Hi there,
Following the tutorial about writing C++/CUDA extensions I’m writing my own one. In this extension I use accessors. I’ve implemented the c++ version and everything work fine on CPU, but when I try to run my code (from the python level, as shown in the tutorial) on CUDA, I get a segfault raised by the attempt to access the elements of the accessors. Are accessors usable with CUDA tensors ?
Thank you.
Did you use packed_accessor for CUDA tensors? Could you share your C++/CUDA extension code?