[PATCH v5 00/13] Centralize format information

[Ville Syrjälä]

On Tue, Oct 18, 2016 at 03:33:29PM +0300, Laurent Pinchart wrote:
> Hi Ville,
> 
> On Tuesday 18 Oct 2016 14:43:18 Ville Syrjälä wrote:
> > On Tue, Oct 18, 2016 at 01:41:08AM +0300, Laurent Pinchart wrote:
> > > Hello,
> > > 
> > > Various pieces of information about DRM formats (number of planes, color
> > > depth, chroma subsampling, ...) are scattered across different helper
> > > functions in the DRM core. Callers of those functions often need to access
> > > more than a single parameter of the format, leading to inefficiencies due
> > > to multiple lookups.
> > > 
> > > This patch series addresses this issue by centralizing all format
> > > information in a single data structure (01/13). It reimplements the
> > > existing format helper functions based on that structure (02/13) and
> > > converts the DRM core code to use the new structure (03/13). The DRM core
> > > now WARNs when a driver tries to query information about an unsupported
> > > format (04/13).
> > > 
> > > The second part of the patch series removes the drm_fb_get_bpp_depth()
> > > legacy function that shouldn't be used directly by drivers. It modifies
> > > all its users to use the appropriate API instead (05/13 to 12/13) and
> > > finally merges the function into its only caller in the DRM core (13/13).
> > > 
> > > The new API is also useful for drivers as shown by the "[PATCH v2 00/20]
> > > OMAP DRM fixes and improvements" patch series previously posted.
> > > 
> > > All patches have been acked, the series is ready to be merged for v4.10.
> > 
> > BTW and idea I had recently is that we could store the pointer to the
> > format info into drm_framebuffer, and then a lot of the
> > drm_format_...() calls could just vanish, or if the information we
> > need isn't directly stored in the info structure we'd just need a small
> > function that takes the entire drm_framebuffer or just the format info
> > struct and computes what we need. Would avoid having to sprinkle
> > drm_format_info() calls into the drivers.
> 
> That's a good idea. I didn't want to add that to my patch series in order to 
> avoid merge delays, but patches are certainly welcome :-)

Sure. I'll preserve the idea for a rainy day, unless someone else beats
me to it that is.

-- 
Ville Syrjälä
Intel OTC