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_u2d19(ratio) == (1 << 19))
12 #define MIN_VIEWPORT_SIZE 12
13
spl_is_yuv420(enum spl_pixel_format format)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
spl_is_rgb8(enum spl_pixel_format format)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
spl_is_video_format(enum spl_pixel_format format)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
spl_is_subsampled_format(enum spl_pixel_format format)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
intersect_rec(const struct spl_rect * r0,const struct spl_rect * r1)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
shift_rec(const struct spl_rect * rec_in,int x,int y)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
calculate_plane_rec_in_timing_active(struct spl_in * spl_in,const struct spl_rect * rec_in)79 static struct spl_rect calculate_plane_rec_in_timing_active(
80 struct spl_in *spl_in,
81 const struct spl_rect *rec_in)
82 {
83 /*
84 * The following diagram shows an example where we map a 1920x1200
85 * desktop to a 2560x1440 timing with a plane rect in the middle
86 * of the screen. To map a plane rect from Stream Source to Timing
87 * Active space, we first multiply stream scaling ratios (i.e 2304/1920
88 * horizontal and 1440/1200 vertical) to the plane's x and y, then
89 * we add stream destination offsets (i.e 128 horizontal, 0 vertical).
90 * This will give us a plane rect's position in Timing Active. However
91 * we have to remove the fractional. The rule is that we find left/right
92 * and top/bottom positions and round the value to the adjacent integer.
93 *
94 * Stream Source Space
95 * ------------
96 * __________________________________________________
97 * |Stream Source (1920 x 1200) ^ |
98 * | y |
99 * | <------- w --------|> |
100 * | __________________V |
101 * |<-- x -->|Plane//////////////| ^ |
102 * | |(pre scale)////////| | |
103 * | |///////////////////| | |
104 * | |///////////////////| h |
105 * | |///////////////////| | |
106 * | |///////////////////| | |
107 * | |///////////////////| V |
108 * | |
109 * | |
110 * |__________________________________________________|
111 *
112 *
113 * Timing Active Space
114 * ---------------------------------
115 *
116 * Timing Active (2560 x 1440)
117 * __________________________________________________
118 * |*****| Stteam Destination (2304 x 1440) |*****|
119 * |*****| |*****|
120 * |<128>| |*****|
121 * |*****| __________________ |*****|
122 * |*****| |Plane/////////////| |*****|
123 * |*****| |(post scale)//////| |*****|
124 * |*****| |//////////////////| |*****|
125 * |*****| |//////////////////| |*****|
126 * |*****| |//////////////////| |*****|
127 * |*****| |//////////////////| |*****|
128 * |*****| |*****|
129 * |*****| |*****|
130 * |*****| |*****|
131 * |*****|______________________________________|*****|
132 *
133 * So the resulting formulas are shown below:
134 *
135 * recout_x = 128 + round(plane_x * 2304 / 1920)
136 * recout_w = 128 + round((plane_x + plane_w) * 2304 / 1920) - recout_x
137 * recout_y = 0 + round(plane_y * 1440 / 1200)
138 * recout_h = 0 + round((plane_y + plane_h) * 1440 / 1200) - recout_y
139 *
140 * NOTE: fixed point division is not error free. To reduce errors
141 * introduced by fixed point division, we divide only after
142 * multiplication is complete.
143 */
144 const struct spl_rect *stream_src = &spl_in->basic_out.src_rect;
145 const struct spl_rect *stream_dst = &spl_in->basic_out.dst_rect;
146 struct spl_rect rec_out = {0};
147 struct spl_fixed31_32 temp;
148
149
150 temp = spl_fixpt_from_fraction(rec_in->x * (long long)stream_dst->width,
151 stream_src->width);
152 rec_out.x = stream_dst->x + spl_fixpt_round(temp);
153
154 temp = spl_fixpt_from_fraction(
155 (rec_in->x + rec_in->width) * (long long)stream_dst->width,
156 stream_src->width);
157 rec_out.width = stream_dst->x + spl_fixpt_round(temp) - rec_out.x;
158
159 temp = spl_fixpt_from_fraction(rec_in->y * (long long)stream_dst->height,
160 stream_src->height);
161 rec_out.y = stream_dst->y + spl_fixpt_round(temp);
162
163 temp = spl_fixpt_from_fraction(
164 (rec_in->y + rec_in->height) * (long long)stream_dst->height,
165 stream_src->height);
166 rec_out.height = stream_dst->y + spl_fixpt_round(temp) - rec_out.y;
167
168 return rec_out;
169 }
170
calculate_mpc_slice_in_timing_active(struct spl_in * spl_in,struct spl_rect * plane_clip_rec)171 static struct spl_rect calculate_mpc_slice_in_timing_active(
172 struct spl_in *spl_in,
173 struct spl_rect *plane_clip_rec)
174 {
175 bool use_recout_width_aligned =
176 spl_in->basic_in.num_h_slices_recout_width_align.use_recout_width_aligned;
177 int mpc_slice_count =
178 spl_in->basic_in.num_h_slices_recout_width_align.num_slices_recout_width.mpc_num_h_slices;
179 int recout_width_align =
180 spl_in->basic_in.num_h_slices_recout_width_align.num_slices_recout_width.mpc_recout_width_align;
181 int mpc_slice_idx = spl_in->basic_in.mpc_h_slice_index;
182 int epimo = mpc_slice_count - plane_clip_rec->width % mpc_slice_count - 1;
183 struct spl_rect mpc_rec;
184
185 if (use_recout_width_aligned) {
186 mpc_rec.width = recout_width_align;
187 if ((mpc_rec.width * (mpc_slice_idx + 1)) > plane_clip_rec->width) {
188 mpc_rec.width = plane_clip_rec->width % recout_width_align;
189 mpc_rec.x = plane_clip_rec->x + recout_width_align * mpc_slice_idx;
190 } else
191 mpc_rec.x = plane_clip_rec->x + mpc_rec.width * mpc_slice_idx;
192 mpc_rec.height = plane_clip_rec->height;
193 mpc_rec.y = plane_clip_rec->y;
194
195 } else {
196 mpc_rec.width = plane_clip_rec->width / mpc_slice_count;
197 mpc_rec.x = plane_clip_rec->x + mpc_rec.width * mpc_slice_idx;
198 mpc_rec.height = plane_clip_rec->height;
199 mpc_rec.y = plane_clip_rec->y;
200 }
201 SPL_ASSERT(mpc_slice_count == 1 ||
202 spl_in->basic_out.view_format != SPL_VIEW_3D_SIDE_BY_SIDE ||
203 mpc_rec.width % 2 == 0);
204
205 /* extra pixels in the division remainder need to go to pipes after
206 * the extra pixel index minus one(epimo) defined here as:
207 */
208 if (mpc_slice_idx > epimo) {
209 mpc_rec.x += mpc_slice_idx - epimo - 1;
210 mpc_rec.width += 1;
211 }
212
213 if (spl_in->basic_out.view_format == SPL_VIEW_3D_TOP_AND_BOTTOM) {
214 SPL_ASSERT(mpc_rec.height % 2 == 0);
215 mpc_rec.height /= 2;
216 }
217 return mpc_rec;
218 }
219
calculate_odm_slice_in_timing_active(struct spl_in * spl_in)220 static struct spl_rect calculate_odm_slice_in_timing_active(struct spl_in *spl_in)
221 {
222 int odm_slice_count = spl_in->basic_out.odm_combine_factor;
223 int odm_slice_idx = spl_in->odm_slice_index;
224 bool is_last_odm_slice = (odm_slice_idx + 1) == odm_slice_count;
225 int h_active = spl_in->basic_out.output_size.width;
226 int v_active = spl_in->basic_out.output_size.height;
227 int odm_slice_width;
228 struct spl_rect odm_rec;
229
230 if (spl_in->basic_out.odm_combine_factor > 0) {
231 odm_slice_width = h_active / odm_slice_count;
232 /*
233 * deprecated, caller must pass in odm slice rect i.e OPP input
234 * rect in timing active for the new interface.
235 */
236 if (spl_in->basic_out.use_two_pixels_per_container && (odm_slice_width % 2))
237 odm_slice_width++;
238
239 odm_rec.x = odm_slice_width * odm_slice_idx;
240 odm_rec.width = is_last_odm_slice ?
241 /* last slice width is the reminder of h_active */
242 h_active - odm_slice_width * (odm_slice_count - 1) :
243 /* odm slice width is the floor of h_active / count */
244 odm_slice_width;
245 odm_rec.y = 0;
246 odm_rec.height = v_active;
247
248 return odm_rec;
249 }
250
251 return spl_in->basic_out.odm_slice_rect;
252 }
253
spl_calculate_recout(struct spl_in * spl_in,struct spl_scratch * spl_scratch,struct spl_out * spl_out)254 static void spl_calculate_recout(struct spl_in *spl_in, struct spl_scratch *spl_scratch, struct spl_out *spl_out)
255 {
256 /*
257 * A plane clip represents the desired plane size and position in Stream
258 * Source Space. Stream Source is the destination where all planes are
259 * blended (i.e. positioned, scaled and overlaid). It is a canvas where
260 * all planes associated with the current stream are drawn together.
261 * After Stream Source is completed, we will further scale and
262 * reposition the entire canvas of the stream source to Stream
263 * Destination in Timing Active Space. This could be due to display
264 * overscan adjustment where we will need to rescale and reposition all
265 * the planes so they can fit into a TV with overscan or downscale
266 * upscale features such as GPU scaling or VSR.
267 *
268 * This two step blending is a virtual procedure in software. In
269 * hardware there is no such thing as Stream Source. all planes are
270 * blended once in Timing Active Space. Software virtualizes a Stream
271 * Source space to decouple the math complicity so scaling param
272 * calculation focuses on one step at a time.
273 *
274 * In the following two diagrams, user applied 10% overscan adjustment
275 * so the Stream Source needs to be scaled down a little before mapping
276 * to Timing Active Space. As a result the Plane Clip is also scaled
277 * down by the same ratio, Plane Clip position (i.e. x and y) with
278 * respect to Stream Source is also scaled down. To map it in Timing
279 * Active Space additional x and y offsets from Stream Destination are
280 * added to Plane Clip as well.
281 *
282 * Stream Source Space
283 * ------------
284 * __________________________________________________
285 * |Stream Source (3840 x 2160) ^ |
286 * | y |
287 * | | |
288 * | __________________V |
289 * |<-- x -->|Plane Clip/////////| |
290 * | |(pre scale)////////| |
291 * | |///////////////////| |
292 * | |///////////////////| |
293 * | |///////////////////| |
294 * | |///////////////////| |
295 * | |///////////////////| |
296 * | |
297 * | |
298 * |__________________________________________________|
299 *
300 *
301 * Timing Active Space (3840 x 2160)
302 * ---------------------------------
303 *
304 * Timing Active
305 * __________________________________________________
306 * | y_____________________________________________ |
307 * |x |Stream Destination (3456 x 1944) | |
308 * | | | |
309 * | | __________________ | |
310 * | | |Plane Clip////////| | |
311 * | | |(post scale)//////| | |
312 * | | |//////////////////| | |
313 * | | |//////////////////| | |
314 * | | |//////////////////| | |
315 * | | |//////////////////| | |
316 * | | | |
317 * | | | |
318 * | |____________________________________________| |
319 * |__________________________________________________|
320 *
321 *
322 * In Timing Active Space a plane clip could be further sliced into
323 * pieces called MPC slices. Each Pipe Context is responsible for
324 * processing only one MPC slice so the plane processing workload can be
325 * distributed to multiple DPP Pipes. MPC slices could be blended
326 * together to a single ODM slice. Each ODM slice is responsible for
327 * processing a portion of Timing Active divided horizontally so the
328 * output pixel processing workload can be distributed to multiple OPP
329 * pipes. All ODM slices are mapped together in ODM block so all MPC
330 * slices belong to different ODM slices could be pieced together to
331 * form a single image in Timing Active. MPC slices must belong to
332 * single ODM slice. If an MPC slice goes across ODM slice boundary, it
333 * needs to be divided into two MPC slices one for each ODM slice.
334 *
335 * In the following diagram the output pixel processing workload is
336 * divided horizontally into two ODM slices one for each OPP blend tree.
337 * OPP0 blend tree is responsible for processing left half of Timing
338 * Active, while OPP2 blend tree is responsible for processing right
339 * half.
340 *
341 * The plane has two MPC slices. However since the right MPC slice goes
342 * across ODM boundary, two DPP pipes are needed one for each OPP blend
343 * tree. (i.e. DPP1 for OPP0 blend tree and DPP2 for OPP2 blend tree).
344 *
345 * Assuming that we have a Pipe Context associated with OPP0 and DPP1
346 * working on processing the plane in the diagram. We want to know the
347 * width and height of the shaded rectangle and its relative position
348 * with respect to the ODM slice0. This is called the recout of the pipe
349 * context.
350 *
351 * Planes can be at arbitrary size and position and there could be an
352 * arbitrary number of MPC and ODM slices. The algorithm needs to take
353 * all scenarios into account.
354 *
355 * Timing Active Space (3840 x 2160)
356 * ---------------------------------
357 *
358 * Timing Active
359 * __________________________________________________
360 * |OPP0(ODM slice0)^ |OPP2(ODM slice1) |
361 * | y | |
362 * | | <- w -> |
363 * | _____V________|____ |
364 * | |DPP0 ^ |DPP1 |DPP2| |
365 * |<------ x |-----|->|/////| | |
366 * | | | |/////| | |
367 * | | h |/////| | |
368 * | | | |/////| | |
369 * | |_____V__|/////|____| |
370 * | | |
371 * | | |
372 * | | |
373 * |_________________________|________________________|
374 *
375 *
376 */
377 struct spl_rect plane_clip;
378 struct spl_rect mpc_slice_of_plane_clip;
379 struct spl_rect odm_slice;
380 struct spl_rect overlapping_area;
381
382 plane_clip = calculate_plane_rec_in_timing_active(spl_in,
383 &spl_in->basic_in.clip_rect);
384 /* guard plane clip from drawing beyond stream dst here */
385 plane_clip = intersect_rec(&plane_clip,
386 &spl_in->basic_out.dst_rect);
387 mpc_slice_of_plane_clip = calculate_mpc_slice_in_timing_active(
388 spl_in, &plane_clip);
389 odm_slice = calculate_odm_slice_in_timing_active(spl_in);
390 overlapping_area = intersect_rec(&mpc_slice_of_plane_clip, &odm_slice);
391
392 if (overlapping_area.height > 0 &&
393 overlapping_area.width > 0) {
394 /* shift the overlapping area so it is with respect to current
395 * ODM slice's position
396 */
397 spl_scratch->scl_data.recout = shift_rec(
398 &overlapping_area,
399 -odm_slice.x, -odm_slice.y);
400 spl_scratch->scl_data.recout.height -=
401 spl_in->debug.visual_confirm_base_offset;
402 spl_scratch->scl_data.recout.height -=
403 spl_in->debug.visual_confirm_dpp_offset;
404 } else
405 /* if there is no overlap, zero recout */
406 memset(&spl_scratch->scl_data.recout, 0,
407 sizeof(struct spl_rect));
408 }
409
410 /* Calculate scaling ratios */
spl_calculate_scaling_ratios(struct spl_in * spl_in,struct spl_scratch * spl_scratch,struct spl_out * spl_out)411 static void spl_calculate_scaling_ratios(struct spl_in *spl_in,
412 struct spl_scratch *spl_scratch,
413 struct spl_out *spl_out)
414 {
415 const int in_w = spl_in->basic_out.src_rect.width;
416 const int in_h = spl_in->basic_out.src_rect.height;
417 const int out_w = spl_in->basic_out.dst_rect.width;
418 const int out_h = spl_in->basic_out.dst_rect.height;
419 struct spl_rect surf_src = spl_in->basic_in.src_rect;
420
421 /*Swap surf_src height and width since scaling ratios are in recout rotation*/
422 if (spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_90 ||
423 spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_270)
424 spl_swap(surf_src.height, surf_src.width);
425
426 spl_scratch->scl_data.ratios.horz = spl_fixpt_from_fraction(
427 surf_src.width,
428 spl_in->basic_in.dst_rect.width);
429 spl_scratch->scl_data.ratios.vert = spl_fixpt_from_fraction(
430 surf_src.height,
431 spl_in->basic_in.dst_rect.height);
432
433 if (spl_in->basic_out.view_format == SPL_VIEW_3D_SIDE_BY_SIDE)
434 spl_scratch->scl_data.ratios.horz.value *= 2;
435 else if (spl_in->basic_out.view_format == SPL_VIEW_3D_TOP_AND_BOTTOM)
436 spl_scratch->scl_data.ratios.vert.value *= 2;
437
438 spl_scratch->scl_data.ratios.vert.value = spl_div64_s64(
439 spl_scratch->scl_data.ratios.vert.value * in_h, out_h);
440 spl_scratch->scl_data.ratios.horz.value = spl_div64_s64(
441 spl_scratch->scl_data.ratios.horz.value * in_w, out_w);
442
443 spl_scratch->scl_data.ratios.horz_c = spl_scratch->scl_data.ratios.horz;
444 spl_scratch->scl_data.ratios.vert_c = spl_scratch->scl_data.ratios.vert;
445
446 if (spl_is_yuv420(spl_in->basic_in.format)) {
447 spl_scratch->scl_data.ratios.horz_c.value /= 2;
448 spl_scratch->scl_data.ratios.vert_c.value /= 2;
449 }
450 spl_scratch->scl_data.ratios.horz = spl_fixpt_truncate(
451 spl_scratch->scl_data.ratios.horz, 19);
452 spl_scratch->scl_data.ratios.vert = spl_fixpt_truncate(
453 spl_scratch->scl_data.ratios.vert, 19);
454 spl_scratch->scl_data.ratios.horz_c = spl_fixpt_truncate(
455 spl_scratch->scl_data.ratios.horz_c, 19);
456 spl_scratch->scl_data.ratios.vert_c = spl_fixpt_truncate(
457 spl_scratch->scl_data.ratios.vert_c, 19);
458
459 /*
460 * Coefficient table and some registers are different based on ratio
461 * that is output/input. Currently we calculate input/output
462 * Store 1/ratio in recip_ratio for those lookups
463 */
464 spl_scratch->scl_data.recip_ratios.horz = spl_fixpt_recip(
465 spl_scratch->scl_data.ratios.horz);
466 spl_scratch->scl_data.recip_ratios.vert = spl_fixpt_recip(
467 spl_scratch->scl_data.ratios.vert);
468 spl_scratch->scl_data.recip_ratios.horz_c = spl_fixpt_recip(
469 spl_scratch->scl_data.ratios.horz_c);
470 spl_scratch->scl_data.recip_ratios.vert_c = spl_fixpt_recip(
471 spl_scratch->scl_data.ratios.vert_c);
472 }
473
474 /* Calculate Viewport size */
spl_calculate_viewport_size(struct spl_in * spl_in,struct spl_scratch * spl_scratch)475 static void spl_calculate_viewport_size(struct spl_in *spl_in, struct spl_scratch *spl_scratch)
476 {
477 spl_scratch->scl_data.viewport.width = spl_fixpt_ceil(spl_fixpt_mul_int(spl_scratch->scl_data.ratios.horz,
478 spl_scratch->scl_data.recout.width));
479 spl_scratch->scl_data.viewport.height = spl_fixpt_ceil(spl_fixpt_mul_int(spl_scratch->scl_data.ratios.vert,
480 spl_scratch->scl_data.recout.height));
481 spl_scratch->scl_data.viewport_c.width = spl_fixpt_ceil(spl_fixpt_mul_int(spl_scratch->scl_data.ratios.horz_c,
482 spl_scratch->scl_data.recout.width));
483 spl_scratch->scl_data.viewport_c.height = spl_fixpt_ceil(spl_fixpt_mul_int(spl_scratch->scl_data.ratios.vert_c,
484 spl_scratch->scl_data.recout.height));
485 if (spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_90 ||
486 spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_270) {
487 spl_swap(spl_scratch->scl_data.viewport.width, spl_scratch->scl_data.viewport.height);
488 spl_swap(spl_scratch->scl_data.viewport_c.width, spl_scratch->scl_data.viewport_c.height);
489 }
490 }
491
spl_get_vp_scan_direction(enum spl_rotation_angle rotation,bool horizontal_mirror,bool * orthogonal_rotation,bool * flip_vert_scan_dir,bool * flip_horz_scan_dir)492 static void spl_get_vp_scan_direction(enum spl_rotation_angle rotation,
493 bool horizontal_mirror,
494 bool *orthogonal_rotation,
495 bool *flip_vert_scan_dir,
496 bool *flip_horz_scan_dir)
497 {
498 *orthogonal_rotation = false;
499 *flip_vert_scan_dir = false;
500 *flip_horz_scan_dir = false;
501 if (rotation == SPL_ROTATION_ANGLE_180) {
502 *flip_vert_scan_dir = true;
503 *flip_horz_scan_dir = true;
504 } else if (rotation == SPL_ROTATION_ANGLE_90) {
505 *orthogonal_rotation = true;
506 *flip_horz_scan_dir = true;
507 } else if (rotation == SPL_ROTATION_ANGLE_270) {
508 *orthogonal_rotation = true;
509 *flip_vert_scan_dir = true;
510 }
511
512 if (horizontal_mirror)
513 *flip_horz_scan_dir = !*flip_horz_scan_dir;
514 }
515
516 /*
517 * We completely calculate vp offset, size and inits here based entirely on scaling
518 * ratios and recout for pixel perfect pipe combine.
519 */
spl_calculate_init_and_vp(bool flip_scan_dir,int recout_offset_within_recout_full,int recout_size,int src_size,int taps,struct spl_fixed31_32 ratio,struct spl_fixed31_32 init_adj,struct spl_fixed31_32 * init,int * vp_offset,int * vp_size)520 static void spl_calculate_init_and_vp(bool flip_scan_dir,
521 int recout_offset_within_recout_full,
522 int recout_size,
523 int src_size,
524 int taps,
525 struct spl_fixed31_32 ratio,
526 struct spl_fixed31_32 init_adj,
527 struct spl_fixed31_32 *init,
528 int *vp_offset,
529 int *vp_size)
530 {
531 struct spl_fixed31_32 temp;
532 int int_part;
533
534 /*
535 * First of the taps starts sampling pixel number <init_int_part> corresponding to recout
536 * pixel 1. Next recout pixel samples int part of <init + scaling ratio> and so on.
537 * All following calculations are based on this logic.
538 *
539 * Init calculated according to formula:
540 * init = (scaling_ratio + number_of_taps + 1) / 2
541 * init_bot = init + scaling_ratio
542 * to get pixel perfect combine add the fraction from calculating vp offset
543 */
544 temp = spl_fixpt_mul_int(ratio, recout_offset_within_recout_full);
545 *vp_offset = spl_fixpt_floor(temp);
546 temp.value &= 0xffffffff;
547 *init = spl_fixpt_add(spl_fixpt_div_int(spl_fixpt_add_int(ratio, taps + 1), 2), temp);
548 *init = spl_fixpt_add(*init, init_adj);
549 *init = spl_fixpt_truncate(*init, 19);
550
551 /*
552 * If viewport has non 0 offset and there are more taps than covered by init then
553 * we should decrease the offset and increase init so we are never sampling
554 * outside of viewport.
555 */
556 int_part = spl_fixpt_floor(*init);
557 if (int_part < taps) {
558 int_part = taps - int_part;
559 if (int_part > *vp_offset)
560 int_part = *vp_offset;
561 *vp_offset -= int_part;
562 *init = spl_fixpt_add_int(*init, int_part);
563 }
564 /*
565 * If taps are sampling outside of viewport at end of recout and there are more pixels
566 * available in the surface we should increase the viewport size, regardless set vp to
567 * only what is used.
568 */
569 temp = spl_fixpt_add(*init, spl_fixpt_mul_int(ratio, recout_size - 1));
570 *vp_size = spl_fixpt_floor(temp);
571 if (*vp_size + *vp_offset > src_size)
572 *vp_size = src_size - *vp_offset;
573
574 /* We did all the math assuming we are scanning same direction as display does,
575 * however mirror/rotation changes how vp scans vs how it is offset. If scan direction
576 * is flipped we simply need to calculate offset from the other side of plane.
577 * Note that outside of viewport all scaling hardware works in recout space.
578 */
579 if (flip_scan_dir)
580 *vp_offset = src_size - *vp_offset - *vp_size;
581 }
582
583 /*Calculate inits and viewport */
spl_calculate_inits_and_viewports(struct spl_in * spl_in,struct spl_scratch * spl_scratch)584 static void spl_calculate_inits_and_viewports(struct spl_in *spl_in,
585 struct spl_scratch *spl_scratch)
586 {
587 struct spl_rect src = spl_in->basic_in.src_rect;
588 struct spl_rect recout_dst_in_active_timing;
589 struct spl_rect recout_clip_in_active_timing;
590 struct spl_rect recout_clip_in_recout_dst;
591 struct spl_rect overlap_in_active_timing;
592 struct spl_rect odm_slice = calculate_odm_slice_in_timing_active(spl_in);
593 int vpc_div = spl_is_subsampled_format(spl_in->basic_in.format) ? 2 : 1;
594 bool orthogonal_rotation, flip_vert_scan_dir, flip_horz_scan_dir;
595 struct spl_fixed31_32 init_adj_h = spl_fixpt_zero;
596 struct spl_fixed31_32 init_adj_v = spl_fixpt_zero;
597
598 recout_clip_in_active_timing = shift_rec(
599 &spl_scratch->scl_data.recout, odm_slice.x, odm_slice.y);
600 recout_dst_in_active_timing = calculate_plane_rec_in_timing_active(
601 spl_in, &spl_in->basic_in.dst_rect);
602 overlap_in_active_timing = intersect_rec(&recout_clip_in_active_timing,
603 &recout_dst_in_active_timing);
604 if (overlap_in_active_timing.width > 0 &&
605 overlap_in_active_timing.height > 0)
606 recout_clip_in_recout_dst = shift_rec(&overlap_in_active_timing,
607 -recout_dst_in_active_timing.x,
608 -recout_dst_in_active_timing.y);
609 else
610 memset(&recout_clip_in_recout_dst, 0, sizeof(struct spl_rect));
611 /*
612 * Work in recout rotation since that requires less transformations
613 */
614 spl_get_vp_scan_direction(
615 spl_in->basic_in.rotation,
616 spl_in->basic_in.horizontal_mirror,
617 &orthogonal_rotation,
618 &flip_vert_scan_dir,
619 &flip_horz_scan_dir);
620
621 if (spl_is_subsampled_format(spl_in->basic_in.format)) {
622 /* this gives the direction of the cositing (negative will move
623 * left, right otherwise)
624 */
625 int sign = 1;
626
627 switch (spl_in->basic_in.cositing) {
628
629 case CHROMA_COSITING_TOPLEFT:
630 init_adj_h = spl_fixpt_from_fraction(sign, 4);
631 init_adj_v = spl_fixpt_from_fraction(sign, 4);
632 break;
633 case CHROMA_COSITING_LEFT:
634 init_adj_h = spl_fixpt_from_fraction(sign, 4);
635 init_adj_v = spl_fixpt_zero;
636 break;
637 case CHROMA_COSITING_NONE:
638 default:
639 init_adj_h = spl_fixpt_zero;
640 init_adj_v = spl_fixpt_zero;
641 break;
642 }
643 }
644
645 if (orthogonal_rotation) {
646 spl_swap(src.width, src.height);
647 spl_swap(flip_vert_scan_dir, flip_horz_scan_dir);
648 spl_swap(init_adj_h, init_adj_v);
649 }
650
651 spl_calculate_init_and_vp(
652 flip_horz_scan_dir,
653 recout_clip_in_recout_dst.x,
654 spl_scratch->scl_data.recout.width,
655 src.width,
656 spl_scratch->scl_data.taps.h_taps,
657 spl_scratch->scl_data.ratios.horz,
658 spl_fixpt_zero,
659 &spl_scratch->scl_data.inits.h,
660 &spl_scratch->scl_data.viewport.x,
661 &spl_scratch->scl_data.viewport.width);
662 spl_calculate_init_and_vp(
663 flip_horz_scan_dir,
664 recout_clip_in_recout_dst.x,
665 spl_scratch->scl_data.recout.width,
666 src.width / vpc_div,
667 spl_scratch->scl_data.taps.h_taps_c,
668 spl_scratch->scl_data.ratios.horz_c,
669 init_adj_h,
670 &spl_scratch->scl_data.inits.h_c,
671 &spl_scratch->scl_data.viewport_c.x,
672 &spl_scratch->scl_data.viewport_c.width);
673 spl_calculate_init_and_vp(
674 flip_vert_scan_dir,
675 recout_clip_in_recout_dst.y,
676 spl_scratch->scl_data.recout.height,
677 src.height,
678 spl_scratch->scl_data.taps.v_taps,
679 spl_scratch->scl_data.ratios.vert,
680 spl_fixpt_zero,
681 &spl_scratch->scl_data.inits.v,
682 &spl_scratch->scl_data.viewport.y,
683 &spl_scratch->scl_data.viewport.height);
684 spl_calculate_init_and_vp(
685 flip_vert_scan_dir,
686 recout_clip_in_recout_dst.y,
687 spl_scratch->scl_data.recout.height,
688 src.height / vpc_div,
689 spl_scratch->scl_data.taps.v_taps_c,
690 spl_scratch->scl_data.ratios.vert_c,
691 init_adj_v,
692 &spl_scratch->scl_data.inits.v_c,
693 &spl_scratch->scl_data.viewport_c.y,
694 &spl_scratch->scl_data.viewport_c.height);
695 if (orthogonal_rotation) {
696 spl_swap(spl_scratch->scl_data.viewport.x, spl_scratch->scl_data.viewport.y);
697 spl_swap(spl_scratch->scl_data.viewport.width, spl_scratch->scl_data.viewport.height);
698 spl_swap(spl_scratch->scl_data.viewport_c.x, spl_scratch->scl_data.viewport_c.y);
699 spl_swap(spl_scratch->scl_data.viewport_c.width, spl_scratch->scl_data.viewport_c.height);
700 }
701 spl_scratch->scl_data.viewport.x += src.x;
702 spl_scratch->scl_data.viewport.y += src.y;
703 SPL_ASSERT(src.x % vpc_div == 0 && src.y % vpc_div == 0);
704 spl_scratch->scl_data.viewport_c.x += src.x / vpc_div;
705 spl_scratch->scl_data.viewport_c.y += src.y / vpc_div;
706 }
707
spl_handle_3d_recout(struct spl_in * spl_in,struct spl_rect * recout)708 static void spl_handle_3d_recout(struct spl_in *spl_in, struct spl_rect *recout)
709 {
710 /*
711 * Handle side by side and top bottom 3d recout offsets after vp calculation
712 * since 3d is special and needs to calculate vp as if there is no recout offset
713 * This may break with rotation, good thing we aren't mixing hw rotation and 3d
714 */
715 if (spl_in->basic_in.mpc_h_slice_index) {
716 SPL_ASSERT(spl_in->basic_in.rotation == SPL_ROTATION_ANGLE_0 ||
717 (spl_in->basic_out.view_format != SPL_VIEW_3D_TOP_AND_BOTTOM &&
718 spl_in->basic_out.view_format != SPL_VIEW_3D_SIDE_BY_SIDE));
719 if (spl_in->basic_out.view_format == SPL_VIEW_3D_TOP_AND_BOTTOM)
720 recout->y += recout->height;
721 else if (spl_in->basic_out.view_format == SPL_VIEW_3D_SIDE_BY_SIDE)
722 recout->x += recout->width;
723 }
724 }
725
spl_clamp_viewport(struct spl_rect * viewport)726 static void spl_clamp_viewport(struct spl_rect *viewport)
727 {
728 /* Clamp minimum viewport size */
729 if (viewport->height < MIN_VIEWPORT_SIZE)
730 viewport->height = MIN_VIEWPORT_SIZE;
731 if (viewport->width < MIN_VIEWPORT_SIZE)
732 viewport->width = MIN_VIEWPORT_SIZE;
733 }
734
spl_get_dscl_mode(const struct spl_in * spl_in,const struct spl_scaler_data * data,bool enable_isharp,bool enable_easf)735 static enum scl_mode spl_get_dscl_mode(const struct spl_in *spl_in,
736 const struct spl_scaler_data *data,
737 bool enable_isharp, bool enable_easf)
738 {
739 const long long one = spl_fixpt_one.value;
740 enum spl_pixel_format pixel_format = spl_in->basic_in.format;
741
742 /* Bypass if ratio is 1:1 with no ISHARP or force scale on */
743 if (data->ratios.horz.value == one
744 && data->ratios.vert.value == one
745 && data->ratios.horz_c.value == one
746 && data->ratios.vert_c.value == one
747 && !spl_in->basic_out.always_scale
748 && !enable_isharp)
749 return SCL_MODE_SCALING_444_BYPASS;
750
751 if (!spl_is_subsampled_format(pixel_format)) {
752 if (spl_is_video_format(pixel_format))
753 return SCL_MODE_SCALING_444_YCBCR_ENABLE;
754 else
755 return SCL_MODE_SCALING_444_RGB_ENABLE;
756 }
757
758 /*
759 * Bypass YUV if Y is 1:1 with no ISHARP
760 * Do not bypass UV at 1:1 for cositing to be applied
761 */
762 if (!enable_isharp) {
763 if (data->ratios.horz.value == one && data->ratios.vert.value == one)
764 return SCL_MODE_SCALING_420_LUMA_BYPASS;
765 }
766
767 return SCL_MODE_SCALING_420_YCBCR_ENABLE;
768 }
769
spl_choose_lls_policy(enum spl_pixel_format format,enum spl_transfer_func_type tf_type,enum spl_transfer_func_predefined tf_predefined_type,enum linear_light_scaling * lls_pref)770 static bool spl_choose_lls_policy(enum spl_pixel_format format,
771 enum spl_transfer_func_type tf_type,
772 enum spl_transfer_func_predefined tf_predefined_type,
773 enum linear_light_scaling *lls_pref)
774 {
775 if (spl_is_video_format(format)) {
776 *lls_pref = LLS_PREF_NO;
777 if ((tf_type == SPL_TF_TYPE_PREDEFINED) ||
778 (tf_type == SPL_TF_TYPE_DISTRIBUTED_POINTS))
779 return true;
780 } else { /* RGB or YUV444 */
781 if ((tf_type == SPL_TF_TYPE_PREDEFINED) ||
782 (tf_type == SPL_TF_TYPE_BYPASS)) {
783 *lls_pref = LLS_PREF_YES;
784 return true;
785 }
786 }
787 *lls_pref = LLS_PREF_NO;
788 return false;
789 }
790
791 /* Enable EASF ?*/
enable_easf(struct spl_in * spl_in,struct spl_scratch * spl_scratch)792 static bool enable_easf(struct spl_in *spl_in, struct spl_scratch *spl_scratch)
793 {
794 int vratio = 0;
795 int hratio = 0;
796 bool skip_easf = false;
797 bool lls_enable_easf = true;
798
799 if (spl_in->disable_easf)
800 skip_easf = true;
801
802 vratio = spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert);
803 hratio = spl_fixpt_ceil(spl_scratch->scl_data.ratios.horz);
804
805 /*
806 * No EASF support for downscaling > 2:1
807 * EASF support for upscaling or downscaling up to 2:1
808 */
809 if ((vratio > 2) || (hratio > 2))
810 skip_easf = true;
811
812 /*
813 * If lls_pref is LLS_PREF_DONT_CARE, then use pixel format and transfer
814 * function to determine whether to use LINEAR or NONLINEAR scaling
815 */
816 if (spl_in->lls_pref == LLS_PREF_DONT_CARE)
817 lls_enable_easf = spl_choose_lls_policy(spl_in->basic_in.format,
818 spl_in->basic_in.tf_type, spl_in->basic_in.tf_predefined_type,
819 &spl_in->lls_pref);
820
821 if (!lls_enable_easf)
822 skip_easf = true;
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 */
spl_is_video_fullscreen(struct spl_in * spl_in)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
spl_get_isharp_en(struct spl_in * spl_in,struct spl_scratch * spl_scratch)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 */
spl_get_taps_non_adaptive_scaler(struct spl_scratch * spl_scratch,const struct spl_taps * in_taps)887 static void spl_get_taps_non_adaptive_scaler(
888 struct spl_scratch *spl_scratch, const struct spl_taps *in_taps)
889 {
890 if (in_taps->h_taps == 0) {
891 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.horz) > 1)
892 spl_scratch->scl_data.taps.h_taps = spl_min(2 * spl_fixpt_ceil(
893 spl_scratch->scl_data.ratios.horz), 8);
894 else
895 spl_scratch->scl_data.taps.h_taps = 4;
896 } else
897 spl_scratch->scl_data.taps.h_taps = in_taps->h_taps;
898
899 if (in_taps->v_taps == 0) {
900 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert) > 1)
901 spl_scratch->scl_data.taps.v_taps = spl_min(2 * spl_fixpt_ceil(
902 spl_scratch->scl_data.ratios.vert), 8);
903 else
904 spl_scratch->scl_data.taps.v_taps = 4;
905 } else
906 spl_scratch->scl_data.taps.v_taps = in_taps->v_taps;
907
908 if (in_taps->v_taps_c == 0) {
909 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert_c) > 1)
910 spl_scratch->scl_data.taps.v_taps_c = spl_min(2 * spl_fixpt_ceil(
911 spl_scratch->scl_data.ratios.vert_c), 8);
912 else
913 spl_scratch->scl_data.taps.v_taps_c = 4;
914 } else
915 spl_scratch->scl_data.taps.v_taps_c = in_taps->v_taps_c;
916
917 if (in_taps->h_taps_c == 0) {
918 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.horz_c) > 1)
919 spl_scratch->scl_data.taps.h_taps_c = spl_min(2 * spl_fixpt_ceil(
920 spl_scratch->scl_data.ratios.horz_c), 8);
921 else
922 spl_scratch->scl_data.taps.h_taps_c = 4;
923 } else if ((in_taps->h_taps_c % 2) != 0 && in_taps->h_taps_c != 1)
924 /* Only 1 and even h_taps_c are supported by hw */
925 spl_scratch->scl_data.taps.h_taps_c = in_taps->h_taps_c - 1;
926 else
927 spl_scratch->scl_data.taps.h_taps_c = in_taps->h_taps_c;
928
929 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz))
930 spl_scratch->scl_data.taps.h_taps = 1;
931 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert))
932 spl_scratch->scl_data.taps.v_taps = 1;
933 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz_c))
934 spl_scratch->scl_data.taps.h_taps_c = 1;
935 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert_c))
936 spl_scratch->scl_data.taps.v_taps_c = 1;
937
938 }
939
940 /* Calculate optimal number of taps */
spl_get_optimal_number_of_taps(int max_downscale_src_width,struct spl_in * spl_in,struct spl_scratch * spl_scratch,const struct spl_taps * in_taps,bool * enable_easf_v,bool * enable_easf_h,bool * enable_isharp)941 static bool spl_get_optimal_number_of_taps(
942 int max_downscale_src_width, struct spl_in *spl_in, struct spl_scratch *spl_scratch,
943 const struct spl_taps *in_taps, bool *enable_easf_v, bool *enable_easf_h,
944 bool *enable_isharp)
945 {
946 int num_part_y, num_part_c;
947 int max_taps_y, max_taps_c;
948 int min_taps_y, min_taps_c;
949 enum lb_memory_config lb_config;
950 bool skip_easf = false;
951 bool is_subsampled = spl_is_subsampled_format(spl_in->basic_in.format);
952
953 if (spl_scratch->scl_data.viewport.width > spl_scratch->scl_data.h_active &&
954 max_downscale_src_width != 0 &&
955 spl_scratch->scl_data.viewport.width > max_downscale_src_width) {
956 spl_get_taps_non_adaptive_scaler(spl_scratch, in_taps);
957 *enable_easf_v = false;
958 *enable_easf_h = false;
959 *enable_isharp = false;
960 return false;
961 }
962
963 /* Disable adaptive scaler and sharpener when integer scaling is enabled */
964 if (spl_in->scaling_quality.integer_scaling) {
965 spl_get_taps_non_adaptive_scaler(spl_scratch, in_taps);
966 *enable_easf_v = false;
967 *enable_easf_h = false;
968 *enable_isharp = false;
969 return true;
970 }
971
972 /* Check if we are using EASF or not */
973 skip_easf = enable_easf(spl_in, spl_scratch);
974
975 /*
976 * Set default taps if none are provided
977 * From programming guide: taps = min{ ceil(2*H_RATIO,1), 8} for downscaling
978 * taps = 4 for upscaling
979 */
980 if (skip_easf)
981 spl_get_taps_non_adaptive_scaler(spl_scratch, in_taps);
982 else {
983 if (spl_is_video_format(spl_in->basic_in.format)) {
984 spl_scratch->scl_data.taps.h_taps = 6;
985 spl_scratch->scl_data.taps.v_taps = 6;
986 spl_scratch->scl_data.taps.h_taps_c = 4;
987 spl_scratch->scl_data.taps.v_taps_c = 4;
988 } else { /* RGB */
989 spl_scratch->scl_data.taps.h_taps = 6;
990 spl_scratch->scl_data.taps.v_taps = 6;
991 spl_scratch->scl_data.taps.h_taps_c = 6;
992 spl_scratch->scl_data.taps.v_taps_c = 6;
993 }
994 }
995
996 /*Ensure we can support the requested number of vtaps*/
997 min_taps_y = spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert);
998 min_taps_c = spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert_c);
999
1000 /* Use LB_MEMORY_CONFIG_3 for 4:2:0 */
1001 if (spl_is_yuv420(spl_in->basic_in.format))
1002 lb_config = LB_MEMORY_CONFIG_3;
1003 else
1004 lb_config = LB_MEMORY_CONFIG_0;
1005 // Determine max vtap support by calculating how much line buffer can fit
1006 spl_in->callbacks.spl_calc_lb_num_partitions(spl_in->basic_out.alpha_en, &spl_scratch->scl_data,
1007 lb_config, &num_part_y, &num_part_c);
1008 /* MAX_V_TAPS = MIN (NUM_LINES - MAX(CEILING(V_RATIO,1)-2, 0), 8) */
1009 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert) > 2)
1010 max_taps_y = num_part_y - (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert) - 2);
1011 else
1012 max_taps_y = num_part_y;
1013
1014 if (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert_c) > 2)
1015 max_taps_c = num_part_c - (spl_fixpt_ceil(spl_scratch->scl_data.ratios.vert_c) - 2);
1016 else
1017 max_taps_c = num_part_c;
1018
1019 if (max_taps_y < min_taps_y)
1020 return false;
1021 else if (max_taps_c < min_taps_c)
1022 return false;
1023
1024 if (spl_scratch->scl_data.taps.v_taps > max_taps_y)
1025 spl_scratch->scl_data.taps.v_taps = max_taps_y;
1026
1027 if (spl_scratch->scl_data.taps.v_taps_c > max_taps_c)
1028 spl_scratch->scl_data.taps.v_taps_c = max_taps_c;
1029
1030 if (!skip_easf) {
1031 /*
1032 * RGB ( L + NL ) and Linear HDR support 6x6, 6x4, 6x3, 4x4, 4x3
1033 * NL YUV420 only supports 6x6, 6x4 for Y and 4x4 for UV
1034 *
1035 * If LB does not support 3, 4, or 6 taps, then disable EASF_V
1036 * and only enable EASF_H. So for RGB, support 6x2, 4x2
1037 * and for NL YUV420, support 6x2 for Y and 4x2 for UV
1038 *
1039 * All other cases, have to disable EASF_V and EASF_H
1040 *
1041 * If optimal no of taps is 5, then set it to 4
1042 * If optimal no of taps is 7 or 8, then fine since max tap is 6
1043 *
1044 */
1045 if (spl_scratch->scl_data.taps.v_taps == 5)
1046 spl_scratch->scl_data.taps.v_taps = 4;
1047
1048 if (spl_scratch->scl_data.taps.v_taps_c == 5)
1049 spl_scratch->scl_data.taps.v_taps_c = 4;
1050
1051 if (spl_scratch->scl_data.taps.h_taps == 5)
1052 spl_scratch->scl_data.taps.h_taps = 4;
1053
1054 if (spl_scratch->scl_data.taps.h_taps_c == 5)
1055 spl_scratch->scl_data.taps.h_taps_c = 4;
1056
1057 if (spl_is_video_format(spl_in->basic_in.format)) {
1058 if (spl_scratch->scl_data.taps.h_taps <= 4) {
1059 *enable_easf_v = false;
1060 *enable_easf_h = false;
1061 } else if (spl_scratch->scl_data.taps.v_taps <= 3) {
1062 *enable_easf_v = false;
1063 *enable_easf_h = true;
1064 } else {
1065 *enable_easf_v = true;
1066 *enable_easf_h = true;
1067 }
1068 SPL_ASSERT((spl_scratch->scl_data.taps.v_taps > 1) &&
1069 (spl_scratch->scl_data.taps.v_taps_c > 1));
1070 } else { /* RGB */
1071 if (spl_scratch->scl_data.taps.h_taps <= 3) {
1072 *enable_easf_v = false;
1073 *enable_easf_h = false;
1074 } else if (spl_scratch->scl_data.taps.v_taps < 3) {
1075 *enable_easf_v = false;
1076 *enable_easf_h = true;
1077 } else {
1078 *enable_easf_v = true;
1079 *enable_easf_h = true;
1080 }
1081 SPL_ASSERT(spl_scratch->scl_data.taps.v_taps > 1);
1082 }
1083 } else {
1084 *enable_easf_v = false;
1085 *enable_easf_h = false;
1086 } // end of if prefer_easf
1087
1088 /* Sharpener requires scaler to be enabled, including for 1:1
1089 * Check if ISHARP can be enabled
1090 * If ISHARP is not enabled, set taps to 1 if ratio is 1:1
1091 * except for chroma taps. Keep previous taps so it can
1092 * handle cositing
1093 */
1094
1095 *enable_isharp = spl_get_isharp_en(spl_in, spl_scratch);
1096 if (!*enable_isharp && !spl_in->basic_out.always_scale) {
1097 if ((IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz)) &&
1098 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert))) {
1099 spl_scratch->scl_data.taps.h_taps = 1;
1100 spl_scratch->scl_data.taps.v_taps = 1;
1101
1102 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz_c) && !is_subsampled)
1103 spl_scratch->scl_data.taps.h_taps_c = 1;
1104
1105 if (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert_c) && !is_subsampled)
1106 spl_scratch->scl_data.taps.v_taps_c = 1;
1107
1108 *enable_easf_v = false;
1109 *enable_easf_h = false;
1110 } else {
1111 if ((!*enable_easf_h) &&
1112 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz)))
1113 spl_scratch->scl_data.taps.h_taps = 1;
1114
1115 if ((!*enable_easf_v) &&
1116 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert)))
1117 spl_scratch->scl_data.taps.v_taps = 1;
1118
1119 if ((!*enable_easf_h) && !is_subsampled &&
1120 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.horz_c)))
1121 spl_scratch->scl_data.taps.h_taps_c = 1;
1122
1123 if ((!*enable_easf_v) && !is_subsampled &&
1124 (IDENTITY_RATIO(spl_scratch->scl_data.ratios.vert_c)))
1125 spl_scratch->scl_data.taps.v_taps_c = 1;
1126 }
1127 }
1128 return true;
1129 }
1130
spl_set_black_color_data(enum spl_pixel_format format,struct scl_black_color * scl_black_color)1131 static void spl_set_black_color_data(enum spl_pixel_format format,
1132 struct scl_black_color *scl_black_color)
1133 {
1134 bool ycbcr = spl_is_video_format(format);
1135 if (ycbcr) {
1136 scl_black_color->offset_rgb_y = BLACK_OFFSET_RGB_Y;
1137 scl_black_color->offset_rgb_cbcr = BLACK_OFFSET_CBCR;
1138 } else {
1139 scl_black_color->offset_rgb_y = 0x0;
1140 scl_black_color->offset_rgb_cbcr = 0x0;
1141 }
1142 }
1143
spl_set_manual_ratio_init_data(struct dscl_prog_data * dscl_prog_data,const struct spl_scaler_data * scl_data)1144 static void spl_set_manual_ratio_init_data(struct dscl_prog_data *dscl_prog_data,
1145 const struct spl_scaler_data *scl_data)
1146 {
1147 struct spl_fixed31_32 bot;
1148
1149 dscl_prog_data->ratios.h_scale_ratio = spl_fixpt_u3d19(scl_data->ratios.horz) << 5;
1150 dscl_prog_data->ratios.v_scale_ratio = spl_fixpt_u3d19(scl_data->ratios.vert) << 5;
1151 dscl_prog_data->ratios.h_scale_ratio_c = spl_fixpt_u3d19(scl_data->ratios.horz_c) << 5;
1152 dscl_prog_data->ratios.v_scale_ratio_c = spl_fixpt_u3d19(scl_data->ratios.vert_c) << 5;
1153 /*
1154 * 0.24 format for fraction, first five bits zeroed
1155 */
1156 dscl_prog_data->init.h_filter_init_frac =
1157 spl_fixpt_u0d19(scl_data->inits.h) << 5;
1158 dscl_prog_data->init.h_filter_init_int =
1159 spl_fixpt_floor(scl_data->inits.h);
1160 dscl_prog_data->init.h_filter_init_frac_c =
1161 spl_fixpt_u0d19(scl_data->inits.h_c) << 5;
1162 dscl_prog_data->init.h_filter_init_int_c =
1163 spl_fixpt_floor(scl_data->inits.h_c);
1164 dscl_prog_data->init.v_filter_init_frac =
1165 spl_fixpt_u0d19(scl_data->inits.v) << 5;
1166 dscl_prog_data->init.v_filter_init_int =
1167 spl_fixpt_floor(scl_data->inits.v);
1168 dscl_prog_data->init.v_filter_init_frac_c =
1169 spl_fixpt_u0d19(scl_data->inits.v_c) << 5;
1170 dscl_prog_data->init.v_filter_init_int_c =
1171 spl_fixpt_floor(scl_data->inits.v_c);
1172
1173 bot = spl_fixpt_add(scl_data->inits.v, scl_data->ratios.vert);
1174 dscl_prog_data->init.v_filter_init_bot_frac = spl_fixpt_u0d19(bot) << 5;
1175 dscl_prog_data->init.v_filter_init_bot_int = spl_fixpt_floor(bot);
1176 bot = spl_fixpt_add(scl_data->inits.v_c, scl_data->ratios.vert_c);
1177 dscl_prog_data->init.v_filter_init_bot_frac_c = spl_fixpt_u0d19(bot) << 5;
1178 dscl_prog_data->init.v_filter_init_bot_int_c = spl_fixpt_floor(bot);
1179 }
1180
spl_set_taps_data(struct dscl_prog_data * dscl_prog_data,const struct spl_scaler_data * scl_data)1181 static void spl_set_taps_data(struct dscl_prog_data *dscl_prog_data,
1182 const struct spl_scaler_data *scl_data)
1183 {
1184 dscl_prog_data->taps.v_taps = scl_data->taps.v_taps - 1;
1185 dscl_prog_data->taps.h_taps = scl_data->taps.h_taps - 1;
1186 dscl_prog_data->taps.v_taps_c = scl_data->taps.v_taps_c - 1;
1187 dscl_prog_data->taps.h_taps_c = scl_data->taps.h_taps_c - 1;
1188 }
1189
1190 /* Populate dscl prog data structure from scaler data calculated by SPL */
spl_set_dscl_prog_data(struct spl_in * spl_in,struct spl_scratch * spl_scratch,struct spl_out * spl_out,bool enable_easf_v,bool enable_easf_h,bool enable_isharp)1191 static void spl_set_dscl_prog_data(struct spl_in *spl_in, struct spl_scratch *spl_scratch,
1192 struct spl_out *spl_out, bool enable_easf_v, bool enable_easf_h, bool enable_isharp)
1193 {
1194 struct dscl_prog_data *dscl_prog_data = spl_out->dscl_prog_data;
1195
1196 const struct spl_scaler_data *data = &spl_scratch->scl_data;
1197
1198 struct scl_black_color *scl_black_color = &dscl_prog_data->scl_black_color;
1199
1200 bool enable_easf = enable_easf_v || enable_easf_h;
1201
1202 // Set values for recout
1203 dscl_prog_data->recout = spl_scratch->scl_data.recout;
1204 // Set values for MPC Size
1205 dscl_prog_data->mpc_size.width = spl_scratch->scl_data.h_active;
1206 dscl_prog_data->mpc_size.height = spl_scratch->scl_data.v_active;
1207
1208 // SCL_MODE - Set SCL_MODE data
1209 dscl_prog_data->dscl_mode = spl_get_dscl_mode(spl_in, data, enable_isharp,
1210 enable_easf);
1211
1212 // SCL_BLACK_COLOR
1213 spl_set_black_color_data(spl_in->basic_in.format, scl_black_color);
1214
1215 /* Manually calculate scale ratio and init values */
1216 spl_set_manual_ratio_init_data(dscl_prog_data, data);
1217
1218 // Set HTaps/VTaps
1219 spl_set_taps_data(dscl_prog_data, data);
1220 // Set viewport
1221 dscl_prog_data->viewport = spl_scratch->scl_data.viewport;
1222 // Set viewport_c
1223 dscl_prog_data->viewport_c = spl_scratch->scl_data.viewport_c;
1224 // Set filters data
1225 spl_set_filters_data(dscl_prog_data, data, enable_easf_v, enable_easf_h);
1226 }
1227
1228 /* Calculate C0-C3 coefficients based on HDR_mult */
spl_calculate_c0_c3_hdr(struct dscl_prog_data * dscl_prog_data,uint32_t sdr_white_level_nits)1229 static void spl_calculate_c0_c3_hdr(struct dscl_prog_data *dscl_prog_data, uint32_t sdr_white_level_nits)
1230 {
1231 struct spl_fixed31_32 hdr_mult, c0_mult, c1_mult, c2_mult;
1232 struct spl_fixed31_32 c0_calc, c1_calc, c2_calc;
1233 struct spl_custom_float_format fmt;
1234 uint32_t hdr_multx100_int;
1235
1236 if ((sdr_white_level_nits >= 80) && (sdr_white_level_nits <= 480))
1237 hdr_multx100_int = sdr_white_level_nits * 100 / 80;
1238 else
1239 hdr_multx100_int = 100; /* default for 80 nits otherwise */
1240
1241 hdr_mult = spl_fixpt_from_fraction((long long)hdr_multx100_int, 100LL);
1242 c0_mult = spl_fixpt_from_fraction(2126LL, 10000LL);
1243 c1_mult = spl_fixpt_from_fraction(7152LL, 10000LL);
1244 c2_mult = spl_fixpt_from_fraction(722LL, 10000LL);
1245
1246 c0_calc = spl_fixpt_mul(hdr_mult, spl_fixpt_mul(c0_mult, spl_fixpt_from_fraction(
1247 16384LL, 125LL)));
1248 c1_calc = spl_fixpt_mul(hdr_mult, spl_fixpt_mul(c1_mult, spl_fixpt_from_fraction(
1249 16384LL, 125LL)));
1250 c2_calc = spl_fixpt_mul(hdr_mult, spl_fixpt_mul(c2_mult, spl_fixpt_from_fraction(
1251 16384LL, 125LL)));
1252
1253 fmt.exponenta_bits = 5;
1254 fmt.mantissa_bits = 10;
1255 fmt.sign = true;
1256
1257 // fp1.5.10, C0 coefficient (LN_rec709: HDR_MULT * 0.212600 * 2^14/125)
1258 spl_convert_to_custom_float_format(c0_calc, &fmt, &dscl_prog_data->easf_matrix_c0);
1259 // fp1.5.10, C1 coefficient (LN_rec709: HDR_MULT * 0.715200 * 2^14/125)
1260 spl_convert_to_custom_float_format(c1_calc, &fmt, &dscl_prog_data->easf_matrix_c1);
1261 // fp1.5.10, C2 coefficient (LN_rec709: HDR_MULT * 0.072200 * 2^14/125)
1262 spl_convert_to_custom_float_format(c2_calc, &fmt, &dscl_prog_data->easf_matrix_c2);
1263 dscl_prog_data->easf_matrix_c3 = 0x0; // fp1.5.10, C3 coefficient
1264 }
1265
1266 /* Set EASF data */
spl_set_easf_data(struct spl_scratch * spl_scratch,struct spl_out * spl_out,bool enable_easf_v,bool enable_easf_h,enum linear_light_scaling lls_pref,enum spl_pixel_format format,enum system_setup setup,uint32_t sdr_white_level_nits)1267 static void spl_set_easf_data(struct spl_scratch *spl_scratch, struct spl_out *spl_out, bool enable_easf_v,
1268 bool enable_easf_h, enum linear_light_scaling lls_pref,
1269 enum spl_pixel_format format, enum system_setup setup,
1270 uint32_t sdr_white_level_nits)
1271 {
1272 struct dscl_prog_data *dscl_prog_data = spl_out->dscl_prog_data;
1273 if (enable_easf_v) {
1274 dscl_prog_data->easf_v_en = true;
1275 dscl_prog_data->easf_v_ring = 0;
1276 dscl_prog_data->easf_v_sharp_factor = 0;
1277 dscl_prog_data->easf_v_bf1_en = 1; // 1-bit, BF1 calculation enable, 0=disable, 1=enable
1278 dscl_prog_data->easf_v_bf2_mode = 0xF; // 4-bit, BF2 calculation mode
1279 /* 2-bit, BF3 chroma mode correction calculation mode */
1280 dscl_prog_data->easf_v_bf3_mode = spl_get_v_bf3_mode(
1281 spl_scratch->scl_data.recip_ratios.vert);
1282 /* FP1.5.10 [ minCoef ]*/
1283 dscl_prog_data->easf_v_ringest_3tap_dntilt_uptilt =
1284 spl_get_3tap_dntilt_uptilt_offset(spl_scratch->scl_data.taps.v_taps,
1285 spl_scratch->scl_data.recip_ratios.vert);
1286 /* FP1.5.10 [ upTiltMaxVal ]*/
1287 dscl_prog_data->easf_v_ringest_3tap_uptilt_max =
1288 spl_get_3tap_uptilt_maxval(spl_scratch->scl_data.taps.v_taps,
1289 spl_scratch->scl_data.recip_ratios.vert);
1290 /* FP1.5.10 [ dnTiltSlope ]*/
1291 dscl_prog_data->easf_v_ringest_3tap_dntilt_slope =
1292 spl_get_3tap_dntilt_slope(spl_scratch->scl_data.taps.v_taps,
1293 spl_scratch->scl_data.recip_ratios.vert);
1294 /* FP1.5.10 [ upTilt1Slope ]*/
1295 dscl_prog_data->easf_v_ringest_3tap_uptilt1_slope =
1296 spl_get_3tap_uptilt1_slope(spl_scratch->scl_data.taps.v_taps,
1297 spl_scratch->scl_data.recip_ratios.vert);
1298 /* FP1.5.10 [ upTilt2Slope ]*/
1299 dscl_prog_data->easf_v_ringest_3tap_uptilt2_slope =
1300 spl_get_3tap_uptilt2_slope(spl_scratch->scl_data.taps.v_taps,
1301 spl_scratch->scl_data.recip_ratios.vert);
1302 /* FP1.5.10 [ upTilt2Offset ]*/
1303 dscl_prog_data->easf_v_ringest_3tap_uptilt2_offset =
1304 spl_get_3tap_uptilt2_offset(spl_scratch->scl_data.taps.v_taps,
1305 spl_scratch->scl_data.recip_ratios.vert);
1306 /* FP1.5.10; (2.0) Ring reducer gain for 4 or 6-tap mode [H_REDUCER_GAIN4] */
1307 dscl_prog_data->easf_v_ringest_eventap_reduceg1 =
1308 spl_get_reducer_gain4(spl_scratch->scl_data.taps.v_taps,
1309 spl_scratch->scl_data.recip_ratios.vert);
1310 /* FP1.5.10; (2.5) Ring reducer gain for 6-tap mode [V_REDUCER_GAIN6] */
1311 dscl_prog_data->easf_v_ringest_eventap_reduceg2 =
1312 spl_get_reducer_gain6(spl_scratch->scl_data.taps.v_taps,
1313 spl_scratch->scl_data.recip_ratios.vert);
1314 /* FP1.5.10; (-0.135742) Ring gain for 6-tap set to -139/1024 */
1315 dscl_prog_data->easf_v_ringest_eventap_gain1 =
1316 spl_get_gainRing4(spl_scratch->scl_data.taps.v_taps,
1317 spl_scratch->scl_data.recip_ratios.vert);
1318 /* FP1.5.10; (-0.024414) Ring gain for 6-tap set to -25/1024 */
1319 dscl_prog_data->easf_v_ringest_eventap_gain2 =
1320 spl_get_gainRing6(spl_scratch->scl_data.taps.v_taps,
1321 spl_scratch->scl_data.recip_ratios.vert);
1322 dscl_prog_data->easf_v_bf_maxa = 63; //Vertical Max BF value A in U0.6 format.Selected if V_FCNTL == 0
1323 dscl_prog_data->easf_v_bf_maxb = 63; //Vertical Max BF value A in U0.6 format.Selected if V_FCNTL == 1
1324 dscl_prog_data->easf_v_bf_mina = 0; //Vertical Min BF value A in U0.6 format.Selected if V_FCNTL == 0
1325 dscl_prog_data->easf_v_bf_minb = 0; //Vertical Min BF value A in U0.6 format.Selected if V_FCNTL == 1
1326 if (lls_pref == LLS_PREF_YES) {
1327 dscl_prog_data->easf_v_bf2_flat1_gain = 4; // U1.3, BF2 Flat1 Gain control
1328 dscl_prog_data->easf_v_bf2_flat2_gain = 8; // U4.0, BF2 Flat2 Gain control
1329 dscl_prog_data->easf_v_bf2_roc_gain = 4; // U2.2, Rate Of Change control
1330
1331 dscl_prog_data->easf_v_bf1_pwl_in_seg0 = 0x600; // S0.10, BF1 PWL Segment 0 = -512
1332 dscl_prog_data->easf_v_bf1_pwl_base_seg0 = 0; // U0.6, BF1 Base PWL Segment 0
1333 dscl_prog_data->easf_v_bf1_pwl_slope_seg0 = 3; // S7.3, BF1 Slope PWL Segment 0
1334 dscl_prog_data->easf_v_bf1_pwl_in_seg1 = 0x7EC; // S0.10, BF1 PWL Segment 1 = -20
1335 dscl_prog_data->easf_v_bf1_pwl_base_seg1 = 12; // U0.6, BF1 Base PWL Segment 1
1336 dscl_prog_data->easf_v_bf1_pwl_slope_seg1 = 326; // S7.3, BF1 Slope PWL Segment 1
1337 dscl_prog_data->easf_v_bf1_pwl_in_seg2 = 0; // S0.10, BF1 PWL Segment 2
1338 dscl_prog_data->easf_v_bf1_pwl_base_seg2 = 63; // U0.6, BF1 Base PWL Segment 2
1339 dscl_prog_data->easf_v_bf1_pwl_slope_seg2 = 0; // S7.3, BF1 Slope PWL Segment 2
1340 dscl_prog_data->easf_v_bf1_pwl_in_seg3 = 16; // S0.10, BF1 PWL Segment 3
1341 dscl_prog_data->easf_v_bf1_pwl_base_seg3 = 63; // U0.6, BF1 Base PWL Segment 3
1342 dscl_prog_data->easf_v_bf1_pwl_slope_seg3 = 0x7C8; // S7.3, BF1 Slope PWL Segment 3 = -56
1343 dscl_prog_data->easf_v_bf1_pwl_in_seg4 = 32; // S0.10, BF1 PWL Segment 4
1344 dscl_prog_data->easf_v_bf1_pwl_base_seg4 = 56; // U0.6, BF1 Base PWL Segment 4
1345 dscl_prog_data->easf_v_bf1_pwl_slope_seg4 = 0x7D0; // S7.3, BF1 Slope PWL Segment 4 = -48
1346 dscl_prog_data->easf_v_bf1_pwl_in_seg5 = 48; // S0.10, BF1 PWL Segment 5
1347 dscl_prog_data->easf_v_bf1_pwl_base_seg5 = 50; // U0.6, BF1 Base PWL Segment 5
1348 dscl_prog_data->easf_v_bf1_pwl_slope_seg5 = 0x710; // S7.3, BF1 Slope PWL Segment 5 = -240
1349 dscl_prog_data->easf_v_bf1_pwl_in_seg6 = 64; // S0.10, BF1 PWL Segment 6
1350 dscl_prog_data->easf_v_bf1_pwl_base_seg6 = 20; // U0.6, BF1 Base PWL Segment 6
1351 dscl_prog_data->easf_v_bf1_pwl_slope_seg6 = 0x760; // S7.3, BF1 Slope PWL Segment 6 = -160
1352 dscl_prog_data->easf_v_bf1_pwl_in_seg7 = 80; // S0.10, BF1 PWL Segment 7
1353 dscl_prog_data->easf_v_bf1_pwl_base_seg7 = 0; // U0.6, BF1 Base PWL Segment 7
1354
1355 dscl_prog_data->easf_v_bf3_pwl_in_set0 = 0x000; // FP0.6.6, BF3 Input value PWL Segment 0
1356 dscl_prog_data->easf_v_bf3_pwl_base_set0 = 63; // S0.6, BF3 Base PWL Segment 0
1357 dscl_prog_data->easf_v_bf3_pwl_slope_set0 = 0x12C5; // FP1.6.6, BF3 Slope PWL Segment 0
1358 dscl_prog_data->easf_v_bf3_pwl_in_set1 =
1359 0x0B37; // FP0.6.6, BF3 Input value PWL Segment 1 (0.0078125 * 125^3)
1360 dscl_prog_data->easf_v_bf3_pwl_base_set1 = 62; // S0.6, BF3 Base PWL Segment 1
1361 dscl_prog_data->easf_v_bf3_pwl_slope_set1 =
1362 0x13B8; // FP1.6.6, BF3 Slope PWL Segment 1
1363 dscl_prog_data->easf_v_bf3_pwl_in_set2 =
1364 0x0BB7; // FP0.6.6, BF3 Input value PWL Segment 2 (0.03125 * 125^3)
1365 dscl_prog_data->easf_v_bf3_pwl_base_set2 = 20; // S0.6, BF3 Base PWL Segment 2
1366 dscl_prog_data->easf_v_bf3_pwl_slope_set2 =
1367 0x1356; // FP1.6.6, BF3 Slope PWL Segment 2
1368 dscl_prog_data->easf_v_bf3_pwl_in_set3 =
1369 0x0BF7; // FP0.6.6, BF3 Input value PWL Segment 3 (0.0625 * 125^3)
1370 dscl_prog_data->easf_v_bf3_pwl_base_set3 = 0; // S0.6, BF3 Base PWL Segment 3
1371 dscl_prog_data->easf_v_bf3_pwl_slope_set3 =
1372 0x136B; // FP1.6.6, BF3 Slope PWL Segment 3
1373 dscl_prog_data->easf_v_bf3_pwl_in_set4 =
1374 0x0C37; // FP0.6.6, BF3 Input value PWL Segment 4 (0.125 * 125^3)
1375 dscl_prog_data->easf_v_bf3_pwl_base_set4 = 0x4E; // S0.6, BF3 Base PWL Segment 4 = -50
1376 dscl_prog_data->easf_v_bf3_pwl_slope_set4 =
1377 0x1200; // FP1.6.6, BF3 Slope PWL Segment 4
1378 dscl_prog_data->easf_v_bf3_pwl_in_set5 =
1379 0x0CF7; // FP0.6.6, BF3 Input value PWL Segment 5 (1.0 * 125^3)
1380 dscl_prog_data->easf_v_bf3_pwl_base_set5 = 0x41; // S0.6, BF3 Base PWL Segment 5 = -63
1381 } else {
1382 dscl_prog_data->easf_v_bf2_flat1_gain = 13; // U1.3, BF2 Flat1 Gain control
1383 dscl_prog_data->easf_v_bf2_flat2_gain = 15; // U4.0, BF2 Flat2 Gain control
1384 dscl_prog_data->easf_v_bf2_roc_gain = 14; // U2.2, Rate Of Change control
1385
1386 dscl_prog_data->easf_v_bf1_pwl_in_seg0 = 0x440; // S0.10, BF1 PWL Segment 0 = -960
1387 dscl_prog_data->easf_v_bf1_pwl_base_seg0 = 0; // U0.6, BF1 Base PWL Segment 0
1388 dscl_prog_data->easf_v_bf1_pwl_slope_seg0 = 2; // S7.3, BF1 Slope PWL Segment 0
1389 dscl_prog_data->easf_v_bf1_pwl_in_seg1 = 0x7C4; // S0.10, BF1 PWL Segment 1 = -60
1390 dscl_prog_data->easf_v_bf1_pwl_base_seg1 = 12; // U0.6, BF1 Base PWL Segment 1
1391 dscl_prog_data->easf_v_bf1_pwl_slope_seg1 = 109; // S7.3, BF1 Slope PWL Segment 1
1392 dscl_prog_data->easf_v_bf1_pwl_in_seg2 = 0; // S0.10, BF1 PWL Segment 2
1393 dscl_prog_data->easf_v_bf1_pwl_base_seg2 = 63; // U0.6, BF1 Base PWL Segment 2
1394 dscl_prog_data->easf_v_bf1_pwl_slope_seg2 = 0; // S7.3, BF1 Slope PWL Segment 2
1395 dscl_prog_data->easf_v_bf1_pwl_in_seg3 = 48; // S0.10, BF1 PWL Segment 3
1396 dscl_prog_data->easf_v_bf1_pwl_base_seg3 = 63; // U0.6, BF1 Base PWL Segment 3
1397 dscl_prog_data->easf_v_bf1_pwl_slope_seg3 = 0x7ED; // S7.3, BF1 Slope PWL Segment 3 = -19
1398 dscl_prog_data->easf_v_bf1_pwl_in_seg4 = 96; // S0.10, BF1 PWL Segment 4
1399 dscl_prog_data->easf_v_bf1_pwl_base_seg4 = 56; // U0.6, BF1 Base PWL Segment 4
1400 dscl_prog_data->easf_v_bf1_pwl_slope_seg4 = 0x7F0; // S7.3, BF1 Slope PWL Segment 4 = -16
1401 dscl_prog_data->easf_v_bf1_pwl_in_seg5 = 144; // S0.10, BF1 PWL Segment 5
1402 dscl_prog_data->easf_v_bf1_pwl_base_seg5 = 50; // U0.6, BF1 Base PWL Segment 5
1403 dscl_prog_data->easf_v_bf1_pwl_slope_seg5 = 0x7B0; // S7.3, BF1 Slope PWL Segment 5 = -80
1404 dscl_prog_data->easf_v_bf1_pwl_in_seg6 = 192; // S0.10, BF1 PWL Segment 6
1405 dscl_prog_data->easf_v_bf1_pwl_base_seg6 = 20; // U0.6, BF1 Base PWL Segment 6
1406 dscl_prog_data->easf_v_bf1_pwl_slope_seg6 = 0x7CB; // S7.3, BF1 Slope PWL Segment 6 = -53
1407 dscl_prog_data->easf_v_bf1_pwl_in_seg7 = 240; // S0.10, BF1 PWL Segment 7
1408 dscl_prog_data->easf_v_bf1_pwl_base_seg7 = 0; // U0.6, BF1 Base PWL Segment 7
1409
1410 dscl_prog_data->easf_v_bf3_pwl_in_set0 = 0x000; // FP0.6.6, BF3 Input value PWL Segment 0
1411 dscl_prog_data->easf_v_bf3_pwl_base_set0 = 63; // S0.6, BF3 Base PWL Segment 0
1412 dscl_prog_data->easf_v_bf3_pwl_slope_set0 = 0x0000; // FP1.6.6, BF3 Slope PWL Segment 0
1413 dscl_prog_data->easf_v_bf3_pwl_in_set1 =
1414 0x06C0; // FP0.6.6, BF3 Input value PWL Segment 1 (0.0625)
1415 dscl_prog_data->easf_v_bf3_pwl_base_set1 = 63; // S0.6, BF3 Base PWL Segment 1
1416 dscl_prog_data->easf_v_bf3_pwl_slope_set1 = 0x1896; // FP1.6.6, BF3 Slope PWL Segment 1
1417 dscl_prog_data->easf_v_bf3_pwl_in_set2 =
1418 0x0700; // FP0.6.6, BF3 Input value PWL Segment 2 (0.125)
1419 dscl_prog_data->easf_v_bf3_pwl_base_set2 = 20; // S0.6, BF3 Base PWL Segment 2
1420 dscl_prog_data->easf_v_bf3_pwl_slope_set2 = 0x1810; // FP1.6.6, BF3 Slope PWL Segment 2
1421 dscl_prog_data->easf_v_bf3_pwl_in_set3 =
1422 0x0740; // FP0.6.6, BF3 Input value PWL Segment 3 (0.25)
1423 dscl_prog_data->easf_v_bf3_pwl_base_set3 = 0; // S0.6, BF3 Base PWL Segment 3
1424 dscl_prog_data->easf_v_bf3_pwl_slope_set3 =
1425 0x1878; // FP1.6.6, BF3 Slope PWL Segment 3
1426 dscl_prog_data->easf_v_bf3_pwl_in_set4 =
1427 0x0761; // FP0.6.6, BF3 Input value PWL Segment 4 (0.375)
1428 dscl_prog_data->easf_v_bf3_pwl_base_set4 = 0x44; // S0.6, BF3 Base PWL Segment 4 = -60
1429 dscl_prog_data->easf_v_bf3_pwl_slope_set4 = 0x1760; // FP1.6.6, BF3 Slope PWL Segment 4
1430 dscl_prog_data->easf_v_bf3_pwl_in_set5 =
1431 0x0780; // FP0.6.6, BF3 Input value PWL Segment 5 (0.5)
1432 dscl_prog_data->easf_v_bf3_pwl_base_set5 = 0x41; // S0.6, BF3 Base PWL Segment 5 = -63
1433 }
1434 } else
1435 dscl_prog_data->easf_v_en = false;
1436
1437 if (enable_easf_h) {
1438 dscl_prog_data->easf_h_en = true;
1439 dscl_prog_data->easf_h_ring = 0;
1440 dscl_prog_data->easf_h_sharp_factor = 0;
1441 dscl_prog_data->easf_h_bf1_en =
1442 1; // 1-bit, BF1 calculation enable, 0=disable, 1=enable
1443 dscl_prog_data->easf_h_bf2_mode =
1444 0xF; // 4-bit, BF2 calculation mode
1445 /* 2-bit, BF3 chroma mode correction calculation mode */
1446 dscl_prog_data->easf_h_bf3_mode = spl_get_h_bf3_mode(
1447 spl_scratch->scl_data.recip_ratios.horz);
1448 /* FP1.5.10; (2.0) Ring reducer gain for 4 or 6-tap mode [H_REDUCER_GAIN4] */
1449 dscl_prog_data->easf_h_ringest_eventap_reduceg1 =
1450 spl_get_reducer_gain4(spl_scratch->scl_data.taps.h_taps,
1451 spl_scratch->scl_data.recip_ratios.horz);
1452 /* FP1.5.10; (2.5) Ring reducer gain for 6-tap mode [V_REDUCER_GAIN6] */
1453 dscl_prog_data->easf_h_ringest_eventap_reduceg2 =
1454 spl_get_reducer_gain6(spl_scratch->scl_data.taps.h_taps,
1455 spl_scratch->scl_data.recip_ratios.horz);
1456 /* FP1.5.10; (-0.135742) Ring gain for 6-tap set to -139/1024 */
1457 dscl_prog_data->easf_h_ringest_eventap_gain1 =
1458 spl_get_gainRing4(spl_scratch->scl_data.taps.h_taps,
1459 spl_scratch->scl_data.recip_ratios.horz);
1460 /* FP1.5.10; (-0.024414) Ring gain for 6-tap set to -25/1024 */
1461 dscl_prog_data->easf_h_ringest_eventap_gain2 =
1462 spl_get_gainRing6(spl_scratch->scl_data.taps.h_taps,
1463 spl_scratch->scl_data.recip_ratios.horz);
1464 dscl_prog_data->easf_h_bf_maxa = 63; //Horz Max BF value A in U0.6 format.Selected if H_FCNTL==0
1465 dscl_prog_data->easf_h_bf_maxb = 63; //Horz Max BF value B in U0.6 format.Selected if H_FCNTL==1
1466 dscl_prog_data->easf_h_bf_mina = 0; //Horz Min BF value B in U0.6 format.Selected if H_FCNTL==0
1467 dscl_prog_data->easf_h_bf_minb = 0; //Horz Min BF value B in U0.6 format.Selected if H_FCNTL==1
1468 if (lls_pref == LLS_PREF_YES) {
1469 dscl_prog_data->easf_h_bf2_flat1_gain = 4; // U1.3, BF2 Flat1 Gain control
1470 dscl_prog_data->easf_h_bf2_flat2_gain = 8; // U4.0, BF2 Flat2 Gain control
1471 dscl_prog_data->easf_h_bf2_roc_gain = 4; // U2.2, Rate Of Change control
1472
1473 dscl_prog_data->easf_h_bf1_pwl_in_seg0 = 0x600; // S0.10, BF1 PWL Segment 0 = -512
1474 dscl_prog_data->easf_h_bf1_pwl_base_seg0 = 0; // U0.6, BF1 Base PWL Segment 0
1475 dscl_prog_data->easf_h_bf1_pwl_slope_seg0 = 3; // S7.3, BF1 Slope PWL Segment 0
1476 dscl_prog_data->easf_h_bf1_pwl_in_seg1 = 0x7EC; // S0.10, BF1 PWL Segment 1 = -20
1477 dscl_prog_data->easf_h_bf1_pwl_base_seg1 = 12; // U0.6, BF1 Base PWL Segment 1
1478 dscl_prog_data->easf_h_bf1_pwl_slope_seg1 = 326; // S7.3, BF1 Slope PWL Segment 1
1479 dscl_prog_data->easf_h_bf1_pwl_in_seg2 = 0; // S0.10, BF1 PWL Segment 2
1480 dscl_prog_data->easf_h_bf1_pwl_base_seg2 = 63; // U0.6, BF1 Base PWL Segment 2
1481 dscl_prog_data->easf_h_bf1_pwl_slope_seg2 = 0; // S7.3, BF1 Slope PWL Segment 2
1482 dscl_prog_data->easf_h_bf1_pwl_in_seg3 = 16; // S0.10, BF1 PWL Segment 3
1483 dscl_prog_data->easf_h_bf1_pwl_base_seg3 = 63; // U0.6, BF1 Base PWL Segment 3
1484 dscl_prog_data->easf_h_bf1_pwl_slope_seg3 = 0x7C8; // S7.3, BF1 Slope PWL Segment 3 = -56
1485 dscl_prog_data->easf_h_bf1_pwl_in_seg4 = 32; // S0.10, BF1 PWL Segment 4
1486 dscl_prog_data->easf_h_bf1_pwl_base_seg4 = 56; // U0.6, BF1 Base PWL Segment 4
1487 dscl_prog_data->easf_h_bf1_pwl_slope_seg4 = 0x7D0; // S7.3, BF1 Slope PWL Segment 4 = -48
1488 dscl_prog_data->easf_h_bf1_pwl_in_seg5 = 48; // S0.10, BF1 PWL Segment 5
1489 dscl_prog_data->easf_h_bf1_pwl_base_seg5 = 50; // U0.6, BF1 Base PWL Segment 5
1490 dscl_prog_data->easf_h_bf1_pwl_slope_seg5 = 0x710; // S7.3, BF1 Slope PWL Segment 5 = -240
1491 dscl_prog_data->easf_h_bf1_pwl_in_seg6 = 64; // S0.10, BF1 PWL Segment 6
1492 dscl_prog_data->easf_h_bf1_pwl_base_seg6 = 20; // U0.6, BF1 Base PWL Segment 6
1493 dscl_prog_data->easf_h_bf1_pwl_slope_seg6 = 0x760; // S7.3, BF1 Slope PWL Segment 6 = -160
1494 dscl_prog_data->easf_h_bf1_pwl_in_seg7 = 80; // S0.10, BF1 PWL Segment 7
1495 dscl_prog_data->easf_h_bf1_pwl_base_seg7 = 0; // U0.6, BF1 Base PWL Segment 7
1496
1497 dscl_prog_data->easf_h_bf3_pwl_in_set0 = 0x000; // FP0.6.6, BF3 Input value PWL Segment 0
1498 dscl_prog_data->easf_h_bf3_pwl_base_set0 = 63; // S0.6, BF3 Base PWL Segment 0
1499 dscl_prog_data->easf_h_bf3_pwl_slope_set0 = 0x12C5; // FP1.6.6, BF3 Slope PWL Segment 0
1500 dscl_prog_data->easf_h_bf3_pwl_in_set1 =
1501 0x0B37; // FP0.6.6, BF3 Input value PWL Segment 1 (0.0078125 * 125^3)
1502 dscl_prog_data->easf_h_bf3_pwl_base_set1 = 62; // S0.6, BF3 Base PWL Segment 1
1503 dscl_prog_data->easf_h_bf3_pwl_slope_set1 = 0x13B8; // FP1.6.6, BF3 Slope PWL Segment 1
1504 dscl_prog_data->easf_h_bf3_pwl_in_set2 =
1505 0x0BB7; // FP0.6.6, BF3 Input value PWL Segment 2 (0.03125 * 125^3)
1506 dscl_prog_data->easf_h_bf3_pwl_base_set2 = 20; // S0.6, BF3 Base PWL Segment 2
1507 dscl_prog_data->easf_h_bf3_pwl_slope_set2 = 0x1356; // FP1.6.6, BF3 Slope PWL Segment 2
1508 dscl_prog_data->easf_h_bf3_pwl_in_set3 =
1509 0x0BF7; // FP0.6.6, BF3 Input value PWL Segment 3 (0.0625 * 125^3)
1510 dscl_prog_data->easf_h_bf3_pwl_base_set3 = 0; // S0.6, BF3 Base PWL Segment 3
1511 dscl_prog_data->easf_h_bf3_pwl_slope_set3 = 0x136B; // FP1.6.6, BF3 Slope PWL Segment 3
1512 dscl_prog_data->easf_h_bf3_pwl_in_set4 =
1513 0x0C37; // FP0.6.6, BF3 Input value PWL Segment 4 (0.125 * 125^3)
1514 dscl_prog_data->easf_h_bf3_pwl_base_set4 = 0x4E; // S0.6, BF3 Base PWL Segment 4 = -50
1515 dscl_prog_data->easf_h_bf3_pwl_slope_set4 = 0x1200; // FP1.6.6, BF3 Slope PWL Segment 4
1516 dscl_prog_data->easf_h_bf3_pwl_in_set5 =
1517 0x0CF7; // FP0.6.6, BF3 Input value PWL Segment 5 (1.0 * 125^3)
1518 dscl_prog_data->easf_h_bf3_pwl_base_set5 = 0x41; // S0.6, BF3 Base PWL Segment 5 = -63
1519 } else {
1520 dscl_prog_data->easf_h_bf2_flat1_gain = 13; // U1.3, BF2 Flat1 Gain control
1521 dscl_prog_data->easf_h_bf2_flat2_gain = 15; // U4.0, BF2 Flat2 Gain control
1522 dscl_prog_data->easf_h_bf2_roc_gain = 14; // U2.2, Rate Of Change control
1523
1524 dscl_prog_data->easf_h_bf1_pwl_in_seg0 = 0x440; // S0.10, BF1 PWL Segment 0 = -960
1525 dscl_prog_data->easf_h_bf1_pwl_base_seg0 = 0; // U0.6, BF1 Base PWL Segment 0
1526 dscl_prog_data->easf_h_bf1_pwl_slope_seg0 = 2; // S7.3, BF1 Slope PWL Segment 0
1527 dscl_prog_data->easf_h_bf1_pwl_in_seg1 = 0x7C4; // S0.10, BF1 PWL Segment 1 = -60
1528 dscl_prog_data->easf_h_bf1_pwl_base_seg1 = 12; // U0.6, BF1 Base PWL Segment 1
1529 dscl_prog_data->easf_h_bf1_pwl_slope_seg1 = 109; // S7.3, BF1 Slope PWL Segment 1
1530 dscl_prog_data->easf_h_bf1_pwl_in_seg2 = 0; // S0.10, BF1 PWL Segment 2
1531 dscl_prog_data->easf_h_bf1_pwl_base_seg2 = 63; // U0.6, BF1 Base PWL Segment 2
1532 dscl_prog_data->easf_h_bf1_pwl_slope_seg2 = 0; // S7.3, BF1 Slope PWL Segment 2
1533 dscl_prog_data->easf_h_bf1_pwl_in_seg3 = 48; // S0.10, BF1 PWL Segment 3
1534 dscl_prog_data->easf_h_bf1_pwl_base_seg3 = 63; // U0.6, BF1 Base PWL Segment 3
1535 dscl_prog_data->easf_h_bf1_pwl_slope_seg3 = 0x7ED; // S7.3, BF1 Slope PWL Segment 3 = -19
1536 dscl_prog_data->easf_h_bf1_pwl_in_seg4 = 96; // S0.10, BF1 PWL Segment 4
1537 dscl_prog_data->easf_h_bf1_pwl_base_seg4 = 56; // U0.6, BF1 Base PWL Segment 4
1538 dscl_prog_data->easf_h_bf1_pwl_slope_seg4 = 0x7F0; // S7.3, BF1 Slope PWL Segment 4 = -16
1539 dscl_prog_data->easf_h_bf1_pwl_in_seg5 = 144; // S0.10, BF1 PWL Segment 5
1540 dscl_prog_data->easf_h_bf1_pwl_base_seg5 = 50; // U0.6, BF1 Base PWL Segment 5
1541 dscl_prog_data->easf_h_bf1_pwl_slope_seg5 = 0x7B0; // S7.3, BF1 Slope PWL Segment 5 = -80
1542 dscl_prog_data->easf_h_bf1_pwl_in_seg6 = 192; // S0.10, BF1 PWL Segment 6
1543 dscl_prog_data->easf_h_bf1_pwl_base_seg6 = 20; // U0.6, BF1 Base PWL Segment 6
1544 dscl_prog_data->easf_h_bf1_pwl_slope_seg6 = 0x7CB; // S7.3, BF1 Slope PWL Segment 6 = -53
1545 dscl_prog_data->easf_h_bf1_pwl_in_seg7 = 240; // S0.10, BF1 PWL Segment 7
1546 dscl_prog_data->easf_h_bf1_pwl_base_seg7 = 0; // U0.6, BF1 Base PWL Segment 7
1547
1548 dscl_prog_data->easf_h_bf3_pwl_in_set0 = 0x000; // FP0.6.6, BF3 Input value PWL Segment 0
1549 dscl_prog_data->easf_h_bf3_pwl_base_set0 = 63; // S0.6, BF3 Base PWL Segment 0
1550 dscl_prog_data->easf_h_bf3_pwl_slope_set0 = 0x0000; // FP1.6.6, BF3 Slope PWL Segment 0
1551 dscl_prog_data->easf_h_bf3_pwl_in_set1 =
1552 0x06C0; // FP0.6.6, BF3 Input value PWL Segment 1 (0.0625)
1553 dscl_prog_data->easf_h_bf3_pwl_base_set1 = 63; // S0.6, BF3 Base PWL Segment 1
1554 dscl_prog_data->easf_h_bf3_pwl_slope_set1 = 0x1896; // FP1.6.6, BF3 Slope PWL Segment 1
1555 dscl_prog_data->easf_h_bf3_pwl_in_set2 =
1556 0x0700; // FP0.6.6, BF3 Input value PWL Segment 2 (0.125)
1557 dscl_prog_data->easf_h_bf3_pwl_base_set2 = 20; // S0.6, BF3 Base PWL Segment 2
1558 dscl_prog_data->easf_h_bf3_pwl_slope_set2 = 0x1810; // FP1.6.6, BF3 Slope PWL Segment 2
1559 dscl_prog_data->easf_h_bf3_pwl_in_set3 =
1560 0x0740; // FP0.6.6, BF3 Input value PWL Segment 3 (0.25)
1561 dscl_prog_data->easf_h_bf3_pwl_base_set3 = 0; // S0.6, BF3 Base PWL Segment 3
1562 dscl_prog_data->easf_h_bf3_pwl_slope_set3 = 0x1878; // FP1.6.6, BF3 Slope PWL Segment 3
1563 dscl_prog_data->easf_h_bf3_pwl_in_set4 =
1564 0x0761; // FP0.6.6, BF3 Input value PWL Segment 4 (0.375)
1565 dscl_prog_data->easf_h_bf3_pwl_base_set4 = 0x44; // S0.6, BF3 Base PWL Segment 4 = -60
1566 dscl_prog_data->easf_h_bf3_pwl_slope_set4 = 0x1760; // FP1.6.6, BF3 Slope PWL Segment 4
1567 dscl_prog_data->easf_h_bf3_pwl_in_set5 =
1568 0x0780; // FP0.6.6, BF3 Input value PWL Segment 5 (0.5)
1569 dscl_prog_data->easf_h_bf3_pwl_base_set5 = 0x41; // S0.6, BF3 Base PWL Segment 5 = -63
1570 } // if (lls_pref == LLS_PREF_YES)
1571 } else
1572 dscl_prog_data->easf_h_en = false;
1573
1574 if (lls_pref == LLS_PREF_YES) {
1575 dscl_prog_data->easf_ltonl_en = 1; // Linear input
1576 if ((setup == HDR_L) && (spl_is_rgb8(format))) {
1577 /* Calculate C0-C3 coefficients based on HDR multiplier */
1578 spl_calculate_c0_c3_hdr(dscl_prog_data, sdr_white_level_nits);
1579 } else { // HDR_L ( DWM ) and SDR_L
1580 dscl_prog_data->easf_matrix_c0 =
1581 0x4EF7; // fp1.5.10, C0 coefficient (LN_rec709: 0.2126 * (2^14)/125 = 27.86590720)
1582 dscl_prog_data->easf_matrix_c1 =
1583 0x55DC; // fp1.5.10, C1 coefficient (LN_rec709: 0.7152 * (2^14)/125 = 93.74269440)
1584 dscl_prog_data->easf_matrix_c2 =
1585 0x48BB; // fp1.5.10, C2 coefficient (LN_rec709: 0.0722 * (2^14)/125 = 9.46339840)
1586 dscl_prog_data->easf_matrix_c3 =
1587 0x0; // fp1.5.10, C3 coefficient
1588 }
1589 } else {
1590 dscl_prog_data->easf_ltonl_en = 0; // Non-Linear input
1591 dscl_prog_data->easf_matrix_c0 =
1592 0x3434; // fp1.5.10, C0 coefficient (LN_BT2020: 0.262695312500000)
1593 dscl_prog_data->easf_matrix_c1 =
1594 0x396D; // fp1.5.10, C1 coefficient (LN_BT2020: 0.678222656250000)
1595 dscl_prog_data->easf_matrix_c2 =
1596 0x2B97; // fp1.5.10, C2 coefficient (LN_BT2020: 0.059295654296875)
1597 dscl_prog_data->easf_matrix_c3 =
1598 0x0; // fp1.5.10, C3 coefficient
1599 }
1600
1601 if (spl_is_subsampled_format(format)) { /* TODO: 0 = RGB, 1 = YUV */
1602 dscl_prog_data->easf_matrix_mode = 1;
1603 /*
1604 * 2-bit, BF3 chroma mode correction calculation mode
1605 * Needs to be disabled for YUV420 mode
1606 * Override lookup value
1607 */
1608 dscl_prog_data->easf_v_bf3_mode = 0;
1609 dscl_prog_data->easf_h_bf3_mode = 0;
1610 } else
1611 dscl_prog_data->easf_matrix_mode = 0;
1612
1613 }
1614
1615 /*Set isharp noise detection */
spl_set_isharp_noise_det_mode(struct dscl_prog_data * dscl_prog_data,const struct spl_scaler_data * data)1616 static void spl_set_isharp_noise_det_mode(struct dscl_prog_data *dscl_prog_data,
1617 const struct spl_scaler_data *data)
1618 {
1619 // ISHARP_NOISEDET_MODE
1620 // 0: 3x5 as VxH
1621 // 1: 4x5 as VxH
1622 // 2:
1623 // 3: 5x5 as VxH
1624 if (data->taps.v_taps == 6)
1625 dscl_prog_data->isharp_noise_det.mode = 3;
1626 else if (data->taps.v_taps == 4)
1627 dscl_prog_data->isharp_noise_det.mode = 1;
1628 else if (data->taps.v_taps == 3)
1629 dscl_prog_data->isharp_noise_det.mode = 0;
1630 };
1631 /* Set Sharpener data */
spl_set_isharp_data(struct dscl_prog_data * dscl_prog_data,struct adaptive_sharpness adp_sharpness,bool enable_isharp,enum linear_light_scaling lls_pref,enum spl_pixel_format format,const struct spl_scaler_data * data,struct spl_fixed31_32 ratio,enum system_setup setup,enum scale_to_sharpness_policy scale_to_sharpness_policy)1632 static void spl_set_isharp_data(struct dscl_prog_data *dscl_prog_data,
1633 struct adaptive_sharpness adp_sharpness, bool enable_isharp,
1634 enum linear_light_scaling lls_pref, enum spl_pixel_format format,
1635 const struct spl_scaler_data *data, struct spl_fixed31_32 ratio,
1636 enum system_setup setup, enum scale_to_sharpness_policy scale_to_sharpness_policy)
1637 {
1638 /* Turn off sharpener if not required */
1639 if (!enable_isharp) {
1640 dscl_prog_data->isharp_en = 0;
1641 return;
1642 }
1643
1644 spl_build_isharp_1dlut_from_reference_curve(ratio, setup, adp_sharpness,
1645 scale_to_sharpness_policy);
1646 memcpy(dscl_prog_data->isharp_delta, spl_get_pregen_filter_isharp_1D_lut(setup),
1647 sizeof(uint32_t) * ISHARP_LUT_TABLE_SIZE);
1648 dscl_prog_data->sharpness_level = adp_sharpness.sharpness_level;
1649
1650 dscl_prog_data->isharp_en = 1; // ISHARP_EN
1651 // Set ISHARP_NOISEDET_MODE if htaps = 6-tap
1652 if (data->taps.h_taps == 6) {
1653 dscl_prog_data->isharp_noise_det.enable = 1; /* ISHARP_NOISEDET_EN */
1654 spl_set_isharp_noise_det_mode(dscl_prog_data, data); /* ISHARP_NOISEDET_MODE */
1655 } else
1656 dscl_prog_data->isharp_noise_det.enable = 0; // ISHARP_NOISEDET_EN
1657 // Program noise detection threshold
1658 dscl_prog_data->isharp_noise_det.uthreshold = 24; // ISHARP_NOISEDET_UTHRE
1659 dscl_prog_data->isharp_noise_det.dthreshold = 4; // ISHARP_NOISEDET_DTHRE
1660 // Program noise detection gain
1661 dscl_prog_data->isharp_noise_det.pwl_start_in = 3; // ISHARP_NOISEDET_PWL_START_IN
1662 dscl_prog_data->isharp_noise_det.pwl_end_in = 13; // ISHARP_NOISEDET_PWL_END_IN
1663 dscl_prog_data->isharp_noise_det.pwl_slope = 1623; // ISHARP_NOISEDET_PWL_SLOPE
1664
1665 if (lls_pref == LLS_PREF_NO) /* ISHARP_FMT_MODE */
1666 dscl_prog_data->isharp_fmt.mode = 1;
1667 else
1668 dscl_prog_data->isharp_fmt.mode = 0;
1669
1670 dscl_prog_data->isharp_fmt.norm = 0x3C00; // ISHARP_FMT_NORM
1671 dscl_prog_data->isharp_lba.mode = 0; // ISHARP_LBA_MODE
1672
1673 if (setup == SDR_L) {
1674 // ISHARP_LBA_PWL_SEG0: ISHARP Local Brightness Adjustment PWL Segment 0
1675 dscl_prog_data->isharp_lba.in_seg[0] = 0; // ISHARP LBA PWL for Seg 0. INPUT value in U0.10 format
1676 dscl_prog_data->isharp_lba.base_seg[0] = 0; // ISHARP LBA PWL for Seg 0. BASE value in U0.6 format
1677 dscl_prog_data->isharp_lba.slope_seg[0] = 62; // ISHARP LBA for Seg 0. SLOPE value in S5.3 format
1678 // ISHARP_LBA_PWL_SEG1: ISHARP LBA PWL Segment 1
1679 dscl_prog_data->isharp_lba.in_seg[1] = 130; // ISHARP LBA PWL for Seg 1. INPUT value in U0.10 format
1680 dscl_prog_data->isharp_lba.base_seg[1] = 63; // ISHARP LBA PWL for Seg 1. BASE value in U0.6 format
1681 dscl_prog_data->isharp_lba.slope_seg[1] = 0; // ISHARP LBA for Seg 1. SLOPE value in S5.3 format
1682 // ISHARP_LBA_PWL_SEG2: ISHARP LBA PWL Segment 2
1683 dscl_prog_data->isharp_lba.in_seg[2] = 450; // ISHARP LBA PWL for Seg 2. INPUT value in U0.10 format
1684 dscl_prog_data->isharp_lba.base_seg[2] = 63; // ISHARP LBA PWL for Seg 2. BASE value in U0.6 format
1685 dscl_prog_data->isharp_lba.slope_seg[2] = 0x18D; // ISHARP LBA for Seg 2. SLOPE value in S5.3 format = -115
1686 // ISHARP_LBA_PWL_SEG3: ISHARP LBA PWL Segment 3
1687 dscl_prog_data->isharp_lba.in_seg[3] = 520; // ISHARP LBA PWL for Seg 3.INPUT value in U0.10 format
1688 dscl_prog_data->isharp_lba.base_seg[3] = 0; // ISHARP LBA PWL for Seg 3. BASE value in U0.6 format
1689 dscl_prog_data->isharp_lba.slope_seg[3] = 0; // ISHARP LBA for Seg 3. SLOPE value in S5.3 format
1690 // ISHARP_LBA_PWL_SEG4: ISHARP LBA PWL Segment 4
1691 dscl_prog_data->isharp_lba.in_seg[4] = 520; // ISHARP LBA PWL for Seg 4.INPUT value in U0.10 format
1692 dscl_prog_data->isharp_lba.base_seg[4] = 0; // ISHARP LBA PWL for Seg 4. BASE value in U0.6 format
1693 dscl_prog_data->isharp_lba.slope_seg[4] = 0; // ISHARP LBA for Seg 4. SLOPE value in S5.3 format
1694 // ISHARP_LBA_PWL_SEG5: ISHARP LBA PWL Segment 5
1695 dscl_prog_data->isharp_lba.in_seg[5] = 520; // ISHARP LBA PWL for Seg 5.INPUT value in U0.10 format
1696 dscl_prog_data->isharp_lba.base_seg[5] = 0; // ISHARP LBA PWL for Seg 5. BASE value in U0.6 format
1697 } else if (setup == HDR_L) {
1698 // ISHARP_LBA_PWL_SEG0: ISHARP Local Brightness Adjustment PWL Segment 0
1699 dscl_prog_data->isharp_lba.in_seg[0] = 0; // ISHARP LBA PWL for Seg 0. INPUT value in U0.10 format
1700 dscl_prog_data->isharp_lba.base_seg[0] = 0; // ISHARP LBA PWL for Seg 0. BASE value in U0.6 format
1701 dscl_prog_data->isharp_lba.slope_seg[0] = 32; // ISHARP LBA for Seg 0. SLOPE value in S5.3 format
1702 // ISHARP_LBA_PWL_SEG1: ISHARP LBA PWL Segment 1
1703 dscl_prog_data->isharp_lba.in_seg[1] = 254; // ISHARP LBA PWL for Seg 1. INPUT value in U0.10 format
1704 dscl_prog_data->isharp_lba.base_seg[1] = 63; // ISHARP LBA PWL for Seg 1. BASE value in U0.6 format
1705 dscl_prog_data->isharp_lba.slope_seg[1] = 0; // ISHARP LBA for Seg 1. SLOPE value in S5.3 format
1706 // ISHARP_LBA_PWL_SEG2: ISHARP LBA PWL Segment 2
1707 dscl_prog_data->isharp_lba.in_seg[2] = 559; // ISHARP LBA PWL for Seg 2. INPUT value in U0.10 format
1708 dscl_prog_data->isharp_lba.base_seg[2] = 63; // ISHARP LBA PWL for Seg 2. BASE value in U0.6 format
1709 dscl_prog_data->isharp_lba.slope_seg[2] = 0x10C; // ISHARP LBA for Seg 2. SLOPE value in S5.3 format = -244
1710 // ISHARP_LBA_PWL_SEG3: ISHARP LBA PWL Segment 3
1711 dscl_prog_data->isharp_lba.in_seg[3] = 592; // ISHARP LBA PWL for Seg 3.INPUT value in U0.10 format
1712 dscl_prog_data->isharp_lba.base_seg[3] = 0; // ISHARP LBA PWL for Seg 3. BASE value in U0.6 format
1713 dscl_prog_data->isharp_lba.slope_seg[3] = 0; // ISHARP LBA for Seg 3. SLOPE value in S5.3 format
1714 // ISHARP_LBA_PWL_SEG4: ISHARP LBA PWL Segment 4
1715 dscl_prog_data->isharp_lba.in_seg[4] = 1023; // ISHARP LBA PWL for Seg 4.INPUT value in U0.10 format
1716 dscl_prog_data->isharp_lba.base_seg[4] = 0; // ISHARP LBA PWL for Seg 4. BASE value in U0.6 format
1717 dscl_prog_data->isharp_lba.slope_seg[4] = 0; // ISHARP LBA for Seg 4. SLOPE value in S5.3 format
1718 // ISHARP_LBA_PWL_SEG5: ISHARP LBA PWL Segment 5
1719 dscl_prog_data->isharp_lba.in_seg[5] = 1023; // ISHARP LBA PWL for Seg 5.INPUT value in U0.10 format
1720 dscl_prog_data->isharp_lba.base_seg[5] = 0; // ISHARP LBA PWL for Seg 5. BASE value in U0.6 format
1721 } else {
1722 // ISHARP_LBA_PWL_SEG0: ISHARP Local Brightness Adjustment PWL Segment 0
1723 dscl_prog_data->isharp_lba.in_seg[0] = 0; // ISHARP LBA PWL for Seg 0. INPUT value in U0.10 format
1724 dscl_prog_data->isharp_lba.base_seg[0] = 0; // ISHARP LBA PWL for Seg 0. BASE value in U0.6 format
1725 dscl_prog_data->isharp_lba.slope_seg[0] = 40; // ISHARP LBA for Seg 0. SLOPE value in S5.3 format
1726 // ISHARP_LBA_PWL_SEG1: ISHARP LBA PWL Segment 1
1727 dscl_prog_data->isharp_lba.in_seg[1] = 204; // ISHARP LBA PWL for Seg 1. INPUT value in U0.10 format
1728 dscl_prog_data->isharp_lba.base_seg[1] = 63; // ISHARP LBA PWL for Seg 1. BASE value in U0.6 format
1729 dscl_prog_data->isharp_lba.slope_seg[1] = 0; // ISHARP LBA for Seg 1. SLOPE value in S5.3 format
1730 // ISHARP_LBA_PWL_SEG2: ISHARP LBA PWL Segment 2
1731 dscl_prog_data->isharp_lba.in_seg[2] = 818; // ISHARP LBA PWL for Seg 2. INPUT value in U0.10 format
1732 dscl_prog_data->isharp_lba.base_seg[2] = 63; // ISHARP LBA PWL for Seg 2. BASE value in U0.6 format
1733 dscl_prog_data->isharp_lba.slope_seg[2] = 0x1D9; // ISHARP LBA for Seg 2. SLOPE value in S5.3 format = -39
1734 // ISHARP_LBA_PWL_SEG3: ISHARP LBA PWL Segment 3
1735 dscl_prog_data->isharp_lba.in_seg[3] = 1023; // ISHARP LBA PWL for Seg 3.INPUT value in U0.10 format
1736 dscl_prog_data->isharp_lba.base_seg[3] = 0; // ISHARP LBA PWL for Seg 3. BASE value in U0.6 format
1737 dscl_prog_data->isharp_lba.slope_seg[3] = 0; // ISHARP LBA for Seg 3. SLOPE value in S5.3 format
1738 // ISHARP_LBA_PWL_SEG4: ISHARP LBA PWL Segment 4
1739 dscl_prog_data->isharp_lba.in_seg[4] = 1023; // ISHARP LBA PWL for Seg 4.INPUT value in U0.10 format
1740 dscl_prog_data->isharp_lba.base_seg[4] = 0; // ISHARP LBA PWL for Seg 4. BASE value in U0.6 format
1741 dscl_prog_data->isharp_lba.slope_seg[4] = 0; // ISHARP LBA for Seg 4. SLOPE value in S5.3 format
1742 // ISHARP_LBA_PWL_SEG5: ISHARP LBA PWL Segment 5
1743 dscl_prog_data->isharp_lba.in_seg[5] = 1023; // ISHARP LBA PWL for Seg 5.INPUT value in U0.10 format
1744 dscl_prog_data->isharp_lba.base_seg[5] = 0; // ISHARP LBA PWL for Seg 5. BASE value in U0.6 format
1745 }
1746
1747 // Program the nldelta soft clip values
1748 if (lls_pref == LLS_PREF_YES) {
1749 dscl_prog_data->isharp_nldelta_sclip.enable_p = 0; /* ISHARP_NLDELTA_SCLIP_EN_P */
1750 dscl_prog_data->isharp_nldelta_sclip.pivot_p = 0; /* ISHARP_NLDELTA_SCLIP_PIVOT_P */
1751 dscl_prog_data->isharp_nldelta_sclip.slope_p = 0; /* ISHARP_NLDELTA_SCLIP_SLOPE_P */
1752 dscl_prog_data->isharp_nldelta_sclip.enable_n = 1; /* ISHARP_NLDELTA_SCLIP_EN_N */
1753 dscl_prog_data->isharp_nldelta_sclip.pivot_n = 71; /* ISHARP_NLDELTA_SCLIP_PIVOT_N */
1754 dscl_prog_data->isharp_nldelta_sclip.slope_n = 16; /* ISHARP_NLDELTA_SCLIP_SLOPE_N */
1755 } else {
1756 dscl_prog_data->isharp_nldelta_sclip.enable_p = 1; /* ISHARP_NLDELTA_SCLIP_EN_P */
1757 dscl_prog_data->isharp_nldelta_sclip.pivot_p = 70; /* ISHARP_NLDELTA_SCLIP_PIVOT_P */
1758 dscl_prog_data->isharp_nldelta_sclip.slope_p = 24; /* ISHARP_NLDELTA_SCLIP_SLOPE_P */
1759 dscl_prog_data->isharp_nldelta_sclip.enable_n = 1; /* ISHARP_NLDELTA_SCLIP_EN_N */
1760 dscl_prog_data->isharp_nldelta_sclip.pivot_n = 70; /* ISHARP_NLDELTA_SCLIP_PIVOT_N */
1761 dscl_prog_data->isharp_nldelta_sclip.slope_n = 24; /* ISHARP_NLDELTA_SCLIP_SLOPE_N */
1762 }
1763
1764 // Set the values as per lookup table
1765 spl_set_blur_scale_data(dscl_prog_data, data);
1766 }
1767
1768 /* Calculate recout, scaling ratio, and viewport, then get optimal number of taps */
spl_calculate_number_of_taps(struct spl_in * spl_in,struct spl_scratch * spl_scratch,struct spl_out * spl_out,bool * enable_easf_v,bool * enable_easf_h,bool * enable_isharp)1769 static bool spl_calculate_number_of_taps(struct spl_in *spl_in, struct spl_scratch *spl_scratch, struct spl_out *spl_out,
1770 bool *enable_easf_v, bool *enable_easf_h, bool *enable_isharp)
1771 {
1772 bool res = false;
1773
1774 memset(spl_scratch, 0, sizeof(struct spl_scratch));
1775 spl_scratch->scl_data.h_active = spl_in->h_active;
1776 spl_scratch->scl_data.v_active = spl_in->v_active;
1777
1778 // All SPL calls
1779 /* recout calculation */
1780 /* depends on h_active */
1781 spl_calculate_recout(spl_in, spl_scratch, spl_out);
1782 /* depends on pixel format */
1783 spl_calculate_scaling_ratios(spl_in, spl_scratch, spl_out);
1784 /* depends on scaling ratios and recout, does not calculate offset yet */
1785 spl_calculate_viewport_size(spl_in, spl_scratch);
1786
1787 res = spl_get_optimal_number_of_taps(
1788 spl_in->basic_out.max_downscale_src_width, spl_in,
1789 spl_scratch, &spl_in->scaling_quality, enable_easf_v,
1790 enable_easf_h, enable_isharp);
1791 return res;
1792 }
1793
1794 /* Calculate scaler parameters */
spl_calculate_scaler_params(struct spl_in * spl_in,struct spl_out * spl_out)1795 bool spl_calculate_scaler_params(struct spl_in *spl_in, struct spl_out *spl_out)
1796 {
1797 bool res = false;
1798 bool enable_easf_v = false;
1799 bool enable_easf_h = false;
1800 int vratio = 0;
1801 int hratio = 0;
1802 struct spl_scratch spl_scratch;
1803 struct spl_fixed31_32 isharp_scale_ratio;
1804 enum system_setup setup;
1805 bool enable_isharp = false;
1806 const struct spl_scaler_data *data = &spl_scratch.scl_data;
1807
1808 res = spl_calculate_number_of_taps(spl_in, &spl_scratch, spl_out,
1809 &enable_easf_v, &enable_easf_h, &enable_isharp);
1810
1811 /*
1812 * Depends on recout, scaling ratios, h_active and taps
1813 * May need to re-check lb size after this in some obscure scenario
1814 */
1815 if (res)
1816 spl_calculate_inits_and_viewports(spl_in, &spl_scratch);
1817 // Handle 3d recout
1818 spl_handle_3d_recout(spl_in, &spl_scratch.scl_data.recout);
1819 // Clamp
1820 spl_clamp_viewport(&spl_scratch.scl_data.viewport);
1821
1822 // Save all calculated parameters in dscl_prog_data structure to program hw registers
1823 spl_set_dscl_prog_data(spl_in, &spl_scratch, spl_out, enable_easf_v, enable_easf_h, enable_isharp);
1824
1825 if (!res)
1826 return res;
1827
1828 if (spl_in->lls_pref == LLS_PREF_YES) {
1829 if (spl_in->is_hdr_on)
1830 setup = HDR_L;
1831 else
1832 setup = SDR_L;
1833 } else {
1834 if (spl_in->is_hdr_on)
1835 setup = HDR_NL;
1836 else
1837 setup = SDR_NL;
1838 }
1839
1840 // Set EASF
1841 spl_set_easf_data(&spl_scratch, spl_out, enable_easf_v, enable_easf_h, spl_in->lls_pref,
1842 spl_in->basic_in.format, setup, spl_in->sdr_white_level_nits);
1843
1844 // Set iSHARP
1845 vratio = spl_fixpt_ceil(spl_scratch.scl_data.ratios.vert);
1846 hratio = spl_fixpt_ceil(spl_scratch.scl_data.ratios.horz);
1847 if (vratio <= hratio)
1848 isharp_scale_ratio = spl_scratch.scl_data.recip_ratios.vert;
1849 else
1850 isharp_scale_ratio = spl_scratch.scl_data.recip_ratios.horz;
1851
1852 spl_set_isharp_data(spl_out->dscl_prog_data, spl_in->adaptive_sharpness, enable_isharp,
1853 spl_in->lls_pref, spl_in->basic_in.format, data, isharp_scale_ratio, setup,
1854 spl_in->debug.scale_to_sharpness_policy);
1855
1856 return res;
1857 }
1858
1859 /* External interface to get number of taps only */
spl_get_number_of_taps(struct spl_in * spl_in,struct spl_out * spl_out)1860 bool spl_get_number_of_taps(struct spl_in *spl_in, struct spl_out *spl_out)
1861 {
1862 bool res = false;
1863 bool enable_easf_v = false;
1864 bool enable_easf_h = false;
1865 bool enable_isharp = false;
1866 struct spl_scratch spl_scratch;
1867 struct dscl_prog_data *dscl_prog_data = spl_out->dscl_prog_data;
1868 const struct spl_scaler_data *data = &spl_scratch.scl_data;
1869
1870 res = spl_calculate_number_of_taps(spl_in, &spl_scratch, spl_out,
1871 &enable_easf_v, &enable_easf_h, &enable_isharp);
1872 spl_set_taps_data(dscl_prog_data, data);
1873 return res;
1874 }
1875