CMP AVX intrinsic (_mm256_cmp_pd)

[Guillaume_S_]

Hey everyone 

I'm testing some intrinsics and checking some results. Unfortunately, I don't understand why when I'm trying to evaluate a > 2, I get "-nan" as an output for the true cases. Maybe I did something wrong but I thought this should return 1 when it's true and 0 when it's false. I looked other internet for this error but I didn't find anything.

Here's my code:

#include <iostream>
#include <immintrin.h>

using namespace std;

void print( double *A, int N )
{
	for( int i(0); i<N; ++i )
		cout << A << "\t" ;
	
	cout << endl;
}

int main( void )
{
	int align = 64;
	int N = 4;
	
	double *A = (double*)_mm_malloc(N*sizeof(double), align);
	__m256d *a =(__m256d*)A;
	
	for( int i(0); i<N; ++i )
		A = i + 1 ;
		
	print( A, N );
		
	__m256d _2_ = _mm256_set1_pd( 2. );
	
	a[0] = _mm256_cmp_pd( a[0], _2_, _CMP_GT_OQ );
	
	print( A, N );
		
	_mm_free( A );
		
	return 0;
}

And the output:

1       2       3       4
0       0       -nan    -nan

My configuration:

 Intel(R) Xeon(R)  E5-2650 v2

icpc version 14.0.2

Any ideas?

Guillaume,