YourWishes/cglm
1
0
Fork 0
mirror of https://github.com/recp/cglm.git synced 2026-02-17 03:39:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

optimize matrix4x4 inverse for SSE2

Browse Source
This commit is contained in:
Recep Aslantas
2016-09-16 20:24:55 +03:00
parent 55c1f93388
commit 253f5ba5f6
3 changed files with 107 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
include/cglm-intrin.h
View File

@@ -16,6 +16,17 @@
_mm_add_ps(_mm_mul_ps(_mm_set1_ps(*(L)), R0), \
_mm_mul_ps(_mm_set1_ps(*(L + 1)), R1))
#define _mm_msub_ps(M00, M01, M10, M11) \
_mm_sub_ps(_mm_mul_ps(M00, M01), \
_mm_mul_ps(M10, M11))
#define _mm_shuffle1_ps(a, z, y, x, w) \
_mm_shuffle_ps(a, a, _MM_SHUFFLE(z, y, x, w))
#define _mm_shuffle2_ps(a, b, z0, y0, x0, w0, z1, y1, x1, w1) \
_mm_shuffle1_ps(_mm_shuffle_ps(a, b, _MM_SHUFFLE(z0, y0, x0, w0)), \
z1, y1, x1, w1);
#define _mm_madd4_ps(L, R0, R1, R2, R3) \
_mm_add_ps(_mm_madd_ps(L, R0, R1), \
_mm_madd_ps(L + 2, R2, R3))