* Optimize CRF decoding with AVX/AVX2 instruction

* Enable the AVX2 flags for compiling

* Clean the code and decrease the count of multiply calculation

* Add the support of AVX512 instruction to optimize CRF Decoding

* Clean the code

* Enable the AVX512f flags for compiling

* Clean the code for the invaluable switch

* Fixed the issue to check AVX512F status

* Clean the code

* Add some explanation of the key points

* cudnn widndows

* "add comment"

* "windows support"

* "fix cmake error"

* Fix bug in conditional_block_op.
* Fix bug and add comments.
* Rename arguments.

Feature/fast executor

fix ProgramToGraph

disable anakin when cuda < 8.0 or cudnn < 7.0

Remove invalid link of use_eigen_doc

sum_op selectedRows dim bug fix

modify tensorrt engine op from cpu mode to gpu 

Add support for global pooling for trt

when while_grad, it writes multiple @EMPTY@ with no VarDesc.

Add dependency to send recv

small fix

Remove unused six and missing six import

* "fix bug"

* "add test case"

fix the paddle script causes 'command not found' error'

Move matplotlib dependencies to MANYLINUX1_DEPS to avoid erase

* add rw lock test

* optimize read_write and wirte_read test

This reverts commit bf3c3496.

Optimize selected rows for dist lookup table with rwlock