* fix bug: wrong match between depend and c_allreduce_sum

(cherry picked from commit 327da8035bdfee3ec2f016e8cda29ec8ee89bc95)

* fix codestyle

(cherry picked from commit bdb1483081adc41aa47d3f7df257f63f1cff399b)

* fix bug

(cherry picked from commit 373ba5253c45ac019ffaa8d69d4ce9e02cb9ae79)

* add c_sync_calc_stream back

(cherry picked from commit 9933d7533ae1f307b76f24a33bf0c59e4c8e8f01)

* fix

(cherry picked from commit abc9a31beaa326f6a566c08749419bb33e209672)

* revert

(cherry picked from commit 07bc98dbf7c9df43910fa6e86a6a2698731dffb2)

* use flag to control

(cherry picked from commit 8e5682a4b99759cbe35a49f3f8c9db735dc8fee4)

* fix for code coverage

(cherry picked from commit fe7e61bdef24fbc43e2f4e1cb67f68963c957cf1)