[Mesa-dev] [PATCH 2/2] i965: Implement ARB_texture_filter_minmax

[Ian Romanick]

On 11/19/2017 01:31 AM, Kenneth Graunke wrote:
> On Thursday, November 16, 2017 11:50:48 AM PST Ian Romanick wrote:
>> On 11/14/2017 02:54 PM, Scott D Phillips wrote:
>>> On gen >= 9, minmax reduction modes are available as a flag in
>>> SAMPLER_STATE.
>>> ---
>>>  docs/features.txt                             |  2 +-
>>>  src/mesa/drivers/dri/i965/brw_formatquery.c   |  4 ++++
>>>  src/mesa/drivers/dri/i965/genX_state_upload.c | 10 ++++++++++
>>>  src/mesa/drivers/dri/i965/intel_extensions.c  |  1 +
>>>  4 files changed, 16 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/docs/features.txt b/docs/features.txt
>>> index 86d07ba80b..9ec3f2b975 100644
>>> --- a/docs/features.txt
>>> +++ b/docs/features.txt
>>> @@ -312,7 +312,7 @@ Khronos, ARB, and OES extensions that are not part of any OpenGL or OpenGL ES ve
>>>    GL_ARB_sparse_texture                                 not started
>>>    GL_ARB_sparse_texture2                                not started
>>>    GL_ARB_sparse_texture_clamp                           not started
>>> -  GL_ARB_texture_filter_minmax                          not started
>>> +  GL_ARB_texture_filter_minmax                          DONE (i965)
>>>    GL_EXT_memory_object                                  DONE (radeonsi)
>>>    GL_EXT_memory_object_fd                               DONE (radeonsi)
>>>    GL_EXT_memory_object_win32                            not started
>>> diff --git a/src/mesa/drivers/dri/i965/brw_formatquery.c b/src/mesa/drivers/dri/i965/brw_formatquery.c
>>> index 4f3b9e467b..bb2281f571 100644
>>> --- a/src/mesa/drivers/dri/i965/brw_formatquery.c
>>> +++ b/src/mesa/drivers/dri/i965/brw_formatquery.c
>>> @@ -107,6 +107,10 @@ brw_query_internal_format(struct gl_context *ctx, GLenum target,
>>>        break;
>>>     }
>>>  
>>> +   case GL_TEXTURE_REDUCTION_MODE_ARB:
>>> +      params[0] = GL_TRUE;
>>> +      break;
>>> +
>>
>> Can Gen9 actually support all formats?  When this was getting discussed
>> in Khronos, I thought we had some format restrictions.  This is the
>> difference between the EXT and the ARB extension.  If we don't actually
>> have any restrictions on Gen9, a good follow-on would be to add support
>> for the EXT... since that doesn't have the OpenGL 3.3 requirement and
>> has interactions with OpenGL ES.
>>
>> I also thought that Gen8 could do some min/max modes.  Perhaps that's
>> where the restrictions were that I was thinking of...
> 
> Skylake has the SAMPLER_STATE reduction mode field.
> 
> Previous generations had sampler_min/max messages that work differently.
> 
> If I recall, we discussed this previously and concluded the old messages
> weren't that useful...or at least, would result in more NOS and awkward
> restrictions.

That's probably what I was thinking of.

So... we think it should "just work" for all formats on SKL?  It sounds
like various test suites only exercise a very small set of
single-channel formats.  I think we should add piglit tests to poke at
least some other formats.  I talked to Scott about this last week, and
he seemed amenable to the idea.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: OpenPGP digital signature
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20171120/b35488b1/attachment-0001.sig>