guezoloic/t3dsr
1
0
Fork 0
mirror of https://github.com/guezoloic/t3dsr.git synced 2026-01-25 10:34:23 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(vec4): force 16 Bytes aligned

Browse Source
This commit is contained in:
Loïc GUEZO
2025-06-11 22:19:18 +02:00
parent e853d6d456
commit e157997d98
2 changed files with 36 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
src/math/vec4.c
View File

@@ -26,14 +26,6 @@ Vec4f_t vec4f_from_array(float *__restrict val)
return vec; return vec;
} }
// clone entire vec4f_t
Vec4f_t vec4f_clone(Vec4f_t *__restrict v)
{
Vec4f_t vec;
memcpy(&vec, v, sizeof(Vec4f_t));
return vec;
}
Vec4f_t vec4f_scalar(float f) Vec4f_t vec4f_scalar(float f)
{ {
Vec4f_t vec4; Vec4f_t vec4;
@@ -57,7 +49,7 @@ Vec4f_t vec4f_scalar(float f)
return vec4; return vec4;
} }
Vec4f_t Vec4f_zero(void) Vec4f_t vec4f_zero(void)
{ {
return vec4f_scalar(0.f); return vec4f_scalar(0.f);
} }
@@ -68,7 +60,7 @@ Vec4f_t vec4f_add_r(Vec4f_t *__restrict out, Vec4f_t a)
__m128 va = _mm_load_ps(a.data); __m128 va = _mm_load_ps(a.data);
__m128 vb = _mm_load_ps(out->data); __m128 vb = _mm_load_ps(out->data);
__m128 vres = _mm_add_ps(va, vb); __m128 vres = _mm_add_ps(va, vb);
_mm_store_ps(out->data, vres); _mm_storeu_ps(out->data, vres);
#elif defined (SIMD_ARCH) #elif defined (SIMD_ARCH)
float32x4_t va = vld1q_f32(a.data); float32x4_t va = vld1q_f32(a.data);
@@ -93,10 +85,10 @@ Vec4f_t vec4f_add(Vec4f_t a, Vec4f_t b)
Vec4f_t vec4f_sub_r(Vec4f_t *__restrict out, Vec4f_t a) Vec4f_t vec4f_sub_r(Vec4f_t *__restrict out, Vec4f_t a)
{ {
#if defined (SIMD_X86) #if defined (SIMD_X86)
__m128 va = _mm_load_ps(a.data); __m128 va = _mm_load_ps(out->data);
__m128 vb = _mm_load_ps(out->data); __m128 vb = _mm_load_ps(a.data);
__m128 vres = _mm_sub_ps(va, vb); __m128 vres = _mm_sub_ps(va, vb);
_mm_store_ps(out->data, vres); _mm_storeu_ps(out->data, vres);
#elif defined (SIMD_ARCH) #elif defined (SIMD_ARCH)
float32x4_t va = vld1q_f32(a.data); float32x4_t va = vld1q_f32(a.data);
@@ -125,7 +117,7 @@ Vec4f_t vec4f_scale_r(Vec4f_t *__restrict out, float scalar)
__m128 va = _mm_load_ps(out->data); __m128 va = _mm_load_ps(out->data);
__m128 vb = _mm_set1_ps(scalar); __m128 vb = _mm_set1_ps(scalar);
__m128 vres = _mm_mul_ps(va, vb); __m128 vres = _mm_mul_ps(va, vb);
_mm_store_ps(out->data, vres); _mm_storeu_ps(out->data, vres);
#elif defined (SIMD_ARCH) #elif defined (SIMD_ARCH)
float32x4_t va = vld1q_f32(out->data); float32x4_t va = vld1q_f32(out->data);
@@ -147,6 +139,21 @@ Vec4f_t vec4f_scale(Vec4f_t a, float scalar)
return vec; return vec;
} }
//float vec4f_dot(Vec4f_t a, Vec4f_t b)
//{
//#if defined (SIMD_X86)
// __m128 va = _mm_load_ps(a.data);
// __m128 vb = _mm_load_ps(b.data);
// __m128 vres = _mm_mul_ps(va, vb);
// return
//
//#elif defined (SIMD_ARCH)
//
//#else
//
//#endif
//}
// float vec4_dot(Vec4_t a, Vec4_t b) // float vec4_dot(Vec4_t a, Vec4_t b)
// { // {
// return a.x * b.x + a.y * b.y + a.z * b.z + a.w * b.w; // return a.x * b.x + a.y * b.y + a.z * b.z + a.w * b.w;

26
src/math/vec4.h
View File

@@ -14,11 +14,12 @@
#else #else
#endif #endif
typedef union // must be aligned by 16 Bytes (less instruction executed for SSE)
typedef union
{ {
struct { float x, y, z, w; }; struct { float x, y, z, w; };
float data[4]; float data[4];
} Vec4f_t; }__attribute__((aligned(16))) Vec4f_t;
Vec4f_t vec4f_from_array(float *__restrict val); Vec4f_t vec4f_from_array(float *__restrict val);
Vec4f_t vec4(float x, float y, float z, float w); Vec4f_t vec4(float x, float y, float z, float w);
@@ -26,8 +27,11 @@ Vec4f_t vec4(float x, float y, float z, float w);
Vec4f_t vec4f_scalar(float f); Vec4f_t vec4f_scalar(float f);
// (0, 0, 0, 0) // (0, 0, 0, 0)
Vec4f_t vec4f_zero(void); Vec4f_t vec4f_zero(void);
Vec4f_t vec4f_clone(Vec4f_t *__restrict v);
inline static Vec4f_t vec4f_clone(Vec4f_t *__restrict v)
{
return *v;
}
Vec4f_t vec4f_add_r(Vec4f_t *__restrict out, Vec4f_t a); Vec4f_t vec4f_add_r(Vec4f_t *__restrict out, Vec4f_t a);
Vec4f_t vec4f_add(Vec4f_t a, Vec4f_t b); Vec4f_t vec4f_add(Vec4f_t a, Vec4f_t b);
@@ -37,20 +41,20 @@ Vec4f_t vec4f_sub(Vec4f_t a, Vec4f_t b);
Vec4f_t vec4f_scale_r(Vec4f_t *__restrict out, float scale); Vec4f_t vec4f_scale_r(Vec4f_t *__restrict out, float scale);
Vec4f_t vec4f_scale(Vec4f_t a, float scale); Vec4f_t vec4f_scale(Vec4f_t a, float scale);
// float vec4_dot(Vec4_t a, Vec4_t b); float vec4_dot(Vec4f_t a, Vec4f_t b);
// float vec4_len(Vec4_t v); float vec4_len(Vec4f_t v);
// Vec4_t vec4_norm(Vec4_t v); Vec4f_t vec4_norm(Vec4f_t v);
// Vec4_t vec4_lerp(Vec4_t a, Vec4_t b, float t); Vec4f_t vec4_lerp(Vec4f_t a, Vec4f_t b, float t);
// float vec4_angle(Vec4_t a, Vec4_t b); float vec4_angle(Vec4f_t a, Vec4f_t b);
// Vec4_t vec4_proj(Vec4_t a, Vec4_t b); Vec4f_t vec4_proj(Vec4f_t a, Vec4f_t b);
// Vec4_t vec4_refl(Vec4_t v, Vec4_t normal); Vec4f_t vec4_refl(Vec4f_t v, Vec4f_t normal);
// float vec4_dist(Vec4_t a, Vec4_t b); float vec4_dist(Vec4f_t a, Vec4f_t b);
#endif // VECTOR4_H #endif // VECTOR4_H