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

Daily update on 20250312-074647 by Nitload Youmu. Additional Message : Change: Temp not want to compile RULE and PROCEDURE

Browse Source 
Add: Add factory method to generate more kind of binary operator node
Add: more SIMD intrinsic C function

Signed-off-by: Nitload <YoumuLiu_NSI@whv0t.onmicrosoft.com>
This commit is contained in:
Nitload
2025-03-12 07:46:48 +08:00
parent 16b813646d
commit 9647c2c68a
29 changed files with 645 additions and 244 deletions
Download Patch File Download Diff File Expand all files Collapse all files
StgSharp.C/mat2.c
+15
View File
@@ -40,3 +40,18 @@ INTERNAL void __cdecl addmatrix2_avx(__2_columnset *left, __2_columnset *right,
__m256 temp = _mm256_add_ps(ALIGN256(left->column[0]), ALIGN256(right->column[0]));
_mm256_storeu_ps(&ans->m[0][0], temp);
}
INTERNAL void __cdecl submatrix2_sse(__2_columnset *left, __2_columnset *right, __2_columnset *ans)
{
__m128 temp0 = _mm_sub_ps(ALIGN(left->column[0]), ALIGN(right->column[0]));
__m128 temp1 = _mm_sub_ps(ALIGN(left->column[1]), ALIGN(right->column[1]));
_mm_storeu_ps(&ans->m[0][0], temp0);
_mm_storeu_ps(&ans->m[0][1], temp1);
}
INTERNAL void __cdecl submatrix2_avx(__2_columnset *left, __2_columnset *right, __2_columnset *ans)
{
__m256 temp = _mm256_sub_ps(ALIGN256(left->column[0]), ALIGN256(right->column[0]));
_mm256_storeu_ps(&ans->m[0][0], temp);
}