__m64_pavg2_nraz

Microsoft Specific

Emits the 2-byte form of the IPF Parallel Average (pavg2) instruction, which computes the average of the two unsigned short integers in each 16-bit block of the input arguments and puts that average in the corresponding 16-bit block of the result.

__m64 __m64_pavg2_nraz( 
   __m64 a, 
   __m64 b 
);

Parameters

• [in] a
  An __m64 union containing an array of four unsigned 16-bit integers.

• [in] b
  An __m64 union containing an array of four unsigned 16-bit integers.

Return Value

An __m64 union containing the array of four unsigned 16-bit integers, each element of which is the average of the corresponding integers in the source arrays.

Requirements

Header file <intrin.h>

Remarks

If fractional, the result of each average is rounded to the nearest odd integer. The suffix nraz means "not rounding away from zero."

Example

// pavg2_nraz.cpp
// processor: IPF
#include <stdio.h>
#include <intrin.h>

#pragma intrinsic(__m64_pavg2_nraz)

void print4u(unsigned __int16* ia)
{
    printf_s("{ %8u, %8u, %8u, %8u }\n", ia[0], ia[1], ia[2], ia[3] );
}


int main()
{
    __m64 m, n, result;
    unsigned __int16 ra[4], i;
    unsigned __int16 ma[4] = { 10, 100, 70, 10000 };
    unsigned __int16 na[4] = { 1, 95, 50,   5000 }; 

    for (i = 0; i < 4; i++)
    {
       m.m64_u16[i] = ma[i];
       n.m64_u16[i] = na[i];
    }
    
    printf_s("a: \n");
    print4u(ma);
    printf_s("b: \n");
    print4u(na);
    printf_s("\n");

    result = __m64_pavg2_nraz(m, n);    
    
    printf_s("Parallel Average (a[]+b[])/2 : \n");
    print4u(result.m64_u16);
    printf_s("\n");
       
    // This operation is commutative.
    result = __m64_pavg2_nraz(n, m);
    
    printf_s("Parallel Average (b[]+a[])/2 : \n");
    print4u(result.m64_u16);
    printf_s("\n");
}

a: 
{       10,      100,       70,    10000 }
b: 
{        1,       95,       50,     5000 }

Parallel Average (a[]+b[])/2 : 
{        5,       97,       60,     7500 }

Parallel Average (b[]+a[])/2 : 
{        5,       97,       60,     7500 }

See Also

Concepts

__m64

Compiler Intrinsics