[PATCH 1/2] drm/i915/display: Refactor handling of dpkgc latency programming

Govindapillai, Vinod vinod.govindapillai at intel.com
Fri Nov 15 11:50:28 UTC 2024


On Fri, 2024-11-15 at 10:28 +0000, Kandpal, Suraj wrote:
> 
> 
> > -----Original Message-----
> > From: Jani Nikula <jani.nikula at linux.intel.com>
> > Sent: Friday, November 15, 2024 3:50 PM
> > To: Kandpal, Suraj <suraj.kandpal at intel.com>; intel-xe at lists.freedesktop.org;
> > intel-gfx at lists.freedesktop.org
> > Cc: Govindapillai, Vinod <vinod.govindapillai at intel.com>; Syrjala, Ville
> > <ville.syrjala at intel.com>; Kandpal, Suraj <suraj.kandpal at intel.com>
> > Subject: Re: [PATCH 1/2] drm/i915/display: Refactor handling of dpkgc latency
> > programming
> > 
> > On Fri, 15 Nov 2024, Suraj Kandpal <suraj.kandpal at intel.com> wrote:
> > > - We want to make sure we have all the required values specially
> > > linetime which is computed after intel_wm_compute, this will also help
> > > implement some WA's which require linetime.
> > > -We do not want to write into any registers during compute_config
> > > phase While we are at it do some more refactors in the function like:
> > > -Use intel_display wherever possible
> > > -Move away from using enable_dpkgc bool and call it fixed_refresh_rate
> > > -Optimize value prepration
> > 
> > Please write proper sentences instead of bullet points. And the general advice
> > is to do one thing per patch.
> 
> Sure will fix that.
> It was previously divided into 6 patches but squashed it after Vinod suggested to do you want me
> to go back to that convention
> https://patchwork.freedesktop.org/series/141200/

Most of the old function handling was changed - from where and when it is called etc. Thats why I
suggested to have one patch as refactoring the whole handling.

> 
> > 
> > > 
> > > --v2
> > > -No need to save anything in intel_display structure [Vinod] -Move
> > > computation and writing into register to intel_atomic_commit_tail
> > > [Vinod]
> > > 
> > > --v3
> > > -Rename the subject [Vinod]
> > > -Rearrange the variable initialization and declaration [Vinod]
> > > -Reaarange condition evaluation for fixed_refresh_rate [Vinod]
> > > 
> > > Signed-off-by: Suraj Kandpal <suraj.kandpal at intel.com>
> > > ---
> > >  drivers/gpu/drm/i915/display/intel_display.c |  2 +
> > >  drivers/gpu/drm/i915/display/intel_wm.c      | 56 ++++++++++++++++++++
> > >  drivers/gpu/drm/i915/display/intel_wm.h      |  1 +
> > >  drivers/gpu/drm/i915/display/skl_watermark.c | 52 ------------------
> > >  4 files changed, 59 insertions(+), 52 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c
> > > b/drivers/gpu/drm/i915/display/intel_display.c
> > > index e790a2de5b3d..d1880e0a5d29 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > > @@ -7826,6 +7826,8 @@ static void intel_atomic_commit_tail(struct
> > intel_atomic_state *state)
> > >         /* Now enable the clocks, plane, pipe, and connectors that we set up.
> > */
> > >         dev_priv->display.funcs.display->commit_modeset_enables(state);
> > > 
> > > +       intel_program_dpkgc_latency(state);
> > > +
> > >         if (state->modeset)
> > >                 intel_set_cdclk_post_plane_update(state);
> > > 
> > > diff --git a/drivers/gpu/drm/i915/display/intel_wm.c
> > > b/drivers/gpu/drm/i915/display/intel_wm.c
> > > index d7dc49aecd27..4a2caf9cb03e 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_wm.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_wm.c
> > > @@ -7,9 +7,18 @@
> > > 
> > >  #include "i915_drv.h"
> > >  #include "i9xx_wm.h"
> > > +#include "intel_de.h"
> > >  #include "intel_display_types.h"
> > >  #include "intel_wm.h"
> > >  #include "skl_watermark.h"
> > > +#include "skl_watermark_regs.h"
> > > +
> > > +/*
> > > + * It is expected that DSB can do posted writes to every register in
> > > + * the pipe and planes within 100us. For flip queue use case, the
> > > + * recommended DSB execution time is 100us + one SAGV block time.
> > > + */
> > > +#define DSB_EXE_TIME 100
> > > 
> > >  /**
> > >   * intel_update_watermarks - update FIFO watermark values based on
> > > current modes @@ -131,6 +140,53 @@ bool intel_wm_plane_visible(const
> > struct intel_crtc_state *crtc_state,
> > >                 return plane_state->uapi.visible;
> > >  }
> > > 
> > > +/*
> > > + * If Fixed Refresh Rate or For VRR case Vmin = Vmax = Flipline:
> > > + * Program DEEP PKG_C_LATENCY Pkg C with highest valid latency from
> > > + * watermark level1 and up and above. If watermark level 1 is
> > > + * invalid program it with all 1's.
> > > + * Program PKG_C_LATENCY Added Wake Time = DSB execution time
> > > + * If Variable Refresh Rate where Vmin != Vmax != Flipline:
> > > + * Program DEEP PKG_C_LATENCY Pkg C with all 1's.
> > > + * Program PKG_C_LATENCY Added Wake Time = 0  */ void
> > > +intel_program_dpkgc_latency(struct intel_atomic_state *state)
> > 
> > The function prefix should match the file name i.e. a function in intel_foo.[ch]
> > should be named intel_foo_bar().
> > 
> 
> Well it was previously intel_wm_program_dpkgc_latency but Vinod had asked to rename the function

Ah.. sorry! I didnt notice that you moved this to intel_wm.c. I guess it has mainly for old
generation wm handling. I am not sure if "skl_watermarks.c" the right place. But atleast it is
better than intel_wm.c. 

> 
> 
> > But why is the function being relocated at all? That's also not explained in the
> > commit message.
> 
> The reason is in the commit message
> "- We want to make sure we have all the required values specially
> linetime which is computed after intel_wm_compute, this will also help
> implement some WA's which require linetime."
> 
> Will reword the commit message into sentences to make it more clearer.
> 
> Regards,
> Suraj Kandpal
> 
> > 
> > > +{
> > > +       struct intel_display *display = to_intel_display(state);
> > > +       struct drm_i915_private *i915 = to_i915(display->drm);
> > > +       struct intel_crtc *crtc;
> > > +       struct intel_crtc_state *new_crtc_state;
> > > +       u32 max_latency = LNL_PKG_C_LATENCY_MASK, added_waketime = 0;
> > > +       bool fixed_refresh_rate = false;
> > > +       u32 clear, val;
> > > +       int i;
> > > +
> > > +       if (DISPLAY_VER(display) < 20)
> > > +               return;
> > > +
> > > +       for_each_new_intel_crtc_in_state(state, crtc, new_crtc_state, i) {
> > > +               if (!new_crtc_state->vrr.enable ||
> > > +                   (new_crtc_state->vrr.vmin == new_crtc_state->vrr.vmax &&
> > > +                    new_crtc_state->vrr.vmin == new_crtc_state->vrr.flipline))
> > > +                       fixed_refresh_rate = true;
> > > +       }
> > > +
> > > +       if (fixed_refresh_rate) {
> > > +               max_latency = skl_watermark_max_latency(i915, 1);
> > > +               if (max_latency == 0)
> > > +                       max_latency = LNL_PKG_C_LATENCY_MASK;
> > > +               added_waketime = DSB_EXE_TIME +
> > > +                       display->sagv.block_time_us;
> > > +       }
> > > +
> > > +       clear = LNL_ADDED_WAKE_TIME_MASK |
> > LNL_PKG_C_LATENCY_MASK;
> > > +       val = REG_FIELD_PREP(LNL_PKG_C_LATENCY_MASK, max_latency) |
> > > +               REG_FIELD_PREP(LNL_ADDED_WAKE_TIME_MASK,
> > added_waketime);
> > > +
> > > +       intel_de_rmw(display, LNL_PKG_C_LATENCY, clear, val); }
> > > +
> > >  void intel_print_wm_latency(struct drm_i915_private *dev_priv,
> > >                             const char *name, const u16 wm[])  { diff --git
> > > a/drivers/gpu/drm/i915/display/intel_wm.h
> > > b/drivers/gpu/drm/i915/display/intel_wm.h
> > > index e97cdca89a5c..07e3d9359d73 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_wm.h
> > > +++ b/drivers/gpu/drm/i915/display/intel_wm.h
> > > @@ -31,5 +31,6 @@ void intel_print_wm_latency(struct drm_i915_private
> > *i915,
> > >                             const char *name, const u16 wm[]);  void
> > intel_wm_init(struct
> > > drm_i915_private *i915);  void intel_wm_debugfs_register(struct
> > > drm_i915_private *i915);
> > > +void intel_program_dpkgc_latency(struct intel_atomic_state *state);
> > > 
> > >  #endif /* __INTEL_WM_H__ */
> > > diff --git a/drivers/gpu/drm/i915/display/skl_watermark.c
> > > b/drivers/gpu/drm/i915/display/skl_watermark.c
> > > index 1a4c1fa24820..d419edb196c6 100644
> > > --- a/drivers/gpu/drm/i915/display/skl_watermark.c
> > > +++ b/drivers/gpu/drm/i915/display/skl_watermark.c
> > > @@ -28,12 +28,6 @@
> > >  #include "skl_watermark.h"
> > >  #include "skl_watermark_regs.h"
> > > 
> > > -/*It is expected that DSB can do posted writes to every register in
> > > - * the pipe and planes within 100us. For flip queue use case, the
> > > - * recommended DSB execution time is 100us + one SAGV block time.
> > > - */
> > > -#define DSB_EXE_TIME 100
> > > -
> > >  static void skl_sagv_disable(struct drm_i915_private *i915);
> > > 
> > >  /* Stores plane specific WM parameters */ @@ -2844,51 +2838,12 @@
> > > static int skl_wm_add_affected_planes(struct intel_atomic_state *state,
> > >         return 0;
> > >  }
> > > 
> > > -/*
> > > - * If Fixed Refresh Rate or For VRR case Vmin = Vmax = Flipline:
> > > - * Program DEEP PKG_C_LATENCY Pkg C with highest valid latency from
> > > - * watermark level1 and up and above. If watermark level 1 is
> > > - * invalid program it with all 1's.
> > > - * Program PKG_C_LATENCY Added Wake Time = DSB execution time
> > > - * If Variable Refresh Rate where Vmin != Vmax != Flipline:
> > > - * Program DEEP PKG_C_LATENCY Pkg C with all 1's.
> > > - * Program PKG_C_LATENCY Added Wake Time = 0
> > > - */
> > > -static void
> > > -skl_program_dpkgc_latency(struct drm_i915_private *i915, bool
> > > enable_dpkgc) -{
> > > -       u32 max_latency = 0;
> > > -       u32 clear = 0, val = 0;
> > > -       u32 added_wake_time = 0;
> > > -
> > > -       if (DISPLAY_VER(i915) < 20)
> > > -               return;
> > > -
> > > -       if (enable_dpkgc) {
> > > -               max_latency = skl_watermark_max_latency(i915, 1);
> > > -               if (max_latency == 0)
> > > -                       max_latency = LNL_PKG_C_LATENCY_MASK;
> > > -               added_wake_time = DSB_EXE_TIME +
> > > -                       i915->display.sagv.block_time_us;
> > > -       } else {
> > > -               max_latency = LNL_PKG_C_LATENCY_MASK;
> > > -               added_wake_time = 0;
> > > -       }
> > > -
> > > -       clear |= LNL_ADDED_WAKE_TIME_MASK |
> > LNL_PKG_C_LATENCY_MASK;
> > > -       val |= REG_FIELD_PREP(LNL_PKG_C_LATENCY_MASK, max_latency);
> > > -       val |= REG_FIELD_PREP(LNL_ADDED_WAKE_TIME_MASK,
> > added_wake_time);
> > > -
> > > -       intel_uncore_rmw(&i915->uncore, LNL_PKG_C_LATENCY, clear, val);
> > > -}
> > > -
> > >  static int
> > >  skl_compute_wm(struct intel_atomic_state *state)  {
> > >         struct intel_crtc *crtc;
> > >         struct intel_crtc_state __maybe_unused *new_crtc_state;
> > >         int ret, i;
> > > -       bool enable_dpkgc = false;
> > > 
> > >         for_each_new_intel_crtc_in_state(state, crtc, new_crtc_state, i) {
> > >                 ret = skl_build_pipe_wm(state, crtc); @@ -2913,15 +2868,8
> > @@
> > > skl_compute_wm(struct intel_atomic_state *state)
> > >                 ret = skl_wm_add_affected_planes(state, crtc);
> > >                 if (ret)
> > >                         return ret;
> > > -
> > > -               if ((new_crtc_state->vrr.vmin == new_crtc_state->vrr.vmax &&
> > > -                    new_crtc_state->vrr.vmin == new_crtc_state->vrr.flipline)
> > > > 
> > > -                   !new_crtc_state->vrr.enable)
> > > -                       enable_dpkgc = true;
> > >         }
> > > 
> > > -       skl_program_dpkgc_latency(to_i915(state->base.dev), enable_dpkgc);
> > > -
> > >         skl_print_wm_changes(state);
> > > 
> > >         return 0;
> > 
> > --
> > Jani Nikula, Intel



More information about the Intel-gfx mailing list