topic Re: Intializing an array in Intel® oneAPI Math Kernel Library

Intializing an array

janzz2k — Tue, 20 May 2008 19:10:36 GMT

Hi,
Suppose I have an array of doubles created using malloc. Is there an efficient (vectorized + optimised) way to set some or all its elements to a specific value, say 1.0? I want to avoid using a for loop to do this element by element. In other words, is there a more optimal way to do the following:

int main()
{
int n=1000;
double *x=(double *) malloc( n*sizeof(double));

int p=20, q=40;
for (int i=p; i < q; ++ i) x=1.0;

return(0);
}

Re: Intializing an array

TimP — Tue, 20 May 2008 21:06:35 GMT

Compilation of your code by an auto-vectorizing compiler may give excellent results. Your loop is probably not long enough for icc #pragma vector nontemporal tobe useful. If you have taken care to get a 16-byte aligned allocation (should be the default on 64-bit OS), icc #pragma vector aligned should speed it up.

You probably spend more time in malloc() than in the initialization, for such a short loop.

If you don't want to use an auto-vectorizing compiler, you should be able to show an advantage from compiling with SSE intrinsics. It seems like overkill on a simple example like this.

Re: Intializing an array

janzz2k — Tue, 20 May 2008 22:11:44 GMT

I used the value n=1000 just as an example. In my actual code, it may be quite large, such as 1 million or 1 billion.