0

Yes, I am having the opposite problem. Python seems to be so, so much lower on my Linux (Pop OS 22.04LTS) than on my Windows 10 install.

The algorithm I'm trying to run is the Iterative deepening Boolean Negamax algorithm. On Windows, the algorithm can solve a game around 20 seconds. However, that same game requires 45 seconds to run on my Linux.

The game is a small (4x4) Go board, and it will try and generate a move when a player requests it.

I'm using dual boot, so they have the same hardware (Ryzen 9 5900X DDR4 3600 64GB). The Python in my Linux is the one shipped with Pop OS 22.04 LTS, while the one in Windows is installed through the Microsoft store. Both are Python 3.10.

If there is any other output that could be useful, let me know

Flyrobot27
  • 43
  • 6
  • What libraries are you using? Numpy in particular can be way slower if compiled poorly – wim Dec 05 '22 at 05:38
  • I've definitely used numpy. How can it be compiled poorly? I'm using pip to install it – Flyrobot27 Dec 05 '22 at 07:03
  • 1
    See [_Find out if/which BLAS library is used by Numpy_](https://stackoverflow.com/q/37184618/674039) – wim Dec 05 '22 at 07:25
  • 1
    Regarding the very small board, it is very likely that the major part of the time is pure overheads. Overheads tends to drastically change from one platform to another. This should not happen in practice since it generally happen only in unoptimized codes. Additionally, there are a lot differences in the two OS that can explain that (management of virtual memory on a NUMA system like yours, different libraries or library versions, different compilers, etc). Please provide a **Minimal Working Example** and the versions of the libraries. A new version of Numpy is likely to be faster (not always) – Jérôme Richard Dec 05 '22 at 10:21

0 Answers0