xref: /linux/drivers/gpu/drm/amd/display/dc/dml/dcn21/display_rq_dlg_calc_21.c (revision 3fd6c59042dbba50391e30862beac979491145fe)
1 /*
2  * Copyright 2017 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 
26 
27 #include "../display_mode_lib.h"
28 #include "../display_mode_vba.h"
29 #include "../dml_inline_defs.h"
30 #include "display_rq_dlg_calc_21.h"
31 
32 /*
33  * NOTE:
34  *   This file is gcc-parseable HW gospel, coming straight from HW engineers.
35  *
36  * It doesn't adhere to Linux kernel style and sometimes will do things in odd
37  * ways. Unless there is something clearly wrong with it the code should
38  * remain as-is as it provides us with a guarantee from HW that it is correct.
39  */
40 
41 static void calculate_ttu_cursor(
42 		struct display_mode_lib *mode_lib,
43 		double *refcyc_per_req_delivery_pre_cur,
44 		double *refcyc_per_req_delivery_cur,
45 		double refclk_freq_in_mhz,
46 		double ref_freq_to_pix_freq,
47 		double hscale_pixel_rate_l,
48 		double hscl_ratio,
49 		double vratio_pre_l,
50 		double vratio_l,
51 		unsigned int cur_width,
52 		enum cursor_bpp cur_bpp);
53 
get_bytes_per_element(enum source_format_class source_format,bool is_chroma)54 static unsigned int get_bytes_per_element(enum source_format_class source_format, bool is_chroma)
55 {
56 	unsigned int ret_val = 1;
57 
58 	if (source_format == dm_444_16) {
59 		if (!is_chroma)
60 			ret_val = 2;
61 	} else if (source_format == dm_444_32) {
62 		if (!is_chroma)
63 			ret_val = 4;
64 	} else if (source_format == dm_444_64) {
65 		if (!is_chroma)
66 			ret_val = 8;
67 	} else if (source_format == dm_420_8) {
68 		if (is_chroma)
69 			ret_val = 2;
70 		else
71 			ret_val = 1;
72 	} else if (source_format == dm_420_10) {
73 		if (is_chroma)
74 			ret_val = 4;
75 		else
76 			ret_val = 2;
77 	} else if (source_format == dm_444_8) {
78 		ret_val = 1;
79 	}
80 	return ret_val;
81 }
82 
is_dual_plane(enum source_format_class source_format)83 static bool is_dual_plane(enum source_format_class source_format)
84 {
85 	bool ret_val = false;
86 
87 	if ((source_format == dm_420_8) || (source_format == dm_420_10))
88 		ret_val = true;
89 
90 	return ret_val;
91 }
92 
get_refcyc_per_delivery(struct display_mode_lib * mode_lib,double refclk_freq_in_mhz,double pclk_freq_in_mhz,bool odm_combine,unsigned int recout_width,unsigned int hactive,double vratio,double hscale_pixel_rate,unsigned int delivery_width,unsigned int req_per_swath_ub)93 static double get_refcyc_per_delivery(
94 		struct display_mode_lib *mode_lib,
95 		double refclk_freq_in_mhz,
96 		double pclk_freq_in_mhz,
97 		bool odm_combine,
98 		unsigned int recout_width,
99 		unsigned int hactive,
100 		double vratio,
101 		double hscale_pixel_rate,
102 		unsigned int delivery_width,
103 		unsigned int req_per_swath_ub)
104 {
105 	double refcyc_per_delivery = 0.0;
106 
107 	if (vratio <= 1.0) {
108 		if (odm_combine)
109 			refcyc_per_delivery = (double) refclk_freq_in_mhz
110 					* dml_min((double) recout_width, (double) hactive / 2.0)
111 					/ pclk_freq_in_mhz / (double) req_per_swath_ub;
112 		else
113 			refcyc_per_delivery = (double) refclk_freq_in_mhz * (double) recout_width
114 					/ pclk_freq_in_mhz / (double) req_per_swath_ub;
115 	} else {
116 		refcyc_per_delivery = (double) refclk_freq_in_mhz * (double) delivery_width
117 				/ (double) hscale_pixel_rate / (double) req_per_swath_ub;
118 	}
119 
120 	dml_print("DML_DLG: %s: refclk_freq_in_mhz = %3.2f\n", __func__, refclk_freq_in_mhz);
121 	dml_print("DML_DLG: %s: pclk_freq_in_mhz   = %3.2f\n", __func__, pclk_freq_in_mhz);
122 	dml_print("DML_DLG: %s: recout_width       = %d\n", __func__, recout_width);
123 	dml_print("DML_DLG: %s: vratio             = %3.2f\n", __func__, vratio);
124 	dml_print("DML_DLG: %s: req_per_swath_ub   = %d\n", __func__, req_per_swath_ub);
125 	dml_print("DML_DLG: %s: refcyc_per_delivery= %3.2f\n", __func__, refcyc_per_delivery);
126 
127 	return refcyc_per_delivery;
128 
129 }
130 
get_blk_size_bytes(const enum source_macro_tile_size tile_size)131 static unsigned int get_blk_size_bytes(const enum source_macro_tile_size tile_size)
132 {
133 	if (tile_size == dm_256k_tile)
134 		return (256 * 1024);
135 	else if (tile_size == dm_64k_tile)
136 		return (64 * 1024);
137 	else
138 		return (4 * 1024);
139 }
140 
extract_rq_sizing_regs(struct display_mode_lib * mode_lib,display_data_rq_regs_st * rq_regs,const display_data_rq_sizing_params_st * rq_sizing)141 static void extract_rq_sizing_regs(
142 		struct display_mode_lib *mode_lib,
143 		display_data_rq_regs_st *rq_regs,
144 		const display_data_rq_sizing_params_st *rq_sizing)
145 {
146 	dml_print("DML_DLG: %s: rq_sizing param\n", __func__);
147 	print__data_rq_sizing_params_st(mode_lib, rq_sizing);
148 
149 	rq_regs->chunk_size = dml_log2(rq_sizing->chunk_bytes) - 10;
150 
151 	if (rq_sizing->min_chunk_bytes == 0)
152 		rq_regs->min_chunk_size = 0;
153 	else
154 		rq_regs->min_chunk_size = dml_log2(rq_sizing->min_chunk_bytes) - 8 + 1;
155 
156 	rq_regs->meta_chunk_size = dml_log2(rq_sizing->meta_chunk_bytes) - 10;
157 	if (rq_sizing->min_meta_chunk_bytes == 0)
158 		rq_regs->min_meta_chunk_size = 0;
159 	else
160 		rq_regs->min_meta_chunk_size = dml_log2(rq_sizing->min_meta_chunk_bytes) - 6 + 1;
161 
162 	rq_regs->dpte_group_size = dml_log2(rq_sizing->dpte_group_bytes) - 6;
163 	rq_regs->mpte_group_size = dml_log2(rq_sizing->mpte_group_bytes) - 6;
164 }
165 
extract_rq_regs(struct display_mode_lib * mode_lib,display_rq_regs_st * rq_regs,const display_rq_params_st * rq_param)166 static void extract_rq_regs(
167 		struct display_mode_lib *mode_lib,
168 		display_rq_regs_st *rq_regs,
169 		const display_rq_params_st *rq_param)
170 {
171 	unsigned int detile_buf_size_in_bytes = mode_lib->ip.det_buffer_size_kbytes * 1024;
172 	unsigned int detile_buf_plane1_addr = 0;
173 
174 	extract_rq_sizing_regs(mode_lib, &(rq_regs->rq_regs_l), &rq_param->sizing.rq_l);
175 
176 	rq_regs->rq_regs_l.pte_row_height_linear = dml_floor(
177 			dml_log2(rq_param->dlg.rq_l.dpte_row_height),
178 			1) - 3;
179 
180 	if (rq_param->yuv420) {
181 		extract_rq_sizing_regs(mode_lib, &(rq_regs->rq_regs_c), &rq_param->sizing.rq_c);
182 		rq_regs->rq_regs_c.pte_row_height_linear = dml_floor(
183 				dml_log2(rq_param->dlg.rq_c.dpte_row_height),
184 				1) - 3;
185 	}
186 
187 	rq_regs->rq_regs_l.swath_height = dml_log2(rq_param->dlg.rq_l.swath_height);
188 	rq_regs->rq_regs_c.swath_height = dml_log2(rq_param->dlg.rq_c.swath_height);
189 
190 	// FIXME: take the max between luma, chroma chunk size?
191 	// okay for now, as we are setting chunk_bytes to 8kb anyways
192 	if (rq_param->sizing.rq_l.chunk_bytes >= 32 * 1024) { //32kb
193 		rq_regs->drq_expansion_mode = 0;
194 	} else {
195 		rq_regs->drq_expansion_mode = 2;
196 	}
197 	rq_regs->prq_expansion_mode = 1;
198 	rq_regs->mrq_expansion_mode = 1;
199 	rq_regs->crq_expansion_mode = 1;
200 
201 	if (rq_param->yuv420) {
202 		if ((double) rq_param->misc.rq_l.stored_swath_bytes
203 				/ (double) rq_param->misc.rq_c.stored_swath_bytes <= 1.5) {
204 			detile_buf_plane1_addr = (detile_buf_size_in_bytes / 2.0 / 64.0); // half to chroma
205 		} else {
206 			detile_buf_plane1_addr = dml_round_to_multiple(
207 					(unsigned int) ((2.0 * detile_buf_size_in_bytes) / 3.0),
208 					256,
209 					0) / 64.0; // 2/3 to chroma
210 		}
211 	}
212 	rq_regs->plane1_base_address = detile_buf_plane1_addr;
213 }
214 
handle_det_buf_split(struct display_mode_lib * mode_lib,display_rq_params_st * rq_param,const display_pipe_source_params_st * pipe_src_param)215 static void handle_det_buf_split(
216 		struct display_mode_lib *mode_lib,
217 		display_rq_params_st *rq_param,
218 		const display_pipe_source_params_st *pipe_src_param)
219 {
220 	unsigned int total_swath_bytes = 0;
221 	unsigned int swath_bytes_l = 0;
222 	unsigned int swath_bytes_c = 0;
223 	unsigned int full_swath_bytes_packed_l = 0;
224 	unsigned int full_swath_bytes_packed_c = 0;
225 	bool req128_l = false;
226 	bool req128_c = false;
227 	bool surf_linear = (pipe_src_param->sw_mode == dm_sw_linear);
228 	bool surf_vert = (pipe_src_param->source_scan == dm_vert);
229 	unsigned int log2_swath_height_l = 0;
230 	unsigned int log2_swath_height_c = 0;
231 	unsigned int detile_buf_size_in_bytes = mode_lib->ip.det_buffer_size_kbytes * 1024;
232 
233 	full_swath_bytes_packed_l = rq_param->misc.rq_l.full_swath_bytes;
234 	full_swath_bytes_packed_c = rq_param->misc.rq_c.full_swath_bytes;
235 
236 	if (rq_param->yuv420_10bpc) {
237 		full_swath_bytes_packed_l = dml_round_to_multiple(
238 				rq_param->misc.rq_l.full_swath_bytes * 2 / 3,
239 				256,
240 				1) + 256;
241 		full_swath_bytes_packed_c = dml_round_to_multiple(
242 				rq_param->misc.rq_c.full_swath_bytes * 2 / 3,
243 				256,
244 				1) + 256;
245 	}
246 
247 	if (rq_param->yuv420) {
248 		total_swath_bytes = 2 * full_swath_bytes_packed_l + 2 * full_swath_bytes_packed_c;
249 
250 		if (total_swath_bytes <= detile_buf_size_in_bytes) { //full 256b request
251 			req128_l = false;
252 			req128_c = false;
253 			swath_bytes_l = full_swath_bytes_packed_l;
254 			swath_bytes_c = full_swath_bytes_packed_c;
255 		} else { //128b request (for luma only for yuv420 8bpc)
256 			req128_l = true;
257 			req128_c = false;
258 			swath_bytes_l = full_swath_bytes_packed_l / 2;
259 			swath_bytes_c = full_swath_bytes_packed_c;
260 		}
261 		// Note: assumption, the config that pass in will fit into
262 		//       the detiled buffer.
263 	} else {
264 		total_swath_bytes = 2 * full_swath_bytes_packed_l;
265 
266 		if (total_swath_bytes <= detile_buf_size_in_bytes)
267 			req128_l = false;
268 		else
269 			req128_l = true;
270 
271 		swath_bytes_l = total_swath_bytes;
272 		swath_bytes_c = 0;
273 	}
274 	rq_param->misc.rq_l.stored_swath_bytes = swath_bytes_l;
275 	rq_param->misc.rq_c.stored_swath_bytes = swath_bytes_c;
276 
277 	if (surf_linear) {
278 		log2_swath_height_l = 0;
279 		log2_swath_height_c = 0;
280 	} else {
281 		unsigned int swath_height_l;
282 		unsigned int swath_height_c;
283 
284 		if (!surf_vert) {
285 			swath_height_l = rq_param->misc.rq_l.blk256_height;
286 			swath_height_c = rq_param->misc.rq_c.blk256_height;
287 		} else {
288 			swath_height_l = rq_param->misc.rq_l.blk256_width;
289 			swath_height_c = rq_param->misc.rq_c.blk256_width;
290 		}
291 
292 		if (swath_height_l > 0)
293 			log2_swath_height_l = dml_log2(swath_height_l);
294 
295 		if (req128_l && log2_swath_height_l > 0)
296 			log2_swath_height_l -= 1;
297 
298 		if (swath_height_c > 0)
299 			log2_swath_height_c = dml_log2(swath_height_c);
300 	}
301 
302 	rq_param->dlg.rq_l.swath_height = 1 << log2_swath_height_l;
303 	rq_param->dlg.rq_c.swath_height = 1 << log2_swath_height_c;
304 
305 	dml_print("DML_DLG: %s: req128_l = %0d\n", __func__, req128_l);
306 	dml_print("DML_DLG: %s: req128_c = %0d\n", __func__, req128_c);
307 	dml_print(
308 			"DML_DLG: %s: full_swath_bytes_packed_l = %0d\n",
309 			__func__,
310 			full_swath_bytes_packed_l);
311 	dml_print(
312 			"DML_DLG: %s: full_swath_bytes_packed_c = %0d\n",
313 			__func__,
314 			full_swath_bytes_packed_c);
315 }
316 
get_meta_and_pte_attr(struct display_mode_lib * mode_lib,display_data_rq_dlg_params_st * rq_dlg_param,display_data_rq_misc_params_st * rq_misc_param,display_data_rq_sizing_params_st * rq_sizing_param,unsigned int vp_width,unsigned int vp_height,unsigned int data_pitch,unsigned int meta_pitch,unsigned int source_format,unsigned int tiling,unsigned int macro_tile_size,unsigned int source_scan,unsigned int hostvm_enable,unsigned int is_chroma)317 static void get_meta_and_pte_attr(
318 		struct display_mode_lib *mode_lib,
319 		display_data_rq_dlg_params_st *rq_dlg_param,
320 		display_data_rq_misc_params_st *rq_misc_param,
321 		display_data_rq_sizing_params_st *rq_sizing_param,
322 		unsigned int vp_width,
323 		unsigned int vp_height,
324 		unsigned int data_pitch,
325 		unsigned int meta_pitch,
326 		unsigned int source_format,
327 		unsigned int tiling,
328 		unsigned int macro_tile_size,
329 		unsigned int source_scan,
330 		unsigned int hostvm_enable,
331 		unsigned int is_chroma)
332 {
333 	bool surf_linear = (tiling == dm_sw_linear);
334 	bool surf_vert = (source_scan == dm_vert);
335 
336 	unsigned int bytes_per_element;
337 	unsigned int bytes_per_element_y = get_bytes_per_element(
338 			(enum source_format_class) (source_format),
339 			false);
340 	unsigned int bytes_per_element_c = get_bytes_per_element(
341 			(enum source_format_class) (source_format),
342 			true);
343 
344 	unsigned int blk256_width = 0;
345 	unsigned int blk256_height = 0;
346 
347 	unsigned int blk256_width_y = 0;
348 	unsigned int blk256_height_y = 0;
349 	unsigned int blk256_width_c = 0;
350 	unsigned int blk256_height_c = 0;
351 	unsigned int log2_bytes_per_element;
352 	unsigned int log2_blk256_width;
353 	unsigned int log2_blk256_height;
354 	unsigned int blk_bytes;
355 	unsigned int log2_blk_bytes;
356 	unsigned int log2_blk_height;
357 	unsigned int log2_blk_width;
358 	unsigned int log2_meta_req_bytes;
359 	unsigned int log2_meta_req_height;
360 	unsigned int log2_meta_req_width;
361 	unsigned int meta_req_width;
362 	unsigned int meta_req_height;
363 	unsigned int log2_meta_row_height;
364 	unsigned int meta_row_width_ub;
365 	unsigned int log2_meta_chunk_bytes;
366 	unsigned int log2_meta_chunk_height;
367 
368 	//full sized meta chunk width in unit of data elements
369 	unsigned int log2_meta_chunk_width;
370 	unsigned int log2_min_meta_chunk_bytes;
371 	unsigned int min_meta_chunk_width;
372 	unsigned int meta_chunk_width;
373 	unsigned int meta_chunk_per_row_int;
374 	unsigned int meta_row_remainder;
375 	unsigned int meta_chunk_threshold;
376 	unsigned int meta_blk_bytes;
377 	unsigned int meta_blk_height;
378 	unsigned int meta_blk_width;
379 	unsigned int meta_surface_bytes;
380 	unsigned int vmpg_bytes;
381 	unsigned int meta_pte_req_per_frame_ub;
382 	unsigned int meta_pte_bytes_per_frame_ub;
383 	const unsigned int log2_vmpg_bytes = dml_log2(mode_lib->soc.vmm_page_size_bytes);
384 	const unsigned int dpte_buf_in_pte_reqs =
385 		mode_lib->ip.dpte_buffer_size_in_pte_reqs_luma + mode_lib->ip.dpte_buffer_size_in_pte_reqs_chroma;
386 	const unsigned int pde_proc_buffer_size_64k_reqs =
387 			mode_lib->ip.pde_proc_buffer_size_64k_reqs;
388 
389 	unsigned int log2_vmpg_height = 0;
390 	unsigned int log2_vmpg_width = 0;
391 	unsigned int log2_dpte_req_height_ptes = 0;
392 	unsigned int log2_dpte_req_height = 0;
393 	unsigned int log2_dpte_req_width = 0;
394 	unsigned int log2_dpte_row_height_linear = 0;
395 	unsigned int log2_dpte_row_height = 0;
396 	unsigned int log2_dpte_group_width = 0;
397 	unsigned int dpte_row_width_ub = 0;
398 	unsigned int dpte_req_height = 0;
399 	unsigned int dpte_req_width = 0;
400 	unsigned int dpte_group_width = 0;
401 	unsigned int log2_dpte_group_bytes = 0;
402 	unsigned int log2_dpte_group_length = 0;
403 	unsigned int pde_buf_entries;
404 	bool yuv420 = (source_format == dm_420_8 || source_format == dm_420_10);
405 
406 	Calculate256BBlockSizes(
407 			(enum source_format_class) (source_format),
408 			(enum dm_swizzle_mode) (tiling),
409 			bytes_per_element_y,
410 			bytes_per_element_c,
411 			&blk256_height_y,
412 			&blk256_height_c,
413 			&blk256_width_y,
414 			&blk256_width_c);
415 
416 	if (!is_chroma) {
417 		blk256_width = blk256_width_y;
418 		blk256_height = blk256_height_y;
419 		bytes_per_element = bytes_per_element_y;
420 	} else {
421 		blk256_width = blk256_width_c;
422 		blk256_height = blk256_height_c;
423 		bytes_per_element = bytes_per_element_c;
424 	}
425 
426 	log2_bytes_per_element = dml_log2(bytes_per_element);
427 
428 	dml_print("DML_DLG: %s: surf_linear        = %d\n", __func__, surf_linear);
429 	dml_print("DML_DLG: %s: surf_vert          = %d\n", __func__, surf_vert);
430 	dml_print("DML_DLG: %s: blk256_width       = %d\n", __func__, blk256_width);
431 	dml_print("DML_DLG: %s: blk256_height      = %d\n", __func__, blk256_height);
432 
433 	log2_blk256_width = dml_log2((double) blk256_width);
434 	log2_blk256_height = dml_log2((double) blk256_height);
435 	blk_bytes = surf_linear ?
436 			256 : get_blk_size_bytes((enum source_macro_tile_size) macro_tile_size);
437 	log2_blk_bytes = dml_log2((double) blk_bytes);
438 
439 	// remember log rule
440 	// "+" in log is multiply
441 	// "-" in log is divide
442 	// "/2" is like square root
443 	// blk is vertical biased
444 	if (tiling != dm_sw_linear)
445 		log2_blk_height = log2_blk256_height
446 				+ dml_ceil((double) (log2_blk_bytes - 8) / 2.0, 1);
447 	else
448 		log2_blk_height = 0;  // blk height of 1
449 
450 	log2_blk_width = log2_blk_bytes - log2_bytes_per_element - log2_blk_height;
451 
452 	if (!surf_vert) {
453 		rq_dlg_param->swath_width_ub = dml_round_to_multiple(vp_width - 1, blk256_width, 1)
454 				+ blk256_width;
455 		rq_dlg_param->req_per_swath_ub = rq_dlg_param->swath_width_ub >> log2_blk256_width;
456 	} else {
457 		rq_dlg_param->swath_width_ub = dml_round_to_multiple(
458 				vp_height - 1,
459 				blk256_height,
460 				1) + blk256_height;
461 		rq_dlg_param->req_per_swath_ub = rq_dlg_param->swath_width_ub >> log2_blk256_height;
462 	}
463 
464 	if (!surf_vert)
465 		rq_misc_param->full_swath_bytes = rq_dlg_param->swath_width_ub * blk256_height
466 				* bytes_per_element;
467 	else
468 		rq_misc_param->full_swath_bytes = rq_dlg_param->swath_width_ub * blk256_width
469 				* bytes_per_element;
470 
471 	rq_misc_param->blk256_height = blk256_height;
472 	rq_misc_param->blk256_width = blk256_width;
473 
474 	// -------
475 	// meta
476 	// -------
477 	log2_meta_req_bytes = 6; // meta request is 64b and is 8x8byte meta element
478 
479 	// each 64b meta request for dcn is 8x8 meta elements and
480 	// a meta element covers one 256b block of the data surface.
481 	log2_meta_req_height = log2_blk256_height + 3; // meta req is 8x8 byte, each byte represent 1 blk256
482 	log2_meta_req_width = log2_meta_req_bytes + 8 - log2_bytes_per_element
483 			- log2_meta_req_height;
484 	meta_req_width = 1 << log2_meta_req_width;
485 	meta_req_height = 1 << log2_meta_req_height;
486 
487 	// the dimensions of a meta row are meta_row_width x meta_row_height in elements.
488 	// calculate upper bound of the meta_row_width
489 	if (!surf_vert) {
490 		log2_meta_row_height = log2_meta_req_height;
491 		meta_row_width_ub = dml_round_to_multiple(vp_width - 1, meta_req_width, 1)
492 				+ meta_req_width;
493 		rq_dlg_param->meta_req_per_row_ub = meta_row_width_ub / meta_req_width;
494 	} else {
495 		log2_meta_row_height = log2_meta_req_width;
496 		meta_row_width_ub = dml_round_to_multiple(vp_height - 1, meta_req_height, 1)
497 				+ meta_req_height;
498 		rq_dlg_param->meta_req_per_row_ub = meta_row_width_ub / meta_req_height;
499 	}
500 	rq_dlg_param->meta_bytes_per_row_ub = rq_dlg_param->meta_req_per_row_ub * 64;
501 
502 	rq_dlg_param->meta_row_height = 1 << log2_meta_row_height;
503 
504 	log2_meta_chunk_bytes = dml_log2(rq_sizing_param->meta_chunk_bytes);
505 	log2_meta_chunk_height = log2_meta_row_height;
506 
507 	//full sized meta chunk width in unit of data elements
508 	log2_meta_chunk_width = log2_meta_chunk_bytes + 8 - log2_bytes_per_element
509 			- log2_meta_chunk_height;
510 	log2_min_meta_chunk_bytes = dml_log2(rq_sizing_param->min_meta_chunk_bytes);
511 	min_meta_chunk_width = 1
512 			<< (log2_min_meta_chunk_bytes + 8 - log2_bytes_per_element
513 					- log2_meta_chunk_height);
514 	meta_chunk_width = 1 << log2_meta_chunk_width;
515 	meta_chunk_per_row_int = (unsigned int) (meta_row_width_ub / meta_chunk_width);
516 	meta_row_remainder = meta_row_width_ub % meta_chunk_width;
517 	meta_chunk_threshold = 0;
518 	meta_blk_bytes = 4096;
519 	meta_blk_height = blk256_height * 64;
520 	meta_blk_width = meta_blk_bytes * 256 / bytes_per_element / meta_blk_height;
521 	meta_surface_bytes = meta_pitch
522 			* (dml_round_to_multiple(vp_height - 1, meta_blk_height, 1)
523 					+ meta_blk_height) * bytes_per_element / 256;
524 	vmpg_bytes = mode_lib->soc.vmm_page_size_bytes;
525 	meta_pte_req_per_frame_ub = (dml_round_to_multiple(
526 			meta_surface_bytes - vmpg_bytes,
527 			8 * vmpg_bytes,
528 			1) + 8 * vmpg_bytes) / (8 * vmpg_bytes);
529 	meta_pte_bytes_per_frame_ub = meta_pte_req_per_frame_ub * 64; //64B mpte request
530 	rq_dlg_param->meta_pte_bytes_per_frame_ub = meta_pte_bytes_per_frame_ub;
531 
532 	dml_print("DML_DLG: %s: meta_blk_height             = %d\n", __func__, meta_blk_height);
533 	dml_print("DML_DLG: %s: meta_blk_width              = %d\n", __func__, meta_blk_width);
534 	dml_print("DML_DLG: %s: meta_surface_bytes          = %d\n", __func__, meta_surface_bytes);
535 	dml_print(
536 			"DML_DLG: %s: meta_pte_req_per_frame_ub   = %d\n",
537 			__func__,
538 			meta_pte_req_per_frame_ub);
539 	dml_print(
540 			"DML_DLG: %s: meta_pte_bytes_per_frame_ub = %d\n",
541 			__func__,
542 			meta_pte_bytes_per_frame_ub);
543 
544 	if (!surf_vert)
545 		meta_chunk_threshold = 2 * min_meta_chunk_width - meta_req_width;
546 	else
547 		meta_chunk_threshold = 2 * min_meta_chunk_width - meta_req_height;
548 
549 	if (meta_row_remainder <= meta_chunk_threshold)
550 		rq_dlg_param->meta_chunks_per_row_ub = meta_chunk_per_row_int + 1;
551 	else
552 		rq_dlg_param->meta_chunks_per_row_ub = meta_chunk_per_row_int + 2;
553 
554 	// ------
555 	// dpte
556 	// ------
557 	if (surf_linear) {
558 		log2_vmpg_height = 0;   // one line high
559 	} else {
560 		log2_vmpg_height = (log2_vmpg_bytes - 8) / 2 + log2_blk256_height;
561 	}
562 	log2_vmpg_width = log2_vmpg_bytes - log2_bytes_per_element - log2_vmpg_height;
563 
564 	// only 3 possible shapes for dpte request in dimensions of ptes: 8x1, 4x2, 2x4.
565 	if (surf_linear) { //one 64B PTE request returns 8 PTEs
566 		log2_dpte_req_height_ptes = 0;
567 		log2_dpte_req_width = log2_vmpg_width + 3;
568 		log2_dpte_req_height = 0;
569 	} else if (log2_blk_bytes == 12) { //4KB tile means 4kB page size
570 					   //one 64B req gives 8x1 PTEs for 4KB tile
571 		log2_dpte_req_height_ptes = 0;
572 		log2_dpte_req_width = log2_blk_width + 3;
573 		log2_dpte_req_height = log2_blk_height + 0;
574 	} else if ((log2_blk_bytes >= 16) && (log2_vmpg_bytes == 12)) { // tile block >= 64KB
575 									//two 64B reqs of 2x4 PTEs give 16 PTEs to cover 64KB
576 		log2_dpte_req_height_ptes = 4;
577 		log2_dpte_req_width = log2_blk256_width + 4; // log2_64KB_width
578 		log2_dpte_req_height = log2_blk256_height + 4; // log2_64KB_height
579 	} else { //64KB page size and must 64KB tile block
580 		 //one 64B req gives 8x1 PTEs for 64KB tile
581 		log2_dpte_req_height_ptes = 0;
582 		log2_dpte_req_width = log2_blk_width + 3;
583 		log2_dpte_req_height = log2_blk_height + 0;
584 	}
585 
586 	// The dpte request dimensions in data elements is dpte_req_width x dpte_req_height
587 	// log2_vmpg_width is how much 1 pte represent, now calculating how much a 64b pte req represent
588 	// That depends on the pte shape (i.e. 8x1, 4x2, 2x4)
589 	//log2_dpte_req_height    = log2_vmpg_height + log2_dpte_req_height_ptes;
590 	//log2_dpte_req_width     = log2_vmpg_width + log2_dpte_req_width_ptes;
591 	dpte_req_height = 1 << log2_dpte_req_height;
592 	dpte_req_width = 1 << log2_dpte_req_width;
593 
594 	// calculate pitch dpte row buffer can hold
595 	// round the result down to a power of two.
596 	pde_buf_entries =
597 			yuv420 ? (pde_proc_buffer_size_64k_reqs >> 1) : pde_proc_buffer_size_64k_reqs;
598 	if (surf_linear) {
599 		unsigned int dpte_row_height;
600 
601 		log2_dpte_row_height_linear = dml_floor(
602 				dml_log2(
603 						dml_min(
604 								64 * 1024 * pde_buf_entries
605 										/ bytes_per_element,
606 								dpte_buf_in_pte_reqs
607 										* dpte_req_width)
608 								/ data_pitch),
609 				1);
610 
611 		ASSERT(log2_dpte_row_height_linear >= 3);
612 
613 		if (log2_dpte_row_height_linear > 7)
614 			log2_dpte_row_height_linear = 7;
615 
616 		log2_dpte_row_height = log2_dpte_row_height_linear;
617 		// For linear, the dpte row is pitch dependent and the pte requests wrap at the pitch boundary.
618 		// the dpte_row_width_ub is the upper bound of data_pitch*dpte_row_height in elements with this unique buffering.
619 		dpte_row_height = 1 << log2_dpte_row_height;
620 		dpte_row_width_ub = dml_round_to_multiple(
621 				data_pitch * dpte_row_height - 1,
622 				dpte_req_width,
623 				1) + dpte_req_width;
624 		rq_dlg_param->dpte_req_per_row_ub = dpte_row_width_ub / dpte_req_width;
625 	} else {
626 		// the upper bound of the dpte_row_width without dependency on viewport position follows.
627 		// for tiled mode, row height is the same as req height and row store up to vp size upper bound
628 		if (!surf_vert) {
629 			log2_dpte_row_height = log2_dpte_req_height;
630 			dpte_row_width_ub = dml_round_to_multiple(vp_width - 1, dpte_req_width, 1)
631 					+ dpte_req_width;
632 			rq_dlg_param->dpte_req_per_row_ub = dpte_row_width_ub / dpte_req_width;
633 		} else {
634 			log2_dpte_row_height =
635 					(log2_blk_width < log2_dpte_req_width) ?
636 							log2_blk_width : log2_dpte_req_width;
637 			dpte_row_width_ub = dml_round_to_multiple(vp_height - 1, dpte_req_height, 1)
638 					+ dpte_req_height;
639 			rq_dlg_param->dpte_req_per_row_ub = dpte_row_width_ub / dpte_req_height;
640 		}
641 	}
642 	if (log2_blk_bytes >= 16 && log2_vmpg_bytes == 12) // tile block >= 64KB
643 		rq_dlg_param->dpte_bytes_per_row_ub = rq_dlg_param->dpte_req_per_row_ub * 128; //2*64B dpte request
644 	else
645 		rq_dlg_param->dpte_bytes_per_row_ub = rq_dlg_param->dpte_req_per_row_ub * 64; //64B dpte request
646 
647 	rq_dlg_param->dpte_row_height = 1 << log2_dpte_row_height;
648 
649 	// the dpte_group_bytes is reduced for the specific case of vertical
650 	// access of a tile surface that has dpte request of 8x1 ptes.
651 
652 	if (hostvm_enable)
653 		rq_sizing_param->dpte_group_bytes = 512;
654 	else {
655 		if (!surf_linear & (log2_dpte_req_height_ptes == 0) & surf_vert) //reduced, in this case, will have page fault within a group
656 			rq_sizing_param->dpte_group_bytes = 512;
657 		else
658 			//full size
659 			rq_sizing_param->dpte_group_bytes = 2048;
660 	}
661 
662 	//since pte request size is 64byte, the number of data pte requests per full sized group is as follows.
663 	log2_dpte_group_bytes = dml_log2(rq_sizing_param->dpte_group_bytes);
664 	log2_dpte_group_length = log2_dpte_group_bytes - 6; //length in 64b requests
665 
666 	// full sized data pte group width in elements
667 	if (!surf_vert)
668 		log2_dpte_group_width = log2_dpte_group_length + log2_dpte_req_width;
669 	else
670 		log2_dpte_group_width = log2_dpte_group_length + log2_dpte_req_height;
671 
672 	//But if the tile block >=64KB and the page size is 4KB, then each dPTE request is 2*64B
673 	if ((log2_blk_bytes >= 16) && (log2_vmpg_bytes == 12)) // tile block >= 64KB
674 		log2_dpte_group_width = log2_dpte_group_width - 1;
675 
676 	dpte_group_width = 1 << log2_dpte_group_width;
677 
678 	// since dpte groups are only aligned to dpte_req_width and not dpte_group_width,
679 	// the upper bound for the dpte groups per row is as follows.
680 	rq_dlg_param->dpte_groups_per_row_ub = dml_ceil(
681 			(double) dpte_row_width_ub / dpte_group_width,
682 			1);
683 }
684 
get_surf_rq_param(struct display_mode_lib * mode_lib,display_data_rq_sizing_params_st * rq_sizing_param,display_data_rq_dlg_params_st * rq_dlg_param,display_data_rq_misc_params_st * rq_misc_param,const display_pipe_params_st * pipe_param,bool is_chroma)685 static void get_surf_rq_param(
686 		struct display_mode_lib *mode_lib,
687 		display_data_rq_sizing_params_st *rq_sizing_param,
688 		display_data_rq_dlg_params_st *rq_dlg_param,
689 		display_data_rq_misc_params_st *rq_misc_param,
690 		const display_pipe_params_st *pipe_param,
691 		bool is_chroma)
692 {
693 	unsigned int vp_width = 0;
694 	unsigned int vp_height = 0;
695 	unsigned int data_pitch = 0;
696 	unsigned int meta_pitch = 0;
697 	unsigned int ppe = 1;
698 
699 	// FIXME check if ppe apply for both luma and chroma in 422 case
700 	if (is_chroma) {
701 		vp_width = pipe_param->src.viewport_width_c / ppe;
702 		vp_height = pipe_param->src.viewport_height_c;
703 		data_pitch = pipe_param->src.data_pitch_c;
704 		meta_pitch = pipe_param->src.meta_pitch_c;
705 	} else {
706 		vp_width = pipe_param->src.viewport_width / ppe;
707 		vp_height = pipe_param->src.viewport_height;
708 		data_pitch = pipe_param->src.data_pitch;
709 		meta_pitch = pipe_param->src.meta_pitch;
710 	}
711 
712 	if (pipe_param->dest.odm_combine) {
713 		unsigned int access_dir;
714 		unsigned int full_src_vp_width;
715 		unsigned int hactive_half;
716 		unsigned int src_hactive_half;
717 		access_dir = (pipe_param->src.source_scan == dm_vert); // vp access direction: horizontal or vertical accessed
718 		hactive_half  = pipe_param->dest.hactive / 2;
719 		if (is_chroma) {
720 			full_src_vp_width = pipe_param->scale_ratio_depth.hscl_ratio_c * pipe_param->dest.full_recout_width;
721 			src_hactive_half  = pipe_param->scale_ratio_depth.hscl_ratio_c * hactive_half;
722 		} else {
723 			full_src_vp_width = pipe_param->scale_ratio_depth.hscl_ratio * pipe_param->dest.full_recout_width;
724 			src_hactive_half  = pipe_param->scale_ratio_depth.hscl_ratio * hactive_half;
725 		}
726 
727 		if (access_dir == 0) {
728 			vp_width = dml_min(full_src_vp_width, src_hactive_half);
729 			dml_print("DML_DLG: %s: vp_width = %d\n", __func__, vp_width);
730 		} else {
731 			vp_height = dml_min(full_src_vp_width, src_hactive_half);
732 			dml_print("DML_DLG: %s: vp_height = %d\n", __func__, vp_height);
733 
734 		}
735 		dml_print("DML_DLG: %s: full_src_vp_width = %d\n", __func__, full_src_vp_width);
736 		dml_print("DML_DLG: %s: hactive_half = %d\n", __func__, hactive_half);
737 		dml_print("DML_DLG: %s: src_hactive_half = %d\n", __func__, src_hactive_half);
738 	}
739 	rq_sizing_param->chunk_bytes = 8192;
740 
741 	if (rq_sizing_param->chunk_bytes == 64 * 1024)
742 		rq_sizing_param->min_chunk_bytes = 0;
743 	else
744 		rq_sizing_param->min_chunk_bytes = 1024;
745 
746 	rq_sizing_param->meta_chunk_bytes = 2048;
747 	rq_sizing_param->min_meta_chunk_bytes = 256;
748 
749 	if (pipe_param->src.hostvm)
750 		rq_sizing_param->mpte_group_bytes = 512;
751 	else
752 		rq_sizing_param->mpte_group_bytes = 2048;
753 
754 	get_meta_and_pte_attr(
755 			mode_lib,
756 			rq_dlg_param,
757 			rq_misc_param,
758 			rq_sizing_param,
759 			vp_width,
760 			vp_height,
761 			data_pitch,
762 			meta_pitch,
763 			pipe_param->src.source_format,
764 			pipe_param->src.sw_mode,
765 			pipe_param->src.macro_tile_size,
766 			pipe_param->src.source_scan,
767 			pipe_param->src.hostvm,
768 			is_chroma);
769 }
770 
dml_rq_dlg_get_rq_params(struct display_mode_lib * mode_lib,display_rq_params_st * rq_param,const display_pipe_params_st * pipe_param)771 static void dml_rq_dlg_get_rq_params(
772 		struct display_mode_lib *mode_lib,
773 		display_rq_params_st *rq_param,
774 		const display_pipe_params_st *pipe_param)
775 {
776 	// get param for luma surface
777 	rq_param->yuv420 = pipe_param->src.source_format == dm_420_8
778 			|| pipe_param->src.source_format == dm_420_10;
779 	rq_param->yuv420_10bpc = pipe_param->src.source_format == dm_420_10;
780 
781 	get_surf_rq_param(
782 			mode_lib,
783 			&(rq_param->sizing.rq_l),
784 			&(rq_param->dlg.rq_l),
785 			&(rq_param->misc.rq_l),
786 			pipe_param,
787 			0);
788 
789 	if (is_dual_plane((enum source_format_class) (pipe_param->src.source_format))) {
790 		// get param for chroma surface
791 		get_surf_rq_param(
792 				mode_lib,
793 				&(rq_param->sizing.rq_c),
794 				&(rq_param->dlg.rq_c),
795 				&(rq_param->misc.rq_c),
796 				pipe_param,
797 				1);
798 	}
799 
800 	// calculate how to split the det buffer space between luma and chroma
801 	handle_det_buf_split(mode_lib, rq_param, &pipe_param->src);
802 	print__rq_params_st(mode_lib, rq_param);
803 }
804 
dml21_rq_dlg_get_rq_reg(struct display_mode_lib * mode_lib,display_rq_regs_st * rq_regs,const display_pipe_params_st * pipe_param)805 void dml21_rq_dlg_get_rq_reg(
806 		struct display_mode_lib *mode_lib,
807 		display_rq_regs_st *rq_regs,
808 		const display_pipe_params_st *pipe_param)
809 {
810 	display_rq_params_st rq_param = {0};
811 
812 	memset(rq_regs, 0, sizeof(*rq_regs));
813 	dml_rq_dlg_get_rq_params(mode_lib, &rq_param, pipe_param);
814 	extract_rq_regs(mode_lib, rq_regs, &rq_param);
815 
816 	print__rq_regs_st(mode_lib, rq_regs);
817 }
818 
819 // Note: currently taken in as is.
820 // Nice to decouple code from hw register implement and extract code that are repeated for luma and chroma.
dml_rq_dlg_get_dlg_params(struct display_mode_lib * mode_lib,const display_e2e_pipe_params_st * e2e_pipe_param,const unsigned int num_pipes,const unsigned int pipe_idx,display_dlg_regs_st * disp_dlg_regs,display_ttu_regs_st * disp_ttu_regs,const display_rq_dlg_params_st * rq_dlg_param,const display_dlg_sys_params_st * dlg_sys_param,const bool cstate_en,const bool pstate_en)821 static void dml_rq_dlg_get_dlg_params(
822 		struct display_mode_lib *mode_lib,
823 		const display_e2e_pipe_params_st *e2e_pipe_param,
824 		const unsigned int num_pipes,
825 		const unsigned int pipe_idx,
826 		display_dlg_regs_st *disp_dlg_regs,
827 		display_ttu_regs_st *disp_ttu_regs,
828 		const display_rq_dlg_params_st *rq_dlg_param,
829 		const display_dlg_sys_params_st *dlg_sys_param,
830 		const bool cstate_en,
831 		const bool pstate_en)
832 {
833 	const display_pipe_source_params_st *src = &e2e_pipe_param[pipe_idx].pipe.src;
834 	const display_pipe_dest_params_st *dst = &e2e_pipe_param[pipe_idx].pipe.dest;
835 	const display_output_params_st *dout = &e2e_pipe_param[pipe_idx].dout;
836 	const display_clocks_and_cfg_st *clks = &e2e_pipe_param[pipe_idx].clks_cfg;
837 	const scaler_ratio_depth_st *scl = &e2e_pipe_param[pipe_idx].pipe.scale_ratio_depth;
838 	const scaler_taps_st *taps = &e2e_pipe_param[pipe_idx].pipe.scale_taps;
839 
840 	// -------------------------
841 	// Section 1.15.2.1: OTG dependent Params
842 	// -------------------------
843 	// Timing
844 	unsigned int htotal = dst->htotal;
845 	//    unsigned int hblank_start = dst.hblank_start; // TODO: Remove
846 	unsigned int hblank_end = dst->hblank_end;
847 	unsigned int vblank_start = dst->vblank_start;
848 	unsigned int vblank_end = dst->vblank_end;
849 	unsigned int min_vblank = mode_lib->ip.min_vblank_lines;
850 
851 	double dppclk_freq_in_mhz = clks->dppclk_mhz;
852 	double dispclk_freq_in_mhz = clks->dispclk_mhz;
853 	double refclk_freq_in_mhz = clks->refclk_mhz;
854 	double pclk_freq_in_mhz = dst->pixel_rate_mhz;
855 	bool interlaced = dst->interlaced;
856 
857 	double ref_freq_to_pix_freq = refclk_freq_in_mhz / pclk_freq_in_mhz;
858 
859 	double min_dcfclk_mhz;
860 	double t_calc_us;
861 	double min_ttu_vblank;
862 
863 	double min_dst_y_ttu_vblank;
864 	unsigned int dlg_vblank_start;
865 	bool dual_plane;
866 	unsigned int access_dir;
867 	unsigned int vp_height_l;
868 	unsigned int vp_width_l;
869 	unsigned int vp_height_c;
870 	unsigned int vp_width_c;
871 
872 	// Scaling
873 	unsigned int htaps_l;
874 	unsigned int htaps_c;
875 	double hratio_l;
876 	double hratio_c;
877 	double vratio_l;
878 	double vratio_c;
879 	bool scl_enable;
880 
881 	double line_time_in_us;
882 	//    double vinit_l;
883 	//    double vinit_c;
884 	//    double vinit_bot_l;
885 	//    double vinit_bot_c;
886 
887 	//    unsigned int swath_height_l;
888 	unsigned int swath_width_ub_l;
889 	//    unsigned int dpte_bytes_per_row_ub_l;
890 	unsigned int dpte_groups_per_row_ub_l;
891 	//    unsigned int meta_pte_bytes_per_frame_ub_l;
892 	//    unsigned int meta_bytes_per_row_ub_l;
893 
894 	//    unsigned int swath_height_c;
895 	unsigned int swath_width_ub_c;
896 	//   unsigned int dpte_bytes_per_row_ub_c;
897 	unsigned int dpte_groups_per_row_ub_c;
898 
899 	unsigned int meta_chunks_per_row_ub_l;
900 	unsigned int meta_chunks_per_row_ub_c;
901 	unsigned int vupdate_offset;
902 	unsigned int vupdate_width;
903 	unsigned int vready_offset;
904 
905 	unsigned int dppclk_delay_subtotal;
906 	unsigned int dispclk_delay_subtotal;
907 	unsigned int pixel_rate_delay_subtotal;
908 
909 	unsigned int vstartup_start;
910 	unsigned int dst_x_after_scaler;
911 	unsigned int dst_y_after_scaler;
912 	double line_wait;
913 	double dst_y_prefetch;
914 	double dst_y_per_vm_vblank;
915 	double dst_y_per_row_vblank;
916 	double dst_y_per_vm_flip;
917 	double dst_y_per_row_flip;
918 	double max_dst_y_per_vm_vblank;
919 	double max_dst_y_per_row_vblank;
920 	double lsw;
921 	double vratio_pre_l;
922 	double vratio_pre_c;
923 	unsigned int req_per_swath_ub_l;
924 	unsigned int req_per_swath_ub_c;
925 	unsigned int meta_row_height_l;
926 	unsigned int meta_row_height_c;
927 	unsigned int swath_width_pixels_ub_l;
928 	unsigned int swath_width_pixels_ub_c;
929 	unsigned int scaler_rec_in_width_l;
930 	unsigned int scaler_rec_in_width_c;
931 	unsigned int dpte_row_height_l;
932 	unsigned int dpte_row_height_c;
933 	double hscale_pixel_rate_l;
934 	double hscale_pixel_rate_c;
935 	double min_hratio_fact_l;
936 	double min_hratio_fact_c;
937 	double refcyc_per_line_delivery_pre_l;
938 	double refcyc_per_line_delivery_pre_c;
939 	double refcyc_per_line_delivery_l;
940 	double refcyc_per_line_delivery_c;
941 
942 	double refcyc_per_req_delivery_pre_l;
943 	double refcyc_per_req_delivery_pre_c;
944 	double refcyc_per_req_delivery_l;
945 	double refcyc_per_req_delivery_c;
946 
947 	unsigned int full_recout_width;
948 	double refcyc_per_req_delivery_pre_cur0;
949 	double refcyc_per_req_delivery_cur0;
950 	double refcyc_per_req_delivery_pre_cur1;
951 	double refcyc_per_req_delivery_cur1;
952 
953 	memset(disp_dlg_regs, 0, sizeof(*disp_dlg_regs));
954 	memset(disp_ttu_regs, 0, sizeof(*disp_ttu_regs));
955 
956 	dml_print("DML_DLG: %s:  cstate_en = %d\n", __func__, cstate_en);
957 	dml_print("DML_DLG: %s:  pstate_en = %d\n", __func__, pstate_en);
958 
959 	dml_print("DML_DLG: %s: dppclk_freq_in_mhz     = %3.2f\n", __func__, dppclk_freq_in_mhz);
960 	dml_print("DML_DLG: %s: dispclk_freq_in_mhz    = %3.2f\n", __func__, dispclk_freq_in_mhz);
961 	dml_print("DML_DLG: %s: refclk_freq_in_mhz     = %3.2f\n", __func__, refclk_freq_in_mhz);
962 	dml_print("DML_DLG: %s: pclk_freq_in_mhz       = %3.2f\n", __func__, pclk_freq_in_mhz);
963 	dml_print("DML_DLG: %s: interlaced             = %d\n", __func__, interlaced);
964 	ASSERT(ref_freq_to_pix_freq < 4.0);
965 
966 	disp_dlg_regs->ref_freq_to_pix_freq =
967 			(unsigned int) (ref_freq_to_pix_freq * dml_pow(2, 19));
968 	disp_dlg_regs->refcyc_per_htotal = (unsigned int) (ref_freq_to_pix_freq * (double) htotal
969 			* dml_pow(2, 8));
970 	disp_dlg_regs->dlg_vblank_end = interlaced ? (vblank_end / 2) : vblank_end; // 15 bits
971 	disp_dlg_regs->refcyc_h_blank_end = (unsigned int) ((double) hblank_end
972 			* (double) ref_freq_to_pix_freq);
973 	ASSERT(disp_dlg_regs->refcyc_h_blank_end < (unsigned int)dml_pow(2, 13));
974 
975 	min_dcfclk_mhz = dlg_sys_param->deepsleep_dcfclk_mhz;
976 	t_calc_us = get_tcalc(mode_lib, e2e_pipe_param, num_pipes);
977 	min_ttu_vblank = get_min_ttu_vblank(mode_lib, e2e_pipe_param, num_pipes, pipe_idx);
978 
979 	min_dst_y_ttu_vblank = min_ttu_vblank * pclk_freq_in_mhz / (double) htotal;
980 	dlg_vblank_start = interlaced ? (vblank_start / 2) : vblank_start;
981 
982 	disp_dlg_regs->min_dst_y_next_start = (unsigned int) (((double) dlg_vblank_start) * dml_pow(2, 2));
983 	ASSERT(disp_dlg_regs->min_dst_y_next_start < (unsigned int)dml_pow(2, 18));
984 
985 	dml_print(
986 			"DML_DLG: %s: min_dcfclk_mhz                         = %3.2f\n",
987 			__func__,
988 			min_dcfclk_mhz);
989 	dml_print(
990 			"DML_DLG: %s: min_ttu_vblank                         = %3.2f\n",
991 			__func__,
992 			min_ttu_vblank);
993 	dml_print(
994 			"DML_DLG: %s: min_dst_y_ttu_vblank                   = %3.2f\n",
995 			__func__,
996 			min_dst_y_ttu_vblank);
997 	dml_print(
998 			"DML_DLG: %s: t_calc_us                              = %3.2f\n",
999 			__func__,
1000 			t_calc_us);
1001 	dml_print(
1002 			"DML_DLG: %s: disp_dlg_regs->min_dst_y_next_start    = 0x%0x\n",
1003 			__func__,
1004 			disp_dlg_regs->min_dst_y_next_start);
1005 	dml_print(
1006 			"DML_DLG: %s: ref_freq_to_pix_freq                   = %3.2f\n",
1007 			__func__,
1008 			ref_freq_to_pix_freq);
1009 
1010 	// -------------------------
1011 	// Section 1.15.2.2: Prefetch, Active and TTU
1012 	// -------------------------
1013 	// Prefetch Calc
1014 	// Source
1015 	//             dcc_en              = src.dcc;
1016 	dual_plane = is_dual_plane((enum source_format_class) (src->source_format));
1017 	access_dir = (src->source_scan == dm_vert); // vp access direction: horizontal or vertical accessed
1018 						    //      bytes_per_element_l = get_bytes_per_element(source_format_class(src.source_format), 0);
1019 						    //      bytes_per_element_c = get_bytes_per_element(source_format_class(src.source_format), 1);
1020 	vp_height_l = src->viewport_height;
1021 	vp_width_l = src->viewport_width;
1022 	vp_height_c = src->viewport_height_c;
1023 	vp_width_c = src->viewport_width_c;
1024 
1025 	// Scaling
1026 	htaps_l = taps->htaps;
1027 	htaps_c = taps->htaps_c;
1028 	hratio_l = scl->hscl_ratio;
1029 	hratio_c = scl->hscl_ratio_c;
1030 	vratio_l = scl->vscl_ratio;
1031 	vratio_c = scl->vscl_ratio_c;
1032 	scl_enable = scl->scl_enable;
1033 
1034 	line_time_in_us = (htotal / pclk_freq_in_mhz);
1035 	swath_width_ub_l = rq_dlg_param->rq_l.swath_width_ub;
1036 	dpte_groups_per_row_ub_l = rq_dlg_param->rq_l.dpte_groups_per_row_ub;
1037 	swath_width_ub_c = rq_dlg_param->rq_c.swath_width_ub;
1038 	dpte_groups_per_row_ub_c = rq_dlg_param->rq_c.dpte_groups_per_row_ub;
1039 
1040 	meta_chunks_per_row_ub_l = rq_dlg_param->rq_l.meta_chunks_per_row_ub;
1041 	meta_chunks_per_row_ub_c = rq_dlg_param->rq_c.meta_chunks_per_row_ub;
1042 	vupdate_offset = dst->vupdate_offset;
1043 	vupdate_width = dst->vupdate_width;
1044 	vready_offset = dst->vready_offset;
1045 
1046 	dppclk_delay_subtotal = mode_lib->ip.dppclk_delay_subtotal;
1047 	dispclk_delay_subtotal = mode_lib->ip.dispclk_delay_subtotal;
1048 
1049 	if (scl_enable)
1050 		dppclk_delay_subtotal += mode_lib->ip.dppclk_delay_scl;
1051 	else
1052 		dppclk_delay_subtotal += mode_lib->ip.dppclk_delay_scl_lb_only;
1053 
1054 	dppclk_delay_subtotal += mode_lib->ip.dppclk_delay_cnvc_formatter
1055 			+ src->num_cursors * mode_lib->ip.dppclk_delay_cnvc_cursor;
1056 
1057 	if (dout->dsc_enable) {
1058 		double dsc_delay = get_dsc_delay(mode_lib, e2e_pipe_param, num_pipes, pipe_idx);
1059 
1060 		dispclk_delay_subtotal += dsc_delay;
1061 	}
1062 
1063 	pixel_rate_delay_subtotal = dppclk_delay_subtotal * pclk_freq_in_mhz / dppclk_freq_in_mhz
1064 			+ dispclk_delay_subtotal * pclk_freq_in_mhz / dispclk_freq_in_mhz;
1065 
1066 	vstartup_start = dst->vstartup_start;
1067 	if (interlaced) {
1068 		if (vstartup_start / 2.0
1069 				- (double) (vready_offset + vupdate_width + vupdate_offset) / htotal
1070 				<= vblank_end / 2.0)
1071 			disp_dlg_regs->vready_after_vcount0 = 1;
1072 		else
1073 			disp_dlg_regs->vready_after_vcount0 = 0;
1074 	} else {
1075 		if (vstartup_start
1076 				- (double) (vready_offset + vupdate_width + vupdate_offset) / htotal
1077 				<= vblank_end)
1078 			disp_dlg_regs->vready_after_vcount0 = 1;
1079 		else
1080 			disp_dlg_regs->vready_after_vcount0 = 0;
1081 	}
1082 
1083 	// TODO: Where is this coming from?
1084 	if (interlaced)
1085 		vstartup_start = vstartup_start / 2;
1086 
1087 	// TODO: What if this min_vblank doesn't match the value in the dml_config_settings.cpp?
1088 	if (vstartup_start >= min_vblank) {
1089 		dml_print(
1090 				"WARNING: DML_DLG: %s: vblank_start=%d vblank_end=%d\n",
1091 				__func__,
1092 				vblank_start,
1093 				vblank_end);
1094 		dml_print(
1095 				"WARNING: DML_DLG: %s: vstartup_start=%d should be less than min_vblank=%d\n",
1096 				__func__,
1097 				vstartup_start,
1098 				min_vblank);
1099 		min_vblank = vstartup_start + 1;
1100 		dml_print(
1101 				"WARNING: DML_DLG: %s: vstartup_start=%d should be less than min_vblank=%d\n",
1102 				__func__,
1103 				vstartup_start,
1104 				min_vblank);
1105 	}
1106 
1107 	dst_x_after_scaler = get_dst_x_after_scaler(mode_lib, e2e_pipe_param, num_pipes, pipe_idx);
1108 	dst_y_after_scaler = get_dst_y_after_scaler(mode_lib, e2e_pipe_param, num_pipes, pipe_idx);
1109 
1110 	dml_print("DML_DLG: %s: htotal                                 = %d\n", __func__, htotal);
1111 	dml_print(
1112 			"DML_DLG: %s: pixel_rate_delay_subtotal              = %d\n",
1113 			__func__,
1114 			pixel_rate_delay_subtotal);
1115 	dml_print(
1116 			"DML_DLG: %s: dst_x_after_scaler                     = %d\n",
1117 			__func__,
1118 			dst_x_after_scaler);
1119 	dml_print(
1120 			"DML_DLG: %s: dst_y_after_scaler                     = %d\n",
1121 			__func__,
1122 			dst_y_after_scaler);
1123 
1124 	// Lwait
1125 	// TODO: Should this be urgent_latency_pixel_mixed_with_vm_data_us?
1126 	line_wait = mode_lib->soc.urgent_latency_pixel_data_only_us;
1127 	if (cstate_en)
1128 		line_wait = dml_max(mode_lib->soc.sr_enter_plus_exit_time_us, line_wait);
1129 	if (pstate_en)
1130 		line_wait = dml_max(
1131 				mode_lib->soc.dram_clock_change_latency_us
1132 						+ mode_lib->soc.urgent_latency_pixel_data_only_us, // TODO: Should this be urgent_latency_pixel_mixed_with_vm_data_us?
1133 				line_wait);
1134 	line_wait = line_wait / line_time_in_us;
1135 
1136 	dst_y_prefetch = get_dst_y_prefetch(mode_lib, e2e_pipe_param, num_pipes, pipe_idx);
1137 	dml_print("DML_DLG: %s: dst_y_prefetch (after rnd) = %3.2f\n", __func__, dst_y_prefetch);
1138 
1139 	dst_y_per_vm_vblank = get_dst_y_per_vm_vblank(
1140 			mode_lib,
1141 			e2e_pipe_param,
1142 			num_pipes,
1143 			pipe_idx);
1144 	dst_y_per_row_vblank = get_dst_y_per_row_vblank(
1145 			mode_lib,
1146 			e2e_pipe_param,
1147 			num_pipes,
1148 			pipe_idx);
1149 	dst_y_per_vm_flip = get_dst_y_per_vm_flip(mode_lib, e2e_pipe_param, num_pipes, pipe_idx);
1150 	dst_y_per_row_flip = get_dst_y_per_row_flip(mode_lib, e2e_pipe_param, num_pipes, pipe_idx);
1151 
1152 	max_dst_y_per_vm_vblank = 32.0;
1153 	max_dst_y_per_row_vblank = 16.0;
1154 
1155 	// magic!
1156 	if (htotal <= 75) {
1157 		min_vblank = 300;
1158 		max_dst_y_per_vm_vblank = 100.0;
1159 		max_dst_y_per_row_vblank = 100.0;
1160 	}
1161 
1162 	dml_print("DML_DLG: %s: dst_y_per_vm_flip    = %3.2f\n", __func__, dst_y_per_vm_flip);
1163 	dml_print("DML_DLG: %s: dst_y_per_row_flip   = %3.2f\n", __func__, dst_y_per_row_flip);
1164 	dml_print("DML_DLG: %s: dst_y_per_vm_vblank  = %3.2f\n", __func__, dst_y_per_vm_vblank);
1165 	dml_print("DML_DLG: %s: dst_y_per_row_vblank = %3.2f\n", __func__, dst_y_per_row_vblank);
1166 
1167 	ASSERT(dst_y_per_vm_vblank < max_dst_y_per_vm_vblank);
1168 	ASSERT(dst_y_per_row_vblank < max_dst_y_per_row_vblank);
1169 
1170 	ASSERT(dst_y_prefetch > (dst_y_per_vm_vblank + dst_y_per_row_vblank));
1171 	lsw = dst_y_prefetch - (dst_y_per_vm_vblank + dst_y_per_row_vblank);
1172 
1173 	dml_print("DML_DLG: %s: lsw = %3.2f\n", __func__, lsw);
1174 
1175 	vratio_pre_l = get_vratio_prefetch_l(mode_lib, e2e_pipe_param, num_pipes, pipe_idx);
1176 	vratio_pre_c = get_vratio_prefetch_c(mode_lib, e2e_pipe_param, num_pipes, pipe_idx);
1177 
1178 	dml_print("DML_DLG: %s: vratio_pre_l=%3.2f\n", __func__, vratio_pre_l);
1179 	dml_print("DML_DLG: %s: vratio_pre_c=%3.2f\n", __func__, vratio_pre_c);
1180 
1181 	// Active
1182 	req_per_swath_ub_l = rq_dlg_param->rq_l.req_per_swath_ub;
1183 	req_per_swath_ub_c = rq_dlg_param->rq_c.req_per_swath_ub;
1184 	meta_row_height_l = rq_dlg_param->rq_l.meta_row_height;
1185 	meta_row_height_c = rq_dlg_param->rq_c.meta_row_height;
1186 	swath_width_pixels_ub_l = 0;
1187 	swath_width_pixels_ub_c = 0;
1188 	scaler_rec_in_width_l = 0;
1189 	scaler_rec_in_width_c = 0;
1190 	dpte_row_height_l = rq_dlg_param->rq_l.dpte_row_height;
1191 	dpte_row_height_c = rq_dlg_param->rq_c.dpte_row_height;
1192 
1193 	swath_width_pixels_ub_l = swath_width_ub_l;
1194 	swath_width_pixels_ub_c = swath_width_ub_c;
1195 
1196 	if (hratio_l <= 1)
1197 		min_hratio_fact_l = 2.0;
1198 	else if (htaps_l <= 6) {
1199 		if ((hratio_l * 2.0) > 4.0)
1200 			min_hratio_fact_l = 4.0;
1201 		else
1202 			min_hratio_fact_l = hratio_l * 2.0;
1203 	} else {
1204 		if (hratio_l > 4.0)
1205 			min_hratio_fact_l = 4.0;
1206 		else
1207 			min_hratio_fact_l = hratio_l;
1208 	}
1209 
1210 	hscale_pixel_rate_l = min_hratio_fact_l * dppclk_freq_in_mhz;
1211 
1212 	if (hratio_c <= 1)
1213 		min_hratio_fact_c = 2.0;
1214 	else if (htaps_c <= 6) {
1215 		if ((hratio_c * 2.0) > 4.0)
1216 			min_hratio_fact_c = 4.0;
1217 		else
1218 			min_hratio_fact_c = hratio_c * 2.0;
1219 	} else {
1220 		if (hratio_c > 4.0)
1221 			min_hratio_fact_c = 4.0;
1222 		else
1223 			min_hratio_fact_c = hratio_c;
1224 	}
1225 
1226 	hscale_pixel_rate_c = min_hratio_fact_c * dppclk_freq_in_mhz;
1227 
1228 	refcyc_per_line_delivery_pre_l = 0.;
1229 	refcyc_per_line_delivery_pre_c = 0.;
1230 	refcyc_per_line_delivery_l = 0.;
1231 	refcyc_per_line_delivery_c = 0.;
1232 
1233 	refcyc_per_req_delivery_pre_l = 0.;
1234 	refcyc_per_req_delivery_pre_c = 0.;
1235 	refcyc_per_req_delivery_l = 0.;
1236 	refcyc_per_req_delivery_c = 0.;
1237 
1238 	full_recout_width = 0;
1239 	// In ODM
1240 	if (src->is_hsplit) {
1241 		// This "hack"  is only allowed (and valid) for MPC combine. In ODM
1242 		// combine, you MUST specify the full_recout_width...according to Oswin
1243 		if (dst->full_recout_width == 0 && !dst->odm_combine) {
1244 			dml_print(
1245 					"DML_DLG: %s: Warning: full_recout_width not set in hsplit mode\n",
1246 					__func__);
1247 			full_recout_width = dst->recout_width * 2; // assume half split for dcn1
1248 		} else
1249 			full_recout_width = dst->full_recout_width;
1250 	} else
1251 		full_recout_width = dst->recout_width;
1252 
1253 	// As of DCN2, mpc_combine and odm_combine are mutually exclusive
1254 	refcyc_per_line_delivery_pre_l = get_refcyc_per_delivery(
1255 			mode_lib,
1256 			refclk_freq_in_mhz,
1257 			pclk_freq_in_mhz,
1258 			dst->odm_combine,
1259 			full_recout_width,
1260 			dst->hactive,
1261 			vratio_pre_l,
1262 			hscale_pixel_rate_l,
1263 			swath_width_pixels_ub_l,
1264 			1); // per line
1265 
1266 	refcyc_per_line_delivery_l = get_refcyc_per_delivery(
1267 			mode_lib,
1268 			refclk_freq_in_mhz,
1269 			pclk_freq_in_mhz,
1270 			dst->odm_combine,
1271 			full_recout_width,
1272 			dst->hactive,
1273 			vratio_l,
1274 			hscale_pixel_rate_l,
1275 			swath_width_pixels_ub_l,
1276 			1); // per line
1277 
1278 	dml_print("DML_DLG: %s: full_recout_width              = %d\n", __func__, full_recout_width);
1279 	dml_print(
1280 			"DML_DLG: %s: hscale_pixel_rate_l            = %3.2f\n",
1281 			__func__,
1282 			hscale_pixel_rate_l);
1283 	dml_print(
1284 			"DML_DLG: %s: refcyc_per_line_delivery_pre_l = %3.2f\n",
1285 			__func__,
1286 			refcyc_per_line_delivery_pre_l);
1287 	dml_print(
1288 			"DML_DLG: %s: refcyc_per_line_delivery_l     = %3.2f\n",
1289 			__func__,
1290 			refcyc_per_line_delivery_l);
1291 
1292 	if (dual_plane) {
1293 		refcyc_per_line_delivery_pre_c = get_refcyc_per_delivery(
1294 				mode_lib,
1295 				refclk_freq_in_mhz,
1296 				pclk_freq_in_mhz,
1297 				dst->odm_combine,
1298 				full_recout_width,
1299 				dst->hactive,
1300 				vratio_pre_c,
1301 				hscale_pixel_rate_c,
1302 				swath_width_pixels_ub_c,
1303 				1); // per line
1304 
1305 		refcyc_per_line_delivery_c = get_refcyc_per_delivery(
1306 				mode_lib,
1307 				refclk_freq_in_mhz,
1308 				pclk_freq_in_mhz,
1309 				dst->odm_combine,
1310 				full_recout_width,
1311 				dst->hactive,
1312 				vratio_c,
1313 				hscale_pixel_rate_c,
1314 				swath_width_pixels_ub_c,
1315 				1);  // per line
1316 
1317 		dml_print(
1318 				"DML_DLG: %s: refcyc_per_line_delivery_pre_c = %3.2f\n",
1319 				__func__,
1320 				refcyc_per_line_delivery_pre_c);
1321 		dml_print(
1322 				"DML_DLG: %s: refcyc_per_line_delivery_c     = %3.2f\n",
1323 				__func__,
1324 				refcyc_per_line_delivery_c);
1325 	}
1326 
1327 	// TTU - Luma / Chroma
1328 	if (access_dir) {  // vertical access
1329 		scaler_rec_in_width_l = vp_height_l;
1330 		scaler_rec_in_width_c = vp_height_c;
1331 	} else {
1332 		scaler_rec_in_width_l = vp_width_l;
1333 		scaler_rec_in_width_c = vp_width_c;
1334 	}
1335 
1336 	refcyc_per_req_delivery_pre_l = get_refcyc_per_delivery(
1337 			mode_lib,
1338 			refclk_freq_in_mhz,
1339 			pclk_freq_in_mhz,
1340 			dst->odm_combine,
1341 			full_recout_width,
1342 			dst->hactive,
1343 			vratio_pre_l,
1344 			hscale_pixel_rate_l,
1345 			scaler_rec_in_width_l,
1346 			req_per_swath_ub_l);  // per req
1347 	refcyc_per_req_delivery_l = get_refcyc_per_delivery(
1348 			mode_lib,
1349 			refclk_freq_in_mhz,
1350 			pclk_freq_in_mhz,
1351 			dst->odm_combine,
1352 			full_recout_width,
1353 			dst->hactive,
1354 			vratio_l,
1355 			hscale_pixel_rate_l,
1356 			scaler_rec_in_width_l,
1357 			req_per_swath_ub_l);  // per req
1358 
1359 	dml_print(
1360 			"DML_DLG: %s: refcyc_per_req_delivery_pre_l = %3.2f\n",
1361 			__func__,
1362 			refcyc_per_req_delivery_pre_l);
1363 	dml_print(
1364 			"DML_DLG: %s: refcyc_per_req_delivery_l     = %3.2f\n",
1365 			__func__,
1366 			refcyc_per_req_delivery_l);
1367 
1368 	ASSERT(refcyc_per_req_delivery_pre_l < dml_pow(2, 13));
1369 	ASSERT(refcyc_per_req_delivery_l < dml_pow(2, 13));
1370 
1371 	if (dual_plane) {
1372 		refcyc_per_req_delivery_pre_c = get_refcyc_per_delivery(
1373 				mode_lib,
1374 				refclk_freq_in_mhz,
1375 				pclk_freq_in_mhz,
1376 				dst->odm_combine,
1377 				full_recout_width,
1378 				dst->hactive,
1379 				vratio_pre_c,
1380 				hscale_pixel_rate_c,
1381 				scaler_rec_in_width_c,
1382 				req_per_swath_ub_c);  // per req
1383 		refcyc_per_req_delivery_c = get_refcyc_per_delivery(
1384 				mode_lib,
1385 				refclk_freq_in_mhz,
1386 				pclk_freq_in_mhz,
1387 				dst->odm_combine,
1388 				full_recout_width,
1389 				dst->hactive,
1390 				vratio_c,
1391 				hscale_pixel_rate_c,
1392 				scaler_rec_in_width_c,
1393 				req_per_swath_ub_c);  // per req
1394 
1395 		dml_print(
1396 				"DML_DLG: %s: refcyc_per_req_delivery_pre_c = %3.2f\n",
1397 				__func__,
1398 				refcyc_per_req_delivery_pre_c);
1399 		dml_print(
1400 				"DML_DLG: %s: refcyc_per_req_delivery_c     = %3.2f\n",
1401 				__func__,
1402 				refcyc_per_req_delivery_c);
1403 
1404 		ASSERT(refcyc_per_req_delivery_pre_c < dml_pow(2, 13));
1405 		ASSERT(refcyc_per_req_delivery_c < dml_pow(2, 13));
1406 	}
1407 
1408 	// TTU - Cursor
1409 	refcyc_per_req_delivery_pre_cur0 = 0.0;
1410 	refcyc_per_req_delivery_cur0 = 0.0;
1411 	if (src->num_cursors > 0) {
1412 		calculate_ttu_cursor(
1413 				mode_lib,
1414 				&refcyc_per_req_delivery_pre_cur0,
1415 				&refcyc_per_req_delivery_cur0,
1416 				refclk_freq_in_mhz,
1417 				ref_freq_to_pix_freq,
1418 				hscale_pixel_rate_l,
1419 				scl->hscl_ratio,
1420 				vratio_pre_l,
1421 				vratio_l,
1422 				src->cur0_src_width,
1423 				(enum cursor_bpp) (src->cur0_bpp));
1424 	}
1425 
1426 	refcyc_per_req_delivery_pre_cur1 = 0.0;
1427 	refcyc_per_req_delivery_cur1 = 0.0;
1428 	if (src->num_cursors > 1) {
1429 		calculate_ttu_cursor(
1430 				mode_lib,
1431 				&refcyc_per_req_delivery_pre_cur1,
1432 				&refcyc_per_req_delivery_cur1,
1433 				refclk_freq_in_mhz,
1434 				ref_freq_to_pix_freq,
1435 				hscale_pixel_rate_l,
1436 				scl->hscl_ratio,
1437 				vratio_pre_l,
1438 				vratio_l,
1439 				src->cur1_src_width,
1440 				(enum cursor_bpp) (src->cur1_bpp));
1441 	}
1442 
1443 	// TTU - Misc
1444 	// all hard-coded
1445 
1446 	// Assignment to register structures
1447 	disp_dlg_regs->dst_y_after_scaler = dst_y_after_scaler; // in terms of line
1448 	disp_dlg_regs->refcyc_x_after_scaler = dst_x_after_scaler * ref_freq_to_pix_freq; // in terms of refclk
1449 	ASSERT(disp_dlg_regs->refcyc_x_after_scaler < (unsigned int)dml_pow(2, 13));
1450 	disp_dlg_regs->dst_y_prefetch = (unsigned int) (dst_y_prefetch * dml_pow(2, 2));
1451 	disp_dlg_regs->dst_y_per_vm_vblank = (unsigned int) (dst_y_per_vm_vblank * dml_pow(2, 2));
1452 	disp_dlg_regs->dst_y_per_row_vblank = (unsigned int) (dst_y_per_row_vblank * dml_pow(2, 2));
1453 	disp_dlg_regs->dst_y_per_vm_flip = (unsigned int) (dst_y_per_vm_flip * dml_pow(2, 2));
1454 	disp_dlg_regs->dst_y_per_row_flip = (unsigned int) (dst_y_per_row_flip * dml_pow(2, 2));
1455 
1456 	disp_dlg_regs->vratio_prefetch = (unsigned int) (vratio_pre_l * dml_pow(2, 19));
1457 	disp_dlg_regs->vratio_prefetch_c = (unsigned int) (vratio_pre_c * dml_pow(2, 19));
1458 
1459 	dml_print("DML_DLG: %s: disp_dlg_regs->dst_y_per_vm_vblank  = 0x%x\n", __func__, disp_dlg_regs->dst_y_per_vm_vblank);
1460 	dml_print("DML_DLG: %s: disp_dlg_regs->dst_y_per_row_vblank = 0x%x\n", __func__, disp_dlg_regs->dst_y_per_row_vblank);
1461 	dml_print("DML_DLG: %s: disp_dlg_regs->dst_y_per_vm_flip    = 0x%x\n", __func__, disp_dlg_regs->dst_y_per_vm_flip);
1462 	dml_print("DML_DLG: %s: disp_dlg_regs->dst_y_per_row_flip   = 0x%x\n", __func__, disp_dlg_regs->dst_y_per_row_flip);
1463 
1464 	disp_dlg_regs->refcyc_per_pte_group_vblank_l =
1465 			(unsigned int) (dst_y_per_row_vblank * (double) htotal
1466 					* ref_freq_to_pix_freq / (double) dpte_groups_per_row_ub_l);
1467 	if ((refclk_freq_in_mhz / ref_freq_to_pix_freq < 28) &&
1468 			disp_dlg_regs->refcyc_per_pte_group_vblank_l >= (unsigned int)dml_pow(2, 13))
1469 		disp_dlg_regs->refcyc_per_pte_group_vblank_l = (1 << 13) - 1;
1470 	else
1471 		ASSERT(disp_dlg_regs->refcyc_per_pte_group_vblank_l < (unsigned int)dml_pow(2, 13));
1472 
1473 	if (dual_plane) {
1474 		disp_dlg_regs->refcyc_per_pte_group_vblank_c = (unsigned int) (dst_y_per_row_vblank
1475 				* (double) htotal * ref_freq_to_pix_freq
1476 				/ (double) dpte_groups_per_row_ub_c);
1477 		if ((refclk_freq_in_mhz / ref_freq_to_pix_freq < 28) &&
1478 				disp_dlg_regs->refcyc_per_pte_group_vblank_c >= (unsigned int)dml_pow(2, 13))
1479 			disp_dlg_regs->refcyc_per_pte_group_vblank_c = (1 << 13) - 1;
1480 		else
1481 			ASSERT(disp_dlg_regs->refcyc_per_pte_group_vblank_c
1482 				< (unsigned int)dml_pow(2, 13));
1483 	}
1484 
1485 	if (src->dcc)
1486 		disp_dlg_regs->refcyc_per_meta_chunk_vblank_l =
1487 			(unsigned int) (dst_y_per_row_vblank * (double) htotal
1488 					* ref_freq_to_pix_freq / (double) meta_chunks_per_row_ub_l);
1489 	else
1490 		disp_dlg_regs->refcyc_per_meta_chunk_vblank_l = 0;
1491 	ASSERT(disp_dlg_regs->refcyc_per_meta_chunk_vblank_l < (unsigned int)dml_pow(2, 13));
1492 
1493 	disp_dlg_regs->refcyc_per_meta_chunk_vblank_c =
1494 			disp_dlg_regs->refcyc_per_meta_chunk_vblank_l; // dcc for 4:2:0 is not supported in dcn1.0.  assigned to be the same as _l for now
1495 
1496 	disp_dlg_regs->refcyc_per_pte_group_flip_l = (unsigned int) (dst_y_per_row_flip * htotal
1497 			* ref_freq_to_pix_freq) / dpte_groups_per_row_ub_l;
1498 	disp_dlg_regs->refcyc_per_meta_chunk_flip_l = (unsigned int) (dst_y_per_row_flip * htotal
1499 			* ref_freq_to_pix_freq) / meta_chunks_per_row_ub_l;
1500 
1501 	if (dual_plane) {
1502 		disp_dlg_regs->refcyc_per_pte_group_flip_c = (unsigned int) (dst_y_per_row_flip
1503 				* htotal * ref_freq_to_pix_freq) / dpte_groups_per_row_ub_c;
1504 		disp_dlg_regs->refcyc_per_meta_chunk_flip_c = (unsigned int) (dst_y_per_row_flip
1505 				* htotal * ref_freq_to_pix_freq) / meta_chunks_per_row_ub_c;
1506 	}
1507 
1508 	disp_dlg_regs->refcyc_per_vm_group_vblank   = get_refcyc_per_vm_group_vblank(mode_lib, e2e_pipe_param, num_pipes, pipe_idx) * refclk_freq_in_mhz;
1509 	disp_dlg_regs->refcyc_per_vm_group_flip     = get_refcyc_per_vm_group_flip(mode_lib, e2e_pipe_param, num_pipes, pipe_idx) * refclk_freq_in_mhz;
1510 	disp_dlg_regs->refcyc_per_vm_req_vblank     = get_refcyc_per_vm_req_vblank(mode_lib, e2e_pipe_param, num_pipes, pipe_idx) * refclk_freq_in_mhz * dml_pow(2, 10);
1511 	disp_dlg_regs->refcyc_per_vm_req_flip       = get_refcyc_per_vm_req_flip(mode_lib, e2e_pipe_param, num_pipes, pipe_idx) * refclk_freq_in_mhz * dml_pow(2, 10);
1512 
1513 	// Clamp to max for now
1514 	if (disp_dlg_regs->refcyc_per_vm_group_vblank >= (unsigned int)dml_pow(2, 23))
1515 		disp_dlg_regs->refcyc_per_vm_group_vblank = dml_pow(2, 23) - 1;
1516 
1517 	if (disp_dlg_regs->refcyc_per_vm_group_flip >= (unsigned int)dml_pow(2, 23))
1518 		disp_dlg_regs->refcyc_per_vm_group_flip = dml_pow(2, 23) - 1;
1519 
1520 	if (disp_dlg_regs->refcyc_per_vm_req_vblank >= (unsigned int)dml_pow(2, 23))
1521 		disp_dlg_regs->refcyc_per_vm_req_vblank = dml_pow(2, 23) - 1;
1522 
1523 	if (disp_dlg_regs->refcyc_per_vm_req_flip >= (unsigned int)dml_pow(2, 23))
1524 		disp_dlg_regs->refcyc_per_vm_req_flip = dml_pow(2, 23) - 1;
1525 	disp_dlg_regs->dst_y_per_pte_row_nom_l = (unsigned int) ((double) dpte_row_height_l
1526 			/ (double) vratio_l * dml_pow(2, 2));
1527 	ASSERT(disp_dlg_regs->dst_y_per_pte_row_nom_l < (unsigned int)dml_pow(2, 17));
1528 
1529 	if (dual_plane) {
1530 		disp_dlg_regs->dst_y_per_pte_row_nom_c = (unsigned int) ((double) dpte_row_height_c
1531 				/ (double) vratio_c * dml_pow(2, 2));
1532 		if (disp_dlg_regs->dst_y_per_pte_row_nom_c >= (unsigned int) dml_pow(2, 17)) {
1533 			dml_print(
1534 					"DML_DLG: %s: Warning dst_y_per_pte_row_nom_c %u larger than supported by register format U15.2 %u\n",
1535 					__func__,
1536 					disp_dlg_regs->dst_y_per_pte_row_nom_c,
1537 					(unsigned int)dml_pow(2, 17) - 1);
1538 		}
1539 	}
1540 
1541 	disp_dlg_regs->dst_y_per_meta_row_nom_l = (unsigned int) ((double) meta_row_height_l
1542 			/ (double) vratio_l * dml_pow(2, 2));
1543 	ASSERT(disp_dlg_regs->dst_y_per_meta_row_nom_l < (unsigned int)dml_pow(2, 17));
1544 
1545 	disp_dlg_regs->dst_y_per_meta_row_nom_c = disp_dlg_regs->dst_y_per_meta_row_nom_l; // TODO: dcc for 4:2:0 is not supported in dcn1.0.  assigned to be the same as _l for now
1546 
1547 	dml_print(
1548 			"DML: Trow: %fus\n",
1549 			line_time_in_us * (double)dpte_row_height_l / (double)vratio_l);
1550 
1551 	disp_dlg_regs->refcyc_per_pte_group_nom_l = (unsigned int) ((double) dpte_row_height_l
1552 			/ (double) vratio_l * (double) htotal * ref_freq_to_pix_freq
1553 			/ (double) dpte_groups_per_row_ub_l);
1554 	if (disp_dlg_regs->refcyc_per_pte_group_nom_l >= (unsigned int) dml_pow(2, 23))
1555 		disp_dlg_regs->refcyc_per_pte_group_nom_l = dml_pow(2, 23) - 1;
1556 	disp_dlg_regs->refcyc_per_meta_chunk_nom_l = (unsigned int) ((double) meta_row_height_l
1557 			/ (double) vratio_l * (double) htotal * ref_freq_to_pix_freq
1558 			/ (double) meta_chunks_per_row_ub_l);
1559 	if (disp_dlg_regs->refcyc_per_meta_chunk_nom_l >= (unsigned int) dml_pow(2, 23))
1560 		disp_dlg_regs->refcyc_per_meta_chunk_nom_l = dml_pow(2, 23) - 1;
1561 
1562 	if (dual_plane) {
1563 		disp_dlg_regs->refcyc_per_pte_group_nom_c =
1564 				(unsigned int) ((double) dpte_row_height_c / (double) vratio_c
1565 						* (double) htotal * ref_freq_to_pix_freq
1566 						/ (double) dpte_groups_per_row_ub_c);
1567 		if (disp_dlg_regs->refcyc_per_pte_group_nom_c >= (unsigned int) dml_pow(2, 23))
1568 			disp_dlg_regs->refcyc_per_pte_group_nom_c = dml_pow(2, 23) - 1;
1569 
1570 		// TODO: Is this the right calculation? Does htotal need to be halved?
1571 		disp_dlg_regs->refcyc_per_meta_chunk_nom_c =
1572 				(unsigned int) ((double) meta_row_height_c / (double) vratio_c
1573 						* (double) htotal * ref_freq_to_pix_freq
1574 						/ (double) meta_chunks_per_row_ub_c);
1575 		if (disp_dlg_regs->refcyc_per_meta_chunk_nom_c >= (unsigned int) dml_pow(2, 23))
1576 			disp_dlg_regs->refcyc_per_meta_chunk_nom_c = dml_pow(2, 23) - 1;
1577 	}
1578 
1579 	disp_dlg_regs->refcyc_per_line_delivery_pre_l = (unsigned int) dml_floor(
1580 			refcyc_per_line_delivery_pre_l, 1);
1581 	disp_dlg_regs->refcyc_per_line_delivery_l = (unsigned int) dml_floor(
1582 			refcyc_per_line_delivery_l, 1);
1583 	ASSERT(disp_dlg_regs->refcyc_per_line_delivery_pre_l < (unsigned int)dml_pow(2, 13));
1584 	ASSERT(disp_dlg_regs->refcyc_per_line_delivery_l < (unsigned int)dml_pow(2, 13));
1585 
1586 	disp_dlg_regs->refcyc_per_line_delivery_pre_c = (unsigned int) dml_floor(
1587 			refcyc_per_line_delivery_pre_c, 1);
1588 	disp_dlg_regs->refcyc_per_line_delivery_c = (unsigned int) dml_floor(
1589 			refcyc_per_line_delivery_c, 1);
1590 	ASSERT(disp_dlg_regs->refcyc_per_line_delivery_pre_c < (unsigned int)dml_pow(2, 13));
1591 	ASSERT(disp_dlg_regs->refcyc_per_line_delivery_c < (unsigned int)dml_pow(2, 13));
1592 
1593 	disp_dlg_regs->chunk_hdl_adjust_cur0 = 3;
1594 	disp_dlg_regs->dst_y_offset_cur0 = 0;
1595 	disp_dlg_regs->chunk_hdl_adjust_cur1 = 3;
1596 	disp_dlg_regs->dst_y_offset_cur1 = 0;
1597 
1598 	disp_dlg_regs->dst_y_delta_drq_limit = 0x7fff; // off
1599 
1600 	disp_ttu_regs->refcyc_per_req_delivery_pre_l = (unsigned int) (refcyc_per_req_delivery_pre_l
1601 			* dml_pow(2, 10));
1602 	disp_ttu_regs->refcyc_per_req_delivery_l = (unsigned int) (refcyc_per_req_delivery_l
1603 			* dml_pow(2, 10));
1604 	disp_ttu_regs->refcyc_per_req_delivery_pre_c = (unsigned int) (refcyc_per_req_delivery_pre_c
1605 			* dml_pow(2, 10));
1606 	disp_ttu_regs->refcyc_per_req_delivery_c = (unsigned int) (refcyc_per_req_delivery_c
1607 			* dml_pow(2, 10));
1608 	disp_ttu_regs->refcyc_per_req_delivery_pre_cur0 =
1609 			(unsigned int) (refcyc_per_req_delivery_pre_cur0 * dml_pow(2, 10));
1610 	disp_ttu_regs->refcyc_per_req_delivery_cur0 = (unsigned int) (refcyc_per_req_delivery_cur0
1611 			* dml_pow(2, 10));
1612 	disp_ttu_regs->refcyc_per_req_delivery_pre_cur1 =
1613 			(unsigned int) (refcyc_per_req_delivery_pre_cur1 * dml_pow(2, 10));
1614 	disp_ttu_regs->refcyc_per_req_delivery_cur1 = (unsigned int) (refcyc_per_req_delivery_cur1
1615 			* dml_pow(2, 10));
1616 	disp_ttu_regs->qos_level_low_wm = 0;
1617 	ASSERT(disp_ttu_regs->qos_level_low_wm < dml_pow(2, 14));
1618 	disp_ttu_regs->qos_level_high_wm = (unsigned int) (4.0 * (double) htotal
1619 			* ref_freq_to_pix_freq);
1620 	ASSERT(disp_ttu_regs->qos_level_high_wm < dml_pow(2, 14));
1621 
1622 	disp_ttu_regs->qos_level_flip = 14;
1623 	disp_ttu_regs->qos_level_fixed_l = 8;
1624 	disp_ttu_regs->qos_level_fixed_c = 8;
1625 	disp_ttu_regs->qos_level_fixed_cur0 = 8;
1626 	disp_ttu_regs->qos_ramp_disable_l = 0;
1627 	disp_ttu_regs->qos_ramp_disable_c = 0;
1628 	disp_ttu_regs->qos_ramp_disable_cur0 = 0;
1629 
1630 	disp_ttu_regs->min_ttu_vblank = min_ttu_vblank * refclk_freq_in_mhz;
1631 	ASSERT(disp_ttu_regs->min_ttu_vblank < dml_pow(2, 24));
1632 
1633 	print__ttu_regs_st(mode_lib, disp_ttu_regs);
1634 	print__dlg_regs_st(mode_lib, disp_dlg_regs);
1635 }
1636 
dml21_rq_dlg_get_dlg_reg(struct display_mode_lib * mode_lib,display_dlg_regs_st * dlg_regs,display_ttu_regs_st * ttu_regs,const display_e2e_pipe_params_st * e2e_pipe_param,const unsigned int num_pipes,const unsigned int pipe_idx,const bool cstate_en,const bool pstate_en,const bool vm_en,const bool ignore_viewport_pos,const bool immediate_flip_support)1637 void dml21_rq_dlg_get_dlg_reg(
1638 		struct display_mode_lib *mode_lib,
1639 		display_dlg_regs_st *dlg_regs,
1640 		display_ttu_regs_st *ttu_regs,
1641 		const display_e2e_pipe_params_st *e2e_pipe_param,
1642 		const unsigned int num_pipes,
1643 		const unsigned int pipe_idx,
1644 		const bool cstate_en,
1645 		const bool pstate_en,
1646 		const bool vm_en,
1647 		const bool ignore_viewport_pos,
1648 		const bool immediate_flip_support)
1649 {
1650 	display_rq_params_st rq_param = {0};
1651 	display_dlg_sys_params_st dlg_sys_param = {0};
1652 
1653 	// Get watermark and Tex.
1654 	dlg_sys_param.t_urg_wm_us = get_wm_urgent(mode_lib, e2e_pipe_param, num_pipes);
1655 	dlg_sys_param.deepsleep_dcfclk_mhz = get_clk_dcf_deepsleep(
1656 			mode_lib,
1657 			e2e_pipe_param,
1658 			num_pipes);
1659 	dlg_sys_param.t_extra_us = get_urgent_extra_latency(mode_lib, e2e_pipe_param, num_pipes);
1660 	dlg_sys_param.mem_trip_us = get_wm_memory_trip(mode_lib, e2e_pipe_param, num_pipes);
1661 	dlg_sys_param.t_mclk_wm_us = get_wm_dram_clock_change(mode_lib, e2e_pipe_param, num_pipes);
1662 	dlg_sys_param.t_sr_wm_us = get_wm_stutter_enter_exit(mode_lib, e2e_pipe_param, num_pipes);
1663 	dlg_sys_param.total_flip_bw = get_total_immediate_flip_bw(
1664 			mode_lib,
1665 			e2e_pipe_param,
1666 			num_pipes);
1667 	dlg_sys_param.total_flip_bytes = get_total_immediate_flip_bytes(
1668 			mode_lib,
1669 			e2e_pipe_param,
1670 			num_pipes);
1671 
1672 	print__dlg_sys_params_st(mode_lib, &dlg_sys_param);
1673 
1674 	// system parameter calculation done
1675 
1676 	dml_print("DML_DLG: Calculation for pipe[%d] start\n\n", pipe_idx);
1677 	dml_rq_dlg_get_rq_params(mode_lib, &rq_param, &e2e_pipe_param[pipe_idx].pipe);
1678 	dml_rq_dlg_get_dlg_params(
1679 			mode_lib,
1680 			e2e_pipe_param,
1681 			num_pipes,
1682 			pipe_idx,
1683 			dlg_regs,
1684 			ttu_regs,
1685 			&rq_param.dlg,
1686 			&dlg_sys_param,
1687 			cstate_en,
1688 			pstate_en);
1689 	dml_print("DML_DLG: Calculation for pipe[%d] end\n", pipe_idx);
1690 }
1691 
calculate_ttu_cursor(struct display_mode_lib * mode_lib,double * refcyc_per_req_delivery_pre_cur,double * refcyc_per_req_delivery_cur,double refclk_freq_in_mhz,double ref_freq_to_pix_freq,double hscale_pixel_rate_l,double hscl_ratio,double vratio_pre_l,double vratio_l,unsigned int cur_width,enum cursor_bpp cur_bpp)1692 static void calculate_ttu_cursor(
1693 		struct display_mode_lib *mode_lib,
1694 		double *refcyc_per_req_delivery_pre_cur,
1695 		double *refcyc_per_req_delivery_cur,
1696 		double refclk_freq_in_mhz,
1697 		double ref_freq_to_pix_freq,
1698 		double hscale_pixel_rate_l,
1699 		double hscl_ratio,
1700 		double vratio_pre_l,
1701 		double vratio_l,
1702 		unsigned int cur_width,
1703 		enum cursor_bpp cur_bpp)
1704 {
1705 	unsigned int cur_src_width = cur_width;
1706 	unsigned int cur_req_size = 0;
1707 	unsigned int cur_req_width = 0;
1708 	double cur_width_ub = 0.0;
1709 	double cur_req_per_width = 0.0;
1710 	double hactive_cur = 0.0;
1711 
1712 	ASSERT(cur_src_width <= 256);
1713 
1714 	*refcyc_per_req_delivery_pre_cur = 0.0;
1715 	*refcyc_per_req_delivery_cur = 0.0;
1716 	if (cur_src_width > 0) {
1717 		unsigned int cur_bit_per_pixel = 0;
1718 
1719 		if (cur_bpp == dm_cur_2bit) {
1720 			cur_req_size = 64; // byte
1721 			cur_bit_per_pixel = 2;
1722 		} else { // 32bit
1723 			cur_bit_per_pixel = 32;
1724 			if (cur_src_width >= 1 && cur_src_width <= 16)
1725 				cur_req_size = 64;
1726 			else if (cur_src_width >= 17 && cur_src_width <= 31)
1727 				cur_req_size = 128;
1728 			else
1729 				cur_req_size = 256;
1730 		}
1731 
1732 		cur_req_width = (double) cur_req_size / ((double) cur_bit_per_pixel / 8.0);
1733 		cur_width_ub = dml_ceil((double) cur_src_width / (double) cur_req_width, 1)
1734 				* (double) cur_req_width;
1735 		cur_req_per_width = cur_width_ub / (double) cur_req_width;
1736 		hactive_cur = (double) cur_src_width / hscl_ratio; // FIXME: oswin to think about what to do for cursor
1737 
1738 		if (vratio_pre_l <= 1.0) {
1739 			*refcyc_per_req_delivery_pre_cur = hactive_cur * ref_freq_to_pix_freq
1740 					/ (double) cur_req_per_width;
1741 		} else {
1742 			*refcyc_per_req_delivery_pre_cur = (double) refclk_freq_in_mhz
1743 					* (double) cur_src_width / hscale_pixel_rate_l
1744 					/ (double) cur_req_per_width;
1745 		}
1746 
1747 		ASSERT(*refcyc_per_req_delivery_pre_cur < dml_pow(2, 13));
1748 
1749 		if (vratio_l <= 1.0) {
1750 			*refcyc_per_req_delivery_cur = hactive_cur * ref_freq_to_pix_freq
1751 					/ (double) cur_req_per_width;
1752 		} else {
1753 			*refcyc_per_req_delivery_cur = (double) refclk_freq_in_mhz
1754 					* (double) cur_src_width / hscale_pixel_rate_l
1755 					/ (double) cur_req_per_width;
1756 		}
1757 
1758 		dml_print(
1759 				"DML_DLG: %s: cur_req_width                     = %d\n",
1760 				__func__,
1761 				cur_req_width);
1762 		dml_print(
1763 				"DML_DLG: %s: cur_width_ub                      = %3.2f\n",
1764 				__func__,
1765 				cur_width_ub);
1766 		dml_print(
1767 				"DML_DLG: %s: cur_req_per_width                 = %3.2f\n",
1768 				__func__,
1769 				cur_req_per_width);
1770 		dml_print(
1771 				"DML_DLG: %s: hactive_cur                       = %3.2f\n",
1772 				__func__,
1773 				hactive_cur);
1774 		dml_print(
1775 				"DML_DLG: %s: refcyc_per_req_delivery_pre_cur   = %3.2f\n",
1776 				__func__,
1777 				*refcyc_per_req_delivery_pre_cur);
1778 		dml_print(
1779 				"DML_DLG: %s: refcyc_per_req_delivery_cur       = %3.2f\n",
1780 				__func__,
1781 				*refcyc_per_req_delivery_cur);
1782 
1783 		ASSERT(*refcyc_per_req_delivery_cur < dml_pow(2, 13));
1784 	}
1785 }
1786 
1787