Skip to content

O
Opencv

Greenplum / Opencv
11 个月 前同步成功

通知 7
Star 0
Fork 0
O
Opencv

体验新版 GitCode，发现更多精彩内容 >>

提交 940f1e79 编写于 作者: I Ilya Lavrenov
浏览文件
操作

interliving / deinterliving

上级 fe371bf6
隐藏空白更改
内联 并排
Showing with 90 addition and 0 deletion
modules/imgproc/src/color.cpp
浏览文件 @ 940f1e79
......@@ -183,6 +183,96 @@ static IppStatus sts = ippInit();
v_b1 = _mm_packus_epi16(_mm_srli_epi16(layer1_chunk4, 8), _mm_srli_epi16(layer1_chunk5, 8)); \
}
#define _MM_DEINTERLIV_EPI16(layer0_chunk0, layer0_chunk1, layer0_chunk2, \
layer0_chunk3, layer0_chunk4, layer0_chunk5) \
{ \
__m128i layer1_chunk0 = _mm_unpacklo_epi16(layer0_chunk0, layer0_chunk3); \
__m128i layer1_chunk1 = _mm_unpackhi_epi16(layer0_chunk0, layer0_chunk3); \
__m128i layer1_chunk2 = _mm_unpacklo_epi16(layer0_chunk1, layer0_chunk4); \
__m128i layer1_chunk3 = _mm_unpackhi_epi16(layer0_chunk1, layer0_chunk4); \
__m128i layer1_chunk4 = _mm_unpacklo_epi16(layer0_chunk2, layer0_chunk5); \
__m128i layer1_chunk5 = _mm_unpackhi_epi16(layer0_chunk2, layer0_chunk5); \
\
__m128i layer2_chunk0 = _mm_unpacklo_epi16(layer1_chunk0, layer1_chunk3); \
__m128i layer2_chunk1 = _mm_unpackhi_epi16(layer1_chunk0, layer1_chunk3); \
__m128i layer2_chunk2 = _mm_unpacklo_epi16(layer1_chunk1, layer1_chunk4); \
__m128i layer2_chunk3 = _mm_unpackhi_epi16(layer1_chunk1, layer1_chunk4); \
__m128i layer2_chunk4 = _mm_unpacklo_epi16(layer1_chunk2, layer1_chunk5); \
__m128i layer2_chunk5 = _mm_unpackhi_epi16(layer1_chunk2, layer1_chunk5); \
\
__m128i layer3_chunk0 = _mm_unpacklo_epi16(layer2_chunk0, layer2_chunk3); \
__m128i layer3_chunk1 = _mm_unpackhi_epi16(layer2_chunk0, layer2_chunk3); \
__m128i layer3_chunk2 = _mm_unpacklo_epi16(layer2_chunk1, layer2_chunk4); \
__m128i layer3_chunk3 = _mm_unpackhi_epi16(layer2_chunk1, layer2_chunk4); \
__m128i layer3_chunk4 = _mm_unpacklo_epi16(layer2_chunk2, layer2_chunk5); \
__m128i layer3_chunk5 = _mm_unpackhi_epi16(layer2_chunk2, layer2_chunk5); \
\
layer0_chunk0 = _mm_unpacklo_epi16(layer3_chunk0, layer3_chunk3); \
layer0_chunk1 = _mm_unpackhi_epi16(layer3_chunk0, layer3_chunk3); \
layer0_chunk2 = _mm_unpacklo_epi16(layer3_chunk1, layer3_chunk4); \
layer0_chunk3 = _mm_unpackhi_epi16(layer3_chunk1, layer3_chunk4); \
layer0_chunk4 = _mm_unpacklo_epi16(layer3_chunk2, layer3_chunk5); \
layer0_chunk5 = _mm_unpackhi_epi16(layer3_chunk2, layer3_chunk5); \
}
#define _MM_INTERLIV_EPI16(v_r0, v_r1, v_g0, v_g1, v_b0, v_b1) \
{ \
__m128i v_mask = _mm_set1_epi32(0x0000ffff); \
\
__m128i layer3_chunk0 = _mm_packus_epi32(_mm_and_si128(v_r0, v_mask), _mm_and_si128(v_r1, v_mask)); \
__m128i layer3_chunk3 = _mm_packus_epi32(_mm_srli_epi32(v_r0, 16), _mm_srli_epi32(v_r1, 16)); \
__m128i layer3_chunk1 = _mm_packus_epi32(_mm_and_si128(v_g0, v_mask), _mm_and_si128(v_g1, v_mask)); \
__m128i layer3_chunk4 = _mm_packus_epi32(_mm_srli_epi32(v_g0, 16), _mm_srli_epi32(v_g1, 16)); \
__m128i layer3_chunk2 = _mm_packus_epi32(_mm_and_si128(v_b0, v_mask), _mm_and_si128(v_b1, v_mask)); \
__m128i layer3_chunk5 = _mm_packus_epi32(_mm_srli_epi32(v_b0, 16), _mm_srli_epi32(v_b1, 16)); \
\
__m128i layer2_chunk0 = _mm_packus_epi32(_mm_and_si128(layer3_chunk0, v_mask), _mm_and_si128(layer3_chunk1, v_mask)); \
__m128i layer2_chunk3 = _mm_packus_epi32(_mm_srli_epi32(layer3_chunk0, 16), _mm_srli_epi32(layer3_chunk1, 16)); \
__m128i layer2_chunk1 = _mm_packus_epi32(_mm_and_si128(layer3_chunk2, v_mask), _mm_and_si128(layer3_chunk3, v_mask)); \
__m128i layer2_chunk4 = _mm_packus_epi32(_mm_srli_epi32(layer3_chunk2, 16), _mm_srli_epi32(layer3_chunk3, 16)); \
__m128i layer2_chunk2 = _mm_packus_epi32(_mm_and_si128(layer3_chunk4, v_mask), _mm_and_si128(layer3_chunk5, v_mask)); \
__m128i layer2_chunk5 = _mm_packus_epi32(_mm_srli_epi32(layer3_chunk4, 16), _mm_srli_epi32(layer3_chunk5, 16)); \
\
__m128i layer1_chunk0 = _mm_packus_epi32(_mm_and_si128(layer2_chunk0, v_mask), _mm_and_si128(layer2_chunk1, v_mask)); \
__m128i layer1_chunk3 = _mm_packus_epi32(_mm_srli_epi32(layer2_chunk0, 16), _mm_srli_epi32(layer2_chunk1, 16)); \
__m128i layer1_chunk1 = _mm_packus_epi32(_mm_and_si128(layer2_chunk2, v_mask), _mm_and_si128(layer2_chunk3, v_mask)); \
__m128i layer1_chunk4 = _mm_packus_epi32(_mm_srli_epi32(layer2_chunk2, 16), _mm_srli_epi32(layer2_chunk3, 16)); \
__m128i layer1_chunk2 = _mm_packus_epi32(_mm_and_si128(layer2_chunk4, v_mask), _mm_and_si128(layer2_chunk5, v_mask)); \
__m128i layer1_chunk5 = _mm_packus_epi32(_mm_srli_epi32(layer2_chunk4, 16), _mm_srli_epi32(layer2_chunk5, 16)); \
\
v_r0 = _mm_packus_epi32(_mm_and_si128(layer1_chunk0, v_mask), _mm_and_si128(layer1_chunk1, v_mask)); \
v_g1 = _mm_packus_epi32(_mm_srli_epi32(layer1_chunk0, 16), _mm_srli_epi32(layer1_chunk1, 16)); \
v_r1 = _mm_packus_epi32(_mm_and_si128(layer1_chunk2, v_mask), _mm_and_si128(layer1_chunk3, v_mask)); \
v_b0 = _mm_packus_epi32(_mm_srli_epi32(layer1_chunk2, 16), _mm_srli_epi32(layer1_chunk3, 16)); \
v_g0 = _mm_packus_epi32(_mm_and_si128(layer1_chunk4, v_mask), _mm_and_si128(layer1_chunk5, v_mask)); \
v_b1 = _mm_packus_epi32(_mm_srli_epi32(layer1_chunk4, 16), _mm_srli_epi32(layer1_chunk5, 16)); \
}
#define _MM_DEINTERLIV_PS(layer0_chunk0, layer0_chunk1, layer0_chunk2, \
layer0_chunk3, layer0_chunk4, layer0_chunk5) \
{ \
__m128 layer1_chunk0 = _mm_unpacklo_ps(layer0_chunk0, layer0_chunk3); \
__m128 layer1_chunk1 = _mm_unpackhi_ps(layer0_chunk0, layer0_chunk3); \
__m128 layer1_chunk2 = _mm_unpacklo_ps(layer0_chunk1, layer0_chunk4); \
__m128 layer1_chunk3 = _mm_unpackhi_ps(layer0_chunk1, layer0_chunk4); \
__m128 layer1_chunk4 = _mm_unpacklo_ps(layer0_chunk2, layer0_chunk5); \
__m128 layer1_chunk5 = _mm_unpackhi_ps(layer0_chunk2, layer0_chunk5); \
\
__m128 layer2_chunk0 = _mm_unpacklo_ps(layer1_chunk0, layer1_chunk3); \
__m128 layer2_chunk1 = _mm_unpackhi_ps(layer1_chunk0, layer1_chunk3); \
__m128 layer2_chunk2 = _mm_unpacklo_ps(layer1_chunk1, layer1_chunk4); \
__m128 layer2_chunk3 = _mm_unpackhi_ps(layer1_chunk1, layer1_chunk4); \
__m128 layer2_chunk4 = _mm_unpacklo_ps(layer1_chunk2, layer1_chunk5); \
__m128 layer2_chunk5 = _mm_unpackhi_ps(layer1_chunk2, layer1_chunk5); \
\
layer0_chunk0 = _mm_unpacklo_ps(layer2_chunk0, layer2_chunk3); \
layer0_chunk1 = _mm_unpackhi_ps(layer2_chunk0, layer2_chunk3); \
layer0_chunk2 = _mm_unpacklo_ps(layer2_chunk1, layer2_chunk4); \
layer0_chunk3 = _mm_unpackhi_ps(layer2_chunk1, layer2_chunk4); \
layer0_chunk4 = _mm_unpacklo_ps(layer2_chunk2, layer2_chunk5); \
layer0_chunk5 = _mm_unpackhi_ps(layer2_chunk2, layer2_chunk5); \
}
#endif
namespace cv
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或