Tim Armstrong has posted comments on this change. ( http://gerrit.cloudera.org:8080/9239 )
Change subject: IMPALA-6258: Uninitialized tuple pointers in row batch for empty rows ...................................................................... Patch Set 2: (1 comment) http://gerrit.cloudera.org:8080/#/c/9239/2//COMMIT_MSG Commit Message: http://gerrit.cloudera.org:8080/#/c/9239/2//COMMIT_MSG@57 PS2, Line 57: 00:SCAN KUDU 3 90.856ms 107.409ms 6.00M 6.00M 512.00 KB 0 tpch_kudu.lineitem I tried to do a similar experiment with a larger Kudu scale factor (I created a new Kudu table like lineitem and expanded it by inserting duplicate data): insert into biglineitem select l_orderkey + max_orderkey, l_partkey, l_suppkey, l_linenumber, l_quantity, l_extendedprice, l_discount, l_tax, l_returnflag, l_linestatus, l_shipdate, l_commitdate, l_receiptdate, l_shipinstruct, l_shipmode, l_comment from biglineitem, (select max(l_orderkey) as max_orderkey from biglineitem) v I can definitely see some time spent in the HandleEmptyProjection() function in "perf top" but the delta in performance seems smaller than your experiment showed. I saw it around 5% slower. The count(*) optimisation sounds good but not sure if the regression is severe enough to block this going in. Maybe Thomas can weigh in on how important he thinks this is. -- To view, visit http://gerrit.cloudera.org:8080/9239 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I298122aaaa7e62eb5971508e0698e189519755de Gerrit-Change-Number: 9239 Gerrit-PatchSet: 2 Gerrit-Owner: Zoltan Borok-Nagy <borokna...@cloudera.com> Gerrit-Reviewer: Tim Armstrong <tarmstr...@cloudera.com> Gerrit-Reviewer: Zoltan Borok-Nagy <borokna...@cloudera.com> Gerrit-Comment-Date: Fri, 09 Feb 2018 23:22:38 +0000 Gerrit-HasComments: Yes