Re: [PATCH 2/2] drm/amd/display: Support uclk switching for DCN2

[Harry Wentland <hwentlan-5C7GfCeVMHo@public.gmane.org>]

On 2019-07-30 9:57 a.m., Nicholas Kazlauskas wrote:
> [Why]
> We were previously forcing the uclk for every state to max and reducing
> the switch time to prevent uclk switching from occuring. This workaround
> was previously needed in order to avoid hangs + underflow under certain
> display configurations.
> 
> Now that DC has the proper fix complete we can drop the hacks and
> improve power for most display configurations.
> 
> [How]
> We still need the function pointers hooked up to grab the real uclk
> states from pplib. The rest of the prior hack can be reverted.
> 
> The key requirements here are really just DC support, updated firmware,
> and support for disabling p-state support when needed in pplib/smu.
> 
> When these requirements are met uclk switching works without underflow
> or hangs.
> 
> Fixes: 02316e963a5a ("drm/amd/display: Force uclk to max for every state")
> 
> Cc: Leo Li <sunpeng.li@amd.com>
> Cc: Harry Wentland <harry.wentland@amd.com>
> Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>

Series is
Reviewed-by: Harry Wentland <harry.wentland@amd.com>

Harry

> ---
>  drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c | 10 ----------
>  1 file changed, 10 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c
> index ff30f5cc4981..42d3666f2037 100644
> --- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c
> +++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c
> @@ -2817,9 +2817,6 @@ static void cap_soc_clocks(
>  						&& max_clocks.uClockInKhz != 0)
>  			bb->clock_limits[i].dram_speed_mts = (max_clocks.uClockInKhz / 1000) * 16;
>  
> -		// HACK: Force every uclk to max for now to "disable" uclk switching.
> -		bb->clock_limits[i].dram_speed_mts = (max_clocks.uClockInKhz / 1000) * 16;
> -
>  		if ((bb->clock_limits[i].fabricclk_mhz > (max_clocks.fabricClockInKhz / 1000))
>  						&& max_clocks.fabricClockInKhz != 0)
>  			bb->clock_limits[i].fabricclk_mhz = (max_clocks.fabricClockInKhz / 1000);
> @@ -3035,8 +3032,6 @@ static bool init_soc_bounding_box(struct dc *dc,
>  				le32_to_cpu(bb->vmm_page_size_bytes);
>  		dcn2_0_soc.dram_clock_change_latency_us =
>  				fixed16_to_double_to_cpu(bb->dram_clock_change_latency_us);
> -		// HACK!! Lower uclock latency switch time so we don't switch
> -		dcn2_0_soc.dram_clock_change_latency_us = 10;
>  		dcn2_0_soc.writeback_dram_clock_change_latency_us =
>  				fixed16_to_double_to_cpu(bb->writeback_dram_clock_change_latency_us);
>  		dcn2_0_soc.return_bus_width_bytes =
> @@ -3078,7 +3073,6 @@ static bool init_soc_bounding_box(struct dc *dc,
>  		struct pp_smu_nv_clock_table max_clocks = {0};
>  		unsigned int uclk_states[8] = {0};
>  		unsigned int num_states = 0;
> -		int i;
>  		enum pp_smu_status status;
>  		bool clock_limits_available = false;
>  		bool uclk_states_available = false;
> @@ -3100,10 +3094,6 @@ static bool init_soc_bounding_box(struct dc *dc,
>  			clock_limits_available = (status == PP_SMU_RESULT_OK);
>  		}
>  
> -		// HACK: Use the max uclk_states value for all elements.
> -		for (i = 0; i < num_states; i++)
> -			uclk_states[i] = uclk_states[num_states - 1];
> -
>  		if (clock_limits_available && uclk_states_available && num_states)
>  			update_bounding_box(dc, &dcn2_0_soc, &max_clocks, uclk_states, num_states);
>  		else if (clock_limits_available)
> 
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx