possibly improve lambda_iterator_t
christiankerl opened this issue · comments
Christian Kerl commented
is there a reason why make_{load_store,load,store}_iterator are not callable from device code and on the contrary why assign_t operators are not callable from host code? I would find it useful to have this functionality.
Sean Baxter commented
On CUDA 7.5 you couldn't host-device tag a lambda. I recall there being issues if you tried to host-device tag the lambda_iterator_t accessor and it called a lambda. I'll take another look and see if this annoyance can be ironed out.