Cristóvão Duarte Sousa

@barche, can you share the version using CUDAdrv.jl, please?

Thanks @barche , but I supposed you had converted the example of "Unified Memory example" to CUDAdrv... I have been trying it myself but I'm getting errors when calling `cuMemPrefetchAsync`...

Ah, thank you very much @barche, that's exactly what I was looking for. :+1: Unfortunately, it is more or less what I was trying and thus throws the same error...

Ok, probably it is because ["Maxwell architectures [..] support a more limited form of Unified Memory"](https://devblogs.nvidia.com/parallelforall/unified-memory-cuda-beginners/) :man_facepalming:. I'll try in a Jetson TX2. Thank you both.

Hi @maleadt and others, I further develop and then packed all the code I had in gists (comments above) into a package that can be easily tested. However, the intention...

Ah, no, it was not my intent to push this into CUDA.jl 😄 I just mentioned it here due to the original request. I say it is not a real...

Hi all, is the problem present in the following code related to this issue? ```julia using StaticArrays, ColorTypes import ForwardDiff, ForwardDiff2, Zygote ForwardDiff.gradient(x -> RGB(x[1], 0.5, 0.5).r, SVector(0.0))[1] ########## Wrong:...

Can I open an issue about making ColorTypes forward diff / dual friendly?

Or maybe this is something that can be workable from ForwardDiff2 with the source-to-source techniques (as this works with Zygote.jl)?