Converts the four single-precision, floating-point values of a to signed 32-bit integer values by rounding (see _mm_cvtps_epi32)
在文件 AkSimd.h 第 399 行定义.