1 | // SPDX-License-Identifier: MIT |
2 | /* |
3 | * Copyright 2022 Advanced Micro Devices, Inc. |
4 | * |
5 | * Permission is hereby granted, free of charge, to any person obtaining a |
6 | * copy of this software and associated documentation files (the "Software"), |
7 | * to deal in the Software without restriction, including without limitation |
8 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
9 | * and/or sell copies of the Software, and to permit persons to whom the |
10 | * Software is furnished to do so, subject to the following conditions: |
11 | * |
12 | * The above copyright notice and this permission notice shall be included in |
13 | * all copies or substantial portions of the Software. |
14 | * |
15 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
16 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
17 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
18 | * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR |
19 | * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, |
20 | * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
21 | * OTHER DEALINGS IN THE SOFTWARE. |
22 | * |
23 | * Authors: AMD |
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_314.h" |
31 | |
32 | static bool CalculateBytePerPixelAnd256BBlockSizes( |
33 | enum source_format_class SourcePixelFormat, |
34 | enum dm_swizzle_mode SurfaceTiling, |
35 | unsigned int *BytePerPixelY, |
36 | unsigned int *BytePerPixelC, |
37 | double *BytePerPixelDETY, |
38 | double *BytePerPixelDETC, |
39 | unsigned int *BlockHeight256BytesY, |
40 | unsigned int *BlockHeight256BytesC, |
41 | unsigned int *BlockWidth256BytesY, |
42 | unsigned int *BlockWidth256BytesC) |
43 | { |
44 | if (SourcePixelFormat == dm_444_64) { |
45 | *BytePerPixelDETY = 8; |
46 | *BytePerPixelDETC = 0; |
47 | *BytePerPixelY = 8; |
48 | *BytePerPixelC = 0; |
49 | } else if (SourcePixelFormat == dm_444_32 || SourcePixelFormat == dm_rgbe) { |
50 | *BytePerPixelDETY = 4; |
51 | *BytePerPixelDETC = 0; |
52 | *BytePerPixelY = 4; |
53 | *BytePerPixelC = 0; |
54 | } else if (SourcePixelFormat == dm_444_16) { |
55 | *BytePerPixelDETY = 2; |
56 | *BytePerPixelDETC = 0; |
57 | *BytePerPixelY = 2; |
58 | *BytePerPixelC = 0; |
59 | } else if (SourcePixelFormat == dm_444_8) { |
60 | *BytePerPixelDETY = 1; |
61 | *BytePerPixelDETC = 0; |
62 | *BytePerPixelY = 1; |
63 | *BytePerPixelC = 0; |
64 | } else if (SourcePixelFormat == dm_rgbe_alpha) { |
65 | *BytePerPixelDETY = 4; |
66 | *BytePerPixelDETC = 1; |
67 | *BytePerPixelY = 4; |
68 | *BytePerPixelC = 1; |
69 | } else if (SourcePixelFormat == dm_420_8) { |
70 | *BytePerPixelDETY = 1; |
71 | *BytePerPixelDETC = 2; |
72 | *BytePerPixelY = 1; |
73 | *BytePerPixelC = 2; |
74 | } else if (SourcePixelFormat == dm_420_12) { |
75 | *BytePerPixelDETY = 2; |
76 | *BytePerPixelDETC = 4; |
77 | *BytePerPixelY = 2; |
78 | *BytePerPixelC = 4; |
79 | } else { |
80 | *BytePerPixelDETY = 4.0 / 3; |
81 | *BytePerPixelDETC = 8.0 / 3; |
82 | *BytePerPixelY = 2; |
83 | *BytePerPixelC = 4; |
84 | } |
85 | |
86 | if ((SourcePixelFormat == dm_444_64 || SourcePixelFormat == dm_444_32 || SourcePixelFormat == dm_444_16 || SourcePixelFormat == dm_444_8 || SourcePixelFormat == dm_mono_16 |
87 | || SourcePixelFormat == dm_mono_8 || SourcePixelFormat == dm_rgbe)) { |
88 | if (SurfaceTiling == dm_sw_linear) |
89 | *BlockHeight256BytesY = 1; |
90 | else if (SourcePixelFormat == dm_444_64) |
91 | *BlockHeight256BytesY = 4; |
92 | else if (SourcePixelFormat == dm_444_8) |
93 | *BlockHeight256BytesY = 16; |
94 | else |
95 | *BlockHeight256BytesY = 8; |
96 | |
97 | *BlockWidth256BytesY = 256U / *BytePerPixelY / *BlockHeight256BytesY; |
98 | *BlockHeight256BytesC = 0; |
99 | *BlockWidth256BytesC = 0; |
100 | } else { |
101 | if (SurfaceTiling == dm_sw_linear) { |
102 | *BlockHeight256BytesY = 1; |
103 | *BlockHeight256BytesC = 1; |
104 | } else if (SourcePixelFormat == dm_rgbe_alpha) { |
105 | *BlockHeight256BytesY = 8; |
106 | *BlockHeight256BytesC = 16; |
107 | } else if (SourcePixelFormat == dm_420_8) { |
108 | *BlockHeight256BytesY = 16; |
109 | *BlockHeight256BytesC = 8; |
110 | } else { |
111 | *BlockHeight256BytesY = 8; |
112 | *BlockHeight256BytesC = 8; |
113 | } |
114 | *BlockWidth256BytesY = 256U / *BytePerPixelY / *BlockHeight256BytesY; |
115 | *BlockWidth256BytesC = 256U / *BytePerPixelC / *BlockHeight256BytesC; |
116 | } |
117 | return true; |
118 | } |
119 | |
120 | static bool is_dual_plane(enum source_format_class source_format) |
121 | { |
122 | bool ret_val = 0; |
123 | |
124 | if ((source_format == dm_420_12) || (source_format == dm_420_8) || (source_format == dm_420_10) || (source_format == dm_rgbe_alpha)) |
125 | ret_val = 1; |
126 | |
127 | return ret_val; |
128 | } |
129 | |
130 | static double get_refcyc_per_delivery( |
131 | struct display_mode_lib *mode_lib, |
132 | double refclk_freq_in_mhz, |
133 | double pclk_freq_in_mhz, |
134 | unsigned int odm_combine, |
135 | unsigned int recout_width, |
136 | unsigned int hactive, |
137 | double vratio, |
138 | double hscale_pixel_rate, |
139 | unsigned int delivery_width, |
140 | unsigned int req_per_swath_ub) |
141 | { |
142 | double refcyc_per_delivery = 0.0; |
143 | |
144 | if (vratio <= 1.0) { |
145 | if (odm_combine) |
146 | refcyc_per_delivery = (double) refclk_freq_in_mhz * (double) ((unsigned int) odm_combine * 2) |
147 | * dml_min(a: (double) recout_width, b: (double) hactive / ((unsigned int) odm_combine * 2)) / pclk_freq_in_mhz / (double) req_per_swath_ub; |
148 | else |
149 | refcyc_per_delivery = (double) refclk_freq_in_mhz * (double) recout_width / pclk_freq_in_mhz / (double) req_per_swath_ub; |
150 | } else { |
151 | refcyc_per_delivery = (double) refclk_freq_in_mhz * (double) delivery_width / (double) hscale_pixel_rate / (double) req_per_swath_ub; |
152 | } |
153 | |
154 | #ifdef __DML_RQ_DLG_CALC_DEBUG__ |
155 | dml_print("DML_DLG: %s: refclk_freq_in_mhz = %3.2f\n" , __func__, refclk_freq_in_mhz); |
156 | dml_print("DML_DLG: %s: pclk_freq_in_mhz = %3.2f\n" , __func__, pclk_freq_in_mhz); |
157 | dml_print("DML_DLG: %s: recout_width = %d\n" , __func__, recout_width); |
158 | dml_print("DML_DLG: %s: vratio = %3.2f\n" , __func__, vratio); |
159 | dml_print("DML_DLG: %s: req_per_swath_ub = %d\n" , __func__, req_per_swath_ub); |
160 | dml_print("DML_DLG: %s: hscale_pixel_rate = %3.2f\n" , __func__, hscale_pixel_rate); |
161 | dml_print("DML_DLG: %s: delivery_width = %d\n" , __func__, delivery_width); |
162 | dml_print("DML_DLG: %s: refcyc_per_delivery= %3.2f\n" , __func__, refcyc_per_delivery); |
163 | #endif |
164 | |
165 | return refcyc_per_delivery; |
166 | |
167 | } |
168 | |
169 | static unsigned int get_blk_size_bytes(const enum source_macro_tile_size tile_size) |
170 | { |
171 | if (tile_size == dm_256k_tile) |
172 | return (256 * 1024); |
173 | else if (tile_size == dm_64k_tile) |
174 | return (64 * 1024); |
175 | else |
176 | return (4 * 1024); |
177 | } |
178 | |
179 | static void (struct display_mode_lib *mode_lib, display_data_rq_regs_st *rq_regs, const display_data_rq_sizing_params_st *rq_sizing) |
180 | { |
181 | print__data_rq_sizing_params_st(mode_lib, rq_sizing); |
182 | |
183 | rq_regs->chunk_size = dml_log2(x: rq_sizing->chunk_bytes) - 10; |
184 | |
185 | if (rq_sizing->min_chunk_bytes == 0) |
186 | rq_regs->min_chunk_size = 0; |
187 | else |
188 | rq_regs->min_chunk_size = dml_log2(x: rq_sizing->min_chunk_bytes) - 8 + 1; |
189 | |
190 | rq_regs->meta_chunk_size = dml_log2(x: rq_sizing->meta_chunk_bytes) - 10; |
191 | if (rq_sizing->min_meta_chunk_bytes == 0) |
192 | rq_regs->min_meta_chunk_size = 0; |
193 | else |
194 | rq_regs->min_meta_chunk_size = dml_log2(x: rq_sizing->min_meta_chunk_bytes) - 6 + 1; |
195 | |
196 | rq_regs->dpte_group_size = dml_log2(x: rq_sizing->dpte_group_bytes) - 6; |
197 | rq_regs->mpte_group_size = dml_log2(x: rq_sizing->mpte_group_bytes) - 6; |
198 | } |
199 | |
200 | static void (struct display_mode_lib *mode_lib, display_rq_regs_st *rq_regs, const display_rq_params_st *rq_param) |
201 | { |
202 | unsigned int detile_buf_size_in_bytes = mode_lib->ip.det_buffer_size_kbytes * 1024; |
203 | unsigned int detile_buf_plane1_addr = 0; |
204 | |
205 | extract_rq_sizing_regs(mode_lib, rq_regs: &(rq_regs->rq_regs_l), rq_sizing: &rq_param->sizing.rq_l); |
206 | |
207 | rq_regs->rq_regs_l.pte_row_height_linear = dml_floor(a: dml_log2(x: rq_param->dlg.rq_l.dpte_row_height), granularity: 1) - 3; |
208 | |
209 | if (rq_param->yuv420) { |
210 | extract_rq_sizing_regs(mode_lib, rq_regs: &(rq_regs->rq_regs_c), rq_sizing: &rq_param->sizing.rq_c); |
211 | rq_regs->rq_regs_c.pte_row_height_linear = dml_floor(a: dml_log2(x: rq_param->dlg.rq_c.dpte_row_height), granularity: 1) - 3; |
212 | } |
213 | |
214 | rq_regs->rq_regs_l.swath_height = dml_log2(x: rq_param->dlg.rq_l.swath_height); |
215 | rq_regs->rq_regs_c.swath_height = dml_log2(x: rq_param->dlg.rq_c.swath_height); |
216 | |
217 | // FIXME: take the max between luma, chroma chunk size? |
218 | // okay for now, as we are setting chunk_bytes to 8kb anyways |
219 | if (rq_param->sizing.rq_l.chunk_bytes >= 32 * 1024 || (rq_param->yuv420 && rq_param->sizing.rq_c.chunk_bytes >= 32 * 1024)) { //32kb |
220 | rq_regs->drq_expansion_mode = 0; |
221 | } else { |
222 | rq_regs->drq_expansion_mode = 2; |
223 | } |
224 | rq_regs->prq_expansion_mode = 1; |
225 | rq_regs->mrq_expansion_mode = 1; |
226 | rq_regs->crq_expansion_mode = 1; |
227 | |
228 | // Note: detile_buf_plane1_addr is in unit of 1KB |
229 | if (rq_param->yuv420) { |
230 | if ((double) rq_param->misc.rq_l.stored_swath_bytes / (double) rq_param->misc.rq_c.stored_swath_bytes <= 1.5) { |
231 | detile_buf_plane1_addr = (detile_buf_size_in_bytes / 2.0 / 1024.0); // half to chroma |
232 | #ifdef __DML_RQ_DLG_CALC_DEBUG__ |
233 | dml_print("DML_DLG: %s: detile_buf_plane1_addr = %0d (1/2 to chroma)\n" , __func__, detile_buf_plane1_addr); |
234 | #endif |
235 | } else { |
236 | detile_buf_plane1_addr = dml_round_to_multiple(num: (unsigned int) ((2.0 * detile_buf_size_in_bytes) / 3.0), multiple: 1024, up: 0) / 1024.0; // 2/3 to luma |
237 | #ifdef __DML_RQ_DLG_CALC_DEBUG__ |
238 | dml_print("DML_DLG: %s: detile_buf_plane1_addr = %0d (1/3 chroma)\n" , __func__, detile_buf_plane1_addr); |
239 | #endif |
240 | } |
241 | } |
242 | rq_regs->plane1_base_address = detile_buf_plane1_addr; |
243 | |
244 | #ifdef __DML_RQ_DLG_CALC_DEBUG__ |
245 | dml_print("DML_DLG: %s: detile_buf_size_in_bytes = %0d\n" , __func__, detile_buf_size_in_bytes); |
246 | dml_print("DML_DLG: %s: detile_buf_plane1_addr = %0d\n" , __func__, detile_buf_plane1_addr); |
247 | dml_print("DML_DLG: %s: plane1_base_address = %0d\n" , __func__, rq_regs->plane1_base_address); |
248 | dml_print("DML_DLG: %s: rq_l.stored_swath_bytes = %0d\n" , __func__, rq_param->misc.rq_l.stored_swath_bytes); |
249 | dml_print("DML_DLG: %s: rq_c.stored_swath_bytes = %0d\n" , __func__, rq_param->misc.rq_c.stored_swath_bytes); |
250 | dml_print("DML_DLG: %s: rq_l.swath_height = %0d\n" , __func__, rq_param->dlg.rq_l.swath_height); |
251 | dml_print("DML_DLG: %s: rq_c.swath_height = %0d\n" , __func__, rq_param->dlg.rq_c.swath_height); |
252 | #endif |
253 | } |
254 | |
255 | static void 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) |
256 | { |
257 | unsigned int total_swath_bytes = 0; |
258 | unsigned int swath_bytes_l = 0; |
259 | unsigned int swath_bytes_c = 0; |
260 | unsigned int full_swath_bytes_packed_l = 0; |
261 | unsigned int full_swath_bytes_packed_c = 0; |
262 | bool req128_l = 0; |
263 | bool req128_c = 0; |
264 | bool surf_linear = (pipe_src_param->sw_mode == dm_sw_linear); |
265 | bool surf_vert = (pipe_src_param->source_scan == dm_vert); |
266 | unsigned int log2_swath_height_l = 0; |
267 | unsigned int log2_swath_height_c = 0; |
268 | unsigned int detile_buf_size_in_bytes = mode_lib->ip.det_buffer_size_kbytes * 1024; |
269 | |
270 | full_swath_bytes_packed_l = rq_param->misc.rq_l.full_swath_bytes; |
271 | full_swath_bytes_packed_c = rq_param->misc.rq_c.full_swath_bytes; |
272 | |
273 | #ifdef __DML_RQ_DLG_CALC_DEBUG__ |
274 | dml_print("DML_DLG: %s: full_swath_bytes_packed_l = %0d\n" , __func__, full_swath_bytes_packed_l); |
275 | dml_print("DML_DLG: %s: full_swath_bytes_packed_c = %0d\n" , __func__, full_swath_bytes_packed_c); |
276 | #endif |
277 | |
278 | if (rq_param->yuv420_10bpc) { |
279 | full_swath_bytes_packed_l = dml_round_to_multiple(num: rq_param->misc.rq_l.full_swath_bytes * 2.0 / 3.0, multiple: 256, up: 1) + 256; |
280 | full_swath_bytes_packed_c = dml_round_to_multiple(num: rq_param->misc.rq_c.full_swath_bytes * 2.0 / 3.0, multiple: 256, up: 1) + 256; |
281 | #ifdef __DML_RQ_DLG_CALC_DEBUG__ |
282 | dml_print("DML_DLG: %s: full_swath_bytes_packed_l = %0d (3-2 packing)\n" , __func__, full_swath_bytes_packed_l); |
283 | dml_print("DML_DLG: %s: full_swath_bytes_packed_c = %0d (3-2 packing)\n" , __func__, full_swath_bytes_packed_c); |
284 | #endif |
285 | } |
286 | |
287 | if (rq_param->yuv420) |
288 | total_swath_bytes = 2 * full_swath_bytes_packed_l + 2 * full_swath_bytes_packed_c; |
289 | else |
290 | total_swath_bytes = 2 * full_swath_bytes_packed_l; |
291 | |
292 | #ifdef __DML_RQ_DLG_CALC_DEBUG__ |
293 | dml_print("DML_DLG: %s: total_swath_bytes = %0d\n" , __func__, total_swath_bytes); |
294 | dml_print("DML_DLG: %s: detile_buf_size_in_bytes = %0d\n" , __func__, detile_buf_size_in_bytes); |
295 | #endif |
296 | |
297 | if (total_swath_bytes <= detile_buf_size_in_bytes) { //full 256b request |
298 | req128_l = 0; |
299 | req128_c = 0; |
300 | swath_bytes_l = full_swath_bytes_packed_l; |
301 | swath_bytes_c = full_swath_bytes_packed_c; |
302 | } else if (!rq_param->yuv420) { |
303 | req128_l = 1; |
304 | req128_c = 0; |
305 | swath_bytes_c = full_swath_bytes_packed_c; |
306 | swath_bytes_l = full_swath_bytes_packed_l / 2; |
307 | } else if ((double) full_swath_bytes_packed_l / (double) full_swath_bytes_packed_c < 1.5) { |
308 | req128_l = 0; |
309 | req128_c = 1; |
310 | swath_bytes_l = full_swath_bytes_packed_l; |
311 | swath_bytes_c = full_swath_bytes_packed_c / 2; |
312 | |
313 | total_swath_bytes = 2 * swath_bytes_l + 2 * swath_bytes_c; |
314 | |
315 | if (total_swath_bytes > detile_buf_size_in_bytes) { |
316 | req128_l = 1; |
317 | swath_bytes_l = full_swath_bytes_packed_l / 2; |
318 | } |
319 | } else { |
320 | req128_l = 1; |
321 | req128_c = 0; |
322 | swath_bytes_l = full_swath_bytes_packed_l / 2; |
323 | swath_bytes_c = full_swath_bytes_packed_c; |
324 | |
325 | total_swath_bytes = 2 * swath_bytes_l + 2 * swath_bytes_c; |
326 | |
327 | if (total_swath_bytes > detile_buf_size_in_bytes) { |
328 | req128_c = 1; |
329 | swath_bytes_c = full_swath_bytes_packed_c / 2; |
330 | } |
331 | } |
332 | |
333 | if (rq_param->yuv420) |
334 | total_swath_bytes = 2 * swath_bytes_l + 2 * swath_bytes_c; |
335 | else |
336 | total_swath_bytes = 2 * swath_bytes_l; |
337 | |
338 | rq_param->misc.rq_l.stored_swath_bytes = swath_bytes_l; |
339 | rq_param->misc.rq_c.stored_swath_bytes = swath_bytes_c; |
340 | |
341 | #ifdef __DML_RQ_DLG_CALC_DEBUG__ |
342 | dml_print("DML_DLG: %s: total_swath_bytes = %0d\n" , __func__, total_swath_bytes); |
343 | dml_print("DML_DLG: %s: rq_l.stored_swath_bytes = %0d\n" , __func__, rq_param->misc.rq_l.stored_swath_bytes); |
344 | dml_print("DML_DLG: %s: rq_c.stored_swath_bytes = %0d\n" , __func__, rq_param->misc.rq_c.stored_swath_bytes); |
345 | #endif |
346 | if (surf_linear) { |
347 | log2_swath_height_l = 0; |
348 | log2_swath_height_c = 0; |
349 | } else { |
350 | unsigned int swath_height_l; |
351 | unsigned int swath_height_c; |
352 | |
353 | if (!surf_vert) { |
354 | swath_height_l = rq_param->misc.rq_l.blk256_height; |
355 | swath_height_c = rq_param->misc.rq_c.blk256_height; |
356 | } else { |
357 | swath_height_l = rq_param->misc.rq_l.blk256_width; |
358 | swath_height_c = rq_param->misc.rq_c.blk256_width; |
359 | } |
360 | |
361 | if (swath_height_l > 0) |
362 | log2_swath_height_l = dml_log2(x: swath_height_l); |
363 | |
364 | if (req128_l && log2_swath_height_l > 0) |
365 | log2_swath_height_l -= 1; |
366 | |
367 | if (swath_height_c > 0) |
368 | log2_swath_height_c = dml_log2(x: swath_height_c); |
369 | |
370 | if (req128_c && log2_swath_height_c > 0) |
371 | log2_swath_height_c -= 1; |
372 | } |
373 | |
374 | rq_param->dlg.rq_l.swath_height = 1 << log2_swath_height_l; |
375 | rq_param->dlg.rq_c.swath_height = 1 << log2_swath_height_c; |
376 | |
377 | #ifdef __DML_RQ_DLG_CALC_DEBUG__ |
378 | dml_print("DML_DLG: %s: req128_l = %0d\n" , __func__, req128_l); |
379 | dml_print("DML_DLG: %s: req128_c = %0d\n" , __func__, req128_c); |
380 | dml_print("DML_DLG: %s: full_swath_bytes_packed_l = %0d\n" , __func__, full_swath_bytes_packed_l); |
381 | dml_print("DML_DLG: %s: full_swath_bytes_packed_c = %0d\n" , __func__, full_swath_bytes_packed_c); |
382 | dml_print("DML_DLG: %s: swath_height luma = %0d\n" , __func__, rq_param->dlg.rq_l.swath_height); |
383 | dml_print("DML_DLG: %s: swath_height chroma = %0d\n" , __func__, rq_param->dlg.rq_c.swath_height); |
384 | #endif |
385 | } |
386 | |
387 | static void get_meta_and_pte_attr( |
388 | struct display_mode_lib *mode_lib, |
389 | display_data_rq_dlg_params_st *rq_dlg_param, |
390 | display_data_rq_misc_params_st *rq_misc_param, |
391 | display_data_rq_sizing_params_st *rq_sizing_param, |
392 | unsigned int vp_width, |
393 | unsigned int vp_height, |
394 | unsigned int data_pitch, |
395 | unsigned int meta_pitch, |
396 | unsigned int source_format, |
397 | unsigned int tiling, |
398 | unsigned int macro_tile_size, |
399 | unsigned int source_scan, |
400 | unsigned int hostvm_enable, |
401 | unsigned int is_chroma, |
402 | unsigned int surface_height) |
403 | { |
404 | bool surf_linear = (tiling == dm_sw_linear); |
405 | bool surf_vert = (source_scan == dm_vert); |
406 | |
407 | unsigned int bytes_per_element; |
408 | unsigned int bytes_per_element_y; |
409 | unsigned int bytes_per_element_c; |
410 | |
411 | unsigned int blk256_width = 0; |
412 | unsigned int blk256_height = 0; |
413 | |
414 | unsigned int blk256_width_y = 0; |
415 | unsigned int blk256_height_y = 0; |
416 | unsigned int blk256_width_c = 0; |
417 | unsigned int blk256_height_c = 0; |
418 | unsigned int log2_bytes_per_element; |
419 | unsigned int log2_blk256_width; |
420 | unsigned int log2_blk256_height; |
421 | unsigned int blk_bytes; |
422 | unsigned int log2_blk_bytes; |
423 | unsigned int log2_blk_height; |
424 | unsigned int log2_blk_width; |
425 | unsigned int log2_meta_req_bytes; |
426 | unsigned int log2_meta_req_height; |
427 | unsigned int log2_meta_req_width; |
428 | unsigned int meta_req_width; |
429 | unsigned int meta_req_height; |
430 | unsigned int log2_meta_row_height; |
431 | unsigned int meta_row_width_ub; |
432 | unsigned int log2_meta_chunk_bytes; |
433 | unsigned int log2_meta_chunk_height; |
434 | |
435 | //full sized meta chunk width in unit of data elements |
436 | unsigned int log2_meta_chunk_width; |
437 | unsigned int log2_min_meta_chunk_bytes; |
438 | unsigned int min_meta_chunk_width; |
439 | unsigned int meta_chunk_width; |
440 | unsigned int meta_chunk_per_row_int; |
441 | unsigned int meta_row_remainder; |
442 | unsigned int meta_chunk_threshold; |
443 | unsigned int meta_blk_height; |
444 | unsigned int meta_surface_bytes; |
445 | unsigned int vmpg_bytes; |
446 | unsigned int meta_pte_req_per_frame_ub; |
447 | unsigned int meta_pte_bytes_per_frame_ub; |
448 | const unsigned int log2_vmpg_bytes = dml_log2(x: mode_lib->soc.gpuvm_min_page_size_bytes); |
449 | const bool dual_plane_en = is_dual_plane(source_format: (enum source_format_class) (source_format)); |
450 | const unsigned int dpte_buf_in_pte_reqs = |
451 | dual_plane_en ? (is_chroma ? mode_lib->ip.dpte_buffer_size_in_pte_reqs_chroma : mode_lib->ip.dpte_buffer_size_in_pte_reqs_luma) : (mode_lib->ip.dpte_buffer_size_in_pte_reqs_luma |
452 | + mode_lib->ip.dpte_buffer_size_in_pte_reqs_chroma); |
453 | |
454 | unsigned int log2_vmpg_height = 0; |
455 | unsigned int log2_vmpg_width = 0; |
456 | unsigned int log2_dpte_req_height_ptes = 0; |
457 | unsigned int log2_dpte_req_height = 0; |
458 | unsigned int log2_dpte_req_width = 0; |
459 | unsigned int log2_dpte_row_height_linear = 0; |
460 | unsigned int log2_dpte_row_height = 0; |
461 | unsigned int log2_dpte_group_width = 0; |
462 | unsigned int dpte_row_width_ub = 0; |
463 | unsigned int dpte_req_height = 0; |
464 | unsigned int dpte_req_width = 0; |
465 | unsigned int dpte_group_width = 0; |
466 | unsigned int log2_dpte_group_bytes = 0; |
467 | unsigned int log2_dpte_group_length = 0; |
468 | double byte_per_pixel_det_y; |
469 | double byte_per_pixel_det_c; |
470 | |
471 | CalculateBytePerPixelAnd256BBlockSizes( |
472 | SourcePixelFormat: (enum source_format_class) (source_format), |
473 | SurfaceTiling: (enum dm_swizzle_mode) (tiling), |
474 | BytePerPixelY: &bytes_per_element_y, |
475 | BytePerPixelC: &bytes_per_element_c, |
476 | BytePerPixelDETY: &byte_per_pixel_det_y, |
477 | BytePerPixelDETC: &byte_per_pixel_det_c, |
478 | BlockHeight256BytesY: &blk256_height_y, |
479 | BlockHeight256BytesC: &blk256_height_c, |
480 | BlockWidth256BytesY: &blk256_width_y, |
481 | BlockWidth256BytesC: &blk256_width_c); |
482 | |
483 | if (!is_chroma) { |
484 | blk256_width = blk256_width_y; |
485 | blk256_height = blk256_height_y; |
486 | bytes_per_element = bytes_per_element_y; |
487 | } else { |
488 | blk256_width = blk256_width_c; |
489 | blk256_height = blk256_height_c; |
490 | bytes_per_element = bytes_per_element_c; |
491 | } |
492 | |
493 | log2_bytes_per_element = dml_log2(x: bytes_per_element); |
494 | |
495 | dml_print("DML_DLG: %s: surf_linear = %d\n" , __func__, surf_linear); |
496 | dml_print("DML_DLG: %s: surf_vert = %d\n" , __func__, surf_vert); |
497 | dml_print("DML_DLG: %s: blk256_width = %d\n" , __func__, blk256_width); |
498 | dml_print("DML_DLG: %s: blk256_height = %d\n" , __func__, blk256_height); |
499 | |
500 | log2_blk256_width = dml_log2(x: (double) blk256_width); |
501 | log2_blk256_height = dml_log2(x: (double) blk256_height); |
502 | blk_bytes = surf_linear ? 256 : get_blk_size_bytes(tile_size: (enum source_macro_tile_size) macro_tile_size); |
503 | log2_blk_bytes = dml_log2(x: (double) blk_bytes); |
504 | log2_blk_height = 0; |
505 | log2_blk_width = 0; |
506 | |
507 | // remember log rule |
508 | // "+" in log is multiply |
509 | // "-" in log is divide |
510 | // "/2" is like square root |
511 | // blk is vertical biased |
512 | if (tiling != dm_sw_linear) |
513 | log2_blk_height = log2_blk256_height + dml_ceil(a: (double) (log2_blk_bytes - 8) / 2.0, granularity: 1); |
514 | else |
515 | log2_blk_height = 0; // blk height of 1 |
516 | |
517 | log2_blk_width = log2_blk_bytes - log2_bytes_per_element - log2_blk_height; |
518 | |
519 | if (!surf_vert) { |
520 | unsigned int temp; |
521 | |
522 | temp = dml_round_to_multiple(num: vp_width - 1, multiple: blk256_width, up: 1) + blk256_width; |
523 | if (data_pitch < blk256_width) { |
524 | dml_print("WARNING: DML_DLG: %s: swath_size calculation ignoring data_pitch=%u < blk256_width=%u\n" , __func__, data_pitch, blk256_width); |
525 | } else { |
526 | if (temp > data_pitch) { |
527 | if (data_pitch >= vp_width) |
528 | temp = data_pitch; |
529 | else |
530 | dml_print("WARNING: DML_DLG: %s: swath_size calculation ignoring data_pitch=%u < vp_width=%u\n" , __func__, data_pitch, vp_width); |
531 | } |
532 | } |
533 | rq_dlg_param->swath_width_ub = temp; |
534 | rq_dlg_param->req_per_swath_ub = temp >> log2_blk256_width; |
535 | } else { |
536 | unsigned int temp; |
537 | |
538 | temp = dml_round_to_multiple(num: vp_height - 1, multiple: blk256_height, up: 1) + blk256_height; |
539 | if (surface_height < blk256_height) { |
540 | dml_print("WARNING: DML_DLG: %s swath_size calculation ignored surface_height=%u < blk256_height=%u\n" , __func__, surface_height, blk256_height); |
541 | } else { |
542 | if (temp > surface_height) { |
543 | if (surface_height >= vp_height) |
544 | temp = surface_height; |
545 | else |
546 | dml_print("WARNING: DML_DLG: %s swath_size calculation ignored surface_height=%u < vp_height=%u\n" , __func__, surface_height, vp_height); |
547 | } |
548 | } |
549 | rq_dlg_param->swath_width_ub = temp; |
550 | rq_dlg_param->req_per_swath_ub = temp >> log2_blk256_height; |
551 | } |
552 | |
553 | if (!surf_vert) |
554 | rq_misc_param->full_swath_bytes = rq_dlg_param->swath_width_ub * blk256_height * bytes_per_element; |
555 | else |
556 | rq_misc_param->full_swath_bytes = rq_dlg_param->swath_width_ub * blk256_width * bytes_per_element; |
557 | |
558 | rq_misc_param->blk256_height = blk256_height; |
559 | rq_misc_param->blk256_width = blk256_width; |
560 | |
561 | // ------- |
562 | // meta |
563 | // ------- |
564 | log2_meta_req_bytes = 6; // meta request is 64b and is 8x8byte meta element |
565 | |
566 | // each 64b meta request for dcn is 8x8 meta elements and |
567 | // a meta element covers one 256b block of the data surface. |
568 | log2_meta_req_height = log2_blk256_height + 3; // meta req is 8x8 byte, each byte represent 1 blk256 |
569 | log2_meta_req_width = log2_meta_req_bytes + 8 - log2_bytes_per_element - log2_meta_req_height; |
570 | meta_req_width = 1 << log2_meta_req_width; |
571 | meta_req_height = 1 << log2_meta_req_height; |
572 | log2_meta_row_height = 0; |
573 | meta_row_width_ub = 0; |
574 | |
575 | // the dimensions of a meta row are meta_row_width x meta_row_height in elements. |
576 | // calculate upper bound of the meta_row_width |
577 | if (!surf_vert) { |
578 | log2_meta_row_height = log2_meta_req_height; |
579 | meta_row_width_ub = dml_round_to_multiple(num: vp_width - 1, multiple: meta_req_width, up: 1) + meta_req_width; |
580 | rq_dlg_param->meta_req_per_row_ub = meta_row_width_ub / meta_req_width; |
581 | } else { |
582 | log2_meta_row_height = log2_meta_req_width; |
583 | meta_row_width_ub = dml_round_to_multiple(num: vp_height - 1, multiple: meta_req_height, up: 1) + meta_req_height; |
584 | rq_dlg_param->meta_req_per_row_ub = meta_row_width_ub / meta_req_height; |
585 | } |
586 | rq_dlg_param->meta_bytes_per_row_ub = rq_dlg_param->meta_req_per_row_ub * 64; |
587 | |
588 | rq_dlg_param->meta_row_height = 1 << log2_meta_row_height; |
589 | |
590 | log2_meta_chunk_bytes = dml_log2(x: rq_sizing_param->meta_chunk_bytes); |
591 | log2_meta_chunk_height = log2_meta_row_height; |
592 | |
593 | //full sized meta chunk width in unit of data elements |
594 | log2_meta_chunk_width = log2_meta_chunk_bytes + 8 - log2_bytes_per_element - log2_meta_chunk_height; |
595 | log2_min_meta_chunk_bytes = dml_log2(x: rq_sizing_param->min_meta_chunk_bytes); |
596 | min_meta_chunk_width = 1 << (log2_min_meta_chunk_bytes + 8 - log2_bytes_per_element - log2_meta_chunk_height); |
597 | meta_chunk_width = 1 << log2_meta_chunk_width; |
598 | meta_chunk_per_row_int = (unsigned int) (meta_row_width_ub / meta_chunk_width); |
599 | meta_row_remainder = meta_row_width_ub % meta_chunk_width; |
600 | meta_chunk_threshold = 0; |
601 | meta_blk_height = blk256_height * 64; |
602 | meta_surface_bytes = meta_pitch * (dml_round_to_multiple(num: vp_height - 1, multiple: meta_blk_height, up: 1) + meta_blk_height) * bytes_per_element / 256; |
603 | vmpg_bytes = mode_lib->soc.gpuvm_min_page_size_bytes; |
604 | meta_pte_req_per_frame_ub = (dml_round_to_multiple(num: meta_surface_bytes - vmpg_bytes, multiple: 8 * vmpg_bytes, up: 1) + 8 * vmpg_bytes) / (8 * vmpg_bytes); |
605 | meta_pte_bytes_per_frame_ub = meta_pte_req_per_frame_ub * 64; //64B mpte request |
606 | rq_dlg_param->meta_pte_bytes_per_frame_ub = meta_pte_bytes_per_frame_ub; |
607 | |
608 | dml_print("DML_DLG: %s: meta_blk_height = %d\n" , __func__, meta_blk_height); |
609 | dml_print("DML_DLG: %s: meta_surface_bytes = %d\n" , __func__, meta_surface_bytes); |
610 | dml_print("DML_DLG: %s: meta_pte_req_per_frame_ub = %d\n" , __func__, meta_pte_req_per_frame_ub); |
611 | dml_print("DML_DLG: %s: meta_pte_bytes_per_frame_ub = %d\n" , __func__, meta_pte_bytes_per_frame_ub); |
612 | |
613 | if (!surf_vert) |
614 | meta_chunk_threshold = 2 * min_meta_chunk_width - meta_req_width; |
615 | else |
616 | meta_chunk_threshold = 2 * min_meta_chunk_width - meta_req_height; |
617 | |
618 | if (meta_row_remainder <= meta_chunk_threshold) |
619 | rq_dlg_param->meta_chunks_per_row_ub = meta_chunk_per_row_int + 1; |
620 | else |
621 | rq_dlg_param->meta_chunks_per_row_ub = meta_chunk_per_row_int + 2; |
622 | |
623 | // ------ |
624 | // dpte |
625 | // ------ |
626 | if (surf_linear) |
627 | log2_vmpg_height = 0; // one line high |
628 | else |
629 | log2_vmpg_height = (log2_vmpg_bytes - 8) / 2 + log2_blk256_height; |
630 | |
631 | log2_vmpg_width = log2_vmpg_bytes - log2_bytes_per_element - log2_vmpg_height; |
632 | |
633 | // only 3 possible shapes for dpte request in dimensions of ptes: 8x1, 4x2, 2x4. |
634 | if (surf_linear) { //one 64B PTE request returns 8 PTEs |
635 | log2_dpte_req_height_ptes = 0; |
636 | log2_dpte_req_width = log2_vmpg_width + 3; |
637 | log2_dpte_req_height = 0; |
638 | } else if (log2_blk_bytes == 12) { //4KB tile means 4kB page size |
639 | //one 64B req gives 8x1 PTEs for 4KB tile |
640 | log2_dpte_req_height_ptes = 0; |
641 | log2_dpte_req_width = log2_blk_width + 3; |
642 | log2_dpte_req_height = log2_blk_height + 0; |
643 | } else if ((log2_blk_bytes >= 16) && (log2_vmpg_bytes == 12)) { // tile block >= 64KB |
644 | //two 64B reqs of 2x4 PTEs give 16 PTEs to cover 64KB |
645 | log2_dpte_req_height_ptes = 4; |
646 | log2_dpte_req_width = log2_blk256_width + 4; // log2_64KB_width |
647 | log2_dpte_req_height = log2_blk256_height + 4; // log2_64KB_height |
648 | } else { //64KB page size and must 64KB tile block |
649 | //one 64B req gives 8x1 PTEs for 64KB tile |
650 | log2_dpte_req_height_ptes = 0; |
651 | log2_dpte_req_width = log2_blk_width + 3; |
652 | log2_dpte_req_height = log2_blk_height + 0; |
653 | } |
654 | |
655 | // The dpte request dimensions in data elements is dpte_req_width x dpte_req_height |
656 | // log2_vmpg_width is how much 1 pte represent, now calculating how much a 64b pte req represent |
657 | // That depends on the pte shape (i.e. 8x1, 4x2, 2x4) |
658 | //log2_dpte_req_height = log2_vmpg_height + log2_dpte_req_height_ptes; |
659 | //log2_dpte_req_width = log2_vmpg_width + log2_dpte_req_width_ptes; |
660 | dpte_req_height = 1 << log2_dpte_req_height; |
661 | dpte_req_width = 1 << log2_dpte_req_width; |
662 | |
663 | // calculate pitch dpte row buffer can hold |
664 | // round the result down to a power of two. |
665 | if (surf_linear) { |
666 | unsigned int dpte_row_height; |
667 | |
668 | log2_dpte_row_height_linear = dml_floor(a: dml_log2(x: dpte_buf_in_pte_reqs * dpte_req_width / data_pitch), granularity: 1); |
669 | |
670 | dml_print("DML_DLG: %s: is_chroma = %d\n" , __func__, is_chroma); |
671 | dml_print("DML_DLG: %s: dpte_buf_in_pte_reqs = %d\n" , __func__, dpte_buf_in_pte_reqs); |
672 | dml_print("DML_DLG: %s: log2_dpte_row_height_linear = %d\n" , __func__, log2_dpte_row_height_linear); |
673 | |
674 | ASSERT(log2_dpte_row_height_linear >= 3); |
675 | |
676 | if (log2_dpte_row_height_linear > 7) |
677 | log2_dpte_row_height_linear = 7; |
678 | |
679 | log2_dpte_row_height = log2_dpte_row_height_linear; |
680 | // For linear, the dpte row is pitch dependent and the pte requests wrap at the pitch boundary. |
681 | // the dpte_row_width_ub is the upper bound of data_pitch*dpte_row_height in elements with this unique buffering. |
682 | dpte_row_height = 1 << log2_dpte_row_height; |
683 | dpte_row_width_ub = dml_round_to_multiple(num: data_pitch * dpte_row_height - 1, multiple: dpte_req_width, up: 1) + dpte_req_width; |
684 | rq_dlg_param->dpte_req_per_row_ub = dpte_row_width_ub / dpte_req_width; |
685 | } else { |
686 | // the upper bound of the dpte_row_width without dependency on viewport position follows. |
687 | // for tiled mode, row height is the same as req height and row store up to vp size upper bound |
688 | if (!surf_vert) { |
689 | log2_dpte_row_height = log2_dpte_req_height; |
690 | dpte_row_width_ub = dml_round_to_multiple(num: vp_width - 1, multiple: dpte_req_width, up: 1) + dpte_req_width; |
691 | rq_dlg_param->dpte_req_per_row_ub = dpte_row_width_ub / dpte_req_width; |
692 | } else { |
693 | log2_dpte_row_height = (log2_blk_width < log2_dpte_req_width) ? log2_blk_width : log2_dpte_req_width; |
694 | dpte_row_width_ub = dml_round_to_multiple(num: vp_height - 1, multiple: dpte_req_height, up: 1) + dpte_req_height; |
695 | rq_dlg_param->dpte_req_per_row_ub = dpte_row_width_ub / dpte_req_height; |
696 | } |
697 | } |
698 | if (log2_blk_bytes >= 16 && log2_vmpg_bytes == 12) // tile block >= 64KB |
699 | rq_dlg_param->dpte_bytes_per_row_ub = rq_dlg_param->dpte_req_per_row_ub * 128; //2*64B dpte request |
700 | else |
701 | rq_dlg_param->dpte_bytes_per_row_ub = rq_dlg_param->dpte_req_per_row_ub * 64; //64B dpte request |
702 | |
703 | rq_dlg_param->dpte_row_height = 1 << log2_dpte_row_height; |
704 | |
705 | // the dpte_group_bytes is reduced for the specific case of vertical |
706 | // access of a tile surface that has dpte request of 8x1 ptes. |
707 | if (hostvm_enable) |
708 | rq_sizing_param->dpte_group_bytes = 512; |
709 | else { |
710 | if (!surf_linear & (log2_dpte_req_height_ptes == 0) & surf_vert) //reduced, in this case, will have page fault within a group |
711 | rq_sizing_param->dpte_group_bytes = 512; |
712 | else |
713 | rq_sizing_param->dpte_group_bytes = 2048; |
714 | } |
715 | |
716 | //since pte request size is 64byte, the number of data pte requests per full sized group is as follows. |
717 | log2_dpte_group_bytes = dml_log2(x: rq_sizing_param->dpte_group_bytes); |
718 | log2_dpte_group_length = log2_dpte_group_bytes - 6; //length in 64b requests |
719 | |
720 | // full sized data pte group width in elements |
721 | if (!surf_vert) |
722 | log2_dpte_group_width = log2_dpte_group_length + log2_dpte_req_width; |
723 | else |
724 | log2_dpte_group_width = log2_dpte_group_length + log2_dpte_req_height; |
725 | |
726 | //But if the tile block >=64KB and the page size is 4KB, then each dPTE request is 2*64B |
727 | if ((log2_blk_bytes >= 16) && (log2_vmpg_bytes == 12)) // tile block >= 64KB |
728 | log2_dpte_group_width = log2_dpte_group_width - 1; |
729 | |
730 | dpte_group_width = 1 << log2_dpte_group_width; |
731 | |
732 | // since dpte groups are only aligned to dpte_req_width and not dpte_group_width, |
733 | // the upper bound for the dpte groups per row is as follows. |
734 | rq_dlg_param->dpte_groups_per_row_ub = dml_ceil(a: (double) dpte_row_width_ub / dpte_group_width, granularity: 1); |
735 | } |
736 | |
737 | static void get_surf_rq_param( |
738 | struct display_mode_lib *mode_lib, |
739 | display_data_rq_sizing_params_st *rq_sizing_param, |
740 | display_data_rq_dlg_params_st *rq_dlg_param, |
741 | display_data_rq_misc_params_st *rq_misc_param, |
742 | const display_pipe_params_st *pipe_param, |
743 | bool is_chroma, |
744 | bool is_alpha) |
745 | { |
746 | bool mode_422 = 0; |
747 | unsigned int vp_width = 0; |
748 | unsigned int vp_height = 0; |
749 | unsigned int data_pitch = 0; |
750 | unsigned int meta_pitch = 0; |
751 | unsigned int surface_height = 0; |
752 | unsigned int ppe = mode_422 ? 2 : 1; |
753 | |
754 | // FIXME check if ppe apply for both luma and chroma in 422 case |
755 | if (is_chroma | is_alpha) { |
756 | vp_width = pipe_param->src.viewport_width_c / ppe; |
757 | vp_height = pipe_param->src.viewport_height_c; |
758 | data_pitch = pipe_param->src.data_pitch_c; |
759 | meta_pitch = pipe_param->src.meta_pitch_c; |
760 | surface_height = pipe_param->src.surface_height_y / 2.0; |
761 | } else { |
762 | vp_width = pipe_param->src.viewport_width / ppe; |
763 | vp_height = pipe_param->src.viewport_height; |
764 | data_pitch = pipe_param->src.data_pitch; |
765 | meta_pitch = pipe_param->src.meta_pitch; |
766 | surface_height = pipe_param->src.surface_height_y; |
767 | } |
768 | |
769 | if (pipe_param->dest.odm_combine) { |
770 | unsigned int access_dir; |
771 | unsigned int full_src_vp_width; |
772 | unsigned int hactive_odm; |
773 | unsigned int src_hactive_odm; |
774 | |
775 | access_dir = (pipe_param->src.source_scan == dm_vert); // vp access direction: horizontal or vertical accessed |
776 | hactive_odm = pipe_param->dest.hactive / ((unsigned int) pipe_param->dest.odm_combine * 2); |
777 | if (is_chroma) { |
778 | full_src_vp_width = pipe_param->scale_ratio_depth.hscl_ratio_c * pipe_param->dest.full_recout_width; |
779 | src_hactive_odm = pipe_param->scale_ratio_depth.hscl_ratio_c * hactive_odm; |
780 | } else { |
781 | full_src_vp_width = pipe_param->scale_ratio_depth.hscl_ratio * pipe_param->dest.full_recout_width; |
782 | src_hactive_odm = pipe_param->scale_ratio_depth.hscl_ratio * hactive_odm; |
783 | } |
784 | |
785 | if (access_dir == 0) { |
786 | vp_width = dml_min(a: full_src_vp_width, b: src_hactive_odm); |
787 | dml_print("DML_DLG: %s: vp_width = %d\n" , __func__, vp_width); |
788 | } else { |
789 | vp_height = dml_min(a: full_src_vp_width, b: src_hactive_odm); |
790 | dml_print("DML_DLG: %s: vp_height = %d\n" , __func__, vp_height); |
791 | |
792 | } |
793 | dml_print("DML_DLG: %s: full_src_vp_width = %d\n" , __func__, full_src_vp_width); |
794 | dml_print("DML_DLG: %s: hactive_odm = %d\n" , __func__, hactive_odm); |
795 | dml_print("DML_DLG: %s: src_hactive_odm = %d\n" , __func__, src_hactive_odm); |
796 | } |
797 | |
798 | rq_sizing_param->chunk_bytes = 8192; |
799 | |
800 | if (is_alpha) |
801 | rq_sizing_param->chunk_bytes = 4096; |
802 | |
803 | if (rq_sizing_param->chunk_bytes == 64 * 1024) |
804 | rq_sizing_param->min_chunk_bytes = 0; |
805 | else |
806 | rq_sizing_param->min_chunk_bytes = 1024; |
807 | |
808 | rq_sizing_param->meta_chunk_bytes = 2048; |
809 | rq_sizing_param->min_meta_chunk_bytes = 256; |
810 | |
811 | if (pipe_param->src.hostvm) |
812 | rq_sizing_param->mpte_group_bytes = 512; |
813 | else |
814 | rq_sizing_param->mpte_group_bytes = 2048; |
815 | |
816 | get_meta_and_pte_attr( |
817 | mode_lib, |
818 | rq_dlg_param, |
819 | rq_misc_param, |
820 | rq_sizing_param, |
821 | vp_width, |
822 | vp_height, |
823 | data_pitch, |
824 | meta_pitch, |
825 | source_format: pipe_param->src.source_format, |
826 | tiling: pipe_param->src.sw_mode, |
827 | macro_tile_size: pipe_param->src.macro_tile_size, |
828 | source_scan: pipe_param->src.source_scan, |
829 | hostvm_enable: pipe_param->src.hostvm, |
830 | is_chroma, |
831 | surface_height); |
832 | } |
833 | |
834 | static void 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) |
835 | { |
836 | // get param for luma surface |
837 | rq_param->yuv420 = pipe_param->src.source_format == dm_420_8 || pipe_param->src.source_format == dm_420_10 || pipe_param->src.source_format == dm_rgbe_alpha |
838 | || pipe_param->src.source_format == dm_420_12; |
839 | |
840 | rq_param->yuv420_10bpc = pipe_param->src.source_format == dm_420_10; |
841 | |
842 | rq_param->rgbe_alpha = (pipe_param->src.source_format == dm_rgbe_alpha) ? 1 : 0; |
843 | |
844 | get_surf_rq_param(mode_lib, rq_sizing_param: &(rq_param->sizing.rq_l), rq_dlg_param: &(rq_param->dlg.rq_l), rq_misc_param: &(rq_param->misc.rq_l), pipe_param, is_chroma: 0, is_alpha: 0); |
845 | |
846 | if (is_dual_plane(source_format: (enum source_format_class) (pipe_param->src.source_format))) { |
847 | // get param for chroma surface |
848 | get_surf_rq_param(mode_lib, rq_sizing_param: &(rq_param->sizing.rq_c), rq_dlg_param: &(rq_param->dlg.rq_c), rq_misc_param: &(rq_param->misc.rq_c), pipe_param, is_chroma: 1, is_alpha: rq_param->rgbe_alpha); |
849 | } |
850 | |
851 | // calculate how to split the det buffer space between luma and chroma |
852 | handle_det_buf_split(mode_lib, rq_param, pipe_src_param: &pipe_param->src); |
853 | print__rq_params_st(mode_lib, rq_param); |
854 | } |
855 | |
856 | void dml314_rq_dlg_get_rq_reg(struct display_mode_lib *mode_lib, display_rq_regs_st *rq_regs, const display_pipe_params_st *pipe_param) |
857 | { |
858 | display_rq_params_st rq_param = {0}; |
859 | |
860 | memset(rq_regs, 0, sizeof(*rq_regs)); |
861 | dml_rq_dlg_get_rq_params(mode_lib, rq_param: &rq_param, pipe_param); |
862 | extract_rq_regs(mode_lib, rq_regs, rq_param: &rq_param); |
863 | |
864 | print__rq_regs_st(mode_lib, rq_regs); |
865 | } |
866 | |
867 | static void calculate_ttu_cursor( |
868 | struct display_mode_lib *mode_lib, |
869 | double *refcyc_per_req_delivery_pre_cur, |
870 | double *refcyc_per_req_delivery_cur, |
871 | double refclk_freq_in_mhz, |
872 | double ref_freq_to_pix_freq, |
873 | double hscale_pixel_rate_l, |
874 | double hscl_ratio, |
875 | double vratio_pre_l, |
876 | double vratio_l, |
877 | unsigned int cur_width, |
878 | enum cursor_bpp cur_bpp) |
879 | { |
880 | unsigned int cur_src_width = cur_width; |
881 | unsigned int cur_req_size = 0; |
882 | unsigned int cur_req_width = 0; |
883 | double cur_width_ub = 0.0; |
884 | double cur_req_per_width = 0.0; |
885 | double hactive_cur = 0.0; |
886 | |
887 | ASSERT(cur_src_width <= 256); |
888 | |
889 | *refcyc_per_req_delivery_pre_cur = 0.0; |
890 | *refcyc_per_req_delivery_cur = 0.0; |
891 | if (cur_src_width > 0) { |
892 | unsigned int cur_bit_per_pixel = 0; |
893 | |
894 | if (cur_bpp == dm_cur_2bit) { |
895 | cur_req_size = 64; // byte |
896 | cur_bit_per_pixel = 2; |
897 | } else { // 32bit |
898 | cur_bit_per_pixel = 32; |
899 | if (cur_src_width >= 1 && cur_src_width <= 16) |
900 | cur_req_size = 64; |
901 | else if (cur_src_width >= 17 && cur_src_width <= 31) |
902 | cur_req_size = 128; |
903 | else |
904 | cur_req_size = 256; |
905 | } |
906 | |
907 | cur_req_width = (double) cur_req_size / ((double) cur_bit_per_pixel / 8.0); |
908 | cur_width_ub = dml_ceil(a: (double) cur_src_width / (double) cur_req_width, granularity: 1) * (double) cur_req_width; |
909 | cur_req_per_width = cur_width_ub / (double) cur_req_width; |
910 | hactive_cur = (double) cur_src_width / hscl_ratio; // FIXME: oswin to think about what to do for cursor |
911 | |
912 | if (vratio_pre_l <= 1.0) |
913 | *refcyc_per_req_delivery_pre_cur = hactive_cur * ref_freq_to_pix_freq / (double) cur_req_per_width; |
914 | else |
915 | *refcyc_per_req_delivery_pre_cur = (double) refclk_freq_in_mhz * (double) cur_src_width / hscale_pixel_rate_l / (double) cur_req_per_width; |
916 | |
917 | ASSERT(*refcyc_per_req_delivery_pre_cur < dml_pow(2, 13)); |
918 | |
919 | if (vratio_l <= 1.0) |
920 | *refcyc_per_req_delivery_cur = hactive_cur * ref_freq_to_pix_freq / (double) cur_req_per_width; |
921 | else |
922 | *refcyc_per_req_delivery_cur = (double) refclk_freq_in_mhz * (double) cur_src_width / hscale_pixel_rate_l / (double) cur_req_per_width; |
923 | |
924 | dml_print("DML_DLG: %s: cur_req_width = %d\n" , __func__, cur_req_width); |
925 | dml_print("DML_DLG: %s: cur_width_ub = %3.2f\n" , __func__, cur_width_ub); |
926 | dml_print("DML_DLG: %s: cur_req_per_width = %3.2f\n" , __func__, cur_req_per_width); |
927 | dml_print("DML_DLG: %s: hactive_cur = %3.2f\n" , __func__, hactive_cur); |
928 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_pre_cur = %3.2f\n" , __func__, *refcyc_per_req_delivery_pre_cur); |
929 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_cur = %3.2f\n" , __func__, *refcyc_per_req_delivery_cur); |
930 | |
931 | ASSERT(*refcyc_per_req_delivery_cur < dml_pow(2, 13)); |
932 | } |
933 | } |
934 | |
935 | // Note: currently taken in as is. |
936 | // Nice to decouple code from hw register implement and extract code that are repeated for luma and chroma. |
937 | static void dml_rq_dlg_get_dlg_params( |
938 | struct display_mode_lib *mode_lib, |
939 | const display_e2e_pipe_params_st *e2e_pipe_param, |
940 | const unsigned int num_pipes, |
941 | const unsigned int pipe_idx, |
942 | display_dlg_regs_st *disp_dlg_regs, |
943 | display_ttu_regs_st *disp_ttu_regs, |
944 | const display_rq_dlg_params_st *rq_dlg_param, |
945 | const display_dlg_sys_params_st *dlg_sys_param, |
946 | const bool cstate_en, |
947 | const bool pstate_en, |
948 | const bool vm_en, |
949 | const bool ignore_viewport_pos, |
950 | const bool immediate_flip_support) |
951 | { |
952 | const display_pipe_source_params_st *src = &e2e_pipe_param[pipe_idx].pipe.src; |
953 | const display_pipe_dest_params_st *dst = &e2e_pipe_param[pipe_idx].pipe.dest; |
954 | const display_clocks_and_cfg_st *clks = &e2e_pipe_param[pipe_idx].clks_cfg; |
955 | const scaler_ratio_depth_st *scl = &e2e_pipe_param[pipe_idx].pipe.scale_ratio_depth; |
956 | const scaler_taps_st *taps = &e2e_pipe_param[pipe_idx].pipe.scale_taps; |
957 | unsigned int pipe_index_in_combine[DC__NUM_PIPES__MAX]; |
958 | |
959 | // ------------------------- |
960 | // Section 1.15.2.1: OTG dependent Params |
961 | // ------------------------- |
962 | // Timing |
963 | unsigned int htotal = dst->htotal; |
964 | unsigned int hblank_end = dst->hblank_end; |
965 | unsigned int vblank_start = dst->vblank_start; |
966 | unsigned int vblank_end = dst->vblank_end; |
967 | |
968 | double dppclk_freq_in_mhz = clks->dppclk_mhz; |
969 | double refclk_freq_in_mhz = clks->refclk_mhz; |
970 | double pclk_freq_in_mhz = dst->pixel_rate_mhz; |
971 | bool interlaced = dst->interlaced; |
972 | double ref_freq_to_pix_freq = refclk_freq_in_mhz / pclk_freq_in_mhz; |
973 | double min_ttu_vblank; |
974 | unsigned int dlg_vblank_start; |
975 | bool dual_plane; |
976 | bool mode_422; |
977 | unsigned int access_dir; |
978 | unsigned int vp_height_l; |
979 | unsigned int vp_width_l; |
980 | unsigned int vp_height_c; |
981 | unsigned int vp_width_c; |
982 | |
983 | // Scaling |
984 | unsigned int htaps_l; |
985 | unsigned int htaps_c; |
986 | double hratio_l; |
987 | double hratio_c; |
988 | double vratio_l; |
989 | double vratio_c; |
990 | |
991 | unsigned int swath_width_ub_l; |
992 | unsigned int dpte_groups_per_row_ub_l; |
993 | unsigned int swath_width_ub_c; |
994 | unsigned int dpte_groups_per_row_ub_c; |
995 | |
996 | unsigned int meta_chunks_per_row_ub_l; |
997 | unsigned int meta_chunks_per_row_ub_c; |
998 | unsigned int vupdate_offset; |
999 | unsigned int vupdate_width; |
1000 | unsigned int vready_offset; |
1001 | |
1002 | unsigned int vstartup_start; |
1003 | unsigned int dst_x_after_scaler; |
1004 | unsigned int dst_y_after_scaler; |
1005 | double dst_y_prefetch; |
1006 | double dst_y_per_vm_vblank; |
1007 | double dst_y_per_row_vblank; |
1008 | double dst_y_per_vm_flip; |
1009 | double dst_y_per_row_flip; |
1010 | double max_dst_y_per_vm_vblank; |
1011 | double max_dst_y_per_row_vblank; |
1012 | double vratio_pre_l; |
1013 | double vratio_pre_c; |
1014 | unsigned int req_per_swath_ub_l; |
1015 | unsigned int req_per_swath_ub_c; |
1016 | unsigned int meta_row_height_l; |
1017 | unsigned int meta_row_height_c; |
1018 | unsigned int swath_width_pixels_ub_l; |
1019 | unsigned int swath_width_pixels_ub_c; |
1020 | unsigned int scaler_rec_in_width_l; |
1021 | unsigned int scaler_rec_in_width_c; |
1022 | unsigned int dpte_row_height_l; |
1023 | unsigned int dpte_row_height_c; |
1024 | double hscale_pixel_rate_l; |
1025 | double hscale_pixel_rate_c; |
1026 | double min_hratio_fact_l; |
1027 | double min_hratio_fact_c; |
1028 | double refcyc_per_line_delivery_pre_l; |
1029 | double refcyc_per_line_delivery_pre_c; |
1030 | double refcyc_per_line_delivery_l; |
1031 | double refcyc_per_line_delivery_c; |
1032 | |
1033 | double refcyc_per_req_delivery_pre_l; |
1034 | double refcyc_per_req_delivery_pre_c; |
1035 | double refcyc_per_req_delivery_l; |
1036 | double refcyc_per_req_delivery_c; |
1037 | |
1038 | unsigned int full_recout_width; |
1039 | double refcyc_per_req_delivery_pre_cur0; |
1040 | double refcyc_per_req_delivery_cur0; |
1041 | double refcyc_per_req_delivery_pre_cur1; |
1042 | double refcyc_per_req_delivery_cur1; |
1043 | unsigned int vba__min_dst_y_next_start = get_min_dst_y_next_start(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // FROM VBA |
1044 | unsigned int vba__vready_after_vcount0 = get_vready_at_or_after_vsync(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1045 | |
1046 | float vba__refcyc_per_line_delivery_pre_l = get_refcyc_per_line_delivery_pre_l_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1047 | float vba__refcyc_per_line_delivery_l = get_refcyc_per_line_delivery_l_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1048 | |
1049 | float vba__refcyc_per_req_delivery_pre_l = get_refcyc_per_req_delivery_pre_l_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1050 | float vba__refcyc_per_req_delivery_l = get_refcyc_per_req_delivery_l_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1051 | |
1052 | memset(disp_dlg_regs, 0, sizeof(*disp_dlg_regs)); |
1053 | memset(disp_ttu_regs, 0, sizeof(*disp_ttu_regs)); |
1054 | |
1055 | dml_print("DML_DLG: %s: cstate_en = %d\n" , __func__, cstate_en); |
1056 | dml_print("DML_DLG: %s: pstate_en = %d\n" , __func__, pstate_en); |
1057 | dml_print("DML_DLG: %s: vm_en = %d\n" , __func__, vm_en); |
1058 | dml_print("DML_DLG: %s: ignore_viewport_pos = %d\n" , __func__, ignore_viewport_pos); |
1059 | dml_print("DML_DLG: %s: immediate_flip_support = %d\n" , __func__, immediate_flip_support); |
1060 | |
1061 | dml_print("DML_DLG: %s: dppclk_freq_in_mhz = %3.2f\n" , __func__, dppclk_freq_in_mhz); |
1062 | dml_print("DML_DLG: %s: refclk_freq_in_mhz = %3.2f\n" , __func__, refclk_freq_in_mhz); |
1063 | dml_print("DML_DLG: %s: pclk_freq_in_mhz = %3.2f\n" , __func__, pclk_freq_in_mhz); |
1064 | dml_print("DML_DLG: %s: interlaced = %d\n" , __func__, interlaced); ASSERT(ref_freq_to_pix_freq < 4.0); |
1065 | |
1066 | disp_dlg_regs->ref_freq_to_pix_freq = (unsigned int) (ref_freq_to_pix_freq * dml_pow(a: 2, exp: 19)); |
1067 | disp_dlg_regs->refcyc_per_htotal = (unsigned int) (ref_freq_to_pix_freq * (double) htotal * dml_pow(a: 2, exp: 8)); |
1068 | disp_dlg_regs->dlg_vblank_end = interlaced ? (vblank_end / 2) : vblank_end; // 15 bits |
1069 | |
1070 | //set_prefetch_mode(mode_lib, cstate_en, pstate_en, ignore_viewport_pos, immediate_flip_support); |
1071 | min_ttu_vblank = get_min_ttu_vblank_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1072 | |
1073 | dlg_vblank_start = interlaced ? (vblank_start / 2) : vblank_start; |
1074 | disp_dlg_regs->min_dst_y_next_start_us = |
1075 | (vba__min_dst_y_next_start * dst->hactive) / (unsigned int) dst->pixel_rate_mhz; |
1076 | disp_dlg_regs->min_dst_y_next_start = vba__min_dst_y_next_start * dml_pow(a: 2, exp: 2); |
1077 | |
1078 | ASSERT(disp_dlg_regs->min_dst_y_next_start < (unsigned int)dml_pow(2, 18)); |
1079 | |
1080 | dml_print("DML_DLG: %s: min_ttu_vblank (us) = %3.2f\n" , __func__, min_ttu_vblank); |
1081 | dml_print("DML_DLG: %s: min_dst_y_next_start = 0x%0x\n" , __func__, disp_dlg_regs->min_dst_y_next_start); |
1082 | dml_print("DML_DLG: %s: dlg_vblank_start = 0x%0x\n" , __func__, dlg_vblank_start); |
1083 | dml_print("DML_DLG: %s: ref_freq_to_pix_freq = %3.2f\n" , __func__, ref_freq_to_pix_freq); |
1084 | dml_print("DML_DLG: %s: vba__min_dst_y_next_start = 0x%0x\n" , __func__, vba__min_dst_y_next_start); |
1085 | |
1086 | //old_impl_vs_vba_impl("min_dst_y_next_start", dlg_vblank_start, vba__min_dst_y_next_start); |
1087 | |
1088 | // ------------------------- |
1089 | // Section 1.15.2.2: Prefetch, Active and TTU |
1090 | // ------------------------- |
1091 | // Prefetch Calc |
1092 | // Source |
1093 | dual_plane = is_dual_plane(source_format: (enum source_format_class) (src->source_format)); |
1094 | mode_422 = 0; |
1095 | access_dir = (src->source_scan == dm_vert); // vp access direction: horizontal or vertical accessed |
1096 | vp_height_l = src->viewport_height; |
1097 | vp_width_l = src->viewport_width; |
1098 | vp_height_c = src->viewport_height_c; |
1099 | vp_width_c = src->viewport_width_c; |
1100 | |
1101 | // Scaling |
1102 | htaps_l = taps->htaps; |
1103 | htaps_c = taps->htaps_c; |
1104 | hratio_l = scl->hscl_ratio; |
1105 | hratio_c = scl->hscl_ratio_c; |
1106 | vratio_l = scl->vscl_ratio; |
1107 | vratio_c = scl->vscl_ratio_c; |
1108 | |
1109 | swath_width_ub_l = rq_dlg_param->rq_l.swath_width_ub; |
1110 | dpte_groups_per_row_ub_l = rq_dlg_param->rq_l.dpte_groups_per_row_ub; |
1111 | swath_width_ub_c = rq_dlg_param->rq_c.swath_width_ub; |
1112 | dpte_groups_per_row_ub_c = rq_dlg_param->rq_c.dpte_groups_per_row_ub; |
1113 | |
1114 | meta_chunks_per_row_ub_l = rq_dlg_param->rq_l.meta_chunks_per_row_ub; |
1115 | meta_chunks_per_row_ub_c = rq_dlg_param->rq_c.meta_chunks_per_row_ub; |
1116 | vupdate_offset = dst->vupdate_offset; |
1117 | vupdate_width = dst->vupdate_width; |
1118 | vready_offset = dst->vready_offset; |
1119 | |
1120 | vstartup_start = dst->vstartup_start; |
1121 | if (interlaced) { |
1122 | if (vstartup_start / 2.0 - (double) (vready_offset + vupdate_width + vupdate_offset) / htotal <= vblank_end / 2.0) |
1123 | disp_dlg_regs->vready_after_vcount0 = 1; |
1124 | else |
1125 | disp_dlg_regs->vready_after_vcount0 = 0; |
1126 | } else { |
1127 | if (vstartup_start - (double) (vready_offset + vupdate_width + vupdate_offset) / htotal <= vblank_end) |
1128 | disp_dlg_regs->vready_after_vcount0 = 1; |
1129 | else |
1130 | disp_dlg_regs->vready_after_vcount0 = 0; |
1131 | } |
1132 | |
1133 | dml_print("DML_DLG: %s: vready_after_vcount0 = %d\n" , __func__, disp_dlg_regs->vready_after_vcount0); |
1134 | dml_print("DML_DLG: %s: vba__vready_after_vcount0 = %d\n" , __func__, vba__vready_after_vcount0); |
1135 | //old_impl_vs_vba_impl("vready_after_vcount0", disp_dlg_regs->vready_after_vcount0, vba__vready_after_vcount0); |
1136 | |
1137 | if (interlaced) |
1138 | vstartup_start = vstartup_start / 2; |
1139 | |
1140 | dst_x_after_scaler = get_dst_x_after_scaler(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1141 | dst_y_after_scaler = get_dst_y_after_scaler(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1142 | |
1143 | // do some adjustment on the dst_after scaler to account for odm combine mode |
1144 | dml_print("DML_DLG: %s: input dst_x_after_scaler = %d\n" , __func__, dst_x_after_scaler); |
1145 | dml_print("DML_DLG: %s: input dst_y_after_scaler = %d\n" , __func__, dst_y_after_scaler); |
1146 | |
1147 | // need to figure out which side of odm combine we're in |
1148 | if (dst->odm_combine) { |
1149 | // figure out which pipes go together |
1150 | bool visited[DC__NUM_PIPES__MAX]; |
1151 | unsigned int i, j, k; |
1152 | |
1153 | for (k = 0; k < num_pipes; ++k) { |
1154 | visited[k] = false; |
1155 | pipe_index_in_combine[k] = 0; |
1156 | } |
1157 | |
1158 | for (i = 0; i < num_pipes; i++) { |
1159 | if (e2e_pipe_param[i].pipe.src.is_hsplit && !visited[i]) { |
1160 | |
1161 | unsigned int grp = e2e_pipe_param[i].pipe.src.hsplit_grp; |
1162 | unsigned int grp_idx = 0; |
1163 | |
1164 | for (j = i; j < num_pipes; j++) { |
1165 | if (e2e_pipe_param[j].pipe.src.hsplit_grp == grp && e2e_pipe_param[j].pipe.src.is_hsplit && !visited[j]) { |
1166 | pipe_index_in_combine[j] = grp_idx; |
1167 | dml_print("DML_DLG: %s: pipe[%d] is in grp %d idx %d\n" , __func__, j, grp, grp_idx); |
1168 | grp_idx++; |
1169 | visited[j] = true; |
1170 | } |
1171 | } |
1172 | } |
1173 | } |
1174 | |
1175 | } |
1176 | |
1177 | if (dst->odm_combine == dm_odm_combine_mode_disabled) { |
1178 | disp_dlg_regs->refcyc_h_blank_end = (unsigned int) ((double) hblank_end * ref_freq_to_pix_freq); |
1179 | } else { |
1180 | 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 |
1181 | unsigned int odm_pipe_index = pipe_index_in_combine[pipe_idx]; |
1182 | |
1183 | 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); |
1184 | } ASSERT(disp_dlg_regs->refcyc_h_blank_end < (unsigned int)dml_pow(2, 13)); |
1185 | |
1186 | dml_print("DML_DLG: %s: htotal = %d\n" , __func__, htotal); |
1187 | dml_print("DML_DLG: %s: dst_x_after_scaler[%d] = %d\n" , __func__, pipe_idx, dst_x_after_scaler); |
1188 | dml_print("DML_DLG: %s: dst_y_after_scaler[%d] = %d\n" , __func__, pipe_idx, dst_y_after_scaler); |
1189 | |
1190 | dst_y_prefetch = get_dst_y_prefetch(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1191 | dst_y_per_vm_vblank = get_dst_y_per_vm_vblank(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1192 | dst_y_per_row_vblank = get_dst_y_per_row_vblank(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1193 | dst_y_per_vm_flip = get_dst_y_per_vm_flip(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1194 | dst_y_per_row_flip = get_dst_y_per_row_flip(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1195 | |
1196 | max_dst_y_per_vm_vblank = 32.0; //U5.2 |
1197 | max_dst_y_per_row_vblank = 16.0; //U4.2 |
1198 | |
1199 | // magic! |
1200 | if (htotal <= 75) { |
1201 | max_dst_y_per_vm_vblank = 100.0; |
1202 | max_dst_y_per_row_vblank = 100.0; |
1203 | } |
1204 | |
1205 | dml_print("DML_DLG: %s: dst_y_prefetch (after rnd) = %3.2f\n" , __func__, dst_y_prefetch); |
1206 | dml_print("DML_DLG: %s: dst_y_per_vm_flip = %3.2f\n" , __func__, dst_y_per_vm_flip); |
1207 | dml_print("DML_DLG: %s: dst_y_per_row_flip = %3.2f\n" , __func__, dst_y_per_row_flip); |
1208 | dml_print("DML_DLG: %s: dst_y_per_vm_vblank = %3.2f\n" , __func__, dst_y_per_vm_vblank); |
1209 | dml_print("DML_DLG: %s: dst_y_per_row_vblank = %3.2f\n" , __func__, dst_y_per_row_vblank); |
1210 | |
1211 | ASSERT(dst_y_per_vm_vblank < max_dst_y_per_vm_vblank); ASSERT(dst_y_per_row_vblank < max_dst_y_per_row_vblank); |
1212 | |
1213 | ASSERT(dst_y_prefetch > (dst_y_per_vm_vblank + dst_y_per_row_vblank)); |
1214 | |
1215 | vratio_pre_l = get_vratio_prefetch_l(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1216 | vratio_pre_c = get_vratio_prefetch_c(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx); // From VBA |
1217 | |
1218 | dml_print("DML_DLG: %s: vratio_pre_l = %3.2f\n" , __func__, vratio_pre_l); |
1219 | dml_print("DML_DLG: %s: vratio_pre_c = %3.2f\n" , __func__, vratio_pre_c); |
1220 | |
1221 | // Active |
1222 | req_per_swath_ub_l = rq_dlg_param->rq_l.req_per_swath_ub; |
1223 | req_per_swath_ub_c = rq_dlg_param->rq_c.req_per_swath_ub; |
1224 | meta_row_height_l = rq_dlg_param->rq_l.meta_row_height; |
1225 | meta_row_height_c = rq_dlg_param->rq_c.meta_row_height; |
1226 | swath_width_pixels_ub_l = 0; |
1227 | swath_width_pixels_ub_c = 0; |
1228 | scaler_rec_in_width_l = 0; |
1229 | scaler_rec_in_width_c = 0; |
1230 | dpte_row_height_l = rq_dlg_param->rq_l.dpte_row_height; |
1231 | dpte_row_height_c = rq_dlg_param->rq_c.dpte_row_height; |
1232 | |
1233 | if (mode_422) { |
1234 | swath_width_pixels_ub_l = swath_width_ub_l * 2; // *2 for 2 pixel per element |
1235 | swath_width_pixels_ub_c = swath_width_ub_c * 2; |
1236 | } else { |
1237 | swath_width_pixels_ub_l = swath_width_ub_l * 1; |
1238 | swath_width_pixels_ub_c = swath_width_ub_c * 1; |
1239 | } |
1240 | |
1241 | hscale_pixel_rate_l = 0.; |
1242 | hscale_pixel_rate_c = 0.; |
1243 | min_hratio_fact_l = 1.0; |
1244 | min_hratio_fact_c = 1.0; |
1245 | |
1246 | if (hratio_l <= 1) |
1247 | min_hratio_fact_l = 2.0; |
1248 | else if (htaps_l <= 6) { |
1249 | if ((hratio_l * 2.0) > 4.0) |
1250 | min_hratio_fact_l = 4.0; |
1251 | else |
1252 | min_hratio_fact_l = hratio_l * 2.0; |
1253 | } else { |
1254 | if (hratio_l > 4.0) |
1255 | min_hratio_fact_l = 4.0; |
1256 | else |
1257 | min_hratio_fact_l = hratio_l; |
1258 | } |
1259 | |
1260 | hscale_pixel_rate_l = min_hratio_fact_l * dppclk_freq_in_mhz; |
1261 | |
1262 | dml_print("DML_DLG: %s: hratio_l = %3.2f\n" , __func__, hratio_l); |
1263 | dml_print("DML_DLG: %s: min_hratio_fact_l = %3.2f\n" , __func__, min_hratio_fact_l); |
1264 | dml_print("DML_DLG: %s: hscale_pixel_rate_l = %3.2f\n" , __func__, hscale_pixel_rate_l); |
1265 | |
1266 | if (hratio_c <= 1) |
1267 | min_hratio_fact_c = 2.0; |
1268 | else if (htaps_c <= 6) { |
1269 | if ((hratio_c * 2.0) > 4.0) |
1270 | min_hratio_fact_c = 4.0; |
1271 | else |
1272 | min_hratio_fact_c = hratio_c * 2.0; |
1273 | } else { |
1274 | if (hratio_c > 4.0) |
1275 | min_hratio_fact_c = 4.0; |
1276 | else |
1277 | min_hratio_fact_c = hratio_c; |
1278 | } |
1279 | |
1280 | hscale_pixel_rate_c = min_hratio_fact_c * dppclk_freq_in_mhz; |
1281 | |
1282 | refcyc_per_line_delivery_pre_l = 0.; |
1283 | refcyc_per_line_delivery_pre_c = 0.; |
1284 | refcyc_per_line_delivery_l = 0.; |
1285 | refcyc_per_line_delivery_c = 0.; |
1286 | |
1287 | refcyc_per_req_delivery_pre_l = 0.; |
1288 | refcyc_per_req_delivery_pre_c = 0.; |
1289 | refcyc_per_req_delivery_l = 0.; |
1290 | refcyc_per_req_delivery_c = 0.; |
1291 | |
1292 | full_recout_width = 0; |
1293 | // In ODM |
1294 | if (src->is_hsplit) { |
1295 | // This "hack" is only allowed (and valid) for MPC combine. In ODM |
1296 | // combine, you MUST specify the full_recout_width...according to Oswin |
1297 | if (dst->full_recout_width == 0 && !dst->odm_combine) { |
1298 | dml_print("DML_DLG: %s: Warning: full_recout_width not set in hsplit mode\n" , __func__); |
1299 | full_recout_width = dst->recout_width * 2; // assume half split for dcn1 |
1300 | } else |
1301 | full_recout_width = dst->full_recout_width; |
1302 | } else |
1303 | full_recout_width = dst->recout_width; |
1304 | |
1305 | // As of DCN2, mpc_combine and odm_combine are mutually exclusive |
1306 | refcyc_per_line_delivery_pre_l = get_refcyc_per_delivery( |
1307 | mode_lib, |
1308 | refclk_freq_in_mhz, |
1309 | pclk_freq_in_mhz, |
1310 | odm_combine: dst->odm_combine, |
1311 | recout_width: full_recout_width, |
1312 | hactive: dst->hactive, |
1313 | vratio: vratio_pre_l, |
1314 | hscale_pixel_rate: hscale_pixel_rate_l, |
1315 | delivery_width: swath_width_pixels_ub_l, |
1316 | req_per_swath_ub: 1); // per line |
1317 | |
1318 | refcyc_per_line_delivery_l = get_refcyc_per_delivery( |
1319 | mode_lib, |
1320 | refclk_freq_in_mhz, |
1321 | pclk_freq_in_mhz, |
1322 | odm_combine: dst->odm_combine, |
1323 | recout_width: full_recout_width, |
1324 | hactive: dst->hactive, |
1325 | vratio: vratio_l, |
1326 | hscale_pixel_rate: hscale_pixel_rate_l, |
1327 | delivery_width: swath_width_pixels_ub_l, |
1328 | req_per_swath_ub: 1); // per line |
1329 | |
1330 | dml_print("DML_DLG: %s: full_recout_width = %d\n" , __func__, full_recout_width); |
1331 | dml_print("DML_DLG: %s: hscale_pixel_rate_l = %3.2f\n" , __func__, hscale_pixel_rate_l); |
1332 | dml_print("DML_DLG: %s: refcyc_per_line_delivery_pre_l = %3.2f\n" , __func__, refcyc_per_line_delivery_pre_l); |
1333 | dml_print("DML_DLG: %s: refcyc_per_line_delivery_l = %3.2f\n" , __func__, refcyc_per_line_delivery_l); |
1334 | dml_print("DML_DLG: %s: vba__refcyc_per_line_delivery_pre_l = %3.2f\n" , __func__, vba__refcyc_per_line_delivery_pre_l); |
1335 | dml_print("DML_DLG: %s: vba__refcyc_per_line_delivery_l = %3.2f\n" , __func__, vba__refcyc_per_line_delivery_l); |
1336 | |
1337 | //old_impl_vs_vba_impl("refcyc_per_line_delivery_pre_l", refcyc_per_line_delivery_pre_l, vba__refcyc_per_line_delivery_pre_l); |
1338 | //old_impl_vs_vba_impl("refcyc_per_line_delivery_l", refcyc_per_line_delivery_l, vba__refcyc_per_line_delivery_l); |
1339 | |
1340 | if (dual_plane) { |
1341 | float vba__refcyc_per_line_delivery_pre_c = get_refcyc_per_line_delivery_pre_c_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1342 | float vba__refcyc_per_line_delivery_c = get_refcyc_per_line_delivery_c_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1343 | |
1344 | refcyc_per_line_delivery_pre_c = get_refcyc_per_delivery( |
1345 | mode_lib, |
1346 | refclk_freq_in_mhz, |
1347 | pclk_freq_in_mhz, |
1348 | odm_combine: dst->odm_combine, |
1349 | recout_width: full_recout_width, |
1350 | hactive: dst->hactive, |
1351 | vratio: vratio_pre_c, |
1352 | hscale_pixel_rate: hscale_pixel_rate_c, |
1353 | delivery_width: swath_width_pixels_ub_c, |
1354 | req_per_swath_ub: 1); // per line |
1355 | |
1356 | refcyc_per_line_delivery_c = get_refcyc_per_delivery( |
1357 | mode_lib, |
1358 | refclk_freq_in_mhz, |
1359 | pclk_freq_in_mhz, |
1360 | odm_combine: dst->odm_combine, |
1361 | recout_width: full_recout_width, |
1362 | hactive: dst->hactive, |
1363 | vratio: vratio_c, |
1364 | hscale_pixel_rate: hscale_pixel_rate_c, |
1365 | delivery_width: swath_width_pixels_ub_c, |
1366 | req_per_swath_ub: 1); // per line |
1367 | |
1368 | dml_print("DML_DLG: %s: refcyc_per_line_delivery_pre_c = %3.2f\n" , __func__, refcyc_per_line_delivery_pre_c); |
1369 | dml_print("DML_DLG: %s: refcyc_per_line_delivery_c = %3.2f\n" , __func__, refcyc_per_line_delivery_c); |
1370 | dml_print("DML_DLG: %s: vba__refcyc_per_line_delivery_pre_c = %3.2f\n" , __func__, vba__refcyc_per_line_delivery_pre_c); |
1371 | dml_print("DML_DLG: %s: vba__refcyc_per_line_delivery_c = %3.2f\n" , __func__, vba__refcyc_per_line_delivery_c); |
1372 | |
1373 | //old_impl_vs_vba_impl("refcyc_per_line_delivery_pre_c", refcyc_per_line_delivery_pre_c, vba__refcyc_per_line_delivery_pre_c); |
1374 | //old_impl_vs_vba_impl("refcyc_per_line_delivery_c", refcyc_per_line_delivery_c, vba__refcyc_per_line_delivery_c); |
1375 | } |
1376 | |
1377 | if (src->dynamic_metadata_enable && src->gpuvm) |
1378 | disp_dlg_regs->refcyc_per_vm_dmdata = get_refcyc_per_vm_dmdata_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1379 | |
1380 | disp_dlg_regs->dmdata_dl_delta = get_dmdata_dl_delta_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1381 | |
1382 | // TTU - Luma / Chroma |
1383 | if (access_dir) { // vertical access |
1384 | scaler_rec_in_width_l = vp_height_l; |
1385 | scaler_rec_in_width_c = vp_height_c; |
1386 | } else { |
1387 | scaler_rec_in_width_l = vp_width_l; |
1388 | scaler_rec_in_width_c = vp_width_c; |
1389 | } |
1390 | |
1391 | refcyc_per_req_delivery_pre_l = get_refcyc_per_delivery( |
1392 | mode_lib, |
1393 | refclk_freq_in_mhz, |
1394 | pclk_freq_in_mhz, |
1395 | odm_combine: dst->odm_combine, |
1396 | recout_width: full_recout_width, |
1397 | hactive: dst->hactive, |
1398 | vratio: vratio_pre_l, |
1399 | hscale_pixel_rate: hscale_pixel_rate_l, |
1400 | delivery_width: scaler_rec_in_width_l, |
1401 | req_per_swath_ub: req_per_swath_ub_l); // per req |
1402 | |
1403 | refcyc_per_req_delivery_l = get_refcyc_per_delivery( |
1404 | mode_lib, |
1405 | refclk_freq_in_mhz, |
1406 | pclk_freq_in_mhz, |
1407 | odm_combine: dst->odm_combine, |
1408 | recout_width: full_recout_width, |
1409 | hactive: dst->hactive, |
1410 | vratio: vratio_l, |
1411 | hscale_pixel_rate: hscale_pixel_rate_l, |
1412 | delivery_width: scaler_rec_in_width_l, |
1413 | req_per_swath_ub: req_per_swath_ub_l); // per req |
1414 | |
1415 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_pre_l = %3.2f\n" , __func__, refcyc_per_req_delivery_pre_l); |
1416 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_l = %3.2f\n" , __func__, refcyc_per_req_delivery_l); |
1417 | dml_print("DML_DLG: %s: vba__refcyc_per_req_delivery_pre_l = %3.2f\n" , __func__, vba__refcyc_per_req_delivery_pre_l); |
1418 | dml_print("DML_DLG: %s: vba__refcyc_per_req_delivery_l = %3.2f\n" , __func__, vba__refcyc_per_req_delivery_l); |
1419 | |
1420 | //old_impl_vs_vba_impl("refcyc_per_req_delivery_pre_l", refcyc_per_req_delivery_pre_l, vba__refcyc_per_req_delivery_pre_l); |
1421 | //old_impl_vs_vba_impl("refcyc_per_req_delivery_l", refcyc_per_req_delivery_l, vba__refcyc_per_req_delivery_l); |
1422 | |
1423 | ASSERT(refcyc_per_req_delivery_pre_l < dml_pow(2, 13)); ASSERT(refcyc_per_req_delivery_l < dml_pow(2, 13)); |
1424 | |
1425 | if (dual_plane) { |
1426 | float vba__refcyc_per_req_delivery_pre_c = get_refcyc_per_req_delivery_pre_c_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1427 | float vba__refcyc_per_req_delivery_c = get_refcyc_per_req_delivery_c_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1428 | |
1429 | refcyc_per_req_delivery_pre_c = get_refcyc_per_delivery( |
1430 | mode_lib, |
1431 | refclk_freq_in_mhz, |
1432 | pclk_freq_in_mhz, |
1433 | odm_combine: dst->odm_combine, |
1434 | recout_width: full_recout_width, |
1435 | hactive: dst->hactive, |
1436 | vratio: vratio_pre_c, |
1437 | hscale_pixel_rate: hscale_pixel_rate_c, |
1438 | delivery_width: scaler_rec_in_width_c, |
1439 | req_per_swath_ub: req_per_swath_ub_c); // per req |
1440 | refcyc_per_req_delivery_c = get_refcyc_per_delivery( |
1441 | mode_lib, |
1442 | refclk_freq_in_mhz, |
1443 | pclk_freq_in_mhz, |
1444 | odm_combine: dst->odm_combine, |
1445 | recout_width: full_recout_width, |
1446 | hactive: dst->hactive, |
1447 | vratio: vratio_c, |
1448 | hscale_pixel_rate: hscale_pixel_rate_c, |
1449 | delivery_width: scaler_rec_in_width_c, |
1450 | req_per_swath_ub: req_per_swath_ub_c); // per req |
1451 | |
1452 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_pre_c = %3.2f\n" , __func__, refcyc_per_req_delivery_pre_c); |
1453 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_c = %3.2f\n" , __func__, refcyc_per_req_delivery_c); |
1454 | dml_print("DML_DLG: %s: vba__refcyc_per_req_delivery_pre_c = %3.2f\n" , __func__, vba__refcyc_per_req_delivery_pre_c); |
1455 | dml_print("DML_DLG: %s: vba__refcyc_per_req_delivery_c = %3.2f\n" , __func__, vba__refcyc_per_req_delivery_c); |
1456 | |
1457 | //old_impl_vs_vba_impl("refcyc_per_req_delivery_pre_c", refcyc_per_req_delivery_pre_c, vba__refcyc_per_req_delivery_pre_c); |
1458 | //old_impl_vs_vba_impl("refcyc_per_req_delivery_c", refcyc_per_req_delivery_c, vba__refcyc_per_req_delivery_c); |
1459 | |
1460 | ASSERT(refcyc_per_req_delivery_pre_c < dml_pow(2, 13)); ASSERT(refcyc_per_req_delivery_c < dml_pow(2, 13)); |
1461 | } |
1462 | |
1463 | // TTU - Cursor |
1464 | refcyc_per_req_delivery_pre_cur0 = 0.0; |
1465 | refcyc_per_req_delivery_cur0 = 0.0; |
1466 | |
1467 | ASSERT(src->num_cursors <= 1); |
1468 | |
1469 | if (src->num_cursors > 0) { |
1470 | float vba__refcyc_per_req_delivery_pre_cur0; |
1471 | float vba__refcyc_per_req_delivery_cur0; |
1472 | |
1473 | calculate_ttu_cursor( |
1474 | mode_lib, |
1475 | refcyc_per_req_delivery_pre_cur: &refcyc_per_req_delivery_pre_cur0, |
1476 | refcyc_per_req_delivery_cur: &refcyc_per_req_delivery_cur0, |
1477 | refclk_freq_in_mhz, |
1478 | ref_freq_to_pix_freq, |
1479 | hscale_pixel_rate_l, |
1480 | hscl_ratio: scl->hscl_ratio, |
1481 | vratio_pre_l, |
1482 | vratio_l, |
1483 | cur_width: src->cur0_src_width, |
1484 | cur_bpp: (enum cursor_bpp) (src->cur0_bpp)); |
1485 | |
1486 | vba__refcyc_per_req_delivery_pre_cur0 = get_refcyc_per_cursor_req_delivery_pre_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1487 | vba__refcyc_per_req_delivery_cur0 = get_refcyc_per_cursor_req_delivery_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1488 | |
1489 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_pre_cur0 = %3.2f\n" , __func__, refcyc_per_req_delivery_pre_cur0); |
1490 | dml_print("DML_DLG: %s: refcyc_per_req_delivery_cur0 = %3.2f\n" , __func__, refcyc_per_req_delivery_cur0); |
1491 | dml_print("DML_DLG: %s: vba__refcyc_per_req_delivery_pre_cur0 = %3.2f\n" , __func__, vba__refcyc_per_req_delivery_pre_cur0); |
1492 | dml_print("DML_DLG: %s: vba__refcyc_per_req_delivery_cur0 = %3.2f\n" , __func__, vba__refcyc_per_req_delivery_cur0); |
1493 | |
1494 | //old_impl_vs_vba_impl("refcyc_per_req_delivery_pre_cur0", refcyc_per_req_delivery_pre_cur0, vba__refcyc_per_req_delivery_pre_cur0); |
1495 | //old_impl_vs_vba_impl("refcyc_per_req_delivery_cur0", refcyc_per_req_delivery_cur0, vba__refcyc_per_req_delivery_cur0); |
1496 | } |
1497 | |
1498 | refcyc_per_req_delivery_pre_cur1 = 0.0; |
1499 | refcyc_per_req_delivery_cur1 = 0.0; |
1500 | |
1501 | // TTU - Misc |
1502 | // all hard-coded |
1503 | |
1504 | // Assignment to register structures |
1505 | disp_dlg_regs->dst_y_after_scaler = dst_y_after_scaler; // in terms of line |
1506 | ASSERT(disp_dlg_regs->dst_y_after_scaler < 8); |
1507 | disp_dlg_regs->refcyc_x_after_scaler = dst_x_after_scaler * ref_freq_to_pix_freq; // in terms of refclk |
1508 | ASSERT(disp_dlg_regs->refcyc_x_after_scaler < (unsigned int)dml_pow(2, 13)); |
1509 | disp_dlg_regs->dst_y_prefetch = (unsigned int) (dst_y_prefetch * dml_pow(a: 2, exp: 2)); |
1510 | disp_dlg_regs->dst_y_per_vm_vblank = (unsigned int) (dst_y_per_vm_vblank * dml_pow(a: 2, exp: 2)); |
1511 | disp_dlg_regs->dst_y_per_row_vblank = (unsigned int) (dst_y_per_row_vblank * dml_pow(a: 2, exp: 2)); |
1512 | disp_dlg_regs->dst_y_per_vm_flip = (unsigned int) (dst_y_per_vm_flip * dml_pow(a: 2, exp: 2)); |
1513 | disp_dlg_regs->dst_y_per_row_flip = (unsigned int) (dst_y_per_row_flip * dml_pow(a: 2, exp: 2)); |
1514 | |
1515 | disp_dlg_regs->vratio_prefetch = (unsigned int) (vratio_pre_l * dml_pow(a: 2, exp: 19)); |
1516 | disp_dlg_regs->vratio_prefetch_c = (unsigned int) (vratio_pre_c * dml_pow(a: 2, exp: 19)); |
1517 | |
1518 | 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); |
1519 | 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); |
1520 | 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); |
1521 | 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); |
1522 | |
1523 | disp_dlg_regs->refcyc_per_pte_group_vblank_l = (unsigned int) (dst_y_per_row_vblank * (double) htotal * ref_freq_to_pix_freq / (double) dpte_groups_per_row_ub_l); |
1524 | ASSERT(disp_dlg_regs->refcyc_per_pte_group_vblank_l < (unsigned int)dml_pow(2, 13)); |
1525 | |
1526 | if (dual_plane) { |
1527 | disp_dlg_regs->refcyc_per_pte_group_vblank_c = (unsigned int) (dst_y_per_row_vblank * (double) htotal * ref_freq_to_pix_freq / (double) dpte_groups_per_row_ub_c); |
1528 | ASSERT(disp_dlg_regs->refcyc_per_pte_group_vblank_c < (unsigned int)dml_pow(2, 13)); |
1529 | } |
1530 | |
1531 | disp_dlg_regs->refcyc_per_meta_chunk_vblank_l = (unsigned int) (dst_y_per_row_vblank * (double) htotal * ref_freq_to_pix_freq / (double) meta_chunks_per_row_ub_l); |
1532 | ASSERT(disp_dlg_regs->refcyc_per_meta_chunk_vblank_l < (unsigned int)dml_pow(2, 13)); |
1533 | |
1534 | disp_dlg_regs->refcyc_per_meta_chunk_vblank_c = 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 |
1535 | |
1536 | disp_dlg_regs->refcyc_per_pte_group_flip_l = (unsigned int) (dst_y_per_row_flip * htotal * ref_freq_to_pix_freq) / dpte_groups_per_row_ub_l; |
1537 | disp_dlg_regs->refcyc_per_meta_chunk_flip_l = (unsigned int) (dst_y_per_row_flip * htotal * ref_freq_to_pix_freq) / meta_chunks_per_row_ub_l; |
1538 | |
1539 | if (dual_plane) { |
1540 | disp_dlg_regs->refcyc_per_pte_group_flip_c = (unsigned int) (dst_y_per_row_flip * htotal * ref_freq_to_pix_freq) / dpte_groups_per_row_ub_c; |
1541 | disp_dlg_regs->refcyc_per_meta_chunk_flip_c = (unsigned int) (dst_y_per_row_flip * htotal * ref_freq_to_pix_freq) / meta_chunks_per_row_ub_c; |
1542 | } |
1543 | |
1544 | disp_dlg_regs->refcyc_per_vm_group_vblank = get_refcyc_per_vm_group_vblank_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1545 | disp_dlg_regs->refcyc_per_vm_group_flip = get_refcyc_per_vm_group_flip_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz; // From VBA |
1546 | disp_dlg_regs->refcyc_per_vm_req_vblank = get_refcyc_per_vm_req_vblank_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz * dml_pow(a: 2, exp: 10); // From VBA |
1547 | disp_dlg_regs->refcyc_per_vm_req_flip = get_refcyc_per_vm_req_flip_in_us(mode_lib, pipes: e2e_pipe_param, num_pipes, which_pipe: pipe_idx) * refclk_freq_in_mhz * dml_pow(a: 2, exp: 10); // From VBA |
1548 | |
1549 | // Clamp to max for now |
1550 | if (disp_dlg_regs->refcyc_per_vm_group_vblank >= (unsigned int) dml_pow(a: 2, exp: 23)) |
1551 | disp_dlg_regs->refcyc_per_vm_group_vblank = dml_pow(a: 2, exp: 23) - 1; |
1552 | |
1553 | if (disp_dlg_regs->refcyc_per_vm_group_flip >= (unsigned int) dml_pow(a: 2, exp: 23)) |
1554 | disp_dlg_regs->refcyc_per_vm_group_flip = dml_pow(a: 2, exp: 23) - 1; |
1555 | |
1556 | if (disp_dlg_regs->refcyc_per_vm_req_vblank >= (unsigned int) dml_pow(a: 2, exp: 23)) |
1557 | disp_dlg_regs->refcyc_per_vm_req_vblank = dml_pow(a: 2, exp: 23) - 1; |
1558 | |
1559 | if (disp_dlg_regs->refcyc_per_vm_req_flip >= (unsigned int) dml_pow(a: 2, exp: 23)) |
1560 | disp_dlg_regs->refcyc_per_vm_req_flip = dml_pow(a: 2, exp: 23) - 1; |
1561 | |
1562 | disp_dlg_regs->dst_y_per_pte_row_nom_l = (unsigned int) ((double) dpte_row_height_l / (double) vratio_l * dml_pow(a: 2, exp: 2)); |
1563 | ASSERT(disp_dlg_regs->dst_y_per_pte_row_nom_l < (unsigned int)dml_pow(2, 17)); |
1564 | if (dual_plane) { |
1565 | disp_dlg_regs->dst_y_per_pte_row_nom_c = (unsigned int) ((double) dpte_row_height_c / (double) vratio_c * dml_pow(a: 2, exp: 2)); |
1566 | if (disp_dlg_regs->dst_y_per_pte_row_nom_c >= (unsigned int) dml_pow(a: 2, exp: 17)) { |
1567 | dml_print( |
1568 | "DML_DLG: %s: Warning dst_y_per_pte_row_nom_c %u larger than supported by register format U15.2 %u\n" , |
1569 | __func__, |
1570 | disp_dlg_regs->dst_y_per_pte_row_nom_c, |
1571 | (unsigned int) dml_pow(2, 17) - 1); |
1572 | } |
1573 | } |
1574 | |
1575 | disp_dlg_regs->dst_y_per_meta_row_nom_l = (unsigned int) ((double) meta_row_height_l / (double) vratio_l * dml_pow(a: 2, exp: 2)); |
1576 | ASSERT(disp_dlg_regs->dst_y_per_meta_row_nom_l < (unsigned int)dml_pow(2, 17)); |
1577 | |
1578 | disp_dlg_regs->dst_y_per_meta_row_nom_c = (unsigned int) ((double) meta_row_height_c / (double) vratio_c * dml_pow(a: 2, exp: 2)); |
1579 | ASSERT(disp_dlg_regs->dst_y_per_meta_row_nom_c < (unsigned int)dml_pow(2, 17)); |
1580 | |
1581 | disp_dlg_regs->refcyc_per_pte_group_nom_l = (unsigned int) ((double) dpte_row_height_l / (double) vratio_l * (double) htotal * ref_freq_to_pix_freq |
1582 | / (double) dpte_groups_per_row_ub_l); |
1583 | if (disp_dlg_regs->refcyc_per_pte_group_nom_l >= (unsigned int) dml_pow(a: 2, exp: 23)) |
1584 | disp_dlg_regs->refcyc_per_pte_group_nom_l = dml_pow(a: 2, exp: 23) - 1; |
1585 | disp_dlg_regs->refcyc_per_meta_chunk_nom_l = (unsigned int) ((double) meta_row_height_l / (double) vratio_l * (double) htotal * ref_freq_to_pix_freq |
1586 | / (double) meta_chunks_per_row_ub_l); |
1587 | if (disp_dlg_regs->refcyc_per_meta_chunk_nom_l >= (unsigned int) dml_pow(a: 2, exp: 23)) |
1588 | disp_dlg_regs->refcyc_per_meta_chunk_nom_l = dml_pow(a: 2, exp: 23) - 1; |
1589 | |
1590 | if (dual_plane) { |
1591 | disp_dlg_regs->refcyc_per_pte_group_nom_c = (unsigned int) ((double) dpte_row_height_c / (double) vratio_c * (double) htotal * ref_freq_to_pix_freq |
1592 | / (double) dpte_groups_per_row_ub_c); |
1593 | if (disp_dlg_regs->refcyc_per_pte_group_nom_c >= (unsigned int) dml_pow(a: 2, exp: 23)) |
1594 | disp_dlg_regs->refcyc_per_pte_group_nom_c = dml_pow(a: 2, exp: 23) - 1; |
1595 | |
1596 | // TODO: Is this the right calculation? Does htotal need to be halved? |
1597 | disp_dlg_regs->refcyc_per_meta_chunk_nom_c = (unsigned int) ((double) meta_row_height_c / (double) vratio_c * (double) htotal * ref_freq_to_pix_freq |
1598 | / (double) meta_chunks_per_row_ub_c); |
1599 | if (disp_dlg_regs->refcyc_per_meta_chunk_nom_c >= (unsigned int) dml_pow(a: 2, exp: 23)) |
1600 | disp_dlg_regs->refcyc_per_meta_chunk_nom_c = dml_pow(a: 2, exp: 23) - 1; |
1601 | } |
1602 | |
1603 | disp_dlg_regs->refcyc_per_line_delivery_pre_l = (unsigned int) dml_floor(a: refcyc_per_line_delivery_pre_l, granularity: 1); |
1604 | disp_dlg_regs->refcyc_per_line_delivery_l = (unsigned int) dml_floor(a: refcyc_per_line_delivery_l, granularity: 1); |
1605 | ASSERT(disp_dlg_regs->refcyc_per_line_delivery_pre_l < (unsigned int)dml_pow(2, 13)); ASSERT(disp_dlg_regs->refcyc_per_line_delivery_l < (unsigned int)dml_pow(2, 13)); |
1606 | |
1607 | disp_dlg_regs->refcyc_per_line_delivery_pre_c = (unsigned int) dml_floor(a: refcyc_per_line_delivery_pre_c, granularity: 1); |
1608 | disp_dlg_regs->refcyc_per_line_delivery_c = (unsigned int) dml_floor(a: refcyc_per_line_delivery_c, granularity: 1); |
1609 | ASSERT(disp_dlg_regs->refcyc_per_line_delivery_pre_c < (unsigned int)dml_pow(2, 13)); ASSERT(disp_dlg_regs->refcyc_per_line_delivery_c < (unsigned int)dml_pow(2, 13)); |
1610 | |
1611 | disp_dlg_regs->chunk_hdl_adjust_cur0 = 3; |
1612 | disp_dlg_regs->dst_y_offset_cur0 = 0; |
1613 | disp_dlg_regs->chunk_hdl_adjust_cur1 = 3; |
1614 | disp_dlg_regs->dst_y_offset_cur1 = 0; |
1615 | |
1616 | disp_dlg_regs->dst_y_delta_drq_limit = 0x7fff; // off |
1617 | |
1618 | disp_ttu_regs->refcyc_per_req_delivery_pre_l = (unsigned int) (refcyc_per_req_delivery_pre_l * dml_pow(a: 2, exp: 10)); |
1619 | disp_ttu_regs->refcyc_per_req_delivery_l = (unsigned int) (refcyc_per_req_delivery_l * dml_pow(a: 2, exp: 10)); |
1620 | disp_ttu_regs->refcyc_per_req_delivery_pre_c = (unsigned int) (refcyc_per_req_delivery_pre_c * dml_pow(a: 2, exp: 10)); |
1621 | disp_ttu_regs->refcyc_per_req_delivery_c = (unsigned int) (refcyc_per_req_delivery_c * dml_pow(a: 2, exp: 10)); |
1622 | disp_ttu_regs->refcyc_per_req_delivery_pre_cur0 = (unsigned int) (refcyc_per_req_delivery_pre_cur0 * dml_pow(a: 2, exp: 10)); |
1623 | disp_ttu_regs->refcyc_per_req_delivery_cur0 = (unsigned int) (refcyc_per_req_delivery_cur0 * dml_pow(a: 2, exp: 10)); |
1624 | disp_ttu_regs->refcyc_per_req_delivery_pre_cur1 = (unsigned int) (refcyc_per_req_delivery_pre_cur1 * dml_pow(a: 2, exp: 10)); |
1625 | disp_ttu_regs->refcyc_per_req_delivery_cur1 = (unsigned int) (refcyc_per_req_delivery_cur1 * dml_pow(a: 2, exp: 10)); |
1626 | |
1627 | disp_ttu_regs->qos_level_low_wm = 0; |
1628 | ASSERT(disp_ttu_regs->qos_level_low_wm < dml_pow(2, 14)); |
1629 | |
1630 | disp_ttu_regs->qos_level_high_wm = (unsigned int) (4.0 * (double) htotal * ref_freq_to_pix_freq); |
1631 | ASSERT(disp_ttu_regs->qos_level_high_wm < dml_pow(2, 14)); |
1632 | |
1633 | disp_ttu_regs->qos_level_flip = 14; |
1634 | disp_ttu_regs->qos_level_fixed_l = 8; |
1635 | disp_ttu_regs->qos_level_fixed_c = 8; |
1636 | disp_ttu_regs->qos_level_fixed_cur0 = 8; |
1637 | disp_ttu_regs->qos_ramp_disable_l = 0; |
1638 | disp_ttu_regs->qos_ramp_disable_c = 0; |
1639 | disp_ttu_regs->qos_ramp_disable_cur0 = 0; |
1640 | |
1641 | disp_ttu_regs->min_ttu_vblank = min_ttu_vblank * refclk_freq_in_mhz; |
1642 | ASSERT(disp_ttu_regs->min_ttu_vblank < dml_pow(2, 24)); |
1643 | |
1644 | print__ttu_regs_st(mode_lib, ttu_regs: disp_ttu_regs); |
1645 | print__dlg_regs_st(mode_lib, dlg_regs: disp_dlg_regs); |
1646 | } |
1647 | |
1648 | void dml314_rq_dlg_get_dlg_reg( |
1649 | struct display_mode_lib *mode_lib, |
1650 | display_dlg_regs_st *dlg_regs, |
1651 | display_ttu_regs_st *ttu_regs, |
1652 | const display_e2e_pipe_params_st *e2e_pipe_param, |
1653 | const unsigned int num_pipes, |
1654 | const unsigned int pipe_idx, |
1655 | const bool cstate_en, |
1656 | const bool pstate_en, |
1657 | const bool vm_en, |
1658 | const bool ignore_viewport_pos, |
1659 | const bool immediate_flip_support) |
1660 | { |
1661 | display_rq_params_st rq_param = {0}; |
1662 | display_dlg_sys_params_st dlg_sys_param = {0}; |
1663 | |
1664 | // Get watermark and Tex. |
1665 | dlg_sys_param.t_urg_wm_us = get_wm_urgent(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1666 | dlg_sys_param.deepsleep_dcfclk_mhz = get_clk_dcf_deepsleep(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1667 | dlg_sys_param.t_extra_us = get_urgent_extra_latency(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1668 | dlg_sys_param.mem_trip_us = get_wm_memory_trip(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1669 | dlg_sys_param.t_mclk_wm_us = get_wm_dram_clock_change(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1670 | dlg_sys_param.t_sr_wm_us = get_wm_stutter_enter_exit(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1671 | dlg_sys_param.total_flip_bw = get_total_immediate_flip_bw(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1672 | dlg_sys_param.total_flip_bytes = get_total_immediate_flip_bytes(mode_lib, pipes: e2e_pipe_param, num_pipes); |
1673 | |
1674 | print__dlg_sys_params_st(mode_lib, dlg_sys_param: &dlg_sys_param); |
1675 | |
1676 | // system parameter calculation done |
1677 | |
1678 | dml_print("DML_DLG: Calculation for pipe[%d] start\n\n" , pipe_idx); |
1679 | dml_rq_dlg_get_rq_params(mode_lib, rq_param: &rq_param, pipe_param: &e2e_pipe_param[pipe_idx].pipe); |
1680 | dml_rq_dlg_get_dlg_params( |
1681 | mode_lib, |
1682 | e2e_pipe_param, |
1683 | num_pipes, |
1684 | pipe_idx, |
1685 | disp_dlg_regs: dlg_regs, |
1686 | disp_ttu_regs: ttu_regs, |
1687 | rq_dlg_param: &rq_param.dlg, |
1688 | dlg_sys_param: &dlg_sys_param, |
1689 | cstate_en, |
1690 | pstate_en, |
1691 | vm_en, |
1692 | ignore_viewport_pos, |
1693 | immediate_flip_support); |
1694 | dml_print("DML_DLG: Calculation for pipe[%d] end\n" , pipe_idx); |
1695 | } |
1696 | |