drm/amd/display: optimize prgoram wm and clks

[Why]
In some display configuration like 1080P monitor playing a 1080P video,
if user use ALT+F4 to exit Movie and TV, there is a chance clocks are
same only water mark changed. Current clock optimization machanism will
result in water mark keeps high after exit Movie and TV app.

[How]
Return if watermark need to be optimized when doing program watermark,
perform the optimization after.

Signed-off-by: Yongqiang Sun <yongqiang.sun@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

1 files changed, 102 insertions, 35 deletions

diff --git a/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_hubbub.c b/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_hubbub.c
index f546260c15b7..8440975206e0 100644
--- a/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_hubbub.c
+++ b/drivers/gpu/drm/amd/display/dc/dcn21/dcn21_hubbub.c
@@ -141,7 +141,7 @@ int hubbub21_init_dchub(struct hubbub *hubbub,
 	return NUM_VMID;
 }
 
-void hubbub21_program_urgent_watermarks(
+bool hubbub21_program_urgent_watermarks(
 		struct hubbub *hubbub,
 		struct dcn_watermark_set *watermarks,
 		unsigned int refclk_mhz,
@@ -149,6 +149,7 @@ void hubbub21_program_urgent_watermarks(
 {
 	struct dcn20_hubbub *hubbub1 = TO_DCN20_HUBBUB(hubbub);
 	uint32_t prog_wm_value;
+	bool wm_pending = false;
 
 	/* Repeat for water mark set A, B, C and D. */
 	/* clock state A */
@@ -163,7 +164,8 @@ void hubbub21_program_urgent_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("URGENCY_WATERMARK_A calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->a.urgent_ns, prog_wm_value);
-	}
+	} else if (watermarks->a.urgent_ns < hubbub1->watermarks.a.urgent_ns)
+		wm_pending = true;
 
 	/* determine the transfer time for a quantity of data for a particular requestor.*/
 	if (safe_to_lower || watermarks->a.frac_urg_bw_flip
@@ -172,7 +174,9 @@ void hubbub21_program_urgent_watermarks(
 
 		REG_SET(DCHUBBUB_ARB_FRAC_URG_BW_FLIP_A, 0,
 				DCHUBBUB_ARB_FRAC_URG_BW_FLIP_A, watermarks->a.frac_urg_bw_flip);
-	}
+	} else if (watermarks->a.frac_urg_bw_flip
+			< hubbub1->watermarks.a.frac_urg_bw_flip)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->a.frac_urg_bw_nom
 			> hubbub1->watermarks.a.frac_urg_bw_nom) {
@@ -180,14 +184,18 @@ void hubbub21_program_urgent_watermarks(
 
 		REG_SET(DCHUBBUB_ARB_FRAC_URG_BW_NOM_A, 0,
 				DCHUBBUB_ARB_FRAC_URG_BW_NOM_A, watermarks->a.frac_urg_bw_nom);
-	}
+	} else if (watermarks->a.frac_urg_bw_nom
+			< hubbub1->watermarks.a.frac_urg_bw_nom)
+		wm_pending = true;
+
 	if (safe_to_lower || watermarks->a.urgent_latency_ns > hubbub1->watermarks.a.urgent_latency_ns) {
 		hubbub1->watermarks.a.urgent_latency_ns = watermarks->a.urgent_latency_ns;
 		prog_wm_value = convert_and_clamp(watermarks->a.urgent_latency_ns,
 				refclk_mhz, 0x1fffff);
 		REG_SET(DCHUBBUB_ARB_REFCYC_PER_TRIP_TO_MEMORY_A, 0,
 				DCHUBBUB_ARB_REFCYC_PER_TRIP_TO_MEMORY_A, prog_wm_value);
-	}
+	} else if (watermarks->a.urgent_latency_ns < hubbub1->watermarks.a.urgent_latency_ns)
+		wm_pending = true;
 
 	/* clock state B */
 	if (safe_to_lower || watermarks->b.urgent_ns > hubbub1->watermarks.b.urgent_ns) {
@@ -201,7 +209,8 @@ void hubbub21_program_urgent_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("URGENCY_WATERMARK_B calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->b.urgent_ns, prog_wm_value);
-	}
+	} else if (watermarks->b.urgent_ns < hubbub1->watermarks.b.urgent_ns)
+		wm_pending = true;
 
 	/* determine the transfer time for a quantity of data for a particular requestor.*/
 	if (safe_to_lower || watermarks->a.frac_urg_bw_flip
@@ -210,7 +219,9 @@ void hubbub21_program_urgent_watermarks(
 
 		REG_SET(DCHUBBUB_ARB_FRAC_URG_BW_FLIP_B, 0,
 				DCHUBBUB_ARB_FRAC_URG_BW_FLIP_B, watermarks->a.frac_urg_bw_flip);
-	}
+	} else if (watermarks->a.frac_urg_bw_flip
+			< hubbub1->watermarks.a.frac_urg_bw_flip)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->a.frac_urg_bw_nom
 			> hubbub1->watermarks.a.frac_urg_bw_nom) {
@@ -218,7 +229,9 @@ void hubbub21_program_urgent_watermarks(
 
 		REG_SET(DCHUBBUB_ARB_FRAC_URG_BW_NOM_B, 0,
 				DCHUBBUB_ARB_FRAC_URG_BW_NOM_B, watermarks->a.frac_urg_bw_nom);
-	}
+	} else if (watermarks->a.frac_urg_bw_nom
+			< hubbub1->watermarks.a.frac_urg_bw_nom)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->b.urgent_latency_ns > hubbub1->watermarks.b.urgent_latency_ns) {
 		hubbub1->watermarks.b.urgent_latency_ns = watermarks->b.urgent_latency_ns;
@@ -226,7 +239,8 @@ void hubbub21_program_urgent_watermarks(
 				refclk_mhz, 0x1fffff);
 		REG_SET(DCHUBBUB_ARB_REFCYC_PER_TRIP_TO_MEMORY_B, 0,
 				DCHUBBUB_ARB_REFCYC_PER_TRIP_TO_MEMORY_B, prog_wm_value);
-	}
+	} else if (watermarks->b.urgent_latency_ns < hubbub1->watermarks.b.urgent_latency_ns)
+		wm_pending = true;
 
 	/* clock state C */
 	if (safe_to_lower || watermarks->c.urgent_ns > hubbub1->watermarks.c.urgent_ns) {
@@ -240,7 +254,8 @@ void hubbub21_program_urgent_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("URGENCY_WATERMARK_C calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->c.urgent_ns, prog_wm_value);
-	}
+	} else if (watermarks->c.urgent_ns < hubbub1->watermarks.c.urgent_ns)
+		wm_pending = true;
 
 	/* determine the transfer time for a quantity of data for a particular requestor.*/
 	if (safe_to_lower || watermarks->a.frac_urg_bw_flip
@@ -249,7 +264,9 @@ void hubbub21_program_urgent_watermarks(
 
 		REG_SET(DCHUBBUB_ARB_FRAC_URG_BW_FLIP_C, 0,
 				DCHUBBUB_ARB_FRAC_URG_BW_FLIP_C, watermarks->a.frac_urg_bw_flip);
-	}
+	} else if (watermarks->a.frac_urg_bw_flip
+			< hubbub1->watermarks.a.frac_urg_bw_flip)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->a.frac_urg_bw_nom
 			> hubbub1->watermarks.a.frac_urg_bw_nom) {
@@ -257,7 +274,9 @@ void hubbub21_program_urgent_watermarks(
 
 		REG_SET(DCHUBBUB_ARB_FRAC_URG_BW_NOM_C, 0,
 				DCHUBBUB_ARB_FRAC_URG_BW_NOM_C, watermarks->a.frac_urg_bw_nom);
-	}
+	} else if (watermarks->a.frac_urg_bw_nom
+			< hubbub1->watermarks.a.frac_urg_bw_nom)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->c.urgent_latency_ns > hubbub1->watermarks.c.urgent_latency_ns) {
 		hubbub1->watermarks.c.urgent_latency_ns = watermarks->c.urgent_latency_ns;
@@ -265,7 +284,8 @@ void hubbub21_program_urgent_watermarks(
 				refclk_mhz, 0x1fffff);
 		REG_SET(DCHUBBUB_ARB_REFCYC_PER_TRIP_TO_MEMORY_C, 0,
 				DCHUBBUB_ARB_REFCYC_PER_TRIP_TO_MEMORY_C, prog_wm_value);
-	}
+	} else if (watermarks->c.urgent_latency_ns < hubbub1->watermarks.c.urgent_latency_ns)
+		wm_pending = true;
 
 	/* clock state D */
 	if (safe_to_lower || watermarks->d.urgent_ns > hubbub1->watermarks.d.urgent_ns) {
@@ -279,7 +299,8 @@ void hubbub21_program_urgent_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("URGENCY_WATERMARK_D calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->d.urgent_ns, prog_wm_value);
-	}
+	} else if (watermarks->d.urgent_ns < hubbub1->watermarks.d.urgent_ns)
+		wm_pending = true;
 
 	/* determine the transfer time for a quantity of data for a particular requestor.*/
 	if (safe_to_lower || watermarks->a.frac_urg_bw_flip
@@ -288,7 +309,9 @@ void hubbub21_program_urgent_watermarks(
 
 		REG_SET(DCHUBBUB_ARB_FRAC_URG_BW_FLIP_D, 0,
 				DCHUBBUB_ARB_FRAC_URG_BW_FLIP_D, watermarks->a.frac_urg_bw_flip);
-	}
+	} else if (watermarks->a.frac_urg_bw_flip
+			< hubbub1->watermarks.a.frac_urg_bw_flip)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->a.frac_urg_bw_nom
 			> hubbub1->watermarks.a.frac_urg_bw_nom) {
@@ -296,7 +319,9 @@ void hubbub21_program_urgent_watermarks(
 
 		REG_SET(DCHUBBUB_ARB_FRAC_URG_BW_NOM_D, 0,
 				DCHUBBUB_ARB_FRAC_URG_BW_NOM_D, watermarks->a.frac_urg_bw_nom);
-	}
+	} else if (watermarks->a.frac_urg_bw_nom
+			< hubbub1->watermarks.a.frac_urg_bw_nom)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->d.urgent_latency_ns > hubbub1->watermarks.d.urgent_latency_ns) {
 		hubbub1->watermarks.d.urgent_latency_ns = watermarks->d.urgent_latency_ns;
@@ -304,10 +329,13 @@ void hubbub21_program_urgent_watermarks(
 				refclk_mhz, 0x1fffff);
 		REG_SET(DCHUBBUB_ARB_REFCYC_PER_TRIP_TO_MEMORY_D, 0,
 				DCHUBBUB_ARB_REFCYC_PER_TRIP_TO_MEMORY_D, prog_wm_value);
-	}
+	} else if (watermarks->d.urgent_latency_ns < hubbub1->watermarks.d.urgent_latency_ns)
+		wm_pending = true;
+
+	return wm_pending;
 }
 
-void hubbub21_program_stutter_watermarks(
+bool hubbub21_program_stutter_watermarks(
 		struct hubbub *hubbub,
 		struct dcn_watermark_set *watermarks,
 		unsigned int refclk_mhz,
@@ -315,6 +343,7 @@ void hubbub21_program_stutter_watermarks(
 {
 	struct dcn20_hubbub *hubbub1 = TO_DCN20_HUBBUB(hubbub);
 	uint32_t prog_wm_value;
+	bool wm_pending = false;
 
 	/* clock state A */
 	if (safe_to_lower || watermarks->a.cstate_pstate.cstate_enter_plus_exit_ns
@@ -330,7 +359,9 @@ void hubbub21_program_stutter_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("SR_ENTER_EXIT_WATERMARK_A calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->a.cstate_pstate.cstate_enter_plus_exit_ns, prog_wm_value);
-	}
+	} else if (watermarks->a.cstate_pstate.cstate_enter_plus_exit_ns
+			< hubbub1->watermarks.a.cstate_pstate.cstate_enter_plus_exit_ns)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->a.cstate_pstate.cstate_exit_ns
 			> hubbub1->watermarks.a.cstate_pstate.cstate_exit_ns) {
@@ -345,7 +376,9 @@ void hubbub21_program_stutter_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("SR_EXIT_WATERMARK_A calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->a.cstate_pstate.cstate_exit_ns, prog_wm_value);
-	}
+	} else if (watermarks->a.cstate_pstate.cstate_exit_ns
+			< hubbub1->watermarks.a.cstate_pstate.cstate_exit_ns)
+		wm_pending = true;
 
 	/* clock state B */
 	if (safe_to_lower || watermarks->b.cstate_pstate.cstate_enter_plus_exit_ns
@@ -361,7 +394,9 @@ void hubbub21_program_stutter_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("SR_ENTER_EXIT_WATERMARK_B calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->b.cstate_pstate.cstate_enter_plus_exit_ns, prog_wm_value);
-	}
+	} else if (watermarks->b.cstate_pstate.cstate_enter_plus_exit_ns
+			< hubbub1->watermarks.b.cstate_pstate.cstate_enter_plus_exit_ns)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->b.cstate_pstate.cstate_exit_ns
 			> hubbub1->watermarks.b.cstate_pstate.cstate_exit_ns) {
@@ -376,7 +411,9 @@ void hubbub21_program_stutter_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("SR_EXIT_WATERMARK_B calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->b.cstate_pstate.cstate_exit_ns, prog_wm_value);
-	}
+	} else if (watermarks->b.cstate_pstate.cstate_exit_ns
+			< hubbub1->watermarks.b.cstate_pstate.cstate_exit_ns)
+		wm_pending = true;
 
 	/* clock state C */
 	if (safe_to_lower || watermarks->c.cstate_pstate.cstate_enter_plus_exit_ns
@@ -392,7 +429,9 @@ void hubbub21_program_stutter_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("SR_ENTER_EXIT_WATERMARK_C calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->c.cstate_pstate.cstate_enter_plus_exit_ns, prog_wm_value);
-	}
+	} else if (watermarks->c.cstate_pstate.cstate_enter_plus_exit_ns
+			< hubbub1->watermarks.c.cstate_pstate.cstate_enter_plus_exit_ns)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->c.cstate_pstate.cstate_exit_ns
 			> hubbub1->watermarks.c.cstate_pstate.cstate_exit_ns) {
@@ -407,7 +446,9 @@ void hubbub21_program_stutter_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("SR_EXIT_WATERMARK_C calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->c.cstate_pstate.cstate_exit_ns, prog_wm_value);
-	}
+	} else if (watermarks->c.cstate_pstate.cstate_exit_ns
+			< hubbub1->watermarks.c.cstate_pstate.cstate_exit_ns)
+		wm_pending = true;
 
 	/* clock state D */
 	if (safe_to_lower || watermarks->d.cstate_pstate.cstate_enter_plus_exit_ns
@@ -423,7 +464,9 @@ void hubbub21_program_stutter_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("SR_ENTER_EXIT_WATERMARK_D calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->d.cstate_pstate.cstate_enter_plus_exit_ns, prog_wm_value);
-	}
+	} else if (watermarks->d.cstate_pstate.cstate_enter_plus_exit_ns
+			< hubbub1->watermarks.d.cstate_pstate.cstate_enter_plus_exit_ns)
+		wm_pending = true;
 
 	if (safe_to_lower || watermarks->d.cstate_pstate.cstate_exit_ns
 			> hubbub1->watermarks.d.cstate_pstate.cstate_exit_ns) {
@@ -438,10 +481,14 @@ void hubbub21_program_stutter_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("SR_EXIT_WATERMARK_D calculated =%d\n"
 			"HW register value = 0x%x\n",
 			watermarks->d.cstate_pstate.cstate_exit_ns, prog_wm_value);
-	}
+	} else if (watermarks->d.cstate_pstate.cstate_exit_ns
+			< hubbub1->watermarks.d.cstate_pstate.cstate_exit_ns)
+		wm_pending = true;
+
+	return wm_pending;
 }
 
-void hubbub21_program_pstate_watermarks(
+bool hubbub21_program_pstate_watermarks(
 		struct hubbub *hubbub,
 		struct dcn_watermark_set *watermarks,
 		unsigned int refclk_mhz,
@@ -450,6 +497,8 @@ void hubbub21_program_pstate_watermarks(
 	struct dcn20_hubbub *hubbub1 = TO_DCN20_HUBBUB(hubbub);
 	uint32_t prog_wm_value;
 
+	bool wm_pending = false;
+
 	/* clock state A */
 	if (safe_to_lower || watermarks->a.cstate_pstate.pstate_change_ns
 			> hubbub1->watermarks.a.cstate_pstate.pstate_change_ns) {
@@ -464,7 +513,9 @@ void hubbub21_program_pstate_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("DRAM_CLK_CHANGE_WATERMARK_A calculated =%d\n"
 			"HW register value = 0x%x\n\n",
 			watermarks->a.cstate_pstate.pstate_change_ns, prog_wm_value);
-	}
+	} else if (watermarks->a.cstate_pstate.pstate_change_ns
+			< hubbub1->watermarks.a.cstate_pstate.pstate_change_ns)
+		wm_pending = true;
 
 	/* clock state B */
 	if (safe_to_lower || watermarks->b.cstate_pstate.pstate_change_ns
@@ -480,7 +531,9 @@ void hubbub21_program_pstate_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("DRAM_CLK_CHANGE_WATERMARK_B calculated =%d\n"
 			"HW register value = 0x%x\n\n",
 			watermarks->b.cstate_pstate.pstate_change_ns, prog_wm_value);
-	}
+	} else if (watermarks->b.cstate_pstate.pstate_change_ns
+			< hubbub1->watermarks.b.cstate_pstate.pstate_change_ns)
+		wm_pending = false;
 
 	/* clock state C */
 	if (safe_to_lower || watermarks->c.cstate_pstate.pstate_change_ns
@@ -496,7 +549,9 @@ void hubbub21_program_pstate_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("DRAM_CLK_CHANGE_WATERMARK_C calculated =%d\n"
 			"HW register value = 0x%x\n\n",
 			watermarks->c.cstate_pstate.pstate_change_ns, prog_wm_value);
-	}
+	} else if (watermarks->c.cstate_pstate.pstate_change_ns
+			< hubbub1->watermarks.c.cstate_pstate.pstate_change_ns)
+		wm_pending = true;
 
 	/* clock state D */
 	if (safe_to_lower || watermarks->d.cstate_pstate.pstate_change_ns
@@ -512,20 +567,30 @@ void hubbub21_program_pstate_watermarks(
 		DC_LOG_BANDWIDTH_CALCS("DRAM_CLK_CHANGE_WATERMARK_D calculated =%d\n"
 			"HW register value = 0x%x\n\n",
 			watermarks->d.cstate_pstate.pstate_change_ns, prog_wm_value);
-	}
+	} else if (watermarks->d.cstate_pstate.pstate_change_ns
+			< hubbub1->watermarks.d.cstate_pstate.pstate_change_ns)
+		wm_pending = true;
+
+	return wm_pending;
 }
 
-void hubbub21_program_watermarks(
+bool hubbub21_program_watermarks(
 		struct hubbub *hubbub,
 		struct dcn_watermark_set *watermarks,
 		unsigned int refclk_mhz,
 		bool safe_to_lower)
 {
 	struct dcn20_hubbub *hubbub1 = TO_DCN20_HUBBUB(hubbub);
+	bool wm_pending = false;
+
+	if (hubbub21_program_urgent_watermarks(hubbub, watermarks, refclk_mhz, safe_to_lower))
+		wm_pending = true;
 
-	hubbub21_program_urgent_watermarks(hubbub, watermarks, refclk_mhz, safe_to_lower);
-	hubbub21_program_stutter_watermarks(hubbub, watermarks, refclk_mhz, safe_to_lower);
-	hubbub21_program_pstate_watermarks(hubbub, watermarks, refclk_mhz, safe_to_lower);
+	if (hubbub21_program_stutter_watermarks(hubbub, watermarks, refclk_mhz, safe_to_lower))
+		wm_pending = true;
+
+	if (hubbub21_program_pstate_watermarks(hubbub, watermarks, refclk_mhz, safe_to_lower))
+		wm_pending = true;
 
 	/*
 	 * The DCHub arbiter has a mechanism to dynamically rate limit the DCHub request stream to the fabric.
@@ -549,6 +614,8 @@ void hubbub21_program_watermarks(
 			DCHUBBUB_ARB_MAX_QOS_COMMIT_THRESHOLD, 0xF);
 
 	hubbub1_allow_self_refresh_control(hubbub, !hubbub->ctx->dc->debug.disable_stutter);
+
+	return wm_pending;
 }
 
 void hubbub21_wm_read_state(struct hubbub *hubbub,