Forgive me for being very new to cuda. I started about a week and a half ago. I am learning with a goal in mind so forgive me if I am trying to do something complicated and do not know some of the basics required.
I want take an array of pointers (I need the host to be pointers for the dynamic memory. I do not know the size of the array at the time of compilation because I am reading data from a file) and then put it into a global device variable for access. Once it is in device memory the size will not change but the values will.
I had hoped I could do this like any other global device variable using cudaMemcpyToSymbol but I get invalid argument so I asssume that multidimensional arrays require a different approach. I have also read somethings about flattening the array but that is not a practical option for me because I must compare patches of data instead of just individual values. They were also from 7 years ago so I dont know if support for copying and traversing multidimensional arrays has been implemented or not. (https://devtalk.nvidia.com/default/topic/401967/how-can-i-allocate-2-dimensional-array-on-the-device-memory-/)
What would the code look like for this? Is this possible in cuda? Is there something I should go back and study? Is there a good tutorial for this?