2 * Copyright 2019-2021 Advanced Micro Devices, Inc.
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:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
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.
28 #include "dcn31/dcn31_resource.h"
29 #include "dcn315/dcn315_resource.h"
30 #include "dcn316/dcn316_resource.h"
32 #include "dml/dcn20/dcn20_fpu.h"
33 #include "dcn31_fpu.h"
36 * DOC: DCN31x FPU manipulation Overview
38 * The DCN architecture relies on FPU operations, which require special
39 * compilation flags and the use of kernel_fpu_begin/end functions; ideally, we
40 * want to avoid spreading FPU access across multiple files. With this idea in
41 * mind, this file aims to centralize all DCN3.1.x functions that require FPU
42 * access in a single place. Code in this file follows the following code
45 * 1. Functions that use FPU operations should be isolated in static functions.
46 * 2. The FPU functions should have the noinline attribute to ensure anything
47 * that deals with FP register is contained within this call.
48 * 3. All function that needs to be accessed outside this file requires a
49 * public interface that not uses any FPU reference.
50 * 4. Developers **must not** use DC_FP_START/END in this file, but they need
51 * to ensure that the caller invokes it before access any function available
52 * in this file. For this reason, public functions in this file must invoke
53 * dc_assert_fp_enabled();
56 struct _vcs_dpi_ip_params_st dcn3_1_ip = {
58 .gpuvm_max_page_table_levels = 1,
60 .hostvm_max_page_table_levels = 2,
61 .rob_buffer_size_kbytes = 64,
62 .det_buffer_size_kbytes = DCN3_1_DEFAULT_DET_SIZE,
63 .config_return_buffer_size_in_kbytes = 1792,
64 .compressed_buffer_segment_size_in_kbytes = 64,
65 .meta_fifo_size_in_kentries = 32,
66 .zero_size_buffer_entries = 512,
67 .compbuf_reserved_space_64b = 256,
68 .compbuf_reserved_space_zs = 64,
69 .dpp_output_buffer_pixels = 2560,
70 .opp_output_buffer_lines = 1,
71 .pixel_chunk_size_kbytes = 8,
72 .meta_chunk_size_kbytes = 2,
73 .min_meta_chunk_size_bytes = 256,
74 .writeback_chunk_size_kbytes = 8,
75 .ptoi_supported = false,
77 .maximum_dsc_bits_per_component = 10,
78 .dsc422_native_support = false,
79 .is_line_buffer_bpp_fixed = true,
80 .line_buffer_fixed_bpp = 48,
81 .line_buffer_size_bits = 789504,
82 .max_line_buffer_lines = 12,
83 .writeback_interface_buffer_size_kbytes = 90,
86 .max_num_hdmi_frl_outputs = 1,
88 .max_dchub_pscl_bw_pix_per_clk = 4,
89 .max_pscl_lb_bw_pix_per_clk = 2,
90 .max_lb_vscl_bw_pix_per_clk = 4,
91 .max_vscl_hscl_bw_pix_per_clk = 4,
96 .dpte_buffer_size_in_pte_reqs_luma = 64,
97 .dpte_buffer_size_in_pte_reqs_chroma = 34,
98 .dispclk_ramp_margin_percent = 1,
99 .max_inter_dcn_tile_repeaters = 8,
100 .cursor_buffer_size = 16,
101 .cursor_chunk_size = 2,
102 .writeback_line_buffer_buffer_size = 0,
103 .writeback_min_hscl_ratio = 1,
104 .writeback_min_vscl_ratio = 1,
105 .writeback_max_hscl_ratio = 1,
106 .writeback_max_vscl_ratio = 1,
107 .writeback_max_hscl_taps = 1,
108 .writeback_max_vscl_taps = 1,
109 .dppclk_delay_subtotal = 46,
110 .dppclk_delay_scl = 50,
111 .dppclk_delay_scl_lb_only = 16,
112 .dppclk_delay_cnvc_formatter = 27,
113 .dppclk_delay_cnvc_cursor = 6,
114 .dispclk_delay_subtotal = 119,
115 .dynamic_metadata_vm_enabled = false,
116 .odm_combine_4to1_supported = false,
117 .dcc_supported = true,
120 static struct _vcs_dpi_soc_bounding_box_st dcn3_1_soc = {
121 /*TODO: correct dispclk/dppclk voltage level determination*/
125 .dispclk_mhz = 1200.0,
126 .dppclk_mhz = 1200.0,
128 .phyclk_d18_mhz = 667.0,
134 .dispclk_mhz = 1200.0,
135 .dppclk_mhz = 1200.0,
137 .phyclk_d18_mhz = 667.0,
143 .dispclk_mhz = 1200.0,
144 .dppclk_mhz = 1200.0,
146 .phyclk_d18_mhz = 667.0,
152 .dispclk_mhz = 1200.0,
153 .dppclk_mhz = 1200.0,
155 .phyclk_d18_mhz = 667.0,
161 .dispclk_mhz = 1200.0,
162 .dppclk_mhz = 1200.0,
164 .phyclk_d18_mhz = 667.0,
170 .sr_exit_time_us = 9.0,
171 .sr_enter_plus_exit_time_us = 11.0,
172 .sr_exit_z8_time_us = 442.0,
173 .sr_enter_plus_exit_z8_time_us = 560.0,
174 .writeback_latency_us = 12.0,
175 .dram_channel_width_bytes = 4,
176 .round_trip_ping_latency_dcfclk_cycles = 106,
177 .urgent_latency_pixel_data_only_us = 4.0,
178 .urgent_latency_pixel_mixed_with_vm_data_us = 4.0,
179 .urgent_latency_vm_data_only_us = 4.0,
180 .urgent_out_of_order_return_per_channel_pixel_only_bytes = 4096,
181 .urgent_out_of_order_return_per_channel_pixel_and_vm_bytes = 4096,
182 .urgent_out_of_order_return_per_channel_vm_only_bytes = 4096,
183 .pct_ideal_sdp_bw_after_urgent = 80.0,
184 .pct_ideal_dram_sdp_bw_after_urgent_pixel_only = 65.0,
185 .pct_ideal_dram_sdp_bw_after_urgent_pixel_and_vm = 60.0,
186 .pct_ideal_dram_sdp_bw_after_urgent_vm_only = 30.0,
187 .max_avg_sdp_bw_use_normal_percent = 60.0,
188 .max_avg_dram_bw_use_normal_percent = 60.0,
189 .fabric_datapath_to_dcn_data_return_bytes = 32,
190 .return_bus_width_bytes = 64,
191 .downspread_percent = 0.38,
192 .dcn_downspread_percent = 0.5,
193 .gpuvm_min_page_size_bytes = 4096,
194 .hostvm_min_page_size_bytes = 4096,
195 .do_urgent_latency_adjustment = false,
196 .urgent_latency_adjustment_fabric_clock_component_us = 0,
197 .urgent_latency_adjustment_fabric_clock_reference_mhz = 0,
200 struct _vcs_dpi_ip_params_st dcn3_15_ip = {
202 .gpuvm_max_page_table_levels = 1,
204 .hostvm_max_page_table_levels = 2,
205 .rob_buffer_size_kbytes = 64,
206 .det_buffer_size_kbytes = DCN3_15_DEFAULT_DET_SIZE,
207 .min_comp_buffer_size_kbytes = 64,
208 .config_return_buffer_size_in_kbytes = 1024,
209 .compressed_buffer_segment_size_in_kbytes = 64,
210 .meta_fifo_size_in_kentries = 32,
211 .zero_size_buffer_entries = 512,
212 .compbuf_reserved_space_64b = 256,
213 .compbuf_reserved_space_zs = 64,
214 .dpp_output_buffer_pixels = 2560,
215 .opp_output_buffer_lines = 1,
216 .pixel_chunk_size_kbytes = 8,
217 .meta_chunk_size_kbytes = 2,
218 .min_meta_chunk_size_bytes = 256,
219 .writeback_chunk_size_kbytes = 8,
220 .ptoi_supported = false,
222 .maximum_dsc_bits_per_component = 10,
223 .dsc422_native_support = false,
224 .is_line_buffer_bpp_fixed = true,
225 .line_buffer_fixed_bpp = 49,
226 .line_buffer_size_bits = 789504,
227 .max_line_buffer_lines = 12,
228 .writeback_interface_buffer_size_kbytes = 90,
231 .max_num_hdmi_frl_outputs = 1,
233 .max_dchub_pscl_bw_pix_per_clk = 4,
234 .max_pscl_lb_bw_pix_per_clk = 2,
235 .max_lb_vscl_bw_pix_per_clk = 4,
236 .max_vscl_hscl_bw_pix_per_clk = 4,
241 .dpte_buffer_size_in_pte_reqs_luma = 64,
242 .dpte_buffer_size_in_pte_reqs_chroma = 34,
243 .dispclk_ramp_margin_percent = 1,
244 .max_inter_dcn_tile_repeaters = 9,
245 .cursor_buffer_size = 16,
246 .cursor_chunk_size = 2,
247 .writeback_line_buffer_buffer_size = 0,
248 .writeback_min_hscl_ratio = 1,
249 .writeback_min_vscl_ratio = 1,
250 .writeback_max_hscl_ratio = 1,
251 .writeback_max_vscl_ratio = 1,
252 .writeback_max_hscl_taps = 1,
253 .writeback_max_vscl_taps = 1,
254 .dppclk_delay_subtotal = 46,
255 .dppclk_delay_scl = 50,
256 .dppclk_delay_scl_lb_only = 16,
257 .dppclk_delay_cnvc_formatter = 27,
258 .dppclk_delay_cnvc_cursor = 6,
259 .dispclk_delay_subtotal = 119,
260 .dynamic_metadata_vm_enabled = false,
261 .odm_combine_4to1_supported = false,
262 .dcc_supported = true,
265 static struct _vcs_dpi_soc_bounding_box_st dcn3_15_soc = {
266 .sr_exit_time_us = 9.0,
267 .sr_enter_plus_exit_time_us = 11.0,
268 .sr_exit_z8_time_us = 50.0,
269 .sr_enter_plus_exit_z8_time_us = 50.0,
270 .writeback_latency_us = 12.0,
271 .dram_channel_width_bytes = 4,
272 .round_trip_ping_latency_dcfclk_cycles = 106,
273 .urgent_latency_pixel_data_only_us = 4.0,
274 .urgent_latency_pixel_mixed_with_vm_data_us = 4.0,
275 .urgent_latency_vm_data_only_us = 4.0,
276 .urgent_out_of_order_return_per_channel_pixel_only_bytes = 4096,
277 .urgent_out_of_order_return_per_channel_pixel_and_vm_bytes = 4096,
278 .urgent_out_of_order_return_per_channel_vm_only_bytes = 4096,
279 .pct_ideal_sdp_bw_after_urgent = 80.0,
280 .pct_ideal_dram_sdp_bw_after_urgent_pixel_only = 65.0,
281 .pct_ideal_dram_sdp_bw_after_urgent_pixel_and_vm = 60.0,
282 .pct_ideal_dram_sdp_bw_after_urgent_vm_only = 30.0,
283 .max_avg_sdp_bw_use_normal_percent = 60.0,
284 .max_avg_dram_bw_use_normal_percent = 60.0,
285 .fabric_datapath_to_dcn_data_return_bytes = 32,
286 .return_bus_width_bytes = 64,
287 .downspread_percent = 0.38,
288 .dcn_downspread_percent = 0.38,
289 .gpuvm_min_page_size_bytes = 4096,
290 .hostvm_min_page_size_bytes = 4096,
291 .do_urgent_latency_adjustment = false,
292 .urgent_latency_adjustment_fabric_clock_component_us = 0,
293 .urgent_latency_adjustment_fabric_clock_reference_mhz = 0,
296 struct _vcs_dpi_ip_params_st dcn3_16_ip = {
298 .gpuvm_max_page_table_levels = 1,
300 .hostvm_max_page_table_levels = 2,
301 .rob_buffer_size_kbytes = 64,
302 .det_buffer_size_kbytes = DCN3_16_DEFAULT_DET_SIZE,
303 .min_comp_buffer_size_kbytes = 64,
304 .config_return_buffer_size_in_kbytes = 1024,
305 .compressed_buffer_segment_size_in_kbytes = 64,
306 .meta_fifo_size_in_kentries = 32,
307 .zero_size_buffer_entries = 512,
308 .compbuf_reserved_space_64b = 256,
309 .compbuf_reserved_space_zs = 64,
310 .dpp_output_buffer_pixels = 2560,
311 .opp_output_buffer_lines = 1,
312 .pixel_chunk_size_kbytes = 8,
313 .meta_chunk_size_kbytes = 2,
314 .min_meta_chunk_size_bytes = 256,
315 .writeback_chunk_size_kbytes = 8,
316 .ptoi_supported = false,
318 .maximum_dsc_bits_per_component = 10,
319 .dsc422_native_support = false,
320 .is_line_buffer_bpp_fixed = true,
321 .line_buffer_fixed_bpp = 48,
322 .line_buffer_size_bits = 789504,
323 .max_line_buffer_lines = 12,
324 .writeback_interface_buffer_size_kbytes = 90,
327 .max_num_hdmi_frl_outputs = 1,
329 .max_dchub_pscl_bw_pix_per_clk = 4,
330 .max_pscl_lb_bw_pix_per_clk = 2,
331 .max_lb_vscl_bw_pix_per_clk = 4,
332 .max_vscl_hscl_bw_pix_per_clk = 4,
337 .dpte_buffer_size_in_pte_reqs_luma = 64,
338 .dpte_buffer_size_in_pte_reqs_chroma = 34,
339 .dispclk_ramp_margin_percent = 1,
340 .max_inter_dcn_tile_repeaters = 8,
341 .cursor_buffer_size = 16,
342 .cursor_chunk_size = 2,
343 .writeback_line_buffer_buffer_size = 0,
344 .writeback_min_hscl_ratio = 1,
345 .writeback_min_vscl_ratio = 1,
346 .writeback_max_hscl_ratio = 1,
347 .writeback_max_vscl_ratio = 1,
348 .writeback_max_hscl_taps = 1,
349 .writeback_max_vscl_taps = 1,
350 .dppclk_delay_subtotal = 46,
351 .dppclk_delay_scl = 50,
352 .dppclk_delay_scl_lb_only = 16,
353 .dppclk_delay_cnvc_formatter = 27,
354 .dppclk_delay_cnvc_cursor = 6,
355 .dispclk_delay_subtotal = 119,
356 .dynamic_metadata_vm_enabled = false,
357 .odm_combine_4to1_supported = false,
358 .dcc_supported = true,
361 static struct _vcs_dpi_soc_bounding_box_st dcn3_16_soc = {
362 /*TODO: correct dispclk/dppclk voltage level determination*/
366 .dispclk_mhz = 556.0,
369 .phyclk_d18_mhz = 445.0,
375 .dispclk_mhz = 625.0,
378 .phyclk_d18_mhz = 667.0,
384 .dispclk_mhz = 625.0,
387 .phyclk_d18_mhz = 667.0,
393 .dispclk_mhz = 1112.0,
394 .dppclk_mhz = 1112.0,
396 .phyclk_d18_mhz = 667.0,
402 .dispclk_mhz = 1250.0,
403 .dppclk_mhz = 1250.0,
405 .phyclk_d18_mhz = 667.0,
411 .sr_exit_time_us = 9.0,
412 .sr_enter_plus_exit_time_us = 11.0,
413 .sr_exit_z8_time_us = 442.0,
414 .sr_enter_plus_exit_z8_time_us = 560.0,
415 .writeback_latency_us = 12.0,
416 .dram_channel_width_bytes = 4,
417 .round_trip_ping_latency_dcfclk_cycles = 106,
418 .urgent_latency_pixel_data_only_us = 4.0,
419 .urgent_latency_pixel_mixed_with_vm_data_us = 4.0,
420 .urgent_latency_vm_data_only_us = 4.0,
421 .urgent_out_of_order_return_per_channel_pixel_only_bytes = 4096,
422 .urgent_out_of_order_return_per_channel_pixel_and_vm_bytes = 4096,
423 .urgent_out_of_order_return_per_channel_vm_only_bytes = 4096,
424 .pct_ideal_sdp_bw_after_urgent = 80.0,
425 .pct_ideal_dram_sdp_bw_after_urgent_pixel_only = 65.0,
426 .pct_ideal_dram_sdp_bw_after_urgent_pixel_and_vm = 60.0,
427 .pct_ideal_dram_sdp_bw_after_urgent_vm_only = 30.0,
428 .max_avg_sdp_bw_use_normal_percent = 60.0,
429 .max_avg_dram_bw_use_normal_percent = 60.0,
430 .fabric_datapath_to_dcn_data_return_bytes = 32,
431 .return_bus_width_bytes = 64,
432 .downspread_percent = 0.38,
433 .dcn_downspread_percent = 0.5,
434 .gpuvm_min_page_size_bytes = 4096,
435 .hostvm_min_page_size_bytes = 4096,
436 .do_urgent_latency_adjustment = false,
437 .urgent_latency_adjustment_fabric_clock_component_us = 0,
438 .urgent_latency_adjustment_fabric_clock_reference_mhz = 0,
441 void dcn31_zero_pipe_dcc_fraction(display_e2e_pipe_params_st *pipes,
444 dc_assert_fp_enabled();
446 pipes[pipe_cnt].pipe.src.dcc_fraction_of_zs_req_luma = 0;
447 pipes[pipe_cnt].pipe.src.dcc_fraction_of_zs_req_chroma = 0;
450 void dcn31_update_soc_for_wm_a(struct dc *dc, struct dc_state *context)
452 dc_assert_fp_enabled();
454 if (dc->clk_mgr->bw_params->wm_table.entries[WM_A].valid) {
455 context->bw_ctx.dml.soc.dram_clock_change_latency_us = dc->clk_mgr->bw_params->wm_table.entries[WM_A].pstate_latency_us;
456 context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us = dc->clk_mgr->bw_params->wm_table.entries[WM_A].sr_enter_plus_exit_time_us;
457 context->bw_ctx.dml.soc.sr_exit_time_us = dc->clk_mgr->bw_params->wm_table.entries[WM_A].sr_exit_time_us;
461 void dcn31_calculate_wm_and_dlg_fp(
462 struct dc *dc, struct dc_state *context,
463 display_e2e_pipe_params_st *pipes,
468 double dcfclk = context->bw_ctx.dml.vba.DCFCLKState[vlevel][context->bw_ctx.dml.vba.maxMpcComb];
470 dc_assert_fp_enabled();
472 if (context->bw_ctx.dml.soc.min_dcfclk > dcfclk)
473 dcfclk = context->bw_ctx.dml.soc.min_dcfclk;
475 /* We don't recalculate clocks for 0 pipe configs, which can block
476 * S0i3 as high clocks will block low power states
477 * Override any clocks that can block S0i3 to min here
480 context->bw_ctx.bw.dcn.clk.dcfclk_khz = dcfclk; // always should be vlevel 0
484 pipes[0].clks_cfg.voltage = vlevel;
485 pipes[0].clks_cfg.dcfclk_mhz = dcfclk;
486 pipes[0].clks_cfg.socclk_mhz = context->bw_ctx.dml.soc.clock_limits[vlevel].socclk_mhz;
492 if (dc->clk_mgr->bw_params->wm_table.nv_entries[WM_B].valid) {
494 pipes[0].clks_cfg.voltage = 1;
495 pipes[0].clks_cfg.dcfclk_mhz = context->bw_ctx.dml.soc.clock_limits[0].dcfclk_mhz;
497 context->bw_ctx.dml.soc.dram_clock_change_latency_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_B].dml_input.pstate_latency_us;
498 context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_B].dml_input.sr_enter_plus_exit_time_us;
499 context->bw_ctx.dml.soc.sr_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_B].dml_input.sr_exit_time_us;
501 context->bw_ctx.bw.dcn.watermarks.b.urgent_ns = get_wm_urgent(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
502 context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
503 context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.cstate_exit_ns = get_wm_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
504 context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.cstate_enter_plus_exit_z8_ns = get_wm_z8_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
505 context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.cstate_exit_z8_ns = get_wm_z8_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
506 context->bw_ctx.bw.dcn.watermarks.b.cstate_pstate.pstate_change_ns = get_wm_dram_clock_change(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
507 context->bw_ctx.bw.dcn.watermarks.b.pte_meta_urgent_ns = get_wm_memory_trip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
508 context->bw_ctx.bw.dcn.watermarks.b.frac_urg_bw_nom = get_fraction_of_urgent_bandwidth(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
509 context->bw_ctx.bw.dcn.watermarks.b.frac_urg_bw_flip = get_fraction_of_urgent_bandwidth_imm_flip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
510 context->bw_ctx.bw.dcn.watermarks.b.urgent_latency_ns = get_urgent_latency(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
512 pipes[0].clks_cfg.voltage = vlevel;
513 pipes[0].clks_cfg.dcfclk_mhz = dcfclk;
518 if (dc->clk_mgr->bw_params->wm_table.nv_entries[WM_C].valid) {
519 context->bw_ctx.dml.soc.dram_clock_change_latency_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_C].dml_input.pstate_latency_us;
520 context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_C].dml_input.sr_enter_plus_exit_time_us;
521 context->bw_ctx.dml.soc.sr_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_C].dml_input.sr_exit_time_us;
523 context->bw_ctx.bw.dcn.watermarks.c.urgent_ns = get_wm_urgent(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
524 context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
525 context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.cstate_exit_ns = get_wm_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
526 context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.cstate_enter_plus_exit_z8_ns = get_wm_z8_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
527 context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.cstate_exit_z8_ns = get_wm_z8_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
528 context->bw_ctx.bw.dcn.watermarks.c.cstate_pstate.pstate_change_ns = get_wm_dram_clock_change(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
529 context->bw_ctx.bw.dcn.watermarks.c.pte_meta_urgent_ns = get_wm_memory_trip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
530 context->bw_ctx.bw.dcn.watermarks.c.frac_urg_bw_nom = get_fraction_of_urgent_bandwidth(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
531 context->bw_ctx.bw.dcn.watermarks.c.frac_urg_bw_flip = get_fraction_of_urgent_bandwidth_imm_flip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
532 context->bw_ctx.bw.dcn.watermarks.c.urgent_latency_ns = get_urgent_latency(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
537 if (dc->clk_mgr->bw_params->wm_table.nv_entries[WM_D].valid) {
538 context->bw_ctx.dml.soc.dram_clock_change_latency_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_D].dml_input.pstate_latency_us;
539 context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_D].dml_input.sr_enter_plus_exit_time_us;
540 context->bw_ctx.dml.soc.sr_exit_time_us = dc->clk_mgr->bw_params->wm_table.nv_entries[WM_D].dml_input.sr_exit_time_us;
542 context->bw_ctx.bw.dcn.watermarks.d.urgent_ns = get_wm_urgent(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
543 context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
544 context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.cstate_exit_ns = get_wm_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
545 context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.pstate_change_ns = get_wm_dram_clock_change(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
546 context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.cstate_enter_plus_exit_z8_ns = get_wm_z8_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
547 context->bw_ctx.bw.dcn.watermarks.d.cstate_pstate.cstate_exit_z8_ns = get_wm_z8_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
548 context->bw_ctx.bw.dcn.watermarks.d.pte_meta_urgent_ns = get_wm_memory_trip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
549 context->bw_ctx.bw.dcn.watermarks.d.frac_urg_bw_nom = get_fraction_of_urgent_bandwidth(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
550 context->bw_ctx.bw.dcn.watermarks.d.frac_urg_bw_flip = get_fraction_of_urgent_bandwidth_imm_flip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
551 context->bw_ctx.bw.dcn.watermarks.d.urgent_latency_ns = get_urgent_latency(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
555 * All clocks min required
557 * Set A calculated last so that following calculations are based on Set A
559 dc->res_pool->funcs->update_soc_for_wm_a(dc, context);
560 context->bw_ctx.bw.dcn.watermarks.a.urgent_ns = get_wm_urgent(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
561 context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.cstate_enter_plus_exit_ns = get_wm_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
562 context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.cstate_exit_ns = get_wm_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
563 context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.pstate_change_ns = get_wm_dram_clock_change(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
564 context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.cstate_enter_plus_exit_z8_ns = get_wm_z8_stutter_enter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
565 context->bw_ctx.bw.dcn.watermarks.a.cstate_pstate.cstate_exit_z8_ns = get_wm_z8_stutter_exit(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
566 context->bw_ctx.bw.dcn.watermarks.a.pte_meta_urgent_ns = get_wm_memory_trip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
567 context->bw_ctx.bw.dcn.watermarks.a.frac_urg_bw_nom = get_fraction_of_urgent_bandwidth(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
568 context->bw_ctx.bw.dcn.watermarks.a.frac_urg_bw_flip = get_fraction_of_urgent_bandwidth_imm_flip(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
569 context->bw_ctx.bw.dcn.watermarks.a.urgent_latency_ns = get_urgent_latency(&context->bw_ctx.dml, pipes, pipe_cnt) * 1000;
571 context->bw_ctx.bw.dcn.watermarks.b = context->bw_ctx.bw.dcn.watermarks.a;
572 context->bw_ctx.bw.dcn.watermarks.c = context->bw_ctx.bw.dcn.watermarks.a;
573 context->bw_ctx.bw.dcn.watermarks.d = context->bw_ctx.bw.dcn.watermarks.a;
576 for (i = 0, pipe_idx = 0; i < dc->res_pool->pipe_count; i++) {
577 if (!context->res_ctx.pipe_ctx[i].stream)
580 pipes[pipe_idx].clks_cfg.dispclk_mhz = get_dispclk_calculated(&context->bw_ctx.dml, pipes, pipe_cnt);
581 pipes[pipe_idx].clks_cfg.dppclk_mhz = get_dppclk_calculated(&context->bw_ctx.dml, pipes, pipe_cnt, pipe_idx);
583 if (dc->config.forced_clocks || dc->debug.max_disp_clk) {
584 pipes[pipe_idx].clks_cfg.dispclk_mhz = context->bw_ctx.dml.soc.clock_limits[0].dispclk_mhz;
585 pipes[pipe_idx].clks_cfg.dppclk_mhz = context->bw_ctx.dml.soc.clock_limits[0].dppclk_mhz;
587 if (dc->debug.min_disp_clk_khz > pipes[pipe_idx].clks_cfg.dispclk_mhz * 1000)
588 pipes[pipe_idx].clks_cfg.dispclk_mhz = dc->debug.min_disp_clk_khz / 1000.0;
589 if (dc->debug.min_dpp_clk_khz > pipes[pipe_idx].clks_cfg.dppclk_mhz * 1000)
590 pipes[pipe_idx].clks_cfg.dppclk_mhz = dc->debug.min_dpp_clk_khz / 1000.0;
595 dcn20_calculate_dlg_params(dc, context, pipes, pipe_cnt, vlevel);
598 void dcn31_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw_params)
600 struct clk_limit_table *clk_table = &bw_params->clk_table;
601 unsigned int i, closest_clk_lvl;
604 dc_assert_fp_enabled();
606 memcpy(&dcn3_1_soc._clock_tmp, &dcn3_1_soc.clock_limits,
607 sizeof(dcn3_1_soc.clock_limits));
609 // Default clock levels are used for diags, which may lead to overclocking.
610 if (!IS_DIAG_DC(dc->ctx->dce_environment)) {
611 int max_dispclk_mhz = 0, max_dppclk_mhz = 0;
613 dcn3_1_ip.max_num_otg = dc->res_pool->res_cap->num_timing_generator;
614 dcn3_1_ip.max_num_dpp = dc->res_pool->pipe_count;
615 dcn3_1_soc.num_chans = bw_params->num_channels;
617 ASSERT(clk_table->num_entries);
619 /* Prepass to find max clocks independent of voltage level. */
620 for (i = 0; i < clk_table->num_entries; ++i) {
621 if (clk_table->entries[i].dispclk_mhz > max_dispclk_mhz)
622 max_dispclk_mhz = clk_table->entries[i].dispclk_mhz;
623 if (clk_table->entries[i].dppclk_mhz > max_dppclk_mhz)
624 max_dppclk_mhz = clk_table->entries[i].dppclk_mhz;
627 for (i = 0; i < clk_table->num_entries; i++) {
629 for (closest_clk_lvl = 0, j = dcn3_1_soc.num_states - 1; j >= 0; j--) {
630 if ((unsigned int) dcn3_1_soc.clock_limits[j].dcfclk_mhz <= clk_table->entries[i].dcfclk_mhz) {
636 dcn3_1_soc._clock_tmp[i].state = i;
638 /* Clocks dependent on voltage level. */
639 dcn3_1_soc._clock_tmp[i].dcfclk_mhz = clk_table->entries[i].dcfclk_mhz;
640 dcn3_1_soc._clock_tmp[i].fabricclk_mhz = clk_table->entries[i].fclk_mhz;
641 dcn3_1_soc._clock_tmp[i].socclk_mhz = clk_table->entries[i].socclk_mhz;
642 dcn3_1_soc._clock_tmp[i].dram_speed_mts = clk_table->entries[i].memclk_mhz * 2 * clk_table->entries[i].wck_ratio;
644 /* Clocks independent of voltage level. */
645 dcn3_1_soc._clock_tmp[i].dispclk_mhz = max_dispclk_mhz ? max_dispclk_mhz :
646 dcn3_1_soc.clock_limits[closest_clk_lvl].dispclk_mhz;
648 dcn3_1_soc._clock_tmp[i].dppclk_mhz = max_dppclk_mhz ? max_dppclk_mhz :
649 dcn3_1_soc.clock_limits[closest_clk_lvl].dppclk_mhz;
651 dcn3_1_soc._clock_tmp[i].dram_bw_per_chan_gbps = dcn3_1_soc.clock_limits[closest_clk_lvl].dram_bw_per_chan_gbps;
652 dcn3_1_soc._clock_tmp[i].dscclk_mhz = dcn3_1_soc.clock_limits[closest_clk_lvl].dscclk_mhz;
653 dcn3_1_soc._clock_tmp[i].dtbclk_mhz = dcn3_1_soc.clock_limits[closest_clk_lvl].dtbclk_mhz;
654 dcn3_1_soc._clock_tmp[i].phyclk_d18_mhz = dcn3_1_soc.clock_limits[closest_clk_lvl].phyclk_d18_mhz;
655 dcn3_1_soc._clock_tmp[i].phyclk_mhz = dcn3_1_soc.clock_limits[closest_clk_lvl].phyclk_mhz;
657 if (clk_table->num_entries) {
658 dcn3_1_soc.num_states = clk_table->num_entries;
662 memcpy(&dcn3_1_soc.clock_limits, &dcn3_1_soc._clock_tmp,
663 sizeof(dcn3_1_soc.clock_limits));
665 dcn3_1_soc.dispclk_dppclk_vco_speed_mhz = dc->clk_mgr->dentist_vco_freq_khz / 1000.0;
666 dc->dml.soc.dispclk_dppclk_vco_speed_mhz = dc->clk_mgr->dentist_vco_freq_khz / 1000.0;
668 if (!IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment))
669 dml_init_instance(&dc->dml, &dcn3_1_soc, &dcn3_1_ip, DML_PROJECT_DCN31);
671 dml_init_instance(&dc->dml, &dcn3_1_soc, &dcn3_1_ip, DML_PROJECT_DCN31_FPGA);
674 void dcn315_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw_params)
676 struct clk_limit_table *clk_table = &bw_params->clk_table;
677 int i, max_dispclk_mhz = 0, max_dppclk_mhz = 0;
679 dc_assert_fp_enabled();
681 dcn3_15_ip.max_num_otg = dc->res_pool->res_cap->num_timing_generator;
682 dcn3_15_ip.max_num_dpp = dc->res_pool->pipe_count;
683 dcn3_15_soc.num_chans = bw_params->num_channels;
685 ASSERT(clk_table->num_entries);
687 /* Setup soc to always use max dispclk/dppclk to avoid odm-to-lower-voltage */
688 for (i = 0; i < clk_table->num_entries; ++i) {
689 if (clk_table->entries[i].dispclk_mhz > max_dispclk_mhz)
690 max_dispclk_mhz = clk_table->entries[i].dispclk_mhz;
691 if (clk_table->entries[i].dppclk_mhz > max_dppclk_mhz)
692 max_dppclk_mhz = clk_table->entries[i].dppclk_mhz;
695 for (i = 0; i < clk_table->num_entries; i++) {
696 dcn3_15_soc.clock_limits[i].state = i;
698 /* Clocks dependent on voltage level. */
699 dcn3_15_soc.clock_limits[i].dcfclk_mhz = clk_table->entries[i].dcfclk_mhz;
700 dcn3_15_soc.clock_limits[i].fabricclk_mhz = clk_table->entries[i].fclk_mhz;
701 dcn3_15_soc.clock_limits[i].socclk_mhz = clk_table->entries[i].socclk_mhz;
702 dcn3_15_soc.clock_limits[i].dram_speed_mts = clk_table->entries[i].memclk_mhz * 2 * clk_table->entries[i].wck_ratio;
704 /* These aren't actually read from smu, but rather set in clk_mgr defaults */
705 dcn3_15_soc.clock_limits[i].dtbclk_mhz = clk_table->entries[i].dtbclk_mhz;
706 dcn3_15_soc.clock_limits[i].phyclk_d18_mhz = clk_table->entries[i].phyclk_d18_mhz;
707 dcn3_15_soc.clock_limits[i].phyclk_mhz = clk_table->entries[i].phyclk_mhz;
709 /* Clocks independent of voltage level. */
710 dcn3_15_soc.clock_limits[i].dispclk_mhz = max_dispclk_mhz;
711 dcn3_15_soc.clock_limits[i].dppclk_mhz = max_dppclk_mhz;
712 dcn3_15_soc.clock_limits[i].dscclk_mhz = max_dispclk_mhz / 3.0;
714 dcn3_15_soc.num_states = clk_table->num_entries;
717 /* Set vco to max_dispclk * 2 to make sure the highest dispclk is always available for dml calcs,
718 * no impact outside of dml validation
720 dcn3_15_soc.dispclk_dppclk_vco_speed_mhz = max_dispclk_mhz * 2;
722 if (!IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment))
723 dml_init_instance(&dc->dml, &dcn3_15_soc, &dcn3_15_ip, DML_PROJECT_DCN31);
725 dml_init_instance(&dc->dml, &dcn3_15_soc, &dcn3_15_ip, DML_PROJECT_DCN31_FPGA);
728 void dcn316_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw_params)
730 struct clk_limit_table *clk_table = &bw_params->clk_table;
731 unsigned int i, closest_clk_lvl;
732 int max_dispclk_mhz = 0, max_dppclk_mhz = 0;
735 dc_assert_fp_enabled();
737 memcpy(&dcn3_16_soc._clock_tmp, &dcn3_16_soc.clock_limits,
738 sizeof(dcn3_16_soc.clock_limits));
740 // Default clock levels are used for diags, which may lead to overclocking.
741 if (!IS_DIAG_DC(dc->ctx->dce_environment)) {
743 dcn3_16_ip.max_num_otg = dc->res_pool->res_cap->num_timing_generator;
744 dcn3_16_ip.max_num_dpp = dc->res_pool->pipe_count;
745 dcn3_16_soc.num_chans = bw_params->num_channels;
747 ASSERT(clk_table->num_entries);
749 /* Prepass to find max clocks independent of voltage level. */
750 for (i = 0; i < clk_table->num_entries; ++i) {
751 if (clk_table->entries[i].dispclk_mhz > max_dispclk_mhz)
752 max_dispclk_mhz = clk_table->entries[i].dispclk_mhz;
753 if (clk_table->entries[i].dppclk_mhz > max_dppclk_mhz)
754 max_dppclk_mhz = clk_table->entries[i].dppclk_mhz;
757 for (i = 0; i < clk_table->num_entries; i++) {
759 for (closest_clk_lvl = 0, j = dcn3_16_soc.num_states - 1; j >= 0; j--) {
760 if ((unsigned int) dcn3_16_soc.clock_limits[j].dcfclk_mhz <= clk_table->entries[i].dcfclk_mhz) {
765 // Ported from DCN315
766 if (clk_table->num_entries == 1) {
767 /*smu gives one DPM level, let's take the highest one*/
768 closest_clk_lvl = dcn3_16_soc.num_states - 1;
771 dcn3_16_soc._clock_tmp[i].state = i;
773 /* Clocks dependent on voltage level. */
774 dcn3_16_soc._clock_tmp[i].dcfclk_mhz = clk_table->entries[i].dcfclk_mhz;
775 if (clk_table->num_entries == 1 &&
776 dcn3_16_soc._clock_tmp[i].dcfclk_mhz < dcn3_16_soc.clock_limits[closest_clk_lvl].dcfclk_mhz) {
777 /*SMU fix not released yet*/
778 dcn3_16_soc._clock_tmp[i].dcfclk_mhz = dcn3_16_soc.clock_limits[closest_clk_lvl].dcfclk_mhz;
780 dcn3_16_soc._clock_tmp[i].fabricclk_mhz = clk_table->entries[i].fclk_mhz;
781 dcn3_16_soc._clock_tmp[i].socclk_mhz = clk_table->entries[i].socclk_mhz;
782 dcn3_16_soc._clock_tmp[i].dram_speed_mts = clk_table->entries[i].memclk_mhz * 2 * clk_table->entries[i].wck_ratio;
784 /* Clocks independent of voltage level. */
785 dcn3_16_soc._clock_tmp[i].dispclk_mhz = max_dispclk_mhz ? max_dispclk_mhz :
786 dcn3_16_soc.clock_limits[closest_clk_lvl].dispclk_mhz;
788 dcn3_16_soc._clock_tmp[i].dppclk_mhz = max_dppclk_mhz ? max_dppclk_mhz :
789 dcn3_16_soc.clock_limits[closest_clk_lvl].dppclk_mhz;
791 dcn3_16_soc._clock_tmp[i].dram_bw_per_chan_gbps = dcn3_16_soc.clock_limits[closest_clk_lvl].dram_bw_per_chan_gbps;
792 dcn3_16_soc._clock_tmp[i].dscclk_mhz = dcn3_16_soc.clock_limits[closest_clk_lvl].dscclk_mhz;
793 dcn3_16_soc._clock_tmp[i].dtbclk_mhz = dcn3_16_soc.clock_limits[closest_clk_lvl].dtbclk_mhz;
794 dcn3_16_soc._clock_tmp[i].phyclk_d18_mhz = dcn3_16_soc.clock_limits[closest_clk_lvl].phyclk_d18_mhz;
795 dcn3_16_soc._clock_tmp[i].phyclk_mhz = dcn3_16_soc.clock_limits[closest_clk_lvl].phyclk_mhz;
797 if (clk_table->num_entries) {
798 dcn3_16_soc.num_states = clk_table->num_entries;
802 memcpy(&dcn3_16_soc.clock_limits, &dcn3_16_soc._clock_tmp,
803 sizeof(dcn3_16_soc.clock_limits));
805 if (max_dispclk_mhz) {
806 dcn3_16_soc.dispclk_dppclk_vco_speed_mhz = max_dispclk_mhz * 2;
807 dc->dml.soc.dispclk_dppclk_vco_speed_mhz = max_dispclk_mhz * 2;
810 if (!IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment))
811 dml_init_instance(&dc->dml, &dcn3_16_soc, &dcn3_16_ip, DML_PROJECT_DCN31);
813 dml_init_instance(&dc->dml, &dcn3_16_soc, &dcn3_16_ip, DML_PROJECT_DCN31_FPGA);