[Mesa-dev] [PATCH] mesa: add support for exposing up to GL4.2

[Matt Turner]

On Thu, Apr 23, 2015 at 8:15 AM, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
> On Thu, Apr 23, 2015 at 11:08 AM, Matt Turner <mattst88 at gmail.com> wrote:
>> On Thu, Apr 23, 2015 at 7:53 AM, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
>>> Add the 4.0/4.1/4.2 extensions lists to compute_version. A coule of
>>> extensions aren't in mesa yet, so those are marked with 0 until they
>>> become supported.
>>>
>>> Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu>
>>> ---
>>>
>>> I wasn't 100% sure about shading_language_packing -- it includes a
>>> couple of functions that don't appear until GL 4.2. However since it's
>>> enabled for all the drivers that matter in mesa, wtvr.
>>
>> What makes you think that?
>>
>> I actually fixed something recently where we were exposing some
>> shading_language_packing functions as GLSL 4.0 instead of 4.2. As far
>> as I'm aware, it's part of 4.2.
>>
>> commit 8d3aa5926b73c67c7dbd4477b7177aaa00c533e5
>> Author: Matt Turner <mattst88 at gmail.com>
>> Date:   Wed Mar 11 18:14:28 2015 -0700
>>
>>     glsl: Expose built-in packing functions under GLSL 4.2.
>>
>>     ARB_shading_language_packing is part of GLSL 4.2, not 4.0 as I
>>     mistakenly believed. The following functions are available only with
>>     ARB_shading_language_packing, GLSL 4.2 (not GLSL 4.0), or ES 3.0:
>>
>>        - packSnorm2x16
>>        - unpackSnorm2x16
>>        - packHalf2x16
>>        - unpackHalf2x16
>>
>>     Reviewed-by: Carl Worth <cworth at cworth.org>
>>     Reviewed-by: Marek Olšák <marek.olsak at amd.com>
>
> Right. Those were the functions from shading_language_packing that
> don't appear until GL 4.2. But other ones it includes are in GL 4.0
> (packUnorm*, packSnorm4x8). In a theoretical world where we wanted to
> support every possibility, we'd have a separate "packing_400" internal
> enable which would imply that only the 4.0 ones are there. But in
> practice, it's all-or-nothing, so who cares. So I included it in the
> 4.0 list.
>
> Buuut... oops. Looks like those actually come in via ARB_gpu_shader5
> as well. And shading_language_packing has those *and* the extra ones
> (for GPUs that don't do gs5). OK. I'll move it to the 4.2 list.

Right, I should have mentioned that.

As you can tell, the packing/unpacking functions are really a
disaster. To make things worse, ES 3.0 exposes a different set of
functions than does gpu_shader5 or shading_language_packing. There are
actually *three* distinct subsets of shading_language_packing that may
be exposed depending on the API and extensions!