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

avx: optimize (re-use) mat4_mul registers

Browse Source
This commit is contained in:
Recep Aslantas
2018-10-30 09:27:55 +03:00
parent 08479f38ce
commit abfa355b84
1 changed files with 21 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
include/cglm/simd/avx/mat4.h
View File

@@ -19,26 +19,32 @@ void
glm_mat4_mul_avx(mat4 m1, mat4 m2, mat4 dest) { glm_mat4_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
/* D = R * L (Column-Major) */ /* D = R * L (Column-Major) */
__m256 y0, y1, y2, y3, y4, y5, y6, y7, y8, y9; __m256 y0, y1, y2, y3, y4, y5, y6, y7, y8, y9;
__m256i yi0, yi1, yi2, yi3;
y0 = glmm_load256(m2[0]); /* h g f e d c b a */ y0 = glmm_load256(m2[0]); /* h g f e d c b a */
y1 = glmm_load256(m2[2]); /* p o n m l k j i */ y1 = glmm_load256(m2[2]); /* p o n m l k j i */
y2 = glmm_load256(m1[0]); /* h g f e d c b a */ y2 = glmm_load256(m1[0]); /* h g f e d c b a */
y3 = glmm_load256(m1[2]); /* p o n m l k j i */ y3 = glmm_load256(m1[2]); /* p o n m l k j i */
/* 0x03: 0b00000011 */ /* 0x03: 0b00000011 */
y4 = _mm256_permute2f128_ps(y2, y2, 0x03); /* d c b a h g f e */ y4 = _mm256_permute2f128_ps(y2, y2, 0x03); /* d c b a h g f e */
y5 = _mm256_permute2f128_ps(y3, y3, 0x03); /* l k j i p o n m */ y5 = _mm256_permute2f128_ps(y3, y3, 0x03); /* l k j i p o n m */
yi0 = _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0);
yi1 = _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2);
yi2 = _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1);
yi3 = _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3);
/* f f f f a a a a */ /* f f f f a a a a */
/* h h h h c c c c */ /* h h h h c c c c */
/* e e e e b b b b */ /* e e e e b b b b */
/* g g g g d d d d */ /* g g g g d d d d */
y6 = _mm256_permutevar_ps(y0, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0)); y6 = _mm256_permutevar_ps(y0, yi0);
y7 = _mm256_permutevar_ps(y0, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2)); y7 = _mm256_permutevar_ps(y0, yi1);
y8 = _mm256_permutevar_ps(y0, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1)); y8 = _mm256_permutevar_ps(y0, yi2);
y9 = _mm256_permutevar_ps(y0, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3)); y9 = _mm256_permutevar_ps(y0, yi3);
glmm_store256(dest[0], glmm_store256(dest[0],
_mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6), _mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
@@ -50,10 +56,10 @@ glm_mat4_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
/* p p p p k k k k */ /* p p p p k k k k */
/* m m m m j j j j */ /* m m m m j j j j */
/* o o o o l l l l */ /* o o o o l l l l */
y6 = _mm256_permutevar_ps(y1, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0)); y6 = _mm256_permutevar_ps(y1, yi0);
y7 = _mm256_permutevar_ps(y1, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2)); y7 = _mm256_permutevar_ps(y1, yi1);
y8 = _mm256_permutevar_ps(y1, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1)); y8 = _mm256_permutevar_ps(y1, yi2);
y9 = _mm256_permutevar_ps(y1, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3)); y9 = _mm256_permutevar_ps(y1, yi3);
glmm_store256(dest[2], glmm_store256(dest[2],
_mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6), _mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),