简体繁体 English

numpy+mkl 比 numpy 快吗？

[英]Is numpy+mkl faster than numpy?

原文 2018-04-24 01:42:53 0 2 python/ numpy/ intel-mkl

If numpy+mkl is faster, how much faster is it than numpy?如果 numpy+mkl 更快，那么它比 numpy 快多少？

I found that the numpy+mkl installation package is much larger than numpy, but I can't feel the difference in their speed.发现numpy+mkl安装包比numpy大很多，但是感觉不到它们的速度差异。

2 个解决方案

numpy's standard pip packages are linked to OpenBLAS. numpy 的标准 pip 包链接到 OpenBLAS。 Mabey OpenBLAS is in the same ballpark for speed as MKL. Mabey OpenBLAS 与 MKL 的速度相同。 For a detailed comparison of their speeds, please see Boosting numpy: Why BLAS Matters and Benchmark OpenBLAS, Intel MKL vs ATLAS有关它们速度的详细比较，请参阅Boosting numpy：Why BLAS Matters and Benchmark OpenBLAS, Intel MKL vs ATLAS

Short answer: No, maybe just slightly.简短回答：不，也许只是一点点。

I just repeated now benchmark from this site: https://jwalton.info/Python-MKL-openBLAS/ .我刚刚重复了这个网站的基准测试： https : //jwalton.info/Python-MKL-openBLAS/ 。 And the results on current version of numpy are similar to those on intel-numpy .并且在当前版本的numpy上的结果与intel-numpy上的结果相似。

Speedup > 1 means MKL is faster. Speedup > 1 意味着 MKL 更快。 Speedup < 0 means "standard" numpy (using openBLAS) is faster. Speedup < 0 意味着“标准”numpy（使用 openBLAS）更快。

在此处输入图片说明

As you can see, differences are small.如您所见，差异很小。 For some functions there is small (~1.1x) speedup.对于某些功能，加速幅度很小（~1.1 倍）。 But for most of them there is no speedup and one function is significantly slower in MKL.但是对于他们中的大多数人来说，没有加速，而且 MKL 中的一个函数明显变慢。