sse::svec< 4, float > Struct Template Reference

data representation and operations on a vector of 4 float. More...

#include <sse4.h>

Public Member Functions
	svec ()
	Default constructor. More...
	svec (__m128 vv)
	For internal use only. More...
	svec (float a, float b, float c, float d)
	Constructor. More...
	svec (float a)
	Constructor. More...
float &	operator[] (int index)
	operator [] to set or get the vector element specified by index. More...
const float	operator[] (int index) const
svec< 4,bool >	operator== (svec< 4, float > a)
	Element-wise compare equal, return a bool vector, e.g., "a == b" *. More...
svec< 4,bool >	operator!= (svec< 4, float > a)
	Element-wise compare not equal, return a bool vector. E.g. "a != b" *. More...
svec< 4,bool >	operator< (svec< 4, float > a)
	Element-wise compare less than, return a bool vector. E.g. "a < b" *. More...
svec< 4,bool >	operator<= (svec< 4, float > a)
	Element-wise compare less equal, return a bool vector. E.g. "a <= b" *. More...
svec< 4,bool >	operator> (svec< 4, float > a)
	Element-wise compare greater than, return a bool vector. E.g. "a > b" *. More...
svec< 4,bool >	operator>= (svec< 4, float > a)
	Element-wise compare greater equal, return a bool vector. E.g. "a >= b" *. More...
svec< 4, float >	operator- ()
	get the neg value, return a VTYPE vector. E.g. "-a" More...
float	reduce_add ()
	Get the sum of all the element values in the vector. return a float scalar. More...
float	reduce_max ()
	Get the max value of all the element values in the vector. return a float scalar. More...
float	reduce_min ()
	Get the min value of all the element values in the vector. return a float scalar. More...
svec< 4, float >	operator+ (svec< 4, float > a)
	Add two vectors. More...
svec< 4, float >	operator+ (float s)
	Add a vector and a scalar. More...
svec< 4, float >	operator- (svec< 4, float > a)
	Sub two vectors. More...
svec< 4, float >	operator- (float s)
	Sub a vector and a scalar. More...
svec< 4, float >	operator* (svec< 4, float > a)
	Multiply two vectors. More...
svec< 4, float >	operator* (float s)
	Multiply a vector and a scalar. More...
svec< 4, float >	operator/ (svec< 4, float > a)
	Divide a vector by a vector. More...
svec< 4, float >	operator/ (float s)
	Divide a vector by a scalar. More...
void	store (svec< 4, float > *p)
	Store the vector to address p. More...
void	masked_store (svec< 4, float > *p, svec< 4, bool > mask)
	Store the vector element's value to pointer p if the mask element is true. More...
void	scatter (svec< 4, void * > ptrs, svec< 4, bool > mask)
	Scatter the vector's elements to the locations pointed by the vector ptrs if the mask element is true. More...
void	scatter_base_offsets (float *b, uint32_t scale, svec< 4, int32_t > offsets, svec< 4, bool > mask)
	Scatter the vector's elements to the addresses ((char*)b + scale * offsets) if the mask element is true. More...
void	scatter_base_offsets (float *b, uint32_t scale, svec< 4, int64_t > offsets, svec< 4, bool > mask)
	Scatter the vector's elements to the addresses ((char*)b + scale * offsets) if the mask element is true. More...
void	scatter_stride (float *b, int32_t off, int32_t stride)
	Scatter the vector's elements to the addresses (b+off, b++off+stride, b+off+2*stride, b+off+3*step). More...
void	scatter_stride (float *b, int64_t off, int64_t stride)
	Scatter the vector's elements to the addresses (b+off, b++off+stride, b+off+2*stride, b+off+3*step). More...
svec< 4, float >	broadcast (int32_t index)
	Return a new vector by setting all the elements of the new vector with this vector's index element. More...
svec< 4, float >	rotate (int32_t index)
	Return a new vector by rotate this vector's elements. e.g. newVec[i] = thisVec[i+index]. More...
svec< 4, float >	shuffle (svec< 4, int32_t > index)
	Return a new vector by shuffle this vector's elements with index vector e.g. newVec[i] = thisVec[index[i]]. More...
svec< 4, float >	abs ()
	Return a new vector of the vector's abs value. More...
svec< 4, float >	round ()
	return the round VTYPE vector. More...
svec< 4, float >	floor ()
	return the floor VTYPE vector. More...
svec< 4, float >	ceil ()
	return the ceil VTYPE vector. More...
svec< 4, float >	sqrt ()
	return the sqrt VTYPE vector. More...
svec< 4, float >	rcp ()
	return the reverse VTYPE vector. (1.0/thisVec) More...
svec< 4, float >	rsqrt ()
	return the reverse sqrt VTYPE vector. ( 1.0/sqrt(thisVec) ) More...
svec< 4, float >	exp ()
	return the exp VTYPE vector. More...
svec< 4, float >	log ()
	return the log VTYPE vector. More...
svec< 4, float >	pow (svec< 4, float > a)
	return the pow VTYPE vector. More...

Static Public Member Functions
static svec< 4, float >	load (svec< 4, float > *p)
	load the vector from the pointer p More...
static svec< 4, float >	masked_load (svec< 4, float > *p, svec< 4, bool > mask)
	Return a new vector by only loading the value from the pointer p if the mask element is true. More...
static svec< 4, float >	load_const (const float *p)
	Construct a vector by loading a scalar value from pointer p, and splat it to all the elements in the vector. More...
static svec< 4, float >	load_and_splat (float *p)
	Construct a vector by loading a scalar value from pointer p, and splat it to all the elements in the vector. More...
static svec< 4, float >	gather (svec< 4, void * > ptrs, svec< 4, bool > mask)
	Gather the elements pointed by the vector ptrs if the mask element is true, and return a vector. More...
static svec< 4, float >	gather_base_offsets (float *b, uint32_t scale, svec< 4, int32_t > offsets, svec< 4, bool > mask)
	Gather the elements pointed by calculating the addresses ((char*)b + scale * offsets) if the mask element is true, and return a vector. More...
static svec< 4, float >	gather_base_offsets (float *b, uint32_t scale, svec< 4, int64_t > offsets, svec< 4, bool > mask)
	Gather the elements pointed by calculating the addresses ((char*)b + scale * offsets) if the mask element is true, and return a vector. More...
static svec< 4, float >	gather_stride (float *b, int32_t off, int32_t stride)
	Gather the elements pointed by (b+off, b++off+stride, b+off+2*stride, b+off+3*step). More...
static svec< 4, float >	gather_stride (float *b, int64_t off, int64_t stride)
	Gather the elements pointed by (b+off, b++off+stride, b+off+2*stride, b+off+3*step). More...

Public Attributes
__m128	v

Friends
std::ostream &	operator<< (std::ostream &out, const svec< 4, float > &v)

Detailed Description

template<>
struct sse::svec< 4, float >

data representation and operations on a vector of 4 float.

Constructor & Destructor Documentation

sse::svec< 4, float >::svec

(

)

Default constructor.

Returns

a vector of 4 undefined float.

sse::svec< 4, float >::svec

(

__m128

vv

)

For internal use only.

Parameters

vv

a __m128.

Returns

a float vector, whose value is from the vv.

sse::svec< 4, float >::svec	(	float	a,
		float	b,
		float	c,
		float	d
	)

Constructor.

Returns

a vector of 4 float: {a,b,c,d}.

sse::svec< 4, float >::svec

(

float

a

)

Constructor.

Parameters

a

a float

Returns

a vector of 4 floats: {a,a,a,a}.

Member Function Documentation

svec< 4, float > sse::svec< 4, float >::abs

(

)

Return a new vector of the vector's abs value.

svec< 4, float > sse::svec< 4, float >::broadcast

(

int32_t

index

)

Return a new vector by setting all the elements of the new vector with this vector's index element.

svec< 4, float > sse::svec< 4, float >::ceil

(

)

return the ceil VTYPE vector.

svec< 4, float > sse::svec< 4, float >::exp

(

)

return the exp VTYPE vector.

svec< 4, float > sse::svec< 4, float >::floor

(

)

return the floor VTYPE vector.

svec< 4, float > sse::svec< 4, float >::gather ( svec< 4, void * >  ptrs, svec< 4, bool >  mask  )

static

Gather the elements pointed by the vector ptrs if the mask element is true, and return a vector.

svec< 4, float > sse::svec< 4, float >::gather_base_offsets ( float *  b, uint32_t  scale, svec< 4, int32_t >  offsets, svec< 4, bool >  mask  )

static

Gather the elements pointed by calculating the addresses ((char*)b + scale * offsets) if the mask element is true, and return a vector.

svec< 4, float > sse::svec< 4, float >::gather_base_offsets ( float *  b, uint32_t  scale, svec< 4, int64_t >  offsets, svec< 4, bool >  mask  )

static

Gather the elements pointed by calculating the addresses ((char*)b + scale * offsets) if the mask element is true, and return a vector.

svec< 4, float > sse::svec< 4, float >::gather_stride ( float *  b, int32_t  off, int32_t  stride  )

static

Gather the elements pointed by (b+off, b++off+stride, b+off+2*stride, b+off+3*step).

svec< 4, float > sse::svec< 4, float >::gather_stride ( float *  b, int64_t  off, int64_t  stride  )

static

Gather the elements pointed by (b+off, b++off+stride, b+off+2*stride, b+off+3*step).

svec< 4, float > sse::svec< 4, float >::load ( svec< 4, float > *  p )

static

load the vector from the pointer p

Parameters

[in]

p

load address

Note

p does not have to be aligned

Returns

no return value. This vector is updated with value loaded from p.

svec< 4, float > sse::svec< 4, float >::load_and_splat ( float *  p )

static

Construct a vector by loading a scalar value from pointer p, and splat it to all the elements in the vector.

svec< 4, float > sse::svec< 4, float >::load_const ( const float *  p )

static

Construct a vector by loading a scalar value from pointer p, and splat it to all the elements in the vector.

svec< 4, float > sse::svec< 4, float >::log

(

)

return the log VTYPE vector.

svec< 4, float > sse::svec< 4, float >::masked_load ( svec< 4, float > *  p, svec< 4, bool >  mask  )

static

Return a new vector by only loading the value from the pointer p if the mask element is true.

void sse::svec< 4, float >::masked_store	(	svec< 4, float > *	p,
		svec< 4, bool >	mask
	)

Store the vector element's value to pointer p if the mask element is true.

svec< 4,bool > sse::svec< 4, float >::operator!=

(

svec< 4, float >

a

)

Element-wise compare not equal, return a bool vector. E.g. "a != b" *.

• Parameters

  [in]

  a

  a vector *

  Returns

  the result of compare not equal as a boolean vector.

svec< 4, float > sse::svec< 4, float >::operator*

(

float

s

)

Multiply a vector and a scalar.

svec< 4, float > sse::svec< 4, float >::operator*

(

svec< 4, float >

a

)

Multiply two vectors.

svec< 4, float > sse::svec< 4, float >::operator+

(

svec< 4, float >

a

)

Add two vectors.

svec< 4, float > sse::svec< 4, float >::operator+

(

float

s

)

Add a vector and a scalar.

svec< 4, float > sse::svec< 4, float >::operator-

(

float

s

)

Sub a vector and a scalar.

svec< 4, float > sse::svec< 4, float >::operator-

(

)

get the neg value, return a VTYPE vector. E.g. "-a"

svec< 4, float > sse::svec< 4, float >::operator-

(

svec< 4, float >

a

)

Sub two vectors.

svec< 4, float > sse::svec< 4, float >::operator/

(

svec< 4, float >

a

)

Divide a vector by a vector.

svec< 4, float > sse::svec< 4, float >::operator/

(

float

s

)

Divide a vector by a scalar.

svec< 4,bool > sse::svec< 4, float >::operator<

(

svec< 4, float >

a

)

Element-wise compare less than, return a bool vector. E.g. "a < b" *.

• Parameters

  [in]

  a

  a vector *

  Returns

  the result of compare less than as a boolean vector.

svec< 4,bool > sse::svec< 4, float >::operator<=

(

svec< 4, float >

a

)

Element-wise compare less equal, return a bool vector. E.g. "a <= b" *.

• Parameters

  [in]

  a

  a vector *

  Returns

  the result of compare less equal as a boolean vector.

svec< 4,bool > sse::svec< 4, float >::operator==

(

svec< 4, float >

a

)

Element-wise compare equal, return a bool vector, e.g., "a == b" *.

• Parameters

  [in]

  a

  a vector *

  Returns

  the result of compare equal as a boolean vector.

svec< 4,bool > sse::svec< 4, float >::operator>

(

svec< 4, float >

a

)

Element-wise compare greater than, return a bool vector. E.g. "a > b" *.

• Parameters

  [in]

  a

  a vector *

  Returns

  the result of compare greater than as a boolean vector.

svec< 4,bool > sse::svec< 4, float >::operator>=

(

svec< 4, float >

a

)

Element-wise compare greater equal, return a bool vector. E.g. "a >= b" *.

• Parameters

  [in]

  a

  a vector *

  Returns

  the result of compare greater equal as a boolean vector.

const float sse::svec< 4, float >::operator[]

(

int

index

)

const

float & sse::svec< 4, float >::operator[]

(

int

index

)

operator [] to set or get the vector element specified by index.

Parameters

index

specifies the index of the element in the vector.

svec< 4, float > sse::svec< 4, float >::pow

(

svec< 4, float >

a

)

return the pow VTYPE vector.

svec< 4, float > sse::svec< 4, float >::rcp

(

)

return the reverse VTYPE vector. (1.0/thisVec)

float sse::svec< 4, float >::reduce_add

(

)

Get the sum of all the element values in the vector. return a float scalar.

float sse::svec< 4, float >::reduce_max

(

)

Get the max value of all the element values in the vector. return a float scalar.

float sse::svec< 4, float >::reduce_min

(

)

Get the min value of all the element values in the vector. return a float scalar.

svec< 4, float > sse::svec< 4, float >::rotate

(

int32_t

index

)

Return a new vector by rotate this vector's elements. e.g. newVec[i] = thisVec[i+index].

svec< 4, float > sse::svec< 4, float >::round

(

)

return the round VTYPE vector.

svec< 4, float > sse::svec< 4, float >::rsqrt

(

)

return the reverse sqrt VTYPE vector. ( 1.0/sqrt(thisVec) )

void sse::svec< 4, float >::scatter	(	svec< 4, void * >	ptrs,
		svec< 4, bool >	mask
	)

Scatter the vector's elements to the locations pointed by the vector ptrs if the mask element is true.

void sse::svec< 4, float >::scatter_base_offsets	(	float *	b,
		uint32_t	scale,
		svec< 4, int64_t >	offsets,
		svec< 4, bool >	mask
	)

Scatter the vector's elements to the addresses ((char*)b + scale * offsets) if the mask element is true.

void sse::svec< 4, float >::scatter_base_offsets	(	float *	b,
		uint32_t	scale,
		svec< 4, int32_t >	offsets,
		svec< 4, bool >	mask
	)

Scatter the vector's elements to the addresses ((char*)b + scale * offsets) if the mask element is true.

void sse::svec< 4, float >::scatter_stride	(	float *	b,
		int64_t	off,
		int64_t	stride
	)

Scatter the vector's elements to the addresses (b+off, b++off+stride, b+off+2*stride, b+off+3*step).

void sse::svec< 4, float >::scatter_stride	(	float *	b,
		int32_t	off,
		int32_t	stride
	)

Scatter the vector's elements to the addresses (b+off, b++off+stride, b+off+2*stride, b+off+3*step).

svec< 4, float > sse::svec< 4, float >::shuffle

(

svec< 4, int32_t >

index

)

Return a new vector by shuffle this vector's elements with index vector e.g. newVec[i] = thisVec[index[i]].

svec< 4, float > sse::svec< 4, float >::sqrt

(

)

return the sqrt VTYPE vector.

void sse::svec< 4, float >::store

(

svec< 4, float > *

p

)

Store the vector to address p.

Parameters

[in]

p

store address

Note

p does not have to be aligned

Friends And Related Function Documentation

std::ostream& operator<< ( std::ostream &  out, const svec< 4, float > &  v  )

friend

Member Data Documentation

__m128 sse::svec< 4, float >::v

---

The documentation for this struct was generated from the following file:

• /cygdrive/c/workspace/ibm-summer/generic_simd/include/sse4.h