[Mesa-dev] [PATCH 5/7] ac/radeonsi: pass bindless bool to load_sampler_desc()

[Marek Olšák]

On Wed, Feb 21, 2018 at 2:21 AM, Timothy Arceri <tarceri at itsqueeze.com> wrote:
>
>
> On 21/02/18 12:10, Marek Olšák wrote:
>>
>> On Wed, Feb 21, 2018 at 12:50 AM, Timothy Arceri <tarceri at itsqueeze.com>
>> wrote:
>>>
>>> On 21/02/18 10:33, Marek Olšák wrote:
>>>>
>>>>
>>>> On Tue, Feb 20, 2018 at 11:51 PM, Timothy Arceri <tarceri at itsqueeze.com>
>>>> wrote:
>>>>>
>>>>>
>>>>> On 21/02/18 09:46, Marek Olšák wrote:
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Tue, Feb 20, 2018 at 11:42 PM, Marek Olšák <maraeo at gmail.com>
>>>>>> wrote:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> For patches 1-5:
>>>>>>>
>>>>>>> Reviewed-by: Marek Olšák <marek.olsak at amd.com>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> Actually no. Only patches 1, 3, 5 are reviewed by me.
>>>>>>
>>>>>> Marek
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> Do you have an issue with patch 4?
>>>>
>>>>
>>>>
>>>> No, I'm just not sure if it's correct. It calls
>>>> st_nir_lookup_parameter_index, but bindless handless are just
>>>> variables. I think it should just visit the whole expression leading
>>>> to the bindless variable in a generic way and not treat it as a
>>>> uniform.
>>>
>>>
>>>
>>> I'm not sure I understand. We use uniform storage for bindless in tgsi
>>> also.
>>
>>
>> A bindless (sampler or buffer) variable is represented as a 64-bit
>> number in the GL API. It can be passed to shaders in many different
>> ways. For example, a bindless sampler2D variable can be a vertex
>> shader input (loaded from a vertex buffer).
>
>
> Right I should have specified this series does not yet handle bindless
> input/output support, that will require more updates to nir itself as those
> shaders currently trip asserts. Patch 4 however is specifically about
> bindless uniforms.

OK. Patch 4 also has my Rb.

Marek