diff mbox

[v2] drm/i915: Audio N value computed for pixel doubled modes

Message ID 1411662396-10935-1-git-send-email-clinton.a.taylor@intel.com (mailing list archive)
State New, archived
Headers show

Commit Message

Taylor, Clinton A Sept. 25, 2014, 4:26 p.m. UTC
From: Clint Taylor <clinton.a.taylor@intel.com>

HDMI audio clock config was incorrectly choosing the default for
pixel doubled interlaced modes. The table was missing pixel clock
values 13.500 (27.000) and 13.513 (27.027). Luckily the default N
value for 25.200 is the same N value for both 27MHz pixel clocks,
a warning message was being printed with drm.debug set.

ver2: Use 13500 * 1001 / 1000 instead of 13513 constant.

Cc: Jani Nikula <jani.nikula at intel.com>

Signed-off-by: Clint Taylor <clinton.a.taylor@intel.com>
---
 drivers/gpu/drm/i915/intel_display.c |    2 ++
 1 file changed, 2 insertions(+)

Comments

Ville Syrjälä Sept. 26, 2014, 4:28 p.m. UTC | #1
On Thu, Sep 25, 2014 at 09:26:36AM -0700, clinton.a.taylor@intel.com wrote:
> From: Clint Taylor <clinton.a.taylor@intel.com>
> 
> HDMI audio clock config was incorrectly choosing the default for
> pixel doubled interlaced modes. The table was missing pixel clock
> values 13.500 (27.000) and 13.513 (27.027). Luckily the default N
> value for 25.200 is the same N value for both 27MHz pixel clocks,
> a warning message was being printed with drm.debug set.
> 
> ver2: Use 13500 * 1001 / 1000 instead of 13513 constant.
> 
> Cc: Jani Nikula <jani.nikula at intel.com>
> 
> Signed-off-by: Clint Taylor <clinton.a.taylor@intel.com>
> ---
>  drivers/gpu/drm/i915/intel_display.c |    2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> index 858011d..e76a4106 100644
> --- a/drivers/gpu/drm/i915/intel_display.c
> +++ b/drivers/gpu/drm/i915/intel_display.c
> @@ -7872,6 +7872,8 @@ static struct {
>  	{ 74250, AUD_CONFIG_PIXEL_CLOCK_HDMI_74250 },
>  	{ DIV_ROUND_UP(148500 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_148352 },
>  	{ 148500, AUD_CONFIG_PIXEL_CLOCK_HDMI_148500 },
> +	{ 13500, AUD_CONFIG_PIXEL_CLOCK_HDMI_27000 },
> +	{ 13500 * 1001 / 1000, AUD_CONFIG_PIXEL_CLOCK_HDMI_27027 },

We have double clocked modes where the non-doubled clock is already
27MHz so this seems like a bandaid for one particular case rather than a
full solution.

The HDMI specification makes it clear that the N/CTS stuff depends on
the TMDS clock and not the pixel clock, but BSpec just talks about pixel
clock without further explaining any of this stuff. So should we look
at port_clock rather than the pixel clock here?

>  };
>  
>  /* get AUD_CONFIG_PIXEL_CLOCK_HDMI_* value for mode */
> -- 
> 1.7.9.5
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Taylor, Clinton A Oct. 6, 2014, 10:01 p.m. UTC | #2
On 09/26/2014 09:28 AM, Ville Syrjälä wrote:
> On Thu, Sep 25, 2014 at 09:26:36AM -0700, clinton.a.taylor@intel.com wrote:
>> From: Clint Taylor <clinton.a.taylor@intel.com>
>>
>> HDMI audio clock config was incorrectly choosing the default for
>> pixel doubled interlaced modes. The table was missing pixel clock
>> values 13.500 (27.000) and 13.513 (27.027). Luckily the default N
>> value for 25.200 is the same N value for both 27MHz pixel clocks,
>> a warning message was being printed with drm.debug set.
>>
>> ver2: Use 13500 * 1001 / 1000 instead of 13513 constant.
>>
>> Cc: Jani Nikula <jani.nikula at intel.com>
>>
>> Signed-off-by: Clint Taylor <clinton.a.taylor@intel.com>
>> ---
>>   drivers/gpu/drm/i915/intel_display.c |    2 ++
>>   1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
>> index 858011d..e76a4106 100644
>> --- a/drivers/gpu/drm/i915/intel_display.c
>> +++ b/drivers/gpu/drm/i915/intel_display.c
>> @@ -7872,6 +7872,8 @@ static struct {
>>   	{ 74250, AUD_CONFIG_PIXEL_CLOCK_HDMI_74250 },
>>   	{ DIV_ROUND_UP(148500 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_148352 },
>>   	{ 148500, AUD_CONFIG_PIXEL_CLOCK_HDMI_148500 },
>> +	{ 13500, AUD_CONFIG_PIXEL_CLOCK_HDMI_27000 },
>> +	{ 13500 * 1001 / 1000, AUD_CONFIG_PIXEL_CLOCK_HDMI_27027 },
>
> We have double clocked modes where the non-doubled clock is already
> 27MHz so this seems like a bandaid for one particular case rather than a
> full solution.
>
> The HDMI specification makes it clear that the N/CTS stuff depends on
> the TMDS clock and not the pixel clock, but BSpec just talks about pixel
> clock without further explaining any of this stuff. So should we look
> at port_clock rather than the pixel clock here?

No, Since we support 12bpc we still need to only worry about the pixel 
clock. TMDS clock and port clock will be multiplied by 1.5 in 12bpc 
mode, but the N value remains the same as the 8bpc clock.

-Clint

>
>>   };
>>
>>   /* get AUD_CONFIG_PIXEL_CLOCK_HDMI_* value for mode */
>> --
>> 1.7.9.5
>>
>> _______________________________________________
>> Intel-gfx mailing list
>> Intel-gfx@lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
>
Ville Syrjälä Oct. 7, 2014, 8:52 a.m. UTC | #3
On Mon, Oct 06, 2014 at 03:01:46PM -0700, Clint Taylor wrote:
> On 09/26/2014 09:28 AM, Ville Syrjälä wrote:
> > On Thu, Sep 25, 2014 at 09:26:36AM -0700, clinton.a.taylor@intel.com wrote:
> >> From: Clint Taylor <clinton.a.taylor@intel.com>
> >>
> >> HDMI audio clock config was incorrectly choosing the default for
> >> pixel doubled interlaced modes. The table was missing pixel clock
> >> values 13.500 (27.000) and 13.513 (27.027). Luckily the default N
> >> value for 25.200 is the same N value for both 27MHz pixel clocks,
> >> a warning message was being printed with drm.debug set.
> >>
> >> ver2: Use 13500 * 1001 / 1000 instead of 13513 constant.
> >>
> >> Cc: Jani Nikula <jani.nikula at intel.com>
> >>
> >> Signed-off-by: Clint Taylor <clinton.a.taylor@intel.com>
> >> ---
> >>   drivers/gpu/drm/i915/intel_display.c |    2 ++
> >>   1 file changed, 2 insertions(+)
> >>
> >> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> >> index 858011d..e76a4106 100644
> >> --- a/drivers/gpu/drm/i915/intel_display.c
> >> +++ b/drivers/gpu/drm/i915/intel_display.c
> >> @@ -7872,6 +7872,8 @@ static struct {
> >>   	{ 74250, AUD_CONFIG_PIXEL_CLOCK_HDMI_74250 },
> >>   	{ DIV_ROUND_UP(148500 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_148352 },
> >>   	{ 148500, AUD_CONFIG_PIXEL_CLOCK_HDMI_148500 },
> >> +	{ 13500, AUD_CONFIG_PIXEL_CLOCK_HDMI_27000 },
> >> +	{ 13500 * 1001 / 1000, AUD_CONFIG_PIXEL_CLOCK_HDMI_27027 },
> >
> > We have double clocked modes where the non-doubled clock is already
> > 27MHz so this seems like a bandaid for one particular case rather than a
> > full solution.
> >
> > The HDMI specification makes it clear that the N/CTS stuff depends on
> > the TMDS clock and not the pixel clock, but BSpec just talks about pixel
> > clock without further explaining any of this stuff. So should we look
> > at port_clock rather than the pixel clock here?
> 
> No, Since we support 12bpc we still need to only worry about the pixel 
> clock. TMDS clock and port clock will be multiplied by 1.5 in 12bpc 
> mode, but the N value remains the same as the 8bpc clock.

Does the hardware then have two different N value tables and it picks
one based on 8bpc vs. 12bpc? Also I'm not sure what happens with
different audio sample rates, but maybe we only support 48kHz? Sadly
the spec is quite lacking when it comes to the audio side.

The recommended N values in the HDMI spec do differ somewhat between
8bpc and 12bpc, so either we end up using a non-recommended N value
in 12bpc, or the hardware does some extra magic.
Taylor, Clinton A Oct. 13, 2014, 11:46 p.m. UTC | #4
On 10/07/2014 01:52 AM, Ville Syrjälä wrote:
> On Mon, Oct 06, 2014 at 03:01:46PM -0700, Clint Taylor wrote:
>> On 09/26/2014 09:28 AM, Ville Syrjälä wrote:
>>> On Thu, Sep 25, 2014 at 09:26:36AM -0700, clinton.a.taylor@intel.com wrote:
>>>> From: Clint Taylor <clinton.a.taylor@intel.com>
>>>>
>>>> HDMI audio clock config was incorrectly choosing the default for
>>>> pixel doubled interlaced modes. The table was missing pixel clock
>>>> values 13.500 (27.000) and 13.513 (27.027). Luckily the default N
>>>> value for 25.200 is the same N value for both 27MHz pixel clocks,
>>>> a warning message was being printed with drm.debug set.
>>>>
>>>> ver2: Use 13500 * 1001 / 1000 instead of 13513 constant.
>>>>
>>>> Cc: Jani Nikula <jani.nikula at intel.com>
>>>>
>>>> Signed-off-by: Clint Taylor <clinton.a.taylor@intel.com>
>>>> ---
>>>>    drivers/gpu/drm/i915/intel_display.c |    2 ++
>>>>    1 file changed, 2 insertions(+)
>>>>
>>>> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
>>>> index 858011d..e76a4106 100644
>>>> --- a/drivers/gpu/drm/i915/intel_display.c
>>>> +++ b/drivers/gpu/drm/i915/intel_display.c
>>>> @@ -7872,6 +7872,8 @@ static struct {
>>>>    	{ 74250, AUD_CONFIG_PIXEL_CLOCK_HDMI_74250 },
>>>>    	{ DIV_ROUND_UP(148500 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_148352 },
>>>>    	{ 148500, AUD_CONFIG_PIXEL_CLOCK_HDMI_148500 },
>>>> +	{ 13500, AUD_CONFIG_PIXEL_CLOCK_HDMI_27000 },
>>>> +	{ 13500 * 1001 / 1000, AUD_CONFIG_PIXEL_CLOCK_HDMI_27027 },
>>>
>>> We have double clocked modes where the non-doubled clock is already
>>> 27MHz so this seems like a bandaid for one particular case rather than a
>>> full solution.
>>>
>>> The HDMI specification makes it clear that the N/CTS stuff depends on
>>> the TMDS clock and not the pixel clock, but BSpec just talks about pixel
>>> clock without further explaining any of this stuff. So should we look
>>> at port_clock rather than the pixel clock here?
>>
>> No, Since we support 12bpc we still need to only worry about the pixel
>> clock. TMDS clock and port clock will be multiplied by 1.5 in 12bpc
>> mode, but the N value remains the same as the 8bpc clock.
>
> Does the hardware then have two different N value tables and it picks
> one based on 8bpc vs. 12bpc? Also I'm not sure what happens with
> different audio sample rates, but maybe we only support 48kHz? Sadly
> the spec is quite lacking when it comes to the audio side.
>
The BSPEC also contains manual n/cts registers for non CEA modes. 
Hopefully the HW automatically selects 12bpc N values. The HW does 
select 8bpc other sample rates, so 12bpc 32 and 44.1 "should" work.

> The recommended N values in the HDMI spec do differ somewhat between
> 8bpc and 12bpc, so either we end up using a non-recommended N value
> in 12bpc, or the hardware does some extra magic.
>
Just saw Appendix D in the HDMI spec..
diff mbox

Patch

diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 858011d..e76a4106 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -7872,6 +7872,8 @@  static struct {
 	{ 74250, AUD_CONFIG_PIXEL_CLOCK_HDMI_74250 },
 	{ DIV_ROUND_UP(148500 * 1000, 1001), AUD_CONFIG_PIXEL_CLOCK_HDMI_148352 },
 	{ 148500, AUD_CONFIG_PIXEL_CLOCK_HDMI_148500 },
+	{ 13500, AUD_CONFIG_PIXEL_CLOCK_HDMI_27000 },
+	{ 13500 * 1001 / 1000, AUD_CONFIG_PIXEL_CLOCK_HDMI_27027 },
 };
 
 /* get AUD_CONFIG_PIXEL_CLOCK_HDMI_* value for mode */