Charles Determan

I don't have much experience with it so I couldn't say for sure. But I believe the code on which it is based in ViennaCL is 'experimental'. You can see...

Not currently, are you concerned about some potential performance implications or have you run in to some problems? In theory it would be possible. Are you interested in a global...

Thinking about this and looking at the code I don't think this would be possible with the implemented operations. I think they are leveraging ViennaCL's autotuning framework so the work...

Unfortunately there isn't a lot at the moment. I am aware that the `chol` implementation is rather slow. This is because such functionality is 'experimental' in the underlying C++ code....

Out of curiosity, what GPU do you have?

So to confirm, when you are doing your benchmarks you are making sure to assign the appropriate context id to make sure the relevant device is actually being used? For...

Yes, that is the same for a global assignment instead of individually. Just wanted to make sure as other users had overlooked that part.

@fafazai Unfortunately no, at the moment. The use of external pointers makes the default parallel backends non-functional for `gpuR` objects. The R objects are not thread safe so when you...

This is proving too complex for the 1.2.0 release. I am also thinking that this functionality (i.e. using multiple GPUs for a single task) may need to have it's own...

@XiaobaoZhu Can you provide your actual device information `gpuInfo()`? You need to understand that the number of cores in a GPU is not the same idea as number of CPUs....