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

simd, sse2: make alignment optional for load operations

Browse Source
This commit is contained in:
Recep Aslantas
2018-05-08 15:25:23 +03:00
parent 0f339c5c03
commit 252bf925fc
9 changed files with 118 additions and 110 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
include/cglm/simd/sse2/affine.h
View File

@@ -18,30 +18,30 @@ glm_mul_sse2(mat4 m1, mat4 m2, mat4 dest) {
/* D = R * L (Column-Major) */
__m128 l0, l1, l2, l3, r;
l0 = _mm_load_ps(m1[0]);
l1 = _mm_load_ps(m1[1]);
l2 = _mm_load_ps(m1[2]);
l3 = _mm_load_ps(m1[3]);
l0 = glmm_load(m1[0]);
l1 = glmm_load(m1[1]);
l2 = glmm_load(m1[2]);
l3 = glmm_load(m1[3]);
r = _mm_load_ps(m2[0]);
r = glmm_load(m2[0]);
_mm_store_ps(dest[0],
_mm_add_ps(_mm_add_ps(_mm_mul_ps(_mm_shuffle1_ps1(r, 0), l0),
_mm_mul_ps(_mm_shuffle1_ps1(r, 1), l1)),
_mm_mul_ps(_mm_shuffle1_ps1(r, 2), l2)));
r = _mm_load_ps(m2[1]);
r = glmm_load(m2[1]);
_mm_store_ps(dest[1],
_mm_add_ps(_mm_add_ps(_mm_mul_ps(_mm_shuffle1_ps1(r, 0), l0),
_mm_mul_ps(_mm_shuffle1_ps1(r, 1), l1)),
_mm_mul_ps(_mm_shuffle1_ps1(r, 2), l2)));
r = _mm_load_ps(m2[2]);
r = glmm_load(m2[2]);
_mm_store_ps(dest[2],
_mm_add_ps(_mm_add_ps(_mm_mul_ps(_mm_shuffle1_ps1(r, 0), l0),
_mm_mul_ps(_mm_shuffle1_ps1(r, 1), l1)),
_mm_mul_ps(_mm_shuffle1_ps1(r, 2), l2)));
r = _mm_load_ps(m2[3]);
r = glmm_load(m2[3]);
_mm_store_ps(dest[3],
_mm_add_ps(_mm_add_ps(_mm_mul_ps(_mm_shuffle1_ps1(r, 0), l0),
_mm_mul_ps(_mm_shuffle1_ps1(r, 1), l1)),
@@ -55,24 +55,24 @@ glm_mul_rot_sse2(mat4 m1, mat4 m2, mat4 dest) {
/* D = R * L (Column-Major) */
__m128 l0, l1, l2, l3, r;
l0 = _mm_load_ps(m1[0]);
l1 = _mm_load_ps(m1[1]);
l2 = _mm_load_ps(m1[2]);
l3 = _mm_load_ps(m1[3]);
l0 = glmm_load(m1[0]);
l1 = glmm_load(m1[1]);
l2 = glmm_load(m1[2]);
l3 = glmm_load(m1[3]);
r = _mm_load_ps(m2[0]);
r = glmm_load(m2[0]);
_mm_store_ps(dest[0],
_mm_add_ps(_mm_add_ps(_mm_mul_ps(_mm_shuffle1_ps1(r, 0), l0),
_mm_mul_ps(_mm_shuffle1_ps1(r, 1), l1)),
_mm_mul_ps(_mm_shuffle1_ps1(r, 2), l2)));
r = _mm_load_ps(m2[1]);
r = glmm_load(m2[1]);
_mm_store_ps(dest[1],
_mm_add_ps(_mm_add_ps(_mm_mul_ps(_mm_shuffle1_ps1(r, 0), l0),
_mm_mul_ps(_mm_shuffle1_ps1(r, 1), l1)),
_mm_mul_ps(_mm_shuffle1_ps1(r, 2), l2)));
r = _mm_load_ps(m2[2]);
r = glmm_load(m2[2]);
_mm_store_ps(dest[2],
_mm_add_ps(_mm_add_ps(_mm_mul_ps(_mm_shuffle1_ps1(r, 0), l0),
_mm_mul_ps(_mm_shuffle1_ps1(r, 1), l1)),
@@ -86,10 +86,10 @@ void
glm_inv_tr_sse2(mat4 mat) {
__m128 r0, r1, r2, r3, x0, x1;
r0 = _mm_load_ps(mat[0]);
r1 = _mm_load_ps(mat[1]);
r2 = _mm_load_ps(mat[2]);
r3 = _mm_load_ps(mat[3]);
r0 = glmm_load(mat[0]);
r1 = glmm_load(mat[1]);
r2 = glmm_load(mat[2]);
r3 = glmm_load(mat[3]);
x1 = _mm_set_ps(1.0f, 0.0f, 0.0f, 0.0f);
_MM_TRANSPOSE4_PS(r0, r1, r2, x1);