1 /*
2 * Copyright 2016 Advanced Micro Devices, Inc.
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 *
22 * Authors: AMD
23 *
24 */
25 #include "dc.h"
26 #include "reg_helper.h"
27 #include "dcn10_dpp.h"
28
29 #include "dcn10_cm_common.h"
30 #include "custom_float.h"
31
32 #define REG(reg) reg
33
34 #define CTX \
35 ctx
36
37 #undef FN
38 #define FN(reg_name, field_name) \
39 reg->shifts.field_name, reg->masks.field_name
40
cm_helper_program_color_matrices(struct dc_context * ctx,const uint16_t * regval,const struct color_matrices_reg * reg)41 void cm_helper_program_color_matrices(
42 struct dc_context *ctx,
43 const uint16_t *regval,
44 const struct color_matrices_reg *reg)
45 {
46 uint32_t cur_csc_reg;
47 unsigned int i = 0;
48
49 for (cur_csc_reg = reg->csc_c11_c12;
50 cur_csc_reg <= reg->csc_c33_c34;
51 cur_csc_reg++) {
52
53 const uint16_t *regval0 = &(regval[2 * i]);
54 const uint16_t *regval1 = &(regval[(2 * i) + 1]);
55
56 REG_SET_2(cur_csc_reg, 0,
57 csc_c11, *regval0,
58 csc_c12, *regval1);
59
60 i++;
61 }
62
63 }
64
cm_helper_program_xfer_func(struct dc_context * ctx,const struct pwl_params * params,const struct xfer_func_reg * reg)65 void cm_helper_program_xfer_func(
66 struct dc_context *ctx,
67 const struct pwl_params *params,
68 const struct xfer_func_reg *reg)
69 {
70 uint32_t reg_region_cur;
71 unsigned int i = 0;
72
73 REG_SET_2(reg->start_cntl_b, 0,
74 exp_region_start, params->arr_points[0].custom_float_x,
75 exp_resion_start_segment, 0);
76 REG_SET_2(reg->start_cntl_g, 0,
77 exp_region_start, params->arr_points[0].custom_float_x,
78 exp_resion_start_segment, 0);
79 REG_SET_2(reg->start_cntl_r, 0,
80 exp_region_start, params->arr_points[0].custom_float_x,
81 exp_resion_start_segment, 0);
82
83 REG_SET(reg->start_slope_cntl_b, 0,
84 field_region_linear_slope, params->arr_points[0].custom_float_slope);
85 REG_SET(reg->start_slope_cntl_g, 0,
86 field_region_linear_slope, params->arr_points[0].custom_float_slope);
87 REG_SET(reg->start_slope_cntl_r, 0,
88 field_region_linear_slope, params->arr_points[0].custom_float_slope);
89
90 REG_SET(reg->start_end_cntl1_b, 0,
91 field_region_end, params->arr_points[1].custom_float_x);
92 REG_SET_2(reg->start_end_cntl2_b, 0,
93 field_region_end_slope, params->arr_points[1].custom_float_slope,
94 field_region_end_base, params->arr_points[1].custom_float_y);
95
96 REG_SET(reg->start_end_cntl1_g, 0,
97 field_region_end, params->arr_points[1].custom_float_x);
98 REG_SET_2(reg->start_end_cntl2_g, 0,
99 field_region_end_slope, params->arr_points[1].custom_float_slope,
100 field_region_end_base, params->arr_points[1].custom_float_y);
101
102 REG_SET(reg->start_end_cntl1_r, 0,
103 field_region_end, params->arr_points[1].custom_float_x);
104 REG_SET_2(reg->start_end_cntl2_r, 0,
105 field_region_end_slope, params->arr_points[1].custom_float_slope,
106 field_region_end_base, params->arr_points[1].custom_float_y);
107
108 for (reg_region_cur = reg->region_start;
109 reg_region_cur <= reg->region_end;
110 reg_region_cur++) {
111
112 const struct gamma_curve *curve0 = &(params->arr_curve_points[2 * i]);
113 const struct gamma_curve *curve1 = &(params->arr_curve_points[(2 * i) + 1]);
114
115 REG_SET_4(reg_region_cur, 0,
116 exp_region0_lut_offset, curve0->offset,
117 exp_region0_num_segments, curve0->segments_num,
118 exp_region1_lut_offset, curve1->offset,
119 exp_region1_num_segments, curve1->segments_num);
120
121 i++;
122 }
123
124 }
125
126
127
cm_helper_convert_to_custom_float(struct pwl_result_data * rgb_resulted,struct curve_points * arr_points,uint32_t hw_points_num,bool fixpoint)128 bool cm_helper_convert_to_custom_float(
129 struct pwl_result_data *rgb_resulted,
130 struct curve_points *arr_points,
131 uint32_t hw_points_num,
132 bool fixpoint)
133 {
134 struct custom_float_format fmt;
135
136 struct pwl_result_data *rgb = rgb_resulted;
137
138 uint32_t i = 0;
139
140 fmt.exponenta_bits = 6;
141 fmt.mantissa_bits = 12;
142 fmt.sign = false;
143
144 if (!convert_to_custom_float_format(arr_points[0].x, &fmt,
145 &arr_points[0].custom_float_x)) {
146 BREAK_TO_DEBUGGER();
147 return false;
148 }
149
150 if (!convert_to_custom_float_format(arr_points[0].offset, &fmt,
151 &arr_points[0].custom_float_offset)) {
152 BREAK_TO_DEBUGGER();
153 return false;
154 }
155
156 if (!convert_to_custom_float_format(arr_points[0].slope, &fmt,
157 &arr_points[0].custom_float_slope)) {
158 BREAK_TO_DEBUGGER();
159 return false;
160 }
161
162 fmt.mantissa_bits = 10;
163 fmt.sign = false;
164
165 if (!convert_to_custom_float_format(arr_points[1].x, &fmt,
166 &arr_points[1].custom_float_x)) {
167 BREAK_TO_DEBUGGER();
168 return false;
169 }
170
171 if (fixpoint == true)
172 arr_points[1].custom_float_y = dc_fixpt_clamp_u0d14(arr_points[1].y);
173 else if (!convert_to_custom_float_format(arr_points[1].y, &fmt,
174 &arr_points[1].custom_float_y)) {
175 BREAK_TO_DEBUGGER();
176 return false;
177 }
178
179 if (!convert_to_custom_float_format(arr_points[1].slope, &fmt,
180 &arr_points[1].custom_float_slope)) {
181 BREAK_TO_DEBUGGER();
182 return false;
183 }
184
185 if (hw_points_num == 0 || rgb_resulted == NULL || fixpoint == true)
186 return true;
187
188 fmt.mantissa_bits = 12;
189 fmt.sign = true;
190
191 while (i != hw_points_num) {
192 if (!convert_to_custom_float_format(rgb->red, &fmt,
193 &rgb->red_reg)) {
194 BREAK_TO_DEBUGGER();
195 return false;
196 }
197
198 if (!convert_to_custom_float_format(rgb->green, &fmt,
199 &rgb->green_reg)) {
200 BREAK_TO_DEBUGGER();
201 return false;
202 }
203
204 if (!convert_to_custom_float_format(rgb->blue, &fmt,
205 &rgb->blue_reg)) {
206 BREAK_TO_DEBUGGER();
207 return false;
208 }
209
210 if (!convert_to_custom_float_format(rgb->delta_red, &fmt,
211 &rgb->delta_red_reg)) {
212 BREAK_TO_DEBUGGER();
213 return false;
214 }
215
216 if (!convert_to_custom_float_format(rgb->delta_green, &fmt,
217 &rgb->delta_green_reg)) {
218 BREAK_TO_DEBUGGER();
219 return false;
220 }
221
222 if (!convert_to_custom_float_format(rgb->delta_blue, &fmt,
223 &rgb->delta_blue_reg)) {
224 BREAK_TO_DEBUGGER();
225 return false;
226 }
227
228 ++rgb;
229 ++i;
230 }
231
232 return true;
233 }
234
235 /* driver uses 32 regions or less, but DCN HW has 34, extra 2 are set to 0 */
236 #define MAX_REGIONS_NUMBER 34
237 #define MAX_LOW_POINT 25
238 #define NUMBER_REGIONS 32
239 #define NUMBER_SW_SEGMENTS 16
240
cm_helper_translate_curve_to_hw_format(const struct dc_transfer_func * output_tf,struct pwl_params * lut_params,bool fixpoint)241 bool cm_helper_translate_curve_to_hw_format(
242 const struct dc_transfer_func *output_tf,
243 struct pwl_params *lut_params, bool fixpoint)
244 {
245 struct curve_points *arr_points;
246 struct pwl_result_data *rgb_resulted;
247 struct pwl_result_data *rgb;
248 struct pwl_result_data *rgb_plus_1;
249 struct fixed31_32 y_r;
250 struct fixed31_32 y_g;
251 struct fixed31_32 y_b;
252 struct fixed31_32 y1_min;
253 struct fixed31_32 y3_max;
254
255 int32_t region_start, region_end;
256 int32_t i;
257 uint32_t j, k, seg_distr[MAX_REGIONS_NUMBER], increment, start_index, hw_points;
258
259 if (output_tf == NULL || lut_params == NULL || output_tf->type == TF_TYPE_BYPASS)
260 return false;
261
262 PERF_TRACE();
263
264 arr_points = lut_params->arr_points;
265 rgb_resulted = lut_params->rgb_resulted;
266 hw_points = 0;
267
268 memset(lut_params, 0, sizeof(struct pwl_params));
269 memset(seg_distr, 0, sizeof(seg_distr));
270
271 if (output_tf->tf == TRANSFER_FUNCTION_PQ) {
272 /* 32 segments
273 * segments are from 2^-25 to 2^7
274 */
275 for (i = 0; i < NUMBER_REGIONS ; i++)
276 seg_distr[i] = 3;
277
278 region_start = -MAX_LOW_POINT;
279 region_end = NUMBER_REGIONS - MAX_LOW_POINT;
280 } else {
281 /* 10 segments
282 * segment is from 2^-10 to 2^0
283 * There are less than 256 points, for optimization
284 */
285 seg_distr[0] = 3;
286 seg_distr[1] = 4;
287 seg_distr[2] = 4;
288 seg_distr[3] = 4;
289 seg_distr[4] = 4;
290 seg_distr[5] = 4;
291 seg_distr[6] = 4;
292 seg_distr[7] = 4;
293 seg_distr[8] = 4;
294 seg_distr[9] = 4;
295 seg_distr[10] = 1;
296
297 region_start = -10;
298 region_end = 1;
299 }
300
301 for (i = region_end - region_start; i < MAX_REGIONS_NUMBER ; i++)
302 seg_distr[i] = -1;
303
304 for (k = 0; k < MAX_REGIONS_NUMBER; k++) {
305 if (seg_distr[k] != -1)
306 hw_points += (1 << seg_distr[k]);
307 }
308
309 j = 0;
310 for (k = 0; k < (region_end - region_start); k++) {
311 increment = NUMBER_SW_SEGMENTS / (1 << seg_distr[k]);
312 start_index = (region_start + k + MAX_LOW_POINT) *
313 NUMBER_SW_SEGMENTS;
314 for (i = start_index; i < start_index + NUMBER_SW_SEGMENTS;
315 i += increment) {
316 if (j == hw_points - 1)
317 break;
318 rgb_resulted[j].red = output_tf->tf_pts.red[i];
319 rgb_resulted[j].green = output_tf->tf_pts.green[i];
320 rgb_resulted[j].blue = output_tf->tf_pts.blue[i];
321 j++;
322 }
323 }
324
325 /* last point */
326 start_index = (region_end + MAX_LOW_POINT) * NUMBER_SW_SEGMENTS;
327 rgb_resulted[hw_points - 1].red = output_tf->tf_pts.red[start_index];
328 rgb_resulted[hw_points - 1].green = output_tf->tf_pts.green[start_index];
329 rgb_resulted[hw_points - 1].blue = output_tf->tf_pts.blue[start_index];
330
331 arr_points[0].x = dc_fixpt_pow(dc_fixpt_from_int(2),
332 dc_fixpt_from_int(region_start));
333 arr_points[1].x = dc_fixpt_pow(dc_fixpt_from_int(2),
334 dc_fixpt_from_int(region_end));
335
336 y_r = rgb_resulted[0].red;
337 y_g = rgb_resulted[0].green;
338 y_b = rgb_resulted[0].blue;
339
340 y1_min = dc_fixpt_min(y_r, dc_fixpt_min(y_g, y_b));
341
342 arr_points[0].y = y1_min;
343 arr_points[0].slope = dc_fixpt_div(arr_points[0].y, arr_points[0].x);
344 y_r = rgb_resulted[hw_points - 1].red;
345 y_g = rgb_resulted[hw_points - 1].green;
346 y_b = rgb_resulted[hw_points - 1].blue;
347
348 /* see comment above, m_arrPoints[1].y should be the Y value for the
349 * region end (m_numOfHwPoints), not last HW point(m_numOfHwPoints - 1)
350 */
351 y3_max = dc_fixpt_max(y_r, dc_fixpt_max(y_g, y_b));
352
353 arr_points[1].y = y3_max;
354
355 arr_points[1].slope = dc_fixpt_zero;
356
357 if (output_tf->tf == TRANSFER_FUNCTION_PQ) {
358 /* for PQ, we want to have a straight line from last HW X point,
359 * and the slope to be such that we hit 1.0 at 10000 nits.
360 */
361 const struct fixed31_32 end_value =
362 dc_fixpt_from_int(125);
363
364 arr_points[1].slope = dc_fixpt_div(
365 dc_fixpt_sub(dc_fixpt_one, arr_points[1].y),
366 dc_fixpt_sub(end_value, arr_points[1].x));
367 }
368
369 lut_params->hw_points_num = hw_points;
370
371 k = 0;
372 for (i = 1; i < MAX_REGIONS_NUMBER; i++) {
373 if (seg_distr[k] != -1) {
374 lut_params->arr_curve_points[k].segments_num =
375 seg_distr[k];
376 lut_params->arr_curve_points[i].offset =
377 lut_params->arr_curve_points[k].offset + (1 << seg_distr[k]);
378 }
379 k++;
380 }
381
382 if (seg_distr[k] != -1)
383 lut_params->arr_curve_points[k].segments_num = seg_distr[k];
384
385 rgb = rgb_resulted;
386 rgb_plus_1 = rgb_resulted + 1;
387
388 i = 1;
389 while (i != hw_points + 1) {
390 if (dc_fixpt_lt(rgb_plus_1->red, rgb->red))
391 rgb_plus_1->red = rgb->red;
392 if (dc_fixpt_lt(rgb_plus_1->green, rgb->green))
393 rgb_plus_1->green = rgb->green;
394 if (dc_fixpt_lt(rgb_plus_1->blue, rgb->blue))
395 rgb_plus_1->blue = rgb->blue;
396
397 rgb->delta_red = dc_fixpt_sub(rgb_plus_1->red, rgb->red);
398 rgb->delta_green = dc_fixpt_sub(rgb_plus_1->green, rgb->green);
399 rgb->delta_blue = dc_fixpt_sub(rgb_plus_1->blue, rgb->blue);
400
401 if (fixpoint == true) {
402 rgb->delta_red_reg = dc_fixpt_clamp_u0d10(rgb->delta_red);
403 rgb->delta_green_reg = dc_fixpt_clamp_u0d10(rgb->delta_green);
404 rgb->delta_blue_reg = dc_fixpt_clamp_u0d10(rgb->delta_blue);
405 rgb->red_reg = dc_fixpt_clamp_u0d14(rgb->red);
406 rgb->green_reg = dc_fixpt_clamp_u0d14(rgb->green);
407 rgb->blue_reg = dc_fixpt_clamp_u0d14(rgb->blue);
408 }
409
410 ++rgb_plus_1;
411 ++rgb;
412 ++i;
413 }
414 cm_helper_convert_to_custom_float(rgb_resulted,
415 lut_params->arr_points,
416 hw_points, fixpoint);
417
418 return true;
419 }
420
421 #define NUM_DEGAMMA_REGIONS 12
422
423
cm_helper_translate_curve_to_degamma_hw_format(const struct dc_transfer_func * output_tf,struct pwl_params * lut_params)424 bool cm_helper_translate_curve_to_degamma_hw_format(
425 const struct dc_transfer_func *output_tf,
426 struct pwl_params *lut_params)
427 {
428 struct curve_points *arr_points;
429 struct pwl_result_data *rgb_resulted;
430 struct pwl_result_data *rgb;
431 struct pwl_result_data *rgb_plus_1;
432 struct fixed31_32 y_r;
433 struct fixed31_32 y_g;
434 struct fixed31_32 y_b;
435 struct fixed31_32 y1_min;
436 struct fixed31_32 y3_max;
437
438 int32_t region_start, region_end;
439 int32_t i;
440 uint32_t j, k, seg_distr[MAX_REGIONS_NUMBER], increment, start_index, hw_points;
441
442 if (output_tf == NULL || lut_params == NULL || output_tf->type == TF_TYPE_BYPASS)
443 return false;
444
445 PERF_TRACE();
446
447 arr_points = lut_params->arr_points;
448 rgb_resulted = lut_params->rgb_resulted;
449 hw_points = 0;
450
451 memset(lut_params, 0, sizeof(struct pwl_params));
452 memset(seg_distr, 0, sizeof(seg_distr));
453
454 region_start = -NUM_DEGAMMA_REGIONS;
455 region_end = 0;
456
457
458 for (i = region_end - region_start; i < MAX_REGIONS_NUMBER ; i++)
459 seg_distr[i] = -1;
460 /* 12 segments
461 * segments are from 2^-12 to 0
462 */
463 for (i = 0; i < NUM_DEGAMMA_REGIONS ; i++)
464 seg_distr[i] = 4;
465
466 for (k = 0; k < MAX_REGIONS_NUMBER; k++) {
467 if (seg_distr[k] != -1)
468 hw_points += (1 << seg_distr[k]);
469 }
470
471 j = 0;
472 for (k = 0; k < (region_end - region_start); k++) {
473 increment = NUMBER_SW_SEGMENTS / (1 << seg_distr[k]);
474 start_index = (region_start + k + MAX_LOW_POINT) *
475 NUMBER_SW_SEGMENTS;
476 for (i = start_index; i < start_index + NUMBER_SW_SEGMENTS;
477 i += increment) {
478 if (j == hw_points - 1)
479 break;
480 rgb_resulted[j].red = output_tf->tf_pts.red[i];
481 rgb_resulted[j].green = output_tf->tf_pts.green[i];
482 rgb_resulted[j].blue = output_tf->tf_pts.blue[i];
483 j++;
484 }
485 }
486
487 /* last point */
488 start_index = (region_end + MAX_LOW_POINT) * NUMBER_SW_SEGMENTS;
489 rgb_resulted[hw_points - 1].red = output_tf->tf_pts.red[start_index];
490 rgb_resulted[hw_points - 1].green = output_tf->tf_pts.green[start_index];
491 rgb_resulted[hw_points - 1].blue = output_tf->tf_pts.blue[start_index];
492
493 arr_points[0].x = dc_fixpt_pow(dc_fixpt_from_int(2),
494 dc_fixpt_from_int(region_start));
495 arr_points[1].x = dc_fixpt_pow(dc_fixpt_from_int(2),
496 dc_fixpt_from_int(region_end));
497
498 y_r = rgb_resulted[0].red;
499 y_g = rgb_resulted[0].green;
500 y_b = rgb_resulted[0].blue;
501
502 y1_min = dc_fixpt_min(y_r, dc_fixpt_min(y_g, y_b));
503
504 arr_points[0].y = y1_min;
505 arr_points[0].slope = dc_fixpt_div(arr_points[0].y, arr_points[0].x);
506 y_r = rgb_resulted[hw_points - 1].red;
507 y_g = rgb_resulted[hw_points - 1].green;
508 y_b = rgb_resulted[hw_points - 1].blue;
509
510 /* see comment above, m_arrPoints[1].y should be the Y value for the
511 * region end (m_numOfHwPoints), not last HW point(m_numOfHwPoints - 1)
512 */
513 y3_max = dc_fixpt_max(y_r, dc_fixpt_max(y_g, y_b));
514
515 arr_points[1].y = y3_max;
516
517 arr_points[1].slope = dc_fixpt_zero;
518
519 if (output_tf->tf == TRANSFER_FUNCTION_PQ) {
520 /* for PQ, we want to have a straight line from last HW X point,
521 * and the slope to be such that we hit 1.0 at 10000 nits.
522 */
523 const struct fixed31_32 end_value =
524 dc_fixpt_from_int(125);
525
526 arr_points[1].slope = dc_fixpt_div(
527 dc_fixpt_sub(dc_fixpt_one, arr_points[1].y),
528 dc_fixpt_sub(end_value, arr_points[1].x));
529 }
530
531 lut_params->hw_points_num = hw_points;
532
533 k = 0;
534 for (i = 1; i < MAX_REGIONS_NUMBER; i++) {
535 if (seg_distr[k] != -1) {
536 lut_params->arr_curve_points[k].segments_num =
537 seg_distr[k];
538 lut_params->arr_curve_points[i].offset =
539 lut_params->arr_curve_points[k].offset + (1 << seg_distr[k]);
540 }
541 k++;
542 }
543
544 if (seg_distr[k] != -1)
545 lut_params->arr_curve_points[k].segments_num = seg_distr[k];
546
547 rgb = rgb_resulted;
548 rgb_plus_1 = rgb_resulted + 1;
549
550 i = 1;
551 while (i != hw_points + 1) {
552 if (dc_fixpt_lt(rgb_plus_1->red, rgb->red))
553 rgb_plus_1->red = rgb->red;
554 if (dc_fixpt_lt(rgb_plus_1->green, rgb->green))
555 rgb_plus_1->green = rgb->green;
556 if (dc_fixpt_lt(rgb_plus_1->blue, rgb->blue))
557 rgb_plus_1->blue = rgb->blue;
558
559 rgb->delta_red = dc_fixpt_sub(rgb_plus_1->red, rgb->red);
560 rgb->delta_green = dc_fixpt_sub(rgb_plus_1->green, rgb->green);
561 rgb->delta_blue = dc_fixpt_sub(rgb_plus_1->blue, rgb->blue);
562
563 ++rgb_plus_1;
564 ++rgb;
565 ++i;
566 }
567 cm_helper_convert_to_custom_float(rgb_resulted,
568 lut_params->arr_points,
569 hw_points, false);
570
571 return true;
572 }
573