| Message ID | 20170718210510.12229-1-eric@anholt.net (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, 18 Jul 2017 14:05:05 -0700 Eric Anholt <eric@anholt.net> wrote: > The incoming mode might have a missing vrefresh field if it came from > drmModeSetCrtc(), which the kernel is supposed to calculate using > drm_mode_vrefresh(). We could either use that or the adjusted_mode's > original vrefresh value. > > However, we can maintain a more exact vrefresh value (not just the > integer approximation), by scaling by the ratio of our clocks. > > v2: Use math suggested by Andrzej Hajda instead. > > Signed-off-by: Eric Anholt <eric@anholt.net> > --- > drivers/gpu/drm/vc4/vc4_dsi.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/vc4/vc4_dsi.c b/drivers/gpu/drm/vc4/vc4_dsi.c > index 629d372633e6..57213f4e3c72 100644 > --- a/drivers/gpu/drm/vc4/vc4_dsi.c > +++ b/drivers/gpu/drm/vc4/vc4_dsi.c > @@ -866,7 +866,8 @@ static bool vc4_dsi_encoder_mode_fixup(struct drm_encoder *encoder, > adjusted_mode->clock = pixel_clock_hz / 1000 + 1; > > /* Given the new pixel clock, adjust HFP to keep vrefresh the same. */ > - adjusted_mode->htotal = pixel_clock_hz / (mode->vrefresh * mode->vtotal); > + adjusted_mode->htotal = (pixel_clock_hz / 1000 * mode->htotal / > + mode->clock); Hm, I'm not sure I understand this. Shouldn't we have something like: adjusted_mode->htotal = (adjusted_mode->clock * mode->htotal) / mode->clock; Is there a reason for doing '+ 1' when you calculate the adjusted pixel clock rate but not here? > adjusted_mode->hsync_end += adjusted_mode->htotal - mode->htotal; > adjusted_mode->hsync_start += adjusted_mode->htotal - mode->htotal; >
Boris Brezillon <boris.brezillon@free-electrons.com> writes: > On Tue, 18 Jul 2017 14:05:05 -0700 > Eric Anholt <eric@anholt.net> wrote: > >> The incoming mode might have a missing vrefresh field if it came from >> drmModeSetCrtc(), which the kernel is supposed to calculate using >> drm_mode_vrefresh(). We could either use that or the adjusted_mode's >> original vrefresh value. >> >> However, we can maintain a more exact vrefresh value (not just the >> integer approximation), by scaling by the ratio of our clocks. >> >> v2: Use math suggested by Andrzej Hajda instead. >> >> Signed-off-by: Eric Anholt <eric@anholt.net> >> --- >> drivers/gpu/drm/vc4/vc4_dsi.c | 3 ++- >> 1 file changed, 2 insertions(+), 1 deletion(-) >> >> diff --git a/drivers/gpu/drm/vc4/vc4_dsi.c b/drivers/gpu/drm/vc4/vc4_dsi.c >> index 629d372633e6..57213f4e3c72 100644 >> --- a/drivers/gpu/drm/vc4/vc4_dsi.c >> +++ b/drivers/gpu/drm/vc4/vc4_dsi.c >> @@ -866,7 +866,8 @@ static bool vc4_dsi_encoder_mode_fixup(struct drm_encoder *encoder, >> adjusted_mode->clock = pixel_clock_hz / 1000 + 1; >> >> /* Given the new pixel clock, adjust HFP to keep vrefresh the same. */ >> - adjusted_mode->htotal = pixel_clock_hz / (mode->vrefresh * mode->vtotal); >> + adjusted_mode->htotal = (pixel_clock_hz / 1000 * mode->htotal / >> + mode->clock); > > Hm, I'm not sure I understand this. Shouldn't we have something like: > > adjusted_mode->htotal = (adjusted_mode->clock * mode->htotal) / > mode->clock; > > Is there a reason for doing '+ 1' when you calculate the adjusted > pixel clock rate but not here? We're actually expecting to get within epsilon of pixel_clock_hz, but we have to bump our clk_set_rate() to a higher value because the clock driver will give you a bad divider if you ask for anything less than the rate it can provide. How about I don't increment the adjusted_mode->clock (since it'll be userspace visible I think), and instead move that and the "Round up" comment to the clk_set_rate()?
Le Fri, 04 Aug 2017 14:15:56 -0700, Eric Anholt <eric@anholt.net> a écrit : > Boris Brezillon <boris.brezillon@free-electrons.com> writes: > > > On Tue, 18 Jul 2017 14:05:05 -0700 > > Eric Anholt <eric@anholt.net> wrote: > > > >> The incoming mode might have a missing vrefresh field if it came from > >> drmModeSetCrtc(), which the kernel is supposed to calculate using > >> drm_mode_vrefresh(). We could either use that or the adjusted_mode's > >> original vrefresh value. > >> > >> However, we can maintain a more exact vrefresh value (not just the > >> integer approximation), by scaling by the ratio of our clocks. > >> > >> v2: Use math suggested by Andrzej Hajda instead. > >> > >> Signed-off-by: Eric Anholt <eric@anholt.net> > >> --- > >> drivers/gpu/drm/vc4/vc4_dsi.c | 3 ++- > >> 1 file changed, 2 insertions(+), 1 deletion(-) > >> > >> diff --git a/drivers/gpu/drm/vc4/vc4_dsi.c b/drivers/gpu/drm/vc4/vc4_dsi.c > >> index 629d372633e6..57213f4e3c72 100644 > >> --- a/drivers/gpu/drm/vc4/vc4_dsi.c > >> +++ b/drivers/gpu/drm/vc4/vc4_dsi.c > >> @@ -866,7 +866,8 @@ static bool vc4_dsi_encoder_mode_fixup(struct drm_encoder *encoder, > >> adjusted_mode->clock = pixel_clock_hz / 1000 + 1; > >> > >> /* Given the new pixel clock, adjust HFP to keep vrefresh the same. */ > >> - adjusted_mode->htotal = pixel_clock_hz / (mode->vrefresh * mode->vtotal); > >> + adjusted_mode->htotal = (pixel_clock_hz / 1000 * mode->htotal / > >> + mode->clock); > > > > Hm, I'm not sure I understand this. Shouldn't we have something like: > > > > adjusted_mode->htotal = (adjusted_mode->clock * mode->htotal) / > > mode->clock; > > > > Is there a reason for doing '+ 1' when you calculate the adjusted > > pixel clock rate but not here? > > We're actually expecting to get within epsilon of pixel_clock_hz, but we > have to bump our clk_set_rate() to a higher value because the clock > driver will give you a bad divider if you ask for anything less than the > rate it can provide. > > How about I don't increment the adjusted_mode->clock (since it'll be > userspace visible I think), and instead move that and the "Round up" > comment to the clk_set_rate()? Sounds good.
diff --git a/drivers/gpu/drm/vc4/vc4_dsi.c b/drivers/gpu/drm/vc4/vc4_dsi.c index 629d372633e6..57213f4e3c72 100644 --- a/drivers/gpu/drm/vc4/vc4_dsi.c +++ b/drivers/gpu/drm/vc4/vc4_dsi.c @@ -866,7 +866,8 @@ static bool vc4_dsi_encoder_mode_fixup(struct drm_encoder *encoder, adjusted_mode->clock = pixel_clock_hz / 1000 + 1; /* Given the new pixel clock, adjust HFP to keep vrefresh the same. */ - adjusted_mode->htotal = pixel_clock_hz / (mode->vrefresh * mode->vtotal); + adjusted_mode->htotal = (pixel_clock_hz / 1000 * mode->htotal / + mode->clock); adjusted_mode->hsync_end += adjusted_mode->htotal - mode->htotal; adjusted_mode->hsync_start += adjusted_mode->htotal - mode->htotal;
The incoming mode might have a missing vrefresh field if it came from drmModeSetCrtc(), which the kernel is supposed to calculate using drm_mode_vrefresh(). We could either use that or the adjusted_mode's original vrefresh value. However, we can maintain a more exact vrefresh value (not just the integer approximation), by scaling by the ratio of our clocks. v2: Use math suggested by Andrzej Hajda instead. Signed-off-by: Eric Anholt <eric@anholt.net> --- drivers/gpu/drm/vc4/vc4_dsi.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)