[Mesa-dev] [PATCH 1/2] gallium: add red-alpha texture formats and a couple of util functions

[Marek Olšák]

On Fri, Feb 8, 2013 at 1:56 AM, Brian Paul <brianp at vmware.com> wrote:
> On 02/07/2013 01:13 PM, Marek Olšák wrote:
>>
>> This is for glGetTexImage and it will be used for samplers only (which
>> some
>> drivers already implement by reading util_format_description).
>> ---
>>   src/gallium/auxiliary/util/u_format.csv |   12 +++
>>   src/gallium/auxiliary/util/u_format.h   |  123
>> +++++++++++++++++++++++++++++++
>>   src/gallium/include/pipe/p_format.h     |   12 +++
>>   3 files changed, 147 insertions(+)
>>
>> diff --git a/src/gallium/auxiliary/util/u_format.csv
>> b/src/gallium/auxiliary/util/u_format.csv
>> index 1d27407..f3925bb 100644
>> --- a/src/gallium/auxiliary/util/u_format.csv
>> +++ b/src/gallium/auxiliary/util/u_format.csv
>> @@ -360,3 +360,15 @@ PIPE_FORMAT_R16G16B16X16_SINT       , plain, 1, 1,
>> sp16, sp16, sp16, x16, xyz1,
>>   PIPE_FORMAT_R32G32B32X32_FLOAT      , plain, 1, 1, f32,  f32,  f32,
>> x32, xyz1, rgb
>>   PIPE_FORMAT_R32G32B32X32_UINT       , plain, 1, 1, up32, up32, up32,
>> x32, xyz1, rgb
>>   PIPE_FORMAT_R32G32B32X32_SINT       , plain, 1, 1, sp32, sp32, sp32,
>> x32, xyz1, rgb
>> +
>> +PIPE_FORMAT_R8A8_SNORM              , plain, 1, 1, sn8  ,  sn8 ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R16A16_UNORM            , plain, 1, 1, un16 , un16 ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R16A16_SNORM            , plain, 1, 1, sn16 , sn16 ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R16A16_FLOAT            , plain, 1, 1, f16  , f16  ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R32A32_FLOAT            , plain, 1, 1, f32  , f32  ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R8A8_UINT               , plain, 1, 1, up8  , up8  ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R8A8_SINT               , plain, 1, 1, sp8  , sp8  ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R16A16_UINT             , plain, 1, 1, up16 , up16 ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R16A16_SINT             , plain, 1, 1, sp16 , sp16 ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R32A32_UINT             , plain, 1, 1, up32 , up32 ,     ,
>> , x00y, rgb
>> +PIPE_FORMAT_R32A32_SINT             , plain, 1, 1, sp32 , sp32 ,     ,
>> , x00y, rgb
>> diff --git a/src/gallium/auxiliary/util/u_format.h
>> b/src/gallium/auxiliary/util/u_format.h
>> index 63015dd..1a5d954 100644
>> --- a/src/gallium/auxiliary/util/u_format.h
>> +++ b/src/gallium/auxiliary/util/u_format.h
>> @@ -888,6 +888,129 @@ util_format_stencil_only(enum pipe_format format)
>>   }
>>
>>   /**
>> + * Converts PIPE_FORMAT_*I* to PIPE_FORMAT_*R*.
>> + * This is identity for non-intensity formats.
>> + */
>> +static INLINE enum pipe_format
>> +util_format_intensity_to_red(enum pipe_format format)
>> +{
>> +   if (!util_format_is_intensity(format))
>> +      return format;
>
>
> If we removed that test, and just did 'return format' for the default case,
> wouldn't that work too?  We seem to be catching all the intensity formats in
> the switch.  That way we could avoid a util_format_description() call.
>
> Then, the default case below could have
> assert(!util_format_is_intensity(format))
> as a sanity check.

You're right. The assertion would suffice, I'll update the patch.

Marek