Ben Thompson

Thanks for bringing this up! The error you got is expected because you’re running the script from within the okada_wrapper folder. As a result, you’re importing the local version of...

> I have tried running the script from my home directory and that's where I get the error message. Oh yeah I missed that line at the top of your...

Two debugging tips if you're interested: - you can check your `sys.path` to see the order of folders Python will look in when you `import` a package. - you can...

Alternatively, I suspect some asynchronous data transfer running simultaneously with the kernel might help make the chunking as fast as the unchunked version. The synchronous chunked version has lots of...

To test this manually: ``` import os import numpy as np import cutde.gpu as gpu gpu_config = dict(float_type='float', verbose=True) gpu.load_gpu( "aca.cu", tmpl_args=gpu_config, tmpl_dir=os.path.join(os.getcwd(), "cutde") ) ```

I agree that this isn't precisely a "bug", but it would be nice to improve the function or the documentation. Looking at the code: https://github.com/tbenthompson/cutde/blob/3ac6a3e19f70ba50fa87a16d81bec4d96760e273/cutde/geometry.py#L4-L28 the current version assumes that...

I made a small update to the README.

Thanks!! So it looks like there's no easy solution at the moment? If that's the case, I'll just leave this issue open as a reminder of the problem.

Ah yeah. I haven't ever used cutde in a multi-GPU settings but I would recommend: - pair one GPU with one process. Either by launching separate python processes or using...

Oh damn. I just realized that `"--use_fast_math"` compiler argument is probably the reason for those NaN bugs you fixed a few weeks back... Without that, I bet the floating point...