opencv-lib - openCV library for Renesas RZ/A

Renesas » Code » opencv-lib

openCV library for Renesas RZ/A

include/opencv2/core/hal/intrin_cpp.hpp@0:0e0631af0305, 2021-01-29 (annotated)

Committer:: RyoheiHagimoto
Date:: Fri Jan 29 04:53:38 2021 +0000
Revision:: 0:0e0631af0305

copied from https://github.com/d-kato/opencv-lib.

Who changed what in which revision?

User	Revision	Line number	New contents of line
RyoheiHagimoto	0:0e0631af0305	1	/*M///////////////////////////////////////////////////////////////////////////////////////
RyoheiHagimoto	0:0e0631af0305	2	//
RyoheiHagimoto	0:0e0631af0305	3	// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
RyoheiHagimoto	0:0e0631af0305	4	//
RyoheiHagimoto	0:0e0631af0305	5	// By downloading, copying, installing or using the software you agree to this license.
RyoheiHagimoto	0:0e0631af0305	6	// If you do not agree to this license, do not download, install,
RyoheiHagimoto	0:0e0631af0305	7	// copy or use the software.
RyoheiHagimoto	0:0e0631af0305	8	//
RyoheiHagimoto	0:0e0631af0305	9	//
RyoheiHagimoto	0:0e0631af0305	10	// License Agreement
RyoheiHagimoto	0:0e0631af0305	11	// For Open Source Computer Vision Library
RyoheiHagimoto	0:0e0631af0305	12	//
RyoheiHagimoto	0:0e0631af0305	13	// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
RyoheiHagimoto	0:0e0631af0305	14	// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
RyoheiHagimoto	0:0e0631af0305	15	// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
RyoheiHagimoto	0:0e0631af0305	16	// Copyright (C) 2015, Itseez Inc., all rights reserved.
RyoheiHagimoto	0:0e0631af0305	17	// Third party copyrights are property of their respective owners.
RyoheiHagimoto	0:0e0631af0305	18	//
RyoheiHagimoto	0:0e0631af0305	19	// Redistribution and use in source and binary forms, with or without modification,
RyoheiHagimoto	0:0e0631af0305	20	// are permitted provided that the following conditions are met:
RyoheiHagimoto	0:0e0631af0305	21	//
RyoheiHagimoto	0:0e0631af0305	22	// * Redistribution's of source code must retain the above copyright notice,
RyoheiHagimoto	0:0e0631af0305	23	// this list of conditions and the following disclaimer.
RyoheiHagimoto	0:0e0631af0305	24	//
RyoheiHagimoto	0:0e0631af0305	25	// * Redistribution's in binary form must reproduce the above copyright notice,
RyoheiHagimoto	0:0e0631af0305	26	// this list of conditions and the following disclaimer in the documentation
RyoheiHagimoto	0:0e0631af0305	27	// and/or other materials provided with the distribution.
RyoheiHagimoto	0:0e0631af0305	28	//
RyoheiHagimoto	0:0e0631af0305	29	// * The name of the copyright holders may not be used to endorse or promote products
RyoheiHagimoto	0:0e0631af0305	30	// derived from this software without specific prior written permission.
RyoheiHagimoto	0:0e0631af0305	31	//
RyoheiHagimoto	0:0e0631af0305	32	// This software is provided by the copyright holders and contributors "as is" and
RyoheiHagimoto	0:0e0631af0305	33	// any express or implied warranties, including, but not limited to, the implied
RyoheiHagimoto	0:0e0631af0305	34	// warranties of merchantability and fitness for a particular purpose are disclaimed.
RyoheiHagimoto	0:0e0631af0305	35	// In no event shall the Intel Corporation or contributors be liable for any direct,
RyoheiHagimoto	0:0e0631af0305	36	// indirect, incidental, special, exemplary, or consequential damages
RyoheiHagimoto	0:0e0631af0305	37	// (including, but not limited to, procurement of substitute goods or services;
RyoheiHagimoto	0:0e0631af0305	38	// loss of use, data, or profits; or business interruption) however caused
RyoheiHagimoto	0:0e0631af0305	39	// and on any theory of liability, whether in contract, strict liability,
RyoheiHagimoto	0:0e0631af0305	40	// or tort (including negligence or otherwise) arising in any way out of
RyoheiHagimoto	0:0e0631af0305	41	// the use of this software, even if advised of the possibility of such damage.
RyoheiHagimoto	0:0e0631af0305	42	//
RyoheiHagimoto	0:0e0631af0305	43	//M*/
RyoheiHagimoto	0:0e0631af0305	44
RyoheiHagimoto	0:0e0631af0305	45	#ifndef OPENCV_HAL_INTRIN_CPP_HPP
RyoheiHagimoto	0:0e0631af0305	46	#define OPENCV_HAL_INTRIN_CPP_HPP
RyoheiHagimoto	0:0e0631af0305	47
RyoheiHagimoto	0:0e0631af0305	48	#include <limits>
RyoheiHagimoto	0:0e0631af0305	49	#include <cstring>
RyoheiHagimoto	0:0e0631af0305	50	#include <algorithm>
RyoheiHagimoto	0:0e0631af0305	51	#include "opencv2/core/saturate.hpp"
RyoheiHagimoto	0:0e0631af0305	52
RyoheiHagimoto	0:0e0631af0305	53	namespace cv
RyoheiHagimoto	0:0e0631af0305	54	{
RyoheiHagimoto	0:0e0631af0305	55
RyoheiHagimoto	0:0e0631af0305	56	/** @addtogroup core_hal_intrin
RyoheiHagimoto	0:0e0631af0305	57
RyoheiHagimoto	0:0e0631af0305	58	"Universal intrinsics" is a types and functions set intended to simplify vectorization of code on
RyoheiHagimoto	0:0e0631af0305	59	different platforms. Currently there are two supported SIMD extensions: __SSE/SSE2__ on x86
RyoheiHagimoto	0:0e0631af0305	60	architectures and __NEON__ on ARM architectures, both allow working with 128 bit registers
RyoheiHagimoto	0:0e0631af0305	61	containing packed values of different types. In case when there is no SIMD extension available
RyoheiHagimoto	0:0e0631af0305	62	during compilation, fallback C++ implementation of intrinsics will be chosen and code will work as
RyoheiHagimoto	0:0e0631af0305	63	expected although it could be slower.
RyoheiHagimoto	0:0e0631af0305	64
RyoheiHagimoto	0:0e0631af0305	65	### Types
RyoheiHagimoto	0:0e0631af0305	66
RyoheiHagimoto	0:0e0631af0305	67	There are several types representing 128-bit register as a vector of packed values, each type is
RyoheiHagimoto	0:0e0631af0305	68	implemented as a structure based on a one SIMD register.
RyoheiHagimoto	0:0e0631af0305	69
RyoheiHagimoto	0:0e0631af0305	70	- cv::v_uint8x16 and cv::v_int8x16: sixteen 8-bit integer values (unsigned/signed) - char
RyoheiHagimoto	0:0e0631af0305	71	- cv::v_uint16x8 and cv::v_int16x8: eight 16-bit integer values (unsigned/signed) - short
RyoheiHagimoto	0:0e0631af0305	72	- cv::v_uint32x4 and cv::v_int32x4: four 32-bit integer values (unsgined/signed) - int
RyoheiHagimoto	0:0e0631af0305	73	- cv::v_uint64x2 and cv::v_int64x2: two 64-bit integer values (unsigned/signed) - int64
RyoheiHagimoto	0:0e0631af0305	74	- cv::v_float32x4: four 32-bit floating point values (signed) - float
RyoheiHagimoto	0:0e0631af0305	75	- cv::v_float64x2: two 64-bit floating point valies (signed) - double
RyoheiHagimoto	0:0e0631af0305	76
RyoheiHagimoto	0:0e0631af0305	77	@note
RyoheiHagimoto	0:0e0631af0305	78	cv::v_float64x2 is not implemented in NEON variant, if you want to use this type, don't forget to
RyoheiHagimoto	0:0e0631af0305	79	check the CV_SIMD128_64F preprocessor definition:
RyoheiHagimoto	0:0e0631af0305	80	@code
RyoheiHagimoto	0:0e0631af0305	81	#if CV_SIMD128_64F
RyoheiHagimoto	0:0e0631af0305	82	//...
RyoheiHagimoto	0:0e0631af0305	83	#endif
RyoheiHagimoto	0:0e0631af0305	84	@endcode
RyoheiHagimoto	0:0e0631af0305	85
RyoheiHagimoto	0:0e0631af0305	86	### Load and store operations
RyoheiHagimoto	0:0e0631af0305	87
RyoheiHagimoto	0:0e0631af0305	88	These operations allow to set contents of the register explicitly or by loading it from some memory
RyoheiHagimoto	0:0e0631af0305	89	block and to save contents of the register to memory block.
RyoheiHagimoto	0:0e0631af0305	90
RyoheiHagimoto	0:0e0631af0305	91	- Constructors:
RyoheiHagimoto	0:0e0631af0305	92	@ref v_reg::v_reg(const _Tp *ptr) "from memory",
RyoheiHagimoto	0:0e0631af0305	93	@ref v_reg::v_reg(_Tp s0, _Tp s1) "from two values", ...
RyoheiHagimoto	0:0e0631af0305	94	- Other create methods:
RyoheiHagimoto	0:0e0631af0305	95	@ref v_setall_s8, @ref v_setall_u8, ...,
RyoheiHagimoto	0:0e0631af0305	96	@ref v_setzero_u8, @ref v_setzero_s8, ...
RyoheiHagimoto	0:0e0631af0305	97	- Memory operations:
RyoheiHagimoto	0:0e0631af0305	98	@ref v_load, @ref v_load_aligned, @ref v_load_halves,
RyoheiHagimoto	0:0e0631af0305	99	@ref v_store, @ref v_store_aligned,
RyoheiHagimoto	0:0e0631af0305	100	@ref v_store_high, @ref v_store_low
RyoheiHagimoto	0:0e0631af0305	101
RyoheiHagimoto	0:0e0631af0305	102	### Value reordering
RyoheiHagimoto	0:0e0631af0305	103
RyoheiHagimoto	0:0e0631af0305	104	These operations allow to reorder or recombine elements in one or multiple vectors.
RyoheiHagimoto	0:0e0631af0305	105
RyoheiHagimoto	0:0e0631af0305	106	- Interleave, deinterleave (2, 3 and 4 channels): @ref v_load_deinterleave, @ref v_store_interleave
RyoheiHagimoto	0:0e0631af0305	107	- Expand: @ref v_load_expand, @ref v_load_expand_q, @ref v_expand
RyoheiHagimoto	0:0e0631af0305	108	- Pack: @ref v_pack, @ref v_pack_u, @ref v_rshr_pack, @ref v_rshr_pack_u,
RyoheiHagimoto	0:0e0631af0305	109	@ref v_pack_store, @ref v_pack_u_store, @ref v_rshr_pack_store, @ref v_rshr_pack_u_store
RyoheiHagimoto	0:0e0631af0305	110	- Recombine: @ref v_zip, @ref v_recombine, @ref v_combine_low, @ref v_combine_high
RyoheiHagimoto	0:0e0631af0305	111	- Extract: @ref v_extract
RyoheiHagimoto	0:0e0631af0305	112
RyoheiHagimoto	0:0e0631af0305	113
RyoheiHagimoto	0:0e0631af0305	114	### Arithmetic, bitwise and comparison operations
RyoheiHagimoto	0:0e0631af0305	115
RyoheiHagimoto	0:0e0631af0305	116	Element-wise binary and unary operations.
RyoheiHagimoto	0:0e0631af0305	117
RyoheiHagimoto	0:0e0631af0305	118	- Arithmetics:
RyoheiHagimoto	0:0e0631af0305	119	@ref operator +(const v_reg &a, const v_reg &b) "+",
RyoheiHagimoto	0:0e0631af0305	120	@ref operator -(const v_reg &a, const v_reg &b) "-",
RyoheiHagimoto	0:0e0631af0305	121	@ref operator (const v_reg &a, const v_reg &b) "",
RyoheiHagimoto	0:0e0631af0305	122	@ref operator /(const v_reg &a, const v_reg &b) "/",
RyoheiHagimoto	0:0e0631af0305	123	@ref v_mul_expand
RyoheiHagimoto	0:0e0631af0305	124
RyoheiHagimoto	0:0e0631af0305	125	- Non-saturating arithmetics: @ref v_add_wrap, @ref v_sub_wrap
RyoheiHagimoto	0:0e0631af0305	126
RyoheiHagimoto	0:0e0631af0305	127	- Bitwise shifts:
RyoheiHagimoto	0:0e0631af0305	128	@ref operator <<(const v_reg &a, int s) "<<",
RyoheiHagimoto	0:0e0631af0305	129	@ref operator >>(const v_reg &a, int s) ">>",
RyoheiHagimoto	0:0e0631af0305	130	@ref v_shl, @ref v_shr
RyoheiHagimoto	0:0e0631af0305	131
RyoheiHagimoto	0:0e0631af0305	132	- Bitwise logic:
RyoheiHagimoto	0:0e0631af0305	133	@ref operator&(const v_reg &a, const v_reg &b) "&",
RyoheiHagimoto	0:0e0631af0305	134	@ref operator \|(const v_reg &a, const v_reg &b) "\|",
RyoheiHagimoto	0:0e0631af0305	135	@ref operator ^(const v_reg &a, const v_reg &b) "^",
RyoheiHagimoto	0:0e0631af0305	136	@ref operator ~(const v_reg &a) "~"
RyoheiHagimoto	0:0e0631af0305	137
RyoheiHagimoto	0:0e0631af0305	138	- Comparison:
RyoheiHagimoto	0:0e0631af0305	139	@ref operator >(const v_reg &a, const v_reg &b) ">",
RyoheiHagimoto	0:0e0631af0305	140	@ref operator >=(const v_reg &a, const v_reg &b) ">=",
RyoheiHagimoto	0:0e0631af0305	141	@ref operator <(const v_reg &a, const v_reg &b) "<",
RyoheiHagimoto	0:0e0631af0305	142	@ref operator <=(const v_reg &a, const v_reg &b) "<=",
RyoheiHagimoto	0:0e0631af0305	143	@ref operator==(const v_reg &a, const v_reg &b) "==",
RyoheiHagimoto	0:0e0631af0305	144	@ref operator !=(const v_reg &a, const v_reg &b) "!="
RyoheiHagimoto	0:0e0631af0305	145
RyoheiHagimoto	0:0e0631af0305	146	- min/max: @ref v_min, @ref v_max
RyoheiHagimoto	0:0e0631af0305	147
RyoheiHagimoto	0:0e0631af0305	148	### Reduce and mask
RyoheiHagimoto	0:0e0631af0305	149
RyoheiHagimoto	0:0e0631af0305	150	Most of these operations return only one value.
RyoheiHagimoto	0:0e0631af0305	151
RyoheiHagimoto	0:0e0631af0305	152	- Reduce: @ref v_reduce_min, @ref v_reduce_max, @ref v_reduce_sum
RyoheiHagimoto	0:0e0631af0305	153	- Mask: @ref v_signmask, @ref v_check_all, @ref v_check_any, @ref v_select
RyoheiHagimoto	0:0e0631af0305	154
RyoheiHagimoto	0:0e0631af0305	155	### Other math
RyoheiHagimoto	0:0e0631af0305	156
RyoheiHagimoto	0:0e0631af0305	157	- Some frequent operations: @ref v_sqrt, @ref v_invsqrt, @ref v_magnitude, @ref v_sqr_magnitude
RyoheiHagimoto	0:0e0631af0305	158	- Absolute values: @ref v_abs, @ref v_absdiff
RyoheiHagimoto	0:0e0631af0305	159
RyoheiHagimoto	0:0e0631af0305	160	### Conversions
RyoheiHagimoto	0:0e0631af0305	161
RyoheiHagimoto	0:0e0631af0305	162	Different type conversions and casts:
RyoheiHagimoto	0:0e0631af0305	163
RyoheiHagimoto	0:0e0631af0305	164	- Rounding: @ref v_round, @ref v_floor, @ref v_ceil, @ref v_trunc,
RyoheiHagimoto	0:0e0631af0305	165	- To float: @ref v_cvt_f32, @ref v_cvt_f64
RyoheiHagimoto	0:0e0631af0305	166	- Reinterpret: @ref v_reinterpret_as_u8, @ref v_reinterpret_as_s8, ...
RyoheiHagimoto	0:0e0631af0305	167
RyoheiHagimoto	0:0e0631af0305	168	### Matrix operations
RyoheiHagimoto	0:0e0631af0305	169
RyoheiHagimoto	0:0e0631af0305	170	In these operations vectors represent matrix rows/columns: @ref v_dotprod, @ref v_matmul, @ref v_transpose4x4
RyoheiHagimoto	0:0e0631af0305	171
RyoheiHagimoto	0:0e0631af0305	172	### Usability
RyoheiHagimoto	0:0e0631af0305	173
RyoheiHagimoto	0:0e0631af0305	174	Most operations are implemented only for some subset of the available types, following matrices
RyoheiHagimoto	0:0e0631af0305	175	shows the applicability of different operations to the types.
RyoheiHagimoto	0:0e0631af0305	176
RyoheiHagimoto	0:0e0631af0305	177	Regular integers:
RyoheiHagimoto	0:0e0631af0305	178
RyoheiHagimoto	0:0e0631af0305	179	\| Operations\\Types \| uint 8x16 \| int 8x16 \| uint 16x8 \| int 16x8 \| uint 32x4 \| int 32x4 \|
RyoheiHagimoto	0:0e0631af0305	180	\|-------------------\|:-:\|:-:\|:-:\|:-:\|:-:\|:-:\|
RyoheiHagimoto	0:0e0631af0305	181	\|load, store \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	182	\|interleave \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	183	\|expand \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	184	\|expand_q \| x \| x \| \| \| \| \|
RyoheiHagimoto	0:0e0631af0305	185	\|add, sub \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	186	\|add_wrap, sub_wrap \| x \| x \| x \| x \| \| \|
RyoheiHagimoto	0:0e0631af0305	187	\|mul \| \| \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	188	\|mul_expand \| \| \| x \| x \| x \| \|
RyoheiHagimoto	0:0e0631af0305	189	\|compare \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	190	\|shift \| \| \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	191	\|dotprod \| \| \| \| x \| \| \|
RyoheiHagimoto	0:0e0631af0305	192	\|logical \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	193	\|min, max \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	194	\|absdiff \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	195	\|reduce \| \| \| \| \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	196	\|mask \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	197	\|pack \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	198	\|pack_u \| x \| \| x \| \| \| \|
RyoheiHagimoto	0:0e0631af0305	199	\|unpack \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	200	\|extract \| x \| x \| x \| x \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	201	\|cvt_flt32 \| \| \| \| \| \| x \|
RyoheiHagimoto	0:0e0631af0305	202	\|cvt_flt64 \| \| \| \| \| \| x \|
RyoheiHagimoto	0:0e0631af0305	203	\|transpose4x4 \| \| \| \| \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	204
RyoheiHagimoto	0:0e0631af0305	205	Big integers:
RyoheiHagimoto	0:0e0631af0305	206
RyoheiHagimoto	0:0e0631af0305	207	\| Operations\\Types \| uint 64x2 \| int 64x2 \|
RyoheiHagimoto	0:0e0631af0305	208	\|-------------------\|:-:\|:-:\|
RyoheiHagimoto	0:0e0631af0305	209	\|load, store \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	210	\|add, sub \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	211	\|shift \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	212	\|logical \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	213	\|extract \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	214
RyoheiHagimoto	0:0e0631af0305	215	Floating point:
RyoheiHagimoto	0:0e0631af0305	216
RyoheiHagimoto	0:0e0631af0305	217	\| Operations\\Types \| float 32x4 \| float 64x2 \|
RyoheiHagimoto	0:0e0631af0305	218	\|-------------------\|:-:\|:-:\|
RyoheiHagimoto	0:0e0631af0305	219	\|load, store \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	220	\|interleave \| x \| \|
RyoheiHagimoto	0:0e0631af0305	221	\|add, sub \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	222	\|mul \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	223	\|div \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	224	\|compare \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	225	\|min, max \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	226	\|absdiff \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	227	\|reduce \| x \| \|
RyoheiHagimoto	0:0e0631af0305	228	\|mask \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	229	\|unpack \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	230	\|cvt_flt32 \| \| x \|
RyoheiHagimoto	0:0e0631af0305	231	\|cvt_flt64 \| x \| \|
RyoheiHagimoto	0:0e0631af0305	232	\|sqrt, abs \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	233	\|float math \| x \| x \|
RyoheiHagimoto	0:0e0631af0305	234	\|transpose4x4 \| x \| \|
RyoheiHagimoto	0:0e0631af0305	235
RyoheiHagimoto	0:0e0631af0305	236
RyoheiHagimoto	0:0e0631af0305	237	@{ */
RyoheiHagimoto	0:0e0631af0305	238
RyoheiHagimoto	0:0e0631af0305	239	template<typename _Tp, int n> struct v_reg
RyoheiHagimoto	0:0e0631af0305	240	{
RyoheiHagimoto	0:0e0631af0305	241	//! @cond IGNORED
RyoheiHagimoto	0:0e0631af0305	242	typedef _Tp lane_type;
RyoheiHagimoto	0:0e0631af0305	243	typedef v_reg<typename V_TypeTraits<_Tp>::int_type, n> int_vec;
RyoheiHagimoto	0:0e0631af0305	244	typedef v_reg<typename V_TypeTraits<_Tp>::abs_type, n> abs_vec;
RyoheiHagimoto	0:0e0631af0305	245	enum { nlanes = n };
RyoheiHagimoto	0:0e0631af0305	246	// !@endcond
RyoheiHagimoto	0:0e0631af0305	247
RyoheiHagimoto	0:0e0631af0305	248	/** @brief Constructor
RyoheiHagimoto	0:0e0631af0305	249
RyoheiHagimoto	0:0e0631af0305	250	Initializes register with data from memory
RyoheiHagimoto	0:0e0631af0305	251	@param ptr pointer to memory block with data for register */
RyoheiHagimoto	0:0e0631af0305	252	explicit v_reg(const _Tp* ptr) { for( int i = 0; i < n; i++ ) s[i] = ptr[i]; }
RyoheiHagimoto	0:0e0631af0305	253
RyoheiHagimoto	0:0e0631af0305	254	/** @brief Constructor
RyoheiHagimoto	0:0e0631af0305	255
RyoheiHagimoto	0:0e0631af0305	256	Initializes register with two 64-bit values */
RyoheiHagimoto	0:0e0631af0305	257	v_reg(_Tp s0, _Tp s1) { s[0] = s0; s[1] = s1; }
RyoheiHagimoto	0:0e0631af0305	258
RyoheiHagimoto	0:0e0631af0305	259	/** @brief Constructor
RyoheiHagimoto	0:0e0631af0305	260
RyoheiHagimoto	0:0e0631af0305	261	Initializes register with four 32-bit values */
RyoheiHagimoto	0:0e0631af0305	262	v_reg(_Tp s0, _Tp s1, _Tp s2, _Tp s3) { s[0] = s0; s[1] = s1; s[2] = s2; s[3] = s3; }
RyoheiHagimoto	0:0e0631af0305	263
RyoheiHagimoto	0:0e0631af0305	264	/** @brief Constructor
RyoheiHagimoto	0:0e0631af0305	265
RyoheiHagimoto	0:0e0631af0305	266	Initializes register with eight 16-bit values */
RyoheiHagimoto	0:0e0631af0305	267	v_reg(_Tp s0, _Tp s1, _Tp s2, _Tp s3,
RyoheiHagimoto	0:0e0631af0305	268	_Tp s4, _Tp s5, _Tp s6, _Tp s7)
RyoheiHagimoto	0:0e0631af0305	269	{
RyoheiHagimoto	0:0e0631af0305	270	s[0] = s0; s[1] = s1; s[2] = s2; s[3] = s3;
RyoheiHagimoto	0:0e0631af0305	271	s[4] = s4; s[5] = s5; s[6] = s6; s[7] = s7;
RyoheiHagimoto	0:0e0631af0305	272	}
RyoheiHagimoto	0:0e0631af0305	273
RyoheiHagimoto	0:0e0631af0305	274	/** @brief Constructor
RyoheiHagimoto	0:0e0631af0305	275
RyoheiHagimoto	0:0e0631af0305	276	Initializes register with sixteen 8-bit values */
RyoheiHagimoto	0:0e0631af0305	277	v_reg(_Tp s0, _Tp s1, _Tp s2, _Tp s3,
RyoheiHagimoto	0:0e0631af0305	278	_Tp s4, _Tp s5, _Tp s6, _Tp s7,
RyoheiHagimoto	0:0e0631af0305	279	_Tp s8, _Tp s9, _Tp s10, _Tp s11,
RyoheiHagimoto	0:0e0631af0305	280	_Tp s12, _Tp s13, _Tp s14, _Tp s15)
RyoheiHagimoto	0:0e0631af0305	281	{
RyoheiHagimoto	0:0e0631af0305	282	s[0] = s0; s[1] = s1; s[2] = s2; s[3] = s3;
RyoheiHagimoto	0:0e0631af0305	283	s[4] = s4; s[5] = s5; s[6] = s6; s[7] = s7;
RyoheiHagimoto	0:0e0631af0305	284	s[8] = s8; s[9] = s9; s[10] = s10; s[11] = s11;
RyoheiHagimoto	0:0e0631af0305	285	s[12] = s12; s[13] = s13; s[14] = s14; s[15] = s15;
RyoheiHagimoto	0:0e0631af0305	286	}
RyoheiHagimoto	0:0e0631af0305	287
RyoheiHagimoto	0:0e0631af0305	288	/** @brief Default constructor
RyoheiHagimoto	0:0e0631af0305	289
RyoheiHagimoto	0:0e0631af0305	290	Does not initialize anything*/
RyoheiHagimoto	0:0e0631af0305	291	v_reg() {}
RyoheiHagimoto	0:0e0631af0305	292
RyoheiHagimoto	0:0e0631af0305	293	/** @brief Copy constructor */
RyoheiHagimoto	0:0e0631af0305	294	v_reg(const v_reg<_Tp, n> & r)
RyoheiHagimoto	0:0e0631af0305	295	{
RyoheiHagimoto	0:0e0631af0305	296	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	297	s[i] = r.s[i];
RyoheiHagimoto	0:0e0631af0305	298	}
RyoheiHagimoto	0:0e0631af0305	299	/** @brief Access first value
RyoheiHagimoto	0:0e0631af0305	300
RyoheiHagimoto	0:0e0631af0305	301	Returns value of the first lane according to register type, for example:
RyoheiHagimoto	0:0e0631af0305	302	@code{.cpp}
RyoheiHagimoto	0:0e0631af0305	303	v_int32x4 r(1, 2, 3, 4);
RyoheiHagimoto	0:0e0631af0305	304	int v = r.get0(); // returns 1
RyoheiHagimoto	0:0e0631af0305	305	v_uint64x2 r(1, 2);
RyoheiHagimoto	0:0e0631af0305	306	uint64_t v = r.get0(); // returns 1
RyoheiHagimoto	0:0e0631af0305	307	@endcode
RyoheiHagimoto	0:0e0631af0305	308	*/
RyoheiHagimoto	0:0e0631af0305	309	_Tp get0() const { return s[0]; }
RyoheiHagimoto	0:0e0631af0305	310
RyoheiHagimoto	0:0e0631af0305	311	//! @cond IGNORED
RyoheiHagimoto	0:0e0631af0305	312	_Tp get(const int i) const { return s[i]; }
RyoheiHagimoto	0:0e0631af0305	313	v_reg<_Tp, n> high() const
RyoheiHagimoto	0:0e0631af0305	314	{
RyoheiHagimoto	0:0e0631af0305	315	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	316	int i;
RyoheiHagimoto	0:0e0631af0305	317	for( i = 0; i < n/2; i++ )
RyoheiHagimoto	0:0e0631af0305	318	{
RyoheiHagimoto	0:0e0631af0305	319	c.s[i] = s[i+(n/2)];
RyoheiHagimoto	0:0e0631af0305	320	c.s[i+(n/2)] = 0;
RyoheiHagimoto	0:0e0631af0305	321	}
RyoheiHagimoto	0:0e0631af0305	322	return c;
RyoheiHagimoto	0:0e0631af0305	323	}
RyoheiHagimoto	0:0e0631af0305	324
RyoheiHagimoto	0:0e0631af0305	325	static v_reg<_Tp, n> zero()
RyoheiHagimoto	0:0e0631af0305	326	{
RyoheiHagimoto	0:0e0631af0305	327	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	328	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	329	c.s[i] = (_Tp)0;
RyoheiHagimoto	0:0e0631af0305	330	return c;
RyoheiHagimoto	0:0e0631af0305	331	}
RyoheiHagimoto	0:0e0631af0305	332
RyoheiHagimoto	0:0e0631af0305	333	static v_reg<_Tp, n> all(_Tp s)
RyoheiHagimoto	0:0e0631af0305	334	{
RyoheiHagimoto	0:0e0631af0305	335	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	336	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	337	c.s[i] = s;
RyoheiHagimoto	0:0e0631af0305	338	return c;
RyoheiHagimoto	0:0e0631af0305	339	}
RyoheiHagimoto	0:0e0631af0305	340
RyoheiHagimoto	0:0e0631af0305	341	template<typename _Tp2, int n2> v_reg<_Tp2, n2> reinterpret_as() const
RyoheiHagimoto	0:0e0631af0305	342	{
RyoheiHagimoto	0:0e0631af0305	343	size_t bytes = std::min(sizeof(_Tp2)n2, sizeof(_Tp)n);
RyoheiHagimoto	0:0e0631af0305	344	v_reg<_Tp2, n2> c;
RyoheiHagimoto	0:0e0631af0305	345	std::memcpy(&c.s[0], &s[0], bytes);
RyoheiHagimoto	0:0e0631af0305	346	return c;
RyoheiHagimoto	0:0e0631af0305	347	}
RyoheiHagimoto	0:0e0631af0305	348
RyoheiHagimoto	0:0e0631af0305	349	_Tp s[n];
RyoheiHagimoto	0:0e0631af0305	350	//! @endcond
RyoheiHagimoto	0:0e0631af0305	351	};
RyoheiHagimoto	0:0e0631af0305	352
RyoheiHagimoto	0:0e0631af0305	353	/** @brief Sixteen 8-bit unsigned integer values */
RyoheiHagimoto	0:0e0631af0305	354	typedef v_reg<uchar, 16> v_uint8x16;
RyoheiHagimoto	0:0e0631af0305	355	/** @brief Sixteen 8-bit signed integer values */
RyoheiHagimoto	0:0e0631af0305	356	typedef v_reg<schar, 16> v_int8x16;
RyoheiHagimoto	0:0e0631af0305	357	/** @brief Eight 16-bit unsigned integer values */
RyoheiHagimoto	0:0e0631af0305	358	typedef v_reg<ushort, 8> v_uint16x8;
RyoheiHagimoto	0:0e0631af0305	359	/** @brief Eight 16-bit signed integer values */
RyoheiHagimoto	0:0e0631af0305	360	typedef v_reg<short, 8> v_int16x8;
RyoheiHagimoto	0:0e0631af0305	361	/** @brief Four 32-bit unsigned integer values */
RyoheiHagimoto	0:0e0631af0305	362	typedef v_reg<unsigned, 4> v_uint32x4;
RyoheiHagimoto	0:0e0631af0305	363	/** @brief Four 32-bit signed integer values */
RyoheiHagimoto	0:0e0631af0305	364	typedef v_reg<int, 4> v_int32x4;
RyoheiHagimoto	0:0e0631af0305	365	/** @brief Four 32-bit floating point values (single precision) */
RyoheiHagimoto	0:0e0631af0305	366	typedef v_reg<float, 4> v_float32x4;
RyoheiHagimoto	0:0e0631af0305	367	/** @brief Two 64-bit floating point values (double precision) */
RyoheiHagimoto	0:0e0631af0305	368	typedef v_reg<double, 2> v_float64x2;
RyoheiHagimoto	0:0e0631af0305	369	/** @brief Two 64-bit unsigned integer values */
RyoheiHagimoto	0:0e0631af0305	370	typedef v_reg<uint64, 2> v_uint64x2;
RyoheiHagimoto	0:0e0631af0305	371	/** @brief Two 64-bit signed integer values */
RyoheiHagimoto	0:0e0631af0305	372	typedef v_reg<int64, 2> v_int64x2;
RyoheiHagimoto	0:0e0631af0305	373
RyoheiHagimoto	0:0e0631af0305	374	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	375	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	376	#define OPENCV_HAL_IMPL_BIN_OP(bin_op) \
RyoheiHagimoto	0:0e0631af0305	377	template<typename _Tp, int n> inline v_reg<_Tp, n> \
RyoheiHagimoto	0:0e0631af0305	378	operator bin_op (const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
RyoheiHagimoto	0:0e0631af0305	379	{ \
RyoheiHagimoto	0:0e0631af0305	380	v_reg<_Tp, n> c; \
RyoheiHagimoto	0:0e0631af0305	381	for( int i = 0; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	382	c.s[i] = saturate_cast<_Tp>(a.s[i] bin_op b.s[i]); \
RyoheiHagimoto	0:0e0631af0305	383	return c; \
RyoheiHagimoto	0:0e0631af0305	384	} \
RyoheiHagimoto	0:0e0631af0305	385	template<typename _Tp, int n> inline v_reg<_Tp, n>& \
RyoheiHagimoto	0:0e0631af0305	386	operator bin_op##= (v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
RyoheiHagimoto	0:0e0631af0305	387	{ \
RyoheiHagimoto	0:0e0631af0305	388	for( int i = 0; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	389	a.s[i] = saturate_cast<_Tp>(a.s[i] bin_op b.s[i]); \
RyoheiHagimoto	0:0e0631af0305	390	return a; \
RyoheiHagimoto	0:0e0631af0305	391	}
RyoheiHagimoto	0:0e0631af0305	392
RyoheiHagimoto	0:0e0631af0305	393	/** @brief Add values
RyoheiHagimoto	0:0e0631af0305	394
RyoheiHagimoto	0:0e0631af0305	395	For all types. */
RyoheiHagimoto	0:0e0631af0305	396	OPENCV_HAL_IMPL_BIN_OP(+)
RyoheiHagimoto	0:0e0631af0305	397
RyoheiHagimoto	0:0e0631af0305	398	/** @brief Subtract values
RyoheiHagimoto	0:0e0631af0305	399
RyoheiHagimoto	0:0e0631af0305	400	For all types. */
RyoheiHagimoto	0:0e0631af0305	401	OPENCV_HAL_IMPL_BIN_OP(-)
RyoheiHagimoto	0:0e0631af0305	402
RyoheiHagimoto	0:0e0631af0305	403	/** @brief Multiply values
RyoheiHagimoto	0:0e0631af0305	404
RyoheiHagimoto	0:0e0631af0305	405	For 16- and 32-bit integer types and floating types. */
RyoheiHagimoto	0:0e0631af0305	406	OPENCV_HAL_IMPL_BIN_OP(*)
RyoheiHagimoto	0:0e0631af0305	407
RyoheiHagimoto	0:0e0631af0305	408	/** @brief Divide values
RyoheiHagimoto	0:0e0631af0305	409
RyoheiHagimoto	0:0e0631af0305	410	For floating types only. */
RyoheiHagimoto	0:0e0631af0305	411	OPENCV_HAL_IMPL_BIN_OP(/)
RyoheiHagimoto	0:0e0631af0305	412
RyoheiHagimoto	0:0e0631af0305	413	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	414	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	415	#define OPENCV_HAL_IMPL_BIT_OP(bit_op) \
RyoheiHagimoto	0:0e0631af0305	416	template<typename _Tp, int n> inline v_reg<_Tp, n> operator bit_op \
RyoheiHagimoto	0:0e0631af0305	417	(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
RyoheiHagimoto	0:0e0631af0305	418	{ \
RyoheiHagimoto	0:0e0631af0305	419	v_reg<_Tp, n> c; \
RyoheiHagimoto	0:0e0631af0305	420	typedef typename V_TypeTraits<_Tp>::int_type itype; \
RyoheiHagimoto	0:0e0631af0305	421	for( int i = 0; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	422	c.s[i] = V_TypeTraits<_Tp>::reinterpret_from_int((itype)(V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) bit_op \
RyoheiHagimoto	0:0e0631af0305	423	V_TypeTraits<_Tp>::reinterpret_int(b.s[i]))); \
RyoheiHagimoto	0:0e0631af0305	424	return c; \
RyoheiHagimoto	0:0e0631af0305	425	} \
RyoheiHagimoto	0:0e0631af0305	426	template<typename _Tp, int n> inline v_reg<_Tp, n>& operator \
RyoheiHagimoto	0:0e0631af0305	427	bit_op##= (v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
RyoheiHagimoto	0:0e0631af0305	428	{ \
RyoheiHagimoto	0:0e0631af0305	429	typedef typename V_TypeTraits<_Tp>::int_type itype; \
RyoheiHagimoto	0:0e0631af0305	430	for( int i = 0; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	431	a.s[i] = V_TypeTraits<_Tp>::reinterpret_from_int((itype)(V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) bit_op \
RyoheiHagimoto	0:0e0631af0305	432	V_TypeTraits<_Tp>::reinterpret_int(b.s[i]))); \
RyoheiHagimoto	0:0e0631af0305	433	return a; \
RyoheiHagimoto	0:0e0631af0305	434	}
RyoheiHagimoto	0:0e0631af0305	435
RyoheiHagimoto	0:0e0631af0305	436	/** @brief Bitwise AND
RyoheiHagimoto	0:0e0631af0305	437
RyoheiHagimoto	0:0e0631af0305	438	Only for integer types. */
RyoheiHagimoto	0:0e0631af0305	439	OPENCV_HAL_IMPL_BIT_OP(&)
RyoheiHagimoto	0:0e0631af0305	440
RyoheiHagimoto	0:0e0631af0305	441	/** @brief Bitwise OR
RyoheiHagimoto	0:0e0631af0305	442
RyoheiHagimoto	0:0e0631af0305	443	Only for integer types. */
RyoheiHagimoto	0:0e0631af0305	444	OPENCV_HAL_IMPL_BIT_OP(\|)
RyoheiHagimoto	0:0e0631af0305	445
RyoheiHagimoto	0:0e0631af0305	446	/** @brief Bitwise XOR
RyoheiHagimoto	0:0e0631af0305	447
RyoheiHagimoto	0:0e0631af0305	448	Only for integer types.*/
RyoheiHagimoto	0:0e0631af0305	449	OPENCV_HAL_IMPL_BIT_OP(^)
RyoheiHagimoto	0:0e0631af0305	450
RyoheiHagimoto	0:0e0631af0305	451	/** @brief Bitwise NOT
RyoheiHagimoto	0:0e0631af0305	452
RyoheiHagimoto	0:0e0631af0305	453	Only for integer types.*/
RyoheiHagimoto	0:0e0631af0305	454	template<typename _Tp, int n> inline v_reg<_Tp, n> operator ~ (const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	455	{
RyoheiHagimoto	0:0e0631af0305	456	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	457	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	458	{
RyoheiHagimoto	0:0e0631af0305	459	c.s[i] = V_TypeTraits<_Tp>::reinterpret_from_int(~V_TypeTraits<_Tp>::reinterpret_int(a.s[i]));
RyoheiHagimoto	0:0e0631af0305	460	}
RyoheiHagimoto	0:0e0631af0305	461	return c;
RyoheiHagimoto	0:0e0631af0305	462	}
RyoheiHagimoto	0:0e0631af0305	463
RyoheiHagimoto	0:0e0631af0305	464	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	465	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	466	#define OPENCV_HAL_IMPL_MATH_FUNC(func, cfunc, _Tp2) \
RyoheiHagimoto	0:0e0631af0305	467	template<typename _Tp, int n> inline v_reg<_Tp2, n> func(const v_reg<_Tp, n>& a) \
RyoheiHagimoto	0:0e0631af0305	468	{ \
RyoheiHagimoto	0:0e0631af0305	469	v_reg<_Tp2, n> c; \
RyoheiHagimoto	0:0e0631af0305	470	for( int i = 0; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	471	c.s[i] = cfunc(a.s[i]); \
RyoheiHagimoto	0:0e0631af0305	472	return c; \
RyoheiHagimoto	0:0e0631af0305	473	}
RyoheiHagimoto	0:0e0631af0305	474
RyoheiHagimoto	0:0e0631af0305	475	/** @brief Square root of elements
RyoheiHagimoto	0:0e0631af0305	476
RyoheiHagimoto	0:0e0631af0305	477	Only for floating point types.*/
RyoheiHagimoto	0:0e0631af0305	478	OPENCV_HAL_IMPL_MATH_FUNC(v_sqrt, std::sqrt, _Tp)
RyoheiHagimoto	0:0e0631af0305	479
RyoheiHagimoto	0:0e0631af0305	480	//! @cond IGNORED
RyoheiHagimoto	0:0e0631af0305	481	OPENCV_HAL_IMPL_MATH_FUNC(v_sin, std::sin, _Tp)
RyoheiHagimoto	0:0e0631af0305	482	OPENCV_HAL_IMPL_MATH_FUNC(v_cos, std::cos, _Tp)
RyoheiHagimoto	0:0e0631af0305	483	OPENCV_HAL_IMPL_MATH_FUNC(v_exp, std::exp, _Tp)
RyoheiHagimoto	0:0e0631af0305	484	OPENCV_HAL_IMPL_MATH_FUNC(v_log, std::log, _Tp)
RyoheiHagimoto	0:0e0631af0305	485	//! @endcond
RyoheiHagimoto	0:0e0631af0305	486
RyoheiHagimoto	0:0e0631af0305	487	/** @brief Absolute value of elements
RyoheiHagimoto	0:0e0631af0305	488
RyoheiHagimoto	0:0e0631af0305	489	Only for floating point types.*/
RyoheiHagimoto	0:0e0631af0305	490	OPENCV_HAL_IMPL_MATH_FUNC(v_abs, (typename V_TypeTraits<_Tp>::abs_type)std::abs,
RyoheiHagimoto	0:0e0631af0305	491	typename V_TypeTraits<_Tp>::abs_type)
RyoheiHagimoto	0:0e0631af0305	492
RyoheiHagimoto	0:0e0631af0305	493	/** @brief Round elements
RyoheiHagimoto	0:0e0631af0305	494
RyoheiHagimoto	0:0e0631af0305	495	Only for floating point types.*/
RyoheiHagimoto	0:0e0631af0305	496	OPENCV_HAL_IMPL_MATH_FUNC(v_round, cvRound, int)
RyoheiHagimoto	0:0e0631af0305	497
RyoheiHagimoto	0:0e0631af0305	498	/** @brief Floor elements
RyoheiHagimoto	0:0e0631af0305	499
RyoheiHagimoto	0:0e0631af0305	500	Only for floating point types.*/
RyoheiHagimoto	0:0e0631af0305	501	OPENCV_HAL_IMPL_MATH_FUNC(v_floor, cvFloor, int)
RyoheiHagimoto	0:0e0631af0305	502
RyoheiHagimoto	0:0e0631af0305	503	/** @brief Ceil elements
RyoheiHagimoto	0:0e0631af0305	504
RyoheiHagimoto	0:0e0631af0305	505	Only for floating point types.*/
RyoheiHagimoto	0:0e0631af0305	506	OPENCV_HAL_IMPL_MATH_FUNC(v_ceil, cvCeil, int)
RyoheiHagimoto	0:0e0631af0305	507
RyoheiHagimoto	0:0e0631af0305	508	/** @brief Truncate elements
RyoheiHagimoto	0:0e0631af0305	509
RyoheiHagimoto	0:0e0631af0305	510	Only for floating point types.*/
RyoheiHagimoto	0:0e0631af0305	511	OPENCV_HAL_IMPL_MATH_FUNC(v_trunc, int, int)
RyoheiHagimoto	0:0e0631af0305	512
RyoheiHagimoto	0:0e0631af0305	513	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	514	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	515	#define OPENCV_HAL_IMPL_MINMAX_FUNC(func, cfunc) \
RyoheiHagimoto	0:0e0631af0305	516	template<typename _Tp, int n> inline v_reg<_Tp, n> func(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
RyoheiHagimoto	0:0e0631af0305	517	{ \
RyoheiHagimoto	0:0e0631af0305	518	v_reg<_Tp, n> c; \
RyoheiHagimoto	0:0e0631af0305	519	for( int i = 0; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	520	c.s[i] = cfunc(a.s[i], b.s[i]); \
RyoheiHagimoto	0:0e0631af0305	521	return c; \
RyoheiHagimoto	0:0e0631af0305	522	}
RyoheiHagimoto	0:0e0631af0305	523
RyoheiHagimoto	0:0e0631af0305	524	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	525	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	526	#define OPENCV_HAL_IMPL_REDUCE_MINMAX_FUNC(func, cfunc) \
RyoheiHagimoto	0:0e0631af0305	527	template<typename _Tp, int n> inline _Tp func(const v_reg<_Tp, n>& a) \
RyoheiHagimoto	0:0e0631af0305	528	{ \
RyoheiHagimoto	0:0e0631af0305	529	_Tp c = a.s[0]; \
RyoheiHagimoto	0:0e0631af0305	530	for( int i = 1; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	531	c = cfunc(c, a.s[i]); \
RyoheiHagimoto	0:0e0631af0305	532	return c; \
RyoheiHagimoto	0:0e0631af0305	533	}
RyoheiHagimoto	0:0e0631af0305	534
RyoheiHagimoto	0:0e0631af0305	535	/** @brief Choose min values for each pair
RyoheiHagimoto	0:0e0631af0305	536
RyoheiHagimoto	0:0e0631af0305	537	Scheme:
RyoheiHagimoto	0:0e0631af0305	538	@code
RyoheiHagimoto	0:0e0631af0305	539	{A1 A2 ...}
RyoheiHagimoto	0:0e0631af0305	540	{B1 B2 ...}
RyoheiHagimoto	0:0e0631af0305	541	--------------
RyoheiHagimoto	0:0e0631af0305	542	{min(A1,B1) min(A2,B2) ...}
RyoheiHagimoto	0:0e0631af0305	543	@endcode
RyoheiHagimoto	0:0e0631af0305	544	For all types except 64-bit integer. */
RyoheiHagimoto	0:0e0631af0305	545	OPENCV_HAL_IMPL_MINMAX_FUNC(v_min, std::min)
RyoheiHagimoto	0:0e0631af0305	546
RyoheiHagimoto	0:0e0631af0305	547	/** @brief Choose max values for each pair
RyoheiHagimoto	0:0e0631af0305	548
RyoheiHagimoto	0:0e0631af0305	549	Scheme:
RyoheiHagimoto	0:0e0631af0305	550	@code
RyoheiHagimoto	0:0e0631af0305	551	{A1 A2 ...}
RyoheiHagimoto	0:0e0631af0305	552	{B1 B2 ...}
RyoheiHagimoto	0:0e0631af0305	553	--------------
RyoheiHagimoto	0:0e0631af0305	554	{max(A1,B1) max(A2,B2) ...}
RyoheiHagimoto	0:0e0631af0305	555	@endcode
RyoheiHagimoto	0:0e0631af0305	556	For all types except 64-bit integer. */
RyoheiHagimoto	0:0e0631af0305	557	OPENCV_HAL_IMPL_MINMAX_FUNC(v_max, std::max)
RyoheiHagimoto	0:0e0631af0305	558
RyoheiHagimoto	0:0e0631af0305	559	/** @brief Find one min value
RyoheiHagimoto	0:0e0631af0305	560
RyoheiHagimoto	0:0e0631af0305	561	Scheme:
RyoheiHagimoto	0:0e0631af0305	562	@code
RyoheiHagimoto	0:0e0631af0305	563	{A1 A2 A3 ...} => min(A1,A2,A3,...)
RyoheiHagimoto	0:0e0631af0305	564	@endcode
RyoheiHagimoto	0:0e0631af0305	565	For 32-bit integer and 32-bit floating point types. */
RyoheiHagimoto	0:0e0631af0305	566	OPENCV_HAL_IMPL_REDUCE_MINMAX_FUNC(v_reduce_min, std::min)
RyoheiHagimoto	0:0e0631af0305	567
RyoheiHagimoto	0:0e0631af0305	568	/** @brief Find one max value
RyoheiHagimoto	0:0e0631af0305	569
RyoheiHagimoto	0:0e0631af0305	570	Scheme:
RyoheiHagimoto	0:0e0631af0305	571	@code
RyoheiHagimoto	0:0e0631af0305	572	{A1 A2 A3 ...} => max(A1,A2,A3,...)
RyoheiHagimoto	0:0e0631af0305	573	@endcode
RyoheiHagimoto	0:0e0631af0305	574	For 32-bit integer and 32-bit floating point types. */
RyoheiHagimoto	0:0e0631af0305	575	OPENCV_HAL_IMPL_REDUCE_MINMAX_FUNC(v_reduce_max, std::max)
RyoheiHagimoto	0:0e0631af0305	576
RyoheiHagimoto	0:0e0631af0305	577	//! @cond IGNORED
RyoheiHagimoto	0:0e0631af0305	578	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	579	inline void v_minmax( const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b,
RyoheiHagimoto	0:0e0631af0305	580	v_reg<_Tp, n>& minval, v_reg<_Tp, n>& maxval )
RyoheiHagimoto	0:0e0631af0305	581	{
RyoheiHagimoto	0:0e0631af0305	582	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	583	{
RyoheiHagimoto	0:0e0631af0305	584	minval.s[i] = std::min(a.s[i], b.s[i]);
RyoheiHagimoto	0:0e0631af0305	585	maxval.s[i] = std::max(a.s[i], b.s[i]);
RyoheiHagimoto	0:0e0631af0305	586	}
RyoheiHagimoto	0:0e0631af0305	587	}
RyoheiHagimoto	0:0e0631af0305	588	//! @endcond
RyoheiHagimoto	0:0e0631af0305	589
RyoheiHagimoto	0:0e0631af0305	590	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	591	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	592	#define OPENCV_HAL_IMPL_CMP_OP(cmp_op) \
RyoheiHagimoto	0:0e0631af0305	593	template<typename _Tp, int n> \
RyoheiHagimoto	0:0e0631af0305	594	inline v_reg<_Tp, n> operator cmp_op(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
RyoheiHagimoto	0:0e0631af0305	595	{ \
RyoheiHagimoto	0:0e0631af0305	596	typedef typename V_TypeTraits<_Tp>::int_type itype; \
RyoheiHagimoto	0:0e0631af0305	597	v_reg<_Tp, n> c; \
RyoheiHagimoto	0:0e0631af0305	598	for( int i = 0; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	599	c.s[i] = V_TypeTraits<_Tp>::reinterpret_from_int((itype)-(int)(a.s[i] cmp_op b.s[i])); \
RyoheiHagimoto	0:0e0631af0305	600	return c; \
RyoheiHagimoto	0:0e0631af0305	601	}
RyoheiHagimoto	0:0e0631af0305	602
RyoheiHagimoto	0:0e0631af0305	603	/** @brief Less-than comparison
RyoheiHagimoto	0:0e0631af0305	604
RyoheiHagimoto	0:0e0631af0305	605	For all types except 64-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	606	OPENCV_HAL_IMPL_CMP_OP(<)
RyoheiHagimoto	0:0e0631af0305	607
RyoheiHagimoto	0:0e0631af0305	608	/** @brief Greater-than comparison
RyoheiHagimoto	0:0e0631af0305	609
RyoheiHagimoto	0:0e0631af0305	610	For all types except 64-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	611	OPENCV_HAL_IMPL_CMP_OP(>)
RyoheiHagimoto	0:0e0631af0305	612
RyoheiHagimoto	0:0e0631af0305	613	/** @brief Less-than or equal comparison
RyoheiHagimoto	0:0e0631af0305	614
RyoheiHagimoto	0:0e0631af0305	615	For all types except 64-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	616	OPENCV_HAL_IMPL_CMP_OP(<=)
RyoheiHagimoto	0:0e0631af0305	617
RyoheiHagimoto	0:0e0631af0305	618	/** @brief Greater-than or equal comparison
RyoheiHagimoto	0:0e0631af0305	619
RyoheiHagimoto	0:0e0631af0305	620	For all types except 64-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	621	OPENCV_HAL_IMPL_CMP_OP(>=)
RyoheiHagimoto	0:0e0631af0305	622
RyoheiHagimoto	0:0e0631af0305	623	/** @brief Equal comparison
RyoheiHagimoto	0:0e0631af0305	624
RyoheiHagimoto	0:0e0631af0305	625	For all types except 64-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	626	OPENCV_HAL_IMPL_CMP_OP(==)
RyoheiHagimoto	0:0e0631af0305	627
RyoheiHagimoto	0:0e0631af0305	628	/** @brief Not equal comparison
RyoheiHagimoto	0:0e0631af0305	629
RyoheiHagimoto	0:0e0631af0305	630	For all types except 64-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	631	OPENCV_HAL_IMPL_CMP_OP(!=)
RyoheiHagimoto	0:0e0631af0305	632
RyoheiHagimoto	0:0e0631af0305	633	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	634	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	635	#define OPENCV_HAL_IMPL_ADD_SUB_OP(func, bin_op, cast_op, _Tp2) \
RyoheiHagimoto	0:0e0631af0305	636	template<typename _Tp, int n> \
RyoheiHagimoto	0:0e0631af0305	637	inline v_reg<_Tp2, n> func(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b) \
RyoheiHagimoto	0:0e0631af0305	638	{ \
RyoheiHagimoto	0:0e0631af0305	639	typedef _Tp2 rtype; \
RyoheiHagimoto	0:0e0631af0305	640	v_reg<rtype, n> c; \
RyoheiHagimoto	0:0e0631af0305	641	for( int i = 0; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	642	c.s[i] = cast_op(a.s[i] bin_op b.s[i]); \
RyoheiHagimoto	0:0e0631af0305	643	return c; \
RyoheiHagimoto	0:0e0631af0305	644	}
RyoheiHagimoto	0:0e0631af0305	645
RyoheiHagimoto	0:0e0631af0305	646	/** @brief Add values without saturation
RyoheiHagimoto	0:0e0631af0305	647
RyoheiHagimoto	0:0e0631af0305	648	For 8- and 16-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	649	OPENCV_HAL_IMPL_ADD_SUB_OP(v_add_wrap, +, (_Tp), _Tp)
RyoheiHagimoto	0:0e0631af0305	650
RyoheiHagimoto	0:0e0631af0305	651	/** @brief Subtract values without saturation
RyoheiHagimoto	0:0e0631af0305	652
RyoheiHagimoto	0:0e0631af0305	653	For 8- and 16-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	654	OPENCV_HAL_IMPL_ADD_SUB_OP(v_sub_wrap, -, (_Tp), _Tp)
RyoheiHagimoto	0:0e0631af0305	655
RyoheiHagimoto	0:0e0631af0305	656	//! @cond IGNORED
RyoheiHagimoto	0:0e0631af0305	657	template<typename T> inline T _absdiff(T a, T b)
RyoheiHagimoto	0:0e0631af0305	658	{
RyoheiHagimoto	0:0e0631af0305	659	return a > b ? a - b : b - a;
RyoheiHagimoto	0:0e0631af0305	660	}
RyoheiHagimoto	0:0e0631af0305	661	//! @endcond
RyoheiHagimoto	0:0e0631af0305	662
RyoheiHagimoto	0:0e0631af0305	663	/** @brief Absolute difference
RyoheiHagimoto	0:0e0631af0305	664
RyoheiHagimoto	0:0e0631af0305	665	Returns \f$ \|a - b\| \f$ converted to corresponding unsigned type.
RyoheiHagimoto	0:0e0631af0305	666	Example:
RyoheiHagimoto	0:0e0631af0305	667	@code{.cpp}
RyoheiHagimoto	0:0e0631af0305	668	v_int32x4 a, b; // {1, 2, 3, 4} and {4, 3, 2, 1}
RyoheiHagimoto	0:0e0631af0305	669	v_uint32x4 c = v_absdiff(a, b); // result is {3, 1, 1, 3}
RyoheiHagimoto	0:0e0631af0305	670	@endcode
RyoheiHagimoto	0:0e0631af0305	671	For 8-, 16-, 32-bit integer source types. */
RyoheiHagimoto	0:0e0631af0305	672	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	673	inline v_reg<typename V_TypeTraits<_Tp>::abs_type, n> v_absdiff(const v_reg<_Tp, n>& a, const v_reg<_Tp, n> & b)
RyoheiHagimoto	0:0e0631af0305	674	{
RyoheiHagimoto	0:0e0631af0305	675	typedef typename V_TypeTraits<_Tp>::abs_type rtype;
RyoheiHagimoto	0:0e0631af0305	676	v_reg<rtype, n> c;
RyoheiHagimoto	0:0e0631af0305	677	const rtype mask = std::numeric_limits<_Tp>::is_signed ? (1 << (sizeof(rtype)*8 - 1)) : 0;
RyoheiHagimoto	0:0e0631af0305	678	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	679	{
RyoheiHagimoto	0:0e0631af0305	680	rtype ua = a.s[i] ^ mask;
RyoheiHagimoto	0:0e0631af0305	681	rtype ub = b.s[i] ^ mask;
RyoheiHagimoto	0:0e0631af0305	682	c.s[i] = _absdiff(ua, ub);
RyoheiHagimoto	0:0e0631af0305	683	}
RyoheiHagimoto	0:0e0631af0305	684	return c;
RyoheiHagimoto	0:0e0631af0305	685	}
RyoheiHagimoto	0:0e0631af0305	686
RyoheiHagimoto	0:0e0631af0305	687	/** @overload
RyoheiHagimoto	0:0e0631af0305	688
RyoheiHagimoto	0:0e0631af0305	689	For 32-bit floating point values */
RyoheiHagimoto	0:0e0631af0305	690	inline v_float32x4 v_absdiff(const v_float32x4& a, const v_float32x4& b)
RyoheiHagimoto	0:0e0631af0305	691	{
RyoheiHagimoto	0:0e0631af0305	692	v_float32x4 c;
RyoheiHagimoto	0:0e0631af0305	693	for( int i = 0; i < c.nlanes; i++ )
RyoheiHagimoto	0:0e0631af0305	694	c.s[i] = _absdiff(a.s[i], b.s[i]);
RyoheiHagimoto	0:0e0631af0305	695	return c;
RyoheiHagimoto	0:0e0631af0305	696	}
RyoheiHagimoto	0:0e0631af0305	697
RyoheiHagimoto	0:0e0631af0305	698	/** @overload
RyoheiHagimoto	0:0e0631af0305	699
RyoheiHagimoto	0:0e0631af0305	700	For 64-bit floating point values */
RyoheiHagimoto	0:0e0631af0305	701	inline v_float64x2 v_absdiff(const v_float64x2& a, const v_float64x2& b)
RyoheiHagimoto	0:0e0631af0305	702	{
RyoheiHagimoto	0:0e0631af0305	703	v_float64x2 c;
RyoheiHagimoto	0:0e0631af0305	704	for( int i = 0; i < c.nlanes; i++ )
RyoheiHagimoto	0:0e0631af0305	705	c.s[i] = _absdiff(a.s[i], b.s[i]);
RyoheiHagimoto	0:0e0631af0305	706	return c;
RyoheiHagimoto	0:0e0631af0305	707	}
RyoheiHagimoto	0:0e0631af0305	708
RyoheiHagimoto	0:0e0631af0305	709	/** @brief Inversed square root
RyoheiHagimoto	0:0e0631af0305	710
RyoheiHagimoto	0:0e0631af0305	711	Returns \f$ 1/sqrt(a) \f$
RyoheiHagimoto	0:0e0631af0305	712	For floating point types only. */
RyoheiHagimoto	0:0e0631af0305	713	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	714	inline v_reg<_Tp, n> v_invsqrt(const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	715	{
RyoheiHagimoto	0:0e0631af0305	716	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	717	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	718	c.s[i] = 1.f/std::sqrt(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	719	return c;
RyoheiHagimoto	0:0e0631af0305	720	}
RyoheiHagimoto	0:0e0631af0305	721
RyoheiHagimoto	0:0e0631af0305	722	/** @brief Magnitude
RyoheiHagimoto	0:0e0631af0305	723
RyoheiHagimoto	0:0e0631af0305	724	Returns \f$ sqrt(a^2 + b^2) \f$
RyoheiHagimoto	0:0e0631af0305	725	For floating point types only. */
RyoheiHagimoto	0:0e0631af0305	726	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	727	inline v_reg<_Tp, n> v_magnitude(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
RyoheiHagimoto	0:0e0631af0305	728	{
RyoheiHagimoto	0:0e0631af0305	729	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	730	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	731	c.s[i] = std::sqrt(a.s[i]a.s[i] + b.s[i]b.s[i]);
RyoheiHagimoto	0:0e0631af0305	732	return c;
RyoheiHagimoto	0:0e0631af0305	733	}
RyoheiHagimoto	0:0e0631af0305	734
RyoheiHagimoto	0:0e0631af0305	735	/** @brief Square of the magnitude
RyoheiHagimoto	0:0e0631af0305	736
RyoheiHagimoto	0:0e0631af0305	737	Returns \f$ a^2 + b^2 \f$
RyoheiHagimoto	0:0e0631af0305	738	For floating point types only. */
RyoheiHagimoto	0:0e0631af0305	739	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	740	inline v_reg<_Tp, n> v_sqr_magnitude(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
RyoheiHagimoto	0:0e0631af0305	741	{
RyoheiHagimoto	0:0e0631af0305	742	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	743	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	744	c.s[i] = a.s[i]a.s[i] + b.s[i]b.s[i];
RyoheiHagimoto	0:0e0631af0305	745	return c;
RyoheiHagimoto	0:0e0631af0305	746	}
RyoheiHagimoto	0:0e0631af0305	747
RyoheiHagimoto	0:0e0631af0305	748	/** @brief Multiply and add
RyoheiHagimoto	0:0e0631af0305	749
RyoheiHagimoto	0:0e0631af0305	750	Returns \f$ a*b + c \f$
RyoheiHagimoto	0:0e0631af0305	751	For floating point types only. */
RyoheiHagimoto	0:0e0631af0305	752	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	753	inline v_reg<_Tp, n> v_muladd(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b,
RyoheiHagimoto	0:0e0631af0305	754	const v_reg<_Tp, n>& c)
RyoheiHagimoto	0:0e0631af0305	755	{
RyoheiHagimoto	0:0e0631af0305	756	v_reg<_Tp, n> d;
RyoheiHagimoto	0:0e0631af0305	757	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	758	d.s[i] = a.s[i]*b.s[i] + c.s[i];
RyoheiHagimoto	0:0e0631af0305	759	return d;
RyoheiHagimoto	0:0e0631af0305	760	}
RyoheiHagimoto	0:0e0631af0305	761
RyoheiHagimoto	0:0e0631af0305	762	/** @brief Dot product of elements
RyoheiHagimoto	0:0e0631af0305	763
RyoheiHagimoto	0:0e0631af0305	764	Multiply values in two registers and sum adjacent result pairs.
RyoheiHagimoto	0:0e0631af0305	765	Scheme:
RyoheiHagimoto	0:0e0631af0305	766	@code
RyoheiHagimoto	0:0e0631af0305	767	{A1 A2 ...} // 16-bit
RyoheiHagimoto	0:0e0631af0305	768	x {B1 B2 ...} // 16-bit
RyoheiHagimoto	0:0e0631af0305	769	-------------
RyoheiHagimoto	0:0e0631af0305	770	{A1B1+A2B2 ...} // 32-bit
RyoheiHagimoto	0:0e0631af0305	771	@endcode
RyoheiHagimoto	0:0e0631af0305	772	Implemented only for 16-bit signed source type (v_int16x8).
RyoheiHagimoto	0:0e0631af0305	773	*/
RyoheiHagimoto	0:0e0631af0305	774	template<typename _Tp, int n> inline v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>
RyoheiHagimoto	0:0e0631af0305	775	v_dotprod(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
RyoheiHagimoto	0:0e0631af0305	776	{
RyoheiHagimoto	0:0e0631af0305	777	typedef typename V_TypeTraits<_Tp>::w_type w_type;
RyoheiHagimoto	0:0e0631af0305	778	v_reg<w_type, n/2> c;
RyoheiHagimoto	0:0e0631af0305	779	for( int i = 0; i < (n/2); i++ )
RyoheiHagimoto	0:0e0631af0305	780	c.s[i] = (w_type)a.s[i2]b.s[i2] + (w_type)a.s[i2+1]b.s[i2+1];
RyoheiHagimoto	0:0e0631af0305	781	return c;
RyoheiHagimoto	0:0e0631af0305	782	}
RyoheiHagimoto	0:0e0631af0305	783
RyoheiHagimoto	0:0e0631af0305	784	/** @brief Multiply and expand
RyoheiHagimoto	0:0e0631af0305	785
RyoheiHagimoto	0:0e0631af0305	786	Multiply values two registers and store results in two registers with wider pack type.
RyoheiHagimoto	0:0e0631af0305	787	Scheme:
RyoheiHagimoto	0:0e0631af0305	788	@code
RyoheiHagimoto	0:0e0631af0305	789	{A B C D} // 32-bit
RyoheiHagimoto	0:0e0631af0305	790	x {E F G H} // 32-bit
RyoheiHagimoto	0:0e0631af0305	791	---------------
RyoheiHagimoto	0:0e0631af0305	792	{AE BF} // 64-bit
RyoheiHagimoto	0:0e0631af0305	793	{CG DH} // 64-bit
RyoheiHagimoto	0:0e0631af0305	794	@endcode
RyoheiHagimoto	0:0e0631af0305	795	Example:
RyoheiHagimoto	0:0e0631af0305	796	@code{.cpp}
RyoheiHagimoto	0:0e0631af0305	797	v_uint32x4 a, b; // {1,2,3,4} and {2,2,2,2}
RyoheiHagimoto	0:0e0631af0305	798	v_uint64x2 c, d; // results
RyoheiHagimoto	0:0e0631af0305	799	v_mul_expand(a, b, c, d); // c, d = {2,4}, {6, 8}
RyoheiHagimoto	0:0e0631af0305	800	@endcode
RyoheiHagimoto	0:0e0631af0305	801	Implemented only for 16- and unsigned 32-bit source types (v_int16x8, v_uint16x8, v_uint32x4).
RyoheiHagimoto	0:0e0631af0305	802	*/
RyoheiHagimoto	0:0e0631af0305	803	template<typename _Tp, int n> inline void v_mul_expand(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b,
RyoheiHagimoto	0:0e0631af0305	804	v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& c,
RyoheiHagimoto	0:0e0631af0305	805	v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& d)
RyoheiHagimoto	0:0e0631af0305	806	{
RyoheiHagimoto	0:0e0631af0305	807	typedef typename V_TypeTraits<_Tp>::w_type w_type;
RyoheiHagimoto	0:0e0631af0305	808	for( int i = 0; i < (n/2); i++ )
RyoheiHagimoto	0:0e0631af0305	809	{
RyoheiHagimoto	0:0e0631af0305	810	c.s[i] = (w_type)a.s[i]*b.s[i];
RyoheiHagimoto	0:0e0631af0305	811	d.s[i] = (w_type)a.s[i+(n/2)]*b.s[i+(n/2)];
RyoheiHagimoto	0:0e0631af0305	812	}
RyoheiHagimoto	0:0e0631af0305	813	}
RyoheiHagimoto	0:0e0631af0305	814
RyoheiHagimoto	0:0e0631af0305	815	//! @cond IGNORED
RyoheiHagimoto	0:0e0631af0305	816	template<typename _Tp, int n> inline void v_hsum(const v_reg<_Tp, n>& a,
RyoheiHagimoto	0:0e0631af0305	817	v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& c)
RyoheiHagimoto	0:0e0631af0305	818	{
RyoheiHagimoto	0:0e0631af0305	819	typedef typename V_TypeTraits<_Tp>::w_type w_type;
RyoheiHagimoto	0:0e0631af0305	820	for( int i = 0; i < (n/2); i++ )
RyoheiHagimoto	0:0e0631af0305	821	{
RyoheiHagimoto	0:0e0631af0305	822	c.s[i] = (w_type)a.s[i2] + a.s[i2+1];
RyoheiHagimoto	0:0e0631af0305	823	}
RyoheiHagimoto	0:0e0631af0305	824	}
RyoheiHagimoto	0:0e0631af0305	825	//! @endcond
RyoheiHagimoto	0:0e0631af0305	826
RyoheiHagimoto	0:0e0631af0305	827	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	828	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	829	#define OPENCV_HAL_IMPL_SHIFT_OP(shift_op) \
RyoheiHagimoto	0:0e0631af0305	830	template<typename _Tp, int n> inline v_reg<_Tp, n> operator shift_op(const v_reg<_Tp, n>& a, int imm) \
RyoheiHagimoto	0:0e0631af0305	831	{ \
RyoheiHagimoto	0:0e0631af0305	832	v_reg<_Tp, n> c; \
RyoheiHagimoto	0:0e0631af0305	833	for( int i = 0; i < n; i++ ) \
RyoheiHagimoto	0:0e0631af0305	834	c.s[i] = (_Tp)(a.s[i] shift_op imm); \
RyoheiHagimoto	0:0e0631af0305	835	return c; \
RyoheiHagimoto	0:0e0631af0305	836	}
RyoheiHagimoto	0:0e0631af0305	837
RyoheiHagimoto	0:0e0631af0305	838	/** @brief Bitwise shift left
RyoheiHagimoto	0:0e0631af0305	839
RyoheiHagimoto	0:0e0631af0305	840	For 16-, 32- and 64-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	841	OPENCV_HAL_IMPL_SHIFT_OP(<<)
RyoheiHagimoto	0:0e0631af0305	842
RyoheiHagimoto	0:0e0631af0305	843	/** @brief Bitwise shift right
RyoheiHagimoto	0:0e0631af0305	844
RyoheiHagimoto	0:0e0631af0305	845	For 16-, 32- and 64-bit integer values. */
RyoheiHagimoto	0:0e0631af0305	846	OPENCV_HAL_IMPL_SHIFT_OP(>>)
RyoheiHagimoto	0:0e0631af0305	847
RyoheiHagimoto	0:0e0631af0305	848	/** @brief Sum packed values
RyoheiHagimoto	0:0e0631af0305	849
RyoheiHagimoto	0:0e0631af0305	850	Scheme:
RyoheiHagimoto	0:0e0631af0305	851	@code
RyoheiHagimoto	0:0e0631af0305	852	{A1 A2 A3 ...} => sum{A1,A2,A3,...}
RyoheiHagimoto	0:0e0631af0305	853	@endcode
RyoheiHagimoto	0:0e0631af0305	854	For 32-bit integer and 32-bit floating point types.*/
RyoheiHagimoto	0:0e0631af0305	855	template<typename _Tp, int n> inline typename V_TypeTraits<_Tp>::sum_type v_reduce_sum(const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	856	{
RyoheiHagimoto	0:0e0631af0305	857	typename V_TypeTraits<_Tp>::sum_type c = a.s[0];
RyoheiHagimoto	0:0e0631af0305	858	for( int i = 1; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	859	c += a.s[i];
RyoheiHagimoto	0:0e0631af0305	860	return c;
RyoheiHagimoto	0:0e0631af0305	861	}
RyoheiHagimoto	0:0e0631af0305	862
RyoheiHagimoto	0:0e0631af0305	863	/** @brief Get negative values mask
RyoheiHagimoto	0:0e0631af0305	864
RyoheiHagimoto	0:0e0631af0305	865	Returned value is a bit mask with bits set to 1 on places corresponding to negative packed values indexes.
RyoheiHagimoto	0:0e0631af0305	866	Example:
RyoheiHagimoto	0:0e0631af0305	867	@code{.cpp}
RyoheiHagimoto	0:0e0631af0305	868	v_int32x4 r; // set to {-1, -1, 1, 1}
RyoheiHagimoto	0:0e0631af0305	869	int mask = v_signmask(r); // mask = 3 <== 00000000 00000000 00000000 00000011
RyoheiHagimoto	0:0e0631af0305	870	@endcode
RyoheiHagimoto	0:0e0631af0305	871	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	872	template<typename _Tp, int n> inline int v_signmask(const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	873	{
RyoheiHagimoto	0:0e0631af0305	874	int mask = 0;
RyoheiHagimoto	0:0e0631af0305	875	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	876	mask \|= (V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) < 0) << i;
RyoheiHagimoto	0:0e0631af0305	877	return mask;
RyoheiHagimoto	0:0e0631af0305	878	}
RyoheiHagimoto	0:0e0631af0305	879
RyoheiHagimoto	0:0e0631af0305	880	/** @brief Check if all packed values are less than zero
RyoheiHagimoto	0:0e0631af0305	881
RyoheiHagimoto	0:0e0631af0305	882	Unsigned values will be casted to signed: `uchar 254 => char -2`.
RyoheiHagimoto	0:0e0631af0305	883	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	884	template<typename _Tp, int n> inline bool v_check_all(const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	885	{
RyoheiHagimoto	0:0e0631af0305	886	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	887	if( V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) >= 0 )
RyoheiHagimoto	0:0e0631af0305	888	return false;
RyoheiHagimoto	0:0e0631af0305	889	return true;
RyoheiHagimoto	0:0e0631af0305	890	}
RyoheiHagimoto	0:0e0631af0305	891
RyoheiHagimoto	0:0e0631af0305	892	/** @brief Check if any of packed values is less than zero
RyoheiHagimoto	0:0e0631af0305	893
RyoheiHagimoto	0:0e0631af0305	894	Unsigned values will be casted to signed: `uchar 254 => char -2`.
RyoheiHagimoto	0:0e0631af0305	895	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	896	template<typename _Tp, int n> inline bool v_check_any(const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	897	{
RyoheiHagimoto	0:0e0631af0305	898	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	899	if( V_TypeTraits<_Tp>::reinterpret_int(a.s[i]) < 0 )
RyoheiHagimoto	0:0e0631af0305	900	return true;
RyoheiHagimoto	0:0e0631af0305	901	return false;
RyoheiHagimoto	0:0e0631af0305	902	}
RyoheiHagimoto	0:0e0631af0305	903
RyoheiHagimoto	0:0e0631af0305	904	/** @brief Bitwise select
RyoheiHagimoto	0:0e0631af0305	905
RyoheiHagimoto	0:0e0631af0305	906	Return value will be built by combining values a and b using the following scheme:
RyoheiHagimoto	0:0e0631af0305	907	If the i-th bit in _mask_ is 1
RyoheiHagimoto	0:0e0631af0305	908	select i-th bit from _a_
RyoheiHagimoto	0:0e0631af0305	909	else
RyoheiHagimoto	0:0e0631af0305	910	select i-th bit from _b_ */
RyoheiHagimoto	0:0e0631af0305	911	template<typename _Tp, int n> inline v_reg<_Tp, n> v_select(const v_reg<_Tp, n>& mask,
RyoheiHagimoto	0:0e0631af0305	912	const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
RyoheiHagimoto	0:0e0631af0305	913	{
RyoheiHagimoto	0:0e0631af0305	914	typedef V_TypeTraits<_Tp> Traits;
RyoheiHagimoto	0:0e0631af0305	915	typedef typename Traits::int_type int_type;
RyoheiHagimoto	0:0e0631af0305	916	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	917	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	918	{
RyoheiHagimoto	0:0e0631af0305	919	int_type m = Traits::reinterpret_int(mask.s[i]);
RyoheiHagimoto	0:0e0631af0305	920	c.s[i] = Traits::reinterpret_from_int((Traits::reinterpret_int(a.s[i]) & m)
RyoheiHagimoto	0:0e0631af0305	921	\| (Traits::reinterpret_int(b.s[i]) & ~m));
RyoheiHagimoto	0:0e0631af0305	922	}
RyoheiHagimoto	0:0e0631af0305	923	return c;
RyoheiHagimoto	0:0e0631af0305	924	}
RyoheiHagimoto	0:0e0631af0305	925
RyoheiHagimoto	0:0e0631af0305	926	/** @brief Expand values to the wider pack type
RyoheiHagimoto	0:0e0631af0305	927
RyoheiHagimoto	0:0e0631af0305	928	Copy contents of register to two registers with 2x wider pack type.
RyoheiHagimoto	0:0e0631af0305	929	Scheme:
RyoheiHagimoto	0:0e0631af0305	930	@code
RyoheiHagimoto	0:0e0631af0305	931	int32x4 int64x2 int64x2
RyoheiHagimoto	0:0e0631af0305	932	{A B C D} ==> {A B} , {C D}
RyoheiHagimoto	0:0e0631af0305	933	@endcode */
RyoheiHagimoto	0:0e0631af0305	934	template<typename _Tp, int n> inline void v_expand(const v_reg<_Tp, n>& a,
RyoheiHagimoto	0:0e0631af0305	935	v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& b0,
RyoheiHagimoto	0:0e0631af0305	936	v_reg<typename V_TypeTraits<_Tp>::w_type, n/2>& b1)
RyoheiHagimoto	0:0e0631af0305	937	{
RyoheiHagimoto	0:0e0631af0305	938	for( int i = 0; i < (n/2); i++ )
RyoheiHagimoto	0:0e0631af0305	939	{
RyoheiHagimoto	0:0e0631af0305	940	b0.s[i] = a.s[i];
RyoheiHagimoto	0:0e0631af0305	941	b1.s[i] = a.s[i+(n/2)];
RyoheiHagimoto	0:0e0631af0305	942	}
RyoheiHagimoto	0:0e0631af0305	943	}
RyoheiHagimoto	0:0e0631af0305	944
RyoheiHagimoto	0:0e0631af0305	945	//! @cond IGNORED
RyoheiHagimoto	0:0e0631af0305	946	template<typename _Tp, int n> inline v_reg<typename V_TypeTraits<_Tp>::int_type, n>
RyoheiHagimoto	0:0e0631af0305	947	v_reinterpret_as_int(const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	948	{
RyoheiHagimoto	0:0e0631af0305	949	v_reg<typename V_TypeTraits<_Tp>::int_type, n> c;
RyoheiHagimoto	0:0e0631af0305	950	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	951	c.s[i] = V_TypeTraits<_Tp>::reinterpret_int(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	952	return c;
RyoheiHagimoto	0:0e0631af0305	953	}
RyoheiHagimoto	0:0e0631af0305	954
RyoheiHagimoto	0:0e0631af0305	955	template<typename _Tp, int n> inline v_reg<typename V_TypeTraits<_Tp>::uint_type, n>
RyoheiHagimoto	0:0e0631af0305	956	v_reinterpret_as_uint(const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	957	{
RyoheiHagimoto	0:0e0631af0305	958	v_reg<typename V_TypeTraits<_Tp>::uint_type, n> c;
RyoheiHagimoto	0:0e0631af0305	959	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	960	c.s[i] = V_TypeTraits<_Tp>::reinterpret_uint(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	961	return c;
RyoheiHagimoto	0:0e0631af0305	962	}
RyoheiHagimoto	0:0e0631af0305	963	//! @endcond
RyoheiHagimoto	0:0e0631af0305	964
RyoheiHagimoto	0:0e0631af0305	965	/** @brief Interleave two vectors
RyoheiHagimoto	0:0e0631af0305	966
RyoheiHagimoto	0:0e0631af0305	967	Scheme:
RyoheiHagimoto	0:0e0631af0305	968	@code
RyoheiHagimoto	0:0e0631af0305	969	{A1 A2 A3 A4}
RyoheiHagimoto	0:0e0631af0305	970	{B1 B2 B3 B4}
RyoheiHagimoto	0:0e0631af0305	971	---------------
RyoheiHagimoto	0:0e0631af0305	972	{A1 B1 A2 B2} and {A3 B3 A4 B4}
RyoheiHagimoto	0:0e0631af0305	973	@endcode
RyoheiHagimoto	0:0e0631af0305	974	For all types except 64-bit.
RyoheiHagimoto	0:0e0631af0305	975	*/
RyoheiHagimoto	0:0e0631af0305	976	template<typename _Tp, int n> inline void v_zip( const v_reg<_Tp, n>& a0, const v_reg<_Tp, n>& a1,
RyoheiHagimoto	0:0e0631af0305	977	v_reg<_Tp, n>& b0, v_reg<_Tp, n>& b1 )
RyoheiHagimoto	0:0e0631af0305	978	{
RyoheiHagimoto	0:0e0631af0305	979	int i;
RyoheiHagimoto	0:0e0631af0305	980	for( i = 0; i < n/2; i++ )
RyoheiHagimoto	0:0e0631af0305	981	{
RyoheiHagimoto	0:0e0631af0305	982	b0.s[i*2] = a0.s[i];
RyoheiHagimoto	0:0e0631af0305	983	b0.s[i*2+1] = a1.s[i];
RyoheiHagimoto	0:0e0631af0305	984	}
RyoheiHagimoto	0:0e0631af0305	985	for( ; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	986	{
RyoheiHagimoto	0:0e0631af0305	987	b1.s[i*2-n] = a0.s[i];
RyoheiHagimoto	0:0e0631af0305	988	b1.s[i*2-n+1] = a1.s[i];
RyoheiHagimoto	0:0e0631af0305	989	}
RyoheiHagimoto	0:0e0631af0305	990	}
RyoheiHagimoto	0:0e0631af0305	991
RyoheiHagimoto	0:0e0631af0305	992	/** @brief Load register contents from memory
RyoheiHagimoto	0:0e0631af0305	993
RyoheiHagimoto	0:0e0631af0305	994	@param ptr pointer to memory block with data
RyoheiHagimoto	0:0e0631af0305	995	@return register object
RyoheiHagimoto	0:0e0631af0305	996
RyoheiHagimoto	0:0e0631af0305	997	@note Returned type will be detected from passed pointer type, for example uchar ==> cv::v_uint8x16, int ==> cv::v_int32x4, etc.
RyoheiHagimoto	0:0e0631af0305	998	*/
RyoheiHagimoto	0:0e0631af0305	999	template<typename _Tp>
RyoheiHagimoto	0:0e0631af0305	1000	inline v_reg<_Tp, V_SIMD128Traits<_Tp>::nlanes> v_load(const _Tp* ptr)
RyoheiHagimoto	0:0e0631af0305	1001	{
RyoheiHagimoto	0:0e0631af0305	1002	return v_reg<_Tp, V_SIMD128Traits<_Tp>::nlanes>(ptr);
RyoheiHagimoto	0:0e0631af0305	1003	}
RyoheiHagimoto	0:0e0631af0305	1004
RyoheiHagimoto	0:0e0631af0305	1005	/** @brief Load register contents from memory (aligned)
RyoheiHagimoto	0:0e0631af0305	1006
RyoheiHagimoto	0:0e0631af0305	1007	similar to cv::v_load, but source memory block should be aligned (to 16-byte boundary)
RyoheiHagimoto	0:0e0631af0305	1008	*/
RyoheiHagimoto	0:0e0631af0305	1009	template<typename _Tp>
RyoheiHagimoto	0:0e0631af0305	1010	inline v_reg<_Tp, V_SIMD128Traits<_Tp>::nlanes> v_load_aligned(const _Tp* ptr)
RyoheiHagimoto	0:0e0631af0305	1011	{
RyoheiHagimoto	0:0e0631af0305	1012	return v_reg<_Tp, V_SIMD128Traits<_Tp>::nlanes>(ptr);
RyoheiHagimoto	0:0e0631af0305	1013	}
RyoheiHagimoto	0:0e0631af0305	1014
RyoheiHagimoto	0:0e0631af0305	1015	/** @brief Load register contents from two memory blocks
RyoheiHagimoto	0:0e0631af0305	1016
RyoheiHagimoto	0:0e0631af0305	1017	@param loptr memory block containing data for first half (0..n/2)
RyoheiHagimoto	0:0e0631af0305	1018	@param hiptr memory block containing data for second half (n/2..n)
RyoheiHagimoto	0:0e0631af0305	1019
RyoheiHagimoto	0:0e0631af0305	1020	@code{.cpp}
RyoheiHagimoto	0:0e0631af0305	1021	int lo[2] = { 1, 2 }, hi[2] = { 3, 4 };
RyoheiHagimoto	0:0e0631af0305	1022	v_int32x4 r = v_load_halves(lo, hi);
RyoheiHagimoto	0:0e0631af0305	1023	@endcode
RyoheiHagimoto	0:0e0631af0305	1024	*/
RyoheiHagimoto	0:0e0631af0305	1025	template<typename _Tp>
RyoheiHagimoto	0:0e0631af0305	1026	inline v_reg<_Tp, V_SIMD128Traits<_Tp>::nlanes> v_load_halves(const _Tp* loptr, const _Tp* hiptr)
RyoheiHagimoto	0:0e0631af0305	1027	{
RyoheiHagimoto	0:0e0631af0305	1028	v_reg<_Tp, V_SIMD128Traits<_Tp>::nlanes> c;
RyoheiHagimoto	0:0e0631af0305	1029	for( int i = 0; i < c.nlanes/2; i++ )
RyoheiHagimoto	0:0e0631af0305	1030	{
RyoheiHagimoto	0:0e0631af0305	1031	c.s[i] = loptr[i];
RyoheiHagimoto	0:0e0631af0305	1032	c.s[i+c.nlanes/2] = hiptr[i];
RyoheiHagimoto	0:0e0631af0305	1033	}
RyoheiHagimoto	0:0e0631af0305	1034	return c;
RyoheiHagimoto	0:0e0631af0305	1035	}
RyoheiHagimoto	0:0e0631af0305	1036
RyoheiHagimoto	0:0e0631af0305	1037	/** @brief Load register contents from memory with double expand
RyoheiHagimoto	0:0e0631af0305	1038
RyoheiHagimoto	0:0e0631af0305	1039	Same as cv::v_load, but result pack type will be 2x wider than memory type.
RyoheiHagimoto	0:0e0631af0305	1040
RyoheiHagimoto	0:0e0631af0305	1041	@code{.cpp}
RyoheiHagimoto	0:0e0631af0305	1042	short buf[4] = {1, 2, 3, 4}; // type is int16
RyoheiHagimoto	0:0e0631af0305	1043	v_int32x4 r = v_load_expand(buf); // r = {1, 2, 3, 4} - type is int32
RyoheiHagimoto	0:0e0631af0305	1044	@endcode
RyoheiHagimoto	0:0e0631af0305	1045	For 8-, 16-, 32-bit integer source types. */
RyoheiHagimoto	0:0e0631af0305	1046	template<typename _Tp>
RyoheiHagimoto	0:0e0631af0305	1047	inline v_reg<typename V_TypeTraits<_Tp>::w_type, V_SIMD128Traits<_Tp>::nlanes / 2>
RyoheiHagimoto	0:0e0631af0305	1048	v_load_expand(const _Tp* ptr)
RyoheiHagimoto	0:0e0631af0305	1049	{
RyoheiHagimoto	0:0e0631af0305	1050	typedef typename V_TypeTraits<_Tp>::w_type w_type;
RyoheiHagimoto	0:0e0631af0305	1051	v_reg<w_type, V_SIMD128Traits<w_type>::nlanes> c;
RyoheiHagimoto	0:0e0631af0305	1052	for( int i = 0; i < c.nlanes; i++ )
RyoheiHagimoto	0:0e0631af0305	1053	{
RyoheiHagimoto	0:0e0631af0305	1054	c.s[i] = ptr[i];
RyoheiHagimoto	0:0e0631af0305	1055	}
RyoheiHagimoto	0:0e0631af0305	1056	return c;
RyoheiHagimoto	0:0e0631af0305	1057	}
RyoheiHagimoto	0:0e0631af0305	1058
RyoheiHagimoto	0:0e0631af0305	1059	/** @brief Load register contents from memory with quad expand
RyoheiHagimoto	0:0e0631af0305	1060
RyoheiHagimoto	0:0e0631af0305	1061	Same as cv::v_load_expand, but result type is 4 times wider than source.
RyoheiHagimoto	0:0e0631af0305	1062	@code{.cpp}
RyoheiHagimoto	0:0e0631af0305	1063	char buf[4] = {1, 2, 3, 4}; // type is int8
RyoheiHagimoto	0:0e0631af0305	1064	v_int32x4 r = v_load_q(buf); // r = {1, 2, 3, 4} - type is int32
RyoheiHagimoto	0:0e0631af0305	1065	@endcode
RyoheiHagimoto	0:0e0631af0305	1066	For 8-bit integer source types. */
RyoheiHagimoto	0:0e0631af0305	1067	template<typename _Tp>
RyoheiHagimoto	0:0e0631af0305	1068	inline v_reg<typename V_TypeTraits<_Tp>::q_type, V_SIMD128Traits<_Tp>::nlanes / 4>
RyoheiHagimoto	0:0e0631af0305	1069	v_load_expand_q(const _Tp* ptr)
RyoheiHagimoto	0:0e0631af0305	1070	{
RyoheiHagimoto	0:0e0631af0305	1071	typedef typename V_TypeTraits<_Tp>::q_type q_type;
RyoheiHagimoto	0:0e0631af0305	1072	v_reg<q_type, V_SIMD128Traits<q_type>::nlanes> c;
RyoheiHagimoto	0:0e0631af0305	1073	for( int i = 0; i < c.nlanes; i++ )
RyoheiHagimoto	0:0e0631af0305	1074	{
RyoheiHagimoto	0:0e0631af0305	1075	c.s[i] = ptr[i];
RyoheiHagimoto	0:0e0631af0305	1076	}
RyoheiHagimoto	0:0e0631af0305	1077	return c;
RyoheiHagimoto	0:0e0631af0305	1078	}
RyoheiHagimoto	0:0e0631af0305	1079
RyoheiHagimoto	0:0e0631af0305	1080	/** @brief Load and deinterleave (2 channels)
RyoheiHagimoto	0:0e0631af0305	1081
RyoheiHagimoto	0:0e0631af0305	1082	Load data from memory deinterleave and store to 2 registers.
RyoheiHagimoto	0:0e0631af0305	1083	Scheme:
RyoheiHagimoto	0:0e0631af0305	1084	@code
RyoheiHagimoto	0:0e0631af0305	1085	{A1 B1 A2 B2 ...} ==> {A1 A2 ...}, {B1 B2 ...}
RyoheiHagimoto	0:0e0631af0305	1086	@endcode
RyoheiHagimoto	0:0e0631af0305	1087	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	1088	template<typename _Tp, int n> inline void v_load_deinterleave(const _Tp* ptr, v_reg<_Tp, n>& a,
RyoheiHagimoto	0:0e0631af0305	1089	v_reg<_Tp, n>& b)
RyoheiHagimoto	0:0e0631af0305	1090	{
RyoheiHagimoto	0:0e0631af0305	1091	int i, i2;
RyoheiHagimoto	0:0e0631af0305	1092	for( i = i2 = 0; i < n; i++, i2 += 2 )
RyoheiHagimoto	0:0e0631af0305	1093	{
RyoheiHagimoto	0:0e0631af0305	1094	a.s[i] = ptr[i2];
RyoheiHagimoto	0:0e0631af0305	1095	b.s[i] = ptr[i2+1];
RyoheiHagimoto	0:0e0631af0305	1096	}
RyoheiHagimoto	0:0e0631af0305	1097	}
RyoheiHagimoto	0:0e0631af0305	1098
RyoheiHagimoto	0:0e0631af0305	1099	/** @brief Load and deinterleave (3 channels)
RyoheiHagimoto	0:0e0631af0305	1100
RyoheiHagimoto	0:0e0631af0305	1101	Load data from memory deinterleave and store to 3 registers.
RyoheiHagimoto	0:0e0631af0305	1102	Scheme:
RyoheiHagimoto	0:0e0631af0305	1103	@code
RyoheiHagimoto	0:0e0631af0305	1104	{A1 B1 C1 A2 B2 C2 ...} ==> {A1 A2 ...}, {B1 B2 ...}, {C1 C2 ...}
RyoheiHagimoto	0:0e0631af0305	1105	@endcode
RyoheiHagimoto	0:0e0631af0305	1106	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	1107	template<typename _Tp, int n> inline void v_load_deinterleave(const _Tp* ptr, v_reg<_Tp, n>& a,
RyoheiHagimoto	0:0e0631af0305	1108	v_reg<_Tp, n>& b, v_reg<_Tp, n>& c)
RyoheiHagimoto	0:0e0631af0305	1109	{
RyoheiHagimoto	0:0e0631af0305	1110	int i, i3;
RyoheiHagimoto	0:0e0631af0305	1111	for( i = i3 = 0; i < n; i++, i3 += 3 )
RyoheiHagimoto	0:0e0631af0305	1112	{
RyoheiHagimoto	0:0e0631af0305	1113	a.s[i] = ptr[i3];
RyoheiHagimoto	0:0e0631af0305	1114	b.s[i] = ptr[i3+1];
RyoheiHagimoto	0:0e0631af0305	1115	c.s[i] = ptr[i3+2];
RyoheiHagimoto	0:0e0631af0305	1116	}
RyoheiHagimoto	0:0e0631af0305	1117	}
RyoheiHagimoto	0:0e0631af0305	1118
RyoheiHagimoto	0:0e0631af0305	1119	/** @brief Load and deinterleave (4 channels)
RyoheiHagimoto	0:0e0631af0305	1120
RyoheiHagimoto	0:0e0631af0305	1121	Load data from memory deinterleave and store to 4 registers.
RyoheiHagimoto	0:0e0631af0305	1122	Scheme:
RyoheiHagimoto	0:0e0631af0305	1123	@code
RyoheiHagimoto	0:0e0631af0305	1124	{A1 B1 C1 D1 A2 B2 C2 D2 ...} ==> {A1 A2 ...}, {B1 B2 ...}, {C1 C2 ...}, {D1 D2 ...}
RyoheiHagimoto	0:0e0631af0305	1125	@endcode
RyoheiHagimoto	0:0e0631af0305	1126	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	1127	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1128	inline void v_load_deinterleave(const _Tp* ptr, v_reg<_Tp, n>& a,
RyoheiHagimoto	0:0e0631af0305	1129	v_reg<_Tp, n>& b, v_reg<_Tp, n>& c,
RyoheiHagimoto	0:0e0631af0305	1130	v_reg<_Tp, n>& d)
RyoheiHagimoto	0:0e0631af0305	1131	{
RyoheiHagimoto	0:0e0631af0305	1132	int i, i4;
RyoheiHagimoto	0:0e0631af0305	1133	for( i = i4 = 0; i < n; i++, i4 += 4 )
RyoheiHagimoto	0:0e0631af0305	1134	{
RyoheiHagimoto	0:0e0631af0305	1135	a.s[i] = ptr[i4];
RyoheiHagimoto	0:0e0631af0305	1136	b.s[i] = ptr[i4+1];
RyoheiHagimoto	0:0e0631af0305	1137	c.s[i] = ptr[i4+2];
RyoheiHagimoto	0:0e0631af0305	1138	d.s[i] = ptr[i4+3];
RyoheiHagimoto	0:0e0631af0305	1139	}
RyoheiHagimoto	0:0e0631af0305	1140	}
RyoheiHagimoto	0:0e0631af0305	1141
RyoheiHagimoto	0:0e0631af0305	1142	/** @brief Interleave and store (2 channels)
RyoheiHagimoto	0:0e0631af0305	1143
RyoheiHagimoto	0:0e0631af0305	1144	Interleave and store data from 2 registers to memory.
RyoheiHagimoto	0:0e0631af0305	1145	Scheme:
RyoheiHagimoto	0:0e0631af0305	1146	@code
RyoheiHagimoto	0:0e0631af0305	1147	{A1 A2 ...}, {B1 B2 ...} ==> {A1 B1 A2 B2 ...}
RyoheiHagimoto	0:0e0631af0305	1148	@endcode
RyoheiHagimoto	0:0e0631af0305	1149	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	1150	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1151	inline void v_store_interleave( _Tp* ptr, const v_reg<_Tp, n>& a,
RyoheiHagimoto	0:0e0631af0305	1152	const v_reg<_Tp, n>& b)
RyoheiHagimoto	0:0e0631af0305	1153	{
RyoheiHagimoto	0:0e0631af0305	1154	int i, i2;
RyoheiHagimoto	0:0e0631af0305	1155	for( i = i2 = 0; i < n; i++, i2 += 2 )
RyoheiHagimoto	0:0e0631af0305	1156	{
RyoheiHagimoto	0:0e0631af0305	1157	ptr[i2] = a.s[i];
RyoheiHagimoto	0:0e0631af0305	1158	ptr[i2+1] = b.s[i];
RyoheiHagimoto	0:0e0631af0305	1159	}
RyoheiHagimoto	0:0e0631af0305	1160	}
RyoheiHagimoto	0:0e0631af0305	1161
RyoheiHagimoto	0:0e0631af0305	1162	/** @brief Interleave and store (3 channels)
RyoheiHagimoto	0:0e0631af0305	1163
RyoheiHagimoto	0:0e0631af0305	1164	Interleave and store data from 3 registers to memory.
RyoheiHagimoto	0:0e0631af0305	1165	Scheme:
RyoheiHagimoto	0:0e0631af0305	1166	@code
RyoheiHagimoto	0:0e0631af0305	1167	{A1 A2 ...}, {B1 B2 ...}, {C1 C2 ...} ==> {A1 B1 C1 A2 B2 C2 ...}
RyoheiHagimoto	0:0e0631af0305	1168	@endcode
RyoheiHagimoto	0:0e0631af0305	1169	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	1170	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1171	inline void v_store_interleave( _Tp* ptr, const v_reg<_Tp, n>& a,
RyoheiHagimoto	0:0e0631af0305	1172	const v_reg<_Tp, n>& b, const v_reg<_Tp, n>& c)
RyoheiHagimoto	0:0e0631af0305	1173	{
RyoheiHagimoto	0:0e0631af0305	1174	int i, i3;
RyoheiHagimoto	0:0e0631af0305	1175	for( i = i3 = 0; i < n; i++, i3 += 3 )
RyoheiHagimoto	0:0e0631af0305	1176	{
RyoheiHagimoto	0:0e0631af0305	1177	ptr[i3] = a.s[i];
RyoheiHagimoto	0:0e0631af0305	1178	ptr[i3+1] = b.s[i];
RyoheiHagimoto	0:0e0631af0305	1179	ptr[i3+2] = c.s[i];
RyoheiHagimoto	0:0e0631af0305	1180	}
RyoheiHagimoto	0:0e0631af0305	1181	}
RyoheiHagimoto	0:0e0631af0305	1182
RyoheiHagimoto	0:0e0631af0305	1183	/** @brief Interleave and store (4 channels)
RyoheiHagimoto	0:0e0631af0305	1184
RyoheiHagimoto	0:0e0631af0305	1185	Interleave and store data from 4 registers to memory.
RyoheiHagimoto	0:0e0631af0305	1186	Scheme:
RyoheiHagimoto	0:0e0631af0305	1187	@code
RyoheiHagimoto	0:0e0631af0305	1188	{A1 A2 ...}, {B1 B2 ...}, {C1 C2 ...}, {D1 D2 ...} ==> {A1 B1 C1 D1 A2 B2 C2 D2 ...}
RyoheiHagimoto	0:0e0631af0305	1189	@endcode
RyoheiHagimoto	0:0e0631af0305	1190	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	1191	template<typename _Tp, int n> inline void v_store_interleave( _Tp* ptr, const v_reg<_Tp, n>& a,
RyoheiHagimoto	0:0e0631af0305	1192	const v_reg<_Tp, n>& b, const v_reg<_Tp, n>& c,
RyoheiHagimoto	0:0e0631af0305	1193	const v_reg<_Tp, n>& d)
RyoheiHagimoto	0:0e0631af0305	1194	{
RyoheiHagimoto	0:0e0631af0305	1195	int i, i4;
RyoheiHagimoto	0:0e0631af0305	1196	for( i = i4 = 0; i < n; i++, i4 += 4 )
RyoheiHagimoto	0:0e0631af0305	1197	{
RyoheiHagimoto	0:0e0631af0305	1198	ptr[i4] = a.s[i];
RyoheiHagimoto	0:0e0631af0305	1199	ptr[i4+1] = b.s[i];
RyoheiHagimoto	0:0e0631af0305	1200	ptr[i4+2] = c.s[i];
RyoheiHagimoto	0:0e0631af0305	1201	ptr[i4+3] = d.s[i];
RyoheiHagimoto	0:0e0631af0305	1202	}
RyoheiHagimoto	0:0e0631af0305	1203	}
RyoheiHagimoto	0:0e0631af0305	1204
RyoheiHagimoto	0:0e0631af0305	1205	/** @brief Store data to memory
RyoheiHagimoto	0:0e0631af0305	1206
RyoheiHagimoto	0:0e0631af0305	1207	Store register contents to memory.
RyoheiHagimoto	0:0e0631af0305	1208	Scheme:
RyoheiHagimoto	0:0e0631af0305	1209	@code
RyoheiHagimoto	0:0e0631af0305	1210	REG {A B C D} ==> MEM {A B C D}
RyoheiHagimoto	0:0e0631af0305	1211	@endcode
RyoheiHagimoto	0:0e0631af0305	1212	Pointer can be unaligned. */
RyoheiHagimoto	0:0e0631af0305	1213	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1214	inline void v_store(_Tp* ptr, const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	1215	{
RyoheiHagimoto	0:0e0631af0305	1216	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1217	ptr[i] = a.s[i];
RyoheiHagimoto	0:0e0631af0305	1218	}
RyoheiHagimoto	0:0e0631af0305	1219
RyoheiHagimoto	0:0e0631af0305	1220	/** @brief Store data to memory (lower half)
RyoheiHagimoto	0:0e0631af0305	1221
RyoheiHagimoto	0:0e0631af0305	1222	Store lower half of register contents to memory.
RyoheiHagimoto	0:0e0631af0305	1223	Scheme:
RyoheiHagimoto	0:0e0631af0305	1224	@code
RyoheiHagimoto	0:0e0631af0305	1225	REG {A B C D} ==> MEM {A B}
RyoheiHagimoto	0:0e0631af0305	1226	@endcode */
RyoheiHagimoto	0:0e0631af0305	1227	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1228	inline void v_store_low(_Tp* ptr, const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	1229	{
RyoheiHagimoto	0:0e0631af0305	1230	for( int i = 0; i < (n/2); i++ )
RyoheiHagimoto	0:0e0631af0305	1231	ptr[i] = a.s[i];
RyoheiHagimoto	0:0e0631af0305	1232	}
RyoheiHagimoto	0:0e0631af0305	1233
RyoheiHagimoto	0:0e0631af0305	1234	/** @brief Store data to memory (higher half)
RyoheiHagimoto	0:0e0631af0305	1235
RyoheiHagimoto	0:0e0631af0305	1236	Store higher half of register contents to memory.
RyoheiHagimoto	0:0e0631af0305	1237	Scheme:
RyoheiHagimoto	0:0e0631af0305	1238	@code
RyoheiHagimoto	0:0e0631af0305	1239	REG {A B C D} ==> MEM {C D}
RyoheiHagimoto	0:0e0631af0305	1240	@endcode */
RyoheiHagimoto	0:0e0631af0305	1241	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1242	inline void v_store_high(_Tp* ptr, const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	1243	{
RyoheiHagimoto	0:0e0631af0305	1244	for( int i = 0; i < (n/2); i++ )
RyoheiHagimoto	0:0e0631af0305	1245	ptr[i] = a.s[i+(n/2)];
RyoheiHagimoto	0:0e0631af0305	1246	}
RyoheiHagimoto	0:0e0631af0305	1247
RyoheiHagimoto	0:0e0631af0305	1248	/** @brief Store data to memory (aligned)
RyoheiHagimoto	0:0e0631af0305	1249
RyoheiHagimoto	0:0e0631af0305	1250	Store register contents to memory.
RyoheiHagimoto	0:0e0631af0305	1251	Scheme:
RyoheiHagimoto	0:0e0631af0305	1252	@code
RyoheiHagimoto	0:0e0631af0305	1253	REG {A B C D} ==> MEM {A B C D}
RyoheiHagimoto	0:0e0631af0305	1254	@endcode
RyoheiHagimoto	0:0e0631af0305	1255	Pointer __should__ be aligned by 16-byte boundary. */
RyoheiHagimoto	0:0e0631af0305	1256	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1257	inline void v_store_aligned(_Tp* ptr, const v_reg<_Tp, n>& a)
RyoheiHagimoto	0:0e0631af0305	1258	{
RyoheiHagimoto	0:0e0631af0305	1259	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1260	ptr[i] = a.s[i];
RyoheiHagimoto	0:0e0631af0305	1261	}
RyoheiHagimoto	0:0e0631af0305	1262
RyoheiHagimoto	0:0e0631af0305	1263	/** @brief Combine vector from first elements of two vectors
RyoheiHagimoto	0:0e0631af0305	1264
RyoheiHagimoto	0:0e0631af0305	1265	Scheme:
RyoheiHagimoto	0:0e0631af0305	1266	@code
RyoheiHagimoto	0:0e0631af0305	1267	{A1 A2 A3 A4}
RyoheiHagimoto	0:0e0631af0305	1268	{B1 B2 B3 B4}
RyoheiHagimoto	0:0e0631af0305	1269	---------------
RyoheiHagimoto	0:0e0631af0305	1270	{A1 A2 B1 B2}
RyoheiHagimoto	0:0e0631af0305	1271	@endcode
RyoheiHagimoto	0:0e0631af0305	1272	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	1273	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1274	inline v_reg<_Tp, n> v_combine_low(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
RyoheiHagimoto	0:0e0631af0305	1275	{
RyoheiHagimoto	0:0e0631af0305	1276	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	1277	for( int i = 0; i < (n/2); i++ )
RyoheiHagimoto	0:0e0631af0305	1278	{
RyoheiHagimoto	0:0e0631af0305	1279	c.s[i] = a.s[i];
RyoheiHagimoto	0:0e0631af0305	1280	c.s[i+(n/2)] = b.s[i];
RyoheiHagimoto	0:0e0631af0305	1281	}
RyoheiHagimoto	0:0e0631af0305	1282	return c;
RyoheiHagimoto	0:0e0631af0305	1283	}
RyoheiHagimoto	0:0e0631af0305	1284
RyoheiHagimoto	0:0e0631af0305	1285	/** @brief Combine vector from last elements of two vectors
RyoheiHagimoto	0:0e0631af0305	1286
RyoheiHagimoto	0:0e0631af0305	1287	Scheme:
RyoheiHagimoto	0:0e0631af0305	1288	@code
RyoheiHagimoto	0:0e0631af0305	1289	{A1 A2 A3 A4}
RyoheiHagimoto	0:0e0631af0305	1290	{B1 B2 B3 B4}
RyoheiHagimoto	0:0e0631af0305	1291	---------------
RyoheiHagimoto	0:0e0631af0305	1292	{A3 A4 B3 B4}
RyoheiHagimoto	0:0e0631af0305	1293	@endcode
RyoheiHagimoto	0:0e0631af0305	1294	For all types except 64-bit. */
RyoheiHagimoto	0:0e0631af0305	1295	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1296	inline v_reg<_Tp, n> v_combine_high(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
RyoheiHagimoto	0:0e0631af0305	1297	{
RyoheiHagimoto	0:0e0631af0305	1298	v_reg<_Tp, n> c;
RyoheiHagimoto	0:0e0631af0305	1299	for( int i = 0; i < (n/2); i++ )
RyoheiHagimoto	0:0e0631af0305	1300	{
RyoheiHagimoto	0:0e0631af0305	1301	c.s[i] = a.s[i+(n/2)];
RyoheiHagimoto	0:0e0631af0305	1302	c.s[i+(n/2)] = b.s[i+(n/2)];
RyoheiHagimoto	0:0e0631af0305	1303	}
RyoheiHagimoto	0:0e0631af0305	1304	return c;
RyoheiHagimoto	0:0e0631af0305	1305	}
RyoheiHagimoto	0:0e0631af0305	1306
RyoheiHagimoto	0:0e0631af0305	1307	/** @brief Combine two vectors from lower and higher parts of two other vectors
RyoheiHagimoto	0:0e0631af0305	1308
RyoheiHagimoto	0:0e0631af0305	1309	@code{.cpp}
RyoheiHagimoto	0:0e0631af0305	1310	low = cv::v_combine_low(a, b);
RyoheiHagimoto	0:0e0631af0305	1311	high = cv::v_combine_high(a, b);
RyoheiHagimoto	0:0e0631af0305	1312	@endcode */
RyoheiHagimoto	0:0e0631af0305	1313	template<typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1314	inline void v_recombine(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b,
RyoheiHagimoto	0:0e0631af0305	1315	v_reg<_Tp, n>& low, v_reg<_Tp, n>& high)
RyoheiHagimoto	0:0e0631af0305	1316	{
RyoheiHagimoto	0:0e0631af0305	1317	for( int i = 0; i < (n/2); i++ )
RyoheiHagimoto	0:0e0631af0305	1318	{
RyoheiHagimoto	0:0e0631af0305	1319	low.s[i] = a.s[i];
RyoheiHagimoto	0:0e0631af0305	1320	low.s[i+(n/2)] = b.s[i];
RyoheiHagimoto	0:0e0631af0305	1321	high.s[i] = a.s[i+(n/2)];
RyoheiHagimoto	0:0e0631af0305	1322	high.s[i+(n/2)] = b.s[i+(n/2)];
RyoheiHagimoto	0:0e0631af0305	1323	}
RyoheiHagimoto	0:0e0631af0305	1324	}
RyoheiHagimoto	0:0e0631af0305	1325
RyoheiHagimoto	0:0e0631af0305	1326	/** @brief Vector extract
RyoheiHagimoto	0:0e0631af0305	1327
RyoheiHagimoto	0:0e0631af0305	1328	Scheme:
RyoheiHagimoto	0:0e0631af0305	1329	@code
RyoheiHagimoto	0:0e0631af0305	1330	{A1 A2 A3 A4}
RyoheiHagimoto	0:0e0631af0305	1331	{B1 B2 B3 B4}
RyoheiHagimoto	0:0e0631af0305	1332	========================
RyoheiHagimoto	0:0e0631af0305	1333	shift = 1 {A2 A3 A4 B1}
RyoheiHagimoto	0:0e0631af0305	1334	shift = 2 {A3 A4 B1 B2}
RyoheiHagimoto	0:0e0631af0305	1335	shift = 3 {A4 B1 B2 B3}
RyoheiHagimoto	0:0e0631af0305	1336	@endcode
RyoheiHagimoto	0:0e0631af0305	1337	Restriction: 0 <= shift < nlanes
RyoheiHagimoto	0:0e0631af0305	1338
RyoheiHagimoto	0:0e0631af0305	1339	Usage:
RyoheiHagimoto	0:0e0631af0305	1340	@code
RyoheiHagimoto	0:0e0631af0305	1341	v_int32x4 a, b, c;
RyoheiHagimoto	0:0e0631af0305	1342	c = v_extract<2>(a, b);
RyoheiHagimoto	0:0e0631af0305	1343	@endcode
RyoheiHagimoto	0:0e0631af0305	1344	For integer types only. */
RyoheiHagimoto	0:0e0631af0305	1345	template<int s, typename _Tp, int n>
RyoheiHagimoto	0:0e0631af0305	1346	inline v_reg<_Tp, n> v_extract(const v_reg<_Tp, n>& a, const v_reg<_Tp, n>& b)
RyoheiHagimoto	0:0e0631af0305	1347	{
RyoheiHagimoto	0:0e0631af0305	1348	v_reg<_Tp, n> r;
RyoheiHagimoto	0:0e0631af0305	1349	const int shift = n - s;
RyoheiHagimoto	0:0e0631af0305	1350	int i = 0;
RyoheiHagimoto	0:0e0631af0305	1351	for (; i < shift; ++i)
RyoheiHagimoto	0:0e0631af0305	1352	r.s[i] = a.s[i+s];
RyoheiHagimoto	0:0e0631af0305	1353	for (; i < n; ++i)
RyoheiHagimoto	0:0e0631af0305	1354	r.s[i] = b.s[i-shift];
RyoheiHagimoto	0:0e0631af0305	1355	return r;
RyoheiHagimoto	0:0e0631af0305	1356	}
RyoheiHagimoto	0:0e0631af0305	1357
RyoheiHagimoto	0:0e0631af0305	1358	/** @brief Round
RyoheiHagimoto	0:0e0631af0305	1359
RyoheiHagimoto	0:0e0631af0305	1360	Rounds each value. Input type is float vector ==> output type is int vector.*/
RyoheiHagimoto	0:0e0631af0305	1361	template<int n> inline v_reg<int, n> v_round(const v_reg<float, n>& a)
RyoheiHagimoto	0:0e0631af0305	1362	{
RyoheiHagimoto	0:0e0631af0305	1363	v_reg<int, n> c;
RyoheiHagimoto	0:0e0631af0305	1364	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1365	c.s[i] = cvRound(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	1366	return c;
RyoheiHagimoto	0:0e0631af0305	1367	}
RyoheiHagimoto	0:0e0631af0305	1368
RyoheiHagimoto	0:0e0631af0305	1369	/** @brief Floor
RyoheiHagimoto	0:0e0631af0305	1370
RyoheiHagimoto	0:0e0631af0305	1371	Floor each value. Input type is float vector ==> output type is int vector.*/
RyoheiHagimoto	0:0e0631af0305	1372	template<int n> inline v_reg<int, n> v_floor(const v_reg<float, n>& a)
RyoheiHagimoto	0:0e0631af0305	1373	{
RyoheiHagimoto	0:0e0631af0305	1374	v_reg<int, n> c;
RyoheiHagimoto	0:0e0631af0305	1375	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1376	c.s[i] = cvFloor(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	1377	return c;
RyoheiHagimoto	0:0e0631af0305	1378	}
RyoheiHagimoto	0:0e0631af0305	1379
RyoheiHagimoto	0:0e0631af0305	1380	/** @brief Ceil
RyoheiHagimoto	0:0e0631af0305	1381
RyoheiHagimoto	0:0e0631af0305	1382	Ceil each value. Input type is float vector ==> output type is int vector.*/
RyoheiHagimoto	0:0e0631af0305	1383	template<int n> inline v_reg<int, n> v_ceil(const v_reg<float, n>& a)
RyoheiHagimoto	0:0e0631af0305	1384	{
RyoheiHagimoto	0:0e0631af0305	1385	v_reg<int, n> c;
RyoheiHagimoto	0:0e0631af0305	1386	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1387	c.s[i] = cvCeil(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	1388	return c;
RyoheiHagimoto	0:0e0631af0305	1389	}
RyoheiHagimoto	0:0e0631af0305	1390
RyoheiHagimoto	0:0e0631af0305	1391	/** @brief Trunc
RyoheiHagimoto	0:0e0631af0305	1392
RyoheiHagimoto	0:0e0631af0305	1393	Truncate each value. Input type is float vector ==> output type is int vector.*/
RyoheiHagimoto	0:0e0631af0305	1394	template<int n> inline v_reg<int, n> v_trunc(const v_reg<float, n>& a)
RyoheiHagimoto	0:0e0631af0305	1395	{
RyoheiHagimoto	0:0e0631af0305	1396	v_reg<int, n> c;
RyoheiHagimoto	0:0e0631af0305	1397	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1398	c.s[i] = (int)(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	1399	return c;
RyoheiHagimoto	0:0e0631af0305	1400	}
RyoheiHagimoto	0:0e0631af0305	1401
RyoheiHagimoto	0:0e0631af0305	1402	/** @overload */
RyoheiHagimoto	0:0e0631af0305	1403	template<int n> inline v_reg<int, n*2> v_round(const v_reg<double, n>& a)
RyoheiHagimoto	0:0e0631af0305	1404	{
RyoheiHagimoto	0:0e0631af0305	1405	v_reg<int, n*2> c;
RyoheiHagimoto	0:0e0631af0305	1406	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1407	{
RyoheiHagimoto	0:0e0631af0305	1408	c.s[i] = cvRound(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	1409	c.s[i+n] = 0;
RyoheiHagimoto	0:0e0631af0305	1410	}
RyoheiHagimoto	0:0e0631af0305	1411	return c;
RyoheiHagimoto	0:0e0631af0305	1412	}
RyoheiHagimoto	0:0e0631af0305	1413
RyoheiHagimoto	0:0e0631af0305	1414	/** @overload */
RyoheiHagimoto	0:0e0631af0305	1415	template<int n> inline v_reg<int, n*2> v_floor(const v_reg<double, n>& a)
RyoheiHagimoto	0:0e0631af0305	1416	{
RyoheiHagimoto	0:0e0631af0305	1417	v_reg<int, n> c;
RyoheiHagimoto	0:0e0631af0305	1418	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1419	{
RyoheiHagimoto	0:0e0631af0305	1420	c.s[i] = cvFloor(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	1421	c.s[i+n] = 0;
RyoheiHagimoto	0:0e0631af0305	1422	}
RyoheiHagimoto	0:0e0631af0305	1423	return c;
RyoheiHagimoto	0:0e0631af0305	1424	}
RyoheiHagimoto	0:0e0631af0305	1425
RyoheiHagimoto	0:0e0631af0305	1426	/** @overload */
RyoheiHagimoto	0:0e0631af0305	1427	template<int n> inline v_reg<int, n*2> v_ceil(const v_reg<double, n>& a)
RyoheiHagimoto	0:0e0631af0305	1428	{
RyoheiHagimoto	0:0e0631af0305	1429	v_reg<int, n> c;
RyoheiHagimoto	0:0e0631af0305	1430	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1431	{
RyoheiHagimoto	0:0e0631af0305	1432	c.s[i] = cvCeil(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	1433	c.s[i+n] = 0;
RyoheiHagimoto	0:0e0631af0305	1434	}
RyoheiHagimoto	0:0e0631af0305	1435	return c;
RyoheiHagimoto	0:0e0631af0305	1436	}
RyoheiHagimoto	0:0e0631af0305	1437
RyoheiHagimoto	0:0e0631af0305	1438	/** @overload */
RyoheiHagimoto	0:0e0631af0305	1439	template<int n> inline v_reg<int, n*2> v_trunc(const v_reg<double, n>& a)
RyoheiHagimoto	0:0e0631af0305	1440	{
RyoheiHagimoto	0:0e0631af0305	1441	v_reg<int, n> c;
RyoheiHagimoto	0:0e0631af0305	1442	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1443	{
RyoheiHagimoto	0:0e0631af0305	1444	c.s[i] = cvCeil(a.s[i]);
RyoheiHagimoto	0:0e0631af0305	1445	c.s[i+n] = 0;
RyoheiHagimoto	0:0e0631af0305	1446	}
RyoheiHagimoto	0:0e0631af0305	1447	return c;
RyoheiHagimoto	0:0e0631af0305	1448	}
RyoheiHagimoto	0:0e0631af0305	1449
RyoheiHagimoto	0:0e0631af0305	1450	/** @brief Convert to float
RyoheiHagimoto	0:0e0631af0305	1451
RyoheiHagimoto	0:0e0631af0305	1452	Supported input type is cv::v_int32x4. */
RyoheiHagimoto	0:0e0631af0305	1453	template<int n> inline v_reg<float, n> v_cvt_f32(const v_reg<int, n>& a)
RyoheiHagimoto	0:0e0631af0305	1454	{
RyoheiHagimoto	0:0e0631af0305	1455	v_reg<float, n> c;
RyoheiHagimoto	0:0e0631af0305	1456	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1457	c.s[i] = (float)a.s[i];
RyoheiHagimoto	0:0e0631af0305	1458	return c;
RyoheiHagimoto	0:0e0631af0305	1459	}
RyoheiHagimoto	0:0e0631af0305	1460
RyoheiHagimoto	0:0e0631af0305	1461	/** @brief Convert to double
RyoheiHagimoto	0:0e0631af0305	1462
RyoheiHagimoto	0:0e0631af0305	1463	Supported input type is cv::v_int32x4. */
RyoheiHagimoto	0:0e0631af0305	1464	template<int n> inline v_reg<double, n> v_cvt_f64(const v_reg<int, n*2>& a)
RyoheiHagimoto	0:0e0631af0305	1465	{
RyoheiHagimoto	0:0e0631af0305	1466	v_reg<double, n> c;
RyoheiHagimoto	0:0e0631af0305	1467	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1468	c.s[i] = (double)a.s[i];
RyoheiHagimoto	0:0e0631af0305	1469	return c;
RyoheiHagimoto	0:0e0631af0305	1470	}
RyoheiHagimoto	0:0e0631af0305	1471
RyoheiHagimoto	0:0e0631af0305	1472	/** @brief Convert to double
RyoheiHagimoto	0:0e0631af0305	1473
RyoheiHagimoto	0:0e0631af0305	1474	Supported input type is cv::v_float32x4. */
RyoheiHagimoto	0:0e0631af0305	1475	template<int n> inline v_reg<double, n> v_cvt_f64(const v_reg<float, n*2>& a)
RyoheiHagimoto	0:0e0631af0305	1476	{
RyoheiHagimoto	0:0e0631af0305	1477	v_reg<double, n> c;
RyoheiHagimoto	0:0e0631af0305	1478	for( int i = 0; i < n; i++ )
RyoheiHagimoto	0:0e0631af0305	1479	c.s[i] = (double)a.s[i];
RyoheiHagimoto	0:0e0631af0305	1480	return c;
RyoheiHagimoto	0:0e0631af0305	1481	}
RyoheiHagimoto	0:0e0631af0305	1482
RyoheiHagimoto	0:0e0631af0305	1483	/** @brief Transpose 4x4 matrix
RyoheiHagimoto	0:0e0631af0305	1484
RyoheiHagimoto	0:0e0631af0305	1485	Scheme:
RyoheiHagimoto	0:0e0631af0305	1486	@code
RyoheiHagimoto	0:0e0631af0305	1487	a0 {A1 A2 A3 A4}
RyoheiHagimoto	0:0e0631af0305	1488	a1 {B1 B2 B3 B4}
RyoheiHagimoto	0:0e0631af0305	1489	a2 {C1 C2 C3 C4}
RyoheiHagimoto	0:0e0631af0305	1490	a3 {D1 D2 D3 D4}
RyoheiHagimoto	0:0e0631af0305	1491	===============
RyoheiHagimoto	0:0e0631af0305	1492	b0 {A1 B1 C1 D1}
RyoheiHagimoto	0:0e0631af0305	1493	b1 {A2 B2 C2 D2}
RyoheiHagimoto	0:0e0631af0305	1494	b2 {A3 B3 C3 D3}
RyoheiHagimoto	0:0e0631af0305	1495	b3 {A4 B4 C4 D4}
RyoheiHagimoto	0:0e0631af0305	1496	@endcode
RyoheiHagimoto	0:0e0631af0305	1497	*/
RyoheiHagimoto	0:0e0631af0305	1498	template<typename _Tp>
RyoheiHagimoto	0:0e0631af0305	1499	inline void v_transpose4x4( v_reg<_Tp, 4>& a0, const v_reg<_Tp, 4>& a1,
RyoheiHagimoto	0:0e0631af0305	1500	const v_reg<_Tp, 4>& a2, const v_reg<_Tp, 4>& a3,
RyoheiHagimoto	0:0e0631af0305	1501	v_reg<_Tp, 4>& b0, v_reg<_Tp, 4>& b1,
RyoheiHagimoto	0:0e0631af0305	1502	v_reg<_Tp, 4>& b2, v_reg<_Tp, 4>& b3 )
RyoheiHagimoto	0:0e0631af0305	1503	{
RyoheiHagimoto	0:0e0631af0305	1504	b0 = v_reg<_Tp, 4>(a0.s[0], a1.s[0], a2.s[0], a3.s[0]);
RyoheiHagimoto	0:0e0631af0305	1505	b1 = v_reg<_Tp, 4>(a0.s[1], a1.s[1], a2.s[1], a3.s[1]);
RyoheiHagimoto	0:0e0631af0305	1506	b2 = v_reg<_Tp, 4>(a0.s[2], a1.s[2], a2.s[2], a3.s[2]);
RyoheiHagimoto	0:0e0631af0305	1507	b3 = v_reg<_Tp, 4>(a0.s[3], a1.s[3], a2.s[3], a3.s[3]);
RyoheiHagimoto	0:0e0631af0305	1508	}
RyoheiHagimoto	0:0e0631af0305	1509
RyoheiHagimoto	0:0e0631af0305	1510	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1511	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1512	#define OPENCV_HAL_IMPL_C_INIT_ZERO(_Tpvec, _Tp, suffix) \
RyoheiHagimoto	0:0e0631af0305	1513	inline _Tpvec v_setzero_##suffix() { return _Tpvec::zero(); }
RyoheiHagimoto	0:0e0631af0305	1514
RyoheiHagimoto	0:0e0631af0305	1515	//! @name Init with zero
RyoheiHagimoto	0:0e0631af0305	1516	//! @{
RyoheiHagimoto	0:0e0631af0305	1517	//! @brief Create new vector with zero elements
RyoheiHagimoto	0:0e0631af0305	1518	OPENCV_HAL_IMPL_C_INIT_ZERO(v_uint8x16, uchar, u8)
RyoheiHagimoto	0:0e0631af0305	1519	OPENCV_HAL_IMPL_C_INIT_ZERO(v_int8x16, schar, s8)
RyoheiHagimoto	0:0e0631af0305	1520	OPENCV_HAL_IMPL_C_INIT_ZERO(v_uint16x8, ushort, u16)
RyoheiHagimoto	0:0e0631af0305	1521	OPENCV_HAL_IMPL_C_INIT_ZERO(v_int16x8, short, s16)
RyoheiHagimoto	0:0e0631af0305	1522	OPENCV_HAL_IMPL_C_INIT_ZERO(v_uint32x4, unsigned, u32)
RyoheiHagimoto	0:0e0631af0305	1523	OPENCV_HAL_IMPL_C_INIT_ZERO(v_int32x4, int, s32)
RyoheiHagimoto	0:0e0631af0305	1524	OPENCV_HAL_IMPL_C_INIT_ZERO(v_float32x4, float, f32)
RyoheiHagimoto	0:0e0631af0305	1525	OPENCV_HAL_IMPL_C_INIT_ZERO(v_float64x2, double, f64)
RyoheiHagimoto	0:0e0631af0305	1526	OPENCV_HAL_IMPL_C_INIT_ZERO(v_uint64x2, uint64, u64)
RyoheiHagimoto	0:0e0631af0305	1527	OPENCV_HAL_IMPL_C_INIT_ZERO(v_int64x2, int64, s64)
RyoheiHagimoto	0:0e0631af0305	1528	//! @}
RyoheiHagimoto	0:0e0631af0305	1529
RyoheiHagimoto	0:0e0631af0305	1530	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1531	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1532	#define OPENCV_HAL_IMPL_C_INIT_VAL(_Tpvec, _Tp, suffix) \
RyoheiHagimoto	0:0e0631af0305	1533	inline _Tpvec v_setall_##suffix(_Tp val) { return _Tpvec::all(val); }
RyoheiHagimoto	0:0e0631af0305	1534
RyoheiHagimoto	0:0e0631af0305	1535	//! @name Init with value
RyoheiHagimoto	0:0e0631af0305	1536	//! @{
RyoheiHagimoto	0:0e0631af0305	1537	//! @brief Create new vector with elements set to a specific value
RyoheiHagimoto	0:0e0631af0305	1538	OPENCV_HAL_IMPL_C_INIT_VAL(v_uint8x16, uchar, u8)
RyoheiHagimoto	0:0e0631af0305	1539	OPENCV_HAL_IMPL_C_INIT_VAL(v_int8x16, schar, s8)
RyoheiHagimoto	0:0e0631af0305	1540	OPENCV_HAL_IMPL_C_INIT_VAL(v_uint16x8, ushort, u16)
RyoheiHagimoto	0:0e0631af0305	1541	OPENCV_HAL_IMPL_C_INIT_VAL(v_int16x8, short, s16)
RyoheiHagimoto	0:0e0631af0305	1542	OPENCV_HAL_IMPL_C_INIT_VAL(v_uint32x4, unsigned, u32)
RyoheiHagimoto	0:0e0631af0305	1543	OPENCV_HAL_IMPL_C_INIT_VAL(v_int32x4, int, s32)
RyoheiHagimoto	0:0e0631af0305	1544	OPENCV_HAL_IMPL_C_INIT_VAL(v_float32x4, float, f32)
RyoheiHagimoto	0:0e0631af0305	1545	OPENCV_HAL_IMPL_C_INIT_VAL(v_float64x2, double, f64)
RyoheiHagimoto	0:0e0631af0305	1546	OPENCV_HAL_IMPL_C_INIT_VAL(v_uint64x2, uint64, u64)
RyoheiHagimoto	0:0e0631af0305	1547	OPENCV_HAL_IMPL_C_INIT_VAL(v_int64x2, int64, s64)
RyoheiHagimoto	0:0e0631af0305	1548	//! @}
RyoheiHagimoto	0:0e0631af0305	1549
RyoheiHagimoto	0:0e0631af0305	1550	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1551	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1552	#define OPENCV_HAL_IMPL_C_REINTERPRET(_Tpvec, _Tp, suffix) \
RyoheiHagimoto	0:0e0631af0305	1553	template<typename _Tp0, int n0> inline _Tpvec \
RyoheiHagimoto	0:0e0631af0305	1554	v_reinterpret_as_##suffix(const v_reg<_Tp0, n0>& a) \
RyoheiHagimoto	0:0e0631af0305	1555	{ return a.template reinterpret_as<_Tp, _Tpvec::nlanes>(); }
RyoheiHagimoto	0:0e0631af0305	1556
RyoheiHagimoto	0:0e0631af0305	1557	//! @name Reinterpret
RyoheiHagimoto	0:0e0631af0305	1558	//! @{
RyoheiHagimoto	0:0e0631af0305	1559	//! @brief Convert vector to different type without modifying underlying data.
RyoheiHagimoto	0:0e0631af0305	1560	OPENCV_HAL_IMPL_C_REINTERPRET(v_uint8x16, uchar, u8)
RyoheiHagimoto	0:0e0631af0305	1561	OPENCV_HAL_IMPL_C_REINTERPRET(v_int8x16, schar, s8)
RyoheiHagimoto	0:0e0631af0305	1562	OPENCV_HAL_IMPL_C_REINTERPRET(v_uint16x8, ushort, u16)
RyoheiHagimoto	0:0e0631af0305	1563	OPENCV_HAL_IMPL_C_REINTERPRET(v_int16x8, short, s16)
RyoheiHagimoto	0:0e0631af0305	1564	OPENCV_HAL_IMPL_C_REINTERPRET(v_uint32x4, unsigned, u32)
RyoheiHagimoto	0:0e0631af0305	1565	OPENCV_HAL_IMPL_C_REINTERPRET(v_int32x4, int, s32)
RyoheiHagimoto	0:0e0631af0305	1566	OPENCV_HAL_IMPL_C_REINTERPRET(v_float32x4, float, f32)
RyoheiHagimoto	0:0e0631af0305	1567	OPENCV_HAL_IMPL_C_REINTERPRET(v_float64x2, double, f64)
RyoheiHagimoto	0:0e0631af0305	1568	OPENCV_HAL_IMPL_C_REINTERPRET(v_uint64x2, uint64, u64)
RyoheiHagimoto	0:0e0631af0305	1569	OPENCV_HAL_IMPL_C_REINTERPRET(v_int64x2, int64, s64)
RyoheiHagimoto	0:0e0631af0305	1570	//! @}
RyoheiHagimoto	0:0e0631af0305	1571
RyoheiHagimoto	0:0e0631af0305	1572	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1573	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1574	#define OPENCV_HAL_IMPL_C_SHIFTL(_Tpvec, _Tp) \
RyoheiHagimoto	0:0e0631af0305	1575	template<int n> inline _Tpvec v_shl(const _Tpvec& a) \
RyoheiHagimoto	0:0e0631af0305	1576	{ return a << n; }
RyoheiHagimoto	0:0e0631af0305	1577
RyoheiHagimoto	0:0e0631af0305	1578	//! @name Left shift
RyoheiHagimoto	0:0e0631af0305	1579	//! @{
RyoheiHagimoto	0:0e0631af0305	1580	//! @brief Shift left
RyoheiHagimoto	0:0e0631af0305	1581	OPENCV_HAL_IMPL_C_SHIFTL(v_uint16x8, ushort)
RyoheiHagimoto	0:0e0631af0305	1582	OPENCV_HAL_IMPL_C_SHIFTL(v_int16x8, short)
RyoheiHagimoto	0:0e0631af0305	1583	OPENCV_HAL_IMPL_C_SHIFTL(v_uint32x4, unsigned)
RyoheiHagimoto	0:0e0631af0305	1584	OPENCV_HAL_IMPL_C_SHIFTL(v_int32x4, int)
RyoheiHagimoto	0:0e0631af0305	1585	OPENCV_HAL_IMPL_C_SHIFTL(v_uint64x2, uint64)
RyoheiHagimoto	0:0e0631af0305	1586	OPENCV_HAL_IMPL_C_SHIFTL(v_int64x2, int64)
RyoheiHagimoto	0:0e0631af0305	1587	//! @}
RyoheiHagimoto	0:0e0631af0305	1588
RyoheiHagimoto	0:0e0631af0305	1589	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1590	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1591	#define OPENCV_HAL_IMPL_C_SHIFTR(_Tpvec, _Tp) \
RyoheiHagimoto	0:0e0631af0305	1592	template<int n> inline _Tpvec v_shr(const _Tpvec& a) \
RyoheiHagimoto	0:0e0631af0305	1593	{ return a >> n; }
RyoheiHagimoto	0:0e0631af0305	1594
RyoheiHagimoto	0:0e0631af0305	1595	//! @name Right shift
RyoheiHagimoto	0:0e0631af0305	1596	//! @{
RyoheiHagimoto	0:0e0631af0305	1597	//! @brief Shift right
RyoheiHagimoto	0:0e0631af0305	1598	OPENCV_HAL_IMPL_C_SHIFTR(v_uint16x8, ushort)
RyoheiHagimoto	0:0e0631af0305	1599	OPENCV_HAL_IMPL_C_SHIFTR(v_int16x8, short)
RyoheiHagimoto	0:0e0631af0305	1600	OPENCV_HAL_IMPL_C_SHIFTR(v_uint32x4, unsigned)
RyoheiHagimoto	0:0e0631af0305	1601	OPENCV_HAL_IMPL_C_SHIFTR(v_int32x4, int)
RyoheiHagimoto	0:0e0631af0305	1602	OPENCV_HAL_IMPL_C_SHIFTR(v_uint64x2, uint64)
RyoheiHagimoto	0:0e0631af0305	1603	OPENCV_HAL_IMPL_C_SHIFTR(v_int64x2, int64)
RyoheiHagimoto	0:0e0631af0305	1604	//! @}
RyoheiHagimoto	0:0e0631af0305	1605
RyoheiHagimoto	0:0e0631af0305	1606	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1607	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1608	#define OPENCV_HAL_IMPL_C_RSHIFTR(_Tpvec, _Tp) \
RyoheiHagimoto	0:0e0631af0305	1609	template<int n> inline _Tpvec v_rshr(const _Tpvec& a) \
RyoheiHagimoto	0:0e0631af0305	1610	{ \
RyoheiHagimoto	0:0e0631af0305	1611	_Tpvec c; \
RyoheiHagimoto	0:0e0631af0305	1612	for( int i = 0; i < _Tpvec::nlanes; i++ ) \
RyoheiHagimoto	0:0e0631af0305	1613	c.s[i] = (_Tp)((a.s[i] + ((_Tp)1 << (n - 1))) >> n); \
RyoheiHagimoto	0:0e0631af0305	1614	return c; \
RyoheiHagimoto	0:0e0631af0305	1615	}
RyoheiHagimoto	0:0e0631af0305	1616
RyoheiHagimoto	0:0e0631af0305	1617	//! @name Rounding shift
RyoheiHagimoto	0:0e0631af0305	1618	//! @{
RyoheiHagimoto	0:0e0631af0305	1619	//! @brief Rounding shift right
RyoheiHagimoto	0:0e0631af0305	1620	OPENCV_HAL_IMPL_C_RSHIFTR(v_uint16x8, ushort)
RyoheiHagimoto	0:0e0631af0305	1621	OPENCV_HAL_IMPL_C_RSHIFTR(v_int16x8, short)
RyoheiHagimoto	0:0e0631af0305	1622	OPENCV_HAL_IMPL_C_RSHIFTR(v_uint32x4, unsigned)
RyoheiHagimoto	0:0e0631af0305	1623	OPENCV_HAL_IMPL_C_RSHIFTR(v_int32x4, int)
RyoheiHagimoto	0:0e0631af0305	1624	OPENCV_HAL_IMPL_C_RSHIFTR(v_uint64x2, uint64)
RyoheiHagimoto	0:0e0631af0305	1625	OPENCV_HAL_IMPL_C_RSHIFTR(v_int64x2, int64)
RyoheiHagimoto	0:0e0631af0305	1626	//! @}
RyoheiHagimoto	0:0e0631af0305	1627
RyoheiHagimoto	0:0e0631af0305	1628	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1629	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1630	#define OPENCV_HAL_IMPL_C_PACK(_Tpvec, _Tpnvec, _Tpn, pack_suffix) \
RyoheiHagimoto	0:0e0631af0305	1631	inline _Tpnvec v_##pack_suffix(const _Tpvec& a, const _Tpvec& b) \
RyoheiHagimoto	0:0e0631af0305	1632	{ \
RyoheiHagimoto	0:0e0631af0305	1633	_Tpnvec c; \
RyoheiHagimoto	0:0e0631af0305	1634	for( int i = 0; i < _Tpvec::nlanes; i++ ) \
RyoheiHagimoto	0:0e0631af0305	1635	{ \
RyoheiHagimoto	0:0e0631af0305	1636	c.s[i] = saturate_cast<_Tpn>(a.s[i]); \
RyoheiHagimoto	0:0e0631af0305	1637	c.s[i+_Tpvec::nlanes] = saturate_cast<_Tpn>(b.s[i]); \
RyoheiHagimoto	0:0e0631af0305	1638	} \
RyoheiHagimoto	0:0e0631af0305	1639	return c; \
RyoheiHagimoto	0:0e0631af0305	1640	}
RyoheiHagimoto	0:0e0631af0305	1641
RyoheiHagimoto	0:0e0631af0305	1642	//! @name Pack
RyoheiHagimoto	0:0e0631af0305	1643	//! @{
RyoheiHagimoto	0:0e0631af0305	1644	//! @brief Pack values from two vectors to one
RyoheiHagimoto	0:0e0631af0305	1645	//!
RyoheiHagimoto	0:0e0631af0305	1646	//! Return vector type have twice more elements than input vector types. Variant with _u_ suffix also
RyoheiHagimoto	0:0e0631af0305	1647	//! converts to corresponding unsigned type.
RyoheiHagimoto	0:0e0631af0305	1648	//!
RyoheiHagimoto	0:0e0631af0305	1649	//! - pack: for 16-, 32- and 64-bit integer input types
RyoheiHagimoto	0:0e0631af0305	1650	//! - pack_u: for 16- and 32-bit signed integer input types
RyoheiHagimoto	0:0e0631af0305	1651	OPENCV_HAL_IMPL_C_PACK(v_uint16x8, v_uint8x16, uchar, pack)
RyoheiHagimoto	0:0e0631af0305	1652	OPENCV_HAL_IMPL_C_PACK(v_int16x8, v_int8x16, schar, pack)
RyoheiHagimoto	0:0e0631af0305	1653	OPENCV_HAL_IMPL_C_PACK(v_uint32x4, v_uint16x8, ushort, pack)
RyoheiHagimoto	0:0e0631af0305	1654	OPENCV_HAL_IMPL_C_PACK(v_int32x4, v_int16x8, short, pack)
RyoheiHagimoto	0:0e0631af0305	1655	OPENCV_HAL_IMPL_C_PACK(v_uint64x2, v_uint32x4, unsigned, pack)
RyoheiHagimoto	0:0e0631af0305	1656	OPENCV_HAL_IMPL_C_PACK(v_int64x2, v_int32x4, int, pack)
RyoheiHagimoto	0:0e0631af0305	1657	OPENCV_HAL_IMPL_C_PACK(v_int16x8, v_uint8x16, uchar, pack_u)
RyoheiHagimoto	0:0e0631af0305	1658	OPENCV_HAL_IMPL_C_PACK(v_int32x4, v_uint16x8, ushort, pack_u)
RyoheiHagimoto	0:0e0631af0305	1659	//! @}
RyoheiHagimoto	0:0e0631af0305	1660
RyoheiHagimoto	0:0e0631af0305	1661	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1662	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1663	#define OPENCV_HAL_IMPL_C_RSHR_PACK(_Tpvec, _Tp, _Tpnvec, _Tpn, pack_suffix) \
RyoheiHagimoto	0:0e0631af0305	1664	template<int n> inline _Tpnvec v_rshr_##pack_suffix(const _Tpvec& a, const _Tpvec& b) \
RyoheiHagimoto	0:0e0631af0305	1665	{ \
RyoheiHagimoto	0:0e0631af0305	1666	_Tpnvec c; \
RyoheiHagimoto	0:0e0631af0305	1667	for( int i = 0; i < _Tpvec::nlanes; i++ ) \
RyoheiHagimoto	0:0e0631af0305	1668	{ \
RyoheiHagimoto	0:0e0631af0305	1669	c.s[i] = saturate_cast<_Tpn>((a.s[i] + ((_Tp)1 << (n - 1))) >> n); \
RyoheiHagimoto	0:0e0631af0305	1670	c.s[i+_Tpvec::nlanes] = saturate_cast<_Tpn>((b.s[i] + ((_Tp)1 << (n - 1))) >> n); \
RyoheiHagimoto	0:0e0631af0305	1671	} \
RyoheiHagimoto	0:0e0631af0305	1672	return c; \
RyoheiHagimoto	0:0e0631af0305	1673	}
RyoheiHagimoto	0:0e0631af0305	1674
RyoheiHagimoto	0:0e0631af0305	1675	//! @name Pack with rounding shift
RyoheiHagimoto	0:0e0631af0305	1676	//! @{
RyoheiHagimoto	0:0e0631af0305	1677	//! @brief Pack values from two vectors to one with rounding shift
RyoheiHagimoto	0:0e0631af0305	1678	//!
RyoheiHagimoto	0:0e0631af0305	1679	//! Values from the input vectors will be shifted right by _n_ bits with rounding, converted to narrower
RyoheiHagimoto	0:0e0631af0305	1680	//! type and returned in the result vector. Variant with _u_ suffix converts to unsigned type.
RyoheiHagimoto	0:0e0631af0305	1681	//!
RyoheiHagimoto	0:0e0631af0305	1682	//! - pack: for 16-, 32- and 64-bit integer input types
RyoheiHagimoto	0:0e0631af0305	1683	//! - pack_u: for 16- and 32-bit signed integer input types
RyoheiHagimoto	0:0e0631af0305	1684	OPENCV_HAL_IMPL_C_RSHR_PACK(v_uint16x8, ushort, v_uint8x16, uchar, pack)
RyoheiHagimoto	0:0e0631af0305	1685	OPENCV_HAL_IMPL_C_RSHR_PACK(v_int16x8, short, v_int8x16, schar, pack)
RyoheiHagimoto	0:0e0631af0305	1686	OPENCV_HAL_IMPL_C_RSHR_PACK(v_uint32x4, unsigned, v_uint16x8, ushort, pack)
RyoheiHagimoto	0:0e0631af0305	1687	OPENCV_HAL_IMPL_C_RSHR_PACK(v_int32x4, int, v_int16x8, short, pack)
RyoheiHagimoto	0:0e0631af0305	1688	OPENCV_HAL_IMPL_C_RSHR_PACK(v_uint64x2, uint64, v_uint32x4, unsigned, pack)
RyoheiHagimoto	0:0e0631af0305	1689	OPENCV_HAL_IMPL_C_RSHR_PACK(v_int64x2, int64, v_int32x4, int, pack)
RyoheiHagimoto	0:0e0631af0305	1690	OPENCV_HAL_IMPL_C_RSHR_PACK(v_int16x8, short, v_uint8x16, uchar, pack_u)
RyoheiHagimoto	0:0e0631af0305	1691	OPENCV_HAL_IMPL_C_RSHR_PACK(v_int32x4, int, v_uint16x8, ushort, pack_u)
RyoheiHagimoto	0:0e0631af0305	1692	//! @}
RyoheiHagimoto	0:0e0631af0305	1693
RyoheiHagimoto	0:0e0631af0305	1694	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1695	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1696	#define OPENCV_HAL_IMPL_C_PACK_STORE(_Tpvec, _Tp, _Tpnvec, _Tpn, pack_suffix) \
RyoheiHagimoto	0:0e0631af0305	1697	inline void v_##pack_suffix##_store(_Tpn* ptr, const _Tpvec& a) \
RyoheiHagimoto	0:0e0631af0305	1698	{ \
RyoheiHagimoto	0:0e0631af0305	1699	for( int i = 0; i < _Tpvec::nlanes; i++ ) \
RyoheiHagimoto	0:0e0631af0305	1700	ptr[i] = saturate_cast<_Tpn>(a.s[i]); \
RyoheiHagimoto	0:0e0631af0305	1701	}
RyoheiHagimoto	0:0e0631af0305	1702
RyoheiHagimoto	0:0e0631af0305	1703	//! @name Pack and store
RyoheiHagimoto	0:0e0631af0305	1704	//! @{
RyoheiHagimoto	0:0e0631af0305	1705	//! @brief Store values from the input vector into memory with pack
RyoheiHagimoto	0:0e0631af0305	1706	//!
RyoheiHagimoto	0:0e0631af0305	1707	//! Values will be stored into memory with saturating conversion to narrower type.
RyoheiHagimoto	0:0e0631af0305	1708	//! Variant with _u_ suffix converts to corresponding unsigned type.
RyoheiHagimoto	0:0e0631af0305	1709	//!
RyoheiHagimoto	0:0e0631af0305	1710	//! - pack: for 16-, 32- and 64-bit integer input types
RyoheiHagimoto	0:0e0631af0305	1711	//! - pack_u: for 16- and 32-bit signed integer input types
RyoheiHagimoto	0:0e0631af0305	1712	OPENCV_HAL_IMPL_C_PACK_STORE(v_uint16x8, ushort, v_uint8x16, uchar, pack)
RyoheiHagimoto	0:0e0631af0305	1713	OPENCV_HAL_IMPL_C_PACK_STORE(v_int16x8, short, v_int8x16, schar, pack)
RyoheiHagimoto	0:0e0631af0305	1714	OPENCV_HAL_IMPL_C_PACK_STORE(v_uint32x4, unsigned, v_uint16x8, ushort, pack)
RyoheiHagimoto	0:0e0631af0305	1715	OPENCV_HAL_IMPL_C_PACK_STORE(v_int32x4, int, v_int16x8, short, pack)
RyoheiHagimoto	0:0e0631af0305	1716	OPENCV_HAL_IMPL_C_PACK_STORE(v_uint64x2, uint64, v_uint32x4, unsigned, pack)
RyoheiHagimoto	0:0e0631af0305	1717	OPENCV_HAL_IMPL_C_PACK_STORE(v_int64x2, int64, v_int32x4, int, pack)
RyoheiHagimoto	0:0e0631af0305	1718	OPENCV_HAL_IMPL_C_PACK_STORE(v_int16x8, short, v_uint8x16, uchar, pack_u)
RyoheiHagimoto	0:0e0631af0305	1719	OPENCV_HAL_IMPL_C_PACK_STORE(v_int32x4, int, v_uint16x8, ushort, pack_u)
RyoheiHagimoto	0:0e0631af0305	1720	//! @}
RyoheiHagimoto	0:0e0631af0305	1721
RyoheiHagimoto	0:0e0631af0305	1722	//! @brief Helper macro
RyoheiHagimoto	0:0e0631af0305	1723	//! @ingroup core_hal_intrin_impl
RyoheiHagimoto	0:0e0631af0305	1724	#define OPENCV_HAL_IMPL_C_RSHR_PACK_STORE(_Tpvec, _Tp, _Tpnvec, _Tpn, pack_suffix) \
RyoheiHagimoto	0:0e0631af0305	1725	template<int n> inline void v_rshr_##pack_suffix##_store(_Tpn* ptr, const _Tpvec& a) \
RyoheiHagimoto	0:0e0631af0305	1726	{ \
RyoheiHagimoto	0:0e0631af0305	1727	for( int i = 0; i < _Tpvec::nlanes; i++ ) \
RyoheiHagimoto	0:0e0631af0305	1728	ptr[i] = saturate_cast<_Tpn>((a.s[i] + ((_Tp)1 << (n - 1))) >> n); \
RyoheiHagimoto	0:0e0631af0305	1729	}
RyoheiHagimoto	0:0e0631af0305	1730
RyoheiHagimoto	0:0e0631af0305	1731	//! @name Pack and store with rounding shift
RyoheiHagimoto	0:0e0631af0305	1732	//! @{
RyoheiHagimoto	0:0e0631af0305	1733	//! @brief Store values from the input vector into memory with pack
RyoheiHagimoto	0:0e0631af0305	1734	//!
RyoheiHagimoto	0:0e0631af0305	1735	//! Values will be shifted _n_ bits right with rounding, converted to narrower type and stored into
RyoheiHagimoto	0:0e0631af0305	1736	//! memory. Variant with _u_ suffix converts to unsigned type.
RyoheiHagimoto	0:0e0631af0305	1737	//!
RyoheiHagimoto	0:0e0631af0305	1738	//! - pack: for 16-, 32- and 64-bit integer input types
RyoheiHagimoto	0:0e0631af0305	1739	//! - pack_u: for 16- and 32-bit signed integer input types
RyoheiHagimoto	0:0e0631af0305	1740	OPENCV_HAL_IMPL_C_RSHR_PACK_STORE(v_uint16x8, ushort, v_uint8x16, uchar, pack)
RyoheiHagimoto	0:0e0631af0305	1741	OPENCV_HAL_IMPL_C_RSHR_PACK_STORE(v_int16x8, short, v_int8x16, schar, pack)
RyoheiHagimoto	0:0e0631af0305	1742	OPENCV_HAL_IMPL_C_RSHR_PACK_STORE(v_uint32x4, unsigned, v_uint16x8, ushort, pack)
RyoheiHagimoto	0:0e0631af0305	1743	OPENCV_HAL_IMPL_C_RSHR_PACK_STORE(v_int32x4, int, v_int16x8, short, pack)
RyoheiHagimoto	0:0e0631af0305	1744	OPENCV_HAL_IMPL_C_RSHR_PACK_STORE(v_uint64x2, uint64, v_uint32x4, unsigned, pack)
RyoheiHagimoto	0:0e0631af0305	1745	OPENCV_HAL_IMPL_C_RSHR_PACK_STORE(v_int64x2, int64, v_int32x4, int, pack)
RyoheiHagimoto	0:0e0631af0305	1746	OPENCV_HAL_IMPL_C_RSHR_PACK_STORE(v_int16x8, short, v_uint8x16, uchar, pack_u)
RyoheiHagimoto	0:0e0631af0305	1747	OPENCV_HAL_IMPL_C_RSHR_PACK_STORE(v_int32x4, int, v_uint16x8, ushort, pack_u)
RyoheiHagimoto	0:0e0631af0305	1748	//! @}
RyoheiHagimoto	0:0e0631af0305	1749
RyoheiHagimoto	0:0e0631af0305	1750	/** @brief Matrix multiplication
RyoheiHagimoto	0:0e0631af0305	1751
RyoheiHagimoto	0:0e0631af0305	1752	Scheme:
RyoheiHagimoto	0:0e0631af0305	1753	@code
RyoheiHagimoto	0:0e0631af0305	1754	{A0 A1 A2 A3} \|V0\|
RyoheiHagimoto	0:0e0631af0305	1755	{B0 B1 B2 B3} \|V1\|
RyoheiHagimoto	0:0e0631af0305	1756	{C0 C1 C2 C3} \|V2\|
RyoheiHagimoto	0:0e0631af0305	1757	{D0 D1 D2 D3} x \|V3\|
RyoheiHagimoto	0:0e0631af0305	1758	====================
RyoheiHagimoto	0:0e0631af0305	1759	{R0 R1 R2 R3}, where:
RyoheiHagimoto	0:0e0631af0305	1760	R0 = A0V0 + A1V1 + A2V2 + A3V3,
RyoheiHagimoto	0:0e0631af0305	1761	R1 = B0V0 + B1V1 + B2V2 + B3V3
RyoheiHagimoto	0:0e0631af0305	1762	...
RyoheiHagimoto	0:0e0631af0305	1763	@endcode
RyoheiHagimoto	0:0e0631af0305	1764	*/
RyoheiHagimoto	0:0e0631af0305	1765	inline v_float32x4 v_matmul(const v_float32x4& v, const v_float32x4& m0,
RyoheiHagimoto	0:0e0631af0305	1766	const v_float32x4& m1, const v_float32x4& m2,
RyoheiHagimoto	0:0e0631af0305	1767	const v_float32x4& m3)
RyoheiHagimoto	0:0e0631af0305	1768	{
RyoheiHagimoto	0:0e0631af0305	1769	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],
RyoheiHagimoto	0:0e0631af0305	1770	v.s[0]m0.s[1] + v.s[1]m1.s[1] + v.s[2]m2.s[1] + v.s[3]m3.s[1],
RyoheiHagimoto	0:0e0631af0305	1771	v.s[0]m0.s[2] + v.s[1]m1.s[2] + v.s[2]m2.s[2] + v.s[3]m3.s[2],
RyoheiHagimoto	0:0e0631af0305	1772	v.s[0]m0.s[3] + v.s[1]m1.s[3] + v.s[2]m2.s[3] + v.s[3]m3.s[3]);
RyoheiHagimoto	0:0e0631af0305	1773	}
RyoheiHagimoto	0:0e0631af0305	1774
RyoheiHagimoto	0:0e0631af0305	1775	//! @}
RyoheiHagimoto	0:0e0631af0305	1776
RyoheiHagimoto	0:0e0631af0305	1777	//! @name Check SIMD support
RyoheiHagimoto	0:0e0631af0305	1778	//! @{
RyoheiHagimoto	0:0e0631af0305	1779	//! @brief Check CPU capability of SIMD operation
RyoheiHagimoto	0:0e0631af0305	1780	static inline bool hasSIMD128()
RyoheiHagimoto	0:0e0631af0305	1781	{
RyoheiHagimoto	0:0e0631af0305	1782	return false;
RyoheiHagimoto	0:0e0631af0305	1783	}
RyoheiHagimoto	0:0e0631af0305	1784
RyoheiHagimoto	0:0e0631af0305	1785	//! @}
RyoheiHagimoto	0:0e0631af0305	1786
RyoheiHagimoto	0:0e0631af0305	1787
RyoheiHagimoto	0:0e0631af0305	1788	}
RyoheiHagimoto	0:0e0631af0305	1789
RyoheiHagimoto	0:0e0631af0305	1790	#endif

Repository toolbox

Export to desktop IDE

Repository details

Type:	Library
Created:	29 Jan 2021
Imports:	11
Forks:	0
Commits:	1
Dependents:	1
Dependencies:	0
Followers:	60

include/opencv2/core/hal/intrin_cpp.hpp@0:0e0631af0305, 2021-01-29 (annotated)

Who changed what in which revision?

Repository toolbox

Repository details

Important Information for this Arm website

Access Warning