[
https://issues.apache.org/jira/browse/IMPALA-12134?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17723881#comment-17723881
]
ASF subversion and git services commented on IMPALA-12134:
----------------------------------------------------------
Commit 607ae742e5b12df12e2e2e64c890190a02beb765 in impala's branch
refs/heads/master from Kurt Deschler
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=607ae742e ]
IMPALA-12134: Optimize row materialization time
This patch improves row materialization time by providing specialized
formatting logic for default date and timestamp formats. For Beeswax
protocol, performance is also improved by caching deserialized column
metadata to avoid unnecessary per-row cost.
Benchmarks:
- Manually tested mixed datatype table showed ~20% reduction in row
materialization time
- Added cases to date-benchmark for new formatters. Date formatting
improved by 3x and timestamp by 2x
Machine Info: Intel(R) Core(TM) i7-8700 CPU @ 3.20GHz
ToYearMonthDay: Function iters/ms
-------------------------------------------------------
TestCctzToYearMonthDay 16.5
TestToYearMonthDay 61.1
TestToYear 280
TestToString 18
TestToString_stringstream 1.86
TestDefaultDateToCharBuf 25.5
TestTimestampToString 11.7
TestDefaultTimestampToCharBuf 15.7
Testing:
- Ran core tests
Change-Id: I1ef5e4137fa6c2d0a5f08b430e01e3fb7de86330
Reviewed-on: http://gerrit.cloudera.org:8080/19875
Reviewed-by: Impala Public Jenkins <[email protected]>
Tested-by: Impala Public Jenkins <[email protected]>
> Optimize row materialization time
> ---------------------------------
>
> Key: IMPALA-12134
> URL: https://issues.apache.org/jira/browse/IMPALA-12134
> Project: IMPALA
> Issue Type: Improvement
> Components: Backend
> Reporter: Kurt Deschler
> Assignee: Kurt Deschler
> Priority: Major
>
> IMPALA-12111 addressed the most significant contributors to slow row
> materialization. However, there is still room for significant improvement
> with the following optimizations:
> * Specialized implementation for default Data and Timestamp formatting.
> * Caching deserialized column metadata
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
