Thx for your awesome work! 
what do i need in order to compile it with compute 3 and 3.5 support?
what do i need in order to compile it with compute 3 and 3.5 support?
TPRUVOT--
You need another branch besides sp_'s version. Tpruvot has a recent and popular branch that maintains compatiblity with v3.5 cards. There are other branches, also.
The code written by sp_ is for Maxwell cards, v5.0 snd 5.2 of CUDA architecture. --scryptr