[d3d9] Determine depth bias r-factors

[K0bin]

Ref: #2892

I've only tested it on Nvidia and it seems to work.

[misyltoad]

Making a gradient?

[K0bin]

Yes.

For floating-point depth attachment, there is no single minimum resolvable
difference. In this case, the minimum resolvable difference for a given polygon is dependent on the
maximum exponent, e, in the range of z values spanned by the primitive. If n is the number of bits
in the floating-point mantissa, the minimum resolvable difference, r, for the given primitive is
defined as: 2^(e-n)

Emphasis mine. But because of that we can't really solve it with a simple factor. I went with the factor for a range from 0-1 as a best effort.

The top left vertex is supposed to be at depth 0 because that's what we read back for the depth bias.

[K0bin]

Unfortunately it doesn't seem to fix some of the depth bias problems and returns rather weird values for AMD.

On Nvidia, I get 1/(1<<23) for D24 and D32 and 1/(1<<15) for D16. Wine calculates the same values.

CME got:

info: Using depth bias r-factor: 1.52588e-05 for format: VK_FORMAT_D16_UNORM
info: Using depth bias r-factor: 1.52588e-05 for format: VK_FORMAT_D16_UNORM_S8_UINT
info: Using depth bias r-factor: 1.21072e-07 for format: VK_FORMAT_D32_SFLOAT
info: Using depth bias r-factor: 1.21072e-07 for format: VK_FORMAT_D32_SFLOAT_S8_UINT

which doesn't match what Wine calculates.

[misyltoad]

@CME42 Can you try with nodcc