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

arm, neon: neon/fma support for glm_mat2_mul()

Browse Source
This commit is contained in:
Recep Aslantas
2021-04-28 22:06:46 +03:00
parent 55ebbdbe40
commit 4e4bff418d
2 changed files with 50 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
include/cglm/mat2.h
View File

@@ -40,6 +40,10 @@
# include "simd/sse2/mat2.h" # include "simd/sse2/mat2.h"
#endif #endif
#ifdef CGLM_NEON_FP
# include "simd/neon/mat2.h"
#endif
#define GLM_MAT2_IDENTITY_INIT {{1.0f, 0.0f}, {0.0f, 1.0f}} #define GLM_MAT2_IDENTITY_INIT {{1.0f, 0.0f}, {0.0f, 1.0f}}
#define GLM_MAT2_ZERO_INIT {{0.0f, 0.0f}, {0.0f, 0.0f}} #define GLM_MAT2_ZERO_INIT {{0.0f, 0.0f}, {0.0f, 0.0f}}
@@ -130,6 +134,8 @@ void
glm_mat2_mul(mat2 m1, mat2 m2, mat2 dest) { glm_mat2_mul(mat2 m1, mat2 m2, mat2 dest) {
#if defined( __SSE__ ) || defined( __SSE2__ ) #if defined( __SSE__ ) || defined( __SSE2__ )
glm_mat2_mul_sse2(m1, m2, dest); glm_mat2_mul_sse2(m1, m2, dest);
#elif defined(CGLM_NEON_FP)
glm_mat2_mul_neon(m1, m2, dest);
#else #else
float a00 = m1[0][0], a01 = m1[0][1], float a00 = m1[0][0], a01 = m1[0][1],
a10 = m1[1][0], a11 = m1[1][1], a10 = m1[1][0], a11 = m1[1][1],

44
include/cglm/simd/neon/mat2.h Normal file
View File

@@ -0,0 +1,44 @@
/*
* Copyright (c), Recep Aslantas.
*
* MIT License (MIT), http://opensource.org/licenses/MIT
* Full license can be found in the LICENSE file
*/
#ifndef cglm_mat2_neon_h
#define cglm_mat2_neon_h
#if defined(__ARM_NEON_FP)
#include "../../common.h"
#include "../intrin.h"
CGLM_INLINE
void
glm_mat2_mul_neon(mat2 m1, mat2 m2, mat2 dest) {
float32x4x2_t a1;
glmm_128 x0, x1, x2;
float32x2_t dc, ba;
x1 = glmm_load(m1[0]); /* d c b a */
x2 = glmm_load(m2[0]); /* h g f e */
dc = vget_high_f32(x1);
ba = vget_low_f32(x1);
/* g g e e, h h f f */
a1 = vtrnq_f32(x2, x2);
/*
dest[0][0] = a * e + c * f;
dest[0][1] = b * e + d * f;
dest[1][0] = a * g + c * h;
dest[1][1] = b * g + d * h;
*/
x0 = glmm_fmadd(vcombine_f32(ba, ba), a1.val[0],
vmulq_f32(vcombine_f32(dc, dc), a1.val[1]));
glmm_store(dest[0], x0);
}
#endif
#endif /* cglm_mat2_neon_h */