SSE指令:字节+短

Question

我有很长的字节数组需要添加到类型short(或int)的目标数组.这样的SSE指令是否存在？或者他们的设定？

Answer 1

您需要将每个8位值向量解包到两个16位值的向量中,然后添加它们.

__m128i v = _mm_set_epi8(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
__m128i vl = _mm_unpacklo_epi8(v, _mm_set1_epi8(0)); // vl = { 7, 6, 5, 4, 3, 2, 1, 0 }
__m128i vh = _mm_unpackhi_epi8(v, _mm_set1_epi8(0)); // vh = { 15, 14, 13, 12, 11, 10, 9, 8 }

其中v是16 x 8位值的向量vl,vh是两个8 x 16位值的解包向量.

注意,我假设8位值是无符号的,因此当解包为16位时,高字节被设置为0(即没有符号扩展).

如果你想总结很多这些向量并获得32位结果,那么一个有用的技巧是使用_mm_madd_epi16乘数为1,例如

__m128i vsuml = _mm_set1_epi32(0);
__m128i vsumh = _mm_set1_epi32(0);
__m128i vsum;
int sum;

for (int i = 0; i < N; i += 16)
{
    __m128i v = _mm_load_si128(&x[i]);
    __m128i vl = _mm_unpacklo_epi8(v, _mm_set1_epi8(0));
    __m128i vh = _mm_unpackhi_epi8(v, _mm_set1_epi8(0));
    vsuml = _mm_add_epi32(vsuml, _mm_madd_epi16(vl, _mm_set1_epi16(1)));
    vsumh = _mm_add_epi32(vsumh, _mm_madd_epi16(vh, _mm_set1_epi16(1)));
}
// do horizontal sum of 4 partial sums and store in scalar int
vsum = _mm_add_epi32(vsuml, vsumh);
vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 8));
vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 4));
sum = _mm_cvtsi128_si32(vsum);