Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Will arrayfire-r be updated soon to allow Windows usage? #11

Open
Despertaferro opened this issue Sep 23, 2018 · 8 comments
Open

Will arrayfire-r be updated soon to allow Windows usage? #11

Despertaferro opened this issue Sep 23, 2018 · 8 comments

Comments

@Despertaferro
Copy link

I see arrayfire-r has been dormant for a few years. In the ArrayFire page it says there is R support. In arrayfire-r, it says Linux is supported, but Windows will be supported "soon".

Is it expected a new arrayfire-r version, which will allow Windows usage?

Thank you for your work.

@9prady9
Copy link
Member

9prady9 commented Sep 23, 2018

@Despertaferro Unfortunately, there isn't any active development going on the R wrapper support for quite some time. We haven't had that many requests for R wrapper, so updating/maintaining this language binding is currently not at the top of the priority list.

@pavanky Any chance that you are working on latest version but haven't published it yet ?

@jeffkeller87
Copy link

You have my vote for an R wrapper. I was actually looking to contribute to a ViennaCL R wrapper when I came across ArrayFire, whose core appears to be much more actively maintained.

@9prady9
Copy link
Member

9prady9 commented May 22, 2019

@jeffkeller87 At the moment, we don't have any active development going on for R bindings. You are welcome to contribute to R wrapper for ArrayFire. We are available on Slack for quick questions. We are also on Google groups for much detailed discussions.

@rstub
Copy link

rstub commented Jun 7, 2019

Getting R and AF to work together on Windows is complicated. If you use the pre-build binaries you have to use MSVC for compiling the R package's library, while normally mingw (from Rtools) is used. Alternatively you could compile AF with the mingw tool chain, but that rules out CUDA. See daqana/rcpparrayfire#3 for a similar discussion also without solution as of now.

@9prady9
Copy link
Member

9prady9 commented Jun 7, 2019

@rstub Can you please explain what do you mean "mingw rules out CUDA".

As long as C-API from our website binaries is used, you wouldn't face any issues even if our binaries are built using MSVC and your target toolchain is mingw.

@rstub
Copy link

rstub commented Jun 7, 2019

@rstub Can you please explain what do you mean "mingw rules out CUDA".

I meant "you cannot build the CUDA backend if you build AF using the mingw tool chain".

As long as C-API from our website binaries is used, you wouldn't face any issues even if our binaries are built using MSVC and your target toolchain is mingw.

I do not remember the details, but I think I had linking problems using the AF's C-API together with mingw. I have not investigated this in detail, though, since I need the C++-API for RcppArrayFire anyway. And that cannot work with mingw due to different C++ name mangling schema.

@9prady9
Copy link
Member

9prady9 commented Jun 7, 2019

That is right, C++ will have the problem you mentioned. But C-API shouldn't be a problem. I believe I have once tried using ArrayFire binaries(website download) with rust mingw toolchain, if I remember correctly, it did work.

To compile AF, you need a C++ compiler. Now, I haven't tried CUDA backend with mingw toolchain, is it a well known issue ? or something you noticed only with ArrayFire ?

@rstub
Copy link

rstub commented Jun 7, 2019

Interesting. I might give it another shot once I have a Windows dev machine running.

AFAIK CUDA on windows requires MSVC, c.f. https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html. I am not sure if it possible to compile the CUDA parts with nvcc, which uses MSVC, and the CPU parts in mingw.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants