generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

used popcnt

Browse Source
This commit is contained in:
Ilya Lavrenov
2015-01-12 10:59:30 +03:00
parent 31827d8dfe
commit fc0869735d
25 changed files with 327 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
modules/core/include/opencv2/core/base.hpp
View File

@@ -13,6 +13,7 @@
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
// Copyright (C) 2014, Itseez Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,

16
modules/core/include/opencv2/core/cvdef.h
View File

@@ -13,6 +13,7 @@
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
// Copyright (C) 2015, Itseez Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,
@@ -157,15 +158,11 @@
# include <nmmintrin.h>
# define CV_SSE4_2 1
# endif
# if defined __FMA__ || (defined _MSC_VER && _MSC_VER >= 1500)
# include <immintrin.h>
# define CV_FMA3 1
# endif
# if defined __POPCNT__ || (defined _MSC_VER && _MSC_VER >= 1500)
# include <popcntintrin.h>
# define CV_POPCNT 1
# endif
# if defined __AVX__ || defined __AVX2__ || (defined _MSC_FULL_VER && _MSC_FULL_VER >= 160040219)
# if defined __AVX__ || (defined _MSC_FULL_VER && _MSC_FULL_VER >= 160040219)
// MS Visual Studio 2010 (2012?) has no macro pre-defined to identify the use of /arch:AVX
// See: http://connect.microsoft.com/VisualStudio/feedback/details/605858/arch-avx-should-define-a-predefined-macro-in-x64-and-set-a-unique-value-for-m-ix86-fp-in-win32
# include <immintrin.h>
@@ -179,6 +176,9 @@
# if defined __AVX2__ || (defined _MSC_FULL_VER && _MSC_FULL_VER >= 160040219)
# include <immintrin.h>
# define CV_AVX2 1
# if defined __FMA__
# define CV_FMA3 1
# endif
# endif
#endif
@@ -194,6 +194,9 @@
#endif // __CUDACC__
#ifndef CV_POPCNT
#define CV_POPCNT 0
#endif
#ifndef CV_MMX
# define CV_MMX 0
#endif
@@ -221,9 +224,6 @@
#ifndef CV_AVX2
# define CV_AVX2 0
#endif
#ifndef CV_POPCNT
#define CV_POPCNT 0
#endif
#ifndef CV_FMA3
# define CV_FMA3 0
#endif

77
modules/core/include/opencv2/core/sse_utils.hpp
View File

@@ -10,7 +10,7 @@
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2015, OpenCV Foundation, all rights reserved.
// Copyright (C) 2015, Itseez Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,
@@ -48,6 +48,34 @@
#if CV_SSE2
inline void _mm_deinterleave_epi8(__m128i & v_r0, __m128i & v_r1, __m128i & v_g0, __m128i & v_g1)
{
__m128i layer1_chunk0 = _mm_unpacklo_epi8(v_r0, v_g0);
__m128i layer1_chunk1 = _mm_unpackhi_epi8(v_r0, v_g0);
__m128i layer1_chunk2 = _mm_unpacklo_epi8(v_r1, v_g1);
__m128i layer1_chunk3 = _mm_unpackhi_epi8(v_r1, v_g1);
__m128i layer2_chunk0 = _mm_unpacklo_epi8(layer1_chunk0, layer1_chunk2);
__m128i layer2_chunk1 = _mm_unpackhi_epi8(layer1_chunk0, layer1_chunk2);
__m128i layer2_chunk2 = _mm_unpacklo_epi8(layer1_chunk1, layer1_chunk3);
__m128i layer2_chunk3 = _mm_unpackhi_epi8(layer1_chunk1, layer1_chunk3);
__m128i layer3_chunk0 = _mm_unpacklo_epi8(layer2_chunk0, layer2_chunk2);
__m128i layer3_chunk1 = _mm_unpackhi_epi8(layer2_chunk0, layer2_chunk2);
__m128i layer3_chunk2 = _mm_unpacklo_epi8(layer2_chunk1, layer2_chunk3);
__m128i layer3_chunk3 = _mm_unpackhi_epi8(layer2_chunk1, layer2_chunk3);
__m128i layer4_chunk0 = _mm_unpacklo_epi8(layer3_chunk0, layer3_chunk2);
__m128i layer4_chunk1 = _mm_unpackhi_epi8(layer3_chunk0, layer3_chunk2);
__m128i layer4_chunk2 = _mm_unpacklo_epi8(layer3_chunk1, layer3_chunk3);
__m128i layer4_chunk3 = _mm_unpackhi_epi8(layer3_chunk1, layer3_chunk3);
v_r0 = _mm_unpacklo_epi8(layer4_chunk0, layer4_chunk2);
v_r1 = _mm_unpackhi_epi8(layer4_chunk0, layer4_chunk2);
v_g0 = _mm_unpacklo_epi8(layer4_chunk1, layer4_chunk3);
v_g1 = _mm_unpackhi_epi8(layer4_chunk1, layer4_chunk3);
}
inline void _mm_deinterleave_epi8(__m128i & v_r0, __m128i & v_r1, __m128i & v_g0,
__m128i & v_g1, __m128i & v_b0, __m128i & v_b1)
{
@@ -228,6 +256,29 @@ inline void _mm_interleavee_epi8(__m128i & v_r0, __m128i & v_r1, __m128i & v_g0,
v_a1 = _mm_packus_epi16(_mm_srli_epi16(layer1_chunk6, 8), _mm_srli_epi16(layer1_chunk7, 8));
}
inline void _mm_deinterleave_epi16(__m128i & v_r0, __m128i & v_r1, __m128i & v_g0, __m128i & v_g1)
{
__m128i layer1_chunk0 = _mm_unpacklo_epi16(v_r0, v_g0);
__m128i layer1_chunk1 = _mm_unpackhi_epi16(v_r0, v_g0);
__m128i layer1_chunk2 = _mm_unpacklo_epi16(v_r1, v_g1);
__m128i layer1_chunk3 = _mm_unpackhi_epi16(v_r1, v_g1);
__m128i layer2_chunk0 = _mm_unpacklo_epi16(layer1_chunk0, layer1_chunk2);
__m128i layer2_chunk1 = _mm_unpackhi_epi16(layer1_chunk0, layer1_chunk2);
__m128i layer2_chunk2 = _mm_unpacklo_epi16(layer1_chunk1, layer1_chunk3);
__m128i layer2_chunk3 = _mm_unpackhi_epi16(layer1_chunk1, layer1_chunk3);
__m128i layer3_chunk0 = _mm_unpacklo_epi16(layer2_chunk0, layer2_chunk2);
__m128i layer3_chunk1 = _mm_unpackhi_epi16(layer2_chunk0, layer2_chunk2);
__m128i layer3_chunk2 = _mm_unpacklo_epi16(layer2_chunk1, layer2_chunk3);
__m128i layer3_chunk3 = _mm_unpackhi_epi16(layer2_chunk1, layer2_chunk3);
v_r0 = _mm_unpacklo_epi16(layer3_chunk0, layer3_chunk2);
v_r1 = _mm_unpackhi_epi16(layer3_chunk0, layer3_chunk2);
v_g0 = _mm_unpacklo_epi16(layer3_chunk1, layer3_chunk3);
v_g1 = _mm_unpackhi_epi16(layer3_chunk1, layer3_chunk3);
}
inline void _mm_deinterleave_epi16(__m128i & v_r0, __m128i & v_r1, __m128i & v_g0,
__m128i & v_g1, __m128i & v_b0, __m128i & v_b1)
{
@@ -300,6 +351,8 @@ inline void _mm_deinterleave_epi16(__m128i & v_r0, __m128i & v_r1, __m128i & v_g
v_a1 = _mm_unpackhi_epi16(layer3_chunk3, layer3_chunk7);
}
#if CV_SSE4_1
inline void _mm_interleave_epi16(__m128i & v_r0, __m128i & v_r1, __m128i & v_g0,
__m128i & v_g1, __m128i & v_b0, __m128i & v_b1)
{
@@ -376,6 +429,26 @@ inline void _mm_interleave_epi16(__m128i & v_r0, __m128i & v_r1, __m128i & v_g0,
v_a1 = _mm_packus_epi32(_mm_srli_epi32(layer1_chunk6, 16), _mm_srli_epi32(layer1_chunk7, 16));
}
#endif // CV_SSE4_1
inline void _mm_deinterleave_ps(__m128 & v_r0, __m128 & v_r1, __m128 & v_g0, __m128 & v_g1)
{
__m128 layer1_chunk0 = _mm_unpacklo_ps(v_r0, v_g0);
__m128 layer1_chunk1 = _mm_unpackhi_ps(v_r0, v_g0);
__m128 layer1_chunk2 = _mm_unpacklo_ps(v_r1, v_g1);
__m128 layer1_chunk3 = _mm_unpackhi_ps(v_r1, v_g1);
__m128 layer2_chunk0 = _mm_unpacklo_ps(layer1_chunk0, layer1_chunk2);
__m128 layer2_chunk1 = _mm_unpackhi_ps(layer1_chunk0, layer1_chunk2);
__m128 layer2_chunk2 = _mm_unpacklo_ps(layer1_chunk1, layer1_chunk3);
__m128 layer2_chunk3 = _mm_unpackhi_ps(layer1_chunk1, layer1_chunk3);
v_r0 = _mm_unpacklo_ps(layer2_chunk0, layer2_chunk2);
v_r1 = _mm_unpackhi_ps(layer2_chunk0, layer2_chunk2);
v_g0 = _mm_unpacklo_ps(layer2_chunk1, layer2_chunk3);
v_g1 = _mm_unpackhi_ps(layer2_chunk1, layer2_chunk3);
}
inline void _mm_deinterleave_ps(__m128 & v_r0, __m128 & v_r1, __m128 & v_g0,
__m128 & v_g1, __m128 & v_b0, __m128 & v_b1)
{
@@ -492,6 +565,6 @@ inline void _mm_interleave_ps(__m128 & v_r0, __m128 & v_r1, __m128 & v_g0, __m12
v_a1 = _mm_shuffle_ps(layer1_chunk6, layer1_chunk7, mask_hi);
}
#endif
#endif // CV_SSE2
#endif //__OPENCV_CORE_SSE_UTILS_HPP__

1
modules/core/include/opencv2/core/utility.hpp
View File

@@ -13,6 +13,7 @@
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
// Copyright (C) 2015, Itseez Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// Redistribution and use in source and binary forms, with or without modification,