CannyLab / tsne-cuda

GPU Accelerated t-SNE for CUDA with Python bindings
BSD 3-Clause "New" or "Revised" License
1.78k stars 126 forks source link

Windows support #130

Open MikeTkachuk opened 7 months ago

MikeTkachuk commented 7 months ago

Disclaimer: It's my first experience with cmake and cuda source, so it's one of the dirties PRs I've ever done. But here is the build that worked for me. Hope this helps somebody run or turn this into a conda-forge package, seems ok for a separate branch Did not benchmark, but it runs tests. My setup:

See windows.txt file for command reference

DavidMChan commented 7 months ago

Hi @MikeTkachuk - Thanks for this contribution! I'm amazed that you could get it working on Windows! To merge this, we'll have to correctly set the cmake flags dependent on the underlying OS (so as to make this platform agnostic) -- the other changes should be able to be preserved as is, since long is underspecified on linux anyways (and int64_t is probably better). Until then, I'll leave this PR open, so people can use this branch/fork if it works for them!

I'm not entirely sure how to update the conda-forge recipe (https://github.com/conda-forge/tsnecuda-feedstock), since @h-vetinari took care of that for us, but that will have to be updated as well.

raulmendez commented 5 months ago

Hi @MikeTkachuk: can you please be more explicit? I do not find the windows.txt file, and need a little bit more instructions (by the way I am on Windows 10, not 11).

Thanks!

MikeTkachuk commented 5 months ago

Hi, this is just a draft of how I made it to work, you can find windows.txt in the PR files. You should still modify a lot of paths until someone has the time to integrate it, Regards

raulmendez commented 5 months ago

Thanks @MikeTkachuk : I saw it https://github.com/MikeTkachuk/tsne-cuda/blob/main/windows.txt . I do not understand very well the questions, are they options for cmake?

MikeTkachuk commented 5 months ago

I reformatted windows.txt file a bit, let me know if it sounds clear now