- It provides a unified style to express graph and kernel for user.
- It provides a unified IR to represent graph and kernel for developer.
- It breaks the boundary between graph and kernel.
- It provides more opportunities to do compile optimization.

Merge pull request !2396 from jjfeing/master

Merge pull request !2394 from kisnwang/add-cpu-outputsync

Merge pull request !2056 from zhoufeng/enable-new-control-sink

Signed-off-by: zhoufeng <zhoufeng54@huawei.com>

Merge pull request !2313 from caifubi/save-profiling-point

Merge pull request !2176 from chenfei_mindspore/code-review-of-session

Merge pull request !2361 from kisnwang/fix-codex-reviewbot

Merge pull request !1990 from caifubi/fix-profiling-stream-id

Merge pull request !2064 from caifubi/add-memory-context-var

Merge pull request !2285 from jjfeing/master

Merge pull request !2292 from limingqi107/master

Merge pull request !2280 from chenjianping/host_reduce

Merge pull request !2222 from wangcong/master

Merge pull request !2247 from yanghaoran/master

Merge pull request !2164 from lianliguang/master

Merge pull request !2210 from limingqi107/max_device_memory_optimize

Merge pull request !2183 from limingqi107/gpu_pynative_optimize

Merge pull request !2215 from caifubi/support-hccl-profiling

Merge pull request !2208 from sunsuodong/review

Merge pull request !2154 from caifubi/fix-profiling-core-dump

Merge pull request !2174 from kisnwang/fix-cpu-reshape-bug

Merge pull request !2125 from lianliguang/master

Merge pull request !2101 from ZPaC/code-review-1

Merge pull request !2130 from limingqi107/master

Merge pull request !2051 from chenjianping/host_reduce