I’m sorry I can not upload those files due to IP issue. I’ll manipulate the file and upload later if necessary.
One reason I suspect is that my code is too long ( more than 6000 lines, and some inline function is more than 3300 lines ), I’m afraid this is the reason. After I commented out most of the code, it can pass. I’ll continue to recover the code to see where is the problem.
Could you share some information on possible limit of code length that current CUDA can handle?
Meanwhile, can we get any information from the following assertion failure:
Assertion failure at line 848 of …/…/be/cg/NVISA/exp_loadstore.cxx
I think this was from nvcc code.
Thank you, Mark