[Mesa-dev] INTEL_map_texture on i915/i965

[Tapani Pälli]

On 08/17/2018 07:37 AM, Jason Ekstrand wrote:
> On Thu, Aug 16, 2018 at 10:39 PM Matt Turner <mattst88 at gmail.com 
> <mailto:mattst88 at gmail.com>> wrote:
> 
>     On Thu, Aug 16, 2018 at 5:50 PM Jason Ekstrand <jason at jlekstrand.net
>     <mailto:jason at jlekstrand.net>> wrote:
>      > If that's true then the extension makes no sense other than maybe
>     "we're a
>      > UMA, we can make things cheap." I think it's more likely that the
>     intention
>      > is to use a fenced map through the GTT to do the detiling. That
>     said, given
>      > how sketchy GTT maps are, it's still a pretty questionable
>     extension at
>      > best and something we don't want to commit to going forward.
> 
>     Why are you speculating about it instead of just reading the extension?
> 
> 
> Replying to e-mails from my phone?  I know, not a good excuse.
> 
>     https://www.khronos.org/registry/OpenGL/extensions/INTEL/INTEL_map_texture.txt
> 
>          This extension allows to create textures with 'linear' layout
>     which allows
>          for simplified access on user side (potentially sacrificing some
>          performance during texture sampling).
> 
> 
> Yeah, that's pretty bad.  I mean, there probably are uses such as if 
> you're doing piles of CPU-side rendering.  If someone really badly wants 
> that trade-off, maybe it's worth implementing but I'm still a skeptic.
> 

One use-case (at least on GLES side with EGL_KHR_lock_surface*) is to 
stream video data to a texture without having to do copy. Maybe using 
texture buffer objects gives this same/similar benefit (?)

// Tapani