Alexander Alekhin
7 years ago
commit
1913482cf5
90 changed files with 6204 additions and 6600 deletions
File diff suppressed because it is too large
Load Diff
@ -0,0 +1,11 @@ |
||||
// This file is part of OpenCV project.
|
||||
// It is subject to the license terms in the LICENSE file found in the top-level directory
|
||||
// of this distribution and at http://opencv.org/license.html
|
||||
|
||||
#include "precomp.hpp" |
||||
#include "arithm_ipp.hpp" |
||||
#include "arithm.simd.hpp" |
||||
#include "arithm.simd_declarations.hpp" |
||||
|
||||
#define ARITHM_DISPATCHING_ONLY |
||||
#include "arithm.simd.hpp" |
||||
File diff suppressed because it is too large
Load Diff
@ -1,629 +0,0 @@ |
||||
/*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_ARITHM_CORE_HPP__ |
||||
#define __OPENCV_ARITHM_CORE_HPP__ |
||||
|
||||
#include "arithm_simd.hpp" |
||||
|
||||
namespace cv { |
||||
|
||||
template<typename T1, typename T2=T1, typename T3=T1> struct OpAdd |
||||
{ |
||||
typedef T1 type1; |
||||
typedef T2 type2; |
||||
typedef T3 rtype; |
||||
T3 operator ()(const T1 a, const T2 b) const { return saturate_cast<T3>(a + b); } |
||||
}; |
||||
|
||||
template<typename T1, typename T2=T1, typename T3=T1> struct OpSub |
||||
{ |
||||
typedef T1 type1; |
||||
typedef T2 type2; |
||||
typedef T3 rtype; |
||||
T3 operator ()(const T1 a, const T2 b) const { return saturate_cast<T3>(a - b); } |
||||
}; |
||||
|
||||
template<typename T1, typename T2=T1, typename T3=T1> struct OpRSub |
||||
{ |
||||
typedef T1 type1; |
||||
typedef T2 type2; |
||||
typedef T3 rtype; |
||||
T3 operator ()(const T1 a, const T2 b) const { return saturate_cast<T3>(b - a); } |
||||
}; |
||||
|
||||
template<typename T> struct OpMin |
||||
{ |
||||
typedef T type1; |
||||
typedef T type2; |
||||
typedef T rtype; |
||||
T operator ()(const T a, const T b) const { return std::min(a, b); } |
||||
}; |
||||
|
||||
template<typename T> struct OpMax |
||||
{ |
||||
typedef T type1; |
||||
typedef T type2; |
||||
typedef T rtype; |
||||
T operator ()(const T a, const T b) const { return std::max(a, b); } |
||||
}; |
||||
|
||||
template<typename T> struct OpAbsDiff |
||||
{ |
||||
typedef T type1; |
||||
typedef T type2; |
||||
typedef T rtype; |
||||
T operator()(T a, T b) const { return a > b ? a - b : b - a; } |
||||
}; |
||||
|
||||
// specializations to prevent "-0" results
|
||||
template<> struct OpAbsDiff<float> |
||||
{ |
||||
typedef float type1; |
||||
typedef float type2; |
||||
typedef float rtype; |
||||
float operator()(float a, float b) const { return std::abs(a - b); } |
||||
}; |
||||
template<> struct OpAbsDiff<double> |
||||
{ |
||||
typedef double type1; |
||||
typedef double type2; |
||||
typedef double rtype; |
||||
double operator()(double a, double b) const { return std::abs(a - b); } |
||||
}; |
||||
|
||||
template<typename T> struct OpAnd |
||||
{ |
||||
typedef T type1; |
||||
typedef T type2; |
||||
typedef T rtype; |
||||
T operator()( T a, T b ) const { return a & b; } |
||||
}; |
||||
|
||||
template<typename T> struct OpOr |
||||
{ |
||||
typedef T type1; |
||||
typedef T type2; |
||||
typedef T rtype; |
||||
T operator()( T a, T b ) const { return a | b; } |
||||
}; |
||||
|
||||
template<typename T> struct OpXor |
||||
{ |
||||
typedef T type1; |
||||
typedef T type2; |
||||
typedef T rtype; |
||||
T operator()( T a, T b ) const { return a ^ b; } |
||||
}; |
||||
|
||||
template<typename T> struct OpNot |
||||
{ |
||||
typedef T type1; |
||||
typedef T type2; |
||||
typedef T rtype; |
||||
T operator()( T a, T ) const { return ~a; } |
||||
}; |
||||
|
||||
//=============================================================================
|
||||
|
||||
template<typename T, class Op, class VOp> |
||||
void vBinOp(const T* src1, size_t step1, const T* src2, size_t step2, T* dst, size_t step, int width, int height) |
||||
{ |
||||
#if CV_SSE2 || CV_NEON |
||||
VOp vop; |
||||
#endif |
||||
Op op; |
||||
|
||||
for( ; height--; src1 = (const T *)((const uchar *)src1 + step1), |
||||
src2 = (const T *)((const uchar *)src2 + step2), |
||||
dst = (T *)((uchar *)dst + step) ) |
||||
{ |
||||
int x = 0; |
||||
|
||||
#if CV_NEON || CV_SSE2 |
||||
#if CV_AVX2 |
||||
if( USE_AVX2 ) |
||||
{ |
||||
for( ; x <= width - 32/(int)sizeof(T); x += 32/sizeof(T) ) |
||||
{ |
||||
typename VLoadStore256<T>::reg_type r0 = VLoadStore256<T>::load(src1 + x); |
||||
r0 = vop(r0, VLoadStore256<T>::load(src2 + x)); |
||||
VLoadStore256<T>::store(dst + x, r0); |
||||
} |
||||
} |
||||
#else |
||||
#if CV_SSE2 |
||||
if( USE_SSE2 ) |
||||
{ |
||||
#endif // CV_SSE2
|
||||
for( ; x <= width - 32/(int)sizeof(T); x += 32/sizeof(T) ) |
||||
{ |
||||
typename VLoadStore128<T>::reg_type r0 = VLoadStore128<T>::load(src1 + x ); |
||||
typename VLoadStore128<T>::reg_type r1 = VLoadStore128<T>::load(src1 + x + 16/sizeof(T)); |
||||
r0 = vop(r0, VLoadStore128<T>::load(src2 + x )); |
||||
r1 = vop(r1, VLoadStore128<T>::load(src2 + x + 16/sizeof(T))); |
||||
VLoadStore128<T>::store(dst + x , r0); |
||||
VLoadStore128<T>::store(dst + x + 16/sizeof(T), r1); |
||||
} |
||||
#if CV_SSE2 |
||||
} |
||||
#endif // CV_SSE2
|
||||
#endif // CV_AVX2
|
||||
#endif // CV_NEON || CV_SSE2
|
||||
|
||||
#if CV_AVX2 |
||||
// nothing
|
||||
#elif CV_SSE2 |
||||
if( USE_SSE2 ) |
||||
{ |
||||
for( ; x <= width - 8/(int)sizeof(T); x += 8/sizeof(T) ) |
||||
{ |
||||
typename VLoadStore64<T>::reg_type r = VLoadStore64<T>::load(src1 + x); |
||||
r = vop(r, VLoadStore64<T>::load(src2 + x)); |
||||
VLoadStore64<T>::store(dst + x, r); |
||||
} |
||||
} |
||||
#endif |
||||
|
||||
#if CV_ENABLE_UNROLLED |
||||
for( ; x <= width - 4; x += 4 ) |
||||
{ |
||||
T v0 = op(src1[x], src2[x]); |
||||
T v1 = op(src1[x+1], src2[x+1]); |
||||
dst[x] = v0; dst[x+1] = v1; |
||||
v0 = op(src1[x+2], src2[x+2]); |
||||
v1 = op(src1[x+3], src2[x+3]); |
||||
dst[x+2] = v0; dst[x+3] = v1; |
||||
} |
||||
#endif |
||||
|
||||
for( ; x < width; x++ ) |
||||
dst[x] = op(src1[x], src2[x]); |
||||
} |
||||
} |
||||
|
||||
template<typename T, class Op, class Op32> |
||||
void vBinOp32(const T* src1, size_t step1, const T* src2, size_t step2, |
||||
T* dst, size_t step, int width, int height) |
||||
{ |
||||
#if CV_SSE2 || CV_NEON |
||||
Op32 op32; |
||||
#endif |
||||
Op op; |
||||
|
||||
for( ; height--; src1 = (const T *)((const uchar *)src1 + step1), |
||||
src2 = (const T *)((const uchar *)src2 + step2), |
||||
dst = (T *)((uchar *)dst + step) ) |
||||
{ |
||||
int x = 0; |
||||
|
||||
#if CV_AVX2 |
||||
if( USE_AVX2 ) |
||||
{ |
||||
if( (((size_t)src1|(size_t)src2|(size_t)dst)&31) == 0 ) |
||||
{ |
||||
for( ; x <= width - 8; x += 8 ) |
||||
{ |
||||
typename VLoadStore256Aligned<T>::reg_type r0 = VLoadStore256Aligned<T>::load(src1 + x); |
||||
r0 = op32(r0, VLoadStore256Aligned<T>::load(src2 + x)); |
||||
VLoadStore256Aligned<T>::store(dst + x, r0); |
||||
} |
||||
} |
||||
} |
||||
#elif CV_SSE2 |
||||
if( USE_SSE2 ) |
||||
{ |
||||
if( (((size_t)src1|(size_t)src2|(size_t)dst)&15) == 0 ) |
||||
{ |
||||
for( ; x <= width - 8; x += 8 ) |
||||
{ |
||||
typename VLoadStore128Aligned<T>::reg_type r0 = VLoadStore128Aligned<T>::load(src1 + x ); |
||||
typename VLoadStore128Aligned<T>::reg_type r1 = VLoadStore128Aligned<T>::load(src1 + x + 4); |
||||
r0 = op32(r0, VLoadStore128Aligned<T>::load(src2 + x )); |
||||
r1 = op32(r1, VLoadStore128Aligned<T>::load(src2 + x + 4)); |
||||
VLoadStore128Aligned<T>::store(dst + x , r0); |
||||
VLoadStore128Aligned<T>::store(dst + x + 4, r1); |
||||
} |
||||
} |
||||
} |
||||
#endif // CV_AVX2
|
||||
|
||||
#if CV_NEON || CV_SSE2 |
||||
#if CV_AVX2 |
||||
if( USE_AVX2 ) |
||||
{ |
||||
for( ; x <= width - 8; x += 8 ) |
||||
{ |
||||
typename VLoadStore256<T>::reg_type r0 = VLoadStore256<T>::load(src1 + x); |
||||
r0 = op32(r0, VLoadStore256<T>::load(src2 + x)); |
||||
VLoadStore256<T>::store(dst + x, r0); |
||||
} |
||||
} |
||||
#else |
||||
#if CV_SSE2 |
||||
if( USE_SSE2 ) |
||||
{ |
||||
#endif // CV_SSE2
|
||||
for( ; x <= width - 8; x += 8 ) |
||||
{ |
||||
typename VLoadStore128<T>::reg_type r0 = VLoadStore128<T>::load(src1 + x ); |
||||
typename VLoadStore128<T>::reg_type r1 = VLoadStore128<T>::load(src1 + x + 4); |
||||
r0 = op32(r0, VLoadStore128<T>::load(src2 + x )); |
||||
r1 = op32(r1, VLoadStore128<T>::load(src2 + x + 4)); |
||||
VLoadStore128<T>::store(dst + x , r0); |
||||
VLoadStore128<T>::store(dst + x + 4, r1); |
||||
} |
||||
#if CV_SSE2 |
||||
} |
||||
#endif // CV_SSE2
|
||||
#endif // CV_AVX2
|
||||
#endif // CV_NEON || CV_SSE2
|
||||
|
||||
#if CV_ENABLE_UNROLLED |
||||
for( ; x <= width - 4; x += 4 ) |
||||
{ |
||||
T v0 = op(src1[x], src2[x]); |
||||
T v1 = op(src1[x+1], src2[x+1]); |
||||
dst[x] = v0; dst[x+1] = v1; |
||||
v0 = op(src1[x+2], src2[x+2]); |
||||
v1 = op(src1[x+3], src2[x+3]); |
||||
dst[x+2] = v0; dst[x+3] = v1; |
||||
} |
||||
#endif |
||||
|
||||
for( ; x < width; x++ ) |
||||
dst[x] = op(src1[x], src2[x]); |
||||
} |
||||
} |
||||
|
||||
|
||||
template<typename T, class Op, class Op64> |
||||
void vBinOp64(const T* src1, size_t step1, const T* src2, size_t step2, |
||||
T* dst, size_t step, int width, int height) |
||||
{ |
||||
#if CV_SSE2 |
||||
Op64 op64; |
||||
#endif |
||||
Op op; |
||||
|
||||
for( ; height--; src1 = (const T *)((const uchar *)src1 + step1), |
||||
src2 = (const T *)((const uchar *)src2 + step2), |
||||
dst = (T *)((uchar *)dst + step) ) |
||||
{ |
||||
int x = 0; |
||||
|
||||
#if CV_AVX2 |
||||
if( USE_AVX2 ) |
||||
{ |
||||
if( (((size_t)src1|(size_t)src2|(size_t)dst)&31) == 0 ) |
||||
{ |
||||
for( ; x <= width - 4; x += 4 ) |
||||
{ |
||||
typename VLoadStore256Aligned<T>::reg_type r0 = VLoadStore256Aligned<T>::load(src1 + x); |
||||
r0 = op64(r0, VLoadStore256Aligned<T>::load(src2 + x)); |
||||
VLoadStore256Aligned<T>::store(dst + x, r0); |
||||
} |
||||
} |
||||
} |
||||
#elif CV_SSE2 |
||||
if( USE_SSE2 ) |
||||
{ |
||||
if( (((size_t)src1|(size_t)src2|(size_t)dst)&15) == 0 ) |
||||
{ |
||||
for( ; x <= width - 4; x += 4 ) |
||||
{ |
||||
typename VLoadStore128Aligned<T>::reg_type r0 = VLoadStore128Aligned<T>::load(src1 + x ); |
||||
typename VLoadStore128Aligned<T>::reg_type r1 = VLoadStore128Aligned<T>::load(src1 + x + 2); |
||||
r0 = op64(r0, VLoadStore128Aligned<T>::load(src2 + x )); |
||||
r1 = op64(r1, VLoadStore128Aligned<T>::load(src2 + x + 2)); |
||||
VLoadStore128Aligned<T>::store(dst + x , r0); |
||||
VLoadStore128Aligned<T>::store(dst + x + 2, r1); |
||||
} |
||||
} |
||||
} |
||||
#endif |
||||
|
||||
for( ; x <= width - 4; x += 4 ) |
||||
{ |
||||
T v0 = op(src1[x], src2[x]); |
||||
T v1 = op(src1[x+1], src2[x+1]); |
||||
dst[x] = v0; dst[x+1] = v1; |
||||
v0 = op(src1[x+2], src2[x+2]); |
||||
v1 = op(src1[x+3], src2[x+3]); |
||||
dst[x+2] = v0; dst[x+3] = v1; |
||||
} |
||||
|
||||
for( ; x < width; x++ ) |
||||
dst[x] = op(src1[x], src2[x]); |
||||
} |
||||
} |
||||
|
||||
template<typename T> static void |
||||
cmp_(const T* src1, size_t step1, const T* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height, int code) |
||||
{ |
||||
step1 /= sizeof(src1[0]); |
||||
step2 /= sizeof(src2[0]); |
||||
if( code == CMP_GE || code == CMP_LT ) |
||||
{ |
||||
std::swap(src1, src2); |
||||
std::swap(step1, step2); |
||||
code = code == CMP_GE ? CMP_LE : CMP_GT; |
||||
} |
||||
|
||||
Cmp_SIMD<T> vop(code); |
||||
|
||||
if( code == CMP_GT || code == CMP_LE ) |
||||
{ |
||||
int m = code == CMP_GT ? 0 : 255; |
||||
for( ; height--; src1 += step1, src2 += step2, dst += step ) |
||||
{ |
||||
int x = vop(src1, src2, dst, width); |
||||
#if CV_ENABLE_UNROLLED |
||||
for( ; x <= width - 4; x += 4 ) |
||||
{ |
||||
int t0, t1; |
||||
t0 = -(src1[x] > src2[x]) ^ m; |
||||
t1 = -(src1[x+1] > src2[x+1]) ^ m; |
||||
dst[x] = (uchar)t0; dst[x+1] = (uchar)t1; |
||||
t0 = -(src1[x+2] > src2[x+2]) ^ m; |
||||
t1 = -(src1[x+3] > src2[x+3]) ^ m; |
||||
dst[x+2] = (uchar)t0; dst[x+3] = (uchar)t1; |
||||
} |
||||
#endif |
||||
for( ; x < width; x++ ) |
||||
dst[x] = (uchar)(-(src1[x] > src2[x]) ^ m); |
||||
} |
||||
} |
||||
else if( code == CMP_EQ || code == CMP_NE ) |
||||
{ |
||||
int m = code == CMP_EQ ? 0 : 255; |
||||
for( ; height--; src1 += step1, src2 += step2, dst += step ) |
||||
{ |
||||
int x = 0; |
||||
#if CV_ENABLE_UNROLLED |
||||
for( ; x <= width - 4; x += 4 ) |
||||
{ |
||||
int t0, t1; |
||||
t0 = -(src1[x] == src2[x]) ^ m; |
||||
t1 = -(src1[x+1] == src2[x+1]) ^ m; |
||||
dst[x] = (uchar)t0; dst[x+1] = (uchar)t1; |
||||
t0 = -(src1[x+2] == src2[x+2]) ^ m; |
||||
t1 = -(src1[x+3] == src2[x+3]) ^ m; |
||||
dst[x+2] = (uchar)t0; dst[x+3] = (uchar)t1; |
||||
} |
||||
#endif |
||||
for( ; x < width; x++ ) |
||||
dst[x] = (uchar)(-(src1[x] == src2[x]) ^ m); |
||||
} |
||||
} |
||||
} |
||||
|
||||
template<typename T, typename WT> static void |
||||
mul_( const T* src1, size_t step1, const T* src2, size_t step2, |
||||
T* dst, size_t step, int width, int height, WT scale ) |
||||
{ |
||||
step1 /= sizeof(src1[0]); |
||||
step2 /= sizeof(src2[0]); |
||||
step /= sizeof(dst[0]); |
||||
|
||||
Mul_SIMD<T, WT> vop; |
||||
|
||||
if( scale == (WT)1. ) |
||||
{ |
||||
for( ; height--; src1 += step1, src2 += step2, dst += step ) |
||||
{ |
||||
int i = vop(src1, src2, dst, width, scale); |
||||
#if CV_ENABLE_UNROLLED |
||||
for(; i <= width - 4; i += 4 ) |
||||
{ |
||||
T t0; |
||||
T t1; |
||||
t0 = saturate_cast<T>(src1[i ] * src2[i ]); |
||||
t1 = saturate_cast<T>(src1[i+1] * src2[i+1]); |
||||
dst[i ] = t0; |
||||
dst[i+1] = t1; |
||||
|
||||
t0 = saturate_cast<T>(src1[i+2] * src2[i+2]); |
||||
t1 = saturate_cast<T>(src1[i+3] * src2[i+3]); |
||||
dst[i+2] = t0; |
||||
dst[i+3] = t1; |
||||
} |
||||
#endif |
||||
for( ; i < width; i++ ) |
||||
dst[i] = saturate_cast<T>(src1[i] * src2[i]); |
||||
} |
||||
} |
||||
else |
||||
{ |
||||
for( ; height--; src1 += step1, src2 += step2, dst += step ) |
||||
{ |
||||
int i = vop(src1, src2, dst, width, scale); |
||||
#if CV_ENABLE_UNROLLED |
||||
for(; i <= width - 4; i += 4 ) |
||||
{ |
||||
T t0 = saturate_cast<T>(scale*(WT)src1[i]*src2[i]); |
||||
T t1 = saturate_cast<T>(scale*(WT)src1[i+1]*src2[i+1]); |
||||
dst[i] = t0; dst[i+1] = t1; |
||||
|
||||
t0 = saturate_cast<T>(scale*(WT)src1[i+2]*src2[i+2]); |
||||
t1 = saturate_cast<T>(scale*(WT)src1[i+3]*src2[i+3]); |
||||
dst[i+2] = t0; dst[i+3] = t1; |
||||
} |
||||
#endif |
||||
for( ; i < width; i++ ) |
||||
dst[i] = saturate_cast<T>(scale*(WT)src1[i]*src2[i]); |
||||
} |
||||
} |
||||
} |
||||
|
||||
|
||||
template<typename T> static void |
||||
div_i( const T* src1, size_t step1, const T* src2, size_t step2, |
||||
T* dst, size_t step, int width, int height, double scale ) |
||||
{ |
||||
step1 /= sizeof(src1[0]); |
||||
step2 /= sizeof(src2[0]); |
||||
step /= sizeof(dst[0]); |
||||
|
||||
Div_SIMD<T> vop; |
||||
float scale_f = (float)scale; |
||||
|
||||
for( ; height--; src1 += step1, src2 += step2, dst += step ) |
||||
{ |
||||
int i = vop(src1, src2, dst, width, scale); |
||||
for( ; i < width; i++ ) |
||||
{ |
||||
T num = src1[i], denom = src2[i]; |
||||
T v = 0; |
||||
if (denom != 0) |
||||
v = saturate_cast<T>(num*scale_f/denom); |
||||
dst[i] = v; |
||||
} |
||||
} |
||||
} |
||||
|
||||
template<typename T> static void |
||||
div_f( const T* src1, size_t step1, const T* src2, size_t step2, |
||||
T* dst, size_t step, int width, int height, double scale ) |
||||
{ |
||||
T scale_f = (T)scale; |
||||
step1 /= sizeof(src1[0]); |
||||
step2 /= sizeof(src2[0]); |
||||
step /= sizeof(dst[0]); |
||||
|
||||
Div_SIMD<T> vop; |
||||
|
||||
for( ; height--; src1 += step1, src2 += step2, dst += step ) |
||||
{ |
||||
int i = vop(src1, src2, dst, width, scale); |
||||
for( ; i < width; i++ ) |
||||
{ |
||||
T num = src1[i], denom = src2[i]; |
||||
dst[i] = saturate_cast<T>(num*scale_f/denom); |
||||
} |
||||
} |
||||
} |
||||
|
||||
template<typename T> static void |
||||
recip_i( const T* src2, size_t step2, |
||||
T* dst, size_t step, int width, int height, double scale ) |
||||
{ |
||||
step2 /= sizeof(src2[0]); |
||||
step /= sizeof(dst[0]); |
||||
|
||||
Recip_SIMD<T> vop; |
||||
float scale_f = (float)scale; |
||||
|
||||
for( ; height--; src2 += step2, dst += step ) |
||||
{ |
||||
int i = vop(src2, dst, width, scale); |
||||
for( ; i < width; i++ ) |
||||
{ |
||||
T denom = src2[i]; |
||||
T v = 0; |
||||
if (denom != 0) |
||||
v = saturate_cast<T>(scale_f/denom); |
||||
dst[i] = v; |
||||
} |
||||
} |
||||
} |
||||
|
||||
template<typename T> static void |
||||
recip_f( const T* src2, size_t step2, |
||||
T* dst, size_t step, int width, int height, double scale ) |
||||
{ |
||||
T scale_f = (T)scale; |
||||
step2 /= sizeof(src2[0]); |
||||
step /= sizeof(dst[0]); |
||||
|
||||
Recip_SIMD<T> vop; |
||||
|
||||
for( ; height--; src2 += step2, dst += step ) |
||||
{ |
||||
int i = vop(src2, dst, width, scale); |
||||
for( ; i < width; i++ ) |
||||
{ |
||||
T denom = src2[i]; |
||||
dst[i] = saturate_cast<T>(scale_f/denom); |
||||
} |
||||
} |
||||
} |
||||
|
||||
template<typename T, typename WT> static void |
||||
addWeighted_( const T* src1, size_t step1, const T* src2, size_t step2, |
||||
T* dst, size_t step, int width, int height, void* _scalars ) |
||||
{ |
||||
const double* scalars = (const double*)_scalars; |
||||
WT alpha = (WT)scalars[0], beta = (WT)scalars[1], gamma = (WT)scalars[2]; |
||||
step1 /= sizeof(src1[0]); |
||||
step2 /= sizeof(src2[0]); |
||||
step /= sizeof(dst[0]); |
||||
|
||||
AddWeighted_SIMD<T, WT> vop; |
||||
|
||||
for( ; height--; src1 += step1, src2 += step2, dst += step ) |
||||
{ |
||||
int x = vop(src1, src2, dst, width, alpha, beta, gamma); |
||||
#if CV_ENABLE_UNROLLED |
||||
for( ; x <= width - 4; x += 4 ) |
||||
{ |
||||
T t0 = saturate_cast<T>(src1[x]*alpha + src2[x]*beta + gamma); |
||||
T t1 = saturate_cast<T>(src1[x+1]*alpha + src2[x+1]*beta + gamma); |
||||
dst[x] = t0; dst[x+1] = t1; |
||||
|
||||
t0 = saturate_cast<T>(src1[x+2]*alpha + src2[x+2]*beta + gamma); |
||||
t1 = saturate_cast<T>(src1[x+3]*alpha + src2[x+3]*beta + gamma); |
||||
dst[x+2] = t0; dst[x+3] = t1; |
||||
} |
||||
#endif |
||||
for( ; x < width; x++ ) |
||||
dst[x] = saturate_cast<T>(src1[x]*alpha + src2[x]*beta + gamma); |
||||
} |
||||
} |
||||
|
||||
} // cv::
|
||||
|
||||
|
||||
#endif // __OPENCV_ARITHM_CORE_HPP__
|
||||
@ -0,0 +1,417 @@ |
||||
// This file is part of OpenCV project.
|
||||
// It is subject to the license terms in the LICENSE file found in the top-level directory
|
||||
// of this distribution and at http://opencv.org/license.html
|
||||
#if ARITHM_USE_IPP |
||||
|
||||
namespace cv { namespace hal { |
||||
|
||||
//=======================================
|
||||
// Arithmetic and logical operations
|
||||
// +, -, *, /, &, |, ^, ~, abs ...
|
||||
//=======================================
|
||||
|
||||
#define ARITHM_IPP_BIN(fun, ...) \ |
||||
do { \
|
||||
if (!CV_IPP_CHECK_COND) \
|
||||
return 0; \
|
||||
if (height == 1) \
|
||||
step1 = step2 = step = width * sizeof(dst[0]); \
|
||||
if (0 <= CV_INSTRUMENT_FUN_IPP(fun, __VA_ARGS__)) \
|
||||
{ \
|
||||
CV_IMPL_ADD(CV_IMPL_IPP); \
|
||||
return 1; \
|
||||
} \
|
||||
setIppErrorStatus(); \
|
||||
return 0; \
|
||||
} while(0) |
||||
|
||||
//=======================================
|
||||
// Addition
|
||||
//=======================================
|
||||
|
||||
inline int arithm_ipp_add8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiAdd_8u_C1RSfs, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height), 0); |
||||
} |
||||
|
||||
inline int arithm_ipp_add16u(const ushort* src1, size_t step1, const ushort* src2, size_t step2, |
||||
ushort* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiAdd_16u_C1RSfs, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height), 0); |
||||
} |
||||
|
||||
inline int arithm_ipp_add16s(const short* src1, size_t step1, const short* src2, size_t step2, |
||||
short* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiAdd_16s_C1RSfs, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height), 0); |
||||
} |
||||
|
||||
inline int arithm_ipp_add32f(const float* src1, size_t step1, const float* src2, size_t step2, |
||||
float* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiAdd_32f_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
#define arithm_ipp_add8s(...) 0 |
||||
#define arithm_ipp_add32s(...) 0 |
||||
#define arithm_ipp_add64f(...) 0 |
||||
|
||||
//=======================================
|
||||
// Subtract
|
||||
//=======================================
|
||||
|
||||
inline int arithm_ipp_sub8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiSub_8u_C1RSfs, src2, (int)step2, src1, (int)step1, dst, (int)step, ippiSize(width, height), 0); |
||||
} |
||||
|
||||
inline int arithm_ipp_sub16u(const ushort* src1, size_t step1, const ushort* src2, size_t step2, |
||||
ushort* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiSub_16u_C1RSfs, src2, (int)step2, src1, (int)step1, dst, (int)step, ippiSize(width, height), 0); |
||||
} |
||||
|
||||
inline int arithm_ipp_sub16s(const short* src1, size_t step1, const short* src2, size_t step2, |
||||
short* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiSub_16s_C1RSfs, src2, (int)step2, src1, (int)step1, dst, (int)step, ippiSize(width, height), 0); |
||||
} |
||||
|
||||
inline int arithm_ipp_sub32f(const float* src1, size_t step1, const float* src2, size_t step2, |
||||
float* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiSub_32f_C1R, src2, (int)step2, src1, (int)step1, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
#define arithm_ipp_sub8s(...) 0 |
||||
#define arithm_ipp_sub32s(...) 0 |
||||
#define arithm_ipp_sub64f(...) 0 |
||||
|
||||
///////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
#define ARITHM_IPP_MIN_MAX(fun, type) \ |
||||
do { \
|
||||
if (!CV_IPP_CHECK_COND) \
|
||||
return 0; \
|
||||
type* s1 = (type*)src1; \
|
||||
type* s2 = (type*)src2; \
|
||||
type* d = dst; \
|
||||
if (height == 1) \
|
||||
step1 = step2 = step = width * sizeof(dst[0]); \
|
||||
int i = 0; \
|
||||
for(; i < height; i++) \
|
||||
{ \
|
||||
if (0 > CV_INSTRUMENT_FUN_IPP(fun, s1, s2, d, width)) \
|
||||
break; \
|
||||
s1 = (type*)((uchar*)s1 + step1); \
|
||||
s2 = (type*)((uchar*)s2 + step2); \
|
||||
d = (type*)((uchar*)d + step); \
|
||||
} \
|
||||
if (i == height) \
|
||||
{ \
|
||||
CV_IMPL_ADD(CV_IMPL_IPP); \
|
||||
return 1; \
|
||||
} \
|
||||
setIppErrorStatus(); \
|
||||
return 0; \
|
||||
} while(0) |
||||
|
||||
//=======================================
|
||||
// Max
|
||||
//=======================================
|
||||
|
||||
inline int arithm_ipp_max8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_MIN_MAX(ippsMaxEvery_8u, uchar); |
||||
} |
||||
|
||||
inline int arithm_ipp_max16u(const ushort* src1, size_t step1, const ushort* src2, size_t step2, |
||||
ushort* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_MIN_MAX(ippsMaxEvery_16u, ushort); |
||||
} |
||||
|
||||
inline int arithm_ipp_max32f(const float* src1, size_t step1, const float* src2, size_t step2, |
||||
float* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_MIN_MAX(ippsMaxEvery_32f, float); |
||||
} |
||||
|
||||
inline int arithm_ipp_max64f(const double* src1, size_t step1, const double* src2, size_t step2, |
||||
double* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_MIN_MAX(ippsMaxEvery_64f, double); |
||||
} |
||||
|
||||
#define arithm_ipp_max8s(...) 0 |
||||
#define arithm_ipp_max16s(...) 0 |
||||
#define arithm_ipp_max32s(...) 0 |
||||
|
||||
//=======================================
|
||||
// Min
|
||||
//=======================================
|
||||
|
||||
inline int arithm_ipp_min8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_MIN_MAX(ippsMinEvery_8u, uchar); |
||||
} |
||||
|
||||
inline int arithm_ipp_min16u(const ushort* src1, size_t step1, const ushort* src2, size_t step2, |
||||
ushort* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_MIN_MAX(ippsMinEvery_16u, ushort); |
||||
} |
||||
|
||||
inline int arithm_ipp_min32f(const float* src1, size_t step1, const float* src2,size_t step2, |
||||
float* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_MIN_MAX(ippsMinEvery_32f, float); |
||||
} |
||||
|
||||
inline int arithm_ipp_min64f(const double* src1, size_t step1, const double* src2, size_t step2, |
||||
double* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_MIN_MAX(ippsMinEvery_64f, double); |
||||
} |
||||
|
||||
#define arithm_ipp_min8s(...) 0 |
||||
#define arithm_ipp_min16s(...) 0 |
||||
#define arithm_ipp_min32s(...) 0 |
||||
|
||||
//=======================================
|
||||
// AbsDiff
|
||||
//=======================================
|
||||
|
||||
inline int arithm_ipp_absdiff8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiAbsDiff_8u_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
inline int arithm_ipp_absdiff16u(const ushort* src1, size_t step1, const ushort* src2, size_t step2, |
||||
ushort* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiAbsDiff_16u_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
inline int arithm_ipp_absdiff32f(const float* src1, size_t step1, const float* src2, size_t step2, |
||||
float* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiAbsDiff_32f_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
#define arithm_ipp_absdiff8s(...) 0 |
||||
#define arithm_ipp_absdiff16s(...) 0 |
||||
#define arithm_ipp_absdiff32s(...) 0 |
||||
#define arithm_ipp_absdiff64f(...) 0 |
||||
|
||||
//=======================================
|
||||
// Logical
|
||||
//=======================================
|
||||
|
||||
inline int arithm_ipp_and8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiAnd_8u_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
inline int arithm_ipp_or8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiOr_8u_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
inline int arithm_ipp_xor8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height) |
||||
{ |
||||
ARITHM_IPP_BIN(ippiXor_8u_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
inline int arithm_ipp_not8u(const uchar* src1, size_t step1, uchar* dst, size_t step, int width, int height) |
||||
{ |
||||
if (!CV_IPP_CHECK_COND) |
||||
return 0; |
||||
if (height == 1) |
||||
step1 = step = width * sizeof(dst[0]); |
||||
if (0 <= CV_INSTRUMENT_FUN_IPP(ippiNot_8u_C1R, src1, (int)step1, dst, (int)step, ippiSize(width, height))) |
||||
{ |
||||
CV_IMPL_ADD(CV_IMPL_IPP); |
||||
return 1; |
||||
} |
||||
setIppErrorStatus(); |
||||
return 0; |
||||
} |
||||
|
||||
//=======================================
|
||||
// Compare
|
||||
//=======================================
|
||||
|
||||
#define ARITHM_IPP_CMP(fun, ...) \ |
||||
do { \
|
||||
if (!CV_IPP_CHECK_COND) \
|
||||
return 0; \
|
||||
IppCmpOp op = arithm_ipp_convert_cmp(cmpop); \
|
||||
if (op < 0) \
|
||||
return 0; \
|
||||
if (height == 1) \
|
||||
step1 = step2 = step = width * sizeof(dst[0]); \
|
||||
if (0 <= CV_INSTRUMENT_FUN_IPP(fun, __VA_ARGS__, op)) \
|
||||
{ \
|
||||
CV_IMPL_ADD(CV_IMPL_IPP); \
|
||||
return 1; \
|
||||
} \
|
||||
setIppErrorStatus(); \
|
||||
return 0; \
|
||||
} while(0) |
||||
|
||||
inline IppCmpOp arithm_ipp_convert_cmp(int cmpop) |
||||
{ |
||||
switch(cmpop) |
||||
{ |
||||
case CMP_EQ: return ippCmpEq; |
||||
case CMP_GT: return ippCmpGreater; |
||||
case CMP_GE: return ippCmpGreaterEq; |
||||
case CMP_LT: return ippCmpLess; |
||||
case CMP_LE: return ippCmpLessEq; |
||||
default: return (IppCmpOp)-1; |
||||
} |
||||
} |
||||
|
||||
inline int arithm_ipp_cmp8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height, int cmpop) |
||||
{ |
||||
ARITHM_IPP_CMP(ippiCompare_8u_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
inline int arithm_ipp_cmp16u(const ushort* src1, size_t step1, const ushort* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height, int cmpop) |
||||
{ |
||||
ARITHM_IPP_CMP(ippiCompare_16u_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
inline int arithm_ipp_cmp16s(const short* src1, size_t step1, const short* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height, int cmpop) |
||||
{ |
||||
ARITHM_IPP_CMP(ippiCompare_16s_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
inline int arithm_ipp_cmp32f(const float* src1, size_t step1, const float* src2, size_t step2, |
||||
uchar* dst, size_t step, int width, int height, int cmpop) |
||||
{ |
||||
ARITHM_IPP_CMP(ippiCompare_32f_C1R, src1, (int)step1, src2, (int)step2, dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
#define arithm_ipp_cmp8s(...) 0 |
||||
#define arithm_ipp_cmp32s(...) 0 |
||||
#define arithm_ipp_cmp64f(...) 0 |
||||
|
||||
//=======================================
|
||||
// Multiply
|
||||
//=======================================
|
||||
|
||||
#define ARITHM_IPP_MUL(fun, ...) \ |
||||
do { \
|
||||
if (!CV_IPP_CHECK_COND) \
|
||||
return 0; \
|
||||
float fscale = (float)scale; \
|
||||
if (std::fabs(fscale - 1) > FLT_EPSILON) \
|
||||
return 0; \
|
||||
if (0 <= CV_INSTRUMENT_FUN_IPP(fun, __VA_ARGS__)) \
|
||||
{ \
|
||||
CV_IMPL_ADD(CV_IMPL_IPP); \
|
||||
return 1; \
|
||||
} \
|
||||
setIppErrorStatus(); \
|
||||
return 0; \
|
||||
} while(0) |
||||
|
||||
inline int arithm_ipp_mul8u(const uchar *src1, size_t step1, const uchar *src2, size_t step2, |
||||
uchar *dst, size_t step, int width, int height, double scale) |
||||
{ |
||||
ARITHM_IPP_MUL(ippiMul_8u_C1RSfs, src1, (int)step1, src2, (int)step2,dst, (int)step, ippiSize(width, height), 0); |
||||
} |
||||
inline int arithm_ipp_mul16u(const ushort *src1, size_t step1, const ushort *src2, size_t step2, |
||||
ushort *dst, size_t step, int width, int height, double scale) |
||||
{ |
||||
ARITHM_IPP_MUL(ippiMul_16u_C1RSfs, src1, (int)step1, src2, (int)step2,dst, (int)step, ippiSize(width, height), 0); |
||||
} |
||||
|
||||
inline int arithm_ipp_mul16s(const short *src1, size_t step1, const short *src2, size_t step2, |
||||
short *dst, size_t step, int width, int height, double scale) |
||||
{ |
||||
ARITHM_IPP_MUL(ippiMul_16s_C1RSfs, src1, (int)step1, src2, (int)step2,dst, (int)step, ippiSize(width, height), 0); |
||||
} |
||||
|
||||
inline int arithm_ipp_mul32f(const float *src1, size_t step1, const float *src2, size_t step2, |
||||
float *dst, size_t step, int width, int height, double scale) |
||||
{ |
||||
ARITHM_IPP_MUL(ippiMul_32f_C1R, src1, (int)step1, src2, (int)step2,dst, (int)step, ippiSize(width, height)); |
||||
} |
||||
|
||||
#define arithm_ipp_mul8s(...) 0 |
||||
#define arithm_ipp_mul32s(...) 0 |
||||
#define arithm_ipp_mul64f(...) 0 |
||||
|
||||
//=======================================
|
||||
// Div
|
||||
//=======================================
|
||||
|
||||
#define arithm_ipp_div8u(...) 0 |
||||
#define arithm_ipp_div8s(...) 0 |
||||
#define arithm_ipp_div16u(...) 0 |
||||
#define arithm_ipp_div16s(...) 0 |
||||
#define arithm_ipp_div32s(...) 0 |
||||
#define arithm_ipp_div32f(...) 0 |
||||
#define arithm_ipp_div64f(...) 0 |
||||
|
||||
//=======================================
|
||||
// AddWeighted
|
||||
//=======================================
|
||||
|
||||
#define arithm_ipp_addWeighted8u(...) 0 |
||||
#define arithm_ipp_addWeighted8s(...) 0 |
||||
#define arithm_ipp_addWeighted16u(...) 0 |
||||
#define arithm_ipp_addWeighted16s(...) 0 |
||||
#define arithm_ipp_addWeighted32s(...) 0 |
||||
#define arithm_ipp_addWeighted32f(...) 0 |
||||
#define arithm_ipp_addWeighted64f(...) 0 |
||||
|
||||
//=======================================
|
||||
// Reciprocial
|
||||
//=======================================
|
||||
|
||||
#define arithm_ipp_recip8u(...) 0 |
||||
#define arithm_ipp_recip8s(...) 0 |
||||
#define arithm_ipp_recip16u(...) 0 |
||||
#define arithm_ipp_recip16s(...) 0 |
||||
#define arithm_ipp_recip32s(...) 0 |
||||
#define arithm_ipp_recip32f(...) 0 |
||||
#define arithm_ipp_recip64f(...) 0 |
||||
|
||||
/** empty block in case if you have "fun"
|
||||
#define arithm_ipp_8u(...) 0 |
||||
#define arithm_ipp_8s(...) 0 |
||||
#define arithm_ipp_16u(...) 0 |
||||
#define arithm_ipp_16s(...) 0 |
||||
#define arithm_ipp_32s(...) 0 |
||||
#define arithm_ipp_32f(...) 0 |
||||
#define arithm_ipp_64f(...) 0 |
||||
**/ |
||||
|
||||
}} // cv::hal::
|
||||
|
||||
#define ARITHM_CALL_IPP(fun, ...) \ |
||||
{ \
|
||||
if (__CV_EXPAND(fun(__VA_ARGS__))) \
|
||||
return; \
|
||||
} |
||||
|
||||
#endif // ARITHM_USE_IPP
|
||||
|
||||
|
||||
#if !ARITHM_USE_IPP |
||||
#define ARITHM_CALL_IPP(...) |
||||
#endif |
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
Loading…
Reference in new issue