[PATCH 386/459] drm/amd/display: Optimize bandwidth validation by adding early return

[Alex Deucher]

From: Joshua Aberback <joshua.aberback at amd.com>

We can split validation into three parts: getting voltage level, getting
watermarks, and rq/dlg calculations. The voltage level is enough to answer
the question "do we support this state", and the rest of it is to determine
what hardware programming is needed to support the state. Most of the calls
to validate_bandwidth only care about the first part, so we added an early
return in that case

Signed-off-by: Joshua Aberback <joshua.aberback at amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng at amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet Lakha at amd.com>
Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
---
 .../gpu/drm/amd/display/dc/calcs/dcn_calcs.c  | 11 +++---
 .../drm/amd/display/dc/dcn20/dcn20_hwseq.c    |  3 +-
 .../drm/amd/display/dc/dcn20/dcn20_resource.c | 35 ++++++++++++++-----
 3 files changed, 31 insertions(+), 18 deletions(-)

diff --git a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c
index ef804948694e..38365dd911a3 100644
--- a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c
+++ b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c
@@ -1116,9 +1116,8 @@ bool dcn_validate_bandwidth(
 
 		context->bw_ctx.bw.dcn.clk.fclk_khz = (int)(bw_consumed * 1000000 /
 				(ddr4_dram_factor_single_Channel * v->number_of_channels));
-		if (bw_consumed == v->fabric_and_dram_bandwidth_vmin0p65) {
+		if (bw_consumed == v->fabric_and_dram_bandwidth_vmin0p65)
 			context->bw_ctx.bw.dcn.clk.fclk_khz = (int)(bw_consumed * 1000000 / 32);
-		}
 
 		context->bw_ctx.bw.dcn.clk.dcfclk_deep_sleep_khz = (int)(v->dcf_clk_deep_sleep * 1000);
 		context->bw_ctx.bw.dcn.clk.dcfclk_khz = (int)(v->dcfclk * 1000);
@@ -1133,7 +1132,8 @@ bool dcn_validate_bandwidth(
 					dc->debug.min_disp_clk_khz;
 		}
 
-		context->bw_ctx.bw.dcn.clk.dppclk_khz = context->bw_ctx.bw.dcn.clk.dispclk_khz / v->dispclk_dppclk_ratio;
+		context->bw_ctx.bw.dcn.clk.dppclk_khz = context->bw_ctx.bw.dcn.clk.dispclk_khz /
+				v->dispclk_dppclk_ratio;
 		context->bw_ctx.bw.dcn.clk.phyclk_khz = v->phyclk_per_state[v->voltage_level];
 		switch (v->voltage_level) {
 		case 0:
@@ -1220,9 +1220,7 @@ bool dcn_validate_bandwidth(
 						/* pipe not split previously needs split */
 						hsplit_pipe = find_idle_secondary_pipe(&context->res_ctx, pool, pipe);
 						ASSERT(hsplit_pipe);
-						split_stream_across_pipes(
-							&context->res_ctx, pool,
-							pipe, hsplit_pipe);
+						split_stream_across_pipes(&context->res_ctx, pool, pipe, hsplit_pipe);
 					}
 
 					dcn_bw_calc_rq_dlg_ttu(dc, v, hsplit_pipe, input_idx);
@@ -1253,7 +1251,6 @@ bool dcn_validate_bandwidth(
 	}
 
 	if (v->voltage_level == 0) {
-
 		context->bw_ctx.dml.soc.sr_enter_plus_exit_time_us =
 				dc->dcn_soc->sr_enter_plus_exit_time;
 		context->bw_ctx.dml.soc.sr_exit_time_us = dc->dcn_soc->sr_exit_time;
diff --git a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c
index 542f144f807f..dc34ce28505c 100644
--- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c
+++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c
@@ -1415,9 +1415,8 @@ bool dcn20_update_bandwidth(
 	int i;
 
 	/* recalculate DML parameters */
-	if (!dc->res_pool->funcs->validate_bandwidth(dc, context, false)) {
+	if (!dc->res_pool->funcs->validate_bandwidth(dc, context, false))
 		return false;
-	}
 
 	/* apply updated bandwidth parameters */
 	dc->hwss.prepare_bandwidth(dc, context);
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 e3cd3e3db2c4..371d96290d86 100644
--- a/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c
+++ b/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c
@@ -1907,10 +1907,11 @@ static bool dcn20_validate_dsc(struct dc *dc, struct dc_state *new_ctx)
 }
 #endif
 
-bool dcn20_validate_bandwidth(struct dc *dc,
-			      struct dc_state *context,
-			      bool fast_validate)
+bool dcn20_validate_bandwidth(struct dc *dc, struct dc_state *context,
+		bool fast_validate)
 {
+	bool out = false;
+
 	int pipe_cnt, i, pipe_idx, vlevel, vlevel_unsplit;
 	int pipe_split_from[MAX_PIPES];
 	bool odm_capable = context->bw_ctx.dml.ip.odm_capable;
@@ -1954,11 +1955,16 @@ bool dcn20_validate_bandwidth(struct dc *dc,
 	}
 
 	pipe_cnt = dcn20_populate_dml_pipes_from_context(dc, &context->res_ctx, pipes);
-	if (!pipe_cnt)
-		goto validate_pass;
+
+	if (!pipe_cnt) {
+		out = true;
+		goto validate_out;
+	}
 
 	context->bw_ctx.dml.ip.odm_capable = 0;
+
 	vlevel = dml_get_voltage_level(&context->bw_ctx.dml, pipes, pipe_cnt);
+
 	context->bw_ctx.dml.ip.odm_capable = odm_capable;
 
 #ifdef CONFIG_DRM_AMD_DC_DSC_SUPPORT
@@ -2111,6 +2117,11 @@ bool dcn20_validate_bandwidth(struct dc *dc,
 	}
 #endif
 
+	if (fast_validate) {
+		out = true;
+		goto validate_out;
+	}
+
 	for (i = 0, pipe_idx = 0, pipe_cnt = 0; i < dc->res_pool->pipe_count; i++) {
 		if (!context->res_ctx.pipe_ctx[i].stream)
 			continue;
@@ -2235,21 +2246,27 @@ bool dcn20_validate_bandwidth(struct dc *dc,
 				pipe_idx,
 				cstate_en,
 				context->bw_ctx.bw.dcn.clk.p_state_change_support);
+
 		context->bw_ctx.dml.funcs.rq_dlg_get_rq_reg(&context->bw_ctx.dml,
 				&context->res_ctx.pipe_ctx[i].rq_regs,
 				pipes[pipe_idx].pipe);
+
 		pipe_idx++;
 	}
 
-validate_pass:
-	kfree(pipes);
-	return true;
+	out = true;
+	goto validate_out;
 
 validate_fail:
 	DC_LOG_WARNING("Mode Validation Warning: %s failed validation.\n",
 		dml_get_status_message(context->bw_ctx.dml.vba.ValidationStatus[context->bw_ctx.dml.vba.soc.num_states]));
+
+	out = false;
+
+validate_out:
 	kfree(pipes);
-	return false;
+
+	return out;
 }
 
 struct pipe_ctx *dcn20_acquire_idle_pipe_for_layer(
-- 
2.20.1