[Mesa-dev] [PATCH] gallium: kill off PIPE_FORMAT_Z32_UNORM with extreme prejudice

[Brian Paul]

On 10/24/2013 12:48 PM, sroland at vmware.com wrote:
> From: Roland Scheidegger <sroland at vmware.com>
>
> This format, while still supported in OpenGL (but optional) and glx, is just
> causing major nuisance everywhere and needs special code in some places,
> because things like 1 << depth_bits don't work.
> It is also the reason why we chose (just like in GL) depth clear values as
> doubles and not floats.
> The format however is just a disaster, no hw (as far as I know) supports it
> (because hw has all float pipelines, and floats don't have enough mantissa
> bits), and while we CLAIM to support it in software rasterizers, fact is it
> cannot (and never will) really work right, unless we wouldn't use floats for
> depth calculations. E.g. depth offsets can't work right (small depth bias will
> disappear with primitive depths > 1/256 due to the float calcs without enough
> mantissa bits), and it is generally useless (since the calculations are all
> float, use a float buffer if you need something better than z24 unorm).


Reviewed-by: Brian Paul <brianp at vmware.com>