SIMD（Single Instruction Multiple Data）是指单指令多数据的运算



方式，这种技术可以使一条指令同时应用于多个数据集，这种处理方式显然在对不同数据集进行同一操作时非常有用，一个简单的例子就是，我们需要把平面空间的圆转换成3D空间的圆（由多个三角形构成），这其中需要大量的三角形计算，并且其中的多个三角形运算都是一些重复性运算，使用浮点SIMD指令显然就可以通过名为衍生计算的算法加快整个运算完成的速度，而在原来如果仅通过FPU浮点单元运算的话，所有数据都需要进行逐个计算，从而大大浪费了CPU资源。-SIMD (Single Instruction Multiple Data) is the single instruction multiple data calculation methods, this technique allows a single instruction to multiple data sets at the same time, this approach is clearly different data sets in the same operation is useful when a The simplest example is, we need to round to round plane into the space in 3D space (composed of a plurality of triangles), which requires a lot of computing triangle, and wherein a plurality of triangular some repetitive arithmetic operation is to use a floating Obviously you can point SIMD instructions derived by calculating algorithm called the accelerated speed of the entire operation is completed, if only in the original floating-point unit FPU operation by then, all the data needs to be individually calculated, thereby greatly wasted CPU resources.