mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-06-07 09:25:45 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #3935 from vpisarev:extending_hal_part1

Browse Source
This commit is contained in:
Vadim Pisarevsky 2015-04-21 14:02:02 +00:00
commit 063e4004ba
30 changed files with 5964 additions and 2173 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
cmake/templates/opencv_abi.xml.in
View File

@ -21,6 +21,7 @@
</libs>
<skip_headers>
opencv2/hal/intrin*
opencv2/core/cuda*
opencv2/core/private*
opencv/cxeigen.hpp

2
modules/calib3d/test/test_fisheye.cpp
View File

@ -381,7 +381,7 @@ TEST_F(fisheyeTest, EtimateUncertainties)
EXPECT_MAT_NEAR(errors.c, cv::Vec2d(0.890439368129246, 0.816096854937896), 1e-10);
EXPECT_MAT_NEAR(errors.k, cv::Vec4d(0.00516248605191506, 0.0168181467500934, 0.0213118690274604, 0.00916010877545648), 1e-10);
EXPECT_MAT_NEAR(err_std, cv::Vec2d(0.187475975266883, 0.185678953263995), 1e-10);
CV_Assert(abs(rms - 0.263782587133546) < 1e-10);
CV_Assert(fabs(rms - 0.263782587133546) < 1e-10);
CV_Assert(errors.alpha == 0);
}

176
modules/core/include/opencv2/core/base.hpp
View File

@ -53,6 +53,7 @@
#include "opencv2/core/cvdef.h"
#include "opencv2/core/cvstd.hpp"
#include "opencv2/hal.hpp"
namespace cv
{
@ -400,136 +401,6 @@ configurations while CV_DbgAssert is only retained in the Debug configuration.
# define CV_DbgAssert(expr)
#endif
/////////////// saturate_cast (used in image & signal processing) ///////////////////
/**
Template function for accurate conversion from one primitive type to another.
The functions saturate_cast resemble the standard C++ cast operations, such as static_cast\<T\>()
and others. They perform an efficient and accurate conversion from one primitive type to another
(see the introduction chapter). saturate in the name means that when the input value v is out of the
range of the target type, the result is not formed just by taking low bits of the input, but instead
the value is clipped. For example:
@code
uchar a = saturate_cast<uchar>(-100); // a = 0 (UCHAR_MIN)
short b = saturate_cast<short>(33333.33333); // b = 32767 (SHRT_MAX)
@endcode
Such clipping is done when the target type is unsigned char , signed char , unsigned short or
signed short . For 32-bit integers, no clipping is done.
When the parameter is a floating-point value and the target type is an integer (8-, 16- or 32-bit),
the floating-point value is first rounded to the nearest integer and then clipped if needed (when
the target type is 8- or 16-bit).
This operation is used in the simplest or most complex image processing functions in OpenCV.
@param v Function parameter.
@sa add, subtract, multiply, divide, Mat::convertTo
*/
template<typename _Tp> static inline _Tp saturate_cast(uchar v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(schar v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(ushort v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(short v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(unsigned v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(int v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(float v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(double v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(int64 v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(uint64 v) { return _Tp(v); }
//! @cond IGNORED
template<> inline uchar saturate_cast<uchar>(schar v) { return (uchar)std::max((int)v, 0); }
template<> inline uchar saturate_cast<uchar>(ushort v) { return (uchar)std::min((unsigned)v, (unsigned)UCHAR_MAX); }
template<> inline uchar saturate_cast<uchar>(int v) { return (uchar)((unsigned)v <= UCHAR_MAX ? v : v > 0 ? UCHAR_MAX : 0); }
template<> inline uchar saturate_cast<uchar>(short v) { return saturate_cast<uchar>((int)v); }
template<> inline uchar saturate_cast<uchar>(unsigned v) { return (uchar)std::min(v, (unsigned)UCHAR_MAX); }
template<> inline uchar saturate_cast<uchar>(float v) { int iv = cvRound(v); return saturate_cast<uchar>(iv); }
template<> inline uchar saturate_cast<uchar>(double v) { int iv = cvRound(v); return saturate_cast<uchar>(iv); }
template<> inline uchar saturate_cast<uchar>(int64 v) { return (uchar)((uint64)v <= (uint64)UCHAR_MAX ? v : v > 0 ? UCHAR_MAX : 0); }
template<> inline uchar saturate_cast<uchar>(uint64 v) { return (uchar)std::min(v, (uint64)UCHAR_MAX); }
template<> inline schar saturate_cast<schar>(uchar v) { return (schar)std::min((int)v, SCHAR_MAX); }
template<> inline schar saturate_cast<schar>(ushort v) { return (schar)std::min((unsigned)v, (unsigned)SCHAR_MAX); }
template<> inline schar saturate_cast<schar>(int v) { return (schar)((unsigned)(v-SCHAR_MIN) <= (unsigned)UCHAR_MAX ? v : v > 0 ? SCHAR_MAX : SCHAR_MIN); }
template<> inline schar saturate_cast<schar>(short v) { return saturate_cast<schar>((int)v); }
template<> inline schar saturate_cast<schar>(unsigned v) { return (schar)std::min(v, (unsigned)SCHAR_MAX); }
template<> inline schar saturate_cast<schar>(float v) { int iv = cvRound(v); return saturate_cast<schar>(iv); }
template<> inline schar saturate_cast<schar>(double v) { int iv = cvRound(v); return saturate_cast<schar>(iv); }
template<> inline schar saturate_cast<schar>(int64 v) { return (schar)((uint64)((int64)v-SCHAR_MIN) <= (uint64)UCHAR_MAX ? v : v > 0 ? SCHAR_MAX : SCHAR_MIN); }
template<> inline schar saturate_cast<schar>(uint64 v) { return (schar)std::min(v, (uint64)SCHAR_MAX); }
template<> inline ushort saturate_cast<ushort>(schar v) { return (ushort)std::max((int)v, 0); }
template<> inline ushort saturate_cast<ushort>(short v) { return (ushort)std::max((int)v, 0); }
template<> inline ushort saturate_cast<ushort>(int v) { return (ushort)((unsigned)v <= (unsigned)USHRT_MAX ? v : v > 0 ? USHRT_MAX : 0); }
template<> inline ushort saturate_cast<ushort>(unsigned v) { return (ushort)std::min(v, (unsigned)USHRT_MAX); }
template<> inline ushort saturate_cast<ushort>(float v) { int iv = cvRound(v); return saturate_cast<ushort>(iv); }
template<> inline ushort saturate_cast<ushort>(double v) { int iv = cvRound(v); return saturate_cast<ushort>(iv); }
template<> inline ushort saturate_cast<ushort>(int64 v) { return (ushort)((uint64)v <= (uint64)USHRT_MAX ? v : v > 0 ? USHRT_MAX : 0); }
template<> inline ushort saturate_cast<ushort>(uint64 v) { return (ushort)std::min(v, (uint64)USHRT_MAX); }
template<> inline short saturate_cast<short>(ushort v) { return (short)std::min((int)v, SHRT_MAX); }
template<> inline short saturate_cast<short>(int v) { return (short)((unsigned)(v - SHRT_MIN) <= (unsigned)USHRT_MAX ? v : v > 0 ? SHRT_MAX : SHRT_MIN); }
template<> inline short saturate_cast<short>(unsigned v) { return (short)std::min(v, (unsigned)SHRT_MAX); }
template<> inline short saturate_cast<short>(float v) { int iv = cvRound(v); return saturate_cast<short>(iv); }
template<> inline short saturate_cast<short>(double v) { int iv = cvRound(v); return saturate_cast<short>(iv); }
template<> inline short saturate_cast<short>(int64 v) { return (short)((uint64)((int64)v - SHRT_MIN) <= (uint64)USHRT_MAX ? v : v > 0 ? SHRT_MAX : SHRT_MIN); }
template<> inline short saturate_cast<short>(uint64 v) { return (short)std::min(v, (uint64)SHRT_MAX); }
template<> inline int saturate_cast<int>(float v) { return cvRound(v); }
template<> inline int saturate_cast<int>(double v) { return cvRound(v); }
// we intentionally do not clip negative numbers, to make -1 become 0xffffffff etc.
template<> inline unsigned saturate_cast<unsigned>(float v) { return cvRound(v); }
template<> inline unsigned saturate_cast<unsigned>(double v) { return cvRound(v); }
//! @endcond
//////////////////////////////// low-level functions ////////////////////////////////
CV_EXPORTS int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n);
CV_EXPORTS int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n);
CV_EXPORTS bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n);
CV_EXPORTS bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n);
CV_EXPORTS int normL1_(const uchar* a, const uchar* b, int n);
CV_EXPORTS float normL1_(const float* a, const float* b, int n);
CV_EXPORTS float normL2Sqr_(const float* a, const float* b, int n);
CV_EXPORTS void exp(const float* src, float* dst, int n);
CV_EXPORTS void log(const float* src, float* dst, int n);
CV_EXPORTS void fastAtan2(const float* y, const float* x, float* dst, int n, bool angleInDegrees);
CV_EXPORTS void magnitude(const float* x, const float* y, float* dst, int n);
/** @brief Computes the cube root of an argument.
The function cubeRoot computes \f$\sqrt[3]{\texttt{val}}\f$. Negative arguments are handled correctly.
NaN and Inf are not handled. The accuracy approaches the maximum possible accuracy for
single-precision data.
@param val A function argument.
*/
CV_EXPORTS_W float cubeRoot(float val);
/** @brief Calculates the angle of a 2D vector in degrees.
The function fastAtan2 calculates the full-range angle of an input 2D vector. The angle is measured
in degrees and varies from 0 to 360 degrees. The accuracy is about 0.3 degrees.
@param x x-coordinate of the vector.
@param y y-coordinate of the vector.
*/
CV_EXPORTS_W float fastAtan2(float y, float x);
/*
* Hamming distance functor - counts the bit differences between two strings - useful for the Brief descriptor
* bit count of A exclusive XOR'ed with B
@ -549,6 +420,11 @@ typedef Hamming HammingLUT;
/////////////////////////////////// inline norms ////////////////////////////////////
template<typename _Tp> inline _Tp cv_abs(_Tp x) { return std::abs(x); }
inline int cv_abs(uchar x) { return x; }
inline int cv_abs(schar x) { return std::abs(x); }
inline int cv_abs(ushort x) { return x; }
inline int cv_abs(short x) { return std::abs(x); }
template<typename _Tp, typename _AccTp> static inline
_AccTp normL2Sqr(const _Tp* a, int n)
@ -578,12 +454,12 @@ _AccTp normL1(const _Tp* a, int n)
#if CV_ENABLE_UNROLLED
for(; i <= n - 4; i += 4 )
{
s += (_AccTp)std::abs(a[i]) + (_AccTp)std::abs(a[i+1]) +
(_AccTp)std::abs(a[i+2]) + (_AccTp)std::abs(a[i+3]);
s += (_AccTp)cv_abs(a[i]) + (_AccTp)cv_abs(a[i+1]) +
(_AccTp)cv_abs(a[i+2]) + (_AccTp)cv_abs(a[i+3]);
}
#endif
for( ; i < n; i++ )
s += std::abs(a[i]);
s += cv_abs(a[i]);
return s;
}
@ -592,7 +468,7 @@ _AccTp normInf(const _Tp* a, int n)
{
_AccTp s = 0;
for( int i = 0; i < n; i++ )
s = std::max(s, (_AccTp)std::abs(a[i]));
s = std::max(s, (_AccTp)cv_abs(a[i]));
return s;
}
@ -616,11 +492,10 @@ _AccTp normL2Sqr(const _Tp* a, const _Tp* b, int n)
return s;
}
template<> inline
float normL2Sqr(const float* a, const float* b, int n)
inline float normL2Sqr(const float* a, const float* b, int n)
{
if( n >= 8 )
return normL2Sqr_(a, b, n);
return hal::normL2Sqr_(a, b, n);
float s = 0;
for( int i = 0; i < n; i++ )
{
@ -650,11 +525,10 @@ _AccTp normL1(const _Tp* a, const _Tp* b, int n)
return s;
}
template<> inline
float normL1(const float* a, const float* b, int n)
inline float normL1(const float* a, const float* b, int n)
{
if( n >= 8 )
return normL1_(a, b, n);
return hal::normL1_(a, b, n);
float s = 0;
for( int i = 0; i < n; i++ )
{
@ -664,10 +538,9 @@ float normL1(const float* a, const float* b, int n)
return s;
}
template<> inline
int normL1(const uchar* a, const uchar* b, int n)
inline int normL1(const uchar* a, const uchar* b, int n)
{
return normL1_(a, b, n);
return hal::normL1_(a, b, n);
}
template<typename _Tp, typename _AccTp> static inline
@ -682,6 +555,23 @@ _AccTp normInf(const _Tp* a, const _Tp* b, int n)
return s;
}
/** @brief Computes the cube root of an argument.
The function cubeRoot computes \f$\sqrt[3]{\texttt{val}}\f$. Negative arguments are handled correctly.
NaN and Inf are not handled. The accuracy approaches the maximum possible accuracy for
single-precision data.
@param val A function argument.
*/
CV_EXPORTS_W float cubeRoot(float val);
/** @brief Calculates the angle of a 2D vector in degrees.
The function fastAtan2 calculates the full-range angle of an input 2D vector. The angle is measured
in degrees and varies from 0 to 360 degrees. The accuracy is about 0.3 degrees.
@param x x-coordinate of the vector.
@param y y-coordinate of the vector.
*/
CV_EXPORTS_W float fastAtan2(float y, float x);
////////////////// forward declarations for important OpenCV types //////////////////

181
modules/core/include/opencv2/core/cvdef.h
View File

@ -70,16 +70,6 @@
# define CV_EXPORTS
#endif
#ifndef CV_INLINE
# if defined __cplusplus
# define CV_INLINE static inline
# elif defined _MSC_VER
# define CV_INLINE __inline
# else
# define CV_INLINE static
# endif
#endif
#ifndef CV_EXTERN_C
# ifdef __cplusplus
# define CV_EXTERN_C extern "C"
@ -186,19 +176,6 @@
#define CV_ELEM_SIZE(type) \
(CV_MAT_CN(type) << ((((sizeof(size_t)/4+1)*16384|0x3a50) >> CV_MAT_DEPTH(type)*2) & 3))
/****************************************************************************************\
* fast math *
\****************************************************************************************/
#if defined __BORLANDC__
# include <fastmath.h>
#elif defined __cplusplus
# include <cmath>
#else
# include <math.h>
#endif
#ifndef MIN
# define MIN(a,b) ((a) > (b) ? (b) : (a))
#endif
@ -207,164 +184,6 @@
# define MAX(a,b) ((a) < (b) ? (b) : (a))
#endif
#ifdef HAVE_TEGRA_OPTIMIZATION
# include "tegra_round.hpp"
#endif
//! @addtogroup core_utils
//! @{
#if CV_VFP
// 1. general scheme
#define ARM_ROUND(_value, _asm_string) \
int res; \
float temp; \
asm(_asm_string : [res] "=r" (res), [temp] "=w" (temp) : [value] "w" (_value)); \
return res;
// 2. version for double
#ifdef __clang__
#define ARM_ROUND_DBL(value) ARM_ROUND(value, "vcvtr.s32.f64 %[temp], %[value] \n vmov %[res], %[temp]")
#else
#define ARM_ROUND_DBL(value) ARM_ROUND(value, "vcvtr.s32.f64 %[temp], %P[value] \n vmov %[res], %[temp]")
#endif
// 3. version for float
#define ARM_ROUND_FLT(value) ARM_ROUND(value, "vcvtr.s32.f32 %[temp], %[value]\n vmov %[res], %[temp]")
#endif // CV_VFP
/** @brief Rounds floating-point number to the nearest integer
@param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
result is not defined.
*/
CV_INLINE int cvRound( double value )
{
#if ((defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__ && defined __SSE2__ && !defined __APPLE__)) && !defined(__CUDACC__)
__m128d t = _mm_set_sd( value );
return _mm_cvtsd_si32(t);
#elif defined _MSC_VER && defined _M_IX86
int t;
__asm
{
fld value;
fistp t;
}
return t;
#elif ((defined _MSC_VER && defined _M_ARM) || defined CV_ICC || defined __GNUC__) && defined HAVE_TEGRA_OPTIMIZATION
TEGRA_ROUND_DBL(value);
#elif defined CV_ICC || defined __GNUC__
# if CV_VFP
ARM_ROUND_DBL(value)
# else
return (int)lrint(value);
# endif
#else
double intpart, fractpart;
fractpart = modf(value, &intpart);
if ((fabs(fractpart) != 0.5) || ((((int)intpart) % 2) != 0))
return (int)(value + (value >= 0 ? 0.5 : -0.5));
else
return (int)intpart;
#endif
}
#ifdef __cplusplus
/** @overload */
CV_INLINE int cvRound(float value)
{
#if defined ANDROID && (defined CV_ICC || defined __GNUC__) && defined HAVE_TEGRA_OPTIMIZATION
TEGRA_ROUND_FLT(value);
#elif CV_VFP && !defined HAVE_TEGRA_OPTIMIZATION
ARM_ROUND_FLT(value)
#else
return cvRound((double)value);
#endif
}
/** @overload */
CV_INLINE int cvRound(int value)
{
return value;
}
#endif // __cplusplus
/** @brief Rounds floating-point number to the nearest integer not larger than the original.
The function computes an integer i such that:
\f[i \le \texttt{value} < i+1\f]
@param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
result is not defined.
*/
CV_INLINE int cvFloor( double value )
{
#if (defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__ && !defined __APPLE__)) && !defined(__CUDACC__)
__m128d t = _mm_set_sd( value );
int i = _mm_cvtsd_si32(t);
return i - _mm_movemask_pd(_mm_cmplt_sd(t, _mm_cvtsi32_sd(t,i)));
#elif defined __GNUC__
int i = (int)value;
return i - (i > value);
#else
int i = cvRound(value);
float diff = (float)(value - i);
return i - (diff < 0);
#endif
}
/** @brief Rounds floating-point number to the nearest integer not larger than the original.
The function computes an integer i such that:
\f[i \le \texttt{value} < i+1\f]
@param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
result is not defined.
*/
CV_INLINE int cvCeil( double value )
{
#if (defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__&& !defined __APPLE__)) && !defined(__CUDACC__)
__m128d t = _mm_set_sd( value );
int i = _mm_cvtsd_si32(t);
return i + _mm_movemask_pd(_mm_cmplt_sd(_mm_cvtsi32_sd(t,i), t));
#elif defined __GNUC__
int i = (int)value;
return i + (i < value);
#else
int i = cvRound(value);
float diff = (float)(i - value);
return i + (diff < 0);
#endif
}
/** @brief Determines if the argument is Not A Number.
@param value The input floating-point value
The function returns 1 if the argument is Not A Number (as defined by IEEE754 standard), 0
otherwise. */
CV_INLINE int cvIsNaN( double value )
{
union { uint64 u; double f; } ieee754;
ieee754.f = value;
return ((unsigned)(ieee754.u >> 32) & 0x7fffffff) +
((unsigned)ieee754.u != 0) > 0x7ff00000;
}
/** @brief Determines if the argument is Infinity.
@param value The input floating-point value
The function returns 1 if the argument is a plus or minus infinity (as defined by IEEE754 standard)
and 0 otherwise. */
CV_INLINE int cvIsInf( double value )
{
union { uint64 u; double f; } ieee754;
ieee754.f = value;
return ((unsigned)(ieee754.u >> 32) & 0x7fffffff) == 0x7ff00000 &&
(unsigned)ieee754.u == 0;
}
//! @} core_utils
/****************************************************************************************\
* exchange-add operation for atomic operations on reference counters *
\****************************************************************************************/

2
modules/core/include/opencv2/core/matx.hpp
View File

@ -427,7 +427,7 @@ template<typename _Tp, int m> struct Matx_DetOp
double operator ()(const Matx<_Tp, m, m>& a) const
{
Matx<_Tp, m, m> temp = a;
double p = LU(temp.val, m*sizeof(_Tp), m, 0, 0, 0);
double p = hal::LU(temp.val, m*sizeof(_Tp), m, 0, 0, 0);
if( p == 0 )
return p;
for( int i = 0; i < m; i++ )

4
modules/core/include/opencv2/core/operations.hpp
View File

@ -72,9 +72,9 @@ template<typename _Tp, int m> struct Matx_FastInvOp
b(i, i) = (_Tp)1;
if( method == DECOMP_CHOLESKY )
return Cholesky(temp.val, m*sizeof(_Tp), m, b.val, m*sizeof(_Tp), m);
return hal::Cholesky(temp.val, m*sizeof(_Tp), m, b.val, m*sizeof(_Tp), m);
return LU(temp.val, m*sizeof(_Tp), m, b.val, m*sizeof(_Tp), m) != 0;
return hal::LU(temp.val, m*sizeof(_Tp), m, b.val, m*sizeof(_Tp), m) != 0;
}
};

8
modules/core/include/opencv2/core/private.hpp
View File

@ -136,14 +136,6 @@ namespace cv
/* the alignment of all the allocated buffers */
#define CV_MALLOC_ALIGN 16
#ifdef __GNUC__
# define CV_DECL_ALIGNED(x) __attribute__ ((aligned (x)))
#elif defined _MSC_VER
# define CV_DECL_ALIGNED(x) __declspec(align(x))
#else
# define CV_DECL_ALIGNED(x)
#endif
/* IEEE754 constants and macros */
#define CV_TOGGLE_FLT(x) ((x)^((int)(x) < 0 ? 0x7fffffff : 0))
#define CV_TOGGLE_DBL(x) ((x)^((int64)(x) < 0 ? CV_BIG_INT(0x7fffffffffffffff) : 0))

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

@ -113,22 +113,6 @@ bytes of the header. In C++ interface the role of CvArr is played by InputArray
*/
typedef void CvArr;
typedef union Cv32suf
{
int i;
unsigned u;
float f;
}
Cv32suf;
typedef union Cv64suf
{
int64 i;
uint64 u;
double f;
}
Cv64suf;
typedef int CVStatus;
/** @see cv::Error::Code */

8
modules/core/src/kmeans.cpp
View File

@ -79,7 +79,7 @@ public:
for ( int i = begin; i<end; i++ )
{
tdist2[i] = std::min(normL2Sqr_(data + step*i, data + stepci, dims), dist[i]);
tdist2[i] = std::min(normL2Sqr(data + step*i, data + stepci, dims), dist[i]);
}
}
@ -114,7 +114,7 @@ static void generateCentersPP(const Mat& _data, Mat& _out_centers,
for( i = 0; i < N; i++ )
{
dist[i] = normL2Sqr_(data + step*i, data + step*centers[0], dims);
dist[i] = normL2Sqr(data + step*i, data + step*centers[0], dims);
sum0 += dist[i];
}
@ -189,7 +189,7 @@ public:
for( int k = 0; k < K; k++ )
{
const float* center = centers.ptr<float>(k);
const double dist = normL2Sqr_(sample, center, dims);
const double dist = normL2Sqr(sample, center, dims);
if( min_dist > dist )
{
@ -384,7 +384,7 @@ double cv::kmeans( InputArray _data, int K,
if( labels[i] != max_k )
continue;
sample = data.ptr<float>(i);
double dist = normL2Sqr_(sample, _old_center, dims);
double dist = normL2Sqr(sample, _old_center, dims);
if( max_dist <= dist )
{

182
modules/core/src/lapack.cpp
View File

@ -50,168 +50,6 @@
namespace cv
{
/****************************************************************************************\
* LU & Cholesky implementation for small matrices *
\****************************************************************************************/
template<typename _Tp> static inline int
LUImpl(_Tp* A, size_t astep, int m, _Tp* b, size_t bstep, int n)
{
int i, j, k, p = 1;
astep /= sizeof(A[0]);
bstep /= sizeof(b[0]);
for( i = 0; i < m; i++ )
{
k = i;
for( j = i+1; j < m; j++ )
if( std::abs(A[j*astep + i]) > std::abs(A[k*astep + i]) )
k = j;
if( std::abs(A[k*astep + i]) < std::numeric_limits<_Tp>::epsilon() )
return 0;
if( k != i )
{
for( j = i; j < m; j++ )
std::swap(A[i*astep + j], A[k*astep + j]);
if( b )
for( j = 0; j < n; j++ )
std::swap(b[i*bstep + j], b[k*bstep + j]);
p = -p;
}
_Tp d = -1/A[i*astep + i];
for( j = i+1; j < m; j++ )
{
_Tp alpha = A[j*astep + i]*d;
for( k = i+1; k < m; k++ )
A[j*astep + k] += alpha*A[i*astep + k];
if( b )
for( k = 0; k < n; k++ )
b[j*bstep + k] += alpha*b[i*bstep + k];
}
A[i*astep + i] = -d;
}
if( b )
{
for( i = m-1; i >= 0; i-- )
for( j = 0; j < n; j++ )
{
_Tp s = b[i*bstep + j];
for( k = i+1; k < m; k++ )
s -= A[i*astep + k]*b[k*bstep + j];
b[i*bstep + j] = s*A[i*astep + i];
}
}
return p;
}
int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n)
{
return LUImpl(A, astep, m, b, bstep, n);
}
int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n)
{
return LUImpl(A, astep, m, b, bstep, n);
}
template<typename _Tp> static inline bool
CholImpl(_Tp* A, size_t astep, int m, _Tp* b, size_t bstep, int n)
{
_Tp* L = A;
int i, j, k;
double s;
astep /= sizeof(A[0]);
bstep /= sizeof(b[0]);
for( i = 0; i < m; i++ )
{
for( j = 0; j < i; j++ )
{
s = A[i*astep + j];
for( k = 0; k < j; k++ )
s -= L[i*astep + k]*L[j*astep + k];
L[i*astep + j] = (_Tp)(s*L[j*astep + j]);
}
s = A[i*astep + i];
for( k = 0; k < j; k++ )
{
double t = L[i*astep + k];
s -= t*t;
}
if( s < std::numeric_limits<_Tp>::epsilon() )
return false;
L[i*astep + i] = (_Tp)(1./std::sqrt(s));
}
if( !b )
return true;
// LLt x = b
// 1: L y = b
// 2. Lt x = y
/*
[ L00 ] y0 b0
[ L10 L11 ] y1 = b1
[ L20 L21 L22 ] y2 b2
[ L30 L31 L32 L33 ] y3 b3
[ L00 L10 L20 L30 ] x0 y0
[ L11 L21 L31 ] x1 = y1
[ L22 L32 ] x2 y2
[ L33 ] x3 y3
*/
for( i = 0; i < m; i++ )
{
for( j = 0; j < n; j++ )
{
s = b[i*bstep + j];
for( k = 0; k < i; k++ )
s -= L[i*astep + k]*b[k*bstep + j];
b[i*bstep + j] = (_Tp)(s*L[i*astep + i]);
}
}
for( i = m-1; i >= 0; i-- )
{
for( j = 0; j < n; j++ )
{
s = b[i*bstep + j];
for( k = m-1; k > i; k-- )
s -= L[k*astep + i]*b[k*bstep + j];
b[i*bstep + j] = (_Tp)(s*L[i*astep + i]);
}
}
return true;
}
bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n)
{
return CholImpl(A, astep, m, b, bstep, n);
}
bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n)
{
return CholImpl(A, astep, m, b, bstep, n);
}
template<typename _Tp> static inline _Tp hypot(_Tp a, _Tp b)
{
a = std::abs(a);
@ -882,7 +720,7 @@ double cv::determinant( InputArray _mat )
Mat a(rows, rows, CV_32F, (uchar*)buffer);
mat.copyTo(a);
result = LU(a.ptr<float>(), a.step, rows, 0, 0, 0);
result = hal::LU(a.ptr<float>(), a.step, rows, 0, 0, 0);
if( result )
{
for( int i = 0; i < rows; i++ )
@ -906,7 +744,7 @@ double cv::determinant( InputArray _mat )
Mat a(rows, rows, CV_64F, (uchar*)buffer);
mat.copyTo(a);
result = LU(a.ptr<double>(), a.step, rows, 0, 0, 0);
result = hal::LU(a.ptr<double>(), a.step, rows, 0, 0, 0);
if( result )
{
for( int i = 0; i < rows; i++ )
@ -1169,13 +1007,13 @@ double cv::invert( InputArray _src, OutputArray _dst, int method )
setIdentity(dst);
if( method == DECOMP_LU && type == CV_32F )
result = LU(src1.ptr<float>(), src1.step, n, dst.ptr<float>(), dst.step, n) != 0;
result = hal::LU(src1.ptr<float>(), src1.step, n, dst.ptr<float>(), dst.step, n) != 0;
else if( method == DECOMP_LU && type == CV_64F )
result = LU(src1.ptr<double>(), src1.step, n, dst.ptr<double>(), dst.step, n) != 0;
result = hal::LU(src1.ptr<double>(), src1.step, n, dst.ptr<double>(), dst.step, n) != 0;
else if( method == DECOMP_CHOLESKY && type == CV_32F )
result = Cholesky(src1.ptr<float>(), src1.step, n, dst.ptr<float>(), dst.step, n);
result = hal::Cholesky(src1.ptr<float>(), src1.step, n, dst.ptr<float>(), dst.step, n);
else
result = Cholesky(src1.ptr<double>(), src1.step, n, dst.ptr<double>(), dst.step, n);
result = hal::Cholesky(src1.ptr<double>(), src1.step, n, dst.ptr<double>(), dst.step, n);
if( !result )
dst = Scalar(0);
@ -1407,16 +1245,16 @@ bool cv::solve( InputArray _src, InputArray _src2arg, OutputArray _dst, int meth
if( method == DECOMP_LU )
{
if( type == CV_32F )
result = LU(a.ptr<float>(), a.step, n, dst.ptr<float>(), dst.step, nb) != 0;
result = hal::LU(a.ptr<float>(), a.step, n, dst.ptr<float>(), dst.step, nb) != 0;
else
result = LU(a.ptr<double>(), a.step, n, dst.ptr<double>(), dst.step, nb) != 0;
result = hal::LU(a.ptr<double>(), a.step, n, dst.ptr<double>(), dst.step, nb) != 0;
}
else if( method == DECOMP_CHOLESKY )
{
if( type == CV_32F )
result = Cholesky(a.ptr<float>(), a.step, n, dst.ptr<float>(), dst.step, nb);
result = hal::Cholesky(a.ptr<float>(), a.step, n, dst.ptr<float>(), dst.step, nb);
else
result = Cholesky(a.ptr<double>(), a.step, n, dst.ptr<double>(), dst.step, nb);
result = hal::Cholesky(a.ptr<double>(), a.step, n, dst.ptr<double>(), dst.step, nb);
}
else
{

1489
modules/core/src/mathfuncs.cpp
View File

File diff suppressed because it is too large Load Diff

166
modules/core/src/stat.cpp
View File

@ -2416,140 +2416,6 @@ void cv::minMaxLoc( InputArray _img, double* minVal, double* maxVal,
namespace cv
{
float normL2Sqr_(const float* a, const float* b, int n)
{
int j = 0; float d = 0.f;
#if CV_SSE
if( USE_SSE2 )
{
float CV_DECL_ALIGNED(16) buf[4];
__m128 d0 = _mm_setzero_ps(), d1 = _mm_setzero_ps();
for( ; j <= n - 8; j += 8 )
{
__m128 t0 = _mm_sub_ps(_mm_loadu_ps(a + j), _mm_loadu_ps(b + j));
__m128 t1 = _mm_sub_ps(_mm_loadu_ps(a + j + 4), _mm_loadu_ps(b + j + 4));
d0 = _mm_add_ps(d0, _mm_mul_ps(t0, t0));
d1 = _mm_add_ps(d1, _mm_mul_ps(t1, t1));
}
_mm_store_ps(buf, _mm_add_ps(d0, d1));
d = buf[0] + buf[1] + buf[2] + buf[3];
}
else
#endif
{
for( ; j <= n - 4; j += 4 )
{
float t0 = a[j] - b[j], t1 = a[j+1] - b[j+1], t2 = a[j+2] - b[j+2], t3 = a[j+3] - b[j+3];
d += t0*t0 + t1*t1 + t2*t2 + t3*t3;
}
}
for( ; j < n; j++ )
{
float t = a[j] - b[j];
d += t*t;
}
return d;
}
float normL1_(const float* a, const float* b, int n)
{
int j = 0; float d = 0.f;
#if CV_SSE
if( USE_SSE2 )
{
float CV_DECL_ALIGNED(16) buf[4];
static const int CV_DECL_ALIGNED(16) absbuf[4] = {0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff};
__m128 d0 = _mm_setzero_ps(), d1 = _mm_setzero_ps();
__m128 absmask = _mm_load_ps((const float*)absbuf);
for( ; j <= n - 8; j += 8 )
{
__m128 t0 = _mm_sub_ps(_mm_loadu_ps(a + j), _mm_loadu_ps(b + j));
__m128 t1 = _mm_sub_ps(_mm_loadu_ps(a + j + 4), _mm_loadu_ps(b + j + 4));
d0 = _mm_add_ps(d0, _mm_and_ps(t0, absmask));
d1 = _mm_add_ps(d1, _mm_and_ps(t1, absmask));
}
_mm_store_ps(buf, _mm_add_ps(d0, d1));
d = buf[0] + buf[1] + buf[2] + buf[3];
}
else
#elif CV_NEON
float32x4_t v_sum = vdupq_n_f32(0.0f);
for ( ; j <= n - 4; j += 4)
v_sum = vaddq_f32(v_sum, vabdq_f32(vld1q_f32(a + j), vld1q_f32(b + j)));
float CV_DECL_ALIGNED(16) buf[4];
vst1q_f32(buf, v_sum);
d = buf[0] + buf[1] + buf[2] + buf[3];
#endif
{
for( ; j <= n - 4; j += 4 )
{
d += std::abs(a[j] - b[j]) + std::abs(a[j+1] - b[j+1]) +
std::abs(a[j+2] - b[j+2]) + std::abs(a[j+3] - b[j+3]);
}
}
for( ; j < n; j++ )
d += std::abs(a[j] - b[j]);
return d;
}
int normL1_(const uchar* a, const uchar* b, int n)
{
int j = 0, d = 0;
#if CV_SSE
if( USE_SSE2 )
{
__m128i d0 = _mm_setzero_si128();
for( ; j <= n - 16; j += 16 )
{
__m128i t0 = _mm_loadu_si128((const __m128i*)(a + j));
__m128i t1 = _mm_loadu_si128((const __m128i*)(b + j));
d0 = _mm_add_epi32(d0, _mm_sad_epu8(t0, t1));
}
for( ; j <= n - 4; j += 4 )
{
__m128i t0 = _mm_cvtsi32_si128(*(const int*)(a + j));
__m128i t1 = _mm_cvtsi32_si128(*(const int*)(b + j));
d0 = _mm_add_epi32(d0, _mm_sad_epu8(t0, t1));
}
d = _mm_cvtsi128_si32(_mm_add_epi32(d0, _mm_unpackhi_epi64(d0, d0)));
}
else
#elif CV_NEON
uint32x4_t v_sum = vdupq_n_u32(0.0f);
for ( ; j <= n - 16; j += 16)
{
uint8x16_t v_dst = vabdq_u8(vld1q_u8(a + j), vld1q_u8(b + j));
uint16x8_t v_low = vmovl_u8(vget_low_u8(v_dst)), v_high = vmovl_u8(vget_high_u8(v_dst));
v_sum = vaddq_u32(v_sum, vaddl_u16(vget_low_u16(v_low), vget_low_u16(v_high)));
v_sum = vaddq_u32(v_sum, vaddl_u16(vget_high_u16(v_low), vget_high_u16(v_high)));
}
uint CV_DECL_ALIGNED(16) buf[4];
vst1q_u32(buf, v_sum);
d = buf[0] + buf[1] + buf[2] + buf[3];
#endif
{
for( ; j <= n - 4; j += 4 )
{
d += std::abs(a[j] - b[j]) + std::abs(a[j+1] - b[j+1]) +
std::abs(a[j+2] - b[j+2]) + std::abs(a[j+3] - b[j+3]);
}
}
for( ; j < n; j++ )
d += std::abs(a[j] - b[j]);
return d;
}
template<typename T, typename ST> int
normInf_(const T* src, const uchar* mask, ST* _result, int len, int cn)
{
@ -2564,7 +2430,7 @@ normInf_(const T* src, const uchar* mask, ST* _result, int len, int cn)
if( mask[i] )
{
for( int k = 0; k < cn; k++ )
result = std::max(result, ST(std::abs(src[k])));
result = std::max(result, ST(cv_abs(src[k])));
}
}
*_result = result;
@ -2585,7 +2451,7 @@ normL1_(const T* src, const uchar* mask, ST* _result, int len, int cn)
if( mask[i] )
{
for( int k = 0; k < cn; k++ )
result += std::abs(src[k]);
result += cv_abs(src[k]);
}
}
*_result = result;
@ -2684,9 +2550,7 @@ normDiffL2_(const T* src1, const T* src2, const uchar* mask, ST* _result, int le
Hamming::ResultType Hamming::operator()( const unsigned char* a, const unsigned char* b, int size ) const
{
int result = 0;
cv::hal::normHamming(a, b, size, result);
return result;
return cv::hal::normHamming(a, b, size);
}
#define CV_DEF_NORM_FUNC(L, suffix, type, ntype) \
@ -3037,16 +2901,12 @@ double cv::norm( InputArray _src, int normType, InputArray _mask )
if( normType == NORM_HAMMING )
{
int result = 0;
cv::hal::normHamming(data, (int)len, result);
return result;
return hal::normHamming(data, (int)len);
}
if( normType == NORM_HAMMING2 )
{
int result = 0;
hal::normHamming(data, (int)len, 2, result);
return result;
return hal::normHamming(data, (int)len, 2);
}
}
}
@ -3072,9 +2932,7 @@ double cv::norm( InputArray _src, int normType, InputArray _mask )
for( size_t i = 0; i < it.nplanes; i++, ++it )
{
int one = 0;
cv::hal::normHamming(ptrs[0], total, cellSize, one);
result += one;
result += hal::normHamming(ptrs[0], total, cellSize);
}
return result;
@ -3558,9 +3416,7 @@ double cv::norm( InputArray _src1, InputArray _src2, int normType, InputArray _m
for( size_t i = 0; i < it.nplanes; i++, ++it )
{
int one = 0;
hal::normHamming(ptrs[0], ptrs[1], total, cellSize, one);
result += one;
result += hal::normHamming(ptrs[0], ptrs[1], total, cellSize);
}
return result;
@ -3698,7 +3554,7 @@ static void batchDistHamming(const uchar* src1, const uchar* src2, size_t step2,
if( !mask )
{
for( int i = 0; i < nvecs; i++ )
hal::normHamming(src1, src2 + step2*i, len, dist[i]);
dist[i] = hal::normHamming(src1, src2 + step2*i, len);
}
else
{
@ -3706,7 +3562,7 @@ static void batchDistHamming(const uchar* src1, const uchar* src2, size_t step2,
for( int i = 0; i < nvecs; i++ )
{
if (mask[i])
hal::normHamming(src1, src2 + step2*i, len, dist[i]);
dist[i] = hal::normHamming(src1, src2 + step2*i, len);
else
dist[i] = val0;
}
@ -3720,7 +3576,7 @@ static void batchDistHamming2(const uchar* src1, const uchar* src2, size_t step2
if( !mask )
{
for( int i = 0; i < nvecs; i++ )
hal::normHamming(src1, src2 + step2*i, len, 2, dist[i]);
dist[i] = hal::normHamming(src1, src2 + step2*i, len, 2);
}
else
{
@ -3728,7 +3584,7 @@ static void batchDistHamming2(const uchar* src1, const uchar* src2, size_t step2
for( int i = 0; i < nvecs; i++ )
{
if (mask[i])
hal::normHamming(src1, src2 + step2*i, len, 2, dist[i]);
dist[i] = hal::normHamming(src1, src2 + step2*i, len, 2);
else
dist[i] = val0;
}

2
modules/features2d/src/kaze/AKAZEFeatures.cpp
View File

@ -812,7 +812,7 @@ void AKAZEFeatures::Compute_Main_Orientation(KeyPoint& kpt, const std::vector<TE
}
}
}
fastAtan2(resY, resX, Ang, ang_size, false);
hal::fastAtan2(resY, resX, Ang, ang_size, false);
// Loop slides pi/3 window around feature point
for (ang1 = 0; ang1 < (float)(2.0 * CV_PI); ang1 += 0.15f) {
ang2 = (ang1 + (float)(CV_PI / 3.0) >(float)(2.0*CV_PI) ? ang1 - (float)(5.0*CV_PI / 3.0) : ang1 + (float)(CV_PI / 3.0));

34
modules/hal/include/opencv2/hal.hpp
View File

@ -55,7 +55,7 @@ namespace cv { namespace hal {
namespace Error {
enum Code
enum
{
Ok = 0,
Unknown = -1
@ -63,11 +63,35 @@ enum Code
}
Error::Code normHamming(const uchar* a, int n, int & result);
Error::Code normHamming(const uchar* a, const uchar* b, int n, int & result);
int normHamming(const uchar* a, int n);
int normHamming(const uchar* a, const uchar* b, int n);
Error::Code normHamming(const uchar* a, int n, int cellSize, int & result);
Error::Code normHamming(const uchar* a, const uchar* b, int n, int cellSize, int & result);
int normHamming(const uchar* a, int n, int cellSize);
int normHamming(const uchar* a, const uchar* b, int n, int cellSize);
//////////////////////////////// low-level functions ////////////////////////////////
int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n);
int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n);
bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n);
bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n);
int normL1_(const uchar* a, const uchar* b, int n);
float normL1_(const float* a, const float* b, int n);
float normL2Sqr_(const float* a, const float* b, int n);
void exp(const float* src, float* dst, int n);
void exp(const double* src, double* dst, int n);
void log(const float* src, float* dst, int n);
void log(const double* src, double* dst, int n);
void fastAtan2(const float* y, const float* x, float* dst, int n, bool angleInDegrees);
void magnitude(const float* x, const float* y, float* dst, int n);
void magnitude(const double* x, const double* y, double* dst, int n);
void sqrt(const float* src, float* dst, int len);
void sqrt(const double* src, double* dst, int len);
void invSqrt(const float* src, float* dst, int len);
void invSqrt(const double* src, double* dst, int len);
}} //cv::hal

393
modules/hal/include/opencv2/hal/defs.h
View File

@ -1,3 +1,4 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
@ -48,6 +49,8 @@
# define _CRT_SECURE_NO_DEPRECATE /* to avoid multiple Visual Studio warnings */
#endif
#include <limits.h>
#if defined __ICL
# define CV_ICC __ICL
#elif defined __ICC
@ -60,12 +63,30 @@
# define CV_ICC __INTEL_COMPILER
#endif
#ifndef CV_INLINE
# if defined __cplusplus
# define CV_INLINE static inline
# elif defined _MSC_VER
# define CV_INLINE __inline
# else
# define CV_INLINE static
# endif
#endif
#if defined CV_ICC && !defined CV_ENABLE_UNROLLED
# define CV_ENABLE_UNROLLED 0
#else
# define CV_ENABLE_UNROLLED 1
#endif
#ifdef __GNUC__
# define CV_DECL_ALIGNED(x) __attribute__ ((aligned (x)))
#elif defined _MSC_VER
# define CV_DECL_ALIGNED(x) __declspec(align(x))
#else
# define CV_DECL_ALIGNED(x)
#endif
/* CPU features and intrinsics support */
#define CV_CPU_NONE 0
#define CV_CPU_MMX 1
@ -99,7 +120,7 @@
// do not include SSE/AVX/NEON headers for NVCC compiler
#ifndef __CUDACC__
#if defined __SSE2__ || defined _M_X64 || (defined _M_IX86_FP && _M_IX86_FP >= 2)
#if defined __SSE2__ || defined _M_X64 || (defined _M_IX86_FP && _M_IX86_FP >= 2)
# include <emmintrin.h>
# define CV_MMX 1
# define CV_SSE 1
@ -281,4 +302,374 @@ typedef signed char schar;
#define CV_2PI 6.283185307179586476925286766559
#define CV_LOG2 0.69314718055994530941723212145818
typedef union Cv32suf
{
int i;
unsigned u;
float f;
}
Cv32suf;
typedef union Cv64suf
{
int64 i;
uint64 u;
double f;
}
Cv64suf;
/****************************************************************************************\
* fast math *
\****************************************************************************************/
#if defined __BORLANDC__
# include <fastmath.h>
#elif defined __cplusplus
# include <cmath>
#else
# include <math.h>
#endif
#ifdef HAVE_TEGRA_OPTIMIZATION
# include "tegra_round.hpp"
#endif
//! @addtogroup core_utils
//! @{
#if CV_VFP
// 1. general scheme
#define ARM_ROUND(_value, _asm_string) \
int res; \
float temp; \
asm(_asm_string : [res] "=r" (res), [temp] "=w" (temp) : [value] "w" (_value)); \
return res
// 2. version for double
#ifdef __clang__
#define ARM_ROUND_DBL(value) ARM_ROUND(value, "vcvtr.s32.f64 %[temp], %[value] \n vmov %[res], %[temp]")
#else
#define ARM_ROUND_DBL(value) ARM_ROUND(value, "vcvtr.s32.f64 %[temp], %P[value] \n vmov %[res], %[temp]")
#endif
// 3. version for float
#define ARM_ROUND_FLT(value) ARM_ROUND(value, "vcvtr.s32.f32 %[temp], %[value]\n vmov %[res], %[temp]")
#endif // CV_VFP
/** @brief Rounds floating-point number to the nearest integer
@param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
result is not defined.
*/
CV_INLINE int
cvRound( double value )
{
#if ((defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__ \
&& defined __SSE2__ && !defined __APPLE__)) && !defined(__CUDACC__)
__m128d t = _mm_set_sd( value );
return _mm_cvtsd_si32(t);
#elif defined _MSC_VER && defined _M_IX86
int t;
__asm
{
fld value;
fistp t;
}
return t;
#elif ((defined _MSC_VER && defined _M_ARM) || defined CV_ICC || \
defined __GNUC__) && defined HAVE_TEGRA_OPTIMIZATION
TEGRA_ROUND_DBL(value);
#elif defined CV_ICC || defined __GNUC__
# if CV_VFP
ARM_ROUND_DBL(value);
# else
return (int)lrint(value);
# endif
#else
/* it's ok if round does not comply with IEEE754 standard;
the tests should allow +/-1 difference when the tested functions use round */
return (int)(value + (value >= 0 ? 0.5 : -0.5));
#endif
}
/** @brief Rounds floating-point number to the nearest integer not larger than the original.
The function computes an integer i such that:
\f[i \le \texttt{value} < i+1\f]
@param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
result is not defined.
*/
CV_INLINE int cvFloor( double value )
{
#if (defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__ && !defined __APPLE__)) && !defined(__CUDACC__)
__m128d t = _mm_set_sd( value );
int i = _mm_cvtsd_si32(t);
return i - _mm_movemask_pd(_mm_cmplt_sd(t, _mm_cvtsi32_sd(t,i)));
#elif defined __GNUC__
int i = (int)value;
return i - (i > value);
#else
int i = cvRound(value);
float diff = (float)(value - i);
return i - (diff < 0);
#endif
}
/** @brief Rounds floating-point number to the nearest integer not larger than the original.
The function computes an integer i such that:
\f[i \le \texttt{value} < i+1\f]
@param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
result is not defined.
*/
CV_INLINE int cvCeil( double value )
{
#if (defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__&& !defined __APPLE__)) && !defined(__CUDACC__)
__m128d t = _mm_set_sd( value );
int i = _mm_cvtsd_si32(t);
return i + _mm_movemask_pd(_mm_cmplt_sd(_mm_cvtsi32_sd(t,i), t));
#elif defined __GNUC__
int i = (int)value;
return i + (i < value);
#else
int i = cvRound(value);
float diff = (float)(i - value);
return i + (diff < 0);
#endif
}
/** @brief Determines if the argument is Not A Number.
@param value The input floating-point value
The function returns 1 if the argument is Not A Number (as defined by IEEE754 standard), 0
otherwise. */
CV_INLINE int cvIsNaN( double value )
{
Cv64suf ieee754;
ieee754.f = value;
return ((unsigned)(ieee754.u >> 32) & 0x7fffffff) +
((unsigned)ieee754.u != 0) > 0x7ff00000;
}
/** @brief Determines if the argument is Infinity.
@param value The input floating-point value
The function returns 1 if the argument is a plus or minus infinity (as defined by IEEE754 standard)
and 0 otherwise. */
CV_INLINE int cvIsInf( double value )
{
Cv64suf ieee754;
ieee754.f = value;
return ((unsigned)(ieee754.u >> 32) & 0x7fffffff) == 0x7ff00000 &&
(unsigned)ieee754.u == 0;
}
#ifdef __cplusplus
/** @overload */
CV_INLINE int cvRound(float value)
{
#if ((defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__ && \
defined __SSE2__ && !defined __APPLE__)) && !defined(__CUDACC__)
__m128 t = _mm_set_ss( value );
return _mm_cvtss_si32(t);
#elif defined _MSC_VER && defined _M_IX86
int t;
__asm
{
fld value;
fistp t;
}
return t;
#elif ((defined _MSC_VER && defined _M_ARM) || defined CV_ICC || \
defined __GNUC__) && defined HAVE_TEGRA_OPTIMIZATION
TEGRA_ROUND_FLT(value);
#elif defined CV_ICC || defined __GNUC__
# if CV_VFP
ARM_ROUND_FLT(value);
# else
return (int)lrintf(value);
# endif
#else
/* it's ok if round does not comply with IEEE754 standard;
the tests should allow +/-1 difference when the tested functions use round */
return (int)(value + (value >= 0 ? 0.5f : -0.5f));
#endif
}
/** @overload */
CV_INLINE int cvRound( int value )
{
return value;
}
/** @overload */
CV_INLINE int cvFloor( float value )
{
#if (defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__ && !defined __APPLE__)) && !defined(__CUDACC__)
__m128 t = _mm_set_ss( value );
int i = _mm_cvtss_si32(t);
return i - _mm_movemask_ps(_mm_cmplt_ss(t, _mm_cvtsi32_ss(t,i)));
#elif defined __GNUC__
int i = (int)value;
return i - (i > value);
#else
int i = cvRound(value);
float diff = (float)(value - i);
return i - (diff < 0);
#endif
}
/** @overload */
CV_INLINE int cvFloor( int value )
{
return value;
}
/** @overload */
CV_INLINE int cvCeil( float value )
{
#if (defined _MSC_VER && defined _M_X64 || (defined __GNUC__ && defined __SSE2__&& !defined __APPLE__)) && !defined(__CUDACC__)
__m128 t = _mm_set_ss( value );
int i = _mm_cvtss_si32(t);
return i + _mm_movemask_ps(_mm_cmplt_ss(_mm_cvtsi32_ss(t,i), t));
#elif defined __GNUC__
int i = (int)value;
return i + (i < value);
#else
int i = cvRound(value);
float diff = (float)(i - value);
return i + (diff < 0);
#endif
}
/** @overload */
CV_INLINE int cvCeil( int value )
{
return value;
}
/** @overload */
CV_INLINE int cvIsNaN( float value )
{
Cv32suf ieee754;
ieee754.f = value;
return (ieee754.u & 0x7fffffff) > 0x7f800000;
}
/** @overload */
CV_INLINE int cvIsInf( float value )
{
Cv32suf ieee754;
ieee754.f = value;
return (ieee754.u & 0x7fffffff) == 0x7f800000;
}
#include <algorithm>
namespace cv
{
/////////////// saturate_cast (used in image & signal processing) ///////////////////
/**
Template function for accurate conversion from one primitive type to another.
The functions saturate_cast resemble the standard C++ cast operations, such as static_cast\<T\>()
and others. They perform an efficient and accurate conversion from one primitive type to another
(see the introduction chapter). saturate in the name means that when the input value v is out of the
range of the target type, the result is not formed just by taking low bits of the input, but instead
the value is clipped. For example:
@code
uchar a = saturate_cast<uchar>(-100); // a = 0 (UCHAR_MIN)
short b = saturate_cast<short>(33333.33333); // b = 32767 (SHRT_MAX)
@endcode
Such clipping is done when the target type is unsigned char , signed char , unsigned short or
signed short . For 32-bit integers, no clipping is done.
When the parameter is a floating-point value and the target type is an integer (8-, 16- or 32-bit),
the floating-point value is first rounded to the nearest integer and then clipped if needed (when
the target type is 8- or 16-bit).
This operation is used in the simplest or most complex image processing functions in OpenCV.
@param v Function parameter.
@sa add, subtract, multiply, divide, Mat::convertTo
*/
template<typename _Tp> static inline _Tp saturate_cast(uchar v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(schar v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(ushort v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(short v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(unsigned v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(int v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(float v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(double v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(int64 v) { return _Tp(v); }
/** @overload */
template<typename _Tp> static inline _Tp saturate_cast(uint64 v) { return _Tp(v); }
//! @cond IGNORED
template<> inline uchar saturate_cast<uchar>(schar v) { return (uchar)std::max((int)v, 0); }
template<> inline uchar saturate_cast<uchar>(ushort v) { return (uchar)std::min((unsigned)v, (unsigned)UCHAR_MAX); }
template<> inline uchar saturate_cast<uchar>(int v) { return (uchar)((unsigned)v <= UCHAR_MAX ? v : v > 0 ? UCHAR_MAX : 0); }
template<> inline uchar saturate_cast<uchar>(short v) { return saturate_cast<uchar>((int)v); }
template<> inline uchar saturate_cast<uchar>(unsigned v) { return (uchar)std::min(v, (unsigned)UCHAR_MAX); }
template<> inline uchar saturate_cast<uchar>(float v) { int iv = cvRound(v); return saturate_cast<uchar>(iv); }
template<> inline uchar saturate_cast<uchar>(double v) { int iv = cvRound(v); return saturate_cast<uchar>(iv); }
template<> inline uchar saturate_cast<uchar>(int64 v) { return (uchar)((uint64)v <= (uint64)UCHAR_MAX ? v : v > 0 ? UCHAR_MAX : 0); }
template<> inline uchar saturate_cast<uchar>(uint64 v) { return (uchar)std::min(v, (uint64)UCHAR_MAX); }
template<> inline schar saturate_cast<schar>(uchar v) { return (schar)std::min((int)v, SCHAR_MAX); }
template<> inline schar saturate_cast<schar>(ushort v) { return (schar)std::min((unsigned)v, (unsigned)SCHAR_MAX); }
template<> inline schar saturate_cast<schar>(int v) { return (schar)((unsigned)(v-SCHAR_MIN) <= (unsigned)UCHAR_MAX ? v : v > 0 ? SCHAR_MAX : SCHAR_MIN); }
template<> inline schar saturate_cast<schar>(short v) { return saturate_cast<schar>((int)v); }
template<> inline schar saturate_cast<schar>(unsigned v) { return (schar)std::min(v, (unsigned)SCHAR_MAX); }
template<> inline schar saturate_cast<schar>(float v) { int iv = cvRound(v); return saturate_cast<schar>(iv); }
template<> inline schar saturate_cast<schar>(double v) { int iv = cvRound(v); return saturate_cast<schar>(iv); }
template<> inline schar saturate_cast<schar>(int64 v) { return (schar)((uint64)((int64)v-SCHAR_MIN) <= (uint64)UCHAR_MAX ? v : v > 0 ? SCHAR_MAX : SCHAR_MIN); }
template<> inline schar saturate_cast<schar>(uint64 v) { return (schar)std::min(v, (uint64)SCHAR_MAX); }
template<> inline ushort saturate_cast<ushort>(schar v) { return (ushort)std::max((int)v, 0); }
template<> inline ushort saturate_cast<ushort>(short v) { return (ushort)std::max((int)v, 0); }
template<> inline ushort saturate_cast<ushort>(int v) { return (ushort)((unsigned)v <= (unsigned)USHRT_MAX ? v : v > 0 ? USHRT_MAX : 0); }
template<> inline ushort saturate_cast<ushort>(unsigned v) { return (ushort)std::min(v, (unsigned)USHRT_MAX); }
template<> inline ushort saturate_cast<ushort>(float v) { int iv = cvRound(v); return saturate_cast<ushort>(iv); }
template<> inline ushort saturate_cast<ushort>(double v) { int iv = cvRound(v); return saturate_cast<ushort>(iv); }
template<> inline ushort saturate_cast<ushort>(int64 v) { return (ushort)((uint64)v <= (uint64)USHRT_MAX ? v : v > 0 ? USHRT_MAX : 0); }
template<> inline ushort saturate_cast<ushort>(uint64 v) { return (ushort)std::min(v, (uint64)USHRT_MAX); }
template<> inline short saturate_cast<short>(ushort v) { return (short)std::min((int)v, SHRT_MAX); }
template<> inline short saturate_cast<short>(int v) { return (short)((unsigned)(v - SHRT_MIN) <= (unsigned)USHRT_MAX ? v : v > 0 ? SHRT_MAX : SHRT_MIN); }
template<> inline short saturate_cast<short>(unsigned v) { return (short)std::min(v, (unsigned)SHRT_MAX); }
template<> inline short saturate_cast<short>(float v) { int iv = cvRound(v); return saturate_cast<short>(iv); }
template<> inline short saturate_cast<short>(double v) { int iv = cvRound(v); return saturate_cast<short>(iv); }
template<> inline short saturate_cast<short>(int64 v) { return (short)((uint64)((int64)v - SHRT_MIN) <= (uint64)USHRT_MAX ? v : v > 0 ? SHRT_MAX : SHRT_MIN); }
template<> inline short saturate_cast<short>(uint64 v) { return (short)std::min(v, (uint64)SHRT_MAX); }
template<> inline int saturate_cast<int>(float v) { return cvRound(v); }
template<> inline int saturate_cast<int>(double v) { return cvRound(v); }
// we intentionally do not clip negative numbers, to make -1 become 0xffffffff etc.
template<> inline unsigned saturate_cast<unsigned>(float v) { return cvRound(v); }
template<> inline unsigned saturate_cast<unsigned>(double v) { return cvRound(v); }
//! @endcond
}
#endif // __cplusplus
//! @} core_utils
#endif //__OPENCV_HAL_H__

292
modules/hal/include/opencv2/hal/intrin.hpp Normal file
View File

@ -0,0 +1,292 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#ifndef __OPENCV_HAL_INTRIN_HPP__
#define __OPENCV_HAL_INTRIN_HPP__
#include <cmath>
#include <float.h>
#include <stdlib.h>
#define OPENCV_HAL_ADD(a, b) ((a) + (b))
#define OPENCV_HAL_AND(a, b) ((a) & (b))
#define OPENCV_HAL_NOP(a) (a)
#define OPENCV_HAL_1ST(a, b) (a)
// unlike HAL API, which is in cv::hall,
// we put intrinsics into cv namespace to make its
// access from within opencv code more accessible
namespace cv {
template<typename _Tp> struct V_TypeTraits
{
typedef _Tp int_type;
typedef _Tp uint_type;
typedef _Tp abs_type;
typedef _Tp sum_type;
enum { delta = 0, shift = 0 };
static int_type reinterpret_int(_Tp x) { return x; }
static uint_type reinterpet_uint(_Tp x) { return x; }
static _Tp reinterpret_from_int(int_type x) { return (_Tp)x; }
};
template<> struct V_TypeTraits<uchar>
{
typedef uchar value_type;
typedef schar int_type;
typedef uchar uint_type;
typedef uchar abs_type;
typedef int sum_type;
typedef ushort w_type;
enum { delta = 128, shift = 8 };
static int_type reinterpret_int(value_type x) { return (int_type)x; }
static uint_type reinterpret_uint(value_type x) { return (uint_type)x; }
static value_type reinterpret_from_int(int_type x) { return (value_type)x; }
};
template<> struct V_TypeTraits<schar>
{
typedef schar value_type;
typedef schar int_type;
typedef uchar uint_type;
typedef uchar abs_type;
typedef int sum_type;
typedef short w_type;
enum { delta = 128, shift = 8 };
static int_type reinterpret_int(value_type x) { return (int_type)x; }
static uint_type reinterpret_uint(value_type x) { return (uint_type)x; }
static value_type reinterpret_from_int(int_type x) { return (value_type)x; }
};
template<> struct V_TypeTraits<ushort>
{
typedef ushort value_type;
typedef short int_type;
typedef ushort uint_type;
typedef ushort abs_type;
typedef int sum_type;
typedef unsigned w_type;
typedef uchar nu_type;
enum { delta = 32768, shift = 16 };
static int_type reinterpret_int(value_type x) { return (int_type)x; }
static uint_type reinterpret_uint(value_type x) { return (uint_type)x; }
static value_type reinterpret_from_int(int_type x) { return (value_type)x; }
};
template<> struct V_TypeTraits<short>
{
typedef short value_type;
typedef short int_type;
typedef ushort uint_type;
typedef ushort abs_type;
typedef int sum_type;
typedef int w_type;
typedef uchar nu_type;
typedef schar n_type;
enum { delta = 128, shift = 8 };
static int_type reinterpret_int(value_type x) { return (int_type)x; }
static uint_type reinterpret_uint(value_type x) { return (uint_type)x; }
static value_type reinterpret_from_int(int_type x) { return (value_type)x; }
};
template<> struct V_TypeTraits<unsigned>
{
typedef unsigned value_type;
typedef int int_type;
typedef unsigned uint_type;
typedef unsigned abs_type;
typedef unsigned sum_type;
typedef uint64 w_type;
typedef ushort nu_type;
static int_type reinterpret_int(value_type x) { return (int_type)x; }
static uint_type reinterpret_uint(value_type x) { return (uint_type)x; }
static value_type reinterpret_from_int(int_type x) { return (value_type)x; }
};
template<> struct V_TypeTraits<int>
{
typedef int value_type;
typedef int int_type;
typedef unsigned uint_type;
typedef unsigned abs_type;
typedef int sum_type;
typedef int64 w_type;
typedef short n_type;
typedef ushort nu_type;
static int_type reinterpret_int(value_type x) { return (int_type)x; }
static uint_type reinterpret_uint(value_type x) { return (uint_type)x; }
static value_type reinterpret_from_int(int_type x) { return (value_type)x; }
};
template<> struct V_TypeTraits<uint64>
{
typedef uint64 value_type;
typedef int64 int_type;
typedef uint64 uint_type;
typedef uint64 abs_type;
typedef uint64 sum_type;
typedef unsigned nu_type;
static int_type reinterpret_int(value_type x) { return (int_type)x; }
static uint_type reinterpret_uint(value_type x) { return (uint_type)x; }
static value_type reinterpret_from_int(int_type x) { return (value_type)x; }
};
template<> struct V_TypeTraits<int64>
{
typedef int64 value_type;
typedef int64 int_type;
typedef uint64 uint_type;
typedef uint64 abs_type;
typedef int64 sum_type;
typedef int nu_type;
static int_type reinterpret_int(value_type x) { return (int_type)x; }
static uint_type reinterpret_uint(value_type x) { return (uint_type)x; }
static value_type reinterpret_from_int(int_type x) { return (value_type)x; }
};
template<> struct V_TypeTraits<float>
{
typedef float value_type;
typedef int int_type;
typedef unsigned uint_type;
typedef float abs_type;
typedef float sum_type;
typedef double w_type;
static int_type reinterpret_int(value_type x)
{
Cv32suf u;
u.f = x;
return u.i;
}
static uint_type reinterpet_uint(value_type x)
{
Cv32suf u;
u.f = x;
return u.u;
}
static value_type reinterpret_from_int(int_type x)
{
Cv32suf u;
u.i = x;
return u.f;
}
};
template<> struct V_TypeTraits<double>
{
typedef double value_type;
typedef int64 int_type;
typedef uint64 uint_type;
typedef double abs_type;
typedef double sum_type;
static int_type reinterpret_int(value_type x)
{
Cv64suf u;
u.f = x;
return u.i;
}
static uint_type reinterpet_uint(value_type x)
{
Cv64suf u;
u.f = x;
return u.u;
}
static value_type reinterpret_from_int(int_type x)
{
Cv64suf u;
u.i = x;
return u.f;
}
};
}
#if CV_SSE2
#include "opencv2/hal/intrin_sse.hpp"
#elif CV_NEON
#include "opencv2/hal/intrin_neon.hpp"
#else
#include "opencv2/hal/intrin_cpp.hpp"
#endif
#ifndef CV_SIMD128
#define CV_SIMD128 0
#endif
#ifndef CV_SIMD128_64F
#define CV_SIMD128_64F 0
#endif
#endif

811
modules/hal/include/opencv2/hal/intrin_cpp.hpp Normal file
View File

@ -0,0 +1,811 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#ifndef __OPENCV_HAL_INTRIN_CPP_HPP__
#define __OPENCV_HAL_INTRIN_CPP_HPP__
namespace cv
{
template<typename _Tp, int n> struct v_reg
{
typedef _Tp lane_type;
typedef v_reg<typename V_TypeTraits<_Tp>::int_type, n> int_vec;
typedef v_reg<typename V_TypeTraits<_Tp>::abs_type, n> abs_vec;
enum { nlanes = n };
explicit v_reg(const _Tp* ptr) { for( int i = 0; i < n; i++ ) s[i] = ptr[i]; }
v_reg(_Tp s0, _Tp s1) { s[0] = s0; s[1] = s1; }
v_reg(_Tp s0, _Tp s1, _Tp s2, _Tp s3) { s[0] = s0; s[1] = s1; s[2] = s2; s[3] = s3; }
v_reg(_Tp s0, _Tp s1, _Tp s2, _Tp s3,
_Tp s4, _Tp s5, _Tp s6, _Tp s7)
{
s[0] = s0; s[1] = s1; s[2] = s2; s[3] = s3;
s[4] = s4; s[5] = s5; s[6] = s6; s[7] = s7;
}
v_reg(_Tp s0, _Tp s1, _Tp s2, _Tp s3,
_Tp s4, _Tp s5, _Tp s6, _Tp s7,
_Tp s8, _Tp s9, _Tp s10, _Tp s11,
_Tp s12, _Tp s13, _Tp s14, _Tp s15)
{
s[0] = s0; s[1] = s1; s[2] = s2; s[3] = s3;
s[4] = s4; s[5] = s5; s[6] = s6; s[7] = s7;
s[8] = s8; s[9] = s9; s[10] = s10; s[11] = s11;
s[12] = s12; s[13] = s13; s[14] = s14; s[15] = s15;
}
v_reg() {}
v_reg(const v_reg<_Tp, n> & r)
{
for( int i = 0; i < n; i++ )
s[i] = r.s[i];
}
_Tp get(const int i) const { return s[i]; }
_Tp get0() const { return s[0]; }
v_reg<_Tp, n> high() const
{
v_reg<_Tp, n> c;
int i;
for( i = 0; i < n/2; i++ )
{
c.s[i] = s[i+(n/2)];
c.s[i+(n/2)] = 0;
}
return c;
}
static v_reg<_Tp, n> zero()
{
v_reg<_Tp, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = (_Tp)0;
return c;
}
static v_reg<_Tp, n> all(_Tp s)
{
v_reg<_Tp, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = s;
return c;
}
template<typename _Tp2, int n2> v_reg<_Tp2, n2> reinterpret_as() const
{
size_t bytes = std::min(sizeof(_Tp2)*n2, sizeof(_Tp)*n);
v_reg<_Tp2, n2> c;
memcpy(&c.s[0], &s[0], bytes);
return c;
}
_Tp s[n];
};
#define OPENCV_HAL_IMPL_BIN_OP(bin_op) \
template<typename _Tp, int n> inline v_reg<_Tp, n> \
operator bin_op (const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
{ \
v_reg<_Tp, n> c; \
for( int i = 0; i < n; i++ ) \
c.s[i] = saturate_cast<_Tp>(a.s[i] bin_op b.s[i]); \
return c; \
} \
template<typename _Tp, int n> inline v_reg<_Tp, n>& \
operator bin_op##= (v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
{ \
for( int i = 0; i < n; i++ ) \
a.s[i] = saturate_cast<_Tp>(a.s[i] bin_op b.s[i]); \
return a; \
}
OPENCV_HAL_IMPL_BIN_OP(+)
OPENCV_HAL_IMPL_BIN_OP(-)
OPENCV_HAL_IMPL_BIN_OP(*)
OPENCV_HAL_IMPL_BIN_OP(/)
#define OPENCV_HAL_IMPL_BIT_OP(bit_op) \
template<typename _Tp, int n> inline v_reg<_Tp, n> operator bit_op \
(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
{ \
v_reg<_Tp, n> c; \
typedef typename V_TypeTraits<_Tp>::int_type itype; \
for( int i = 0; i < n; i++ ) \
c.s[i] = V_TypeTraits<_Tp>::reinterpret_from_int((itype)(V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) bit_op \
V_TypeTraits<_Tp>::reinterpret_int(b.s[i]))); \
return c; \
} \
template<typename _Tp, int n> inline v_reg<_Tp, n>& operator \
bit_op##= (v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
{ \
typedef typename V_TypeTraits<_Tp>::int_type itype; \
for( int i = 0; i < n; i++ ) \
a.s[i] = V_TypeTraits<_Tp>::reinterpret_from_int((itype)(V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) bit_op \
V_TypeTraits<_Tp>::reinterpret_int(b.s[i]))); \
return a; \
}
OPENCV_HAL_IMPL_BIT_OP(&)
OPENCV_HAL_IMPL_BIT_OP(|)
OPENCV_HAL_IMPL_BIT_OP(^)
template<typename _Tp, int n> inline v_reg<_Tp, n> operator ~ (const v_reg<_Tp, n>& a)
{
v_reg<_Tp, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = V_TypeTraits<_Tp>::reinterpret_from_int(~V_TypeTraits<_Tp>::reinterpret_int(a.s[i]));
return c;
}
#define OPENCV_HAL_IMPL_MATH_FUNC(func, cfunc, _Tp2) \
template<typename _Tp, int n> inline v_reg<_Tp2, n> func(const v_reg<_Tp, n>& a) \
{ \
v_reg<_Tp2, n> c; \
for( int i = 0; i < n; i++ ) \
c.s[i] = cfunc(a.s[i]); \
return c; \
}
OPENCV_HAL_IMPL_MATH_FUNC(v_sqrt, std::sqrt, _Tp)
OPENCV_HAL_IMPL_MATH_FUNC(v_sin, std::sin, _Tp)
OPENCV_HAL_IMPL_MATH_FUNC(v_cos, std::cos, _Tp)
OPENCV_HAL_IMPL_MATH_FUNC(v_exp, std::exp, _Tp)
OPENCV_HAL_IMPL_MATH_FUNC(v_log, std::log, _Tp)
OPENCV_HAL_IMPL_MATH_FUNC(v_abs, (typename V_TypeTraits<_Tp>::abs_type)std::abs,
typename V_TypeTraits<_Tp>::abs_type)
OPENCV_HAL_IMPL_MATH_FUNC(v_round, cvRound, int)
OPENCV_HAL_IMPL_MATH_FUNC(v_floor, cvFloor, int)
OPENCV_HAL_IMPL_MATH_FUNC(v_ceil, cvCeil, int)
OPENCV_HAL_IMPL_MATH_FUNC(v_trunc, int, int)
#define OPENCV_HAL_IMPL_MINMAX_FUNC(func, hfunc, cfunc) \
template<typename _Tp, int n> inline v_reg<_Tp, n> func(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
{ \
v_reg<_Tp, n> c; \
for( int i = 0; i < n; i++ ) \
c.s[i] = cfunc(a.s[i], b.s[i]); \
return c; \
} \
template<typename _Tp, int n> inline _Tp hfunc(const v_reg<_Tp, n>& a) \
{ \
_Tp c = a.s[0]; \
for( int i = 1; i < n; i++ ) \
c = cfunc(c, a.s[i]); \
return c; \
}
OPENCV_HAL_IMPL_MINMAX_FUNC(v_min, v_reduce_min, std::min)
OPENCV_HAL_IMPL_MINMAX_FUNC(v_max, v_reduce_max, std::max)
template<typename _Tp, int n>
inline void v_minmax( const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b,
v_reg<_Tp, n>& minval, v_reg<_Tp, n>& maxval )
{
for( int i = 0; i < n; i++ )
{
minval.s[i] = std::min(a.s[i], b.s[i]);
maxval.s[i] = std::max(a.s[i], b.s[i]);
}
}
#define OPENCV_HAL_IMPL_CMP_OP(cmp_op) \
template<typename _Tp, int n> \
inline v_reg<_Tp, n> operator cmp_op(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
{ \
typedef typename V_TypeTraits<_Tp>::int_type itype; \
v_reg<_Tp, n> c; \
for( int i = 0; i < n; i++ ) \
c.s[i] = V_TypeTraits<_Tp>::reinterpret_from_int((itype)-(int)(a.s[i] cmp_op b.s[i])); \
return c; \
}
OPENCV_HAL_IMPL_CMP_OP(<)
OPENCV_HAL_IMPL_CMP_OP(>)
OPENCV_HAL_IMPL_CMP_OP(<=)
OPENCV_HAL_IMPL_CMP_OP(>=)
OPENCV_HAL_IMPL_CMP_OP(==)
OPENCV_HAL_IMPL_CMP_OP(!=)
#define OPENCV_HAL_IMPL_ADD_SUB_OP(func, bin_op, cast_op, _Tp2) \
template<typename _Tp, int n> \
inline v_reg<_Tp2, n> func(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
{ \
typedef _Tp2 rtype; \
v_reg<rtype, n> c; \
for( int i = 0; i < n; i++ ) \
c.s[i] = cast_op(a.s[i] bin_op b.s[i]); \
return c; \
}
OPENCV_HAL_IMPL_ADD_SUB_OP(v_add_wrap, +, (_Tp), _Tp)
OPENCV_HAL_IMPL_ADD_SUB_OP(v_sub_wrap, -, (_Tp), _Tp)
OPENCV_HAL_IMPL_ADD_SUB_OP(v_absdiff, -, (rtype)std::abs, typename V_TypeTraits<_Tp>::abs_type)
template<typename _Tp, int n>
inline v_reg<_Tp, n> v_invsqrt(const v_reg<_Tp, n>& a)
{
v_reg<_Tp, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = 1.f/std::sqrt(a.s[i]);
return c;
}
template<typename _Tp, int n>
inline v_reg<_Tp, n> v_magnitude(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
{
v_reg<_Tp, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = std::sqrt(a.s[i]*a.s[i] + b.s[i]*b.s[i]);
return c;
}
template<typename _Tp, int n>
inline v_reg<_Tp, n> v_sqr_magnitude(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
{
v_reg<_Tp, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = a.s[i]*a.s[i] + b.s[i]*b.s[i];
return c;
}
template<typename _Tp, int n>
inline v_reg<_Tp, n> v_muladd(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b,
const v_reg<_Tp, n>& c)
{
v_reg<_Tp, n> d;
for( int i = 0; i < n; i++ )
d.s[i] = a.s[i]*b.s[i] + c.s[i];
return d;
}
template<typename _Tp, int n> inline v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>
v_dotprod(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
{
typedef typename V_TypeTraits<_Tp>::w_type w_type;
v_reg<w_type, n/2> c;
for( int i = 0; i < (n/2); i++ )
c.s[i] = (w_type)a.s[i*2]*b.s[i*2] + (w_type)a.s[i*2+1]*b.s[i*2+1];
return c;
}
template<typename _Tp, int n> inline void v_mul_expand(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b,
v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& c,
v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& d)
{
typedef typename V_TypeTraits<_Tp>::w_type w_type;
for( int i = 0; i < (n/2); i++ )
{
c.s[i] = (w_type)a.s[i]*b.s[i]*2;
d.s[i] = (w_type)a.s[i+(n/2)]*b.s[i+(n/2)];
}
}
template<typename _Tp, int n> inline void v_hsum(const v_reg<_Tp, n>& a,
v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& c)
{
typedef typename V_TypeTraits<_Tp>::w_type w_type;
for( int i = 0; i < (n/2); i++ )
{
c.s[i] = (w_type)a.s[i*2] + a.s[i*2+1];
}
}
#define OPENCV_HAL_IMPL_SHIFT_OP(shift_op) \
template<typename _Tp, int n> inline v_reg<_Tp, n> operator shift_op(const v_reg<_Tp, n>& a, int imm) \
{ \
v_reg<_Tp, n> c; \
for( int i = 0; i < n; i++ ) \
c.s[i] = (_Tp)(a.s[i] shift_op imm); \
return c; \
}
OPENCV_HAL_IMPL_SHIFT_OP(<<)
OPENCV_HAL_IMPL_SHIFT_OP(>>)
template<typename _Tp, int n> inline typename V_TypeTraits<_Tp>::sum_type v_reduce_sum(const v_reg<_Tp, n>& a)
{
typename V_TypeTraits<_Tp>::sum_type c = a.s[0];
for( int i = 1; i < n; i++ )
c += a.s[i];
return c;
}
template<typename _Tp, int n> inline int v_signmask(const v_reg<_Tp, n>& a)
{
int mask = 0;
for( int i = 0; i < n; i++ )
mask |= (V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) < 0) << i;
return mask;
}
template<typename _Tp, int n> inline bool v_check_all(const v_reg<_Tp, n>& a)
{
for( int i = 0; i < n; i++ )
if( V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) >= 0 )
return false;
return true;
}
template<typename _Tp, int n> inline bool v_check_any(const v_reg<_Tp, n>& a)
{
for( int i = 0; i < n; i++ )
if( V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) < 0 )
return true;
return false;
}
template<typename _Tp, int n> inline v_reg<_Tp, n> v_select(const v_reg<_Tp, n>& mask,
const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
{
v_reg<_Tp, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = V_TypeTraits<_Tp>::reinterpret_int(mask.s[i]) < 0 ? b.s[i] : a.s[i];
return c;
}
template<typename _Tp, int n> inline void v_expand(const v_reg<_Tp, n>& a,
v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& b0,
v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& b1)
{
for( int i = 0; i < (n/2); i++ )
{
b0.s[i] = a.s[i];
b1.s[i] = a.s[i+(n/2)];
}
}
template<typename _Tp, int n> inline v_reg<typename V_TypeTraits<_Tp>::int_type, n>
v_reinterpret_as_int(const v_reg<_Tp, n>& a)
{
v_reg<typename V_TypeTraits<_Tp>::int_type, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = V_TypeTraits<_Tp>::reinterpret_int(a.s[i]);
return c;
}
template<typename _Tp, int n> inline v_reg<typename V_TypeTraits<_Tp>::uint_type, n>
v_reinterpret_as_uint(const v_reg<_Tp, n>& a)
{
v_reg<typename V_TypeTraits<_Tp>::uint_type, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = V_TypeTraits<_Tp>::reinterpret_uint(a.s[i]);
return c;
}
template<typename _Tp, int n> inline void v_zip( const v_reg<_Tp, n>& a0, const v_reg<_Tp, n>& a1,
v_reg<_Tp, n>& b0, v_reg<_Tp, n>& b1 )
{
int i;
for( i = 0; i < n/2; i++ )
{
b0.s[i*2] = a0.s[i];
b0.s[i*2+1] = a1.s[i];
}
for( ; i < n; i++ )
{
b1.s[i*2-n] = a0.s[i];
b1.s[i*2-n+1] = a1.s[i];
}
}
template<typename _Tp, int n> inline v_reg<_Tp, n> v_load(const _Tp* ptr)
{
return v_reg<_Tp, n>(ptr);
}
template<typename _Tp, int n> inline v_reg<_Tp, n> v_load_aligned(const _Tp* ptr)
{
return v_reg<_Tp, n>(ptr);
}
template<typename _Tp, int n> inline void v_load_halves(const _Tp* loptr, const _Tp* hiptr)
{
v_reg<_Tp, n> c;
for( int i = 0; i < n/2; i++ )
{
c.s[i] = loptr[i];
c.s[i+n/2] = hiptr[i];
}
return c;
}
template<typename _Tp, int n> inline v_reg<typename V_TypeTraits<_Tp>::w_type, n> v_load_expand(const _Tp* ptr)
{
typedef typename V_TypeTraits<_Tp>::w_type w_type;
v_reg<w_type, n> c;
for( int i = 0; i < n; i++ )
{
c.s[i] = ptr[i];
}
return c;
}
template<typename _Tp, int n> inline v_reg<typename
V_TypeTraits<typename V_TypeTraits<_Tp>::w_type>::w_type, n> v_load_expand_q(const _Tp* ptr)
{
typedef typename V_TypeTraits<typename V_TypeTraits<_Tp>::w_type>::w_type w_type;
v_reg<w_type, n> c;
for( int i = 0; i < n; i++ )
{
c.s[i] = ptr[i];
}
return c;
}
template<typename _Tp, int n> inline void v_load_deinterleave(const _Tp* ptr, v_reg<_Tp, n>& a,
v_reg<_Tp, n>& b, v_reg<_Tp, n>& c)
{
int i, i3;
for( i = i3 = 0; i < n; i++, i3 += 3 )
{
a.s[i] = ptr[i3];
b.s[i] = ptr[i3+1];
c.s[i] = ptr[i3+2];
}
}
template<typename _Tp, int n>
inline void v_load_deinterleave(const _Tp* ptr, v_reg<_Tp, n>& a,
v_reg<_Tp, n>& b, v_reg<_Tp, n>& c,
v_reg<_Tp, n>& d)
{
int i, i4;
for( i = i4 = 0; i < n; i++, i4 += 4 )
{
a.s[i] = ptr[i4];
b.s[i] = ptr[i4+1];
c.s[i] = ptr[i4+2];
d.s[i] = ptr[i4+3];
}
}
template<typename _Tp, int n>
inline void v_store_interleave( _Tp* ptr, const v_reg<_Tp, n>& a,
const v_reg<_Tp, n>& b, const v_reg<_Tp, n>& c)
{
int i, i3;
for( i = i3 = 0; i < n; i++, i3 += 3 )
{
ptr[i3] = a.s[i];
ptr[i3+1] = b.s[i];
ptr[i3+2] = c.s[i];
}
}
template<typename _Tp, int n> inline void v_store_interleave( _Tp* ptr, const v_reg<_Tp, n>& a,
const v_reg<_Tp, n>& b, const v_reg<_Tp, n>& c,
const v_reg<_Tp, n>& d)
{
int i, i4;
for( i = i4 = 0; i < n; i++, i4 += 4 )
{
ptr[i4] = a.s[i];
ptr[i4+1] = b.s[i];
ptr[i4+2] = c.s[i];
ptr[i4+3] = d.s[i];
}
}
template<typename _Tp, int n>
inline void v_store(_Tp* ptr, const v_reg<_Tp, n>& a)
{
for( int i = 0; i < n; i++ )
ptr[i] = a.s[i];
}
template<typename _Tp, int n>
inline void v_store_low(_Tp* ptr, const v_reg<_Tp, n>& a)
{
for( int i = 0; i < (n/2); i++ )
ptr[i] = a.s[i];
}
template<typename _Tp, int n>
inline void v_store_high(_Tp* ptr, const v_reg<_Tp, n>& a)
{
for( int i = 0; i < (n/2); i++ )
ptr[i] = a.s[i+(n/2)];
}
template<typename _Tp, int n>
inline void v_store_aligned(_Tp* ptr, const v_reg<_Tp, n>& a)
{
for( int i = 0; i < n; i++ )
ptr[i] = a.s[i];
}
template<typename _Tp, int n>
inline v_reg<_Tp, n> v_combine_low(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
{
v_reg<_Tp, n> c;
for( int i = 0; i < (n/2); i++ )
{
c.s[i] = a.s[i];
c.s[i+(n/2)] = b.s[i];
}
}
template<typename _Tp, int n>
inline v_reg<_Tp, n> v_combine_high(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
{
v_reg<_Tp, n> c;
for( int i = 0; i < (n/2); i++ )
{
c.s[i] = a.s[i+(n/2)];
c.s[i+(n/2)] = b.s[i+(n/2)];
}
}
template<typename _Tp, int n>
inline void v_recombine(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b,
v_reg<_Tp, n>& low, v_reg<_Tp, n>& high)
{
for( int i = 0; i < (n/2); i++ )
{
low.s[i] = a.s[i];
low.s[i+(n/2)] = b.s[i];
high.s[i] = a.s[i+(n/2)];
high.s[i+(n/2)] = b.s[i+(n/2)];
}
}
template<int n> inline v_reg<int, n> v_round(const v_reg<float, n>& a)
{
v_reg<int, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = cvRound(a.s[i]);
return c;
}
template<int n> inline v_reg<int, n> v_floor(const v_reg<float, n>& a)
{
v_reg<int, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = cvFloor(a.s[i]);
return c;
}
template<int n> inline v_reg<int, n> v_ceil(const v_reg<float, n>& a)
{
v_reg<int, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = cvCeil(a.s[i]);
return c;
}
template<int n> inline v_reg<int, n> v_trunc(const v_reg<float, n>& a)
{
v_reg<int, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = (int)(a.s[i]);
return c;
}
template<int n> inline v_reg<int, n*2> v_round(const v_reg<double, n>& a)
{
v_reg<int, n*2> c;
for( int i = 0; i < n; i++ )
{
c.s[i] = cvRound(a.s[i]);
c.s[i+n] = 0;
}
return c;
}
template<int n> inline v_reg<int, n*2> v_floor(const v_reg<double, n>& a)
{
v_reg<int, n> c;
for( int i = 0; i < n; i++ )
{
c.s[i] = cvFloor(a.s[i]);
c.s[i+n] = 0;
}
return c;
}
template<int n> inline v_reg<int, n*2> v_ceil(const v_reg<double, n>& a)
{
v_reg<int, n> c;
for( int i = 0; i < n; i++ )
{
c.s[i] = cvCeil(a.s[i]);
c.s[i+n] = 0;
}
return c;
}
template<int n> inline v_reg<int, n*2> v_trunc(const v_reg<double, n>& a)
{
v_reg<int, n> c;
for( int i = 0; i < n; i++ )
{
c.s[i] = cvCeil(a.s[i]);
c.s[i+n] = 0;
}
return c;
}
template<int n> inline v_reg<float, n> v_cvt_f32(const v_reg<int, n>& a)
{
v_reg<float, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = (float)a.s[i];
return c;
}
template<int n> inline v_reg<double, n> v_cvt_f64(const v_reg<int, n*2>& a)
{
v_reg<double, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = (double)a.s[i];
return c;
}
template<int n> inline v_reg<double, n> v_cvt_f64(const v_reg<float, n*2>& a)
{
v_reg<double, n> c;
for( int i = 0; i < n; i++ )
c.s[i] = (double)a.s[i];
return c;
}
template<typename _Tp>
inline void v_transpose4x4( v_reg<_Tp, 4>& a0, const v_reg<_Tp, 4>& a1,
const v_reg<_Tp, 4>& a2, const v_reg<_Tp, 4>& a3,
v_reg<_Tp, 4>& b0, v_reg<_Tp, 4>& b1,
v_reg<_Tp, 4>& b2, v_reg<_Tp, 4>& b3 )
{
b0 = v_reg<_Tp, 4>(a0.s[0], a1.s[0], a2.s[0], a3.s[0]);
b1 = v_reg<_Tp, 4>(a0.s[1], a1.s[1], a2.s[1], a3.s[1]);
b2 = v_reg<_Tp, 4>(a0.s[2], a1.s[2], a2.s[2], a3.s[2]);
b3 = v_reg<_Tp, 4>(a0.s[3], a1.s[3], a2.s[3], a3.s[3]);
}
typedef v_reg<uchar, 16> v_uint8x16;
typedef v_reg<schar, 16> v_int8x16;
typedef v_reg<ushort, 8> v_uint16x8;
typedef v_reg<short, 8> v_int16x8;
typedef v_reg<unsigned, 4> v_uint32x4;
typedef v_reg<int, 4> v_int32x4;
typedef v_reg<float, 4> v_float32x4;
typedef v_reg<float, 8> v_float32x8;
typedef v_reg<double, 2> v_float64x2;
typedef v_reg<uint64, 2> v_uint64x2;
typedef v_reg<int64, 2> v_int64x2;
#define OPENCV_HAL_IMPL_C_INIT(_Tpvec, _Tp, suffix) \
inline _Tpvec v_setzero_##suffix() { return _Tpvec::zero(); } \
inline _Tpvec v_setall_##suffix(_Tp val) { return _Tpvec::all(val); } \
template<typename _Tp0, int n0> inline _Tpvec \
v_reinterpret_as_##suffix(const v_reg<_Tp0, n0>& a) \
{ return a.template reinterpret_as<_Tp, _Tpvec::nlanes>(a); }
OPENCV_HAL_IMPL_C_INIT(v_uint8x16, uchar, u8)
OPENCV_HAL_IMPL_C_INIT(v_int8x16, schar, s8)
OPENCV_HAL_IMPL_C_INIT(v_uint16x8, ushort, u16)
OPENCV_HAL_IMPL_C_INIT(v_int16x8, short, s16)
OPENCV_HAL_IMPL_C_INIT(v_uint32x4, unsigned, u32)
OPENCV_HAL_IMPL_C_INIT(v_int32x4, int, s32)
OPENCV_HAL_IMPL_C_INIT(v_float32x4, float, f32)
OPENCV_HAL_IMPL_C_INIT(v_float64x2, double, f64)
OPENCV_HAL_IMPL_C_INIT(v_uint64x2, uint64, u64)
OPENCV_HAL_IMPL_C_INIT(v_uint64x2, int64, s64)
#define OPENCV_HAL_IMPL_C_SHIFT(_Tpvec, _Tp) \
template<int n> inline _Tpvec v_shl(const _Tpvec& a) \
{ return a << n; } \
template<int n> inline _Tpvec v_shr(const _Tpvec& a) \
{ return a >> n; } \
template<int n> inline _Tpvec v_rshr(const _Tpvec& a) \
{ \
_Tpvec c; \
for( int i = 0; i < _Tpvec::nlanes; i++ ) \
c.s[i] = (_Tp)((a.s[i] + ((_Tp)1 << (n - 1))) >> n); \
return c; \
}
OPENCV_HAL_IMPL_C_SHIFT(v_uint16x8, ushort)
OPENCV_HAL_IMPL_C_SHIFT(v_int16x8, short)
OPENCV_HAL_IMPL_C_SHIFT(v_uint32x4, unsigned)
OPENCV_HAL_IMPL_C_SHIFT(v_int32x4, int)
OPENCV_HAL_IMPL_C_SHIFT(v_uint64x2, uint64)
OPENCV_HAL_IMPL_C_SHIFT(v_int64x2, int64)
#define OPENCV_HAL_IMPL_C_PACK(_Tpvec, _Tp, _Tpnvec, _Tpn, pack_suffix) \
inline _Tpnvec v_##pack_suffix(const _Tpvec& a, const _Tpvec& b) \
{ \
_Tpnvec c; \
for( int i = 0; i < _Tpvec::nlanes; i++ ) \
{ \
c.s[i] = saturate_cast<_Tpn>(a.s[i]); \
c.s[i+_Tpvec::nlanes] = saturate_cast<_Tpn>(b.s[i]); \
} \
return c; \
} \
template<int n> inline _Tpnvec v_rshr_##pack_suffix(const _Tpvec& a, const _Tpvec& b) \
{ \
_Tpnvec c; \
for( int i = 0; i < _Tpvec::nlanes; i++ ) \
{ \
c.s[i] = saturate_cast<_Tpn>((a.s[i] + ((_Tp)1 << (n - 1))) >> n); \
c.s[i+_Tpvec::nlanes] = saturate_cast<_Tpn>((b.s[i] + ((_Tp)1 << (n - 1))) >> n); \
} \
return c; \
} \
inline void v_##pack_suffix##_store(_Tpn* ptr, const _Tpvec& a) \
{ \
for( int i = 0; i < _Tpvec::nlanes; i++ ) \
ptr[i] = saturate_cast<_Tpn>(a.s[i]); \
} \
template<int n> inline void v_rshr_##pack_suffix##_store(_Tpn* ptr, const _Tpvec& a) \
{ \
for( int i = 0; i < _Tpvec::nlanes; i++ ) \
ptr[i] = saturate_cast<_Tpn>((a.s[i] + ((_Tp)1 << (n - 1))) >> n); \
}
OPENCV_HAL_IMPL_C_PACK(v_uint16x8, ushort, v_uint8x16, uchar, pack)
OPENCV_HAL_IMPL_C_PACK(v_int16x8, short, v_int8x16, schar, pack)
OPENCV_HAL_IMPL_C_PACK(v_int16x8, short, v_uint8x16, uchar, pack_u)
OPENCV_HAL_IMPL_C_PACK(v_uint32x4, unsigned, v_uint16x8, ushort, pack)
OPENCV_HAL_IMPL_C_PACK(v_int32x4, int, v_int16x8, short, pack)
OPENCV_HAL_IMPL_C_PACK(v_int32x4, int, v_uint16x8, ushort, pack_u)
OPENCV_HAL_IMPL_C_PACK(v_uint64x2, uint64, v_uint32x4, unsigned, pack)
OPENCV_HAL_IMPL_C_PACK(v_int64x2, int64, v_int32x4, int, pack)
inline v_float32x4 v_matmul(const v_float32x4& v, const v_float32x4& m0,
const v_float32x4& m1, const v_float32x4& m2,
const v_float32x4& m3)
{
return v_float32x4(v.s[0]*m0.s[0] + v.s[1]*m1.s[0] + v.s[2]*m2.s[0] + v.s[3]*m3.s[0],
v.s[0]*m0.s[1] + v.s[1]*m1.s[1] + v.s[2]*m2.s[1] + v.s[3]*m3.s[1],
v.s[0]*m0.s[2] + v.s[1]*m1.s[2] + v.s[2]*m2.s[2] + v.s[3]*m3.s[2],
v.s[0]*m0.s[3] + v.s[1]*m1.s[3] + v.s[2]*m2.s[3] + v.s[3]*m3.s[3]);
}
}
#endif

823
modules/hal/include/opencv2/hal/intrin_neon.hpp Normal file
View File

@ -0,0 +1,823 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#ifndef __OPENCV_HAL_INTRIN_NEON_HPP__
#define __OPENCV_HAL_INTRIN_NEON_HPP__
namespace cv
{
#define CV_SIMD128 1
struct v_uint8x16
{
typedef uchar lane_type;
enum { nlanes = 16 };
v_uint8x16() {}
explicit v_uint8x16(uint8x16_t v) : val(v) {}
v_uint8x16(uchar v0, uchar v1, uchar v2, uchar v3, uchar v4, uchar v5, uchar v6, uchar v7,
uchar v8, uchar v9, uchar v10, uchar v11, uchar v12, uchar v13, uchar v14, uchar v15)
{
uchar v[] = {v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15};
val = vld1q_u8(v);
}
uchar get0() const
{
return vgetq_lane_u8(val, 0);
}
uint8x16_t val;
};
struct v_int8x16
{
typedef schar lane_type;
enum { nlanes = 16 };
v_int8x16() {}
explicit v_int8x16(int8x16_t v) : val(v) {}
v_int8x16(schar v0, schar v1, schar v2, schar v3, schar v4, schar v5, schar v6, schar v7,
schar v8, schar v9, schar v10, schar v11, schar v12, schar v13, schar v14, schar v15)
{
schar v[] = {v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15};
val = vld1q_s8(v);
}
schar get0() const
{
return vgetq_lane_s8(val, 0);
}
int8x16_t val;
};
struct v_uint16x8
{
typedef ushort lane_type;
enum { nlanes = 8 };
v_uint16x8() {}
explicit v_uint16x8(uint16x8_t v) : val(v) {}
v_uint16x8(ushort v0, ushort v1, ushort v2, ushort v3, ushort v4, ushort v5, ushort v6, ushort v7)
{
ushort v[] = {v0, v1, v2, v3, v4, v5, v6, v7};
val = vld1q_u16(v);
}
ushort get0() const
{
return vgetq_lane_u16(val, 0);
}
uint16x8_t val;
};
struct v_int16x8
{
typedef short lane_type;
enum { nlanes = 8 };
v_int16x8() {}
explicit v_int16x8(int16x8_t v) : val(v) {}
v_int16x8(short v0, short v1, short v2, short v3, short v4, short v5, short v6, short v7)
{
short v[] = {v0, v1, v2, v3, v4, v5, v6, v7};
val = vld1q_s16(v);
}
short get0() const
{
return vgetq_lane_s16(val, 0);
}
int16x8_t val;
};
struct v_uint32x4
{
typedef unsigned lane_type;
enum { nlanes = 4 };
v_uint32x4() {}
explicit v_uint32x4(uint32x4_t v) : val(v) {}
v_uint32x4(unsigned v0, unsigned v1, unsigned v2, unsigned v3)
{
unsigned v[] = {v0, v1, v2, v3};
val = vld1q_u32(v);
}
unsigned get0() const
{
return vgetq_lane_u32(val, 0);
}
uint32x4_t val;
};
struct v_int32x4
{
typedef int lane_type;
enum { nlanes = 4 };
v_int32x4() {}
explicit v_int32x4(int32x4_t v) : val(v) {}
v_int32x4(int v0, int v1, int v2, int v3)
{
int v[] = {v0, v1, v2, v3};
val = vld1q_s32(v);
}
int get0() const
{
return vgetq_lane_s32(val, 0);
}
int32x4_t val;
};
struct v_float32x4
{
typedef float lane_type;
enum { nlanes = 4 };
v_float32x4() {}
explicit v_float32x4(float32x4_t v) : val(v) {}
v_float32x4(float v0, float v1, float v2, float v3)
{
float v[] = {v0, v1, v2, v3};
val = vld1q_f32(v);
}
float get0() const
{
return vgetq_lane_f32(val, 0);
}
float32x4_t val;
};
struct v_uint64x2
{
typedef uint64 lane_type;
enum { nlanes = 2 };
v_uint64x2() {}
explicit v_uint64x2(uint64x2_t v) : val(v) {}
v_uint64x2(unsigned v0, unsigned v1)
{
uint64 v[] = {v0, v1};
val = vld1q_u64(v);
}
uint64 get0() const
{
return vgetq_lane_u64(val, 0);
}
uint64x2_t val;
};
struct v_int64x2
{
typedef int64 lane_type;
enum { nlanes = 2 };
v_int64x2() {}
explicit v_int64x2(int64x2_t v) : val(v) {}
v_int64x2(int v0, int v1)
{
int64 v[] = {v0, v1};
val = vld1q_s64(v);
}
int64 get0() const
{
return vgetq_lane_s64(val, 0);
}
int64x2_t val;
};
#define OPENCV_HAL_IMPL_NEON_INIT(_Tpv, _Tp, suffix) \
inline v_##_Tpv v_setzero_##suffix() { return v_##_Tpv(vdupq_n_##suffix((_Tp)0)); } \
inline v_##_Tpv v_setall_##suffix(_Tp v) { return v_##_Tpv(vdupq_n_##suffix(v)); } \
inline _Tpv##_t vreinterpretq_##suffix##_##suffix(_Tpv##_t v) { return v; } \
inline v_uint8x16 v_reinterpret_as_u8(const v_##_Tpv& v) { return v_uint8x16(vreinterpretq_u8_##suffix(v.val)); } \
inline v_int8x16 v_reinterpret_as_s8(const v_##_Tpv& v) { return v_int8x16(vreinterpretq_s8_##suffix(v.val)); } \
inline v_uint16x8 v_reinterpret_as_u16(const v_##_Tpv& v) { return v_uint16x8(vreinterpretq_u16_##suffix(v.val)); } \
inline v_int16x8 v_reinterpret_as_s16(const v_##_Tpv& v) { return v_int16x8(vreinterpretq_s16_##suffix(v.val)); } \
inline v_uint32x4 v_reinterpret_as_u32(const v_##_Tpv& v) { return v_uint32x4(vreinterpretq_u32_##suffix(v.val)); } \
inline v_int32x4 v_reinterpret_as_s32(const v_##_Tpv& v) { return v_int32x4(vreinterpretq_s32_##suffix(v.val)); } \
inline v_uint64x2 v_reinterpret_as_u64(const v_##_Tpv& v) { return v_uint64x2(vreinterpretq_u64_##suffix(v.val)); } \
inline v_int64x2 v_reinterpret_as_s64(const v_##_Tpv& v) { return v_int64x2(vreinterpretq_s64_##suffix(v.val)); } \
inline v_float32x4 v_reinterpret_as_f32(const v_##_Tpv& v) { return v_float32x4(vreinterpretq_f32_##suffix(v.val)); }
OPENCV_HAL_IMPL_NEON_INIT(uint8x16, uchar, u8)
OPENCV_HAL_IMPL_NEON_INIT(int8x16, schar, s8)
OPENCV_HAL_IMPL_NEON_INIT(uint16x8, ushort, u16)
OPENCV_HAL_IMPL_NEON_INIT(int16x8, short, s16)
OPENCV_HAL_IMPL_NEON_INIT(uint32x4, unsigned, u32)
OPENCV_HAL_IMPL_NEON_INIT(int32x4, int, s32)
OPENCV_HAL_IMPL_NEON_INIT(uint64x2, uint64, u64)
OPENCV_HAL_IMPL_NEON_INIT(int64x2, int64, s64)
OPENCV_HAL_IMPL_NEON_INIT(float32x4, float, f32)
#define OPENCV_HAL_IMPL_NEON_PACK(_Tpvec, _Tp, hreg, suffix, _Tpwvec, wsuffix, pack, op) \
inline _Tpvec v_##pack(const _Tpwvec& a, const _Tpwvec& b) \
{ \
hreg a1 = vqmov##op##_##wsuffix(a.val), b1 = vqmov##op##_##wsuffix(b.val); \
return _Tpvec(vcombine_##suffix(a1, b1)); \
} \
inline void v_##pack##_store(_Tp* ptr, const _Tpwvec& a) \
{ \
hreg a1 = vqmov##op##_##wsuffix(a.val); \
vst1_##suffix(ptr, a1); \
} \
template<int n> inline \
_Tpvec v_rshr_##pack(const _Tpwvec& a, const _Tpwvec& b) \
{ \
hreg a1 = vqrshr##op##_n_##wsuffix(a.val, n); \
hreg b1 = vqrshr##op##_n_##wsuffix(b.val, n); \
return _Tpvec(vcombine_##suffix(a1, b1)); \
} \
template<int n> inline \
void v_rshr_##pack##_store(_Tp* ptr, const _Tpwvec& a) \
{ \
hreg a1 = vqrshr##op##_n_##wsuffix(a.val, n); \
vst1_##suffix(ptr, a1); \
}
OPENCV_HAL_IMPL_NEON_PACK(v_uint8x16, uchar, uint8x8_t, u8, v_uint16x8, u16, pack, n)
OPENCV_HAL_IMPL_NEON_PACK(v_uint8x16, uchar, uint8x8_t, u8, v_int16x8, s16, pack_u, un)
OPENCV_HAL_IMPL_NEON_PACK(v_int8x16, schar, int8x8_t, s8, v_int16x8, s16, pack, n)
OPENCV_HAL_IMPL_NEON_PACK(v_uint16x8, ushort, uint16x4_t, u16, v_uint32x4, u32, pack, n)
OPENCV_HAL_IMPL_NEON_PACK(v_uint16x8, ushort, uint16x4_t, u16, v_int32x4, s32, pack_u, un)
OPENCV_HAL_IMPL_NEON_PACK(v_int16x8, short, int16x4_t, s16, v_int32x4, s32, pack, n)
OPENCV_HAL_IMPL_NEON_PACK(v_uint32x4, unsigned, uint32x2_t, u32, v_uint64x2, u64, pack, n)
OPENCV_HAL_IMPL_NEON_PACK(v_int32x4, int, int32x2_t, s32, v_int64x2, s64, pack, n)
inline v_float32x4 v_matmul(const v_float32x4& v, const v_float32x4& m0,
const v_float32x4& m1, const v_float32x4& m2,
const v_float32x4& m3)
{
float32x2_t vl = vget_low_f32(v.val), vh = vget_high_f32(v.val);
float32x4_t res = vmulq_lane_f32(m0.val, vl, 0);
res = vmlaq_lane_f32(res, m1.val, vl, 1);
res = vmlaq_lane_f32(res, m2.val, vh, 0);
res = vmlaq_lane_f32(res, m3.val, vh, 1);
return v_float32x4(res);
}
#define OPENCV_HAL_IMPL_NEON_BIN_OP(bin_op, _Tpvec, intrin) \
inline _Tpvec operator bin_op (const _Tpvec& a, const _Tpvec& b) \
{ \
return _Tpvec(intrin(a.val, b.val)); \
} \
inline _Tpvec& operator bin_op##= (_Tpvec& a, const _Tpvec& b) \
{ \
a.val = intrin(a.val, b.val); \
return a; \
}
OPENCV_HAL_IMPL_NEON_BIN_OP(+, v_uint8x16, vqaddq_u8)
OPENCV_HAL_IMPL_NEON_BIN_OP(-, v_uint8x16, vqsubq_u8)
OPENCV_HAL_IMPL_NEON_BIN_OP(+, v_int8x16, vqaddq_s8)
OPENCV_HAL_IMPL_NEON_BIN_OP(-, v_int8x16, vqsubq_s8)
OPENCV_HAL_IMPL_NEON_BIN_OP(+, v_uint16x8, vqaddq_u16)
OPENCV_HAL_IMPL_NEON_BIN_OP(-, v_uint16x8, vqsubq_u16)
OPENCV_HAL_IMPL_NEON_BIN_OP(*, v_uint16x8, vmulq_u16)
OPENCV_HAL_IMPL_NEON_BIN_OP(+, v_int16x8, vqaddq_s16)
OPENCV_HAL_IMPL_NEON_BIN_OP(-, v_int16x8, vqsubq_s16)
OPENCV_HAL_IMPL_NEON_BIN_OP(*, v_int16x8, vmulq_s16)
OPENCV_HAL_IMPL_NEON_BIN_OP(+, v_int32x4, vaddq_s32)
OPENCV_HAL_IMPL_NEON_BIN_OP(-, v_int32x4, vsubq_s32)
OPENCV_HAL_IMPL_NEON_BIN_OP(*, v_int32x4, vmulq_s32)
OPENCV_HAL_IMPL_NEON_BIN_OP(+, v_float32x4, vaddq_f32)
OPENCV_HAL_IMPL_NEON_BIN_OP(-, v_float32x4, vsubq_f32)
OPENCV_HAL_IMPL_NEON_BIN_OP(*, v_float32x4, vmulq_f32)
OPENCV_HAL_IMPL_NEON_BIN_OP(+, v_int64x2, vaddq_s64)
OPENCV_HAL_IMPL_NEON_BIN_OP(-, v_int64x2, vsubq_s64)
OPENCV_HAL_IMPL_NEON_BIN_OP(+, v_uint64x2, vaddq_u64)
OPENCV_HAL_IMPL_NEON_BIN_OP(-, v_uint64x2, vsubq_u64)
inline v_float32x4 operator / (const v_float32x4& a, const v_float32x4& b)
{
float32x4_t reciprocal = vrecpeq_f32(b.val);
reciprocal = vmulq_f32(vrecpsq_f32(b.val, reciprocal), reciprocal);
reciprocal = vmulq_f32(vrecpsq_f32(b.val, reciprocal), reciprocal);
return v_float32x4(vmulq_f32(a.val, reciprocal));
}
inline v_float32x4& operator /= (v_float32x4& a, const v_float32x4& b)
{
float32x4_t reciprocal = vrecpeq_f32(b.val);
reciprocal = vmulq_f32(vrecpsq_f32(b.val, reciprocal), reciprocal);
reciprocal = vmulq_f32(vrecpsq_f32(b.val, reciprocal), reciprocal);
a.val = vmulq_f32(a.val, reciprocal);
return a;
}
inline void v_mul_expand(const v_int16x8& a, const v_int16x8& b,
v_int32x4& c, v_int32x4& d)
{
c.val = vmull_s16(vget_low_s16(a.val), vget_low_s16(b.val));
d.val = vmull_s16(vget_high_s16(a.val), vget_high_s16(b.val));
}
inline void v_mul_expand(const v_uint16x8& a, const v_uint16x8& b,
v_uint32x4& c, v_uint32x4& d)
{
c.val = vmull_u16(vget_low_u16(a.val), vget_low_u16(b.val));
d.val = vmull_u16(vget_high_u16(a.val), vget_high_u16(b.val));
}
inline void v_mul_expand(const v_uint32x4& a, const v_uint32x4& b,
v_uint64x2& c, v_uint64x2& d)
{
c.val = vmull_u32(vget_low_u32(a.val), vget_low_u32(b.val));
d.val = vmull_u32(vget_high_u32(a.val), vget_high_u32(b.val));
}
inline v_int32x4 v_dotprod(const v_int16x8& a, const v_int16x8& b)
{
int32x4_t c = vmull_s16(vget_low_s16(a.val), vget_low_s16(b.val));
int32x4_t d = vmull_s16(vget_high_s16(a.val), vget_high_s16(b.val));
int32x4x2_t cd = vtrnq_s32(c, d);
return v_int32x4(vaddq_s32(cd.val[0], cd.val[1]));
}
#define OPENCV_HAL_IMPL_NEON_LOGIC_OP(_Tpvec, suffix) \
OPENCV_HAL_IMPL_NEON_BIN_OP(&, _Tpvec, vandq_##suffix) \
OPENCV_HAL_IMPL_NEON_BIN_OP(|, _Tpvec, vorrq_##suffix) \
OPENCV_HAL_IMPL_NEON_BIN_OP(^, _Tpvec, veorq_##suffix) \
inline _Tpvec operator ~ (const _Tpvec& a) \
{ \
return _Tpvec(vreinterpretq_##suffix##_u8(vmvnq_u8(vreinterpretq_u8_##suffix(a.val)))); \
}
OPENCV_HAL_IMPL_NEON_LOGIC_OP(v_uint8x16, u8)
OPENCV_HAL_IMPL_NEON_LOGIC_OP(v_int8x16, s8)
OPENCV_HAL_IMPL_NEON_LOGIC_OP(v_uint16x8, u16)
OPENCV_HAL_IMPL_NEON_LOGIC_OP(v_int16x8, s16)
OPENCV_HAL_IMPL_NEON_LOGIC_OP(v_uint32x4, u32)
OPENCV_HAL_IMPL_NEON_LOGIC_OP(v_int32x4, s32)
OPENCV_HAL_IMPL_NEON_LOGIC_OP(v_uint64x2, u64)
OPENCV_HAL_IMPL_NEON_LOGIC_OP(v_int64x2, s64)
#define OPENCV_HAL_IMPL_NEON_FLT_BIT_OP(bin_op, intrin) \
inline v_float32x4 operator bin_op (const v_float32x4& a, const v_float32x4& b) \
{ \
return v_float32x4(vreinterpretq_f32_s32(intrin(vreinterpretq_s32_f32(a.val), vreinterpretq_s32_f32(b.val)))); \
} \
inline v_float32x4& operator bin_op##= (v_float32x4& a, const v_float32x4& b) \
{ \
a.val = vreinterpretq_f32_s32(intrin(vreinterpretq_s32_f32(a.val), vreinterpretq_s32_f32(b.val))); \
return a; \
}
OPENCV_HAL_IMPL_NEON_FLT_BIT_OP(&, vandq_s32)
OPENCV_HAL_IMPL_NEON_FLT_BIT_OP(|, vorrq_s32)
OPENCV_HAL_IMPL_NEON_FLT_BIT_OP(^, veorq_s32)
inline v_float32x4 operator ~ (const v_float32x4& a)
{
return v_float32x4(vreinterpretq_f32_s32(vmvnq_s32(vreinterpretq_s32_f32(a.val))));
}
inline v_float32x4 v_sqrt(const v_float32x4& x)
{
float32x4_t x1 = vmaxq_f32(x.val, vdupq_n_f32(FLT_MIN));
float32x4_t e = vrsqrteq_f32(x1);
e = vmulq_f32(vrsqrtsq_f32(vmulq_f32(x1, e), e), e);
e = vmulq_f32(vrsqrtsq_f32(vmulq_f32(x1, e), e), e);
return v_float32x4(vmulq_f32(x.val, e));
}
inline v_float32x4 v_invsqrt(const v_float32x4& x)
{
float32x4_t e = vrsqrteq_f32(x.val);
e = vmulq_f32(vrsqrtsq_f32(vmulq_f32(x.val, e), e), e);
e = vmulq_f32(vrsqrtsq_f32(vmulq_f32(x.val, e), e), e);
return v_float32x4(e);
}
inline v_float32x4 v_abs(v_float32x4 x)
{ return v_float32x4(vabsq_f32(x.val)); }
// TODO: exp, log, sin, cos
#define OPENCV_HAL_IMPL_NEON_BIN_FUNC(_Tpvec, func, intrin) \
inline _Tpvec func(const _Tpvec& a, const _Tpvec& b) \
{ \
return _Tpvec(intrin(a.val, b.val)); \
}
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint8x16, v_min, vminq_u8)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint8x16, v_max, vmaxq_u8)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int8x16, v_min, vminq_s8)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int8x16, v_max, vmaxq_s8)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint16x8, v_min, vminq_u16)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint16x8, v_max, vmaxq_u16)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int16x8, v_min, vminq_s16)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int16x8, v_max, vmaxq_s16)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint32x4, v_min, vminq_u32)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint32x4, v_max, vmaxq_u32)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int32x4, v_min, vminq_s32)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int32x4, v_max, vmaxq_s32)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_float32x4, v_min, vminq_f32)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_float32x4, v_max, vmaxq_f32)
#define OPENCV_HAL_IMPL_NEON_INT_CMP_OP(_Tpvec, cast, suffix, not_suffix) \
inline _Tpvec operator == (const _Tpvec& a, const _Tpvec& b) \
{ return _Tpvec(cast(vceqq_##suffix(a.val, b.val))); } \
inline _Tpvec operator != (const _Tpvec& a, const _Tpvec& b) \
{ return _Tpvec(cast(vmvnq_##not_suffix(vceqq_##suffix(a.val, b.val)))); } \
inline _Tpvec operator < (const _Tpvec& a, const _Tpvec& b) \
{ return _Tpvec(cast(vcltq_##suffix(a.val, b.val))); } \
inline _Tpvec operator > (const _Tpvec& a, const _Tpvec& b) \
{ return _Tpvec(cast(vcgtq_##suffix(a.val, b.val))); } \
inline _Tpvec operator <= (const _Tpvec& a, const _Tpvec& b) \
{ return _Tpvec(cast(vcleq_##suffix(a.val, b.val))); } \
inline _Tpvec operator >= (const _Tpvec& a, const _Tpvec& b) \
{ return _Tpvec(cast(vcgeq_##suffix(a.val, b.val))); }
OPENCV_HAL_IMPL_NEON_INT_CMP_OP(v_uint8x16, OPENCV_HAL_NOP, u8, u8)
OPENCV_HAL_IMPL_NEON_INT_CMP_OP(v_int8x16, vreinterpretq_s8_u8, s8, u8)
OPENCV_HAL_IMPL_NEON_INT_CMP_OP(v_uint16x8, OPENCV_HAL_NOP, u16, u16)
OPENCV_HAL_IMPL_NEON_INT_CMP_OP(v_int16x8, vreinterpretq_s16_u16, s16, u16)
OPENCV_HAL_IMPL_NEON_INT_CMP_OP(v_uint32x4, OPENCV_HAL_NOP, u32, u32)
OPENCV_HAL_IMPL_NEON_INT_CMP_OP(v_int32x4, vreinterpretq_s32_u32, s32, u32)
OPENCV_HAL_IMPL_NEON_INT_CMP_OP(v_float32x4, vreinterpretq_f32_u32, f32, u32)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint8x16, v_add_wrap, vaddq_u8)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int8x16, v_add_wrap, vaddq_s8)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint16x8, v_add_wrap, vaddq_u16)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int16x8, v_add_wrap, vaddq_s16)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint8x16, v_sub_wrap, vsubq_u8)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int8x16, v_sub_wrap, vsubq_s8)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint16x8, v_sub_wrap, vsubq_u16)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_int16x8, v_sub_wrap, vsubq_s16)
// TODO: absdiff for signed integers
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint8x16, v_absdiff, vabdq_u8)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint16x8, v_absdiff, vabdq_u16)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_uint32x4, v_absdiff, vabdq_u32)
OPENCV_HAL_IMPL_NEON_BIN_FUNC(v_float32x4, v_absdiff, vabdq_f32)
inline v_float32x4 v_magnitude(const v_float32x4& a, const v_float32x4& b)
{
v_float32x4 x(vmlaq_f32(vmulq_f32(a.val, a.val), b.val, b.val));
return v_sqrt(x);
}
inline v_float32x4 v_sqr_magnitude(const v_float32x4& a, const v_float32x4& b)
{
return v_float32x4(vmlaq_f32(vmulq_f32(a.val, a.val), b.val, b.val));
}
inline v_float32x4 v_muladd(const v_float32x4& a, const v_float32x4& b, const v_float32x4& c)
{
return v_float32x4(vmlaq_f32(c.val, a.val, b.val));
}
// trade efficiency for convenience
#define OPENCV_HAL_IMPL_NEON_SHIFT_OP(_Tpvec, suffix, _Tps, ssuffix) \
inline _Tpvec operator << (const _Tpvec& a, int n) \
{ return _Tpvec(vshlq_##suffix(a.val, vdupq_n_##ssuffix((_Tps)n))); } \
inline _Tpvec operator >> (const _Tpvec& a, int n) \
{ return _Tpvec(vshlq_##suffix(a.val, vdupq_n_##ssuffix((_Tps)-n))); } \
template<int n> inline _Tpvec v_shl(const _Tpvec& a) \
{ return _Tpvec(vshlq_n_##suffix(a.val, n)); } \
template<int n> inline _Tpvec v_shr(const _Tpvec& a) \
{ return _Tpvec(vshrq_n_##suffix(a.val, n)); } \
template<int n> inline _Tpvec v_rshr(const _Tpvec& a) \
{ return _Tpvec(vrshrq_n_##suffix(a.val, n)); }
OPENCV_HAL_IMPL_NEON_SHIFT_OP(v_uint8x16, u8, schar, s8)
OPENCV_HAL_IMPL_NEON_SHIFT_OP(v_int8x16, s8, schar, s8)
OPENCV_HAL_IMPL_NEON_SHIFT_OP(v_uint16x8, u16, short, s16)
OPENCV_HAL_IMPL_NEON_SHIFT_OP(v_int16x8, s16, short, s16)
OPENCV_HAL_IMPL_NEON_SHIFT_OP(v_uint32x4, u32, int, s32)
OPENCV_HAL_IMPL_NEON_SHIFT_OP(v_int32x4, s32, int, s32)
OPENCV_HAL_IMPL_NEON_SHIFT_OP(v_uint64x2, u64, int64, s64)
OPENCV_HAL_IMPL_NEON_SHIFT_OP(v_int64x2, s64, int64, s64)
#define OPENCV_HAL_IMPL_NEON_LOADSTORE_OP(_Tpvec, _Tp, suffix) \
inline _Tpvec v_load(const _Tp* ptr) \
{ return _Tpvec(vld1q_##suffix(ptr)); } \
inline _Tpvec v_load_aligned(const _Tp* ptr) \
{ return _Tpvec(vld1q_##suffix(ptr)); } \
inline _Tpvec v_load_halves(const _Tp* ptr0, const _Tp* ptr1) \
{ return _Tpvec(vcombine_##suffix(vld1_##suffix(ptr0), vld1_##suffix(ptr1))); } \
inline void v_store(_Tp* ptr, const _Tpvec& a) \
{ vst1q_##suffix(ptr, a.val); } \
inline void v_store_aligned(_Tp* ptr, const _Tpvec& a) \
{ vst1q_##suffix(ptr, a.val); } \
inline void v_store_low(_Tp* ptr, const _Tpvec& a) \
{ vst1_##suffix(ptr, vget_low_##suffix(a.val)); } \
inline void v_store_high(_Tp* ptr, const _Tpvec& a) \
{ vst1_##suffix(ptr, vget_high_##suffix(a.val)); }
OPENCV_HAL_IMPL_NEON_LOADSTORE_OP(v_uint8x16, uchar, u8)
OPENCV_HAL_IMPL_NEON_LOADSTORE_OP(v_int8x16, schar, s8)
OPENCV_HAL_IMPL_NEON_LOADSTORE_OP(v_uint16x8, ushort, u16)
OPENCV_HAL_IMPL_NEON_LOADSTORE_OP(v_int16x8, short, s16)
OPENCV_HAL_IMPL_NEON_LOADSTORE_OP(v_uint32x4, unsigned, u32)
OPENCV_HAL_IMPL_NEON_LOADSTORE_OP(v_int32x4, int, s32)
OPENCV_HAL_IMPL_NEON_LOADSTORE_OP(v_float32x4, float, f32)
#define OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(_Tpvec, scalartype, func, scalar_func) \
inline scalartype v_reduce_##func(const _Tpvec& a) \
{ \
scalartype CV_DECL_ALIGNED(16) buf[4]; \
v_store_aligned(buf, a); \
scalartype s0 = scalar_func(buf[0], buf[1]); \
scalartype s1 = scalar_func(buf[2], buf[3]); \
return scalar_func(s0, s1); \
}
OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_uint32x4, unsigned, sum, OPENCV_HAL_ADD)
OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_uint32x4, unsigned, max, std::max)
OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_uint32x4, unsigned, min, std::min)
OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_int32x4, int, sum, OPENCV_HAL_ADD)
OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_int32x4, int, max, std::max)
OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_int32x4, int, min, std::min)
OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_float32x4, float, sum, OPENCV_HAL_ADD)
OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_float32x4, float, max, std::max)
OPENCV_HAL_IMPL_NEON_REDUCE_OP_4(v_float32x4, float, min, std::min)
inline int v_signmask(const v_uint8x16& a)
{
int8x8_t m0 = vcreate_s8(CV_BIG_UINT(0x0706050403020100));
uint8x16_t v0 = vshlq_u8(vshrq_n_u8(a.val, 7), vcombine_s8(m0, m0));
uint64x2_t v1 = vpaddlq_u32(vpaddlq_u16(vpaddlq_u8(v0)));
return (int)vgetq_lane_u64(v1, 0) + ((int)vgetq_lane_u64(v1, 1) << 8);
}
inline int v_signmask(const v_int8x16& a)
{ return v_signmask(v_reinterpret_as_u8(a)); }
inline int v_signmask(const v_uint16x8& a)
{
int16x4_t m0 = vcreate_s16(CV_BIG_UINT(0x0003000200010000));
uint16x8_t v0 = vshlq_u16(vshrq_n_u16(a.val, 15), vcombine_s16(m0, m0));
uint64x2_t v1 = vpaddlq_u32(vpaddlq_u16(v0));
return (int)vgetq_lane_u64(v1, 0) + ((int)vgetq_lane_u64(v1, 1) << 4);
}
inline int v_signmask(const v_int16x8& a)
{ return v_signmask(v_reinterpret_as_u16(a)); }
inline int v_signmask(const v_uint32x4& a)
{
int32x2_t m0 = vcreate_s32(CV_BIG_UINT(0x0000000100000000));
uint32x4_t v0 = vshlq_u32(vshrq_n_u32(a.val, 31), vcombine_s32(m0, m0));
uint64x2_t v1 = vpaddlq_u32(v0);
return (int)vgetq_lane_u64(v1, 0) + ((int)vgetq_lane_u64(v1, 1) << 2);
}
inline int v_signmask(const v_int32x4& a)
{ return v_signmask(v_reinterpret_as_u32(a)); }
inline int v_signmask(const v_float32x4& a)
{ return v_signmask(v_reinterpret_as_u32(a)); }
#define OPENCV_HAL_IMPL_NEON_CHECK_ALLANY(_Tpvec, suffix, shift) \
inline bool v_check_all(const v_##_Tpvec& a) \
{ \
_Tpvec##_t v0 = vshrq_n_##suffix(vmvnq_##suffix(a.val), shift); \
uint64x2_t v1 = vreinterpretq_u64_##suffix(v0); \
return (vgetq_lane_u64(v1, 0) | vgetq_lane_u64(v1, 1)) == 0; \
} \
inline bool v_check_any(const v_##_Tpvec& a) \
{ \
_Tpvec##_t v0 = vshrq_n_##suffix(a.val, shift); \
uint64x2_t v1 = vreinterpretq_u64_##suffix(v0); \
return (vgetq_lane_u64(v1, 0) | vgetq_lane_u64(v1, 1)) != 0; \
}
OPENCV_HAL_IMPL_NEON_CHECK_ALLANY(uint8x16, u8, 7)
OPENCV_HAL_IMPL_NEON_CHECK_ALLANY(uint16x8, u16, 15)
OPENCV_HAL_IMPL_NEON_CHECK_ALLANY(uint32x4, u32, 31)
inline bool v_check_all(const v_int8x16& a)
{ return v_check_all(v_reinterpret_as_u8(a)); }
inline bool v_check_all(const v_int16x8& a)
{ return v_check_all(v_reinterpret_as_u16(a)); }
inline bool v_check_all(const v_int32x4& a)
{ return v_check_all(v_reinterpret_as_u32(a)); }
inline bool v_check_all(const v_float32x4& a)
{ return v_check_all(v_reinterpret_as_u32(a)); }
inline bool v_check_any(const v_int8x16& a)
{ return v_check_all(v_reinterpret_as_u8(a)); }
inline bool v_check_any(const v_int16x8& a)
{ return v_check_all(v_reinterpret_as_u16(a)); }
inline bool v_check_any(const v_int32x4& a)
{ return v_check_all(v_reinterpret_as_u32(a)); }
inline bool v_check_any(const v_float32x4& a)
{ return v_check_all(v_reinterpret_as_u32(a)); }
#define OPENCV_HAL_IMPL_NEON_SELECT(_Tpvec, suffix, usuffix) \
inline _Tpvec v_select(const _Tpvec& mask, const _Tpvec& a, const _Tpvec& b) \
{ \
return _Tpvec(vbslq_##suffix(vreinterpretq_##usuffix##_##suffix(mask.val), a.val, b.val)); \
}
OPENCV_HAL_IMPL_NEON_SELECT(v_uint8x16, u8, u8)
OPENCV_HAL_IMPL_NEON_SELECT(v_int8x16, s8, u8)
OPENCV_HAL_IMPL_NEON_SELECT(v_uint16x8, u16, u16)
OPENCV_HAL_IMPL_NEON_SELECT(v_int16x8, s16, u16)
OPENCV_HAL_IMPL_NEON_SELECT(v_uint32x4, u32, u32)
OPENCV_HAL_IMPL_NEON_SELECT(v_int32x4, s32, u32)
OPENCV_HAL_IMPL_NEON_SELECT(v_float32x4, f32, u32)
#define OPENCV_HAL_IMPL_NEON_EXPAND(_Tpvec, _Tpwvec, _Tp, suffix) \
inline void v_expand(const _Tpvec& a, _Tpwvec& b0, _Tpwvec& b1) \
{ \
b0.val = vmovl_##suffix(vget_low_##suffix(a.val)); \
b1.val = vmovl_##suffix(vget_high_##suffix(a.val)); \
} \
inline _Tpwvec v_load_expand(const _Tp* ptr) \
{ \
return _Tpwvec(vmovl_##suffix(vld1_##suffix(ptr))); \
}
OPENCV_HAL_IMPL_NEON_EXPAND(v_uint8x16, v_uint16x8, uchar, u8)
OPENCV_HAL_IMPL_NEON_EXPAND(v_int8x16, v_int16x8, schar, s8)
OPENCV_HAL_IMPL_NEON_EXPAND(v_uint16x8, v_uint32x4, ushort, u16)
OPENCV_HAL_IMPL_NEON_EXPAND(v_int16x8, v_int32x4, short, s16)
inline v_uint32x4 v_load_expand_q(const uchar* ptr)
{
uint8x8_t v0 = vcreate_u8(*(unsigned*)ptr);
uint16x4_t v1 = vget_low_u16(vmovl_u8(v0));
return v_uint32x4(vmovl_u16(v1));
}
inline v_int32x4 v_load_expand_q(const schar* ptr)
{
int8x8_t v0 = vcreate_s8(*(unsigned*)ptr);
int16x4_t v1 = vget_low_s16(vmovl_s8(v0));
return v_int32x4(vmovl_s16(v1));
}
#define OPENCV_HAL_IMPL_NEON_UNPACKS(_Tpvec, suffix) \
inline void v_zip(const v_##_Tpvec& a0, const v_##_Tpvec& a1, v_##_Tpvec& b0, v_##_Tpvec& b1) \
{ \
_Tpvec##x2_t p = vzipq_##suffix(a0.val, a1.val); \
b0.val = p.val[0]; \
b1.val = p.val[1]; \
} \
inline v_##_Tpvec v_combine_low(const v_##_Tpvec& a, const v_##_Tpvec& b) \
{ \
return v_##_Tpvec(vcombine_##suffix(vget_low_##suffix(a.val), vget_low_##suffix(b.val))); \
} \
inline v_##_Tpvec v_combine_high(const v_##_Tpvec& a, const v_##_Tpvec& b) \
{ \
return v_##_Tpvec(vcombine_##suffix(vget_high_##suffix(a.val), vget_high_##suffix(b.val))); \
} \
inline void v_recombine(const v_##_Tpvec& a, const v_##_Tpvec& b, v_##_Tpvec& c, v_##_Tpvec& d) \
{ \
c.val = vcombine_##suffix(vget_low_##suffix(a.val), vget_low_##suffix(b.val)); \
d.val = vcombine_##suffix(vget_high_##suffix(a.val), vget_high_##suffix(b.val)); \
}
OPENCV_HAL_IMPL_NEON_UNPACKS(uint8x16, u8)
OPENCV_HAL_IMPL_NEON_UNPACKS(int8x16, s8)
OPENCV_HAL_IMPL_NEON_UNPACKS(uint16x8, u16)
OPENCV_HAL_IMPL_NEON_UNPACKS(int16x8, s16)
OPENCV_HAL_IMPL_NEON_UNPACKS(uint32x4, u32)
OPENCV_HAL_IMPL_NEON_UNPACKS(int32x4, s32)
OPENCV_HAL_IMPL_NEON_UNPACKS(float32x4, f32)
inline v_int32x4 v_round(const v_float32x4& a)
{
static const int32x4_t v_sign = vdupq_n_s32(1 << 31),
v_05 = vreinterpretq_s32_f32(vdupq_n_f32(0.5f));
int32x4_t v_addition = vorrq_s32(v_05, vandq_s32(v_sign, vreinterpretq_s32_f32(a.val)));
return v_int32x4(vcvtq_s32_f32(vaddq_f32(a.val, vreinterpretq_f32_s32(v_addition))));
}
inline v_int32x4 v_floor(const v_float32x4& a)
{
int32x4_t a1 = vcvtq_s32_f32(a.val);
uint32x4_t mask = vcgtq_f32(vcvtq_f32_s32(a1), a.val);
return v_int32x4(vaddq_s32(a1, vreinterpretq_s32_u32(mask)));
}
inline v_int32x4 v_ceil(const v_float32x4& a)
{
int32x4_t a1 = vcvtq_s32_f32(a.val);
uint32x4_t mask = vcgtq_f32(a.val, vcvtq_f32_s32(a1));
return v_int32x4(vsubq_s32(a1, vreinterpretq_s32_u32(mask)));
}
inline v_int32x4 v_trunc(const v_float32x4& a)
{ return v_int32x4(vcvtq_s32_f32(a.val)); }
#define OPENCV_HAL_IMPL_NEON_TRANSPOSE4x4(_Tpvec, suffix) \
inline void transpose4x4(const v_##_Tpvec& a0, const v_##_Tpvec& a1, \
const v_##_Tpvec& a2, const v_##_Tpvec& a3, \
v_##_Tpvec& b0, v_##_Tpvec& b1, \
v_##_Tpvec& b2, v_##_Tpvec& b3) \
{ \
/* m00 m01 m02 m03 */ \
/* m10 m11 m12 m13 */ \
/* m20 m21 m22 m23 */ \
/* m30 m31 m32 m33 */ \
_Tpvec##x2_t t0 = vtrnq_##suffix(a0.val, a1.val); \
_Tpvec##x2_t t1 = vtrnq_##suffix(a2.val, a3.val); \
/* m00 m10 m02 m12 */ \
/* m01 m11 m03 m13 */ \
/* m20 m30 m22 m32 */ \
/* m21 m31 m23 m33 */ \
b0.val = vcombine_##suffix(vget_low_##suffix(t0.val[0]), vget_low_##suffix(t1.val[0])); \
b1.val = vcombine_##suffix(vget_low_##suffix(t0.val[1]), vget_low_##suffix(t1.val[1])); \
b2.val = vcombine_##suffix(vget_high_##suffix(t0.val[0]), vget_high_##suffix(t1.val[0])); \
b3.val = vcombine_##suffix(vget_high_##suffix(t0.val[1]), vget_high_##suffix(t1.val[1])); \
}
OPENCV_HAL_IMPL_NEON_TRANSPOSE4x4(uint32x4, u32)
OPENCV_HAL_IMPL_NEON_TRANSPOSE4x4(int32x4, s32)
OPENCV_HAL_IMPL_NEON_TRANSPOSE4x4(float32x4, f32)
#define OPENCV_HAL_IMPL_NEON_INTERLEAVED(_Tpvec, _Tp, suffix) \
inline void v_load_deinterleave(const _Tp* ptr, v_##_Tpvec& a, v_##_Tpvec& b, v_##_Tpvec& c) \
{ \
_Tpvec##x3_t v = vld3q_##suffix(ptr); \
a.val = v.val[0]; \
b.val = v.val[1]; \
c.val = v.val[2]; \
} \
inline void v_load_deinterleave(const _Tp* ptr, v_##_Tpvec& a, v_##_Tpvec& b, \
v_##_Tpvec& c, v_##_Tpvec& d) \
{ \
_Tpvec##x4_t v = vld4q_##suffix(ptr); \
a.val = v.val[0]; \
b.val = v.val[1]; \
c.val = v.val[2]; \
d.val = v.val[3]; \
} \
inline void v_store_interleave( _Tp* ptr, const v_##_Tpvec& a, const v_##_Tpvec& b, const v_##_Tpvec& c) \
{ \
_Tpvec##x3_t v; \
v.val[0] = a.val; \
v.val[1] = b.val; \
v.val[2] = c.val; \
vst3q_##suffix(ptr, v); \
} \
inline void v_store_interleave( _Tp* ptr, const v_##_Tpvec& a, const v_##_Tpvec& b, \
const v_##_Tpvec& c, const v_##_Tpvec& d) \
{ \
_Tpvec##x4_t v; \
v.val[0] = a.val; \
v.val[1] = b.val; \
v.val[2] = c.val; \
v.val[3] = d.val; \
vst4q_##suffix(ptr, v); \
}
OPENCV_HAL_IMPL_NEON_INTERLEAVED(uint8x16, uchar, u8)
OPENCV_HAL_IMPL_NEON_INTERLEAVED(int8x16, schar, s8)
OPENCV_HAL_IMPL_NEON_INTERLEAVED(uint16x8, ushort, u16)
OPENCV_HAL_IMPL_NEON_INTERLEAVED(int16x8, short, s16)
OPENCV_HAL_IMPL_NEON_INTERLEAVED(uint32x4, unsigned, u32)
OPENCV_HAL_IMPL_NEON_INTERLEAVED(int32x4, int, s32)
OPENCV_HAL_IMPL_NEON_INTERLEAVED(float32x4, float, f32)
inline v_float32x4 v_cvt_f32(const v_int32x4& a)
{
return v_float32x4(vcvtq_f32_s32(a.val));
}
}
#endif

1544
modules/hal/include/opencv2/hal/intrin_sse.hpp Normal file
View File

File diff suppressed because it is too large Load Diff

47
modules/hal/src/arithm.cpp Normal file
View File

@ -0,0 +1,47 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009-2011, Willow Garage 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "precomp.hpp"
namespace cv { namespace hal {
}}

47
modules/hal/src/color.cpp Normal file
View File

@ -0,0 +1,47 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009-2011, Willow Garage 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "precomp.hpp"
namespace cv { namespace hal {
}}

47
modules/hal/src/filter.cpp Normal file
View File

@ -0,0 +1,47 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009-2011, Willow Garage 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "precomp.hpp"
namespace cv { namespace hal {
}}

1352
modules/hal/src/mathfuncs.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

208
modules/hal/src/matrix.cpp Normal file
View File

@ -0,0 +1,208 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009-2011, Willow Garage 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "precomp.hpp"
namespace cv { namespace hal {
/****************************************************************************************\
* LU & Cholesky implementation for small matrices *
\****************************************************************************************/
template<typename _Tp> static inline int
LUImpl(_Tp* A, size_t astep, int m, _Tp* b, size_t bstep, int n)
{
int i, j, k, p = 1;
astep /= sizeof(A[0]);
bstep /= sizeof(b[0]);
for( i = 0; i < m; i++ )
{
k = i;
for( j = i+1; j < m; j++ )
if( std::abs(A[j*astep + i]) > std::abs(A[k*astep + i]) )
k = j;
if( std::abs(A[k*astep + i]) < std::numeric_limits<_Tp>::epsilon() )
return 0;
if( k != i )
{
for( j = i; j < m; j++ )
std::swap(A[i*astep + j], A[k*astep + j]);
if( b )
for( j = 0; j < n; j++ )
std::swap(b[i*bstep + j], b[k*bstep + j]);
p = -p;
}
_Tp d = -1/A[i*astep + i];
for( j = i+1; j < m; j++ )
{
_Tp alpha = A[j*astep + i]*d;
for( k = i+1; k < m; k++ )
A[j*astep + k] += alpha*A[i*astep + k];
if( b )
for( k = 0; k < n; k++ )
b[j*bstep + k] += alpha*b[i*bstep + k];
}
A[i*astep + i] = -d;
}
if( b )
{
for( i = m-1; i >= 0; i-- )
for( j = 0; j < n; j++ )
{
_Tp s = b[i*bstep + j];
for( k = i+1; k < m; k++ )
s -= A[i*astep + k]*b[k*bstep + j];
b[i*bstep + j] = s*A[i*astep + i];
}
}
return p;
}
int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n)
{
return LUImpl(A, astep, m, b, bstep, n);
}
int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n)
{
return LUImpl(A, astep, m, b, bstep, n);
}
template<typename _Tp> static inline bool
CholImpl(_Tp* A, size_t astep, int m, _Tp* b, size_t bstep, int n)
{
_Tp* L = A;
int i, j, k;
double s;
astep /= sizeof(A[0]);
bstep /= sizeof(b[0]);
for( i = 0; i < m; i++ )
{
for( j = 0; j < i; j++ )
{
s = A[i*astep + j];
for( k = 0; k < j; k++ )
s -= L[i*astep + k]*L[j*astep + k];
L[i*astep + j] = (_Tp)(s*L[j*astep + j]);
}
s = A[i*astep + i];
for( k = 0; k < j; k++ )
{
double t = L[i*astep + k];
s -= t*t;
}
if( s < std::numeric_limits<_Tp>::epsilon() )
return false;
L[i*astep + i] = (_Tp)(1./std::sqrt(s));
}
if( !b )
return true;
// LLt x = b
// 1: L y = b
// 2. Lt x = y
/*
[ L00 ] y0 b0
[ L10 L11 ] y1 = b1
[ L20 L21 L22 ] y2 b2
[ L30 L31 L32 L33 ] y3 b3
[ L00 L10 L20 L30 ] x0 y0
[ L11 L21 L31 ] x1 = y1
[ L22 L32 ] x2 y2
[ L33 ] x3 y3
*/
for( i = 0; i < m; i++ )
{
for( j = 0; j < n; j++ )
{
s = b[i*bstep + j];
for( k = 0; k < i; k++ )
s -= L[i*astep + k]*b[k*bstep + j];
b[i*bstep + j] = (_Tp)(s*L[i*astep + i]);
}
}
for( i = m-1; i >= 0; i-- )
{
for( j = 0; j < n; j++ )
{
s = b[i*bstep + j];
for( k = m-1; k > i; k-- )
s -= L[k*astep + i]*b[k*bstep + j];
b[i*bstep + j] = (_Tp)(s*L[i*astep + i]);
}
}
return true;
}
bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n)
{
return CholImpl(A, astep, m, b, bstep, n);
}
bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n)
{
return CholImpl(A, astep, m, b, bstep, n);
}
}}

47
modules/hal/src/precomp.hpp
View File

@ -1,2 +1,49 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009-2011, Willow Garage 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "opencv2/hal.hpp"
#include "opencv2/hal/intrin.hpp"
#include <algorithm>
#include <cmath>
#include <cstdlib>
#include <limits>
#include <float.h>

47
modules/hal/src/resize.cpp Normal file
View File

@ -0,0 +1,47 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009-2011, Willow Garage 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "precomp.hpp"
namespace cv { namespace hal {
}}

154
modules/hal/src/norm.cpp → modules/hal/src/stat.cpp
View File

@ -80,10 +80,10 @@ static const uchar popCountTable4[] =
1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2
};
Error::Code normHamming(const uchar* a, int n, int & result)
int normHamming(const uchar* a, int n)
{
int i = 0;
result = 0;
int result = 0;
#if CV_NEON
{
uint32x4_t bits = vmovq_n_u32(0);
@ -104,13 +104,13 @@ Error::Code normHamming(const uchar* a, int n, int & result)
popCountTable[a[i+2]] + popCountTable[a[i+3]];
for( ; i < n; i++ )
result += popCountTable[a[i]];
return Error::Ok;
return result;
}
Error::Code normHamming(const uchar* a, const uchar* b, int n, int & result)
int normHamming(const uchar* a, const uchar* b, int n)
{
int i = 0;
result = 0;
int result = 0;
#if CV_NEON
{
uint32x4_t bits = vmovq_n_u32(0);
@ -133,44 +133,44 @@ Error::Code normHamming(const uchar* a, const uchar* b, int n, int & result)
popCountTable[a[i+2] ^ b[i+2]] + popCountTable[a[i+3] ^ b[i+3]];
for( ; i < n; i++ )
result += popCountTable[a[i] ^ b[i]];
return Error::Ok;
return result;
}
Error::Code normHamming(const uchar* a, int n, int cellSize, int & result)
int normHamming(const uchar* a, int n, int cellSize)
{
if( cellSize == 1 )
return normHamming(a, n, result);
return normHamming(a, n);
const uchar* tab = 0;
if( cellSize == 2 )
tab = popCountTable2;
else if( cellSize == 4 )
tab = popCountTable4;
else
return Error::Unknown;
return -1;
int i = 0;
result = 0;
int result = 0;
#if CV_ENABLE_UNROLLED
for( ; i <= n - 4; i += 4 )
result += tab[a[i]] + tab[a[i+1]] + tab[a[i+2]] + tab[a[i+3]];
#endif
for( ; i < n; i++ )
result += tab[a[i]];
return Error::Ok;
return result;
}
Error::Code normHamming(const uchar* a, const uchar* b, int n, int cellSize, int & result)
int normHamming(const uchar* a, const uchar* b, int n, int cellSize)
{
if( cellSize == 1 )
return normHamming(a, b, n, result);
return normHamming(a, b, n);
const uchar* tab = 0;
if( cellSize == 2 )
tab = popCountTable2;
else if( cellSize == 4 )
tab = popCountTable4;
else
return Error::Unknown;
return -1;
int i = 0;
result = 0;
int result = 0;
#if CV_ENABLE_UNROLLED
for( ; i <= n - 4; i += 4 )
result += tab[a[i] ^ b[i]] + tab[a[i+1] ^ b[i+1]] +
@ -178,7 +178,129 @@ Error::Code normHamming(const uchar* a, const uchar* b, int n, int cellSize, int
#endif
for( ; i < n; i++ )
result += tab[a[i] ^ b[i]];
return Error::Ok;
return result;
}
float normL2Sqr_(const float* a, const float* b, int n)
{
int j = 0; float d = 0.f;
#if CV_SSE
float CV_DECL_ALIGNED(16) buf[4];
__m128 d0 = _mm_setzero_ps(), d1 = _mm_setzero_ps();
for( ; j <= n - 8; j += 8 )
{
__m128 t0 = _mm_sub_ps(_mm_loadu_ps(a + j), _mm_loadu_ps(b + j));
__m128 t1 = _mm_sub_ps(_mm_loadu_ps(a + j + 4), _mm_loadu_ps(b + j + 4));
d0 = _mm_add_ps(d0, _mm_mul_ps(t0, t0));
d1 = _mm_add_ps(d1, _mm_mul_ps(t1, t1));
}
_mm_store_ps(buf, _mm_add_ps(d0, d1));
d = buf[0] + buf[1] + buf[2] + buf[3];
#endif
{
for( ; j <= n - 4; j += 4 )
{
float t0 = a[j] - b[j], t1 = a[j+1] - b[j+1], t2 = a[j+2] - b[j+2], t3 = a[j+3] - b[j+3];
d += t0*t0 + t1*t1 + t2*t2 + t3*t3;
}
}
for( ; j < n; j++ )
{
float t = a[j] - b[j];
d += t*t;
}
return d;
}
float normL1_(const float* a, const float* b, int n)
{
int j = 0; float d = 0.f;
#if CV_SSE
float CV_DECL_ALIGNED(16) buf[4];
static const int CV_DECL_ALIGNED(16) absbuf[4] = {0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff};
__m128 d0 = _mm_setzero_ps(), d1 = _mm_setzero_ps();
__m128 absmask = _mm_load_ps((const float*)absbuf);
for( ; j <= n - 8; j += 8 )
{
__m128 t0 = _mm_sub_ps(_mm_loadu_ps(a + j), _mm_loadu_ps(b + j));
__m128 t1 = _mm_sub_ps(_mm_loadu_ps(a + j + 4), _mm_loadu_ps(b + j + 4));
d0 = _mm_add_ps(d0, _mm_and_ps(t0, absmask));
d1 = _mm_add_ps(d1, _mm_and_ps(t1, absmask));
}
_mm_store_ps(buf, _mm_add_ps(d0, d1));
d = buf[0] + buf[1] + buf[2] + buf[3];
#elif CV_NEON
float32x4_t v_sum = vdupq_n_f32(0.0f);
for ( ; j <= n - 4; j += 4)
v_sum = vaddq_f32(v_sum, vabdq_f32(vld1q_f32(a + j), vld1q_f32(b + j)));
float CV_DECL_ALIGNED(16) buf[4];
vst1q_f32(buf, v_sum);
d = buf[0] + buf[1] + buf[2] + buf[3];
#endif
{
for( ; j <= n - 4; j += 4 )
{
d += std::abs(a[j] - b[j]) + std::abs(a[j+1] - b[j+1]) +
std::abs(a[j+2] - b[j+2]) + std::abs(a[j+3] - b[j+3]);
}
}
for( ; j < n; j++ )
d += std::abs(a[j] - b[j]);
return d;
}
int normL1_(const uchar* a, const uchar* b, int n)
{
int j = 0, d = 0;
#if CV_SSE
__m128i d0 = _mm_setzero_si128();
for( ; j <= n - 16; j += 16 )
{
__m128i t0 = _mm_loadu_si128((const __m128i*)(a + j));
__m128i t1 = _mm_loadu_si128((const __m128i*)(b + j));
d0 = _mm_add_epi32(d0, _mm_sad_epu8(t0, t1));
}
for( ; j <= n - 4; j += 4 )
{
__m128i t0 = _mm_cvtsi32_si128(*(const int*)(a + j));
__m128i t1 = _mm_cvtsi32_si128(*(const int*)(b + j));
d0 = _mm_add_epi32(d0, _mm_sad_epu8(t0, t1));
}
d = _mm_cvtsi128_si32(_mm_add_epi32(d0, _mm_unpackhi_epi64(d0, d0)));
#elif CV_NEON
uint32x4_t v_sum = vdupq_n_u32(0.0f);
for ( ; j <= n - 16; j += 16)
{
uint8x16_t v_dst = vabdq_u8(vld1q_u8(a + j), vld1q_u8(b + j));
uint16x8_t v_low = vmovl_u8(vget_low_u8(v_dst)), v_high = vmovl_u8(vget_high_u8(v_dst));
v_sum = vaddq_u32(v_sum, vaddl_u16(vget_low_u16(v_low), vget_low_u16(v_high)));
v_sum = vaddq_u32(v_sum, vaddl_u16(vget_high_u16(v_low), vget_high_u16(v_high)));
}
uint CV_DECL_ALIGNED(16) buf[4];
vst1q_u32(buf, v_sum);
d = buf[0] + buf[1] + buf[2] + buf[3];
#endif
{
for( ; j <= n - 4; j += 4 )
{
d += std::abs(a[j] - b[j]) + std::abs(a[j+1] - b[j+1]) +
std::abs(a[j+2] - b[j+2]) + std::abs(a[j+3] - b[j+3]);
}
}
for( ; j < n; j++ )
d += std::abs(a[j] - b[j]);
return d;
}
}} //cv::hal

47
modules/hal/src/warp.cpp Normal file
View File

@ -0,0 +1,47 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2009-2011, Willow Garage 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,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors "as is" and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#include "precomp.hpp"
namespace cv { namespace hal {
}}

5
modules/photo/src/arrays.hpp
View File

@ -44,6 +44,9 @@
#ifndef __OPENCV_DENOISING_ARRAYS_HPP__
#define __OPENCV_DENOISING_ARRAYS_HPP__
namespace cv
{
template <class T>
struct Array2d
{
@ -176,4 +179,6 @@ struct Array4d
}
};
}
#endif

2
modules/stitching/src/autocalib.cpp
View File

@ -49,7 +49,7 @@ namespace {
template<typename _Tp> static inline bool
decomposeCholesky(_Tp* A, size_t astep, int m)
{
if (!Cholesky(A, astep, m, 0, 0, 0))
if (!hal::Cholesky(A, astep, m, 0, 0, 0))
return false;
astep /= sizeof(A[0]);
for (int i = 0; i < m; ++i)