ilikecats/stk-code_catmod
1
1
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Revert "Revert "Use pointer arithmetic to enforce matrix 16 bytes alignment""

Browse Source 
This reverts commit e83bd11891.
This commit is contained in:
Vincent Lejeune 2014-09-07 00:44:49 +02:00
parent e83bd11891
commit b9a595e666
1 changed files with 11 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
lib/irrlicht/include/matrix4.h
View File

@ -47,6 +47,8 @@ namespace core
template <class T>
class CMatrix4
{
private:
float M_raw[24];
public:
//! Constructor Flags
@ -406,7 +408,7 @@ namespace core
private:
//! Matrix data, stored in row-major order
T M[16];
T* M = (T*)((uintptr_t)&M_raw[4] & ~0xF);
#if defined ( USE_MATRIX_TEST )
//! Flag is this matrix is identity matrix
mutable u32 definitelyIdentityMatrix;
@ -667,10 +669,10 @@ namespace core
const float *matA = other_a.pointer();
const __m128 a = _mm_loadu_ps(matA); // First row
const __m128 b = _mm_loadu_ps(&matA[4]); // Second row
const __m128 c = _mm_loadu_ps(&matA[8]); // Third row
const __m128 d = _mm_loadu_ps(&matA[12]); // Fourth row
const __m128 a = _mm_load_ps(matA); // First row
const __m128 b = _mm_load_ps(&matA[4]); // Second row
const __m128 c = _mm_load_ps(&matA[8]); // Third row
const __m128 d = _mm_load_ps(&matA[12]); // Fourth row
__m128 t1 = _mm_set1_ps(m2[0]);
__m128 t2 = _mm_mul_ps(a, t1);
@ -680,7 +682,7 @@ namespace core
t2 = _mm_add_ps(_mm_mul_ps(c, t1), t2);
t1 = _mm_set1_ps(m2[3]);
t2 = _mm_add_ps(_mm_mul_ps(d, t1), t2);
_mm_storeu_ps(&M[0], t2);
_mm_store_ps(&M[0], t2);
t1 = _mm_set1_ps(m2[4]);
t2 = _mm_mul_ps(a, t1);
@ -690,7 +692,7 @@ namespace core
t2 = _mm_add_ps(_mm_mul_ps(c, t1), t2);
t1 = _mm_set1_ps(m2[7]);
t2 = _mm_add_ps(_mm_mul_ps(d, t1), t2);
_mm_storeu_ps(&M[4], t2);
_mm_store_ps(&M[4], t2);
t1 = _mm_set1_ps(m2[8]);
t2 = _mm_mul_ps(a, t1);
@ -700,7 +702,7 @@ namespace core
t2 = _mm_add_ps(_mm_mul_ps(c, t1), t2);
t1 = _mm_set1_ps(m2[11]);
t2 = _mm_add_ps(_mm_mul_ps(d, t1), t2);
_mm_storeu_ps(&M[8], t2);
_mm_store_ps(&M[8], t2);
t1 = _mm_set1_ps(m2[12]);
t2 = _mm_mul_ps(a, t1);
@ -710,7 +712,7 @@ namespace core
t2 = _mm_add_ps(_mm_mul_ps(c, t1), t2);
t1 = _mm_set1_ps(m2[15]);
t2 = _mm_add_ps(_mm_mul_ps(d, t1), t2);
_mm_storeu_ps(&M[12], t2);
_mm_store_ps(&M[12], t2);
#else
M[0] = m1[0]*m2[0] + m1[4]*m2[1] + m1[8]*m2[2] + m1[12]*m2[3];