共用方式為


_mm_blendv_ps

Microsoft Specific

Emits the Streaming SIMD Extensions 4 (SSE4) instruction blendvps. This function blends packed single precision floating point values.

__m128 _mm_blendv_ps( 
   __m128 a,
   __m128 b,
   __m128 mask 
);

Parameters

  • [in] a
    A 128-bit parameter that contains four 32-bit floating point values to be blended.

  • [in] b
    A 128-bit parameter that contains four 32-bit floating point values to be blended.

  • [in] mask
    A 128-bit mask that selects components for the blend operation.

Result value

r0 := (mask0 & 0x80000000) ? b0 : a0
r1 := (mask1 & 0x80000000) ? b1 : a1
r2 := (mask2 & 0x80000000) ? b2 : a2
r3 := (mask3 & 0x80000000) ? b3 : a3

Requirements

Intrinsic

Architecture

_mm_blendv_ps

x86, x64

Header file <smmintrin.h>

Remarks

The return value r and parameters a, b, and mask each consist of 128 bits. r0-r3, a0-a3, b0-b3, and mask0-mask3 are the sequentially ordered 32-bit components of these parameters, where r0, a0, b0, and mask0 indicate the least significant 32 bits.

Before using this intrinsic, software must ensure that the processor supports the instruction.

Example

#include <stdio.h>
#include <smmintrin.h>

int main () {
    __m128 a, b, mask;
    unsigned __int64 maskValue[4];

    maskValue[3] = 0x80000000;
    maskValue[2] = 0x80000000;
    maskValue[1] = 0x00000000;
    maskValue[0] = 0x00000000;

    mask.m128_f32[3] = *reinterpret_cast<float *>(&maskValue[3]);
    mask.m128_f32[2] = *reinterpret_cast<float *>(&maskValue[2]);
    mask.m128_f32[1] = *reinterpret_cast<float *>(&maskValue[1]);
    mask.m128_f32[0] = *reinterpret_cast<float *>(&maskValue[0]);

    a.m128_f32[3] = -10.25;
    a.m128_f32[2] = -20;
    a.m128_f32[1] = -900;
    a.m128_f32[0] = -32786;
    b.m128_f32[3] = 36;
    b.m128_f32[2] = 0;
    b.m128_f32[1] = 3.25;
    b.m128_f32[0] = 78.75;

    __m128 res = _mm_blendv_ps( a, b, mask );

    printf_s("Original a: %14f %14f %14f %14f\nOriginal b: %14f %14f %14f %14f\n",
                a.m128_f32[3], a.m128_f32[2], a.m128_f32[1], a.m128_f32[0],
                b.m128_f32[3], b.m128_f32[2], b.m128_f32[1], b.m128_f32[0]);

    printf_s("Result res: %14f %14f %14f %14f\n",
                res.m128_f32[3], res.m128_f32[2], res.m128_f32[1], res.m128_f32[0]);

    return 0;
}
Original a:     -10.250000     -20.000000    -900.000000  -32786.000000
Original b:      36.000000       0.000000       3.250000      78.750000
Result res:      36.000000       0.000000    -900.000000  -32786.000000

See Also

Reference

Compiler Intrinsics