* Support Sync Batch Norm.
* Note, do not enable it in one device.

Usage:

build_strategy = fluid.BuildStrategy()
build_strategy.sync_batch_norm = True
binary = fluid.compiler.CompiledProgram(tp).with_data_parallel(
        loss_name=loss_mean.name,
        build_strategy=build_strategy)

* Impl fp16 compute kernel for slice_op

test=develop

* Use data() to replace mutable_data()

Fix cross_entropy2_op numeric error

test=develop

test=develop

Revert "PaddingRNN model memory optimize"

test=develop

test=develop

test=develop

2. remove lod check in reshape op
test=develop

* Add cpu_quantize_squash_pass for C-API quantization

test=develop

* add cpu_quantize_squash_pass teste

* fix lint: add include memory unorderd_map and unordered_set

test=develop

* lint fix 2

* fixes

test=develop

* refactored

test=develop

* fix windows ci

test=develop

PaddingRNN model memory optimize

* make timeline.py compatible
test=develop

* Follow comment
test=develop

add all_kernels_must_compute_runtime_shape example for speedup infershape

Fix api.spec

test=develop

* resolve #15618
Backgroud: the PR #15398 raised the box_coder op performance regression, we optimized the code via the more efficency leveraging opemmp.

test=develop

Add the inverse trigonometric function

* fix broadcast with mp mode

* polish code test=develop

* fix bcast strategy test=develop

* fic cpplint test=develop

* fix py3 failed test=develop

* fix comment test=develop

* update comment test=develop

* remove concast_cast and refactor code test=develop

* reduce flag use test=develop

test=develop

test=develop

Fix NaN bugs for static quantization strategy (mutil-cards train). 

test=develop

Imperative transfer gru unit

Add clone function for IrGraph.

test=develop

Add channel wise quant op and channel wise dequant op.

test=develop

test=develop