Converts the four single-precision, floating-point values of a to signed 32-bit integer values by truncating (see _mm_cvttps_epi32)
在文件 AkSimd.h 第 206 行定义.