c - 没有SSE4.1的高效SSE FP`floor()`/`ceil()`/`round()`舍入函数？

如何将浮点向量向上/向下舍入到最接近的整数，就像这些函数一样？
Round - __m128
Ceil - roundf()或SSE4.1ceilf()。
Floor - _mm_ceil_ps或SSE4.1floorf()。
我需要在没有SSE4.1_mm_floor_ps（roundps/_mm_floor_ps/_mm_ceil_ps的情况下执行此操作_mm_round_ps(x, _MM_FROUND_TO_NEAREST_INT |_MM_FROUND_NO_EXC)也可以截断为零，但对于这个应用程序，我不需要这样做。
我可以用SSE3和更早的（无SSSE3或SSE4）
所以函数声明如下：
roundps、__m128 RoundSse( __m128 x )和__m128 CeilSse( __m128 x )。

最佳答案

我将从http://dss.stephanierct.com/DevBlog/?p=8发布代码：
它应该采用按值形式（我刚从代码中删除了&，不确定它是否正常）：

static inline __m128 FloorSse(const __m128 x) {
    __m128i v0 = _mm_setzero_si128();
    __m128i v1 = _mm_cmpeq_epi32(v0, v0);
    __m128i ji = _mm_srli_epi32(v1, 25);
    __m128i tmp = _mm_slli_epi32(ji, 23); // I edited this (Added tmp) not sure about it
    __m128 j = _mm_castsi128_ps(tmp); //create vector 1.0f // I edited this not sure about it
    __m128i i = _mm_cvttps_epi32(x);
    __m128 fi = _mm_cvtepi32_ps(i);
    __m128 igx = _mm_cmpgt_ps(fi, x);
    j = _mm_and_ps(igx, j);
    return _mm_sub_ps(fi, j);
}

static inline __m128 CeilSse(const __m128 x) {
    __m128i v0 = _mm_setzero_si128();
    __m128i v1 = _mm_cmpeq_epi32(v0, v0);
    __m128i ji = _mm_srli_epi32(v1, 25);
    __m128i tmp = _mm_slli_epi32(ji, 23); // I edited this (Added tmp) not sure about it
    __m128 j = _mm_castsi128_ps(tmp); //create vector 1.0f // I edited this not sure about it
    __m128i i = _mm_cvttps_epi32(x);
    __m128 fi = _mm_cvtepi32_ps(i);
    __m128 igx = _mm_cmplt_ps(fi, x);
    j = _mm_and_ps(igx, j);
    return _mm_add_ps(fi, j);
}

static inline __m128 RoundSse(const __m128 a) {
    __m128 v0 = _mm_setzero_ps();             //generate the highest value < 2
    __m128 v1 = _mm_cmpeq_ps(v0, v0);
    __m128i tmp = _mm_castps_si128(v1); // I edited this (Added tmp) not sure about it
    tmp = _mm_srli_epi32(tmp, 2); // I edited this (Added tmp) not sure about it
    __m128 vNearest2 = _mm_castsi128_ps(tmp); // I edited this (Added tmp) not sure about it
    __m128i i = _mm_cvttps_epi32(a);
    __m128 aTrunc = _mm_cvtepi32_ps(i);        // truncate a
    __m128 rmd = _mm_sub_ps(a, aTrunc);        // get remainder
    __m128 rmd2 = _mm_mul_ps(rmd, vNearest2); // mul remainder by near 2 will yield the needed offset
    __m128i rmd2i = _mm_cvttps_epi32(rmd2);    // after being truncated of course
    __m128 rmd2Trunc = _mm_cvtepi32_ps(rmd2i);
    __m128 r = _mm_add_ps(aTrunc, rmd2Trunc);
    return r;
}


inline __m128 ModSee(const __m128 a, const __m128 aDiv) {
    __m128 c = _mm_div_ps(a, aDiv);
    __m128i i = _mm_cvttps_epi32(c);
    __m128 cTrunc = _mm_cvtepi32_ps(i);
    __m128 base = _mm_mul_ps(cTrunc, aDiv);
    __m128 r = _mm_sub_ps(a, base);
    return r;
}