Intel-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Govindapillai, Vinod" <vinod.govindapillai@intel.com>
To: "Lisovskiy, Stanislav" <stanislav.lisovskiy@intel.com>,
	"intel-gfx@lists.freedesktop.org"
	<intel-gfx@lists.freedesktop.org>
Cc: "Saarinen, Jani" <jani.saarinen@intel.com>,
	"ville.syrjala@linux.intel.com" <ville.syrjala@linux.intel.com>
Subject: Re: [PATCH 2/3] drm/i915: Extract code required to calculate max qgv/psf gv point
Date: Wed, 20 Mar 2024 22:07:59 +0000	[thread overview]
Message-ID: <d9fe3b795d06ed81c482a659d46504fec02aaaaf.camel@intel.com> (raw)
In-Reply-To: <20240220093136.32181-3-stanislav.lisovskiy@intel.com>

Hi Stan.


On Tue, 2024-02-20 at 11:31 +0200, Stanislav Lisovskiy wrote:
> We need that in order to force disable SAGV in next patch.
> Also it is beneficial to separate that code, as in majority cases,
> when SAGV is enabled, we don't even need those calculations.
> Also we probably need to determine max PSF GV point as well, however
> currently we don't do that when we disable SAGV, which might be
> actually causing some issues in that case.
> 
> v2: - Introduce helper adl_qgv_bw(counterpart to adl_psf_bw)
>       (Ville Syrjälä)
>     - Don't restrict psf gv points for SAGV disable case
>       (Ville Syrjälä)
> 
> Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
> ---
>  drivers/gpu/drm/i915/display/intel_bw.c | 81 ++++++++++++++++---------
>  1 file changed, 53 insertions(+), 28 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/display/intel_bw.c b/drivers/gpu/drm/i915/display/intel_bw.c
> index 77886cc21211..7baa1c13eccd 100644
> --- a/drivers/gpu/drm/i915/display/intel_bw.c
> +++ b/drivers/gpu/drm/i915/display/intel_bw.c
> @@ -652,15 +652,31 @@ static unsigned int tgl_max_bw_index(struct drm_i915_private *dev_priv,
>         return 0;
>  }
>  
> -static unsigned int adl_psf_bw(struct drm_i915_private *dev_priv,
> +static unsigned int adl_psf_bw(struct drm_i915_private *i915,
>                                int psf_gv_point)
>  {
>         const struct intel_bw_info *bi =
> -                       &dev_priv->display.bw.max[0];
> +                       &i915->display.bw.max[0];
>  
>         return bi->psf_bw[psf_gv_point];
>  }
This is probably not related to this patch..

>  
> +static unsigned int adl_qgv_bw(struct drm_i915_private *i915,
> +                              int qgv_point, int num_active_planes)
In the next patch you are chaging the order of these parameters and calling with adl_qgv_bw(i915,
num_active_planes, i). As you are adding this functions in this patch, I think you should fix the
position of parameters in this patch itself and next patch call this normally

Also about the naming of this function, should this be icl_* as this is called from icl_* functions?
> +{
> +       unsigned int idx;
> +
> +       if (DISPLAY_VER(i915) > 11)
This is just fine.. but just for the sake of easy readability, wonder DISPLAY_VER(i915) >= 12 is
better as TGL is display version 12.

BR
vinod

> +               idx = tgl_max_bw_index(i915, num_active_planes, qgv_point);
> +       else
> +               idx = icl_max_bw_index(i915, num_active_planes, qgv_point);
> +
> +       if (idx >= ARRAY_SIZE(i915->display.bw.max))
> +               return 0;
> +
> +       return i915->display.bw.max[idx].deratedbw[qgv_point];
> +}
> +
>  void intel_bw_init_hw(struct drm_i915_private *dev_priv)
>  {
>         if (!HAS_DISPLAY(dev_priv))
> @@ -806,6 +822,36 @@ intel_atomic_get_bw_state(struct intel_atomic_state *state)
>         return to_intel_bw_state(bw_state);
>  }
>  
> +static unsigned int icl_max_bw_qgv_point(struct drm_i915_private *i915,
> +                                        int num_active_planes)
> +{
> +       unsigned int max_bw_point = 0;
> +       unsigned int max_bw = 0;
> +       unsigned int num_qgv_points = i915->display.bw.max[0].num_qgv_points;
> +       int i;
> +
> +       for (i = 0; i < num_qgv_points; i++) {
> +               unsigned int max_data_rate;
> +
> +               max_data_rate = adl_qgv_bw(i915, i, num_active_planes);
> +
> +               /*
> +                * We need to know which qgv point gives us
> +                * maximum bandwidth in order to disable SAGV
> +                * if we find that we exceed SAGV block time
> +                * with watermarks. By that moment we already
> +                * have those, as it is calculated earlier in
> +                * intel_atomic_check,
> +                */
> +               if (max_data_rate > max_bw) {
> +                       max_bw_point = i;
> +                       max_bw = max_data_rate;
> +               }
> +       }
> +
> +       return max_bw_point;
> +}
> +
>  static int mtl_find_qgv_points(struct drm_i915_private *i915,
>                                unsigned int data_rate,
>                                unsigned int num_active_planes,
> @@ -883,8 +929,6 @@ static int icl_find_qgv_points(struct drm_i915_private *i915,
>                                const struct intel_bw_state *old_bw_state,
>                                struct intel_bw_state *new_bw_state)
>  {
> -       unsigned int max_bw_point = 0;
> -       unsigned int max_bw = 0;
>         unsigned int num_psf_gv_points = i915->display.bw.max[0].num_psf_gv_points;
>         unsigned int num_qgv_points = i915->display.bw.max[0].num_qgv_points;
>         u16 psf_points = 0;
> @@ -897,31 +941,10 @@ static int icl_find_qgv_points(struct drm_i915_private *i915,
>                 return ret;
>  
>         for (i = 0; i < num_qgv_points; i++) {
> -               unsigned int idx;
>                 unsigned int max_data_rate;
>  
> -               if (DISPLAY_VER(i915) >= 12)
> -                       idx = tgl_max_bw_index(i915, num_active_planes, i);
> -               else
> -                       idx = icl_max_bw_index(i915, num_active_planes, i);
> -
> -               if (idx >= ARRAY_SIZE(i915->display.bw.max))
> -                       continue;
> -
> -               max_data_rate = i915->display.bw.max[idx].deratedbw[i];
> +               max_data_rate = adl_qgv_bw(i915, i, num_active_planes);
>  
> -               /*
> -                * We need to know which qgv point gives us
> -                * maximum bandwidth in order to disable SAGV
> -                * if we find that we exceed SAGV block time
> -                * with watermarks. By that moment we already
> -                * have those, as it is calculated earlier in
> -                * intel_atomic_check,
> -                */
> -               if (max_data_rate > max_bw) {
> -                       max_bw_point = i;
> -                       max_bw = max_data_rate;
> -               }
>                 if (max_data_rate >= data_rate)
>                         qgv_points |= BIT(i);
>  
> @@ -965,9 +988,11 @@ static int icl_find_qgv_points(struct drm_i915_private *i915,
>          * cause.
>          */
>         if (!intel_can_enable_sagv(i915, new_bw_state)) {
> -               qgv_points = BIT(max_bw_point);
> +               unsigned int max_bw_qgv_point = icl_max_bw_qgv_point(i915, num_active_planes);
> +
> +               qgv_points = BIT(max_bw_qgv_point);
>                 drm_dbg_kms(&i915->drm, "No SAGV, using single QGV point %d\n",
> -                           max_bw_point);
> +                           max_bw_qgv_point);
>         }
>  
>         /*


  reply	other threads:[~2024-03-20 22:08 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-02-20  9:31 [PATCH 0/3] QGV/SAGV related fixes Stanislav Lisovskiy
2024-02-20  9:31 ` [PATCH 1/3] drm/i915: Add meaningful traces for QGV point info error handling Stanislav Lisovskiy
2024-03-20 17:09   ` Govindapillai, Vinod
2024-02-20  9:31 ` [PATCH 2/3] drm/i915: Extract code required to calculate max qgv/psf gv point Stanislav Lisovskiy
2024-03-20 22:07   ` Govindapillai, Vinod [this message]
2024-02-20  9:31 ` [PATCH 3/3] drm/i915: Disable SAGV on bw init, to force QGV point recalculation Stanislav Lisovskiy
2024-03-20 22:33   ` Govindapillai, Vinod
2024-02-20 11:55 ` ✗ Fi.CI.CHECKPATCH: warning for QGV/SAGV related fixes (rev7) Patchwork
2024-02-20 11:55 ` ✗ Fi.CI.SPARSE: " Patchwork
2024-02-20 12:06 ` ✓ Fi.CI.BAT: success " Patchwork
2024-02-20 13:45 ` ✓ Fi.CI.IGT: " Patchwork
  -- strict thread matches above, loose matches on Subject: below --
2024-02-19  9:18 [PATCH 0/3] QGV/SAGV related fixes Stanislav Lisovskiy
2024-02-19  9:18 ` [PATCH 2/3] drm/i915: Extract code required to calculate max qgv/psf gv point Stanislav Lisovskiy
2024-01-17 15:57 [PATCH 0/3] QGV/SAGV related fixes Stanislav Lisovskiy
2024-01-17 15:57 ` [PATCH 2/3] drm/i915: Extract code required to calculate max qgv/psf gv point Stanislav Lisovskiy
2024-01-18  8:24   ` Ville Syrjälä
2023-11-28  8:37 [Intel-gfx] [PATCH 0/3] QGV/SAGV related fixes Stanislav Lisovskiy
2023-11-28  8:37 ` [Intel-gfx] [PATCH 2/3] drm/i915: Extract code required to calculate max qgv/psf gv point Stanislav Lisovskiy
2024-01-12 17:35   ` Ville Syrjälä
2024-01-17 10:12     ` Lisovskiy, Stanislav
2024-01-17 11:12       ` Ville Syrjälä
2024-01-17 11:23         ` Lisovskiy, Stanislav

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=d9fe3b795d06ed81c482a659d46504fec02aaaaf.camel@intel.com \
    --to=vinod.govindapillai@intel.com \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=jani.saarinen@intel.com \
    --cc=stanislav.lisovskiy@intel.com \
    --cc=ville.syrjala@linux.intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox