[Intel-gfx] [PATCH v5 5/9] drm/i915/tgl: Add helper function to prefer dc3co over dc5
Imre Deak
imre.deak at intel.com
Mon Aug 19 13:23:12 UTC 2019
On Mon, Aug 19, 2019 at 06:27:48PM +0530, Gupta, Anshuman wrote:
>
>
> On 8/13/2019 9:17 PM, Imre Deak wrote:
> > On Sat, Aug 10, 2019 at 12:02:19AM +0530, Anshuman Gupta wrote:
> > > We need to have a S/W flag based upon which driver can switch to DC3CO.
> > > If it is only edp display connected and it has psr2 capability,
> > > then set a prefer_dc3co flag to true, which will be used to
> > > switch to dc3co as well as to program DC3CO PSR2 transcoder
> > > early exitline event.
> > >
> > > Cc: Jani Nikula <jani.nikula at intel.com>
> > > Cc: Imre Deak <imre.deak at intel.com>
> > > Cc: Animesh Manna <animesh.manna at intel.com>
> > > Signed-off-by: Anshuman Gupta <anshuman.gupta at intel.com>
> > > ---
> > > drivers/gpu/drm/i915/display/intel_display.c | 5 +
> > > .../drm/i915/display/intel_display_power.c | 105 ++++++++++++++++++
> > > .../drm/i915/display/intel_display_power.h | 5 +
> > > drivers/gpu/drm/i915/i915_drv.h | 1 +
> > > drivers/gpu/drm/i915/intel_pm.c | 2 +-
> > > drivers/gpu/drm/i915/intel_pm.h | 2 +
> > > 6 files changed, 119 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> > > index 647f49ca86ff..1ec204c14a10 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > > @@ -6448,6 +6448,9 @@ static void haswell_crtc_enable(struct intel_crtc_state *pipe_config,
> > > if (WARN_ON(intel_crtc->active))
> > > return;
> > > + /* Enable PSR2 transcoder exit line */
> > > + if (pipe_config->has_psr2 && dev_priv->csr.prefer_dc3co)
> > > + tgl_enable_psr2_transcoder_exitline(pipe_config);
> >
> > This is part of PSR2 programming, so should be done somewhere below
> > intel_psr_enable() imo.
> >
> > > intel_encoders_pre_pll_enable(intel_crtc, pipe_config, state);
> > > @@ -13685,6 +13688,8 @@ static int intel_atomic_check(struct drm_device *dev,
> > > "[modeset]" : "[fastset]");
> > > }
> > > + tgl_prefer_dc3co_over_dc5_check(dev_priv, state);
> >
> > I think this belongs to intel_modeset_checks(), where we could also do
> > all necessary pipe locking (since now I can't see how we would prevent
> > allowing DC3CO when an external output is enabled asynchronously). This
> > would be akin to CDCLK rate change, but I defer on this to Ville.
> Hi Ville,
> Could you please provide your inputs on DC3CO design perspective,
> whether we should have pipe locking in order to maintain consistent state
> for dc3co.
>
> I feel there can be two way to maintain consistent state.
>
> 1. Ensure necessary locking on current crtc before calculating prefer_dc3co
> state, this will block the asynchronous commit.
>
> 2. We can make the prefer_dc3co state to false from enable powerwell
> callback which will trigger by getting a reference count for PG2 power
> domains if any external output is enabled.
>
> (DC3CO will be disallowed from power well enable callback if any external
> output is enabled akin to DC5/DC6, here we want to make prefer_dc3co atomic
> state consistent).
Actually the latest idea based on further discussions was: since all DC
states will be disabled if there is an external output enabled, we
wouldn't need to compute prefer_dc3co. We could just enable/disable
DC3CO (with a new enable_dc3co() API) whenever enabling/disabling PSR2
and during a page-flip/from the page-idle work. That together with the
current state of the dc_off power well determines if any of DC5/6 or
DC3co is enabled atm.
>
> Thanks ,
> Anshuman Gupta.
> >
> > > +
> > > return 0;
> > > fail:
> > > diff --git a/drivers/gpu/drm/i915/display/intel_display_power.c b/drivers/gpu/drm/i915/display/intel_display_power.c
> > > index 167839060154..04a02c88ff93 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_display_power.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_display_power.c
> > > @@ -18,6 +18,7 @@
> > > #include "intel_hotplug.h"
> > > #include "intel_sideband.h"
> > > #include "intel_tc.h"
> > > +#include "intel_pm.h"
> > > bool intel_display_power_well_is_enabled(struct drm_i915_private *dev_priv,
> > > enum i915_power_well_id power_well_id);
> > > @@ -791,6 +792,110 @@ static void gen9_set_dc_state(struct drm_i915_private *dev_priv, u32 state)
> > > dev_priv->csr.dc_state = val & mask;
> > > }
> > > +void tgl_enable_psr2_transcoder_exitline(struct intel_crtc_state *cstate)
> > > +{
> > > + u32 linetime_us, val, exit_scanlines;
> > > + u32 crtc_vdisplay = cstate->base.adjusted_mode.crtc_vdisplay;
> > > + struct drm_i915_private *dev_priv = to_i915(cstate->base.crtc->dev);
> > > +
> > > + if (WARN_ON(cstate->cpu_transcoder != TRANSCODER_A))
> > > + return;
> >
> > Where's the TRANSCODER-A restriction coming from?
> >
> > > +
> > > + linetime_us = fixed16_to_u32_round_up(intel_get_linetime_us(cstate));
> > > + if (WARN_ON(!linetime_us))
> > > + return;
> > > + /*
> > > + * DC3CO Exit time 200us B.Spec 49196
> > > + * PSR2 transcoder Early Exit scanlines = ROUNDUP(200 / line time) + 1
> > > + * Exit line event need to program above calculated scan lines before
> > > + * next VBLANK.
> > > + */
> > > + exit_scanlines = DIV_ROUND_UP(200, linetime_us) + 1;
> > > + if (WARN_ON(exit_scanlines > crtc_vdisplay))
> > > + return;
> > > +
> > > + exit_scanlines = crtc_vdisplay - exit_scanlines;
> > > + exit_scanlines <<= EXITLINE_SHIFT;
> > > + val = I915_READ(EXITLINE(cstate->cpu_transcoder));
> > > + val &= ~(EXITLINE_MASK | EXITLINE_ENABLE);
> > > + val |= exit_scanlines;
> > > + val |= EXITLINE_ENABLE;
> > > + I915_WRITE(EXITLINE(cstate->cpu_transcoder), val);
> > > +}
> > > +
> > > +static bool tgl_is_only_edp_connected(struct intel_crtc_state *crtc_state)
> > > +{
> > > + struct drm_atomic_state *state = crtc_state->base.state;
> > > + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc);
> > > + struct drm_connector *connector, *edp_connector = NULL;
> > > + struct drm_connector_state *connector_state;
> > > + int i;
> > > +
> > > + for_each_new_connector_in_state(state, connector, connector_state, i) {
> > > + if (connector_state->crtc != &crtc->base)
> > > + continue;
> > > +
> > > + if (connector->status == connector_status_connected &&
> > > + connector->connector_type != DRM_MODE_CONNECTOR_eDP)
> > > + return false;
> > > + else if (connector->connector_type == DRM_MODE_CONNECTOR_eDP &&
> > > + connector->status == connector_status_connected)
> > > + edp_connector = connector;
> > > + }
> > > +
> > > + if (edp_connector)
> > > + return true;
> > > +
> > > + return false;
> > > +}
> > > +
> > > +/*
> > > + * tgl_prefer_dc3co_over_dc5_check check whether it is worth to choose
> > > + * DC3CO over DC5. Currently it just check crtc psr2 capebilty and only
> > > + * edp display should be connected.
> > > + * TODO: Prefer DC3CO over DC5 only in video playback.
> > > + */
> > > +void tgl_prefer_dc3co_over_dc5_check(struct drm_i915_private *dev_priv,
> > > + struct intel_atomic_state *state)
> > > +{
> > > + struct intel_crtc_state *crtc_state, *mode_changed_cstate;
> > > + struct intel_crtc *crtc;
> > > + int i;
> > > + u32 val;
> > > +
> > > + dev_priv->csr.prefer_dc3co = false;
> > > +
> > > + if (!IS_TIGERLAKE(dev_priv))
> > > + return;
> > > +
> > > + if (!(dev_priv->csr.allowed_dc_mask & DC_STATE_EN_DC3CO))
> > > + return;
> > > +
> > > + for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) {
> > > + if (crtc->pipe == PIPE_A)
> > > + mode_changed_cstate = crtc_state;
> >
> > Where is the PIPE-A restriction coming from? Couldn't we simply pick the
> > crtc state that has the PSR2 sink?
> >
> > > + if (!crtc_state->has_psr2 && crtc_state->base.active) {
> > > + dev_priv->csr.prefer_dc3co = false;
> > > + return;
> > > + } else if (crtc_state->has_psr2) {
> > > + if (tgl_is_only_edp_connected(crtc_state) &&
> > > + crtc_state->base.active) {
> > > + dev_priv->csr.prefer_dc3co = true;
> > > + continue;
> > > + } else {
> > > + dev_priv->csr.prefer_dc3co = false;
> > > + return;
> > > + }
> > > + }
> > > + }
> > > +
> > > + if (dev_priv->csr.prefer_dc3co) {
> > > + val = I915_READ(EXITLINE(mode_changed_cstate->cpu_transcoder));
> >
> > No HW access during the modeset check phase is the rule. You could just
> > use the old prefer_dc3co value, no?
> >
> > > + if (!(val & EXITLINE_ENABLE))
> > > + mode_changed_cstate->base.mode_changed = true;
> > > + }
> > > +}
> > > +
> > > static void tgl_allow_dc3co(struct drm_i915_private *dev_priv)
> > > {
> > > gen9_set_dc_state(dev_priv, DC_STATE_EN_DC3CO);
> > > diff --git a/drivers/gpu/drm/i915/display/intel_display_power.h b/drivers/gpu/drm/i915/display/intel_display_power.h
> > > index 97f2562fc5d3..46e1bcfa490a 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_display_power.h
> > > +++ b/drivers/gpu/drm/i915/display/intel_display_power.h
> > > @@ -12,6 +12,8 @@
> > > struct drm_i915_private;
> > > struct intel_encoder;
> > > +struct intel_crtc_state;
> > > +struct intel_atomic_state;
> > > enum intel_display_power_domain {
> > > POWER_DOMAIN_DISPLAY_CORE,
> > > @@ -246,6 +248,9 @@ void intel_display_power_suspend_late(struct drm_i915_private *i915);
> > > void intel_display_power_resume_early(struct drm_i915_private *i915);
> > > void intel_display_power_suspend(struct drm_i915_private *i915);
> > > void intel_display_power_resume(struct drm_i915_private *i915);
> > > +void tgl_prefer_dc3co_over_dc5_check(struct drm_i915_private *dev_priv,
> > > + struct intel_atomic_state *state);
> > > +void tgl_enable_psr2_transcoder_exitline(struct intel_crtc_state *cstate);
> > > const char *
> > > intel_display_power_domain_str(struct drm_i915_private *i915,
> > > diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> > > index 3cdb5bf489f2..7ca0703209a4 100644
> > > --- a/drivers/gpu/drm/i915/i915_drv.h
> > > +++ b/drivers/gpu/drm/i915/i915_drv.h
> > > @@ -332,6 +332,7 @@ struct intel_csr {
> > > u32 dc_state;
> > > u32 allowed_dc_mask;
> > > intel_wakeref_t wakeref;
> > > + bool prefer_dc3co;
> > > };
> > > enum i915_cache_level {
> > > diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> > > index 780df8db2eba..634e43219164 100644
> > > --- a/drivers/gpu/drm/i915/intel_pm.c
> > > +++ b/drivers/gpu/drm/i915/intel_pm.c
> > > @@ -4584,7 +4584,7 @@ skl_wm_method2(u32 pixel_rate, u32 pipe_htotal, u32 latency,
> > > return ret;
> > > }
> > > -static uint_fixed_16_16_t
> > > +uint_fixed_16_16_t
> > > intel_get_linetime_us(const struct intel_crtc_state *crtc_state)
> > > {
> > > u32 pixel_rate;
> > > diff --git a/drivers/gpu/drm/i915/intel_pm.h b/drivers/gpu/drm/i915/intel_pm.h
> > > index e3573e1e16e3..454e92c06dff 100644
> > > --- a/drivers/gpu/drm/i915/intel_pm.h
> > > +++ b/drivers/gpu/drm/i915/intel_pm.h
> > > @@ -8,6 +8,7 @@
> > > #include <linux/types.h>
> > > +#include "i915_drv.h"
> > > #include "i915_reg.h"
> > > struct drm_device;
> > > @@ -76,6 +77,7 @@ u64 intel_rc6_residency_ns(struct drm_i915_private *dev_priv, i915_reg_t reg);
> > > u64 intel_rc6_residency_us(struct drm_i915_private *dev_priv, i915_reg_t reg);
> > > u32 intel_get_cagf(struct drm_i915_private *dev_priv, u32 rpstat1);
> > > +uint_fixed_16_16_t intel_get_linetime_us(const struct intel_crtc_state *cstate);
> > > unsigned long i915_chipset_val(struct drm_i915_private *dev_priv);
> > > unsigned long i915_mch_val(struct drm_i915_private *dev_priv);
> > > --
> > > 2.21.0
> > >
More information about the Intel-gfx
mailing list