Computes the maximums of the four single-precision, floating-point values of a and b (see _mm_max_ps)
在文件 AkSimd.h 第 214 行定义.