Henry Robinson has posted comments on this change. Change subject: IMPALA-2076: Correct execution time tracking for DataStreamSender. ......................................................................
Patch Set 5: (14 comments) Looks pretty good to me. http://gerrit.cloudera.org:8080/#/c/2578/5/be/src/runtime/backend-client.h File be/src/runtime/backend-client.h: Line 66: /// Common cache / connection types remove this comment (there's only one cache / connection type here etc.) http://gerrit.cloudera.org:8080/#/c/2578/5/be/src/runtime/data-stream-test.cc File be/src/runtime/data-stream-test.cc: Line 630: client.DoRpc(&ImpalaBackendClient::TransmitData, params, &result); nit: one line? http://gerrit.cloudera.org:8080/#/c/2578/5/be/src/runtime/exec-env.cc File be/src/runtime/exec-env.cc: Line 200: new ImpalaBackendClientCache( one line? http://gerrit.cloudera.org:8080/#/c/2578/5/be/src/service/fragment-exec-state.cc File be/src/service/fragment-exec-state.cc: Line 118: coord.DoRpc(&ImpalaBackendClient::ReportExecStatus, params, &res); nit: one line? http://gerrit.cloudera.org:8080/#/c/2578/5/be/src/util/runtime-profile-test.cc File be/src/util/runtime-profile-test.cc: Line 541: : timercounter_(TUnit::TIME_NS) {} nit: one line? Line 548: NULL just for simplicity, why not assign this in the constructor? Then line 582 can do all the initialisation in one statement. Line 605: 20000000 why did this change so much? Line 605: MAX_TIMER_ERROR add units to the variable name (ns?) Line 622: TimerCounterTest& timer this should be a pointer if it's going to have non-const methods called on it http://gerrit.cloudera.org:8080/#/c/2578/5/be/src/util/runtime-profile.h File be/src/util/runtime-profile.h: Line 382: : Counter(unit) {} nit: one line? Line 392: uint64_t LapTime() { return csw_.LapTime(); } needs a comment http://gerrit.cloudera.org:8080/#/c/2578/5/be/src/util/stopwatch.h File be/src/util/stopwatch.h: Line 182: /// Utility class to measure multiple threads concurrent running time. I think it's important to say *wall* time here, otherwise users might expect this to return the sum of the wall times for each thread. Line 207: and reset lap_time_ : /// back to 0. can remove this bit of the comment: it's an implementation detail (the behaviour is captured by the first part of the sentence). Line 228: /// Lock with busy_threads_. Nit: put the lock above the thing it protects (easiest thing is to swap msw_ and busy_threads_). -- To view, visit http://gerrit.cloudera.org:8080/2578 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I9c530b2056d852c8bcac6263e9e6b1a6bede1047 Gerrit-PatchSet: 5 Gerrit-Project: Impala Gerrit-Branch: cdh5-trunk Gerrit-Owner: Juan Yu <[email protected]> Gerrit-Reviewer: Henry Robinson <[email protected]> Gerrit-Reviewer: Juan Yu <[email protected]> Gerrit-Reviewer: Sailesh Mukil <[email protected]> Gerrit-HasComments: Yes
