mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-06-07 17:44:04 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #15025 from alalek:issue_14281

Browse Source
This commit is contained in:
Alexander Alekhin 2019-07-12 15:28:44 +00:00
commit 6aa07cdc7e
1 changed files with 4 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
modules/imgproc/src/filter.simd.hpp
View File

@ -84,6 +84,7 @@ Ptr<BaseFilter> getLinearFilter(
#ifndef CV_CPU_OPTIMIZATION_DECLARATIONS_ONLY
typedef int CV_DECL_ALIGNED(1) unaligned_int;
#define VEC_ALIGN CV_MALLOC_ALIGN
int FilterEngine__start(FilterEngine& this_, const Size &_wholeSize, const Size &sz, const Point &ofs)
@ -1049,7 +1050,7 @@ struct SymmColumnVec_32s8u
s0 = v_muladd(v_cvt_f32(v_load(src[k] + i) + v_load(src[-k] + i)), v_setall_f32(ky[k]), s0);
v_int32x4 s32 = v_round(s0);
v_int16x8 s16 = v_pack(s32, s32);
*(int*)(dst + i) = v_reinterpret_as_s32(v_pack_u(s16, s16)).get0();
*(unaligned_int*)(dst + i) = v_reinterpret_as_s32(v_pack_u(s16, s16)).get0();
i += v_int32x4::nlanes;
}
}
@ -1104,7 +1105,7 @@ struct SymmColumnVec_32s8u
s0 = v_muladd(v_cvt_f32(v_load(src[k] + i) - v_load(src[-k] + i)), v_setall_f32(ky[k]), s0);
v_int32x4 s32 = v_round(s0);
v_int16x8 s16 = v_pack(s32, s32);
*(int*)(dst + i) = v_reinterpret_as_s32(v_pack_u(s16, s16)).get0();
*(unaligned_int*)(dst + i) = v_reinterpret_as_s32(v_pack_u(s16, s16)).get0();
i += v_int32x4::nlanes;
}
}
@ -2129,7 +2130,7 @@ struct FilterVec_8u
s0 = v_muladd(v_cvt_f32(v_reinterpret_as_s32(v_load_expand_q(src[k] + i))), v_setall_f32(kf[k]), s0);
v_int32x4 s32 = v_round(s0);
v_int16x8 s16 = v_pack(s32, s32);
*(int*)(dst + i) = v_reinterpret_as_s32(v_pack_u(s16, s16)).get0();
*(unaligned_int*)(dst + i) = v_reinterpret_as_s32(v_pack_u(s16, s16)).get0();
i += v_int32x4::nlanes;
}
return i;