'How to switch between low-level math libraries linked to in scientific computing packages used in MATLAB/Python, on Windows?
I have an AMD Ryzen 9 5900 cpu running on a Windows 10 build. Partly because of the known issues with Intel's MKL running on AMD CPUs, and partly because I just want to try out AMD's Optimizing CPU Libraries (AOCL), I'm looking to install AOCL and do some benchmarking to see how much performance changes on my build for some matlab and python (numpy and scipy) functions.
AMD has simple instructions on how to install the AOCL binaries on Windows, but I couldn't find a more detailed explanation of how to changes the paths to allow functions in higher-level programming languages (e.g. matlab and python) to link to AOCL over MKL (and whatever other OME defaults).
Are there, for example, settings in MATLAB that a user can change to specify which lower-level math libraries to use? Similarly, for python packages, how can I specify which lower-level math libraries the package uses (I am using conda as my python env and package manager, but is there a general way to do this, or only python distribution and package manager specific)?
TLDR; I want to be able to switch between which lower-level libraries are linked to at runtime - is this possible and does anyone know how to do so?
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|