Export (0) Print
Expand All
Expand Minimize
This topic has not yet been rated - Rate this topic

__m64_pmpy2l, __m64_pmpy2r

Microsoft Specific

Emit the two forms of the IPF Parallel Multiply (pmpy2) instruction, which performs multiplication in parallel on multiple integers.

__m64 __m64_pmpy2l( 
   __m64 a, 
    __m64 b 
);
__m64 __m64_pmpy2r( 
   __m64 a, 
   __m64 b 
);
[in] a

An __m64 union containing an array of four 16-bit signed integers.

[in] b

An __m64 union containing an array of four 16-bit signed integers.

An __m64 union containing two 32-bit signed integers representing the results of the multiplication.

Intrinsic

Architecture

__m64_pmpy2l

IPF

__m64_pmpy2r

IPF

Header file <intrin.h>

The __m64_pmpy2l (left) intrinsic multiplies a[1]*b[1] and a[3]*b[3]. The __m64_pmpy2r (right) intrinsic multiplies a[0]*b[0] and a[2]*b[2]. In both cases, the result is two 32-bit signed integers, which are placed in elements 0 and 1 of the result.

// pmpy2.cpp
// processor: IPF
#include <stdio.h>
#include <intrin.h>

#pragma intrinsic(__m64_pmpy2l, __m64_pmpy2r)

void print16(__int16* ia)
{
    printf_s("{ %8d, %8d, %8d, %8d }\n", ia[0], ia[1], ia[2], ia[3]);
}

void print32(__int32* ia)
{
    printf_s("{ %8d, %8d }\n", ia[0], ia[1]);
}

int main()
{
    __int16 a[4] = { -1, 1, -2, 6000 };
    __int16 b[4] = { -1, 2, 10, 15000 };
    __int32 cl[2], cr[2], i;
    __m64 m, n, result;

    printf_s("a: \n");
    print16(a);
    printf_s("b: \n");
    print16(b);

    for (int i = 0; i < 4; i++)
    {
       m.m64_i16[i] = a[i];
       n.m64_i16[i] = b[i];
    }

    printf_s("__m64_pmpy2r computes a[0]*b[0] and a[2]*b[2] : ");
    result = __m64_pmpy2r(m, n);
    print32(result.m64_i32);
    
    result = __m64_pmpy2l(m, n);
    printf_s("__m64_pmpy2l computes a[1]*b[1] and a[3]*b[3] : ");
    print32(result.m64_i32);

}
a: 
{       -1,        1,       -2,     6000 }
b: 
{       -1,        2,       10,    15000 }
__m64_pmpy2r computes a[0]*b[0] and a[2]*b[2] : {        1,      -20 }
__m64_pmpy2l computes a[1]*b[1] and a[3]*b[3] : {        2, 90000000 }
Did you find this helpful?
(1500 characters remaining)
Thank you for your feedback

Community Additions

ADD
Show:
© 2014 Microsoft. All rights reserved.