[Mesa-dev] [PATCH 5/7] i965/image_load_store: Add a brw_image_format_for_gl_format helper

[Ilia Mirkin]

On Wed, Nov 4, 2015 at 9:47 PM, Jason Ekstrand <jason at jlekstrand.net> wrote:
> On Wed, Nov 4, 2015 at 5:10 PM, Ilia Mirkin <imirkin at alum.mit.edu> wrote:
>> On Wed, Nov 4, 2015 at 8:03 PM, Jason Ekstrand <jason at jlekstrand.net> wrote:
>>> ---
>>>  src/mesa/drivers/dri/i965/brw_image_load_store.c | 49 ++++++++++++++++++++++++
>>>  src/mesa/drivers/dri/i965/brw_image_load_store.h |  2 +
>>>  2 files changed, 51 insertions(+)
>>>
>>> diff --git a/src/mesa/drivers/dri/i965/brw_image_load_store.c b/src/mesa/drivers/dri/i965/brw_image_load_store.c
>>> index b78d6b2..f2e2e36 100644
>>> --- a/src/mesa/drivers/dri/i965/brw_image_load_store.c
>>> +++ b/src/mesa/drivers/dri/i965/brw_image_load_store.c
>>> @@ -72,6 +72,55 @@ const struct brw_image_format_info brw_image_format_info[] = {
>>>  #undef IF
>>>
>>>  uint32_t
>>> +brw_image_format_for_gl_format(uint32_t gl_format)
>>> +{
>>> +   switch (gl_format) {
>>> +   case GL_R8:             return BRW_SURFACEFORMAT_R8_UNORM;
>>> +   case GL_R8_SNORM:       return BRW_SURFACEFORMAT_R8_SNORM;
>>> +   case GL_R8UI:           return BRW_SURFACEFORMAT_R8_UINT;
>>> +   case GL_R8I:            return BRW_SURFACEFORMAT_R8_SINT;
>>> +   case GL_RG8:            return BRW_SURFACEFORMAT_R8G8_UNORM;
>>> +   case GL_RG8_SNORM:      return BRW_SURFACEFORMAT_R8G8_SNORM;
>>> +   case GL_RG8UI:          return BRW_SURFACEFORMAT_R8G8_UINT;
>>> +   case GL_RG8I:           return BRW_SURFACEFORMAT_R8G8_SINT;
>>> +   case GL_RGBA8:          return BRW_SURFACEFORMAT_R8G8B8A8_UNORM;
>>> +   case GL_RGBA8_SNORM:    return BRW_SURFACEFORMAT_R8G8B8A8_SNORM;
>>> +   case GL_RGBA8UI:        return BRW_SURFACEFORMAT_R8G8B8A8_UINT;
>>> +   case GL_RGBA8I:         return BRW_SURFACEFORMAT_R8G8B8A8_SINT;
>>> +   case GL_R11F_G11F_B10F: return BRW_SURFACEFORMAT_R11G11B10_FLOAT;
>>> +   case GL_RGB10_A2:       return BRW_SURFACEFORMAT_R10G10B10A2_UNORM;
>>> +   case GL_RGB10_A2UI:     return BRW_SURFACEFORMAT_R10G10B10A2_UINT;
>>> +   case GL_R16:            return BRW_SURFACEFORMAT_R16_UNORM;
>>> +   case GL_R16_SNORM:      return BRW_SURFACEFORMAT_R16_SNORM;
>>> +   case GL_R16F:           return BRW_SURFACEFORMAT_R16_FLOAT;
>>> +   case GL_R16UI:          return BRW_SURFACEFORMAT_R16_UINT;
>>> +   case GL_R16I:           return BRW_SURFACEFORMAT_R16_SINT;
>>> +   case GL_RG16:           return BRW_SURFACEFORMAT_R16G16_UNORM;
>>> +   case GL_RG16_SNORM:     return BRW_SURFACEFORMAT_R16G16_SNORM;
>>> +   case GL_RG16F:          return BRW_SURFACEFORMAT_R16G16_FLOAT;
>>> +   case GL_RG16UI:         return BRW_SURFACEFORMAT_R16G16_UINT;
>>> +   case GL_RG16I:          return BRW_SURFACEFORMAT_R16G16_SINT;
>>> +   case GL_RGBA16:         return BRW_SURFACEFORMAT_R16G16B16A16_UNORM;
>>> +   case GL_RGBA16_SNORM:   return BRW_SURFACEFORMAT_R16G16B16A16_SNORM;
>>> +   case GL_RGBA16F:        return BRW_SURFACEFORMAT_R16G16B16A16_FLOAT;
>>> +   case GL_RGBA16UI:       return BRW_SURFACEFORMAT_R16G16B16A16_UINT;
>>> +   case GL_RGBA16I:        return BRW_SURFACEFORMAT_R16G16B16A16_SINT;
>>> +   case GL_R32F:           return BRW_SURFACEFORMAT_R32_FLOAT;
>>> +   case GL_R32UI:          return BRW_SURFACEFORMAT_R32_UINT;
>>> +   case GL_R32I:           return BRW_SURFACEFORMAT_R32_SINT;
>>> +   case GL_RG32F:          return BRW_SURFACEFORMAT_R32G32_FLOAT;
>>> +   case GL_RG32UI:         return BRW_SURFACEFORMAT_R32G32_UINT;
>>> +   case GL_RG32I:          return BRW_SURFACEFORMAT_R32G32_SINT;
>>> +   case GL_RGBA32F:        return BRW_SURFACEFORMAT_R32G32B32A32_FLOAT;
>>> +   case GL_RGBA32UI:       return BRW_SURFACEFORMAT_R32G32B32A32_UINT;
>>> +   case GL_RGBA32I:        return BRW_SURFACEFORMAT_R32G32B32A32_SINT;
>>> +   default:
>>> +      assert(!"Invalid image format");
>>> +      return GL_NONE;
>>
>> Seems odd to be returning GL_NONE here where all the other returns are
>> BRW_SURFACEFORMAT_*.
>
> Oops.  Yeah, that should be BRW_SURFACEFORMAT_RAW.  I'll get it fixed
> locally.  However, if anyone ever hits that, they'll hit the assert so
> it doesn't matter too much.

Asserts fall through in opt builds. (You probably knew that, but just
for posterity...)

  -ilia