Kirill Ignatiev

The message gets printed before the error in to_host, so that's cuda_ptrs just before the crash.

``` julia> Pkg.test("CUDArt") INFO: Testing CUDArt juliarc = "/***/.juliarc.jl" INFO: CUDArt tests passed INFO: No packages to install, update or remove ~ $ nvcc --version nvcc: NVIDIA (R) Cuda compiler...

In that example, the device is still initialized (i.e. cudaDeviceReset was not called, which is a real problem): ``` function isInitialized(dev) device(dev) try CUDArt.rt.cudaSetDeviceFlags(0) return false catch ex @show ex...

> There's a little explanation for the current behavior in the first paragraph of this section. To be a little more explicit, I settled on this design as a way...

Okay, here is an example where CUDArt is not reset correctly. I think this can be addressed by ignoring all errors in the finalizers in device_reset, because cudaFree and other...

No, I don't. If I understand correctly from documentation, cuLaunchKernel copies the arguments itself when those arguments are passed by value. For example, if you pass it a pointer to...

It knows the kernel function definition, and the definitions of types of its arguments, like a C compiler would, so that's how it knows how many bytes to copy and...

The allocation happens here: https://github.com/JuliaLang/julia/blob/c094a89b9bf5ae9dccbf5d85c3925d619cfe1dcf/base/io.jl#L766, because in order to read from a stream julia allocates a `Ref{Int32}`, and reads into it. Also, while `read(io,UInt8)` doesn't allocate, it is still much...

Maybe detect whether the initial position is empty or not (ignoring handicap stones)? It seems like an awkward special case, I could be missing something there.

The only think I could find was a similar issue in grpc (https://github.com/grpc/grpc-node/issues/1407), there are also some links there. It seems node at some point decided to no longer allow...