mixed-precision-programing
programing with gpu on low precision methods in cuda, cublas, cudnn and so on.
2018.05.14：(finished)
#add p100/v100/p4 and tensorcore on/off using low precision programing test cases with cublas.
2018.05.21:(finished)
#add v100/p4 low precision(int8) programing test cases with cudnn.
2018.06.09:(finished)
#add v100/p100(fp16) low precision programing test cases with cudnn.