youmuliu/stgsharp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

add double compute kernel

Browse Source
This commit is contained in:
Youmu Liu
2025-12-26 17:16:12 +08:00
parent e5539bcc6d
commit c6d53d97ea
16 changed files with 821 additions and 300 deletions
Download Patch File Download Diff File Expand all files Collapse all files
StgSharp.Native/sn-src/vector.c
+4 -4
View File
@@ -3,7 +3,7 @@
INTERNAL void SN_DECL f32_normalize(VEC(float) *restrict source, VEC(float) *restrict target)
{
register __m128 s = _mm_load_ps(&source->xmm);
register __m128 s = _mm_load_ps(&source->f32_x);
register __m128 a_length_sq = _mm_mul_ps(s, s);
a_length_sq = _mm_add_ps(a_length_sq,
_mm_shuffle_ps(a_length_sq, a_length_sq, _MM_SHUFFLE(2, 3, 0, 1)));
@@ -13,13 +13,13 @@ INTERNAL void SN_DECL f32_normalize(VEC(float) *restrict source, VEC(float) *res
__m128 a_inv_length = _mm_rsqrt_ps(a_length_sq);
s = _mm_mul_ps(s, a_inv_length);
__m128 dot_product = _mm_mul_ps(target->xmm[0], s);
__m128 dot_product = _mm_mul_ps(target->f32_x[0], s);
dot_product = _mm_add_ps(dot_product,
_mm_shuffle_ps(dot_product, dot_product, _MM_SHUFFLE(2, 3, 0, 1)));
dot_product = _mm_add_ss(dot_product,
_mm_shuffle_ps(dot_product, dot_product, _MM_SHUFFLE(0, 0, 3, 2)));
__m128 b_orthogonal = _mm_sub_ps(target->xmm[0], _mm_mul_ps(s, dot_product));
__m128 b_orthogonal = _mm_sub_ps(target->f32_x[0], _mm_mul_ps(s, dot_product));
__m128 b_length_sq = _mm_mul_ps(b_orthogonal, b_orthogonal);
b_length_sq = _mm_add_ps(b_length_sq,
@@ -28,5 +28,5 @@ INTERNAL void SN_DECL f32_normalize(VEC(float) *restrict source, VEC(float) *res
_mm_shuffle_ps(b_length_sq, b_length_sq, _MM_SHUFFLE(0, 0, 3, 2)));
__m128 b_inv_length = _mm_rsqrt_ps(b_length_sq);
target->xmm[0] = _mm_mul_ps(b_orthogonal, b_inv_length);
target->f32_x[0] = _mm_mul_ps(b_orthogonal, b_inv_length);
}