fabi/Xenia-Canary
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[D3D12] Truncate depth to float24 in EDRAM range ownership transfers and resolves by default

Browse Source 
Doesn't ruin the "greater or equal" depth test in subsequent rendering passes if precision is lost, unlike rounding to the nearest
This commit is contained in:
Triang3l
2022-06-22 12:53:09 +03:00
parent e2f632f8fa
commit 7869b080d3
14 changed files with 193 additions and 185 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/xenia/gpu/draw_util.cc
View File

@@ -532,8 +532,10 @@ void GetHostViewportInfo(const RegisterFile& regs,
// interpolated Z instead if conversion can't be done exactly, without
// modifying clipping bounds by adjusting Z in vertex shaders, as that
// may cause polygons placed explicitly at Z = 0 or Z = W to be clipped.
z_min = xenos::Float20e4To32(xenos::Float32To20e4(z_min));
z_max = xenos::Float20e4To32(xenos::Float32To20e4(z_max));
// Rounding the bounds to the nearest even regardless of the depth
// rounding mode not to add even more error by truncating twice.
z_min = xenos::Float20e4To32(xenos::Float32To20e4(z_min, true));
z_max = xenos::Float20e4To32(xenos::Float32To20e4(z_max, true));
}
if (full_float24_in_0_to_1) {
// Remap the full [0...2) float24 range to [0...1) support data round-trip