Re: [PATCH] drm/i915/skl: Buffer translation improvements

From: David Weinehall <david.weinehall@linux.intel.com>
To: "Jindal, Sonika" <sonika.jindal@intel.com>
Cc: intel-gfx <intel-gfx@lists.freedesktop.org>
Subject: Re: [PATCH] drm/i915/skl: Buffer translation improvements
Date: Thu, 25 Jun 2015 13:18:06 +0300	[thread overview]
Message-ID: <20150625101806.GP14570@boom> (raw)
In-Reply-To: <558A3666.6020803@intel.com>

On Wed, Jun 24, 2015 at 10:17:34AM +0530, Jindal, Sonika wrote:
> 
> 
> On 6/23/2015 4:42 PM, David Weinehall wrote:
> >On Thu, Jun 18, 2015 at 05:05:21PM +0200, Daniel Vetter wrote:
> >>On Thu, Jun 18, 2015 at 12:50:33PM +0300, David Weinehall wrote:
> >>>@@ -3520,6 +3545,9 @@ intel_dp_set_signal_levels(struct intel_dp *intel_dp, uint32_t *DP)
> >>>  	} else if (HAS_DDI(dev)) {
> >>>  		signal_levels = hsw_signal_levels(train_set);
> >>>  		mask = DDI_BUF_EMP_MASK;
> >>>+
> >>>+		if (IS_SKYLAKE(dev))
> >>>+			skl_set_iboost(intel_dp);
> >>
> >>Imo this should be put into hsw_signal_levels and then hsw_signal_levels
> >>be moved into intel_ddi.c - that way everything related to low-level ddi
> >>DP signal level code in intel_ddi.c.
> >
> >I'm guessing the BXT code should be moved there too
> >and preferably folded in under HAS_DDI(dev)?
> No, it is not required. The Vswing programming for SKL and BXT is completely
> different. So, better keep it separate.

It's not a matter of whether it's required, but whether it makes sense
from a consistency perspective.  From a logical point of view it makes
sense, and keeping them in the same place also makes it possible to
reuse quite a lot of code.  Hence my v2 does so.

Kind regards, David
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx