shuffling upper 32 bits with lower 32 bits in m128

Question

I'm working with C intrinsics (SSE/SSE2 only) right now, and i have a m128 value with 4 floats in it. Are there any possibility of shifting / shuffling / moving the most upper 32 bits to most lower 32 bits?

Example : I have {1.0f, 2.0f, 3.0f, 4.0f} in m128 and i want to make {4.0f, 2.0f, 3.0f, 1.0f} out of it. (the values in beetween may be erased).

Answer 1

You can do that via shufps xmm, xmm, imm8 instruction, with which you can statically select which input word should be stored for each output word.

#include <stdio.h>
#include <xmmintrin.h>

int main(void) {
    float array[4] = {1.0f, 2.0f, 3.0f, 4.0f};
    __m128 data;
    printf("before : %.1f %.1f %.1f %.1f\n", array[0], array[1], array[2], array[3]);
    data = _mm_loadu_ps(array);
    data = _mm_shuffle_ps(data, data, 0x27);
    _mm_storeu_ps(array, data);
    printf("after  : %.1f %.1f %.1f %.1f\n", array[0], array[1], array[2], array[3]);
    return 0;
}