Bulk Operation

IMPALA-3902 Parallelise flush in data stream sender

IMPALA-3902 Add general mechanism to find DataSink from other fragments

IMPALA-3902 testMtDopValidationWithHDFSNumRowsEstDisabled toggles isTestEnv() changing table loading behaviour

IMPALA-3902 Fix cancellation of RuntimeFilter::WaitForArrival()

IMPALA-3902 Update planner decisions to factor in mt_dop

IMPALA-3902 Fix runtime filter bugs with mt_dop

IMPALA-3902 Use better algorithm for allocating scan ranges to finstances within a daemon in schedule

IMPALA-3902 Parallelise all plans, including UNION, when mt_dop > 1

IMPALA-3902 Admission control accounting for mt_dop

IMPALA-3902 Allow graceful fallback to mt_dop=0

IMPALA-3902 Experimental flag for running all queries with mt_dop

IMPALA-3902 COMPUTE STATS uses MT_DOP=4 by default

IMPALA-3902 ReleaseResources() should not destroy control structures

IMPALA-3902 Single-threaded KuduScanNode

IMPALA-3902 Adjust maximum size of row batch queue for non-Parquet scans with MT_DOP>0.

IMPALA-3902 COMPUTE STATS on Parquet tables uses MT_DOP=4 by default

IMPALA-3902 Standardize on MT-related data structures in the coordinator and scheduler

IMPALA-3902 Planner should disallow queries with mt_dop > 0 that are not executable.

IMPALA-3902 Create tests for multi-threaded query execution

IMPALA-3902 Introduce query-wide execution state.

IMPALA-3902 Generate codegen module without operator/query-specific constants

IMPALA-3902 Single-threaded scan node

IMPALA-3902 Scheduler improvements for running multiple fragment instances on a single backend

IMPALA-3902 Factor out join build-side

IMPALA-3902 Switch to per-query exec rpc