Closed bhlevca closed 4 months ago
It appears that it has something to do with the CPU. I have an AMD Ryzen. NumPy is slower when using the blas lib. On Intel platforms numpy is using MKL
@bhlevca can we agree that this is not directly related to MIKE IO?
Describe the bug DFSU read(x,y) is 20-30 times slower on Linux when compared with Windows from a large DFSU
To Reproduce load a large dfsu file like in Dfsu - 3D sigma-z.ipynb
Expected behavior I would expect similar timing not such a huge performance hit on Linux. I am curious if you can reproduce the Linux slowness. Everything else works fast on my Linux machine.
System information: