[Bug 105301] The big SKQP bug

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=105301

--- Comment #50 from Dongseong Hwang <dongseong.hwang at intel.com> ---
(In reply to Tapani Pälli from comment #47)
> (In reply to Kenneth Graunke from comment #46)
> > (In reply to Dongseong Hwang from comment #45)
> > > On the other hands, from implementation side, I think Aditya's change makes
> > > sense and this change would be safe because
> > > 1. it passes Piglit and deqp
> > > 2. it restricts available functionality.
> > 
> > How does removing a restriction "restrict available functionality"?
> 
> I guess what he means is that when using TEXTURE_EXTERNAL_OES handle, there
> is less functionality exposed. Of course at that at same time client can
> still use the TEXTURE_2D handle with all functionality. So we are not really
> restricting any functionality here.

Yes, correct. EGL image based on EGL_KHR_gl_image has the full potential of
functionality. The only read functionality is exposed with TEXTURE_EXTERNAL_OES
target.

> I'm not sure if we have fast clear and compression state stored so that one
> could do resolve then switch them off (and never turn on again). If we allow
> this then this is something we should probably do in this use case? From
> client perspective this will result in un-optimal performance I'm not sure
> if there are sensible options here :/

If unfortunately it's un-optimal performance from mesa driver perspective (we
don't know it's true tho), the driver should respect the spec and the spec
explicitly says TEXTURE_EXTERNAL_OES target can be used "ANY EGL image"
described by GL_OES_EGL_image.

https://www.khronos.org/registry/OpenGL/extensions/OES/OES_EGL_image_external.txt
Dependencies on GL_OES_EGL_image

    If GL_OES_EGL_image is supported then change the text in both extensions
    to allow either TEXTURE_2D or TEXTURE_EXTERNAL_OES to be passed as the
    <target> parameter to EGLImageTargetTexture2DOES().  When <target> is
    TEXTURE_2D, behavior of EGLImageTargetTexture2DOES() is as described in
    the GL_OES_EGL_image spec.  When <target> is TEXTURE_EXTERNAL_OES,
    behavior of EGLImageTargetTexture2DOES() is as described in this spec.

In my opinion, Intel has no excuse to Google about this conformance test
failure, and a major product has been broken.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20181106/400691a7/attachment.html>