矩阵乘法的多线程并行实现，包括：simple，采用openmp对索引顺序ijk（一般写法）的循环实现并行；opt，采用openmp对索引顺序ikj（向量化友好）的循环实现并行；mkl，调用高性能数学库实现矩阵乘法。(Multi-thread parallelization of matrix-matrix multiplication: "simple" case, ijk loop (usual pattern); "opt" case, ikj (vectorization friendly); "mkl" case, calling Intel MKL.)