Create Vec based on Managed Memory

[kpp]

I named it MVec (M stands for Managed wheres U stands for Uniformed (I guess so))

[termoshtt]

Manged memory reserved by cudaMalloc cannot be accessed from host code since it is allocated on the device. I agree to create an interface for managed memory, but we should not implement Derefand Index for it. I think the approach by thrust, i.e. creating host_vector and device_vector will be better. Your MVec seems to corresponds to device_vector. Could you create memcpy wrapper from Vec to MVec?

[kpp]

Sure! I am working on it.

[kpp]

---- uvec::tests::uvec_new stdout ----
	thread 'uvec::tests::uvec_new' panicked at 'called `Result::unwrap()` on an `Err` value: cudaRuntimeError(InsufficientDriver)', libcore/result.rs:945:5

seems like you did not set up your CI and build env correctly. I can write the code but I can't check it locally.

[termoshtt]

seems like you did not set up your CI and build env correctly. I can write the code but I can't check it locally.

CI with GPU is not running currently #9

[kpp]

Could you create memcpy wrapper from Vec to MVec?

done

[kpp]

@termoshtt would you merge it?

[termoshtt]

Sorry for later replay. I will merge it after I confirm it works on my env.

[termoshtt]

Fix host vector init, and it works well. Thanks :)