> -----Original Message-----
> From: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
> Sent: Saturday, November 23, 2024 1:32 AM
> To: Garg, Nemesa <nemesa.garg@xxxxxxxxx>
> Cc: intel-gfx@xxxxxxxxxxxxxxxxxxxxx; intel-xe@xxxxxxxxxxxxxxxxxxxxx
> Subject: Re: [PATCH 3/6] drm/i915/display: Replace adjusted mode with pipe
> mode
>
> On Wed, Nov 20, 2024 at 11:08:35AM +0530, Nemesa Garg wrote:
> > Replace adjusted_mode with pipe_mode in pch_panel_fitting so as to
> > that final pipe src width and height is used after joiner calculation.
> > De-couple the current intel_panel_fitting function, one pre-ilk and
> > one post-ilk, as post-ilk pch_panel_fitting is called from
> > pipe_config.
> >
> > Signed-off-by: Nemesa Garg <nemesa.garg@xxxxxxxxx>
> > ---
> > drivers/gpu/drm/i915/display/intel_display.c | 7 +++
> > drivers/gpu/drm/i915/display/intel_pfit.c | 59 +++++++++++++-------
> > 2 files changed, 47 insertions(+), 19 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/display/intel_display.c
> > b/drivers/gpu/drm/i915/display/intel_display.c
> > index 9db255bb1230..6ad47cf0d419 100644
> > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > @@ -4865,6 +4865,13 @@ copy_joiner_crtc_state_modeset(struct
> intel_atomic_state *state,
> > drm_dp_tunnel_ref_get(primary_crtc_state-
> >dp_tunnel_ref.tunnel,
> > &secondary_crtc_state->dp_tunnel_ref);
> >
> > + if (secondary_crtc_state->pch_pfit.enabled) {
> > + struct drm_rect *dst = &secondary_crtc_state->pch_pfit.dst;
> > + int y = dst->y1;
> > +
> > + drm_rect_translate_to(dst, 0, y);
> > + }
> > +
> > copy_joiner_crtc_state_nomodeset(state, secondary_crtc);
> >
> > secondary_crtc_state->uapi.mode_changed =
> > primary_crtc_state->uapi.mode_changed;
> > diff --git a/drivers/gpu/drm/i915/display/intel_pfit.c
> > b/drivers/gpu/drm/i915/display/intel_pfit.c
> > index fb6387b537b7..b2619e1c4021 100644
> > --- a/drivers/gpu/drm/i915/display/intel_pfit.c
> > +++ b/drivers/gpu/drm/i915/display/intel_pfit.c
> > @@ -181,18 +181,19 @@ static int pch_panel_fitting(struct intel_crtc_state
> *crtc_state,
> > const struct drm_connector_state *conn_state) {
> > struct intel_display *display = to_intel_display(crtc_state);
> > - const struct drm_display_mode *adjusted_mode =
> > - &crtc_state->hw.adjusted_mode;
> > + const struct drm_display_mode *pipe_mode =
> > + &crtc_state->hw.pipe_mode;
> > int pipe_src_w = drm_rect_width(&crtc_state->pipe_src);
> > int pipe_src_h = drm_rect_height(&crtc_state->pipe_src);
> > - int ret, x, y, width, height;
> > + int num_pipes = intel_crtc_num_joined_pipes(crtc_state);
> > + int ret, x, y, width, height, hdisplay_full, src_w_full;
> >
> > if (crtc_state->joiner_pipes)
> > return 0;
> >
> > /* Native modes don't need fitting */
> > - if (adjusted_mode->crtc_hdisplay == pipe_src_w &&
> > - adjusted_mode->crtc_vdisplay == pipe_src_h &&
> > + if (pipe_mode->crtc_hdisplay == pipe_src_w &&
> > + pipe_mode->crtc_vdisplay == pipe_src_h &&
> > crtc_state->output_format != INTEL_OUTPUT_FORMAT_YCBCR420)
> > return 0;
> >
> > @@ -200,46 +201,66 @@ static int pch_panel_fitting(struct intel_crtc_state
> *crtc_state,
> > case DRM_MODE_SCALE_CENTER:
> > width = pipe_src_w;
> > height = pipe_src_h;
> > - x = (adjusted_mode->crtc_hdisplay - width + 1)/2;
> > - y = (adjusted_mode->crtc_vdisplay - height + 1)/2;
> > + x = (pipe_mode->crtc_hdisplay - width + 1) / 2;
> > + y = (pipe_mode->crtc_vdisplay - height + 1) / 2;
> > + /*
> > + * The x-coordinate for Primary should be calculated in such a
> way
> > + * that it remains consistent whether the pipes are joined or
> not.
> > + * This means we need to consider the full width of the
> display even
> > + * when the pipes are joined. The x-coordinate for
> secondaries is 0
> > + * because it starts at the leftmost point of its own display
> area,
> > + * ensuring that the framebuffer is centered within Pipe B’s
> portion
> > + * of the overall display.
> > + */
> > + if (intel_crtc_is_joiner_primary(crtc_state)) {
> > + hdisplay_full = pipe_mode->crtc_hdisplay *
> num_pipes;
> > + src_w_full = width * num_pipes;
> > + x = (hdisplay_full - src_w_full + 1) / 2;
> > + }
>
> Ugh. This stuff is getting quite nasty. I'm thinking it'll be a lot cleaner if we in
> fact don't move the pch_panel_fitting() call from its current location, and
> instead do something like this later:
>
> joiner_adjust_pfit_dst()
> {
> area = pipe_area();
>
> if (!pit.enable)
> return;
>
> rect_intersect(pfit.dst, area);
> rect_translate(pfit.dst, -area.x1, -area.y1); }
>
> where pipe_area() (or whatever we want to call it) gives you per-pipe output
> area, relative to the origing of the full joined output. It should look quite
> similar to a combination of
> intel_joiner_adjust_pipe_src()+intel_joiner_adjust_pipe_src()
> except using crtc_hdisplay/vdisplay instead of pipe_src.
>
> I'm also wondering what would happed if we use mode=CENTER but the pfit
> dst window doesn't extend into all of the joined pipes? Dunno if there's any
> way to have a zero sized pipe_src for some of the pipes or not. If not I
> suppose we'll just have to return an error...
>
Hi Ville,
Thank you for suggestion. I got your point as it make the code a lot cleaner and simple. I will send a new patch with above changes. Also with this new change we wont be requiring other patches as we are not moving pch_panel_fitting. Only the new patch and cleanup patch will be sufficient.
Thanks and Regards,
Nemesa
> --
> Ville Syrjälä
> Intel
