BASE value in U0.6 format 1718 dscl_prog_data->isharp_lba.slope_seg[4] = 0; // ISHARP LBA for Seg 4. SLOPE value in S5.3 format 1719 // ISHARP_LBA_PWL_SEG5: ISHARP LBA PWL Segment 5 1720 dscl_prog_data->isharp_lba.in_seg[5] = 520; // ISHARP LBA PWL for Seg 5.INPUT value in U0.10 format 1721 dscl_prog_data->isharp_lba.base_seg[5] = 0; // ISHARP LBA PWL for Seg 5. BASE value in U0.6 format 1722 } else if (setup == HDR_L) { 1723 // ISHARP_LBA_PWL_SEG0: ISHARP Local Brightness Adjustment PWL Segment 0 1724 dscl_prog_data->isharp_lba.in_seg[0] = 0; // ISHARP LBA PWL for Seg 0. INPUT value in U0.10 format 1725 dscl_prog_data->isharp_lba.base_seg[0] = 0; // ISHARP LBA PWL for Seg 0. BASE value in U0.6 format 1726 dscl_prog_data->isharp_lba.slope_seg[0] = 32; // ISHARP LBA for Seg 0. SLOPE value in S5.3 format 1727 // ISHARP_LBA_PWL_SEG1: ISHARP LBA PWL Segment 1 1728 dscl_prog_data->isharp_lba.in_seg[1] = 254; // ISHARP LBA PWL for Seg 1. INPUT value in U0.10 format 1729 dscl_prog_data->isharp_lba.base_seg[1] = 63; // ISHARP LBA PWL for Seg 1. BASE value in U0.6 format 1730 dscl_prog_data->isharp_lba.slope_seg[1] = 0; // ISHARP LBA for Seg 1. SLOPE value in S5.3 format 1731 // ISHARP_LBA_PWL_SEG2: ISHARP LBA PWL Segment 2 1732 dscl_prog_data->isharp_lba.in_seg[2] = 559; // ISHARP LBA PWL for Seg 2. INPUT value in U0.10 format 1733 dscl_prog_data->isharp_lba.base_seg[2] = 63; // ISHARP LBA PWL for Seg 2. BASE value in U0.6 format 1734 dscl_prog_data->isharp_lba.slope_seg[2] = 0x10C; // ISHARP LBA for Seg 2. SLOPE value in S5.3 format = -244 1735 // ISHARP_LBA_PWL_SEG3: ISHARP LBA PWL Segment 3 1736 dscl_prog_data->isharp_lba.in_seg[3] = 592; // ISHARP LBA PWL for Seg 3.INPUT value in U0.10 format 1737 dscl_prog_data->isharp_lba.base_seg[3] = 0; // ISHARP LBA PWL for Seg 3. BASE value in U0.6 format 1738 dscl_prog_data->isharp_lba.slope_seg[3] = 0; // ISHARP LBA for Seg 3. SLOPE value in S5.3 format 1739 // ISHARP_LBA_PWL_SEG4: ISHARP LBA PWL Segment 4 1740 dscl_prog_data->isharp_lba.in_seg[4] = 1023; // ISHARP LBA PWL for Seg 4.INPUT value in U0.10 format 1741 dscl_prog_data->isharp_lba.base_seg[4] = 0; // ISHARP LBA PWL for Seg 4. BASE value in U0.6 format 1742 dscl_prog_data->isharp_lba.slope_seg[4] = 0; // ISHARP LBA for Seg 4. SLOPE value in S5.3 format 1743 // ISHARP_LBA_PWL_SEG5: ISHARP LBA PWL Segment 5 1744 dscl_prog_data->isharp_lba.in_seg[5] = 1023; // ISHARP LBA PWL for Seg 5.INPUT value in U0.10 format 1745 dscl_prog_data->isharp_lba.base_seg[5] = 0; // ISHARP LBA PWL for Seg 5. BASE value in U0.6 format 1746 } else { 1747 // ISHARP_LBA_PWL_SEG0: ISHARP Local Brightness Adjustment PWL Segment 0 1748 dscl_prog_data->isharp_lba.in_seg[0] = 0; // ISHARP LBA PWL for Seg 0. INPUT value in U0.10 format 1749 dscl_prog_data->isharp_lba.base_seg[0] = 0; // ISHARP LBA PWL for Seg 0. BASE value in U0.6 format 1750 dscl_prog_data->isharp_lba.slope_seg[0] = 40; // ISHARP LBA for Seg 0. SLOPE value in S5.3 format 1751 // ISHARP_LBA_PWL_SEG1: ISHARP LBA PWL Segment 1 1752 dscl_prog_data->isharp_lba.in_seg[1] = 204; // ISHARP LBA PWL for Seg 1. INPUT value in U0.10 format 1753 dscl_prog_data->isharp_lba.base_seg[1] = 63; // ISHARP LBA PWL for Seg 1. BASE value in U0.6 format 1754 dscl_prog_data->isharp_lba.slope_seg[1] = 0; // ISHARP LBA for Seg 1. SLOPE value in S5.3 format 1755 // ISHARP_LBA_PWL_SEG2: ISHARP LBA PWL Segment 2 1756 dscl_prog_data->isharp_lba.in_seg[2] = 818; // ISHARP LBA PWL for Seg 2. INPUT value in U0.10 format 1757 dscl_prog_data->isharp_lba.base_seg[2] = 63; // ISHARP LBA PWL for Seg 2. BASE value in U0.6 format 1758 dscl_prog_data->isharp_lba.slope_seg[2] = 0x1D9; // ISHARP LBA for Seg 2. SLOPE value in S5.3 format = -39 1759 // ISHARP_LBA_PWL_SEG3: ISHARP LBA PWL Segment 3 1760 dscl_prog_data->isharp_lba.in_seg[3] = 1023; // ISHARP LBA PWL for Seg 3.INPUT value in U0.10 format 1761 dscl_prog_data->isharp_lba.base_seg[3] = 0; // ISHARP LBA PWL for Seg 3. BASE value in U0.6 format 1762 dscl_prog_data->isharp_lba.slope_seg[3] = 0; // ISHARP LBA for Seg 3. SLOPE value in S5.3 format 1763 // ISHARP_LBA_PWL_SEG4: ISHARP LBA PWL Segment 4 1764 dscl_prog_data->isharp_lba.in_seg[4] = 1023; // ISHARP LBA PWL for Seg 4.INPUT value in U0.10 format 1765 dscl_prog_data->isharp_lba.base_seg[4] = 0; // ISHARP LBA PWL for Seg 4. BASE value in U0.6 format 1766 dscl_prog_data->isharp_lba.slope_seg[4] = 0; // ISHARP LBA for Seg 4. SLOPE value in S5.3 format 1767 // ISHARP_LBA_PWL_SEG5: ISHARP LBA PWL Segment 5 1768 dscl_prog_data->isharp_lba.in_seg[5] = 1023; // ISHARP LBA PWL for Seg 5.INPUT value in U0.10 format 1769 dscl_prog_data->isharp_lba.base_seg[5] = 0; // ISHARP LBA PWL for Seg 5. BASE value in U0.6 format 1770 } 1771 1772 // Program the nldelta soft clip values 1773 if (lls_pref == LLS_PREF_YES) { 1774 dscl_prog_data->isharp_nldelta_sclip.enable_p = 0; /* ISHARP_NLDELTA_SCLIP_EN_P */ 1775 dscl_prog_data->isharp_nldelta_sclip.pivot_p = 0; /* ISHARP_NLDELTA_SCLIP_PIVOT_P */ 1776 dscl_prog_data->isharp_nldelta_sclip.slope_p = 0; /* ISHARP_NLDELTA_SCLIP_SLOPE_P */ 1777 dscl_prog_data->isharp_nldelta_sclip.enable_n = 1; /* ISHARP_NLDELTA_SCLIP_EN_N */ 1778 dscl_prog_data->isharp_nldelta_sclip.pivot_n = 71; /* ISHARP_NLDELTA_SCLIP_PIVOT_N */ 1779 dscl_prog_data->isharp_nldelta_sclip.slope_n = 16; /* ISHARP_NLDELTA_SCLIP_SLOPE_N */ 1780 } else { 1781 dscl_prog_data->isharp_nldelta_sclip.enable_p = 1; /* ISHARP_NLDELTA_SCLIP_EN_P */ 1782 dscl_prog_data->isharp_nldelta_sclip.pivot_p = 70; /* ISHARP_NLDELTA_SCLIP_PIVOT_P */ 1783 dscl_prog_data->isharp_nldelta_sclip.slope_p = 24; /* ISHARP_NLDELTA_SCLIP_SLOPE_P */ 1784 dscl_prog_data->isharp_nldelta_sclip.enable_n = 1; /* ISHARP_NLDELTA_SCLIP_EN_N */ 1785 dscl_prog_data->isharp_nldelta_sclip.pivot_n = 70; /* ISHARP_NLDELTA_SCLIP_PIVOT_N */ 1786 dscl_prog_data->isharp_nldelta_sclip.slope_n = 24; /* ISHARP_NLDELTA_SCLIP_SLOPE_N */ 1787 } 1788 1789 // Set the values as per lookup table 1790 spl_set_blur_scale_data(dscl_prog_data, data); 1791 } 1792 1793 /* Calculate recout, scaling ratio, and viewport, then get optimal number of taps */ 1794 static bool spl_calculate_number_of_taps(struct spl_in *spl_in, struct spl_scratch *spl_scratch, struct spl_out *spl_out, 1795 bool *enable_easf_v, bool *enable_easf_h, bool *enable_isharp) 1796 { 1797 bool res = false; 1798 1799 memset(spl_scratch, 0, sizeof(struct spl_scratch)); 1800 spl_scratch->scl_data.h_active = spl_in->h_active; 1801 spl_scratch->scl_data.v_active = spl_in->v_active; 1802 1803 // All SPL calls 1804 /* recout calculation */ 1805 /* depends on h_active */ 1806 spl_calculate_recout(spl_in, spl_scratch, spl_out); 1807 /* depends on pixel format */ 1808 spl_calculate_scaling_ratios(spl_in, spl_scratch, spl_out); 1809 /* Adjust recout for opp if needed */ 1810 spl_opp_adjust_rect(&spl_scratch->scl_data.recout, &spl_in->basic_in.opp_recout_adjust); 1811 /* depends on scaling ratios and recout, does not calculate offset yet */ 1812 spl_calculate_viewport_size(spl_in, spl_scratch); 1813 1814 res = spl_get_optimal_number_of_taps( 1815 spl_in->basic_out.max_downscale_src_width, spl_in, 1816 spl_scratch, &spl_in->scaling_quality, enable_easf_v, 1817 enable_easf_h, enable_isharp); 1818 return res; 1819 } 1820 1821 /* Calculate scaler parameters */ 1822 bool SPL_NAMESPACE(spl_calculate_scaler_params(struct spl_in *spl_in, struct spl_out *spl_out)) 1823 { 1824 bool res = false; 1825 bool enable_easf_v = false; 1826 bool enable_easf_h = false; 1827 int vratio = 0; 1828 int hratio = 0; 1829 struct spl_scratch spl_scratch; 1830 struct spl_fixed31_32 isharp_scale_ratio; 1831 enum system_setup setup; 1832 bool enable_isharp = false; 1833 const struct spl_scaler_data *data = &spl_scratch.scl_data; 1834 1835 res = spl_calculate_number_of_taps(spl_in, &spl_scratch, spl_out, 1836 &enable_easf_v, &enable_easf_h, &enable_isharp); 1837 1838 /* 1839 * Depends on recout, scaling ratios, h_active and taps 1840 * May need to re-check lb size after this in some obscure scenario 1841 */ 1842 if (res) 1843 spl_calculate_inits_and_viewports(spl_in, &spl_scratch); 1844 // Handle 3d recout 1845 spl_handle_3d_recout(spl_in, &spl_scratch.scl_data.recout); 1846 // Clamp 1847 spl_clamp_viewport(&spl_scratch.scl_data.viewport, spl_in->min_viewport_size); 1848 1849 // Save all calculated parameters in dscl_prog_data structure to program hw registers 1850 spl_set_dscl_prog_data(spl_in, &spl_scratch, spl_out, enable_easf_v, enable_easf_h, enable_isharp); 1851 1852 if (!res) 1853 return res; 1854 1855 if (spl_in->lls_pref == LLS_PREF_YES) { 1856 if (spl_in->is_hdr_on) 1857 setup = HDR_L; 1858 else 1859 setup = SDR_L; 1860 } else { 1861 if (spl_in->is_hdr_on) 1862 setup = HDR_NL; 1863 else 1864 setup = SDR_NL; 1865 } 1866 1867 // Set EASF 1868 spl_set_easf_data(&spl_scratch, spl_out, enable_easf_v, enable_easf_h, spl_in->lls_pref, 1869 spl_in->basic_in.format, setup, spl_in->sdr_white_level_nits); 1870 1871 // Set iSHARP 1872 vratio = spl_fixpt_ceil(spl_scratch.scl_data.ratios.vert); 1873 hratio = spl_fixpt_ceil(spl_scratch.scl_data.ratios.horz); 1874 if (vratio <= hratio) 1875 isharp_scale_ratio = spl_scratch.scl_data.recip_ratios.vert; 1876 else 1877 isharp_scale_ratio = spl_scratch.scl_data.recip_ratios.horz; 1878 1879 spl_set_isharp_data(spl_out->dscl_prog_data, spl_in->adaptive_sharpness, enable_isharp, 1880 spl_in->lls_pref, spl_in->basic_in.format, data, isharp_scale_ratio, setup, 1881 spl_in->debug.scale_to_sharpness_policy); 1882 1883 return res; 1884 } 1885 1886 /* External interface to get number of taps only */ 1887 bool SPL_NAMESPACE(spl_get_number_of_taps(struct spl_in *spl_in, struct spl_out *spl_out)) 1888 { 1889 bool res = false; 1890 bool enable_easf_v = false; 1891 bool enable_easf_h = false; 1892 bool enable_isharp = false; 1893 struct spl_scratch spl_scratch; 1894 struct dscl_prog_data *dscl_prog_data = spl_out->dscl_prog_data; 1895 const struct spl_scaler_data *data = &spl_scratch.scl_data; 1896 1897 res = spl_calculate_number_of_taps(spl_in, &spl_scratch, spl_out, 1898 &enable_easf_v, &enable_easf_h, &enable_isharp); 1899 spl_set_taps_data(dscl_prog_data, data); 1900 return res; 1901 } 1902