mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-06-11 11:45:30 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #12121 from maver1:amatyuko/sse2_convert_with_saturation_fix

Browse Source
This commit is contained in:
Vadim Pisarevsky 2018-08-06 14:26:37 +00:00
commit 23022f3ffb
2 changed files with 9 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
modules/core/include/opencv2/core/hal/intrin_sse.hpp
View File

@ -494,7 +494,12 @@ void v_rshr_pack_store(ushort* ptr, const v_uint32x4& a)
inline v_uint16x8 v_pack_u(const v_int32x4& a, const v_int32x4& b)
{
__m128i delta32 = _mm_set1_epi32(32768);
__m128i r = _mm_packs_epi32(_mm_sub_epi32(a.val, delta32), _mm_sub_epi32(b.val, delta32));
// preliminary saturate negative values to zero
__m128i a1 = _mm_and_si128(a.val, _mm_cmpgt_epi32(a.val, _mm_set1_epi32(0)));
__m128i b1 = _mm_and_si128(b.val, _mm_cmpgt_epi32(b.val, _mm_set1_epi32(0)));
__m128i r = _mm_packs_epi32(_mm_sub_epi32(a1, delta32), _mm_sub_epi32(b1, delta32));
return v_uint16x8(_mm_sub_epi16(r, _mm_set1_epi16(-32768)));
}

6
modules/core/src/convert.cpp
View File

@ -453,9 +453,9 @@ struct Cvt_SIMD<int, uchar>
{
v_int32x4 v_src1 = v_load(src + x), v_src2 = v_load(src + x + cWidth);
v_int32x4 v_src3 = v_load(src + x + cWidth * 2), v_src4 = v_load(src + x + cWidth * 3);
v_uint16x8 v_dst1 = v_pack_u(v_src1, v_src2);
v_uint16x8 v_dst2 = v_pack_u(v_src3, v_src4);
v_store(dst + x, v_pack(v_dst1, v_dst2));
v_int16x8 v_dst1 = v_pack(v_src1, v_src2);
v_int16x8 v_dst2 = v_pack(v_src3, v_src4);
v_store(dst + x, v_pack_u(v_dst1, v_dst2));
}
}
return x;