Does Numpy on Google App Engine use LAPACK or ATLAS

Question

I have an app on google app engine written in python 2.7 that uses the numpy third party library that is included with app engine. I am using numpy.linalg.solve() to solve a linear algebra equation. Does numpy on app engine make use of ATLAS, or LAPACK or BLAS to make the numpy.linalg.solve() function to run faster or is this function only executing in Python. The reason I ask is that I noticed that the numpy.linalg.solve() routine runs about 10X slower on app engine than a comparable solver on MATLAB on my computer. MATLAB uses ATLAS, LAPACK and BLAS under the hood so I was wondering numpy on app engine does also. The python command numpy.config.show() on app engine lists information about ATLAS, LAPACK and BLAS but I would like to know is does numpy.linalg.solve() make use of these libraries under the hood, or is it just executing python code?

Answer 1

The documentation for the function says:

The solutions are computed using LAPACK routine _gesv

Is that what you're after?

More generally, numpy comprises C code as well as python, which is why it's one of the libraries you can opt in to using - you're not allowed to upload it yourself.

In terms of the relative performance to your computer, the likelihood is that your computer is significantly more powerful than one of the standard appengine instances - the default for a frontend instance, the F1, is just 600MHz, with 128MB RAM.