diff options
Diffstat (limited to 'drivers/gpu/drm/amd/display/dc/dcn302')
| -rw-r--r-- | drivers/gpu/drm/amd/display/dc/dcn302/Makefile | 31 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.c | 335 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.h | 3 |
3 files changed, 25 insertions, 344 deletions
diff --git a/drivers/gpu/drm/amd/display/dc/dcn302/Makefile b/drivers/gpu/drm/amd/display/dc/dcn302/Makefile index 101620a8867a..ebd01cb467b7 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn302/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dcn302/Makefile @@ -1,43 +1,12 @@ # # (c) Copyright 2020 Advanced Micro Devices, Inc. All the rights reserved # -# All rights reserved. This notice is intended as a precaution against -# inadvertent publication and does not imply publication or any waiver -# of confidentiality. The year included in the foregoing notice is the -# year of creation of the work. -# # Authors: AMD # # Makefile for dcn302. DCN3_02 = dcn302_init.o dcn302_hwseq.o dcn302_resource.o -ifdef CONFIG_X86 -CFLAGS_$(AMDDALPATH)/dc/dcn302/dcn302_resource.o := -msse -endif - -ifdef CONFIG_PPC64 -CFLAGS_$(AMDDALPATH)/dc/dcn302/dcn302_resource.o := -mhard-float -maltivec -endif - -ifdef CONFIG_CC_IS_GCC -ifeq ($(call cc-ifversion, -lt, 0701, y), y) -IS_OLD_GCC = 1 -endif -CFLAGS_$(AMDDALPATH)/dc/dcn302/dcn302_resource.o += -mhard-float -endif - -ifdef CONFIG_X86 -ifdef IS_OLD_GCC -# Stack alignment mismatch, proceed with caution. -# GCC < 7.1 cannot compile code using `double` and -mpreferred-stack-boundary=3 -# (8B stack alignment). -CFLAGS_$(AMDDALPATH)/dc/dcn302/dcn302_resource.o += -mpreferred-stack-boundary=4 -else -CFLAGS_$(AMDDALPATH)/dc/dcn302/dcn302_resource.o += -msse2 -endif -endif - AMD_DAL_DCN3_02 = $(addprefix $(AMDDALPATH)/dc/dcn302/,$(DCN3_02)) AMD_DISPLAY_FILES += $(AMD_DAL_DCN3_02) diff --git a/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.c b/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.c index 2e9cbfa7663b..b925b6ddde5a 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.c @@ -43,6 +43,8 @@ #include "dcn20/dcn20_dsc.h" #include "dcn20/dcn20_resource.h" +#include "dml/dcn30/dcn30_fpu.h" + #include "dcn10/dcn10_resource.h" #include "dce/dce_abm.h" @@ -61,150 +63,18 @@ #include "resource.h" #include "vm_helper.h" +#include "dml/dcn302/dcn302_fpu.h" + #include "dimgrey_cavefish_ip_offset.h" #include "dcn/dcn_3_0_2_offset.h" #include "dcn/dcn_3_0_2_sh_mask.h" -#include "dcn/dpcs_3_0_0_offset.h" -#include "dcn/dpcs_3_0_0_sh_mask.h" +#include "dpcs/dpcs_3_0_0_offset.h" +#include "dpcs/dpcs_3_0_0_sh_mask.h" #include "nbio/nbio_7_4_offset.h" #include "amdgpu_socbb.h" #define DC_LOGGER_INIT(logger) -struct _vcs_dpi_ip_params_st dcn3_02_ip = { - .use_min_dcfclk = 0, - .clamp_min_dcfclk = 0, - .odm_capable = 1, - .gpuvm_enable = 1, - .hostvm_enable = 0, - .gpuvm_max_page_table_levels = 4, - .hostvm_max_page_table_levels = 4, - .hostvm_cached_page_table_levels = 0, - .pte_group_size_bytes = 2048, - .num_dsc = 5, - .rob_buffer_size_kbytes = 184, - .det_buffer_size_kbytes = 184, - .dpte_buffer_size_in_pte_reqs_luma = 64, - .dpte_buffer_size_in_pte_reqs_chroma = 34, - .pde_proc_buffer_size_64k_reqs = 48, - .dpp_output_buffer_pixels = 2560, - .opp_output_buffer_lines = 1, - .pixel_chunk_size_kbytes = 8, - .pte_enable = 1, - .max_page_table_levels = 2, - .pte_chunk_size_kbytes = 2, // ? - .meta_chunk_size_kbytes = 2, - .writeback_chunk_size_kbytes = 8, - .line_buffer_size_bits = 789504, - .is_line_buffer_bpp_fixed = 0, // ? - .line_buffer_fixed_bpp = 0, // ? - .dcc_supported = true, - .writeback_interface_buffer_size_kbytes = 90, - .writeback_line_buffer_buffer_size = 0, - .max_line_buffer_lines = 12, - .writeback_luma_buffer_size_kbytes = 12, // writeback_line_buffer_buffer_size = 656640 - .writeback_chroma_buffer_size_kbytes = 8, - .writeback_chroma_line_buffer_width_pixels = 4, - .writeback_max_hscl_ratio = 1, - .writeback_max_vscl_ratio = 1, - .writeback_min_hscl_ratio = 1, - .writeback_min_vscl_ratio = 1, - .writeback_max_hscl_taps = 1, - .writeback_max_vscl_taps = 1, - .writeback_line_buffer_luma_buffer_size = 0, - .writeback_line_buffer_chroma_buffer_size = 14643, - .cursor_buffer_size = 8, - .cursor_chunk_size = 2, - .max_num_otg = 5, - .max_num_dpp = 5, - .max_num_wb = 1, - .max_dchub_pscl_bw_pix_per_clk = 4, - .max_pscl_lb_bw_pix_per_clk = 2, - .max_lb_vscl_bw_pix_per_clk = 4, - .max_vscl_hscl_bw_pix_per_clk = 4, - .max_hscl_ratio = 6, - .max_vscl_ratio = 6, - .hscl_mults = 4, - .vscl_mults = 4, - .max_hscl_taps = 8, - .max_vscl_taps = 8, - .dispclk_ramp_margin_percent = 1, - .underscan_factor = 1.11, - .min_vblank_lines = 32, - .dppclk_delay_subtotal = 46, - .dynamic_metadata_vm_enabled = true, - .dppclk_delay_scl_lb_only = 16, - .dppclk_delay_scl = 50, - .dppclk_delay_cnvc_formatter = 27, - .dppclk_delay_cnvc_cursor = 6, - .dispclk_delay_subtotal = 119, - .dcfclk_cstate_latency = 5.2, // SRExitTime - .max_inter_dcn_tile_repeaters = 8, - .max_num_hdmi_frl_outputs = 1, - .odm_combine_4to1_supported = true, - - .xfc_supported = false, - .xfc_fill_bw_overhead_percent = 10.0, - .xfc_fill_constant_bytes = 0, - .gfx7_compat_tiling_supported = 0, - .number_of_cursors = 1, -}; - -struct _vcs_dpi_soc_bounding_box_st dcn3_02_soc = { - .clock_limits = { - { - .state = 0, - .dispclk_mhz = 562.0, - .dppclk_mhz = 300.0, - .phyclk_mhz = 300.0, - .phyclk_d18_mhz = 667.0, - .dscclk_mhz = 405.6, - }, - }, - - .min_dcfclk = 500.0, /* TODO: set this to actual min DCFCLK */ - .num_states = 1, - .sr_exit_time_us = 26.5, - .sr_enter_plus_exit_time_us = 31, - .urgent_latency_us = 4.0, - .urgent_latency_pixel_data_only_us = 4.0, - .urgent_latency_pixel_mixed_with_vm_data_us = 4.0, - .urgent_latency_vm_data_only_us = 4.0, - .urgent_out_of_order_return_per_channel_pixel_only_bytes = 4096, - .urgent_out_of_order_return_per_channel_pixel_and_vm_bytes = 4096, - .urgent_out_of_order_return_per_channel_vm_only_bytes = 4096, - .pct_ideal_dram_sdp_bw_after_urgent_pixel_only = 80.0, - .pct_ideal_dram_sdp_bw_after_urgent_pixel_and_vm = 60.0, - .pct_ideal_dram_sdp_bw_after_urgent_vm_only = 40.0, - .max_avg_sdp_bw_use_normal_percent = 60.0, - .max_avg_dram_bw_use_normal_percent = 40.0, - .writeback_latency_us = 12.0, - .max_request_size_bytes = 256, - .fabric_datapath_to_dcn_data_return_bytes = 64, - .dcn_downspread_percent = 0.5, - .downspread_percent = 0.38, - .dram_page_open_time_ns = 50.0, - .dram_rw_turnaround_time_ns = 17.5, - .dram_return_buffer_per_channel_bytes = 8192, - .round_trip_ping_latency_dcfclk_cycles = 156, - .urgent_out_of_order_return_per_channel_bytes = 4096, - .channel_interleave_bytes = 256, - .num_banks = 8, - .gpuvm_min_page_size_bytes = 4096, - .hostvm_min_page_size_bytes = 4096, - .dram_clock_change_latency_us = 404, - .dummy_pstate_latency_us = 5, - .writeback_dram_clock_change_latency_us = 23.0, - .return_bus_width_bytes = 64, - .dispclk_dppclk_vco_speed_mhz = 3650, - .xfc_bus_transport_time_us = 20, // ? - .xfc_xbuf_latency_tolerance_us = 4, // ? - .use_urgent_burst_bw = 1, // ? - .do_urgent_latency_adjustment = true, - .urgent_latency_adjustment_fabric_clock_component_us = 1.0, - .urgent_latency_adjustment_fabric_clock_reference_mhz = 1000, -}; - static const struct dc_debug_options debug_defaults_drv = { .disable_dmcu = true, .force_abm_enable = false, @@ -223,7 +93,8 @@ static const struct dc_debug_options debug_defaults_drv = { .underflow_assert_delay_us = 0xFFFFFFFF, .dwb_fi_phase = -1, // -1 = disable, .dmub_command_table = true, - .use_max_lb = true + .use_max_lb = true, + .exit_idle_opt_for_cursor_updates = true }; static const struct dc_debug_options debug_defaults_diags = { @@ -1021,7 +892,9 @@ static const struct dcn10_link_enc_hpd_registers link_enc_hpd_regs[] = { hpd_regs(4) }; -static struct link_encoder *dcn302_link_encoder_create(const struct encoder_init_data *enc_init_data) +static struct link_encoder *dcn302_link_encoder_create( + struct dc_context *ctx, + const struct encoder_init_data *enc_init_data) { struct dcn20_link_encoder *enc20 = kzalloc(sizeof(struct dcn20_link_encoder), GFP_KERNEL); @@ -1105,24 +978,19 @@ static bool init_soc_bounding_box(struct dc *dc, struct resource_pool *pool) loaded_ip->max_num_otg = pool->pipe_count; loaded_ip->max_num_dpp = pool->pipe_count; loaded_ip->clamp_min_dcfclk = dc->config.clamp_min_dcfclk; + DC_FP_START(); dcn20_patch_bounding_box(dc, loaded_bb); + DC_FP_END(); if (dc->ctx->dc_bios->funcs->get_soc_bb_info) { struct bp_soc_bb_info bb_info = { 0 }; if (dc->ctx->dc_bios->funcs->get_soc_bb_info( dc->ctx->dc_bios, &bb_info) == BP_RESULT_OK) { - if (bb_info.dram_clock_change_latency_100ns > 0) - dcn3_02_soc.dram_clock_change_latency_us = - bb_info.dram_clock_change_latency_100ns * 10; - - if (bb_info.dram_sr_enter_exit_latency_100ns > 0) - dcn3_02_soc.sr_enter_plus_exit_time_us = - bb_info.dram_sr_enter_exit_latency_100ns * 10; - if (bb_info.dram_sr_exit_latency_100ns > 0) - dcn3_02_soc.sr_exit_time_us = - bb_info.dram_sr_exit_latency_100ns * 10; + DC_FP_START(); + dcn302_fpu_init_soc_bounding_box(bb_info); + DC_FP_END(); } } @@ -1257,170 +1125,11 @@ static void dcn302_destroy_resource_pool(struct resource_pool **pool) *pool = NULL; } -static void dcn302_get_optimal_dcfclk_fclk_for_uclk(unsigned int uclk_mts, - unsigned int *optimal_dcfclk, - unsigned int *optimal_fclk) -{ - double bw_from_dram, bw_from_dram1, bw_from_dram2; - - bw_from_dram1 = uclk_mts * dcn3_02_soc.num_chans * - dcn3_02_soc.dram_channel_width_bytes * (dcn3_02_soc.max_avg_dram_bw_use_normal_percent / 100); - bw_from_dram2 = uclk_mts * dcn3_02_soc.num_chans * - dcn3_02_soc.dram_channel_width_bytes * (dcn3_02_soc.max_avg_sdp_bw_use_normal_percent / 100); - - bw_from_dram = (bw_from_dram1 < bw_from_dram2) ? bw_from_dram1 : bw_from_dram2; - - if (optimal_fclk) - *optimal_fclk = bw_from_dram / - (dcn3_02_soc.fabric_datapath_to_dcn_data_return_bytes * (dcn3_02_soc.max_avg_sdp_bw_use_normal_percent / 100)); - - if (optimal_dcfclk) - *optimal_dcfclk = bw_from_dram / - (dcn3_02_soc.return_bus_width_bytes * (dcn3_02_soc.max_avg_sdp_bw_use_normal_percent / 100)); -} - void dcn302_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw_params) { - unsigned int i, j; - unsigned int num_states = 0; - - unsigned int dcfclk_mhz[DC__VOLTAGE_STATES] = {0}; - unsigned int dram_speed_mts[DC__VOLTAGE_STATES] = {0}; - unsigned int optimal_uclk_for_dcfclk_sta_targets[DC__VOLTAGE_STATES] = {0}; - unsigned int optimal_dcfclk_for_uclk[DC__VOLTAGE_STATES] = {0}; - - unsigned int dcfclk_sta_targets[DC__VOLTAGE_STATES] = {694, 875, 1000, 1200}; - unsigned int num_dcfclk_sta_targets = 4; - unsigned int num_uclk_states; - - - if (dc->ctx->dc_bios->vram_info.num_chans) - dcn3_02_soc.num_chans = dc->ctx->dc_bios->vram_info.num_chans; - - if (dc->ctx->dc_bios->vram_info.dram_channel_width_bytes) - dcn3_02_soc.dram_channel_width_bytes = dc->ctx->dc_bios->vram_info.dram_channel_width_bytes; - - dcn3_02_soc.dispclk_dppclk_vco_speed_mhz = dc->clk_mgr->dentist_vco_freq_khz / 1000.0; - dc->dml.soc.dispclk_dppclk_vco_speed_mhz = dc->clk_mgr->dentist_vco_freq_khz / 1000.0; - - if (bw_params->clk_table.entries[0].memclk_mhz) { - int max_dcfclk_mhz = 0, max_dispclk_mhz = 0, max_dppclk_mhz = 0, max_phyclk_mhz = 0; - - for (i = 0; i < MAX_NUM_DPM_LVL; i++) { - if (bw_params->clk_table.entries[i].dcfclk_mhz > max_dcfclk_mhz) - max_dcfclk_mhz = bw_params->clk_table.entries[i].dcfclk_mhz; - if (bw_params->clk_table.entries[i].dispclk_mhz > max_dispclk_mhz) - max_dispclk_mhz = bw_params->clk_table.entries[i].dispclk_mhz; - if (bw_params->clk_table.entries[i].dppclk_mhz > max_dppclk_mhz) - max_dppclk_mhz = bw_params->clk_table.entries[i].dppclk_mhz; - if (bw_params->clk_table.entries[i].phyclk_mhz > max_phyclk_mhz) - max_phyclk_mhz = bw_params->clk_table.entries[i].phyclk_mhz; - } - if (!max_dcfclk_mhz) - max_dcfclk_mhz = dcn3_02_soc.clock_limits[0].dcfclk_mhz; - if (!max_dispclk_mhz) - max_dispclk_mhz = dcn3_02_soc.clock_limits[0].dispclk_mhz; - if (!max_dppclk_mhz) - max_dppclk_mhz = dcn3_02_soc.clock_limits[0].dppclk_mhz; - if (!max_phyclk_mhz) - max_phyclk_mhz = dcn3_02_soc.clock_limits[0].phyclk_mhz; - - if (max_dcfclk_mhz > dcfclk_sta_targets[num_dcfclk_sta_targets-1]) { - /* If max DCFCLK is greater than the max DCFCLK STA target, insert into the DCFCLK STA target array */ - dcfclk_sta_targets[num_dcfclk_sta_targets] = max_dcfclk_mhz; - num_dcfclk_sta_targets++; - } else if (max_dcfclk_mhz < dcfclk_sta_targets[num_dcfclk_sta_targets-1]) { - /* If max DCFCLK is less than the max DCFCLK STA target, cap values and remove duplicates */ - for (i = 0; i < num_dcfclk_sta_targets; i++) { - if (dcfclk_sta_targets[i] > max_dcfclk_mhz) { - dcfclk_sta_targets[i] = max_dcfclk_mhz; - break; - } - } - /* Update size of array since we "removed" duplicates */ - num_dcfclk_sta_targets = i + 1; - } - - num_uclk_states = bw_params->clk_table.num_entries; - - /* Calculate optimal dcfclk for each uclk */ - for (i = 0; i < num_uclk_states; i++) { - dcn302_get_optimal_dcfclk_fclk_for_uclk(bw_params->clk_table.entries[i].memclk_mhz * 16, - &optimal_dcfclk_for_uclk[i], NULL); - if (optimal_dcfclk_for_uclk[i] < bw_params->clk_table.entries[0].dcfclk_mhz) { - optimal_dcfclk_for_uclk[i] = bw_params->clk_table.entries[0].dcfclk_mhz; - } - } - - /* Calculate optimal uclk for each dcfclk sta target */ - for (i = 0; i < num_dcfclk_sta_targets; i++) { - for (j = 0; j < num_uclk_states; j++) { - if (dcfclk_sta_targets[i] < optimal_dcfclk_for_uclk[j]) { - optimal_uclk_for_dcfclk_sta_targets[i] = - bw_params->clk_table.entries[j].memclk_mhz * 16; - break; - } - } - } - - i = 0; - j = 0; - /* create the final dcfclk and uclk table */ - while (i < num_dcfclk_sta_targets && j < num_uclk_states && num_states < DC__VOLTAGE_STATES) { - if (dcfclk_sta_targets[i] < optimal_dcfclk_for_uclk[j] && i < num_dcfclk_sta_targets) { - dcfclk_mhz[num_states] = dcfclk_sta_targets[i]; - dram_speed_mts[num_states++] = optimal_uclk_for_dcfclk_sta_targets[i++]; - } else { - if (j < num_uclk_states && optimal_dcfclk_for_uclk[j] <= max_dcfclk_mhz) { - dcfclk_mhz[num_states] = optimal_dcfclk_for_uclk[j]; - dram_speed_mts[num_states++] = bw_params->clk_table.entries[j++].memclk_mhz * 16; - } else { - j = num_uclk_states; - } - } - } - - while (i < num_dcfclk_sta_targets && num_states < DC__VOLTAGE_STATES) { - dcfclk_mhz[num_states] = dcfclk_sta_targets[i]; - dram_speed_mts[num_states++] = optimal_uclk_for_dcfclk_sta_targets[i++]; - } - - while (j < num_uclk_states && num_states < DC__VOLTAGE_STATES && - optimal_dcfclk_for_uclk[j] <= max_dcfclk_mhz) { - dcfclk_mhz[num_states] = optimal_dcfclk_for_uclk[j]; - dram_speed_mts[num_states++] = bw_params->clk_table.entries[j++].memclk_mhz * 16; - } - - dcn3_02_soc.num_states = num_states; - for (i = 0; i < dcn3_02_soc.num_states; i++) { - dcn3_02_soc.clock_limits[i].state = i; - dcn3_02_soc.clock_limits[i].dcfclk_mhz = dcfclk_mhz[i]; - dcn3_02_soc.clock_limits[i].fabricclk_mhz = dcfclk_mhz[i]; - dcn3_02_soc.clock_limits[i].dram_speed_mts = dram_speed_mts[i]; - - /* Fill all states with max values of all other clocks */ - dcn3_02_soc.clock_limits[i].dispclk_mhz = max_dispclk_mhz; - dcn3_02_soc.clock_limits[i].dppclk_mhz = max_dppclk_mhz; - dcn3_02_soc.clock_limits[i].phyclk_mhz = max_phyclk_mhz; - /* Populate from bw_params for DTBCLK, SOCCLK */ - if (!bw_params->clk_table.entries[i].dtbclk_mhz && i > 0) - dcn3_02_soc.clock_limits[i].dtbclk_mhz = dcn3_02_soc.clock_limits[i-1].dtbclk_mhz; - else - dcn3_02_soc.clock_limits[i].dtbclk_mhz = bw_params->clk_table.entries[i].dtbclk_mhz; - if (!bw_params->clk_table.entries[i].socclk_mhz && i > 0) - dcn3_02_soc.clock_limits[i].socclk_mhz = dcn3_02_soc.clock_limits[i-1].socclk_mhz; - else - dcn3_02_soc.clock_limits[i].socclk_mhz = bw_params->clk_table.entries[i].socclk_mhz; - /* These clocks cannot come from bw_params, always fill from dcn3_02_soc[1] */ - /* FCLK, PHYCLK_D18, DSCCLK */ - dcn3_02_soc.clock_limits[i].phyclk_d18_mhz = dcn3_02_soc.clock_limits[0].phyclk_d18_mhz; - dcn3_02_soc.clock_limits[i].dscclk_mhz = dcn3_02_soc.clock_limits[0].dscclk_mhz; - } - /* re-init DML with updated bb */ - dml_init_instance(&dc->dml, &dcn3_02_soc, &dcn3_02_ip, DML_PROJECT_DCN30); - if (dc->current_state) - dml_init_instance(&dc->current_state->bw_ctx.dml, &dcn3_02_soc, &dcn3_02_ip, DML_PROJECT_DCN30); - } + DC_FP_START(); + dcn302_fpu_update_bw_bounding_box(dc, bw_params); + DC_FP_END(); } static struct resource_funcs dcn302_res_pool_funcs = { @@ -1515,9 +1224,9 @@ static bool dcn302_resource_construct( /* total size = mall per channel * num channels * 1024 * 1024 */ dc->caps.mall_size_total = dc->caps.mall_size_per_mem_channel * dc->ctx->dc_bios->vram_info.num_chans * 1048576; dc->caps.cursor_cache_size = dc->caps.max_cursor_size * dc->caps.max_cursor_size * 8; - dc->caps.max_slave_planes = 1; - dc->caps.max_slave_yuv_planes = 1; - dc->caps.max_slave_rgb_planes = 1; + dc->caps.max_slave_planes = 2; + dc->caps.max_slave_yuv_planes = 2; + dc->caps.max_slave_rgb_planes = 2; dc->caps.post_blend_color_processing = true; dc->caps.force_dp_tps4_for_cp2520 = true; dc->caps.extended_aux_timeout_support = true; diff --git a/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.h b/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.h index 42d2c73e30bc..9f24e73b92b3 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.h +++ b/drivers/gpu/drm/amd/display/dc/dcn302/dcn302_resource.h @@ -28,6 +28,9 @@ #include "core_types.h" +extern struct _vcs_dpi_ip_params_st dcn3_02_ip; +extern struct _vcs_dpi_soc_bounding_box_st dcn3_02_soc; + struct resource_pool *dcn302_create_resource_pool(const struct dc_init_data *init_data, struct dc *dc); void dcn302_update_bw_bounding_box(struct dc *dc, struct clk_bw_params *bw_params); |