nrm2: Vector 2 norm

\(||x||_2\) More...

Functions
template<typename T >
real_type< T >	blas::nrm2 (int64_t n, T const *x, int64_t incx)
void	blas::nrm2 (int64_t n, float const *x, int64_t incx, float *result, blas::Queue &queue)
	GPU device, float version.
void	blas::nrm2 (int64_t n, double const *x, int64_t incx, double *result, blas::Queue &queue)
	GPU device, double version.
void	blas::nrm2 (int64_t n, std::complex< float > const *x, int64_t incx, float *result, blas::Queue &queue)
	GPU device, complex<float> version.
void	blas::nrm2 (int64_t n, std::complex< double > const *x, int64_t incx, double *result, blas::Queue &queue)
	GPU device, complex<double> version.
float	blas::nrm2 (int64_t n, float const *x, int64_t incx)
	CPU, float version.
double	blas::nrm2 (int64_t n, double const *x, int64_t incx)
	CPU, double version.
float	blas::nrm2 (int64_t n, std::complex< float > const *x, int64_t incx)
	CPU, complex<float> version.
double	blas::nrm2 (int64_t n, std::complex< double > const *x, int64_t incx)
	CPU, complex<double> version.

Detailed Description

\(||x||_2\)

Function Documentation

nrm2() [1/2]

void blas::nrm2	(	int64_t	n,
		float const *	x,
		int64_t	incx,
		float *	result,
		blas::Queue &	queue
	)

GPU device, float version.

Unlike CPU version, here result is an output parameter, to store the result when the asynchronous execution completes.

nrm2() [2/2]

template<typename T >

real_type< T > blas::nrm2	(	int64_t	n,
		T const *	x,
		int64_t	incx
	)

Returns

2-norm of vector, \(|| x ||_2 = (\sum_{i=0}^{n-1} |x_i|^2)^{1/2}\).

Generic implementation for arbitrary data types. TODO: generic implementation does not currently scale to avoid over- or underflow.

Parameters

[in]	n	Number of elements in x. n >= 0.
[in]	x	The n-element vector x, in an array of length (n-1)*incx + 1.
[in]	incx	Stride between elements of x. incx > 0.