test=develop

test=develop

test=develop

test=develop

* move to platform

* "move init from framework to platform"

* "remove used init"

* "fix ci"

* "fix ci"

* "fix generic"

* "fix ci"

* "fix ci"

* "fix ci"

* "disable fragile test"

* Fix CPPLint issues in init

* Fix compilation

* Fix typo in init.cc

* Fix CPPLint issues in library_type.h

* Fix compilation in init.h

On k40 with 4 devices, time reduces from ~4.0 to ~3.8+, should be
more obvious on better hardware

* "fix gpu init"

* "set env variable default value for share gpu"

* "fix ci"

* "removed CUDA_VISIBLE_DEVICES default"

* "removed"

glog will not hold ARGV[0] inside.

* "init use all default devices"

* "fix init test"

* "add c++ side kernel selection"

* "add multiple kernel op test"

* "kernel selection only support cudnn"

* "better formatter"

* "small fix with UseCPU"

* "depends on change interface Get(Place, Library)"

* "fix CI"

* "fix python cudnn test"

* "leave the register cudnn op to another PR"

* "fix CI"

* "use all kernel by default"

* "fix CI"