The compressor does not currently correctly handle sRGB alpha component values, which should decompress to an fp16 value, or reflect the decode_unorm8 extension behavior when writing to an 8-bit per component output image.
This PR attempts to improve this so that:
sRGB alpha will decompress to the correct fp16 intermediate when storing to an fp16 output image.
LDR decompression uses the correct output specified for decode_unorm8 when writing to an 8-bit per component output image.
The compressor does not currently correctly handle sRGB alpha component values, which should decompress to an fp16 value, or reflect the decode_unorm8 extension behavior when writing to an 8-bit per component output image.
This PR attempts to improve this so that: