OrdinaryMagician/vkdoom_m
OrdinaryMagician/vkdoom_m
1
Fork 0
Code Activity

- improve triangle setup performance a little bit

Browse source
This commit is contained in:
Magnus Norddahl 2017-03-24 20:00:53 +01:00
commit ac518e23bf
3 changed files with 237 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

15
src/polyrenderer/math/tri_matrix.cpp
View file

@ -175,6 +175,7 @@ TriMatrix TriMatrix::operator*(const TriMatrix &mult) const
ShadedTriVertex TriMatrix::operator*(TriVertex v) const
{
#ifdef NO_SSE
float vx = matrix[0 * 4 + 0] * v.x + matrix[1 * 4 + 0] * v.y + matrix[2 * 4 + 0] * v.z + matrix[3 * 4 + 0] * v.w;
float vy = matrix[0 * 4 + 1] * v.x + matrix[1 * 4 + 1] * v.y + matrix[2 * 4 + 1] * v.z + matrix[3 * 4 + 1] * v.w;
float vz = matrix[0 * 4 + 2] * v.x + matrix[1 * 4 + 2] * v.y + matrix[2 * 4 + 2] * v.z + matrix[3 * 4 + 2] * v.w;
@ -184,6 +185,20 @@ ShadedTriVertex TriMatrix::operator*(TriVertex v) const
sv.y = vy;
sv.z = vz;
sv.w = vw;
#else
__m128 m0 = _mm_loadu_ps(matrix);
__m128 m1 = _mm_loadu_ps(matrix + 4);
__m128 m2 = _mm_loadu_ps(matrix + 8);
__m128 m3 = _mm_loadu_ps(matrix + 12);
__m128 mv = _mm_loadu_ps(&v.x);
m0 = _mm_mul_ps(m0, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(0, 0, 0, 0)));
m1 = _mm_mul_ps(m1, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(1, 1, 1, 1)));
m2 = _mm_mul_ps(m2, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(2, 2, 2, 2)));
m3 = _mm_mul_ps(m3, _mm_shuffle_ps(mv, mv, _MM_SHUFFLE(3, 3, 3, 3)));
mv = _mm_add_ps(_mm_add_ps(_mm_add_ps(m0, m1), m2), m3);
ShadedTriVertex sv;
_mm_storeu_ps(&sv.x, mv);
#endif
for (int i = 0; i < TriVertex::NumVarying; i++)
sv.varying[i] = v.varying[i];
return sv;