Kwafoor commented on issue #5467: URL: https://github.com/apache/kyuubi/issues/5467#issuecomment-1831229505
> for testing, Kyuubi provides spark tpc-h/ds connector and queries, would be good cases When I conducted the TPCH tests using Gluten, I identified two issues. Due to these, I am considering segregating Kyuubi Gluten TPC-H/DS into a separate issue. Would this be acceptable? @pan3793 1. While testing for `tpch.tiny` data, a discrepancy in the query result of q9 gluten+velox+spark3.4 in comparison to spark3.4. Gluten was `[SAUDI ARABIA,1995,62668.73]`, Spark3.4 was `[SAUDI ARABIA,1995,62668.72]`. 2. Gluten+velox+spark3.4 was observed to execute slower than spark3.4. I'd have to further explore to understand the specific reasons behind this. Here are the related logs. ```log ## gluten+velox+spark3.4 03:43:56.634 broadcast-exchange-30 INFO DAGScheduler: Job 73 finished: collect at SparkPlanExecApiImpl.scala:284, took 6.998831 s 03:43:59.349 broadcast-exchange-28 INFO DAGScheduler: Job 74 finished: collect at SparkPlanExecApiImpl.scala:284, took 9.710332 s 03:44:19.072 broadcast-exchange-31 INFO DAGScheduler: Job 75 finished: collect at SparkPlanExecApiImpl.scala:284, took 29.406944 s ``` ```log ## spark3.4 09:48:41.929 broadcast-exchange-29 INFO DAGScheduler: Job 71 finished: $anonfun$withThreadLocalCaptured$1 at FutureTask.java:266, took 0.054058 s 09:48:41.946 broadcast-exchange-28 INFO DAGScheduler: Job 72 finished: $anonfun$withThreadLocalCaptured$1 at FutureTask.java:266, took 0.067240 s 09:48:41.974 broadcast-exchange-30 INFO DAGScheduler: Job 73 finished: $anonfun$withThreadLocalCaptured$1 at FutureTask.java:266, took 0.092720 s 09:48:42.013 broadcast-exchange-32 INFO DAGScheduler: Job 75 finished: $anonfun$withThreadLocalCaptured$1 at FutureTask.java:266, took 0.130110 s 09:48:42.019 broadcast-exchange-31 INFO DAGScheduler: Job 74 finished: $anonfun$withThreadLocalCaptured$1 at FutureTask.java:266, took 0.137744 s ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
