youmuliu/stgsharp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

daily update on 20241226 by nitload youmu

Browse Source
This commit is contained in:
Nitload
2024-12-26 03:09:46 +08:00
parent 7c8d462ca5
commit 592ad55d57
94 changed files with 3767 additions and 1484 deletions
Download Patch File Download Diff File Expand all files Collapse all files
内核源代码/StgSharpC/mat3.c
+21 -37
View File
@@ -2,51 +2,38 @@
#include "ssc_intrinsic.h"
#include <stdio.h>
INTERNAL void _cdecl transpose32(__3_columnset *source,
__2_columnset *target)
INTERNAL void _cdecl transpose32(__3_columnset *source, __2_columnset *target)
{
__m128 t0 = _mm_shuffle_ps(ALIGN(source->column[0]),
ALIGN(source->column[1]),
_MM_SHUFFLE(1, 0, 1, 0));
__m128 t2 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec,
__m128 t0 = _mm_shuffle_ps(ALIGN(source->column[0]), ALIGN(source->column[1]),
_MM_SHUFFLE(1, 0, 1, 0));
__m128 t2 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec, _MM_SHUFFLE(1, 0, 1, 0));
target->column[0] = _mm_shuffle_ps(t0, t2, _MM_SHUFFLE(2, 0, 2, 0));
target->column[1] = _mm_shuffle_ps(t0, t2, _MM_SHUFFLE(3, 1, 3, 1));
}
INTERNAL void _cdecl transpose33(__3_columnset *source,
__3_columnset *target)
INTERNAL void _cdecl transpose33(__3_columnset *source, __3_columnset *target)
{
__m128 t0 = _mm_shuffle_ps(ALIGN(source->column[0]),
ALIGN(source->column[1]),
__m128 t0 = _mm_shuffle_ps(ALIGN(source->column[0]), ALIGN(source->column[1]),
_MM_SHUFFLE(1, 0, 1, 0));
__m128 t1 = _mm_shuffle_ps(ALIGN(source->column[0]),
ALIGN(source->column[1]),
_MM_SHUFFLE(3, 2, 3, 2));
__m128 t2 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec,
_MM_SHUFFLE(1, 0, 1, 0));
__m128 t3 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec,
__m128 t1 = _mm_shuffle_ps(ALIGN(source->column[0]), ALIGN(source->column[1]),
_MM_SHUFFLE(3, 2, 3, 2));
__m128 t2 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec, _MM_SHUFFLE(1, 0, 1, 0));
__m128 t3 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec, _MM_SHUFFLE(3, 2, 3, 2));
target->column[0] = _mm_shuffle_ps(t0, t2, _MM_SHUFFLE(2, 0, 2, 0));
target->column[1] = _mm_shuffle_ps(t0, t2, _MM_SHUFFLE(3, 1, 3, 1));
target->column[2] = _mm_shuffle_ps(t1, t3, _MM_SHUFFLE(2, 0, 2, 0));
}
INTERNAL void _cdecl transpose34(__3_columnset *source,
__4_columnset *target)
INTERNAL void _cdecl transpose34(__3_columnset *source, __4_columnset *target)
{
__m128 t0 = _mm_shuffle_ps(ALIGN(source->column[0]),
ALIGN(source->column[1]),
__m128 t0 = _mm_shuffle_ps(ALIGN(source->column[0]), ALIGN(source->column[1]),
_MM_SHUFFLE(1, 0, 1, 0));
__m128 t1 = _mm_shuffle_ps(ALIGN(source->column[0]),
ALIGN(source->column[1]),
_MM_SHUFFLE(3, 2, 3, 2));
__m128 t2 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec,
_MM_SHUFFLE(1, 0, 1, 0));
__m128 t3 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec,
__m128 t1 = _mm_shuffle_ps(ALIGN(source->column[0]), ALIGN(source->column[1]),
_MM_SHUFFLE(3, 2, 3, 2));
__m128 t2 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec, _MM_SHUFFLE(1, 0, 1, 0));
__m128 t3 = _mm_shuffle_ps(ALIGN(source->column[2]), zeroVec, _MM_SHUFFLE(3, 2, 3, 2));
target->column[0] = _mm_shuffle_ps(t0, t2, _MM_SHUFFLE(2, 0, 2, 0));
target->column[1] = _mm_shuffle_ps(t0, t2, _MM_SHUFFLE(3, 1, 3, 1));
@@ -54,9 +41,7 @@ INTERNAL void _cdecl transpose34(__3_columnset *source,
target->column[3] = _mm_shuffle_ps(t1, t3, _MM_SHUFFLE(3, 1, 3, 1));
}
INTERNAL float _cdecl det_mat3_internal(__m128 const c0,
__m128 const c1,
__m128 const c2)
INTERNAL float _cdecl det_mat3_internal(__m128 const c0, __m128 const c1, __m128 const c2)
{
//let adding row aligned
__m128 addc1 = _mm_shuffle_ps(c1, c1, _MM_SHUFFLE(3, 0, 2, 1));
@@ -83,16 +68,14 @@ INTERNAL float _cdecl det_mat3_internal(__m128 const c0,
return ret.m128_f32[2];
}
INTERNAL float _cdecl det_mat3(__3_columnset *source,
__3_columnset *transpose)
INTERNAL float _cdecl det_mat3(__3_columnset *source, __3_columnset *transpose)
{
return det_mat3_internal(source->column[0], source->column[1],
source->column[2]);
return det_mat3_internal(source->column[0], source->column[1], source->column[2]);
}
#pragma region add
INTERNAL void __cdecl addmatrix3_sse(__3_columnset* left, __3_columnset* right, __3_columnset* ans)
INTERNAL void __cdecl addmatrix3_sse(__3_columnset *left, __3_columnset *right, __3_columnset *ans)
{
__m128 temp0 = _mm_add_ps(ALIGN(left->column[0]), ALIGN(right->column[0]));
__m128 temp1 = _mm_add_ps(ALIGN(left->column[1]), ALIGN(right->column[1]));
@@ -103,7 +86,7 @@ INTERNAL void __cdecl addmatrix3_sse(__3_columnset* left, __3_columnset* right,
_mm_storeu_ps(&ans->m[0][2], temp2);
}
INTERNAL void __cdecl addmatrix3_sse(__3_columnset* left, __3_columnset* right, __3_columnset* ans)
INTERNAL void __cdecl addmatrix3_avx(__3_columnset *left, __3_columnset *right, __3_columnset *ans)
{
__m256 temp0 = _mm256_add_ps(ALIGN256(left->part), ALIGN256(right->part));
__m128 temp1 = _mm_add_ps(ALIGN(left->column[2]), ALIGN(right->column[2]));
@@ -112,9 +95,10 @@ INTERNAL void __cdecl addmatrix3_sse(__3_columnset* left, __3_columnset* right,
_mm_storeu_ps(&ans->m[0][2], temp1);
}
INTERNAL void __cdecl addmatrix3_sse(__3_columnset* left, __3_columnset* right, __3_columnset* ans)
INTERNAL void __cdecl addmatrix3_512(__3_columnset *left, __3_columnset *right, __3_columnset *ans)
{
__m512
__m512 temp = _mm512_add_ps(_mm512_loadu_ps(left), _mm512_loadu_ps(right));
_mm512_mask_store_ps(ans, 0x0fff, temp);
}
#pragma endregion