External/cglm
1
0
Fork 0
mirror of https://github.com/recp/cglm.git synced 2025-10-04 01:00:46 +00:00
Code Issues Packages Projects Releases Wiki Activity

avx: implement mat4 avx mul

Browse Source
This commit is contained in:
Recep Aslantas
2016-09-21 16:15:48 +03:00
parent ce74bc4a82
commit b8589ae0e2
2 changed files with 67 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

63
include/cglm-mat-simd-avx.h Normal file
View File

@@ -0,0 +1,63 @@
/*
* Copyright (c), Recep Aslantas.
*
* MIT License (MIT), http://opensource.org/licenses/MIT
* Full license can be found in the LICENSE file
*/
#ifndef cglm_mat_simd_avx_h
#define cglm_mat_simd_avx_h
#ifdef __AVX__
#include "cglm-intrin.h"
#include <immintrin.h>
CGLM_INLINE
void
glm_mat4_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
/* D = R * L (Column-Major) */
__m256 y0, y1, y2, y3, y4, y5, y6, y7, y8, y9;
y0 = _mm256_load_ps(m2[0]); /* h g f e d c b a */
y1 = _mm256_load_ps(m2[2]); /* p o n m l k j i */
y2 = _mm256_load_ps(m1[0]); /* h g f e d c b a */
y3 = _mm256_load_ps(m1[2]); /* p o n m l k j i */
y4 = _mm256_permute2f128_ps(y0, y0, 0b00000011); /* d c b a h g f e */
y5 = _mm256_permute2f128_ps(y3, y1, 0b00000011); /* l k j i p o n m */
/* f f f f a a a a */
/* h h h h c c c c */
/* e e e e b b b b */
/* g g g g d d d d */
y6 = _mm256_permutevar_ps(y0, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0));
y7 = _mm256_permutevar_ps(y0, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2));
y8 = _mm256_permutevar_ps(y0, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1));
y9 = _mm256_permutevar_ps(y0, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3));
_mm256_store_ps(dest[0],
_mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
_mm256_mul_ps(y3, y7)),
_mm256_add_ps(_mm256_mul_ps(y4, y8),
_mm256_mul_ps(y5, y9))));
/* n n n n i i i i */
/* p p p p k k k k */
/* m m m m j j j j */
/* o o o o l l l l */
y6 = _mm256_permutevar_ps(y1, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0));
y7 = _mm256_permutevar_ps(y1, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2));
y8 = _mm256_permutevar_ps(y1, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1));
y9 = _mm256_permutevar_ps(y1, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3));
_mm256_store_ps(dest[2],
_mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
_mm256_mul_ps(y3, y7)),
_mm256_add_ps(_mm256_mul_ps(y4, y8),
_mm256_mul_ps(y5, y9))));
}
#endif
#endif /* cglm_mat_simd_avx_h */

5
include/cglm-mat.h
View File

@@ -10,6 +10,7 @@
#include "cglm.h" #include "cglm.h"
#include "cglm-mat-simd.h" #include "cglm-mat-simd.h"
#include "cglm-mat-simd-avx.h"
#include <assert.h> #include <assert.h>
#define GLM_MAT4_IDENTITY_INIT {1.0f, 0.0f, 0.0f, 0.0f, \ #define GLM_MAT4_IDENTITY_INIT {1.0f, 0.0f, 0.0f, 0.0f, \
@@ -22,7 +23,9 @@
CGLM_INLINE CGLM_INLINE
void void
glm_mat4_mul(mat4 l, mat4 r, mat4 d) { glm_mat4_mul(mat4 l, mat4 r, mat4 d) {
#if defined( __SSE__ ) || defined( __SSE2__ ) #ifdef __AVX__
glm_mat4_mul_avx(l, r, d);
#elif defined( __SSE__ ) || defined( __SSE2__ )
glm_mat4_mul_sse2(l, r, d); glm_mat4_mul_sse2(l, r, d);
#else #else
d[0][0] = l[0][0] * r[0][0] + l[1][0] * r[0][1] + d[0][0] = l[0][0] * r[0][0] + l[1][0] * r[0][1] +