OrdinaryMagician/vkdoom_m
OrdinaryMagician/vkdoom_m
1
Fork 0
Code Activity

Switch to the lower precision rsqrtss instruction as it is faster, especially on older Intel architectures and on AMD

Browse source
This commit is contained in:
Magnus Norddahl 2016-12-22 21:21:57 +01:00
commit 9a529192b0
4 changed files with 17 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

2
tools/drawergen/fixedfunction/drawwallcodegen.cpp
View file

@ -244,7 +244,7 @@ SSAVec4i DrawWallCodegen::Shade(SSAVec4i fg, int index, bool isSimpleShade)
// attenuation = 1 - MIN(dist * (1/radius), 1)
SSAFloat Lxy2 = light_x; // L.x*L.x + L.y*L.y
SSAFloat Lz = light_z - z;
SSAFloat dist = SSAFloat::sqrt(Lxy2 + Lz * Lz);
SSAFloat dist = SSAFloat::fastsqrt(Lxy2 + Lz * Lz);
SSAInt attenuation = SSAInt(SSAFloat(256.0f) - SSAFloat::MIN(dist * light_rcp_radius, SSAFloat(256.0f)), true);
SSAVec4i contribution = (light_color * fg * attenuation) >> 16;