1 | /* |
2 | * Copyright 2020 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 | #include "../display_mode_lib.h" |
27 | #include "../display_mode_vba.h" |
28 | #include "../dml_inline_defs.h" |
29 | #include "display_rq_dlg_calc_30.h" |
30 | #include "display_mode_vba_30.h" |
31 | |
32 | static bool is_dual_plane(enum source_format_class source_format) |
33 | { |
34 | bool ret_val = 0; |
35 | |
36 | if ((source_format == dm_420_12) || (source_format == dm_420_8) || (source_format == dm_420_10) || (source_format == dm_rgbe_alpha)) |
37 | ret_val = 1; |
38 | |
39 | return ret_val; |
40 | } |
41 | |
42 | static double get_refcyc_per_delivery(struct display_mode_lib *mode_lib, |
43 | double refclk_freq_in_mhz, |
44 | double pclk_freq_in_mhz, |
45 | unsigned int odm_combine, |
46 | unsigned int recout_width, |
47 | unsigned int hactive, |
48 | double vratio, |
49 | double hscale_pixel_rate, |
50 | unsigned int delivery_width, |
51 | unsigned int req_per_swath_ub) |
52 | { |
53 | double refcyc_per_delivery = 0.0; |
54 | |
55 | if (vratio <= 1.0) { |
56 | if (odm_combine) |
57 | refcyc_per_delivery = (double)refclk_freq_in_mhz * (double)((unsigned int)odm_combine*2) |
58 | * dml_min(a: (double)recout_width, b: (double)hactive / ((unsigned int)odm_combine*2)) |
59 | / pclk_freq_in_mhz / (double)req_per_swath_ub; |
60 | else |
61 | refcyc_per_delivery = (double)refclk_freq_in_mhz * (double)recout_width |
62 | / pclk_freq_in_mhz / (double)req_per_swath_ub; |
63 | } else { |
64 | refcyc_per_delivery = (double)refclk_freq_in_mhz * (double)delivery_width |
65 | / (double)hscale_pixel_rate / (double)req_per_swath_ub; |
66 | } |
67 | |
68 | dml_print("DML_DLG: %s: refclk_freq_in_mhz = %3.2f\n" , __func__, refclk_freq_in_mhz); |
69 | dml_print("DML_DLG: %s: pclk_freq_in_mhz = %3.2f\n" , __func__, pclk_freq_in_mhz); |
70 | dml_print("DML_DLG: %s: recout_width = %d\n" , __func__, recout_width); |
71 | dml_print("DML_DLG: %s: vratio = %3.2f\n" , __func__, vratio); |
72 | dml_print("DML_DLG: %s: req_per_swath_ub = %d\n" , __func__, req_per_swath_ub); |
73 | dml_print("DML_DLG: %s: refcyc_per_delivery= %3.2f\n" , __func__, refcyc_per_delivery); |
74 | |
75 | return refcyc_per_delivery; |
76 | |
77 | } |
78 | |
79 | static unsigned int get_blk_size_bytes(const enum source_macro_tile_size tile_size) |
80 | { |
81 | if (tile_size == dm_256k_tile) |
82 | return (256 * 1024); |
83 | else if (tile_size == dm_64k_tile) |
84 | return (64 * 1024); |
85 | else |
86 | return (4 * 1024); |
87 | } |
88 | |
89 | static void (struct display_mode_lib *mode_lib, |
90 | display_data_rq_regs_st *rq_regs, |
91 | const display_data_rq_sizing_params_st *rq_sizing) |
92 | { |
93 | dml_print("DML_DLG: %s: rq_sizing param\n" , __func__); |
94 | print__data_rq_sizing_params_st(mode_lib, rq_sizing); |
95 | |
96 | rq_regs->chunk_size = dml_log2(x: rq_sizing->chunk_bytes) - 10; |
97 | |
98 | if (rq_sizing->min_chunk_bytes == 0) |
99 | rq_regs->min_chunk_size = 0; |
100 | else |
101 | rq_regs->min_chunk_size = dml_log2(x: rq_sizing->min_chunk_bytes) - 8 + 1; |
102 | |
103 | rq_regs->meta_chunk_size = dml_log2(x: rq_sizing->meta_chunk_bytes) - 10; |
104 | if (rq_sizing->min_meta_chunk_bytes == 0) |
105 | rq_regs->min_meta_chunk_size = 0; |
106 | else |
107 | rq_regs->min_meta_chunk_size = dml_log2(x: rq_sizing->min_meta_chunk_bytes) - 6 + 1; |
108 | |
109 | rq_regs->dpte_group_size = dml_log2(x: rq_sizing->dpte_group_bytes) - 6; |
110 | rq_regs->mpte_group_size = dml_log2(x: rq_sizing->mpte_group_bytes) - 6; |
111 | } |
112 | |
113 | static void (struct display_mode_lib *mode_lib, |
114 | display_rq_regs_st *rq_regs, |
115 | const display_rq_params_st *rq_param) |
116 | { |
117 | unsigned int detile_buf_size_in_bytes = mode_lib->ip.det_buffer_size_kbytes * 1024; |
118 | unsigned int detile_buf_plane1_addr = 0; |
119 | |
120 | extract_rq_sizing_regs(mode_lib, rq_regs: &(rq_regs->rq_regs_l), rq_sizing: &rq_param->sizing.rq_l); |
121 | |
122 | rq_regs->rq_regs_l.pte_row_height_linear = dml_floor(a: dml_log2(x: rq_param->dlg.rq_l.dpte_row_height), |
123 | granularity: 1) - 3; |
124 | |
125 | if (rq_param->yuv420) { |
126 | extract_rq_sizing_regs(mode_lib, rq_regs: &(rq_regs->rq_regs_c), rq_sizing: &rq_param->sizing.rq_c); |
127 | rq_regs->rq_regs_c.pte_row_height_linear = dml_floor(a: dml_log2(x: rq_param->dlg.rq_c.dpte_row_height), |
128 | granularity: 1) - 3; |
129 | } |
130 | |
131 | rq_regs->rq_regs_l.swath_height = dml_log2(x: rq_param->dlg.rq_l.swath_height); |
132 | rq_regs->rq_regs_c.swath_height = dml_log2(x: rq_param->dlg.rq_c.swath_height); |
133 | |
134 | // FIXME: take the max between luma, chroma chunk size? |
135 | // okay for now, as we are setting chunk_bytes to 8kb anyways |
136 | if (rq_param->sizing.rq_l.chunk_bytes >= 32 * 1024 || (rq_param->yuv420 && rq_param->sizing.rq_c.chunk_bytes >= 32 * 1024)) { //32kb |
137 | rq_regs->drq_expansion_mode = 0; |
138 | } else { |
139 | rq_regs->drq_expansion_mode = 2; |
140 | } |
141 | rq_regs->prq_expansion_mode = 1; |
142 | rq_regs->mrq_expansion_mode = 1; |
143 | rq_regs->crq_expansion_mode = 1; |
144 | |
145 | if (rq_param->yuv420) { |
146 | if ((double)rq_param->misc.rq_l.stored_swath_bytes |
147 | / (double)rq_param->misc.rq_c.stored_swath_bytes <= 1.5) { |
148 | detile_buf_plane1_addr = (detile_buf_size_in_bytes / 2.0 / 64.0); // half to chroma |
149 | } else { |
150 | detile_buf_plane1_addr = dml_round_to_multiple(num: (unsigned int)((2.0 * detile_buf_size_in_bytes) / 3.0), |
151 | multiple: 256, |
152 | up: 0) / 64.0; // 2/3 to chroma |
153 | } |
154 | } |
155 | rq_regs->plane1_base_address = detile_buf_plane1_addr; |
156 | } |
157 | |
158 | static void handle_det_buf_split(struct display_mode_lib *mode_lib, |
159 | display_rq_params_st *rq_param, |
160 | const display_pipe_source_params_st *pipe_src_param) |
161 | { |
162 | unsigned int total_swath_bytes = 0; |
163 | unsigned int swath_bytes_l = 0; |
164 | unsigned int swath_bytes_c = 0; |
165 | unsigned int full_swath_bytes_packed_l = 0; |
166 | unsigned int full_swath_bytes_packed_c = 0; |
167 | bool req128_l = false; |
168 | bool req128_c = false; |
169 | bool surf_linear = (pipe_src_param->sw_mode == dm_sw_linear); |
170 | bool surf_vert = (pipe_src_param->source_scan == dm_vert); |
171 | unsigned int log2_swath_height_l = 0; |
172 | unsigned int log2_swath_height_c = 0; |
173 | unsigned int detile_buf_size_in_bytes = mode_lib->ip.det_buffer_size_kbytes * 1024; |
174 | |
175 | full_swath_bytes_packed_l = rq_param->misc.rq_l.full_swath_bytes; |
176 | full_swath_bytes_packed_c = rq_param->misc.rq_c.full_swath_bytes; |
177 | |
178 | if (rq_param->yuv420_10bpc) { |
179 | full_swath_bytes_packed_l = dml_round_to_multiple(num: rq_param->misc.rq_l.full_swath_bytes * 2.0 / 3.0, |
180 | multiple: 256, |
181 | up: 1) + 256; |
182 | full_swath_bytes_packed_c = dml_round_to_multiple(num: rq_param->misc.rq_c.full_swath_bytes * 2.0 / 3.0, |
183 | multiple: 256, |
184 | up: 1) + 256; |
185 | } |
186 | |
187 | if (rq_param->yuv420) |
188 | total_swath_bytes = 2 * full_swath_bytes_packed_l + 2 * full_swath_bytes_packed_c; |
189 | else |
190 | total_swath_bytes = 2 * full_swath_bytes_packed_l; |
191 | |
192 | if (total_swath_bytes <= detile_buf_size_in_bytes) { //full 256b request |
193 | req128_l = false; |
194 | req128_c = false; |
195 | swath_bytes_l = full_swath_bytes_packed_l; |
196 | swath_bytes_c = full_swath_bytes_packed_c; |
197 | } else if (!rq_param->yuv420) { |
198 | req128_l = true; |
199 | req128_c = false; |
200 | swath_bytes_c = full_swath_bytes_packed_c; |
201 | swath_bytes_l = full_swath_bytes_packed_l / 2; |
202 | } else if ((double)full_swath_bytes_packed_l / (double)full_swath_bytes_packed_c < 1.5) { |
203 | req128_l = false; |
204 | req128_c = true; |
205 | swath_bytes_l = full_swath_bytes_packed_l; |
206 | swath_bytes_c = full_swath_bytes_packed_c / 2; |
207 | |
208 | total_swath_bytes = 2 * swath_bytes_l + 2 * swath_bytes_c; |
209 | |
210 | if (total_swath_bytes > detile_buf_size_in_bytes) { |
211 | req128_l = true; |
212 | swath_bytes_l = full_swath_bytes_packed_l / 2; |
213 | } |
214 | } else { |
215 | req128_l = true; |
216 | req128_c = false; |
217 | swath_bytes_l = full_swath_bytes_packed_l/2; |
218 | swath_bytes_c = full_swath_bytes_packed_c; |
219 | |
220 | total_swath_bytes = 2 * swath_bytes_l + 2 * swath_bytes_c; |
221 | |
222 | if (total_swath_bytes > detile_buf_size_in_bytes) { |
223 | req128_c = true; |
224 | swath_bytes_c = full_swath_bytes_packed_c/2; |
225 | } |
226 | } |
227 | |
228 | if (rq_param->yuv420) |
229 | total_swath_bytes = 2 * swath_bytes_l + 2 * swath_bytes_c; |
230 | else |
231 | total_swath_bytes = 2 * swath_bytes_l; |
232 | |
233 | rq_param->misc.rq_l.stored_swath_bytes = swath_bytes_l; |
234 | rq_param->misc.rq_c.stored_swath_bytes = swath_bytes_c; |
235 | |
236 | if (surf_linear) { |
237 | log2_swath_height_l = 0; |
238 | log2_swath_height_c = 0; |
239 | } else { |
240 | unsigned int swath_height_l; |
241 | unsigned int swath_height_c; |
242 | |
243 | if (!surf_vert) { |
244 | swath_height_l = rq_param->misc.rq_l.blk256_height; |
245 | swath_height_c = rq_param->misc.rq_c.blk256_height; |
246 | } else { |
247 | swath_height_l = rq_param->misc.rq_l.blk256_width; |
248 | swath_height_c = rq_param->misc.rq_c.blk256_width; |
249 | } |
250 | |
251 | if (swath_height_l > 0) |
252 | log2_swath_height_l = dml_log2(x: swath_height_l); |
253 | |
254 | if (req128_l && log2_swath_height_l > 0) |
255 | log2_swath_height_l -= 1; |
256 | |
257 | if (swath_height_c > 0) |
258 | log2_swath_height_c = dml_log2(x: swath_height_c); |
259 | |
260 | if (req128_c && log2_swath_height_c > 0) |
261 | log2_swath_height_c -= 1; |
262 | } |
263 | |
264 | rq_param->dlg.rq_l.swath_height = 1 << log2_swath_height_l; |
265 | rq_param->dlg.rq_c.swath_height = 1 << log2_swath_height_c; |
266 | |
267 | dml_print("DML_DLG: %s: req128_l = %0d\n" , __func__, req128_l); |
268 | dml_print("DML_DLG: %s: req128_c = %0d\n" , __func__, req128_c); |
269 | dml_print("DML_DLG: %s: full_swath_bytes_packed_l = %0d\n" , |
270 | __func__, |
271 | full_swath_bytes_packed_l); |
272 | dml_print("DML_DLG: %s: full_swath_bytes_packed_c = %0d\n" , |
273 | __func__, |
274 | full_swath_bytes_packed_c); |
275 | } |
276 | |
277 | static void get_meta_and_pte_attr(struct display_mode_lib *mode_lib, |
278 | display_data_rq_dlg_params_st *rq_dlg_param, |
279 | display_data_rq_misc_params_st *rq_misc_param, |
280 | display_data_rq_sizing_params_st *rq_sizing_param, |
281 | unsigned int vp_width, |
282 | unsigned int vp_height, |
283 | unsigned int data_pitch, |
284 | unsigned int meta_pitch, |
285 | unsigned int source_format, |
286 | unsigned int tiling, |
287 | unsigned int macro_tile_size, |
288 | unsigned int source_scan, |
289 | unsigned int hostvm_enable, |
290 | unsigned int is_chroma, |
291 | unsigned int surface_height) |
292 | { |
293 | bool surf_linear = (tiling == dm_sw_linear); |
294 | bool surf_vert = (source_scan == dm_vert); |
295 | |
296 | unsigned int bytes_per_element = 0; |
297 | unsigned int bytes_per_element_y = 0; |
298 | unsigned int bytes_per_element_c = 0; |
299 | |
300 | unsigned int blk256_width = 0; |
301 | unsigned int blk256_height = 0; |
302 | |
303 | unsigned int blk256_width_y = 0; |
304 | unsigned int blk256_height_y = 0; |
305 | unsigned int blk256_width_c = 0; |
306 | unsigned int blk256_height_c = 0; |
307 | unsigned int log2_bytes_per_element = 0; |
308 | unsigned int log2_blk256_width = 0; |
309 | unsigned int log2_blk256_height = 0; |
310 | unsigned int blk_bytes = 0; |
311 | unsigned int log2_blk_bytes = 0; |
312 | unsigned int log2_blk_height = 0; |
313 | unsigned int log2_blk_width = 0; |
314 | unsigned int log2_meta_req_bytes = 0; |
315 | unsigned int log2_meta_req_height = 0; |
316 | unsigned int log2_meta_req_width = 0; |
317 | unsigned int meta_req_width = 0; |
318 | unsigned int meta_req_height = 0; |
319 | unsigned int log2_meta_row_height = 0; |
320 | unsigned int meta_row_width_ub = 0; |
321 | unsigned int log2_meta_chunk_bytes = 0; |
322 | unsigned int log2_meta_chunk_height = 0; |
323 | |
324 | //full sized meta chunk width in unit of data elements |
325 | unsigned int log2_meta_chunk_width = 0; |
326 | unsigned int log2_min_meta_chunk_bytes = 0; |
327 | unsigned int min_meta_chunk_width = 0; |
328 | unsigned int meta_chunk_width = 0; |
329 | unsigned int meta_chunk_per_row_int = 0; |
330 | unsigned int meta_row_remainder = 0; |
331 | unsigned int meta_chunk_threshold = 0; |
332 | unsigned int meta_blk_bytes = 0; |
333 | unsigned int meta_blk_height = 0; |
334 | unsigned int meta_blk_width = 0; |
335 | unsigned int meta_surface_bytes = 0; |
336 | unsigned int vmpg_bytes = 0; |
337 | unsigned int meta_pte_req_per_frame_ub = 0; |
338 | unsigned int meta_pte_bytes_per_frame_ub = 0; |
339 | const unsigned int log2_vmpg_bytes = dml_log2(x: mode_lib->soc.gpuvm_min_page_size_bytes); |
340 | const bool dual_plane_en = is_dual_plane(source_format: (enum source_format_class)(source_format)); |
341 | const unsigned int dpte_buf_in_pte_reqs = dual_plane_en ? |
342 | (is_chroma ? mode_lib->ip.dpte_buffer_size_in_pte_reqs_chroma : mode_lib->ip.dpte_buffer_size_in_pte_reqs_luma) |
343 | : (mode_lib->ip.dpte_buffer_size_in_pte_reqs_luma + mode_lib->ip.dpte_buffer_size_in_pte_reqs_chroma); |
344 | |
345 | unsigned int log2_vmpg_height = 0; |
346 | unsigned int log2_vmpg_width = 0; |
347 | unsigned int log2_dpte_req_height_ptes = 0; |
348 | unsigned int log2_dpte_req_height = 0; |
349 | unsigned int log2_dpte_req_width = 0; |
350 | unsigned int log2_dpte_row_height_linear = 0; |
351 | unsigned int log2_dpte_row_height = 0; |
352 | unsigned int log2_dpte_group_width = 0; |
353 | unsigned int dpte_row_width_ub = 0; |
354 | unsigned int dpte_req_height = 0; |
355 | unsigned int dpte_req_width = 0; |
356 | unsigned int dpte_group_width = 0; |
357 | unsigned int log2_dpte_group_bytes = 0; |
358 | unsigned int log2_dpte_group_length = 0; |
359 | double byte_per_pixel_det_y = 0; |
360 | double byte_per_pixel_det_c = 0; |
361 | |
362 | dml30_CalculateBytePerPixelAnd256BBlockSizes(SourcePixelFormat: (enum source_format_class)(source_format), |
363 | SurfaceTiling: (enum dm_swizzle_mode)(tiling), |
364 | BytePerPixelY: &bytes_per_element_y, |
365 | BytePerPixelC: &bytes_per_element_c, |
366 | BytePerPixelDETY: &byte_per_pixel_det_y, |
367 | BytePerPixelDETC: &byte_per_pixel_det_c, |
368 | BlockHeight256BytesY: &blk256_height_y, |
369 | BlockHeight256BytesC: &blk256_height_c, |
370 | BlockWidth256BytesY: &blk256_width_y, |
371 | BlockWidth256BytesC: &blk256_width_c); |
372 | |
373 | if (!is_chroma) { |
374 | blk256_width = blk256_width_y; |
375 | blk256_height = blk256_height_y; |
376 | bytes_per_element = bytes_per_element_y; |
377 | } else { |
378 | blk256_width = blk256_width_c; |
379 | blk256_height = blk256_height_c; |
380 | bytes_per_element = bytes_per_element_c; |
381 | } |
382 | |
383 | log2_bytes_per_element = dml_log2(x: bytes_per_element); |
384 | |
385 | dml_print("DML_DLG: %s: surf_linear = %d\n" , __func__, surf_linear); |
386 | dml_print("DML_DLG: %s: surf_vert = %d\n" , __func__, surf_vert); |
387 | dml_print("DML_DLG: %s: blk256_width = %d\n" , __func__, blk256_width); |
388 | dml_print("DML_DLG: %s: blk256_height = %d\n" , __func__, blk256_height); |
389 | |
390 | log2_blk256_width = dml_log2(x: (double)blk256_width); |
391 | log2_blk256_height = dml_log2(x: (double)blk256_height); |
392 | blk_bytes = surf_linear ? |
393 | 256 : get_blk_size_bytes(tile_size: (enum source_macro_tile_size) macro_tile_size); |
394 | log2_blk_bytes = dml_log2(x: (double)blk_bytes); |
395 | log2_blk_height = 0; |
396 | log2_blk_width = 0; |
397 | |
398 | // remember log rule |
399 | // "+" in log is multiply |
400 | // "-" in log is divide |
401 | // "/2" is like square root |
402 | // blk is vertical biased |
403 | if (tiling != dm_sw_linear) |
404 | log2_blk_height = log2_blk256_height |
405 | + dml_ceil(a: (double)(log2_blk_bytes - 8) / 2.0, granularity: 1); |
406 | else |
407 | log2_blk_height = 0; // blk height of 1 |
408 | |
409 | log2_blk_width = log2_blk_bytes - log2_bytes_per_element - log2_blk_height; |
410 | |
411 | if (!surf_vert) { |
412 | int unsigned temp = 0; |
413 | |
414 | temp = dml_round_to_multiple(num: vp_width - 1, multiple: blk256_width, up: 1) + blk256_width; |
415 | if (data_pitch < blk256_width) { |
416 | dml_print("WARNING: DML_DLG: %s: swath_size calculation ignoring data_pitch=%u < blk256_width=%u\n" , __func__, data_pitch, blk256_width); |
417 | } else { |
418 | if (temp > data_pitch) { |
419 | if (data_pitch >= vp_width) |
420 | temp = data_pitch; |
421 | else |
422 | dml_print("WARNING: DML_DLG: %s: swath_size calculation ignoring data_pitch=%u < vp_width=%u\n" , __func__, data_pitch, vp_width); |
423 | } |
424 | } |
425 | rq_dlg_param->swath_width_ub = temp; |
426 | rq_dlg_param->req_per_swath_ub = temp >> log2_blk256_width; |
427 | } else { |
428 | int unsigned temp = 0; |
429 | |
430 | temp = dml_round_to_multiple(num: vp_height - 1, multiple: blk256_height, up: 1) + blk256_height; |
431 | if (surface_height < blk256_height) { |
432 | dml_print("WARNING: DML_DLG: %s swath_size calculation ignored surface_height=%u < blk256_height=%u\n" , __func__, surface_height, blk256_height); |
433 | } else { |
434 | if (temp > surface_height) { |
435 | if (surface_height >= vp_height) |
436 | temp = surface_height; |
437 | else |
438 | dml_print("WARNING: DML_DLG: %s swath_size calculation ignored surface_height=%u < vp_height=%u\n" , __func__, surface_height, vp_height); |
439 | } |
440 | } |
441 | rq_dlg_param->swath_width_ub = temp; |
442 | rq_dlg_param->req_per_swath_ub = temp >> log2_blk256_height; |
443 | } |
444 | |
445 | if (!surf_vert) |
446 | rq_misc_param->full_swath_bytes = rq_dlg_param->swath_width_ub * blk256_height |
447 | * bytes_per_element; |
448 | else |
449 | rq_misc_param->full_swath_bytes = rq_dlg_param->swath_width_ub * blk256_width |
450 | * bytes_per_element; |
451 | |
452 | rq_misc_param->blk256_height = blk256_height; |
453 | rq_misc_param->blk256_width = blk256_width; |
454 | |
455 | // ------- |
456 | // meta |
457 | // ------- |
458 | log2_meta_req_bytes = 6; // meta request is 64b and is 8x8byte meta element |
459 | |
460 | // each 64b meta request for dcn is 8x8 meta elements and |
461 | // a meta element covers one 256b block of the data surface. |
462 | log2_meta_req_height = log2_blk256_height + 3; // meta req is 8x8 byte, each byte represent 1 blk256 |
463 | log2_meta_req_width = log2_meta_req_bytes + 8 - log2_bytes_per_element |
464 | - log2_meta_req_height; |
465 | meta_req_width = 1 << log2_meta_req_width; |
466 | meta_req_height = 1 << log2_meta_req_height; |
467 | log2_meta_row_height = 0; |
468 | meta_row_width_ub = 0; |
469 | |
470 | // the dimensions of a meta row are meta_row_width x meta_row_height in elements. |
471 | // calculate upper bound of the meta_row_width |
472 | if (!surf_vert) { |
473 | log2_meta_row_height = log2_meta_req_height; |
474 | meta_row_width_ub = dml_round_to_multiple(num: vp_width - 1, multiple: meta_req_width, up: 1) |
475 | + meta_req_width; |
476 | rq_dlg_param->meta_req_per_row_ub = meta_row_width_ub / meta_req_width; |
477 | } else { |
478 | log2_meta_row_height = log2_meta_req_width; |
479 | meta_row_width_ub = dml_round_to_multiple(num: vp_height - 1, multiple: meta_req_height, up: 1) |
480 | + meta_req_height; |
481 | rq_dlg_param->meta_req_per_row_ub = meta_row_width_ub / meta_req_height; |
482 | } |
483 | rq_dlg_param->meta_bytes_per_row_ub = rq_dlg_param->meta_req_per_row_ub * 64; |
484 | |
485 | rq_dlg_param->meta_row_height = 1 << log2_meta_row_height; |
486 | |
487 | log2_meta_chunk_bytes = dml_log2(x: rq_sizing_param->meta_chunk_bytes); |
488 | log2_meta_chunk_height = log2_meta_row_height; |
489 | |
490 | //full sized meta chunk width in unit of data elements |
491 | log2_meta_chunk_width = log2_meta_chunk_bytes + 8 - log2_bytes_per_element |
492 | - log2_meta_chunk_height; |
493 | log2_min_meta_chunk_bytes = dml_log2(x: rq_sizing_param->min_meta_chunk_bytes); |
494 | min_meta_chunk_width = 1 |
495 | << (log2_min_meta_chunk_bytes + 8 - log2_bytes_per_element |
496 | - log2_meta_chunk_height); |
497 | meta_chunk_width = 1 << log2_meta_chunk_width; |
498 | meta_chunk_per_row_int = (unsigned int)(meta_row_width_ub / meta_chunk_width); |
499 | meta_row_remainder = meta_row_width_ub % meta_chunk_width; |
500 | meta_chunk_threshold = 0; |
501 | meta_blk_bytes = 4096; |
502 | meta_blk_height = blk256_height * 64; |
503 | meta_blk_width = meta_blk_bytes * 256 / bytes_per_element / meta_blk_height; |
504 | meta_surface_bytes = meta_pitch |
505 | * (dml_round_to_multiple(num: vp_height - 1, multiple: meta_blk_height, up: 1) + meta_blk_height) |
506 | * bytes_per_element / 256; |
507 | vmpg_bytes = mode_lib->soc.gpuvm_min_page_size_bytes; |
508 | meta_pte_req_per_frame_ub = (dml_round_to_multiple(num: meta_surface_bytes - vmpg_bytes, |
509 | multiple: 8 * vmpg_bytes, |
510 | up: 1) + 8 * vmpg_bytes) / (8 * vmpg_bytes); |
511 | meta_pte_bytes_per_frame_ub = meta_pte_req_per_frame_ub * 64; //64B mpte request |
512 | rq_dlg_param->meta_pte_bytes_per_frame_ub = meta_pte_bytes_per_frame_ub; |
513 | |
514 | dml_print("DML_DLG: %s: meta_blk_height = %d\n" , __func__, meta_blk_height); |
515 | dml_print("DML_DLG: %s: meta_blk_width = %d\n" , __func__, meta_blk_width); |
516 | dml_print("DML_DLG: %s: meta_surface_bytes = %d\n" , __func__, meta_surface_bytes); |
517 | dml_print("DML_DLG: %s: meta_pte_req_per_frame_ub = %d\n" , |
518 | __func__, |
519 | meta_pte_req_per_frame_ub); |
520 | dml_print("DML_DLG: %s: meta_pte_bytes_per_frame_ub = %d\n" , |
521 | __func__, |
522 | meta_pte_bytes_per_frame_ub); |
523 | |
524 | if (!surf_vert) |
525 | meta_chunk_threshold = 2 * min_meta_chunk_width - meta_req_width; |
526 | else |
527 | meta_chunk_threshold = 2 * min_meta_chunk_width - meta_req_height; |
528 | |
529 | if (meta_row_remainder <= meta_chunk_threshold) |
530 | rq_dlg_param->meta_chunks_per_row_ub = meta_chunk_per_row_int + 1; |
531 | else |
532 | rq_dlg_param->meta_chunks_per_row_ub = meta_chunk_per_row_int + 2; |
533 | |
534 | // ------ |
535 | // dpte |
536 | // ------ |
537 | if (surf_linear) { |
538 | log2_vmpg_height = 0; // one line high |
539 | } else { |
540 | log2_vmpg_height = (log2_vmpg_bytes - 8) / 2 + log2_blk256_height; |
541 | } |
542 | log2_vmpg_width = log2_vmpg_bytes - log2_bytes_per_element - log2_vmpg_height; |
543 | |
544 | // only 3 possible shapes for dpte request in dimensions of ptes: 8x1, 4x2, 2x4. |
545 | if (surf_linear) { //one 64B PTE request returns 8 PTEs |
546 | log2_dpte_req_height_ptes = 0; |
547 | log2_dpte_req_width = log2_vmpg_width + 3; |
548 | log2_dpte_req_height = 0; |
549 | } else if (log2_blk_bytes == 12) { //4KB tile means 4kB page size |
550 | //one 64B req gives 8x1 PTEs for 4KB tile |
551 | log2_dpte_req_height_ptes = 0; |
552 | log2_dpte_req_width = log2_blk_width + 3; |
553 | log2_dpte_req_height = log2_blk_height + 0; |
554 | } else if ((log2_blk_bytes >= 16) && (log2_vmpg_bytes == 12)) { // tile block >= 64KB |
555 | //two 64B reqs of 2x4 PTEs give 16 PTEs to cover 64KB |
556 | log2_dpte_req_height_ptes = 4; |
557 | log2_dpte_req_width = log2_blk256_width + 4; // log2_64KB_width |
558 | log2_dpte_req_height = log2_blk256_height + 4; // log2_64KB_height |
559 | } else { //64KB page size and must 64KB tile block |
560 | //one 64B req gives 8x1 PTEs for 64KB tile |
561 | log2_dpte_req_height_ptes = 0; |
562 | log2_dpte_req_width = log2_blk_width + 3; |
563 | log2_dpte_req_height = log2_blk_height + 0; |
564 | } |
565 | |
566 | // The dpte request dimensions in data elements is dpte_req_width x dpte_req_height |
567 | // log2_vmpg_width is how much 1 pte represent, now calculating how much a 64b pte req represent |
568 | // That depends on the pte shape (i.e. 8x1, 4x2, 2x4) |
569 | //log2_dpte_req_height = log2_vmpg_height + log2_dpte_req_height_ptes; |
570 | //log2_dpte_req_width = log2_vmpg_width + log2_dpte_req_width_ptes; |
571 | dpte_req_height = 1 << log2_dpte_req_height; |
572 | dpte_req_width = 1 << log2_dpte_req_width; |
573 | |
574 | // calculate pitch dpte row buffer can hold |
575 | // round the result down to a power of two. |
576 | if (surf_linear) { |
577 | unsigned int dpte_row_height = 0; |
578 | |
579 | log2_dpte_row_height_linear = dml_floor(a: dml_log2(x: dpte_buf_in_pte_reqs * dpte_req_width / data_pitch), granularity: 1); |
580 | |
581 | dml_print("DML_DLG: %s: is_chroma = %d\n" , __func__, is_chroma); |
582 | dml_print("DML_DLG: %s: dpte_buf_in_pte_reqs = %d\n" , __func__, dpte_buf_in_pte_reqs); |
583 | dml_print("DML_DLG: %s: log2_dpte_row_height_linear = %d\n" , __func__, log2_dpte_row_height_linear); |
584 | |
585 | ASSERT(log2_dpte_row_height_linear >= 3); |
586 | |
587 | if (log2_dpte_row_height_linear > 7) |
588 | log2_dpte_row_height_linear = 7; |
589 | |
590 | log2_dpte_row_height = log2_dpte_row_height_linear; |
591 | // For linear, the dpte row is pitch dependent and the pte requests wrap at the pitch boundary. |
592 | // the dpte_row_width_ub is the upper bound of data_pitch*dpte_row_height in elements with this unique buffering. |
593 | dpte_row_height = 1 << log2_dpte_row_height; |
594 | dpte_row_width_ub = dml_round_to_multiple(num: data_pitch * dpte_row_height - 1, |
595 | multiple: dpte_req_width, |
596 | up: 1) + dpte_req_width; |
597 | rq_dlg_param->dpte_req_per_row_ub = dpte_row_width_ub / dpte_req_width; |
598 | } else { |
599 | // the upper bound of the dpte_row_width without dependency on viewport position follows. |
600 | // for tiled mode, row height is the same as req height and row store up to vp size upper bound |
601 | if (!surf_vert) { |
602 | log2_dpte_row_height = log2_dpte_req_height; |
603 | dpte_row_width_ub = dml_round_to_multiple(num: vp_width - 1, multiple: dpte_req_width, up: 1) |
604 | + dpte_req_width; |
605 | rq_dlg_param->dpte_req_per_row_ub = dpte_row_width_ub / dpte_req_width; |
606 | } else { |
607 | log2_dpte_row_height = |
608 | (log2_blk_width < log2_dpte_req_width) ? |
609 | log2_blk_width : log2_dpte_req_width; |
610 | dpte_row_width_ub = dml_round_to_multiple(num: vp_height - 1, multiple: dpte_req_height, up: 1) |
611 | + dpte_req_height; |
612 | rq_dlg_param->dpte_req_per_row_ub = dpte_row_width_ub / dpte_req_height; |
613 | } |
614 | } |
615 | if (log2_blk_bytes >= 16 && log2_vmpg_bytes == 12) // tile block >= 64KB |
616 | rq_dlg_param->dpte_bytes_per_row_ub = rq_dlg_param->dpte_req_per_row_ub * 128; //2*64B dpte request |
617 | else |
618 | rq_dlg_param->dpte_bytes_per_row_ub = rq_dlg_param->dpte_req_per_row_ub * 64; //64B dpte request |
619 | |
620 | rq_dlg_param->dpte_row_height = 1 << log2_dpte_row_height; |
621 | |
622 | // the dpte_group_bytes is reduced for the specific case of vertical |
623 | // access of a tile surface that has dpte request of 8x1 ptes. |
624 | if (hostvm_enable) |
625 | rq_sizing_param->dpte_group_bytes = 512; |
626 | else { |
627 | if (!surf_linear & (log2_dpte_req_height_ptes == 0) & surf_vert) //reduced, in this case, will have page fault within a group |
628 | rq_sizing_param->dpte_group_bytes = 512; |
629 | else |
630 | rq_sizing_param->dpte_group_bytes = 2048; |
631 | } |
632 | |
633 | //since pte request size is 64byte, the number of data pte requests per full sized group is as follows. |
634 | log2_dpte_group_bytes = dml_log2(x: rq_sizing_param->dpte_group_bytes); |
635 | log2_dpte_group_length = log2_dpte_group_bytes - 6; //length in 64b requests |
636 | |
637 | // full sized data pte group width in elements |
638 | if (!surf_vert) |
639 | log2_dpte_group_width = log2_dpte_group_length + log2_dpte_req_width; |
640 | else |
641 | log2_dpte_group_width = log2_dpte_group_length + log2_dpte_req_height; |
642 | |
643 | //But if the tile block >=64KB and the page size is 4KB, then each dPTE request is 2*64B |
644 | if ((log2_blk_bytes >= 16) && (log2_vmpg_bytes == 12)) // tile block >= 64KB |
645 | log2_dpte_group_width = log2_dpte_group_width - 1; |
646 | |
647 | dpte_group_width = 1 << log2_dpte_group_width; |
648 | |
649 | // since dpte groups are only aligned to dpte_req_width and not dpte_group_width, |
650 | // the upper bound for the dpte groups per row is as follows. |
651 | rq_dlg_param->dpte_groups_per_row_ub = dml_ceil(a: (double)dpte_row_width_ub / dpte_group_width, |
652 | granularity: 1); |
653 | } |
654 | |
655 | static void get_surf_rq_param(struct display_mode_lib *mode_lib, |
656 | display_data_rq_sizing_params_st *rq_sizing_param, |
657 | display_data_rq_dlg_params_st *rq_dlg_param, |
658 | display_data_rq_misc_params_st *rq_misc_param, |
659 | const display_pipe_params_st *pipe_param, |
660 | bool is_chroma, |
661 | bool is_alpha) |
662 | { |
663 | bool mode_422 = 0; |
664 | unsigned int vp_width = 0; |
665 | unsigned int vp_height = 0; |
666 | unsigned int data_pitch = 0; |
667 | unsigned int meta_pitch = 0; |
668 | unsigned int surface_height = 0; |
669 | unsigned int ppe = mode_422 ? 2 : 1; |
670 | |
671 | // FIXME check if ppe apply for both luma and chroma in 422 case |
672 | if (is_chroma | is_alpha) { |
673 | vp_width = pipe_param->src.viewport_width_c / ppe; |
674 | vp_height = pipe_param->src.viewport_height_c; |
675 | data_pitch = pipe_param->src.data_pitch_c; |
676 | meta_pitch = pipe_param->src.meta_pitch_c; |
677 | surface_height = pipe_param->src.surface_height_y / 2.0; |
678 | } else { |
679 | vp_width = pipe_param->src.viewport_width / ppe; |
680 | vp_height = pipe_param->src.viewport_height; |
681 | data_pitch = pipe_param->src.data_pitch; |
682 | meta_pitch = pipe_param->src.meta_pitch; |
683 | surface_height = pipe_param->src.surface_height_y; |
684 | } |
685 | |
686 | if (pipe_param->dest.odm_combine) { |
687 | unsigned int access_dir = 0; |
688 | unsigned int full_src_vp_width = 0; |
689 | unsigned int hactive_odm = 0; |
690 | unsigned int src_hactive_odm = 0; |
691 | access_dir = (pipe_param->src.source_scan == dm_vert); // vp access direction: horizontal or vertical accessed |
692 | hactive_odm = pipe_param->dest.hactive / ((unsigned int) pipe_param->dest.odm_combine*2); |
693 | if (is_chroma) { |
694 | full_src_vp_width = pipe_param->scale_ratio_depth.hscl_ratio_c * pipe_param->dest.full_recout_width; |
695 | src_hactive_odm = pipe_param->scale_ratio_depth.hscl_ratio_c * hactive_odm; |
696 | } else { |
697 | full_src_vp_width = pipe_param->scale_ratio_depth.hscl_ratio * pipe_param->dest.full_recout_width; |
698 | src_hactive_odm = pipe_param->scale_ratio_depth.hscl_ratio * hactive_odm; |
699 | } |
700 | |
701 | if (access_dir == 0) { |
702 | vp_width = dml_min(a: full_src_vp_width, b: src_hactive_odm); |
703 | dml_print("DML_DLG: %s: vp_width = %d\n" , __func__, vp_width); |
704 | } else { |
705 | vp_height = dml_min(a: full_src_vp_width, b: src_hactive_odm); |
706 | dml_print("DML_DLG: %s: vp_height = %d\n" , __func__, vp_height); |
707 | } |
708 | dml_print("DML_DLG: %s: full_src_vp_width = %d\n" , __func__, full_src_vp_width); |
709 | dml_print("DML_DLG: %s: hactive_odm = %d\n" , __func__, hactive_odm); |
710 | dml_print("DML_DLG: %s: src_hactive_odm = %d\n" , __func__, src_hactive_odm); |
711 | } |
712 | |
713 | rq_sizing_param->chunk_bytes = 8192; |
714 | |
715 | if (is_alpha) { |
716 | rq_sizing_param->chunk_bytes = 4096; |
717 | } |
718 | |
719 | if (rq_sizing_param->chunk_bytes == 64 * 1024) |
720 | rq_sizing_param->min_chunk_bytes = 0; |
721 | else |
722 | rq_sizing_param->min_chunk_bytes = 1024; |
723 | |
724 | rq_sizing_param->meta_chunk_bytes = 2048; |
725 | rq_sizing_param->min_meta_chunk_bytes = 256; |
726 | |
727 | if (pipe_param->src.hostvm) |
728 | rq_sizing_param->mpte_group_bytes = 512; |
729 | else |
730 | rq_sizing_param->mpte_group_bytes = 2048; |
731 | |
732 | get_meta_and_pte_attr(mode_lib, |
733 | rq_dlg_param, |
734 | rq_misc_param, |
735 | rq_sizing_param, |
736 | vp_width, |
737 | vp_height, |
738 | data_pitch, |
739 | meta_pitch, |
740 | source_format: pipe_param->src.source_format, |
741 | tiling: pipe_param->src.sw_mode, |
742 | macro_tile_size: pipe_param->src.macro_tile_size, |
743 | source_scan: pipe_param->src.source_scan, |
744 | hostvm_enable: pipe_param->src.hostvm, |
745 | is_chroma, |
746 | surface_height); |
747 | } |
748 | |
749 | static void dml_rq_dlg_get_rq_params(struct display_mode_lib *mode_lib, |
750 | display_rq_params_st *rq_param, |
751 | const display_pipe_params_st *pipe_param) |
752 | { |
753 | // get param for luma surface |
754 | rq_param->yuv420 = pipe_param->src.source_format == dm_420_8 |
755 | || pipe_param->src.source_format == dm_420_10 |
756 | || pipe_param->src.source_format == dm_rgbe_alpha |
757 | || pipe_param->src.source_format == dm_420_12; |
758 | |
759 | rq_param->yuv420_10bpc = pipe_param->src.source_format == dm_420_10; |
760 | |
761 | rq_param->rgbe_alpha = (pipe_param->src.source_format == dm_rgbe_alpha)?1:0; |
762 | |
763 | get_surf_rq_param(mode_lib, |
764 | rq_sizing_param: &(rq_param->sizing.rq_l), |
765 | rq_dlg_param: &(rq_param->dlg.rq_l), |
766 | rq_misc_param: &(rq_param->misc.rq_l), |
767 | pipe_param, |
768 | is_chroma: 0, |
769 | is_alpha: 0); |
770 | |
771 | if (is_dual_plane(source_format: (enum source_format_class)(pipe_param->src.source_format))) { |
772 | // get param for chroma surface |
773 | get_surf_rq_param(mode_lib, |
774 | rq_sizing_param: &(rq_param->sizing.rq_c), |
775 | rq_dlg_param: &(rq_param->dlg.rq_c), |
776 | rq_misc_param: &(rq_param->misc.rq_c), |
777 | pipe_param, |
778 | is_chroma: 1, |
779 | is_alpha: rq_param->rgbe_alpha); |
780 | } |
781 | |
782 | // calculate how to split the det buffer space between luma and chroma |
783 | handle_det_buf_split(mode_lib, rq_param, pipe_src_param: &pipe_param->src); |
784 | print__rq_params_st(mode_lib, rq_param); |
785 | } |
786 | |
787 | void dml30_rq_dlg_get_rq_reg(struct display_mode_lib *mode_lib, |
788 | display_rq_regs_st *rq_regs, |
789 | const display_pipe_params_st *pipe_param) |
790 | { |
791 | display_rq_params_st rq_param = { 0 }; |
792 | |
793 | memset(rq_regs, 0, sizeof(*rq_regs)); |
794 | dml_rq_dlg_get_rq_params(mode_lib, rq_param: &rq_param, pipe_param); |
795 | extract_rq_regs(mode_lib, rq_regs, rq_param: &rq_param); |
796 | |
797 | print__rq_regs_st(mode_lib, rq_regs); |
798 | } |
799 | |
800 | static void calculate_ttu_cursor(struct display_mode_lib *mode_lib, |
801 | double *refcyc_per_req_delivery_pre_cur, |
802 | double *refcyc_per_req_delivery_cur, |
803 | double refclk_freq_in_mhz, |
804 | double ref_freq_to_pix_freq, |
805 | double hscale_pixel_rate_l, |
806 | double hscl_ratio, |
807 | double vratio_pre_l, |
808 | double vratio_l, |
809 | unsigned int cur_width, |
810 | enum cursor_bpp cur_bpp) |
811 | { |
812 | unsigned int cur_src_width = cur_width; |
813 | unsigned int cur_req_size = 0; |
814 | unsigned int cur_req_width = 0; |
815 | double cur_width_ub = 0.0; |
816 | double cur_req_per_width = 0.0; |
817 | double hactive_cur = 0.0; |
818 | |
819 | ASSERT(cur_src_width <= 256); |
820 | |
821 | *refcyc_per_req_delivery_pre_cur = 0.0; |
822 | *refcyc_per_req_delivery_cur = 0.0; |
823 | if (cur_src_width > 0) { |
824 | unsigned int cur_bit_per_pixel = 0; |
825 | |
826 | if (cur_bpp == dm_cur_2bit) { |
827 | cur_req_size = 64; // byte |
828 | cur_bit_per_pixel = 2; |
829 | } else { // 32bit |
830 | cur_bit_per_pixel = 32; |
831 | if (cur_src_width >= 1 && cur_src_width <= 16) |
832 | cur_req_size = 64; |
833 | else if (cur_src_width >= 17 && cur_src_width <= 31) |
834 | cur_req_size = 128; |
835 | else |
836 | cur_req_size = 256; |
837 | } |
838 | |
839 | cur_req_width = (double)cur_req_size / ((double)cur_bit_per_pixel / 8.0); |
840 | cur_width_ub = dml_ceil(a: (double)cur_src_width / (double)cur_req_width, granularity: 1) |
841 | * (double)cur_req_width; |
842 | cur_req_per_width = cur_width_ub / (double)cur_req_width; |
843 | hactive_cur = (double)cur_src_width / hscl_ratio; // FIXME: oswin to think about what to do for cursor |
844 | |
845 | if (vratio_pre_l <= 1.0) { |
846 | *refcyc_per_req_delivery_pre_cur = hactive_cur * ref_freq_to_pix_freq |
847 | / (double)cur_req_per_width; |
848 | } else { |
849 | *refcyc_per_req_delivery_pre_cur = (double)refclk_freq_in_mhz |
850 | * (double)cur_src_width / hscale_pixel_rate_l |
851 | / (double)cur_req_per_width; |
852 | } |
853 | |
854 | ASSERT(*refcyc_per_req_delivery_pre_cur < dml_pow(2, 13)); |
855 | |
856 | if (vratio_l <= 1.0) { |
857 | *refcyc_per_req_delivery_cur = hactive_cur * ref_freq_to_pix_freq |
858 | / (double)cur_req_per_width; |
859 | } else { |
860 | *refcyc_per_req_delivery_cur = (double)refclk_freq_in_mhz |
861 | * (double)cur_src_width / hscale_pixel_rate_l |
862 | / (double)cur_req_per_width; |
863 | } |
864 | |
865 | dml_print("DML_DLG: %s: cur_req_width = %d\n" , |
866 | __func__, |
867 | cur_req_width); |
868 | dml_print("DML_DLG: %s: cur_width_ub = %3.2f\n" , |
869 | __func__, |
870 | cur_width_ub); |
871 | dml_print("DML_DLG: %s: cur_req_per_width = %3.2f\n" , |
872 | __func__, |
873 | cur_req_per_width); |
874 | dml_print("DML_DLG: %s: hactive_cur = %3.2f\n" , |
875 | __func__, |
876 | hactive_cur); |
877 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_pre_cur = %3.2f\n" , |
878 | __func__, |
879 | *refcyc_per_req_delivery_pre_cur); |
880 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_cur = %3.2f\n" , |
881 | __func__, |
882 | *refcyc_per_req_delivery_cur); |
883 | |
884 | ASSERT(*refcyc_per_req_delivery_cur < dml_pow(2, 13)); |
885 | } |
886 | } |
887 | |
888 | // Note: currently taken in as is. |
889 | // Nice to decouple code from hw register implement and extract code that are repeated for luma and chroma. |
890 | static void dml_rq_dlg_get_dlg_params(struct display_mode_lib *mode_lib, |
891 | const display_e2e_pipe_params_st *e2e_pipe_param, |
892 | const unsigned int num_pipes, |
893 | const unsigned int pipe_idx, |
894 | display_dlg_regs_st *disp_dlg_regs, |
895 | display_ttu_regs_st *disp_ttu_regs, |
896 | const display_rq_dlg_params_st rq_dlg_param, |
897 | const display_dlg_sys_params_st dlg_sys_param, |
898 | const bool cstate_en, |
899 | const bool pstate_en, |
900 | const bool vm_en, |
901 | const bool ignore_viewport_pos, |
902 | const bool immediate_flip_support) |
903 | { |
904 | const display_pipe_source_params_st *src = &e2e_pipe_param[pipe_idx].pipe.src; |
905 | const display_pipe_dest_params_st *dst = &e2e_pipe_param[pipe_idx].pipe.dest; |
906 | const display_output_params_st *dout = &e2e_pipe_param[pipe_idx].dout; |
907 | const display_clocks_and_cfg_st *clks = &e2e_pipe_param[pipe_idx].clks_cfg; |
908 | const scaler_ratio_depth_st *scl = &e2e_pipe_param[pipe_idx].pipe.scale_ratio_depth; |
909 | const scaler_taps_st *taps = &e2e_pipe_param[pipe_idx].pipe.scale_taps; |
910 | |
911 | // ------------------------- |
912 | // Section 1.15.2.1: OTG dependent Params |
913 | // ------------------------- |
914 | // Timing |
915 | unsigned int htotal = dst->htotal; |
916 | // unsigned int hblank_start = dst.hblank_start; // TODO: Remove |
917 | unsigned int hblank_end = dst->hblank_end; |
918 | unsigned int vblank_start = dst->vblank_start; |
919 | unsigned int vblank_end = dst->vblank_end; |
920 | unsigned int min_vblank = mode_lib->ip.min_vblank_lines; |
921 | |
922 | double dppclk_freq_in_mhz = clks->dppclk_mhz; |
923 | double dispclk_freq_in_mhz = clks->dispclk_mhz; |
924 | double refclk_freq_in_mhz = clks->refclk_mhz; |
925 | double pclk_freq_in_mhz = dst->pixel_rate_mhz; |
926 | bool interlaced = dst->interlaced; |
927 | |
928 | double ref_freq_to_pix_freq = refclk_freq_in_mhz / pclk_freq_in_mhz; |
929 | |
930 | double min_dcfclk_mhz = 0; |
931 | double t_calc_us = 0; |
932 | double min_ttu_vblank = 0; |
933 | |
934 | double min_dst_y_ttu_vblank = 0; |
935 | unsigned int dlg_vblank_start = 0; |
936 | bool dual_plane = false; |
937 | bool mode_422 = false; |
938 | unsigned int access_dir = 0; |
939 | unsigned int vp_height_l = 0; |
940 | unsigned int vp_width_l = 0; |
941 | unsigned int vp_height_c = 0; |
942 | unsigned int vp_width_c = 0; |
943 | |
944 | // Scaling |
945 | unsigned int htaps_l = 0; |
946 | unsigned int htaps_c = 0; |
947 | double hratio_l = 0; |
948 | double hratio_c = 0; |
949 | double vratio_l = 0; |
950 | double vratio_c = 0; |
951 | bool scl_enable = false; |
952 | |
953 | double line_time_in_us = 0; |
954 | // double vinit_l; |
955 | // double vinit_c; |
956 | // double vinit_bot_l; |
957 | // double vinit_bot_c; |
958 | |
959 | // unsigned int swath_height_l; |
960 | unsigned int swath_width_ub_l = 0; |
961 | // unsigned int dpte_bytes_per_row_ub_l; |
962 | unsigned int dpte_groups_per_row_ub_l = 0; |
963 | // unsigned int meta_pte_bytes_per_frame_ub_l; |
964 | // unsigned int meta_bytes_per_row_ub_l; |
965 | |
966 | // unsigned int swath_height_c; |
967 | unsigned int swath_width_ub_c = 0; |
968 | // unsigned int dpte_bytes_per_row_ub_c; |
969 | unsigned int dpte_groups_per_row_ub_c = 0; |
970 | |
971 | unsigned int meta_chunks_per_row_ub_l = 0; |
972 | unsigned int meta_chunks_per_row_ub_c = 0; |
973 | unsigned int vupdate_offset = 0; |
974 | unsigned int vupdate_width = 0; |
975 | unsigned int vready_offset = 0; |
976 | |
977 | unsigned int dppclk_delay_subtotal = 0; |
978 | unsigned int dispclk_delay_subtotal = 0; |
979 | unsigned int pixel_rate_delay_subtotal = 0; |
980 | |
981 | unsigned int vstartup_start = 0; |
982 | unsigned int dst_x_after_scaler = 0; |
983 | int dst_y_after_scaler = 0; |
984 | double line_wait = 0; |
985 | double dst_y_prefetch = 0; |
986 | double dst_y_per_vm_vblank = 0; |
987 | double dst_y_per_row_vblank = 0; |
988 | double dst_y_per_vm_flip = 0; |
989 | double dst_y_per_row_flip = 0; |
990 | double max_dst_y_per_vm_vblank = 0; |
991 | double max_dst_y_per_row_vblank = 0; |
992 | double lsw = 0; |
993 | double vratio_pre_l = 0; |
994 | double vratio_pre_c = 0; |
995 | unsigned int req_per_swath_ub_l = 0; |
996 | unsigned int req_per_swath_ub_c = 0; |
997 | unsigned int meta_row_height_l = 0; |
998 | unsigned int meta_row_height_c = 0; |
999 | unsigned int swath_width_pixels_ub_l = 0; |
1000 | unsigned int swath_width_pixels_ub_c = 0; |
1001 | unsigned int scaler_rec_in_width_l = 0; |
1002 | unsigned int scaler_rec_in_width_c = 0; |
1003 | unsigned int dpte_row_height_l = 0; |
1004 | unsigned int dpte_row_height_c = 0; |
1005 | double hscale_pixel_rate_l = 0; |
1006 | double hscale_pixel_rate_c = 0; |
1007 | double min_hratio_fact_l = 0; |
1008 | double min_hratio_fact_c = 0; |
1009 | double refcyc_per_line_delivery_pre_l = 0; |
1010 | double refcyc_per_line_delivery_pre_c = 0; |
1011 | double refcyc_per_line_delivery_l = 0; |
1012 | double refcyc_per_line_delivery_c = 0; |
1013 | |
1014 | double refcyc_per_req_delivery_pre_l = 0; |
1015 | double refcyc_per_req_delivery_pre_c = 0; |
1016 | double refcyc_per_req_delivery_l = 0; |
1017 | double refcyc_per_req_delivery_c = 0; |
1018 | |
1019 | unsigned int full_recout_width = 0; |
1020 | double refcyc_per_req_delivery_pre_cur0 = 0; |
1021 | double refcyc_per_req_delivery_cur0 = 0; |
1022 | double refcyc_per_req_delivery_pre_cur1 = 0; |
1023 | double refcyc_per_req_delivery_cur1 = 0; |
1024 | |
1025 | unsigned int pipe_index_in_combine[DC__NUM_PIPES__MAX] = { 0 }; |
1026 | |
1027 | memset(disp_dlg_regs, 0, sizeof(*disp_dlg_regs)); |
1028 | memset(disp_ttu_regs, 0, sizeof(*disp_ttu_regs)); |
1029 | |
1030 | dml_print("DML_DLG: %s: cstate_en = %d\n" , __func__, cstate_en); |
1031 | dml_print("DML_DLG: %s: pstate_en = %d\n" , __func__, pstate_en); |
1032 | dml_print("DML_DLG: %s: vm_en = %d\n" , __func__, vm_en); |
1033 | dml_print("DML_DLG: %s: ignore_viewport_pos = %d\n" , __func__, ignore_viewport_pos); |
1034 | dml_print("DML_DLG: %s: immediate_flip_support = %d\n" , __func__, immediate_flip_support); |
1035 | |
1036 | dml_print("DML_DLG: %s: dppclk_freq_in_mhz = %3.2f\n" , __func__, dppclk_freq_in_mhz); |
1037 | dml_print("DML_DLG: %s: dispclk_freq_in_mhz = %3.2f\n" , __func__, dispclk_freq_in_mhz); |
1038 | dml_print("DML_DLG: %s: refclk_freq_in_mhz = %3.2f\n" , __func__, refclk_freq_in_mhz); |
1039 | dml_print("DML_DLG: %s: pclk_freq_in_mhz = %3.2f\n" , __func__, pclk_freq_in_mhz); |
1040 | dml_print("DML_DLG: %s: interlaced = %d\n" , __func__, interlaced); |
1041 | ASSERT(ref_freq_to_pix_freq < 4.0); |
1042 | |
1043 | disp_dlg_regs->ref_freq_to_pix_freq = |
1044 | (unsigned int)(ref_freq_to_pix_freq * dml_pow(a: 2, exp: 19)); |
1045 | disp_dlg_regs->refcyc_per_htotal = (unsigned int)(ref_freq_to_pix_freq * (double)htotal |
1046 | * dml_pow(a: 2, exp: 8)); |
1047 | disp_dlg_regs->dlg_vblank_end = interlaced ? (vblank_end / 2) : vblank_end; // 15 bits |
1048 | |
1049 | min_dcfclk_mhz = dlg_sys_param.deepsleep_dcfclk_mhz; |
1050 | t_calc_us = get_tcalc(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1051 | min_ttu_vblank = get_min_ttu_vblank(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); |
1052 | |
1053 | min_dst_y_ttu_vblank = min_ttu_vblank * pclk_freq_in_mhz / (double)htotal; |
1054 | dlg_vblank_start = interlaced ? (vblank_start / 2) : vblank_start; |
1055 | |
1056 | disp_dlg_regs->min_dst_y_next_start = (unsigned int)(((double)dlg_vblank_start |
1057 | ) * dml_pow(a: 2, exp: 2)); |
1058 | ASSERT(disp_dlg_regs->min_dst_y_next_start < (unsigned int)dml_pow(2, 18)); |
1059 | |
1060 | dml_print("DML_DLG: %s: min_dcfclk_mhz = %3.2f\n" , |
1061 | __func__, |
1062 | min_dcfclk_mhz); |
1063 | dml_print("DML_DLG: %s: min_ttu_vblank = %3.2f\n" , |
1064 | __func__, |
1065 | min_ttu_vblank); |
1066 | dml_print("DML_DLG: %s: min_dst_y_ttu_vblank = %3.2f\n" , |
1067 | __func__, |
1068 | min_dst_y_ttu_vblank); |
1069 | dml_print("DML_DLG: %s: t_calc_us = %3.2f\n" , |
1070 | __func__, |
1071 | t_calc_us); |
1072 | dml_print("DML_DLG: %s: disp_dlg_regs->min_dst_y_next_start = 0x%0x\n" , |
1073 | __func__, |
1074 | disp_dlg_regs->min_dst_y_next_start); |
1075 | dml_print("DML_DLG: %s: ref_freq_to_pix_freq = %3.2f\n" , |
1076 | __func__, |
1077 | ref_freq_to_pix_freq); |
1078 | |
1079 | // ------------------------- |
1080 | // Section 1.15.2.2: Prefetch, Active and TTU |
1081 | // ------------------------- |
1082 | // Prefetch Calc |
1083 | // Source |
1084 | // dcc_en = src.dcc; |
1085 | dual_plane = is_dual_plane(source_format: (enum source_format_class)(src->source_format)); |
1086 | mode_422 = false; // TODO |
1087 | access_dir = (src->source_scan == dm_vert); // vp access direction: horizontal or vertical accessed |
1088 | vp_height_l = src->viewport_height; |
1089 | vp_width_l = src->viewport_width; |
1090 | vp_height_c = src->viewport_height_c; |
1091 | vp_width_c = src->viewport_width_c; |
1092 | |
1093 | // Scaling |
1094 | htaps_l = taps->htaps; |
1095 | htaps_c = taps->htaps_c; |
1096 | hratio_l = scl->hscl_ratio; |
1097 | hratio_c = scl->hscl_ratio_c; |
1098 | vratio_l = scl->vscl_ratio; |
1099 | vratio_c = scl->vscl_ratio_c; |
1100 | scl_enable = scl->scl_enable; |
1101 | |
1102 | line_time_in_us = (htotal / pclk_freq_in_mhz); |
1103 | swath_width_ub_l = rq_dlg_param.rq_l.swath_width_ub; |
1104 | dpte_groups_per_row_ub_l = rq_dlg_param.rq_l.dpte_groups_per_row_ub; |
1105 | swath_width_ub_c = rq_dlg_param.rq_c.swath_width_ub; |
1106 | dpte_groups_per_row_ub_c = rq_dlg_param.rq_c.dpte_groups_per_row_ub; |
1107 | |
1108 | meta_chunks_per_row_ub_l = rq_dlg_param.rq_l.meta_chunks_per_row_ub; |
1109 | meta_chunks_per_row_ub_c = rq_dlg_param.rq_c.meta_chunks_per_row_ub; |
1110 | vupdate_offset = dst->vupdate_offset; |
1111 | vupdate_width = dst->vupdate_width; |
1112 | vready_offset = dst->vready_offset; |
1113 | |
1114 | dppclk_delay_subtotal = mode_lib->ip.dppclk_delay_subtotal; |
1115 | dispclk_delay_subtotal = mode_lib->ip.dispclk_delay_subtotal; |
1116 | |
1117 | if (scl_enable) |
1118 | dppclk_delay_subtotal += mode_lib->ip.dppclk_delay_scl; |
1119 | else |
1120 | dppclk_delay_subtotal += mode_lib->ip.dppclk_delay_scl_lb_only; |
1121 | |
1122 | dppclk_delay_subtotal += mode_lib->ip.dppclk_delay_cnvc_formatter |
1123 | + src->num_cursors * mode_lib->ip.dppclk_delay_cnvc_cursor; |
1124 | |
1125 | if (dout->dsc_enable) { |
1126 | double dsc_delay = get_dsc_delay(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); |
1127 | |
1128 | dispclk_delay_subtotal += dsc_delay; |
1129 | } |
1130 | |
1131 | pixel_rate_delay_subtotal = dppclk_delay_subtotal * pclk_freq_in_mhz / dppclk_freq_in_mhz |
1132 | + dispclk_delay_subtotal * pclk_freq_in_mhz / dispclk_freq_in_mhz; |
1133 | |
1134 | vstartup_start = dst->vstartup_start; |
1135 | if (interlaced) { |
1136 | if (vstartup_start / 2.0 |
1137 | - (double)(vready_offset + vupdate_width + vupdate_offset) / htotal |
1138 | <= vblank_end / 2.0) |
1139 | disp_dlg_regs->vready_after_vcount0 = 1; |
1140 | else |
1141 | disp_dlg_regs->vready_after_vcount0 = 0; |
1142 | } else { |
1143 | if (vstartup_start |
1144 | - (double)(vready_offset + vupdate_width + vupdate_offset) / htotal |
1145 | <= vblank_end) |
1146 | disp_dlg_regs->vready_after_vcount0 = 1; |
1147 | else |
1148 | disp_dlg_regs->vready_after_vcount0 = 0; |
1149 | } |
1150 | |
1151 | // TODO: Where is this coming from? |
1152 | if (interlaced) |
1153 | vstartup_start = vstartup_start / 2; |
1154 | |
1155 | // TODO: What if this min_vblank doesn't match the value in the dml_config_settings.cpp? |
1156 | if (vstartup_start >= min_vblank) { |
1157 | dml_print("WARNING: DML_DLG: %s: vblank_start=%d vblank_end=%d\n" , |
1158 | __func__, |
1159 | vblank_start, |
1160 | vblank_end); |
1161 | dml_print("WARNING: DML_DLG: %s: vstartup_start=%d should be less than min_vblank=%d\n" , |
1162 | __func__, |
1163 | vstartup_start, |
1164 | min_vblank); |
1165 | min_vblank = vstartup_start + 1; |
1166 | dml_print("WARNING: DML_DLG: %s: vstartup_start=%d should be less than min_vblank=%d\n" , |
1167 | __func__, |
1168 | vstartup_start, |
1169 | min_vblank); |
1170 | } |
1171 | |
1172 | dst_x_after_scaler = get_dst_x_after_scaler(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); |
1173 | dst_y_after_scaler = get_dst_y_after_scaler(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); |
1174 | if (dst_y_after_scaler < 0) |
1175 | dst_y_after_scaler = 0; |
1176 | |
1177 | // do some adjustment on the dst_after scaler to account for odm combine mode |
1178 | dml_print("DML_DLG: %s: input dst_x_after_scaler = %d\n" , |
1179 | __func__, |
1180 | dst_x_after_scaler); |
1181 | dml_print("DML_DLG: %s: input dst_y_after_scaler = %d\n" , |
1182 | __func__, |
1183 | dst_y_after_scaler); |
1184 | |
1185 | // need to figure out which side of odm combine we're in |
1186 | if (dst->odm_combine) { |
1187 | // figure out which pipes go together |
1188 | bool visited[DC__NUM_PIPES__MAX] = { false }; |
1189 | unsigned int i, j, k; |
1190 | |
1191 | for (k = 0; k < num_pipes; ++k) { |
1192 | visited[k] = false; |
1193 | pipe_index_in_combine[k] = 0; |
1194 | } |
1195 | |
1196 | for (i = 0; i < num_pipes; i++) { |
1197 | if (e2e_pipe_param[i].pipe.src.is_hsplit && !visited[i]) { |
1198 | |
1199 | unsigned int grp = e2e_pipe_param[i].pipe.src.hsplit_grp; |
1200 | unsigned int grp_idx = 0; |
1201 | |
1202 | for (j = i; j < num_pipes; j++) { |
1203 | if (e2e_pipe_param[j].pipe.src.hsplit_grp == grp |
1204 | && e2e_pipe_param[j].pipe.src.is_hsplit && !visited[j]) { |
1205 | pipe_index_in_combine[j] = grp_idx; |
1206 | dml_print("DML_DLG: %s: pipe[%d] is in grp %d idx %d\n" , __func__, j, grp, grp_idx); |
1207 | grp_idx++; |
1208 | visited[j] = true; |
1209 | } |
1210 | } |
1211 | } |
1212 | } |
1213 | |
1214 | } |
1215 | |
1216 | if (dst->odm_combine == dm_odm_combine_mode_disabled) { |
1217 | disp_dlg_regs->refcyc_h_blank_end = (unsigned int)((double) hblank_end * ref_freq_to_pix_freq); |
1218 | } else { |
1219 | unsigned int odm_combine_factor = (dst->odm_combine == dm_odm_combine_mode_2to1 ? 2 : 4); // TODO: We should really check that 4to1 is supported before setting it to 4 |
1220 | unsigned int odm_pipe_index = pipe_index_in_combine[pipe_idx]; |
1221 | disp_dlg_regs->refcyc_h_blank_end = (unsigned int)(((double) hblank_end + odm_pipe_index * (double) dst->hactive / odm_combine_factor) * ref_freq_to_pix_freq); |
1222 | } |
1223 | ASSERT(disp_dlg_regs->refcyc_h_blank_end < (unsigned int)dml_pow(2, 13)); |
1224 | |
1225 | dml_print("DML_DLG: %s: htotal = %d\n" , __func__, htotal); |
1226 | dml_print("DML_DLG: %s: pixel_rate_delay_subtotal = %d\n" , |
1227 | __func__, |
1228 | pixel_rate_delay_subtotal); |
1229 | dml_print("DML_DLG: %s: dst_x_after_scaler[%d] = %d\n" , |
1230 | __func__, |
1231 | pipe_idx, |
1232 | dst_x_after_scaler); |
1233 | dml_print("DML_DLG: %s: dst_y_after_scaler[%d] = %d\n" , |
1234 | __func__, |
1235 | pipe_idx, |
1236 | dst_y_after_scaler); |
1237 | |
1238 | // Lwait |
1239 | // TODO: Should this be urgent_latency_pixel_mixed_with_vm_data_us? |
1240 | line_wait = mode_lib->soc.urgent_latency_pixel_data_only_us; |
1241 | if (cstate_en) |
1242 | line_wait = dml_max(a: mode_lib->soc.sr_enter_plus_exit_time_us, b: line_wait); |
1243 | if (pstate_en) |
1244 | line_wait = dml_max(a: mode_lib->soc.dram_clock_change_latency_us |
1245 | + mode_lib->soc.urgent_latency_pixel_data_only_us, // TODO: Should this be urgent_latency_pixel_mixed_with_vm_data_us? |
1246 | b: line_wait); |
1247 | line_wait = line_wait / line_time_in_us; |
1248 | |
1249 | dst_y_prefetch = get_dst_y_prefetch(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); |
1250 | dml_print("DML_DLG: %s: dst_y_prefetch (after rnd) = %3.2f\n" , __func__, dst_y_prefetch); |
1251 | |
1252 | dst_y_per_vm_vblank = get_dst_y_per_vm_vblank(mode_lib, |
1253 | pipes: e2e_pipe_param, |
1254 | num_pipes, |
1255 | which_pipe: pipe_idx); |
1256 | dst_y_per_row_vblank = get_dst_y_per_row_vblank(mode_lib, |
1257 | pipes: e2e_pipe_param, |
1258 | num_pipes, |
1259 | which_pipe: pipe_idx); |
1260 | dst_y_per_vm_flip = get_dst_y_per_vm_flip(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); |
1261 | dst_y_per_row_flip = get_dst_y_per_row_flip(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); |
1262 | |
1263 | max_dst_y_per_vm_vblank = 32.0; //U5.2 |
1264 | max_dst_y_per_row_vblank = 16.0; //U4.2 |
1265 | |
1266 | // magic! |
1267 | if (htotal <= 75) { |
1268 | min_vblank = 300; |
1269 | max_dst_y_per_vm_vblank = 100.0; |
1270 | max_dst_y_per_row_vblank = 100.0; |
1271 | } |
1272 | |
1273 | dml_print("DML_DLG: %s: dst_y_per_vm_flip = %3.2f\n" , __func__, dst_y_per_vm_flip); |
1274 | dml_print("DML_DLG: %s: dst_y_per_row_flip = %3.2f\n" , __func__, dst_y_per_row_flip); |
1275 | dml_print("DML_DLG: %s: dst_y_per_vm_vblank = %3.2f\n" , __func__, dst_y_per_vm_vblank); |
1276 | dml_print("DML_DLG: %s: dst_y_per_row_vblank = %3.2f\n" , __func__, dst_y_per_row_vblank); |
1277 | |
1278 | ASSERT(dst_y_per_vm_vblank < max_dst_y_per_vm_vblank); |
1279 | ASSERT(dst_y_per_row_vblank < max_dst_y_per_row_vblank); |
1280 | |
1281 | ASSERT(dst_y_prefetch > (dst_y_per_vm_vblank + dst_y_per_row_vblank)); |
1282 | lsw = dst_y_prefetch - (dst_y_per_vm_vblank + dst_y_per_row_vblank); |
1283 | |
1284 | dml_print("DML_DLG: %s: lsw = %3.2f\n" , __func__, lsw); |
1285 | |
1286 | vratio_pre_l = get_vratio_prefetch_l(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); |
1287 | vratio_pre_c = get_vratio_prefetch_c(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); |
1288 | |
1289 | dml_print("DML_DLG: %s: vratio_pre_l=%3.2f\n" , __func__, vratio_pre_l); |
1290 | dml_print("DML_DLG: %s: vratio_pre_c=%3.2f\n" , __func__, vratio_pre_c); |
1291 | |
1292 | // Active |
1293 | req_per_swath_ub_l = rq_dlg_param.rq_l.req_per_swath_ub; |
1294 | req_per_swath_ub_c = rq_dlg_param.rq_c.req_per_swath_ub; |
1295 | meta_row_height_l = rq_dlg_param.rq_l.meta_row_height; |
1296 | meta_row_height_c = rq_dlg_param.rq_c.meta_row_height; |
1297 | swath_width_pixels_ub_l = 0; |
1298 | swath_width_pixels_ub_c = 0; |
1299 | scaler_rec_in_width_l = 0; |
1300 | scaler_rec_in_width_c = 0; |
1301 | dpte_row_height_l = rq_dlg_param.rq_l.dpte_row_height; |
1302 | dpte_row_height_c = rq_dlg_param.rq_c.dpte_row_height; |
1303 | |
1304 | if (mode_422) { |
1305 | swath_width_pixels_ub_l = swath_width_ub_l * 2; // *2 for 2 pixel per element |
1306 | swath_width_pixels_ub_c = swath_width_ub_c * 2; |
1307 | } else { |
1308 | swath_width_pixels_ub_l = swath_width_ub_l * 1; |
1309 | swath_width_pixels_ub_c = swath_width_ub_c * 1; |
1310 | } |
1311 | |
1312 | hscale_pixel_rate_l = 0.; |
1313 | hscale_pixel_rate_c = 0.; |
1314 | min_hratio_fact_l = 1.0; |
1315 | min_hratio_fact_c = 1.0; |
1316 | |
1317 | if (hratio_l <= 1) |
1318 | min_hratio_fact_l = 2.0; |
1319 | else if (htaps_l <= 6) { |
1320 | if ((hratio_l * 2.0) > 4.0) |
1321 | min_hratio_fact_l = 4.0; |
1322 | else |
1323 | min_hratio_fact_l = hratio_l * 2.0; |
1324 | } else { |
1325 | if (hratio_l > 4.0) |
1326 | min_hratio_fact_l = 4.0; |
1327 | else |
1328 | min_hratio_fact_l = hratio_l; |
1329 | } |
1330 | |
1331 | hscale_pixel_rate_l = min_hratio_fact_l * dppclk_freq_in_mhz; |
1332 | |
1333 | if (hratio_c <= 1) |
1334 | min_hratio_fact_c = 2.0; |
1335 | else if (htaps_c <= 6) { |
1336 | if ((hratio_c * 2.0) > 4.0) |
1337 | min_hratio_fact_c = 4.0; |
1338 | else |
1339 | min_hratio_fact_c = hratio_c * 2.0; |
1340 | } else { |
1341 | if (hratio_c > 4.0) |
1342 | min_hratio_fact_c = 4.0; |
1343 | else |
1344 | min_hratio_fact_c = hratio_c; |
1345 | } |
1346 | |
1347 | hscale_pixel_rate_c = min_hratio_fact_c * dppclk_freq_in_mhz; |
1348 | |
1349 | refcyc_per_line_delivery_pre_l = 0.; |
1350 | refcyc_per_line_delivery_pre_c = 0.; |
1351 | refcyc_per_line_delivery_l = 0.; |
1352 | refcyc_per_line_delivery_c = 0.; |
1353 | |
1354 | refcyc_per_req_delivery_pre_l = 0.; |
1355 | refcyc_per_req_delivery_pre_c = 0.; |
1356 | refcyc_per_req_delivery_l = 0.; |
1357 | refcyc_per_req_delivery_c = 0.; |
1358 | |
1359 | full_recout_width = 0; |
1360 | // In ODM |
1361 | if (src->is_hsplit) { |
1362 | // This "hack" is only allowed (and valid) for MPC combine. In ODM |
1363 | // combine, you MUST specify the full_recout_width...according to Oswin |
1364 | if (dst->full_recout_width == 0 && !dst->odm_combine) { |
1365 | dml_print("DML_DLG: %s: Warning: full_recout_width not set in hsplit mode\n" , |
1366 | __func__); |
1367 | full_recout_width = dst->recout_width * 2; // assume half split for dcn1 |
1368 | } else |
1369 | full_recout_width = dst->full_recout_width; |
1370 | } else |
1371 | full_recout_width = dst->recout_width; |
1372 | |
1373 | // As of DCN2, mpc_combine and odm_combine are mutually exclusive |
1374 | refcyc_per_line_delivery_pre_l = get_refcyc_per_delivery(mode_lib, |
1375 | refclk_freq_in_mhz, |
1376 | pclk_freq_in_mhz, |
1377 | odm_combine: dst->odm_combine, |
1378 | recout_width: full_recout_width, |
1379 | hactive: dst->hactive, |
1380 | vratio: vratio_pre_l, |
1381 | hscale_pixel_rate: hscale_pixel_rate_l, |
1382 | delivery_width: swath_width_pixels_ub_l, |
1383 | req_per_swath_ub: 1); // per line |
1384 | |
1385 | refcyc_per_line_delivery_l = get_refcyc_per_delivery(mode_lib, |
1386 | refclk_freq_in_mhz, |
1387 | pclk_freq_in_mhz, |
1388 | odm_combine: dst->odm_combine, |
1389 | recout_width: full_recout_width, |
1390 | hactive: dst->hactive, |
1391 | vratio: vratio_l, |
1392 | hscale_pixel_rate: hscale_pixel_rate_l, |
1393 | delivery_width: swath_width_pixels_ub_l, |
1394 | req_per_swath_ub: 1); // per line |
1395 | |
1396 | dml_print("DML_DLG: %s: full_recout_width = %d\n" , |
1397 | __func__, |
1398 | full_recout_width); |
1399 | dml_print("DML_DLG: %s: hscale_pixel_rate_l = %3.2f\n" , |
1400 | __func__, |
1401 | hscale_pixel_rate_l); |
1402 | dml_print("DML_DLG: %s: refcyc_per_line_delivery_pre_l = %3.2f\n" , |
1403 | __func__, |
1404 | refcyc_per_line_delivery_pre_l); |
1405 | dml_print("DML_DLG: %s: refcyc_per_line_delivery_l = %3.2f\n" , |
1406 | __func__, |
1407 | refcyc_per_line_delivery_l); |
1408 | |
1409 | if (dual_plane) { |
1410 | refcyc_per_line_delivery_pre_c = get_refcyc_per_delivery(mode_lib, |
1411 | refclk_freq_in_mhz, |
1412 | pclk_freq_in_mhz, |
1413 | odm_combine: dst->odm_combine, |
1414 | recout_width: full_recout_width, |
1415 | hactive: dst->hactive, |
1416 | vratio: vratio_pre_c, |
1417 | hscale_pixel_rate: hscale_pixel_rate_c, |
1418 | delivery_width: swath_width_pixels_ub_c, |
1419 | req_per_swath_ub: 1); // per line |
1420 | |
1421 | refcyc_per_line_delivery_c = get_refcyc_per_delivery(mode_lib, |
1422 | refclk_freq_in_mhz, |
1423 | pclk_freq_in_mhz, |
1424 | odm_combine: dst->odm_combine, |
1425 | recout_width: full_recout_width, |
1426 | hactive: dst->hactive, |
1427 | vratio: vratio_c, |
1428 | hscale_pixel_rate: hscale_pixel_rate_c, |
1429 | delivery_width: swath_width_pixels_ub_c, |
1430 | req_per_swath_ub: 1); // per line |
1431 | |
1432 | dml_print("DML_DLG: %s: refcyc_per_line_delivery_pre_c = %3.2f\n" , |
1433 | __func__, |
1434 | refcyc_per_line_delivery_pre_c); |
1435 | dml_print("DML_DLG: %s: refcyc_per_line_delivery_c = %3.2f\n" , |
1436 | __func__, |
1437 | refcyc_per_line_delivery_c); |
1438 | } |
1439 | |
1440 | // smehta: this is a hack added until we get the real dml, sorry, need to make progress |
1441 | if (src->dynamic_metadata_enable && src->gpuvm) { |
1442 | unsigned int levels = mode_lib->ip.gpuvm_max_page_table_levels; |
1443 | double ref_cycles; |
1444 | |
1445 | if (src->hostvm) |
1446 | levels = levels * (mode_lib->ip.hostvm_max_page_table_levels+1); |
1447 | |
1448 | ref_cycles = (levels * mode_lib->soc.urgent_latency_vm_data_only_us) * refclk_freq_in_mhz; |
1449 | dml_print("BENyamin: dst_y_prefetch = %f %d %f %f \n" , |
1450 | ref_cycles, levels, mode_lib->soc.urgent_latency_vm_data_only_us, refclk_freq_in_mhz); |
1451 | disp_dlg_regs->refcyc_per_vm_dmdata = (unsigned int) ref_cycles; |
1452 | } |
1453 | dml_print("BENyamin: dmdta_en vm = %d %d \n" , |
1454 | src->dynamic_metadata_enable, src->vm); |
1455 | // TTU - Luma / Chroma |
1456 | if (access_dir) { // vertical access |
1457 | scaler_rec_in_width_l = vp_height_l; |
1458 | scaler_rec_in_width_c = vp_height_c; |
1459 | } else { |
1460 | scaler_rec_in_width_l = vp_width_l; |
1461 | scaler_rec_in_width_c = vp_width_c; |
1462 | } |
1463 | |
1464 | refcyc_per_req_delivery_pre_l = get_refcyc_per_delivery(mode_lib, |
1465 | refclk_freq_in_mhz, |
1466 | pclk_freq_in_mhz, |
1467 | odm_combine: dst->odm_combine, |
1468 | recout_width: full_recout_width, |
1469 | hactive: dst->hactive, |
1470 | vratio: vratio_pre_l, |
1471 | hscale_pixel_rate: hscale_pixel_rate_l, |
1472 | delivery_width: scaler_rec_in_width_l, |
1473 | req_per_swath_ub: req_per_swath_ub_l); // per req |
1474 | refcyc_per_req_delivery_l = get_refcyc_per_delivery(mode_lib, |
1475 | refclk_freq_in_mhz, |
1476 | pclk_freq_in_mhz, |
1477 | odm_combine: dst->odm_combine, |
1478 | recout_width: full_recout_width, |
1479 | hactive: dst->hactive, |
1480 | vratio: vratio_l, |
1481 | hscale_pixel_rate: hscale_pixel_rate_l, |
1482 | delivery_width: scaler_rec_in_width_l, |
1483 | req_per_swath_ub: req_per_swath_ub_l); // per req |
1484 | |
1485 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_pre_l = %3.2f\n" , |
1486 | __func__, |
1487 | refcyc_per_req_delivery_pre_l); |
1488 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_l = %3.2f\n" , |
1489 | __func__, |
1490 | refcyc_per_req_delivery_l); |
1491 | |
1492 | ASSERT(refcyc_per_req_delivery_pre_l < dml_pow(2, 13)); |
1493 | ASSERT(refcyc_per_req_delivery_l < dml_pow(2, 13)); |
1494 | |
1495 | if (dual_plane) { |
1496 | refcyc_per_req_delivery_pre_c = get_refcyc_per_delivery(mode_lib, |
1497 | refclk_freq_in_mhz, |
1498 | pclk_freq_in_mhz, |
1499 | odm_combine: dst->odm_combine, |
1500 | recout_width: full_recout_width, |
1501 | hactive: dst->hactive, |
1502 | vratio: vratio_pre_c, |
1503 | hscale_pixel_rate: hscale_pixel_rate_c, |
1504 | delivery_width: scaler_rec_in_width_c, |
1505 | req_per_swath_ub: req_per_swath_ub_c); // per req |
1506 | refcyc_per_req_delivery_c = get_refcyc_per_delivery(mode_lib, |
1507 | refclk_freq_in_mhz, |
1508 | pclk_freq_in_mhz, |
1509 | odm_combine: dst->odm_combine, |
1510 | recout_width: full_recout_width, |
1511 | hactive: dst->hactive, |
1512 | vratio: vratio_c, |
1513 | hscale_pixel_rate: hscale_pixel_rate_c, |
1514 | delivery_width: scaler_rec_in_width_c, |
1515 | req_per_swath_ub: req_per_swath_ub_c); // per req |
1516 | |
1517 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_pre_c = %3.2f\n" , |
1518 | __func__, |
1519 | refcyc_per_req_delivery_pre_c); |
1520 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_c = %3.2f\n" , |
1521 | __func__, |
1522 | refcyc_per_req_delivery_c); |
1523 | |
1524 | ASSERT(refcyc_per_req_delivery_pre_c < dml_pow(2, 13)); |
1525 | ASSERT(refcyc_per_req_delivery_c < dml_pow(2, 13)); |
1526 | } |
1527 | |
1528 | // TTU - Cursor |
1529 | refcyc_per_req_delivery_pre_cur0 = 0.0; |
1530 | refcyc_per_req_delivery_cur0 = 0.0; |
1531 | if (src->num_cursors > 0) { |
1532 | calculate_ttu_cursor(mode_lib, |
1533 | refcyc_per_req_delivery_pre_cur: &refcyc_per_req_delivery_pre_cur0, |
1534 | refcyc_per_req_delivery_cur: &refcyc_per_req_delivery_cur0, |
1535 | refclk_freq_in_mhz, |
1536 | ref_freq_to_pix_freq, |
1537 | hscale_pixel_rate_l, |
1538 | hscl_ratio: scl->hscl_ratio, |
1539 | vratio_pre_l, |
1540 | vratio_l, |
1541 | cur_width: src->cur0_src_width, |
1542 | cur_bpp: (enum cursor_bpp)(src->cur0_bpp)); |
1543 | } |
1544 | |
1545 | refcyc_per_req_delivery_pre_cur1 = 0.0; |
1546 | refcyc_per_req_delivery_cur1 = 0.0; |
1547 | if (src->num_cursors > 1) { |
1548 | calculate_ttu_cursor(mode_lib, |
1549 | refcyc_per_req_delivery_pre_cur: &refcyc_per_req_delivery_pre_cur1, |
1550 | refcyc_per_req_delivery_cur: &refcyc_per_req_delivery_cur1, |
1551 | refclk_freq_in_mhz, |
1552 | ref_freq_to_pix_freq, |
1553 | hscale_pixel_rate_l, |
1554 | hscl_ratio: scl->hscl_ratio, |
1555 | vratio_pre_l, |
1556 | vratio_l, |
1557 | cur_width: src->cur1_src_width, |
1558 | cur_bpp: (enum cursor_bpp)(src->cur1_bpp)); |
1559 | } |
1560 | |
1561 | // TTU - Misc |
1562 | // all hard-coded |
1563 | |
1564 | // Assignment to register structures |
1565 | disp_dlg_regs->dst_y_after_scaler = dst_y_after_scaler; // in terms of line |
1566 | ASSERT(disp_dlg_regs->dst_y_after_scaler < (unsigned int)8); |
1567 | disp_dlg_regs->refcyc_x_after_scaler = dst_x_after_scaler * ref_freq_to_pix_freq; // in terms of refclk |
1568 | ASSERT(disp_dlg_regs->refcyc_x_after_scaler < (unsigned int)dml_pow(2, 13)); |
1569 | disp_dlg_regs->dst_y_prefetch = (unsigned int)(dst_y_prefetch * dml_pow(a: 2, exp: 2)); |
1570 | disp_dlg_regs->dst_y_per_vm_vblank = (unsigned int)(dst_y_per_vm_vblank * dml_pow(a: 2, exp: 2)); |
1571 | disp_dlg_regs->dst_y_per_row_vblank = (unsigned int)(dst_y_per_row_vblank * dml_pow(a: 2, exp: 2)); |
1572 | disp_dlg_regs->dst_y_per_vm_flip = (unsigned int)(dst_y_per_vm_flip * dml_pow(a: 2, exp: 2)); |
1573 | disp_dlg_regs->dst_y_per_row_flip = (unsigned int)(dst_y_per_row_flip * dml_pow(a: 2, exp: 2)); |
1574 | |
1575 | disp_dlg_regs->vratio_prefetch = (unsigned int)(vratio_pre_l * dml_pow(a: 2, exp: 19)); |
1576 | disp_dlg_regs->vratio_prefetch_c = (unsigned int)(vratio_pre_c * dml_pow(a: 2, exp: 19)); |
1577 | |
1578 | 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); |
1579 | 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); |
1580 | 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); |
1581 | 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); |
1582 | disp_dlg_regs->refcyc_per_pte_group_vblank_l = |
1583 | (unsigned int)(dst_y_per_row_vblank * (double)htotal |
1584 | * ref_freq_to_pix_freq / (double)dpte_groups_per_row_ub_l); |
1585 | ASSERT(disp_dlg_regs->refcyc_per_pte_group_vblank_l < (unsigned int)dml_pow(2, 13)); |
1586 | |
1587 | if (dual_plane) { |
1588 | disp_dlg_regs->refcyc_per_pte_group_vblank_c = (unsigned int)(dst_y_per_row_vblank |
1589 | * (double)htotal * ref_freq_to_pix_freq |
1590 | / (double)dpte_groups_per_row_ub_c); |
1591 | ASSERT(disp_dlg_regs->refcyc_per_pte_group_vblank_c |
1592 | < (unsigned int)dml_pow(2, 13)); |
1593 | } |
1594 | |
1595 | disp_dlg_regs->refcyc_per_meta_chunk_vblank_l = |
1596 | (unsigned int)(dst_y_per_row_vblank * (double)htotal |
1597 | * ref_freq_to_pix_freq / (double)meta_chunks_per_row_ub_l); |
1598 | ASSERT(disp_dlg_regs->refcyc_per_meta_chunk_vblank_l < (unsigned int)dml_pow(2, 13)); |
1599 | |
1600 | disp_dlg_regs->refcyc_per_meta_chunk_vblank_c = |
1601 | 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 |
1602 | |
1603 | disp_dlg_regs->refcyc_per_pte_group_flip_l = (unsigned int)(dst_y_per_row_flip * htotal |
1604 | * ref_freq_to_pix_freq) / dpte_groups_per_row_ub_l; |
1605 | disp_dlg_regs->refcyc_per_meta_chunk_flip_l = (unsigned int)(dst_y_per_row_flip * htotal |
1606 | * ref_freq_to_pix_freq) / meta_chunks_per_row_ub_l; |
1607 | |
1608 | if (dual_plane) { |
1609 | disp_dlg_regs->refcyc_per_pte_group_flip_c = (unsigned int)(dst_y_per_row_flip |
1610 | * htotal * ref_freq_to_pix_freq) / dpte_groups_per_row_ub_c; |
1611 | disp_dlg_regs->refcyc_per_meta_chunk_flip_c = (unsigned int)(dst_y_per_row_flip |
1612 | * htotal * ref_freq_to_pix_freq) / meta_chunks_per_row_ub_c; |
1613 | } |
1614 | |
1615 | disp_dlg_regs->refcyc_per_vm_group_vblank = get_refcyc_per_vm_group_vblank(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; |
1616 | disp_dlg_regs->refcyc_per_vm_group_flip = get_refcyc_per_vm_group_flip(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; |
1617 | disp_dlg_regs->refcyc_per_vm_req_vblank = get_refcyc_per_vm_req_vblank(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz * dml_pow(a: 2, exp: 10); |
1618 | disp_dlg_regs->refcyc_per_vm_req_flip = get_refcyc_per_vm_req_flip(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz * dml_pow(a: 2, exp: 10); |
1619 | |
1620 | // Clamp to max for now |
1621 | if (disp_dlg_regs->refcyc_per_vm_group_vblank >= (unsigned int)dml_pow(a: 2, exp: 23)) |
1622 | disp_dlg_regs->refcyc_per_vm_group_vblank = dml_pow(a: 2, exp: 23) - 1; |
1623 | |
1624 | if (disp_dlg_regs->refcyc_per_vm_group_flip >= (unsigned int)dml_pow(a: 2, exp: 23)) |
1625 | disp_dlg_regs->refcyc_per_vm_group_flip = dml_pow(a: 2, exp: 23) - 1; |
1626 | |
1627 | if (disp_dlg_regs->refcyc_per_vm_req_vblank >= (unsigned int)dml_pow(a: 2, exp: 23)) |
1628 | disp_dlg_regs->refcyc_per_vm_req_vblank = dml_pow(a: 2, exp: 23) - 1; |
1629 | |
1630 | if (disp_dlg_regs->refcyc_per_vm_req_flip >= (unsigned int)dml_pow(a: 2, exp: 23)) |
1631 | disp_dlg_regs->refcyc_per_vm_req_flip = dml_pow(a: 2, exp: 23) - 1; |
1632 | |
1633 | disp_dlg_regs->dst_y_per_pte_row_nom_l = (unsigned int)((double)dpte_row_height_l |
1634 | / (double)vratio_l * dml_pow(a: 2, exp: 2)); |
1635 | ASSERT(disp_dlg_regs->dst_y_per_pte_row_nom_l < (unsigned int)dml_pow(2, 17)); |
1636 | |
1637 | if (dual_plane) { |
1638 | disp_dlg_regs->dst_y_per_pte_row_nom_c = (unsigned int)((double)dpte_row_height_c |
1639 | / (double)vratio_c * dml_pow(a: 2, exp: 2)); |
1640 | if (disp_dlg_regs->dst_y_per_pte_row_nom_c >= (unsigned int)dml_pow(a: 2, exp: 17)) { |
1641 | dml_print("DML_DLG: %s: Warning dst_y_per_pte_row_nom_c %u larger than supported by register format U15.2 %u\n" , |
1642 | __func__, |
1643 | disp_dlg_regs->dst_y_per_pte_row_nom_c, |
1644 | (unsigned int)dml_pow(2, 17) - 1); |
1645 | } |
1646 | } |
1647 | |
1648 | disp_dlg_regs->dst_y_per_meta_row_nom_l = (unsigned int)((double)meta_row_height_l |
1649 | / (double)vratio_l * dml_pow(a: 2, exp: 2)); |
1650 | ASSERT(disp_dlg_regs->dst_y_per_meta_row_nom_l < (unsigned int)dml_pow(2, 17)); |
1651 | |
1652 | 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 |
1653 | |
1654 | dml_print("DML: Trow: %fus\n" , line_time_in_us * (double)dpte_row_height_l / (double)vratio_l); |
1655 | |
1656 | disp_dlg_regs->refcyc_per_pte_group_nom_l = (unsigned int)((double)dpte_row_height_l |
1657 | / (double)vratio_l * (double)htotal * ref_freq_to_pix_freq |
1658 | / (double)dpte_groups_per_row_ub_l); |
1659 | if (disp_dlg_regs->refcyc_per_pte_group_nom_l >= (unsigned int)dml_pow(a: 2, exp: 23)) |
1660 | disp_dlg_regs->refcyc_per_pte_group_nom_l = dml_pow(a: 2, exp: 23) - 1; |
1661 | disp_dlg_regs->refcyc_per_meta_chunk_nom_l = (unsigned int)((double)meta_row_height_l |
1662 | / (double)vratio_l * (double)htotal * ref_freq_to_pix_freq |
1663 | / (double)meta_chunks_per_row_ub_l); |
1664 | if (disp_dlg_regs->refcyc_per_meta_chunk_nom_l >= (unsigned int)dml_pow(a: 2, exp: 23)) |
1665 | disp_dlg_regs->refcyc_per_meta_chunk_nom_l = dml_pow(a: 2, exp: 23) - 1; |
1666 | |
1667 | if (dual_plane) { |
1668 | disp_dlg_regs->refcyc_per_pte_group_nom_c = |
1669 | (unsigned int)((double)dpte_row_height_c / (double)vratio_c |
1670 | * (double)htotal * ref_freq_to_pix_freq |
1671 | / (double)dpte_groups_per_row_ub_c); |
1672 | if (disp_dlg_regs->refcyc_per_pte_group_nom_c >= (unsigned int)dml_pow(a: 2, exp: 23)) |
1673 | disp_dlg_regs->refcyc_per_pte_group_nom_c = dml_pow(a: 2, exp: 23) - 1; |
1674 | |
1675 | // TODO: Is this the right calculation? Does htotal need to be halved? |
1676 | disp_dlg_regs->refcyc_per_meta_chunk_nom_c = |
1677 | (unsigned int)((double)meta_row_height_c / (double)vratio_c |
1678 | * (double)htotal * ref_freq_to_pix_freq |
1679 | / (double)meta_chunks_per_row_ub_c); |
1680 | if (disp_dlg_regs->refcyc_per_meta_chunk_nom_c >= (unsigned int)dml_pow(a: 2, exp: 23)) |
1681 | disp_dlg_regs->refcyc_per_meta_chunk_nom_c = dml_pow(a: 2, exp: 23) - 1; |
1682 | } |
1683 | |
1684 | disp_dlg_regs->refcyc_per_line_delivery_pre_l = (unsigned int)dml_floor(a: refcyc_per_line_delivery_pre_l, |
1685 | granularity: 1); |
1686 | disp_dlg_regs->refcyc_per_line_delivery_l = (unsigned int)dml_floor(a: refcyc_per_line_delivery_l, |
1687 | granularity: 1); |
1688 | ASSERT(disp_dlg_regs->refcyc_per_line_delivery_pre_l < (unsigned int)dml_pow(2, 13)); |
1689 | ASSERT(disp_dlg_regs->refcyc_per_line_delivery_l < (unsigned int)dml_pow(2, 13)); |
1690 | |
1691 | disp_dlg_regs->refcyc_per_line_delivery_pre_c = (unsigned int)dml_floor(a: refcyc_per_line_delivery_pre_c, |
1692 | granularity: 1); |
1693 | disp_dlg_regs->refcyc_per_line_delivery_c = (unsigned int)dml_floor(a: refcyc_per_line_delivery_c, |
1694 | granularity: 1); |
1695 | ASSERT(disp_dlg_regs->refcyc_per_line_delivery_pre_c < (unsigned int)dml_pow(2, 13)); |
1696 | ASSERT(disp_dlg_regs->refcyc_per_line_delivery_c < (unsigned int)dml_pow(2, 13)); |
1697 | |
1698 | disp_dlg_regs->chunk_hdl_adjust_cur0 = 3; |
1699 | disp_dlg_regs->dst_y_offset_cur0 = 0; |
1700 | disp_dlg_regs->chunk_hdl_adjust_cur1 = 3; |
1701 | disp_dlg_regs->dst_y_offset_cur1 = 0; |
1702 | |
1703 | disp_dlg_regs->dst_y_delta_drq_limit = 0x7fff; // off |
1704 | |
1705 | disp_ttu_regs->refcyc_per_req_delivery_pre_l = (unsigned int)(refcyc_per_req_delivery_pre_l |
1706 | * dml_pow(a: 2, exp: 10)); |
1707 | disp_ttu_regs->refcyc_per_req_delivery_l = (unsigned int)(refcyc_per_req_delivery_l |
1708 | * dml_pow(a: 2, exp: 10)); |
1709 | disp_ttu_regs->refcyc_per_req_delivery_pre_c = (unsigned int)(refcyc_per_req_delivery_pre_c |
1710 | * dml_pow(a: 2, exp: 10)); |
1711 | disp_ttu_regs->refcyc_per_req_delivery_c = (unsigned int)(refcyc_per_req_delivery_c |
1712 | * dml_pow(a: 2, exp: 10)); |
1713 | disp_ttu_regs->refcyc_per_req_delivery_pre_cur0 = |
1714 | (unsigned int)(refcyc_per_req_delivery_pre_cur0 * dml_pow(a: 2, exp: 10)); |
1715 | disp_ttu_regs->refcyc_per_req_delivery_cur0 = (unsigned int)(refcyc_per_req_delivery_cur0 |
1716 | * dml_pow(a: 2, exp: 10)); |
1717 | disp_ttu_regs->refcyc_per_req_delivery_pre_cur1 = |
1718 | (unsigned int)(refcyc_per_req_delivery_pre_cur1 * dml_pow(a: 2, exp: 10)); |
1719 | disp_ttu_regs->refcyc_per_req_delivery_cur1 = (unsigned int)(refcyc_per_req_delivery_cur1 |
1720 | * dml_pow(a: 2, exp: 10)); |
1721 | disp_ttu_regs->qos_level_low_wm = 0; |
1722 | ASSERT(disp_ttu_regs->qos_level_low_wm < dml_pow(2, 14)); |
1723 | disp_ttu_regs->qos_level_high_wm = (unsigned int)(4.0 * (double)htotal |
1724 | * ref_freq_to_pix_freq); |
1725 | ASSERT(disp_ttu_regs->qos_level_high_wm < dml_pow(2, 14)); |
1726 | |
1727 | disp_ttu_regs->qos_level_flip = 14; |
1728 | disp_ttu_regs->qos_level_fixed_l = 8; |
1729 | disp_ttu_regs->qos_level_fixed_c = 8; |
1730 | disp_ttu_regs->qos_level_fixed_cur0 = 8; |
1731 | disp_ttu_regs->qos_ramp_disable_l = 0; |
1732 | disp_ttu_regs->qos_ramp_disable_c = 0; |
1733 | disp_ttu_regs->qos_ramp_disable_cur0 = 0; |
1734 | |
1735 | disp_ttu_regs->min_ttu_vblank = min_ttu_vblank * refclk_freq_in_mhz; |
1736 | ASSERT(disp_ttu_regs->min_ttu_vblank < dml_pow(2, 24)); |
1737 | |
1738 | print__ttu_regs_st(mode_lib, ttu_regs: disp_ttu_regs); |
1739 | print__dlg_regs_st(mode_lib, dlg_regs: disp_dlg_regs); |
1740 | } |
1741 | |
1742 | void dml30_rq_dlg_get_dlg_reg(struct display_mode_lib *mode_lib, |
1743 | display_dlg_regs_st *dlg_regs, |
1744 | display_ttu_regs_st *ttu_regs, |
1745 | const display_e2e_pipe_params_st *e2e_pipe_param, |
1746 | const unsigned int num_pipes, |
1747 | const unsigned int pipe_idx, |
1748 | const bool cstate_en, |
1749 | const bool pstate_en, |
1750 | const bool vm_en, |
1751 | const bool ignore_viewport_pos, |
1752 | const bool immediate_flip_support) |
1753 | { |
1754 | display_rq_params_st rq_param = { 0 }; |
1755 | display_dlg_sys_params_st dlg_sys_param = { 0 }; |
1756 | |
1757 | // Get watermark and Tex. |
1758 | dlg_sys_param.t_urg_wm_us = get_wm_urgent(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1759 | dlg_sys_param.deepsleep_dcfclk_mhz = get_clk_dcf_deepsleep(mode_lib, |
1760 | pipes: e2e_pipe_param, |
1761 | num_pipes); |
1762 | dlg_sys_param.t_extra_us = get_urgent_extra_latency(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1763 | dlg_sys_param.mem_trip_us = get_wm_memory_trip(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1764 | dlg_sys_param.t_mclk_wm_us = get_wm_dram_clock_change(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1765 | dlg_sys_param.t_sr_wm_us = get_wm_stutter_enter_exit(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1766 | dlg_sys_param.total_flip_bw = get_total_immediate_flip_bw(mode_lib, |
1767 | pipes: e2e_pipe_param, |
1768 | num_pipes); |
1769 | dlg_sys_param.total_flip_bytes = get_total_immediate_flip_bytes(mode_lib, |
1770 | pipes: e2e_pipe_param, |
1771 | num_pipes); |
1772 | |
1773 | print__dlg_sys_params_st(mode_lib, dlg_sys_param: &dlg_sys_param); |
1774 | |
1775 | // system parameter calculation done |
1776 | |
1777 | dml_print("DML_DLG: Calculation for pipe[%d] start\n\n" , pipe_idx); |
1778 | dml_rq_dlg_get_rq_params(mode_lib, rq_param: &rq_param, pipe_param: &e2e_pipe_param[pipe_idx].pipe); |
1779 | dml_rq_dlg_get_dlg_params(mode_lib, |
1780 | e2e_pipe_param, |
1781 | num_pipes, |
1782 | pipe_idx, |
1783 | disp_dlg_regs: dlg_regs, |
1784 | disp_ttu_regs: ttu_regs, |
1785 | rq_dlg_param: rq_param.dlg, |
1786 | dlg_sys_param, |
1787 | cstate_en, |
1788 | pstate_en, |
1789 | vm_en, |
1790 | ignore_viewport_pos, |
1791 | immediate_flip_support); |
1792 | dml_print("DML_DLG: Calculation for pipe[%d] end\n" , pipe_idx); |
1793 | } |
1794 | |
1795 | |