Optimze Gelu with MKL Erf function#15770

Merged

tensor-tang merged 10 commits intoPaddlePaddle:developfrom

yihuaxu:develop_a6910f900_gelu_mkl_opt

Feb 22, 2019

Commits on Jan 28, 2019

Optimize for gelu operator
yihuaxu
committed

Commits on Jan 31, 2019

Set up the low accuracy mode of MKL ERF function.
yihuaxu
committed

Commits on Feb 14, 2019

Only enable MKLML ERF when OS is linux
yihuaxu
committed

Commits on Feb 18, 2019

Use the speical mklml version included vmsErf function to verify gelu mkl kernel.
yihuaxu
committed
Merge branch 'develop' into develop_a6910f900_gelu_mkl_opt
yihuaxu
committed

Commits on Feb 19, 2019

Commits on Feb 20, 2019

Commits on Feb 22, 2019

Merge branch 'develop' into develop_a6910f900_gelu_mkl_opt
yihuaxu
committed