[
https://issues.apache.org/jira/browse/CALCITE-2847?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16767230#comment-16767230
]
Hongze Zhang edited comment on CALCITE-2847 at 2/13/19 2:16 PM:
----------------------------------------------------------------
A PR opened at: [https://github.com/apache/calcite/pull/1041]
I've run a more precise benchmark (by increasing the "warm up iteration" and
"measurement iteration" from 7 to 50 in
[ParserBenchmark.java|https://github.com/apache/calcite/blob/2102f1f5442fa271c258b7754da8df07d65847ec/ubenchmark/src/main/java/org/apache/calcite/benchmarks/ParserBenchmark.java#L48]),
and the result seems to be positive too. ParserBenchmark.parseCached has been
reduced from 1500.814 ± 224.844 us/op[1] to 179.250 ± 6.004 us/op[2].
[1] [https://www.dropbox.com/s/euit7gjxqyxihm4/bench_la2_50itr.log?dl=0]
[2] [https://www.dropbox.com/s/2ka88g3k6yy9pvs/bench_la1_50itr.log?dl=0]
was (Author: zhztheplayer):
A PR opened at: [https://github.com/apache/calcite/pull/1041]
I've run a more precise benchmark (by increasing the "warm up iteration" and
"measurement iteration" from 7 to 50 in
[ParserBenchmark.java|https://github.com/apache/calcite/blob/2102f1f5442fa271c258b7754da8df07d65847ec/ubenchmark/src/main/java/org/apache/calcite/benchmarks/ParserBenchmark.java#L48]),
and the result seems to be positive too. ParserBenchmark.parseCached has been
reduced from 1500.814 ± 224.844 us/op[1] to 179.250 ± 6.004 us/op[2].
[1] [https://www.dropbox.com/s/2ka88g3k6yy9pvs/bench_la1_50itr.log?dl=0]
[2] [https://www.dropbox.com/s/euit7gjxqyxihm4/bench_la2_50itr.log?dl=0]
> Optimize global LOOKAHEAD for SQL parsers
> -----------------------------------------
>
> Key: CALCITE-2847
> URL: https://issues.apache.org/jira/browse/CALCITE-2847
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Reporter: Hongze Zhang
> Assignee: Julian Hyde
> Priority: Major
> Labels: pull-request-available
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Currently global LOOKAHEAD is set to 2 for the built-in SQL parsers[1], we'd
> like to optimize LOOKAHEAD to 1 to enable performance enhancement if possible.
> [1]https://github.com/apache/calcite/blob/883666929478aabe07ee5b9e572c43a6f1a703e2/core/pom.xml#L304
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
