[Mesa-dev] [PATCH 2/3] mesa: enable ARB_texture_buffer_* extensions in the Compatibility profile

[Mark Janes]

Dylan Baker <dylan at pnwbakers.com> writes:

> I haven't run the CTS tests, but both the deqp and the piglit test pass on my
> skl with Marek's patches applied.

I must have tested with only the patch 2 applied.  Running with all
three patches in the series, I see no piglit/deqp regressions.

However, GLES CTS fails thousands of ES31-CTS.functional.texture tests,
eg:

ES31-CTS.functional.texture.format.buffer.rg8i_npot
glGetIntegerv() failed: glGetError() returned GL_INVALID_ENUM at gluContextInfo.cpp:229

-Mark

>
> Dylan
>
> Quoting Marek Olšák (2017-10-31 18:16:51)
>> This is how I run piglit on i965:
>> WAFFLE_GBM_DEVICE=/dev/dri/renderD128 PIGLIT_PLATFORM=gbm piglit/bin/glinfo|head
>> GL_RENDERER = Mesa DRI Intel(R) Ivybridge Desktop
>> 
>> Of course I have Radeon as my main GPU.
>> 
>> Marek
>> 
>> On Wed, Nov 1, 2017 at 12:52 AM, Mark Janes <mark.a.janes at intel.com> wrote:
>> > No, please don't.  We can't have thousands of failures added to CI.
>> >
>> > Dylan, can you double check this for Marek?  I'm surprised that it
>> > doesn't reproduce for him.
>> >
>> > -Mark
>> >
>> > Marek Olšák <maraeo at gmail.com> writes:
>> >
>> >> I can't reproduce the issue on my IVB. How about I just push this and
>> >> you guys fix i965 when you have time?
>> >>
>> >> Marek
>> >>
>> >> On Thu, Oct 26, 2017 at 5:46 PM, Mark Janes <mark.a.janes at intel.com> wrote:
>> >>> That fixed many issues, but:
>> >>>
>> >>> On deqp EGL tests, we get:
>> >>>
>> >>> dEQP-EGL.functional.create_context_ext.gl_42.rgba8888_no_depth_no_stencil
>> >>> deqp-egl: main/api_exec.c:135: _mesa_initialize_exec_table: Assertion `ctx->Version > 0' failed.
>> >>>
>> >>> On ES31 CTS tests:
>> >>> ES31-CTS.functional.texture.texture_buffer.render.as_vertex_array_as_vertex_texture_as_fragment_texture.range_size_513
>> >>> glGetIntegerv() failed: glGetError() returned GL_INVALID_ENUM at
>> >>> gluContextInfo.cpp:229
>> >>>
>> >>> on GL46 CTS (27 failures):
>> >>> src/mesa/main/teximage.c:595: _mesa_tex_target_is_layered: Assertion
>> >>> `!"Invalid texture target."' failed.
>> >>>
>> >>> Piglit fails 3.5k tests with the same `ctx->Version > 0' assertion as
>> >>> deqp-egl.
>> >>>
>> >>> -Mark
>> >>>
>> >>> Dylan Baker <dylan at pnwbakers.com> writes:
>> >>>
>> >>>> I've pushed it to our CI, I'll let you know tomorrow what it looks like
>> >>>>
>> >>>> Dylan
>> >>>>
>> >>>> Quoting Marek Olšák (2017-10-25 19:52:43)
>> >>>>> Hi Dylan,
>> >>>>>
>> >>>>> Can you please add this and re-test?
>> >>>>>
>> >>>>> diff --git a/src/mesa/drivers/dri/i965/intel_extensions.c b/src/mesa/drivers/
>> >>>>> dri/i965/intel_extensions.c
>> >>>>> index 21cf632..4d17393 100644
>> >>>>> --- a/src/mesa/drivers/dri/i965/intel_extensions.c
>> >>>>> +++ b/src/mesa/drivers/dri/i965/intel_extensions.c
>> >>>>> @@ -178,7 +178,7 @@ intelInitExtensions(struct gl_context *ctx)
>> >>>>>        ctx->Extensions.ARB_pipeline_statistics_query = true;
>> >>>>>        ctx->Extensions.ARB_sample_shading = true;
>> >>>>>        ctx->Extensions.ARB_shading_language_420pack = true;
>> >>>>> -      if (ctx->API == API_OPENGL_CORE) {
>> >>>>> +      if (ctx->API != API_OPENGL_COMPAT) {
>> >>>>>           ctx->Extensions.ARB_texture_buffer_object = true;
>> >>>>>           ctx->Extensions.ARB_texture_buffer_object_rgb32 = true;
>> >>>>>           ctx->Extensions.ARB_texture_buffer_range = true;
>> >>>>>
>> >>>>>
>> >>>>> Marek
>> >>>>>
>> >>>>> On Wed, Oct 25, 2017 at 11:42 PM, Dylan Baker <dylanx.c.baker at intel.com> wrote:
>> >>>>>
>> >>>>>     There are a significant number of i965 regressions from
>> >>>>>     d96c68146a781c79a23f5181d7050174f1070d90, largely related to texturing (I
>> >>>>>     can
>> >>>>>     send you a complete list of regressions if you care, but due to the large
>> >>>>>     number
>> >>>>>     of them I suspect it's something fairly simple).
>> >>>>>
>> >>>>>     For example:
>> >>>>>     ES31-CTS.functional.texture.format.buffer.r32ui_npot
>> >>>>>
>> >>>>>     glGetIntegerv() failed: glGetError() returned GL_INVALID_ENUM at
>> >>>>>     gluContextInfo.cpp:229
>> >>>>>
>> >>>>>     dEQP-GLES31.functional.state_query.integer.texture_buffer_
>> >>>>>     binding_getinteger:
>> >>>>>
>> >>>>>     glGetIntegerv(GL_TEXTURE_BUFFER_BINDING, 0x00007ffee0c43834);
>> >>>>>     // data = { -555819298 }
>> >>>>>     glGetError();
>> >>>>>     // GL_INVALID_ENUM returned
>> >>>>>      // ERROR: glGetIntegerv: glGetError() returned GL_INVALID_ENUM
>> >>>>>
>> >>>>>     Dylan
>> >>>>>
>> >>>>>     Quoting Emil Velikov (2017-10-24 09:30:03)
>> >>>>>     > Hi Marek,
>> >>>>>     >
>> >>>>>     > On 21 October 2017 at 13:54, Marek Olšák <maraeo at gmail.com> wrote:
>> >>>>>     > > From: Marek Olšák <marek.olsak at amd.com>
>> >>>>>     > >
>> >>>>>     > > We already have piglit tests testing alpha, luminance, and intensity
>> >>>>>     > > formats. They were skipped by piglit until now.
>> >>>>>     > >
>> >>>>>     > > Additionally, I'm enabling one ARB_texture_buffer_range piglit test to
>> >>>>>     run
>> >>>>>     > > with the compat profile.
>> >>>>>     >
>> >>>>>     > Can you please mention that ARB_texture_buffer_* on i965 is unchanged
>> >>>>>     > - aka still enabled only for core profiles.
>> >>>>>     > Out of curiosity - can you tried the series with anything more than
>> >>>>>     piglit?
>> >>>>>     >
>> >>>>>     > The Intel guys can run the lot through CTS, dEQP... admittedly only on
>> >>>>>     > Intel hardware.
>> >>>>>     > Still it should help catch if a piece is missing somewhere.
>> >>>>>     >
>> >>>>>     >
>> >>>>>     > -Emil
>> >>>>>     > _______________________________________________
>> >>>>>     > mesa-dev mailing list
>> >>>>>     > mesa-dev at lists.freedesktop.org
>> >>>>>     > https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>> >>>>>
>> >>>>>