Skip to contentSkip to conversation

drm/amdgpu/ci: disable mclk switching for high refresh rates

PATCH In review Alex Deucher 2 participants 6 messages May 11, 2017 – May 12, 2017 4 messages not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/4 drm/amdgpu/ci: disable mclk switching for high refresh rates

Even if the vblank period would allow it, it still seems to be problematic on some cards.

---
 drivers/gpu/drm/amd/amdgpu/ci_dpm.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/gpu/drm/amd/amdgpu/ci_dpm.c +6 -0
index 6dc1410..eb5d03b 100644
@@ -906,6 +906,12 @@ static bool ci_dpm_vblank_too_short(struct amdgpu_device *adev)
 	u32 vblank_time = amdgpu_dpm_get_vblank_time(adev);
 	u32 switch_limit = adev->mc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ? 450 : 300;
 
+	/* disable mclk switching if the refresh is >120Hz, even if the
+	 * blanking period would allow it
+	 */
+	if (amdgpu_dpm_get_vrefresh(adev) > 120)
+		return false;
+
 	if (vblank_time < switch_limit)
 		return true;
 	else
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1494525521-20864-1-git-send-email-alexander.deucher-5C7GfCeVMHo@public.gmane.org

#2 lore raw

PATCH 2/4 drm/radeon/ci: disable mclk switching for high refresh rates

Even if the vblank period would allow it, it still seems to be problematic on some cards.

---
 drivers/gpu/drm/radeon/ci_dpm.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/gpu/drm/radeon/ci_dpm.c +6 -0
index 7ba4508..5ced733 100644
@@ -776,6 +776,12 @@ bool ci_dpm_vblank_too_short(struct radeon_device *rdev)
 	u32 vblank_time = r600_dpm_get_vblank_time(rdev);
 	u32 switch_limit = pi->mem_gddr5 ? 450 : 300;
 
+	/* disable mclk switching if the refresh is >120Hz, even if the
+        * blanking period would allow it
+        */
+	if (r600_dpm_get_vrefresh(rdev) > 120)
+		return false;
+
 	if (vblank_time < switch_limit)
 		return true;
 	else
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#3 lore raw

PATCH 3/4 drm/amd/powerplay/smu7: add vblank check for mclk switching

In reply to a message not in the archive

Check to make sure the vblank period is long enough to support mclk switching.

---
 drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c | 31 +++++++++++++++++++++---
 1 file changed, 27 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c +27 -4
index a74a3db..d745065 100644
@@ -2655,6 +2655,28 @@ static int smu7_get_power_state_size(struct pp_hwmgr *hwmgr)
 	return sizeof(struct smu7_power_state);
 }
 
+static int smu7_vblank_too_short(struct pp_hwmgr *hwmgr,
+				 uint32_t vblank_time_us)
+{
+	struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend);
+	uint32_t switch_limit_us = data->is_memory_gddr5 ? 450 : 150;
+
+	switch (hwmgr->chip_id) {
+	case CHIP_POLARIS10:
+	case CHIP_POLARIS11:
+	case CHIP_POLARIS12:
+		switch_limit_us = data->is_memory_gddr5 ? 190 : 150;
+		break;
+	default:
+		switch_limit_us = data->is_memory_gddr5 ? 450 : 150;
+		break;
+	}
+
+	if (vblank_time_us < switch_limit_us)
+		return true;
+	else
+		return false;
+}
 
 static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,
 				struct pp_power_state *request_ps,
@@ -2669,6 +2691,7 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,
 	bool disable_mclk_switching;
 	bool disable_mclk_switching_for_frame_lock;
 	struct cgs_display_info info = {0};
+	struct cgs_mode_info mode_info = {0};
 	const struct phm_clock_and_voltage_limits *max_limits;
 	uint32_t i;
 	struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend);
@@ -2677,6 +2700,7 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,
 	int32_t count;
 	int32_t stable_pstate_sclk = 0, stable_pstate_mclk = 0;
 
+	info.mode_info = &mode_info;
 	data->battery_state = (PP_StateUILabel_Battery ==
 			request_ps->classification.ui_label);
 
@@ -2703,8 +2727,6 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,
 
 	cgs_get_active_displays_info(hwmgr->device, &info);
 
-	/*TO DO result = PHM_CheckVBlankTime(hwmgr, &vblankTooShort);*/
-
 	minimum_clocks.engineClock = hwmgr->display_config.min_core_set_clock;
 	minimum_clocks.memoryClock = hwmgr->display_config.min_mem_set_clock;
 
@@ -2769,8 +2791,9 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,
 				    PHM_PlatformCaps_DisableMclkSwitchingForFrameLock);
 
 
-	disable_mclk_switching = (1 < info.display_count) ||
-				    disable_mclk_switching_for_frame_lock;
+	disable_mclk_switching = ((1 < info.display_count) ||
+				  disable_mclk_switching_for_frame_lock ||
+				  smu7_vblank_too_short(hwmgr, mode_info.vblank_time_us));
 
 	sclk = smu7_ps->performance_levels[0].engine_clock;
 	mclk = smu7_ps->performance_levels[0].memory_clock;
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1494525521-20864-3-git-send-email-alexander.deucher-5C7GfCeVMHo@public.gmane.org

#5 lore raw

Re: [PATCH 3/4] drm/amd/powerplay/smu7: add vblank check for mclk switching

Hi Alex,

Comment below

Den 11 maj 2017 7:59 em skrev "Alex Deucher" <[email protected]>:

Check to make sure the vblank period is long enough to support mclk switching.

---
 drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c | 31
+++++++++++++++++++++---
 1 file changed, 27 insertions(+), 4 deletions(-)
(quoted diff) +0 -0

b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c index a74a3db..d745065 100644

drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c +0 -0
@@ -2655,6 +2655,28 @@ static int smu7_get_power_state_size(struct pp_hwmgr

*hwmgr)

        return sizeof(struct smu7_power_state);

}

+static int smu7_vblank_too_short(struct pp_hwmgr *hwmgr,
+ uint32_t vblank_time_us)
+{
+ struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend);
+ uint32_t switch_limit_us = data->is_memory_gddr5 ? 450 : 150;

This assignment is overwritten below.

BR
Nils

+

+       switch (hwmgr->chip_id) {
+       case CHIP_POLARIS10:
+       case CHIP_POLARIS11:
+       case CHIP_POLARIS12:

+ switch_limit_us = data->is_memory_gddr5 ? 190 : 150;
+ break;
+ default:
+ switch_limit_us = data->is_memory_gddr5 ? 450 : 150;
+ break;
+ }
+
+ if (vblank_time_us < switch_limit_us)
+ return true;
+ else
+ return false;
+}

static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,
struct pp_power_state *request_ps,
@@ -2669,6 +2691,7 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,

        bool disable_mclk_switching;
        bool disable_mclk_switching_for_frame_lock;
        struct cgs_display_info info = {0};
+       struct cgs_mode_info mode_info = {0};
        const struct phm_clock_and_voltage_limits *max_limits;
        uint32_t i;
        struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend);
@@ -2677,6 +2700,7 @@ static int smu7_apply_state_adjust_rules(struct
pp_hwmgr *hwmgr,
        int32_t count;
        int32_t stable_pstate_sclk = 0, stable_pstate_mclk = 0;
+       info.mode_info = &mode_info;
        data->battery_state = (PP_StateUILabel_Battery ==
                        request_ps->classification.ui_label);

@@ -2703,8 +2727,6 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,

        cgs_get_active_displays_info(hwmgr->device, &info);

- /*TO DO result = PHM_CheckVBlankTime(hwmgr, &vblankTooShort);*/
-

        minimum_clocks.engineClock = hwmgr->display_config.min_
core_set_clock;
        minimum_clocks.memoryClock = hwmgr->display_config.min_mem_

set_clock;

@@ -2769,8 +2791,9 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,
PHM_PlatformCaps_
DisableMclkSwitchingForFrameLock);

- disable_mclk_switching = (1 < info.display_count) ||
- disable_mclk_switching_for_frame_lock;
+ disable_mclk_switching = ((1 < info.display_count) ||
+ disable_mclk_switching_for_frame_lock ||
+ smu7_vblank_too_short(hwmgr,
mode_info.vblank_time_us));

        sclk = smu7_ps->performance_levels[0].engine_clock;
        mclk = smu7_ps->performance_levels[0].memory_clock;

--
2.5.5

_______________________________________________
amd-gfx mailing list [email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#6 lore raw

PATCH 4/4 drm/amd/powerplay/smu7: disable mclk switching for high refresh rates

In reply to a message not in the archive

Even if the vblank period would allow it, it still seems to be problematic on some cards.

---
 drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c +2 -1
index d745065..98a1463 100644
@@ -2793,7 +2793,8 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,
 
 	disable_mclk_switching = ((1 < info.display_count) ||
 				  disable_mclk_switching_for_frame_lock ||
-				  smu7_vblank_too_short(hwmgr, mode_info.vblank_time_us));
+				  smu7_vblank_too_short(hwmgr, mode_info.vblank_time_us) ||
+				  (mode_info.refresh_rate > 120));
 
 	sclk = smu7_ps->performance_levels[0].engine_clock;
 	mclk = smu7_ps->performance_levels[0].memory_clock;
Signature · 6 lines
2.5.5

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): CA+nq7DsJ=Bd4_yM=0xsv4a3MB2dyyVF_UfJE1h5Y4980cdF2nw@mail.gmail.com

A message that is not in the archive (sent off-list or not yet received): CA+nq7DsJ=Bd4_yM=0xsv4a3MB2dyyVF_UfJE1h5Y4980cdF2nw-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#9 lore raw

Hi, Alex

Shouldn't the function return true when the refresh rate > 120? If so, same comment on patch 2.

BR
Nils

Den 11 maj 2017 7:59 em skrev "Alex Deucher" <[email protected]>:

Even if the vblank period would allow it, it still seems to be problematic on some cards.

---
 drivers/gpu/drm/amd/amdgpu/ci_dpm.c | 6 ++++++
 1 file changed, 6 insertions(+)
(quoted diff) +0 -0

b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
index 6dc1410..eb5d03b 100644

drivers/gpu/drm/amd/amdgpu/ci_dpm.c +0 -0
@@ -906,6 +906,12 @@ static bool ci_dpm_vblank_too_short(struct

amdgpu_device *adev)

        u32 vblank_time = amdgpu_dpm_get_vblank_time(adev);
        u32 switch_limit = adev->mc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ?

450 : 300;

+ /* disable mclk switching if the refresh is >120Hz, even if the
+ * blanking period would allow it
+ */
+ if (amdgpu_dpm_get_vrefresh(adev) > 120)
+ return false;
+
if (vblank_time < switch_limit)
return true;
else
--
2.5.5

_______________________________________________
amd-gfx mailing list [email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx