External/cglm
1
0
Fork 0
mirror of https://github.com/recp/cglm.git synced 2025-10-04 09:08:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

vec4: optimize vec4 norm and norm2

Browse Source
This commit is contained in:
Recep Aslantas
2018-04-13 11:18:42 +03:00
parent 0eb37da8bb
commit 79f8b1ebf8
4 changed files with 42 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
include/cglm/vec4.h
View File

@@ -146,7 +146,15 @@ glm_vec4_dot(vec4 a, vec4 b) {
CGLM_INLINE CGLM_INLINE
float float
glm_vec4_norm2(vec4 v) { glm_vec4_norm2(vec4 v) {
return glm_vec4_dot(v, v); #if defined( __SSE__ ) || defined( __SSE2__ )
__m128 x0;
x0 = _mm_load_ps(v);
x0 = _mm_mul_ps(x0, x0);
x0 = _mm_add_ps(x0, _mm_shuffle1_ps(x0, 1, 0, 3, 2));
return _mm_cvtss_f32(_mm_add_ss(x0, _mm_shuffle1_ps(x0, 0, 1, 0, 1)));
#else
return a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + a[3] * b[3];
#endif
} }
/*! /*!
@@ -159,7 +167,13 @@ glm_vec4_norm2(vec4 v) {
CGLM_INLINE CGLM_INLINE
float float
glm_vec4_norm(vec4 vec) { glm_vec4_norm(vec4 vec) {
#if defined( __SSE__ ) || defined( __SSE2__ )
__m128 x0;
x0 = _mm_load_ps(vec);
return _mm_cvtss_f32(_mm_sqrt_ss(glm_simd_dot(x0, x0)));
#else
return sqrtf(glm_vec4_norm2(vec)); return sqrtf(glm_vec4_norm2(vec));
#endif
} }
/*! /*!

5
test/src/test_common.c
View File

@@ -84,6 +84,11 @@ test_assert_mat4_eq2(mat4 m1, mat4 m2, float eps) {
} }
} }
void
test_assert_eqf(float a, float b) {
assert_true(fabsf(a - b) <= 0.000009); /* rounding errors */
}
void void
test_assert_vec3_eq(vec3 v1, vec3 v2) { test_assert_vec3_eq(vec3 v1, vec3 v2) {
assert_true(fabsf(v1[0] - v2[0]) <= 0.000009); /* rounding errors */ assert_true(fabsf(v1[0] - v2[0]) <= 0.000009); /* rounding errors */

3
test/src/test_common.h
View File

@@ -25,6 +25,9 @@
void void
test_rand_mat4(mat4 dest); test_rand_mat4(mat4 dest);
void
test_assert_eqf(float a, float b);
void void
test_assert_mat4_eq(mat4 m1, mat4 m2); test_assert_mat4_eq(mat4 m1, mat4 m2);

20
test/src/test_vec4.c
View File

@@ -28,6 +28,16 @@ test_vec4_normalize_to(vec4 vec, vec4 dest) {
glm_vec4_scale(vec, 1.0f / norm, dest); glm_vec4_scale(vec, 1.0f / norm, dest);
} }
float
test_vec4_norm2(vec4 vec) {
return test_vec4_dot(vec, vec);
}
float
test_vec4_norm(vec4 vec) {
return sqrtf(test_vec4_dot(vec, vec));
}
void void
test_vec4(void **state) { test_vec4(void **state) {
vec4 v, v1, v2; vec4 v, v1, v2;
@@ -35,7 +45,7 @@ test_vec4(void **state) {
float d1, d2; float d1, d2;
for (i = 0; i < 100; i++) { for (i = 0; i < 1000; i++) {
/* 1. test SSE/SIMD dot product */ /* 1. test SSE/SIMD dot product */
test_rand_vec4(v); test_rand_vec4(v);
d1 = glm_vec4_dot(v, v); d1 = glm_vec4_dot(v, v);
@@ -51,5 +61,13 @@ test_vec4(void **state) {
/* all must be same */ /* all must be same */
test_assert_vec4_eq(v1, v2); test_assert_vec4_eq(v1, v2);
test_assert_vec4_eq(v, v2); test_assert_vec4_eq(v, v2);
/* 3. test SIMD norm */
test_rand_vec4(v);
test_assert_eqf(test_vec4_norm(v), glm_vec4_norm(v));
/* 3. test SIMD norm2 */
test_rand_vec4(v);
test_assert_eqf(test_vec4_norm2(v), glm_vec4_norm2(v));
} }
} }