valarray construction from C-array

[klucar]

I'll start off differently by showing what I know how to do:

valarray<float> va(100, 0.0f);
float* fp;
fp = &va[0];
some_c_function( fp, va.size() );

What I want to do though is quite the opposite:

float* fp;
valarray<float> va;

&va[0] = fp; // this doesn't work on many levels

What I am trying to show is that I want to construct a valarray, but
use some pointer (from shared memory for example) to initialize the
valarray. I want to overlay a valarray on a c-style array so I don't
suffer the penalty of copying all the data from the array like the
following constructor does:

float af1 [] = {0, 1, 2, 3};
valarray <float> vf1 (af1, 4);

Is this possible with just STL or do I need to dive in to some other
C++ numerics library?

[Daniel T.]

In article <11*********************@g47g2000cwa.googlegroups. com>,
"klucar" <kl****@gmail.com> wrote:

I'll start off differently by showing what I know how to do:

valarray<float> va(100, 0.0f);
float* fp;
fp = &va[0];
some_c_function( fp, va.size() );

What I want to do though is quite the opposite:

float* fp;
valarray<float> va;

&va[0] = fp; // this doesn't work on many levels

What I am trying to show is that I want to construct a valarray, but
use some pointer (from shared memory for example) to initialize the
valarray. I want to overlay a valarray on a c-style array so I don't
suffer the penalty of copying all the data from the array like the
following constructor does:

float af1 [] = {0, 1, 2, 3};
valarray <float> vf1 (af1, 4);

Is this possible with just STL or do I need to dive in to some other
C++ numerics library?

int array[] = { 3, 6, 18, 3, 22 };

// initialize valarray by elements of an ordinary array
std::valarray<int> va3(array, sizeof (array)/sizeof (array[0]));

// initialize by the second to the fourth element
std::valarray<int> va4(array+1, 3);

--
Magic depends on tradition and belief. It does not welcome observation,
nor does it profit by experiment. On the other hand, science is based
on experience; it is open to correction by observation and experiment.

[Daniel T.]

In article <11*********************@g47g2000cwa.googlegroups. com>,
"klucar" <kl****@gmail.com> wrote:

I'll start off differently by showing what I know how to do:

valarray<float> va(100, 0.0f);
float* fp;
fp = &va[0];
some_c_function( fp, va.size() );

What I want to do though is quite the opposite:

float* fp;
valarray<float> va;

&va[0] = fp; // this doesn't work on many levels

What I am trying to show is that I want to construct a valarray, but
use some pointer (from shared memory for example) to initialize the
valarray. I want to overlay a valarray on a c-style array so I don't
suffer the penalty of copying all the data from the array like the
following constructor does:

float af1 [] = {0, 1, 2, 3};
valarray <float> vf1 (af1, 4);

Is this possible with just STL or do I need to dive in to some other
C++ numerics library?

Sorry, my last post was premature. valarray holds its data by value just
like other containers.

--
Magic depends on tradition and belief. It does not welcome observation,
nor does it profit by experiment. On the other hand, science is based
on experience; it is open to correction by observation and experiment.

[klucar]

So is my only hope to get a pointer to a valarray object? It just
seems bad that C++ containers have to be in charge. I guess a memory
manager object would have to use valarrays to store data and then pass
out valarray pointers to C++ objects and C pointers to C functions
using &va[0].