[
https://issues.apache.org/jira/browse/CALCITE-1012?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15049372#comment-15049372
]
Vladimir Sitnikov edited comment on CALCITE-1012 at 12/9/15 9:04 PM:
---------------------------------------------------------------------
Here's quick pick on "reuse" approach (exactly the same SQL):
{noformat}
Benchmark (comments) (length) Mode
Cnt Score Error Units
ParserBenchmark.parseCached true 1000 avgt
7 132,824 ± 5,348 us/op
ParserBenchmark.parseCached:·gc.alloc.rate.norm true 1000 avgt
7 51265,006 ± 5,617 B/op
ParserBenchmark.parseNonCached true 1000 avgt
7 146,007 ± 14,061 us/op
ParserBenchmark.parseNonCached:·gc.alloc.rate.norm true 1000 avgt
7 97756,885 ± 112,912 B/op
{noformat}
It does improve performance a bit (previous "degradation" was due to different
SQLs generated for different runs)
was (Author: vladimirsitnikov):
Here's quick pick on "reuse" approach:
{noformat}
Benchmark (comments) (length) Mode
Cnt Score Error Units
ParserBenchmark.parseCached true 1000 avgt
7 125,794 ± 6,069 us/op
ParserBenchmark.parseCached:·gc.alloc.rate.norm true 1000 avgt
7 51872,058 ± 0,003 B/op
ParserBenchmark.parseNonCached true 1000 avgt
7 109,976 ± 3,401 us/op
ParserBenchmark.parseNonCached:·gc.alloc.rate.norm true 1000 avgt
7 96102,688 ± 89,184 B/op
{noformat}
Less allocations, but *worse* performance.
> Benchmark SQL parser
> --------------------
>
> Key: CALCITE-1012
> URL: https://issues.apache.org/jira/browse/CALCITE-1012
> Project: Calcite
> Issue Type: Bug
> Reporter: Julian Hyde
> Assignee: Julian Hyde
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
