[Mesa-dev] [PATCH] i965: Lift restriction in external textures for EGLImage support

Chad Versace chadversary at chromium.org
Thu Nov 8 20:38:35 UTC 2018


On Thu 08 Nov 2018, Chad Versace wrote:
> On Wed 31 Oct 2018, Aditya Swarup wrote:
> > For Intel platforms, we support external textures only for EGLImages
> > created with EGL_EXT_image_dma_buf_import. This restriction seems to
> > be Intel specific and not present for other platforms.
> > 
> > While running SKQP test - unitTest_EGLImageTest, GL_INVALID is sent
> > to the test because of this restriction.
> > 
> > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105301
> > Fixes Skqp's unitTest_EGLImageTest test.
> > 
> > Change-Id: I54a162db534d54858319fdb98ba502314d28fc27
> > Signed-off-by: Aditya Swarup <aditya.swarup at intel.com>
> > ---
> >  src/mesa/drivers/dri/i965/intel_image.h     |  3 ---
> >  src/mesa/drivers/dri/i965/intel_screen.c    |  2 --
> >  src/mesa/drivers/dri/i965/intel_tex_image.c | 10 ----------
> >  3 files changed, 15 deletions(-)
> 
> I'm the person who originally insisted that we restrict
> GL_OES_EGL_image_external to only dma_buf-imported EGLImages. The
> pivotal comment below is mine, according git-blame, from 2015-04-09.
> 
> > -   /* We support external textures only for EGLImages created with
> > -    * EGL_EXT_image_dma_buf_import. We may lift that restriction in the future.
> > -    */
> 
> It is now time to lift that restriction :-)
> Reviewed-by: Chad Versace <chadversary at chromium.org>
> 
> The original reason was that the driver was not yet robust enough to
> handle all the corner cases of external images, auxiliary surfaces, etc.
> But, I believe the driver became sufficiently robust enough near
> September last year.

To clarify... I believe the driver is robust regarding auxiliary
surfaces and EGLImages (solution: The driver disables auxiliary surfaces
on EGLImages created via this codepath), but not necessarily fully
correct around all the non-compression corner cases of the EGLImage api.
But, as Ken said, we probably still have problems lurking around the
orphaning logic, and we can fix those corner cases as we find them.


More information about the mesa-dev mailing list