Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit e4d8d46c authored by Huaibin Yang's avatar Huaibin Yang Committed by Ingrid Gallardo
Browse files

msm: mdss: reduce mdp bandwidth for single pipe use case 



For single layer use case, mdp prefill bandwidth request can be
reduced because mdp pipe operates like a fifo and there is no need to
fill all buffers before the active region.

Change-Id: Ic087f394b2551658543cff7f99789eace0388e07
Signed-off-by: default avatarHuaibin Yang <huaibiny@codeaurora.org>
Signed-off-by: default avatarIngrid Gallardo <ingridg@codeaurora.org>
parent 2ab645b4

drivers/video/msm/mdss/mdss_mdp_ctl.c

+76 −7
Original line number Diff line number Diff line
@@ -86,6 +86,7 @@ struct mdss_mdp_prefill_params { 
	bool is_bwc;

	bool is_tile;

	bool is_hflip;

	bool is_cmd;

};



static inline bool mdss_mdp_perf_is_caf(struct mdss_mdp_pipe *pipe)
@@ -291,6 +292,66 @@ static u32 mdss_mdp_perf_calc_pipe_prefill_cmd(struct mdss_mdp_prefill_params 
	return prefill_bytes;

}



u32 mdss_mdp_perf_calc_pipe_prefill_single(struct mdss_mdp_prefill_params

	*params)

{

	struct mdss_data_type *mdata = mdss_mdp_get_mdata();

	struct mdss_prefill_data *prefill = &mdata->prefill_data;

	u32 prefill_bytes;

	u32 latency_lines, latency_buf_bytes;

	u32 y_scaler_bytes;

	u32 fbc_cmd_lines = 0, fbc_cmd_bytes = 0;



	if (params->is_bwc || params->is_tile)

		/* can start processing after receiving 4 lines */

		latency_lines = 4;

	else if (!params->is_caf && params->is_hflip)

		/* need oneline before reading backwards */

		latency_lines = 1;

	else

		latency_lines = 0;

	latency_buf_bytes = params->src_w * params->bpp * latency_lines;

	prefill_bytes = latency_buf_bytes;



	y_scaler_bytes = mdss_mdp_calc_y_scaler_bytes(params, prefill);

	prefill_bytes += y_scaler_bytes;



	if (params->is_cmd)

		fbc_cmd_lines++;

	if (params->is_fbc)

		fbc_cmd_lines++;



	if (fbc_cmd_lines) {

		fbc_cmd_bytes = params->bpp * params->dst_w * fbc_cmd_lines;

		fbc_cmd_bytes = mdss_mdp_calc_scaling_w_h(fbc_cmd_bytes,

			params->src_h, params->dst_h, params->src_w,

			params->dst_w);

		prefill_bytes += fbc_cmd_bytes;

	}



	return prefill_bytes;

}



static inline bool mdss_mdp_is_single_pipe_per_mixer(

	struct mdss_mdp_mixer *mixer)

{

	bool is_single = true;

	int cnt = 0;

	int i;



	for (i = 0; i < MDSS_MDP_MAX_STAGE; i++) {

		struct mdss_mdp_pipe *pipe = mixer->stage_pipe[i];

		if (pipe) {

			cnt++;

			if (cnt > 1) {

				is_single = false;

				break;

			}

		}

	}

	return is_single;

}



/**

 * mdss_mdp_perf_calc_pipe() - calculate performance numbers required by pipe

 * @pipe:	Source pipe struct containing updated pipe params
@@ -389,6 +450,11 @@ int mdss_mdp_perf_calc_pipe(struct mdss_mdp_pipe *pipe, 
	else

		perf->mdp_clk_rate = rate;



	if (mixer->ctl->intf_num == MDSS_MDP_NO_INTF) {

		perf->prefill_bytes = 0;

		return 0;

	}



	prefill_params.smp_bytes = mdss_mdp_smp_get_size(pipe);

	prefill_params.xres = xres;

	prefill_params.src_w = src.w;
@@ -403,14 +469,17 @@ int mdss_mdp_perf_calc_pipe(struct mdss_mdp_pipe *pipe, 
	prefill_params.is_bwc = pipe->bwc_mode;

	prefill_params.is_tile = pipe->src_fmt->tile;

	prefill_params.is_hflip = pipe->flags & MDP_FLIP_LR;



	if (mixer->type == MDSS_MDP_MIXER_TYPE_INTF) {

		perf->prefill_bytes = (mixer->ctl->is_video_mode) ?

			mdss_mdp_perf_calc_pipe_prefill_video(&prefill_params) :

			mdss_mdp_perf_calc_pipe_prefill_cmd(&prefill_params);

	}

	prefill_params.is_cmd = !mixer->ctl->is_video_mode;



	if (mdss_mdp_is_single_pipe_per_mixer(mixer))

		perf->prefill_bytes =

			mdss_mdp_perf_calc_pipe_prefill_single(&prefill_params);

	else if (!prefill_params.is_cmd)

		perf->prefill_bytes =

			mdss_mdp_perf_calc_pipe_prefill_video(&prefill_params);

	else

		perf->prefill_bytes = 0;

		perf->prefill_bytes =

			mdss_mdp_perf_calc_pipe_prefill_cmd(&prefill_params);



	pr_debug("mixer=%d pnum=%d clk_rate=%u bw_overlap=%llu prefill=%d\n",

		 mixer->num, pipe->num, perf->mdp_clk_rate, perf->bw_overlap,