在Linux下分别用Ptread线程和一般的方法编矩阵乘法，比较两者的差异。-we use Pthread and the general method to implement the matrix multiplication,and compare the difference between the two methods.

机器学习irect和unrolling乘法数一模一样，所以图上只有direct的。unrolling方法就是为blas/gemm设计的，至于blas/gemm把矩阵乘复杂度、乘法数降到多少还没见，但是那就到blas优化的范畴了。 -Machine Learning irect and unrolling exactly the number of multiplications, so the figure of only direct. unrolling method is to bla