[
https://issues.apache.org/jira/browse/IGNITE-7931?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ivan Fedotov updated IGNITE-7931:
---------------------------------
Description:
When creating ConcurrentHashMap:
{code:java|title=BinaryUtils.java}
new ConcurrentHashMap<>(U.capacity((map).size()));{code}
[1], [2] `U.capacity` returns capacity that is sufficient to keep the map from
being resized. In ConcurrentHashMap this is unnecessary because recalculation
already performs in the constructor.
Similar problem in GridConcurrentHashSet because inside it implements
ConcurrentHashMap:
{code:java|title=DataStreamerImpl.java}
keys = new GridConcurrentHashSet<>(entries.size(), U.capacity(entries.size()),
1);{code}
[3],[4] result of `U.capacity` is passed as `loadFactor` value. When loadFactor
== U.capacity, initial size of table is 1.
This leads to performance penalty due to rehashing of internal map.
[1[https://github.com/1vanan/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/binary/BinaryUtils.java#L670]
[2][https://github.com/1vanan/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/binary/BinaryUtils.java#L688|https://github.com/1vanan/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/binary/BinaryUtils.java#L670]
[3][https://github.com/apache/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/processors/datastreamer/DataStreamerImpl.java#L633]
[4][https://github.com/apache/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/processors/datastreamer/DataStreamerImpl.java#L574]
was:
When creating
{code:java|title=DataStreamerImpl.java}
keys = new GridConcurrentHashSet<>(entries.size(), U.capacity(entries.size()),
1);{code}
[1],[2] result of `U.capacity` is passed as `loadFactor` value. When loadFactor
== U.capacity, initial size of table is 1. This leads to performance penalty
due to rehashing of internal map.
[1][https://github.com/apache/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/processors/datastreamer/DataStreamerImpl.java#L633]
[2][https://github.com/apache/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/processors/datastreamer/DataStreamerImpl.java#L574]
> Wrong arguments for ConcurrentHashMap
> --------------------------------------
>
> Key: IGNITE-7931
> URL: https://issues.apache.org/jira/browse/IGNITE-7931
> Project: Ignite
> Issue Type: Improvement
> Reporter: Ivan Fedotov
> Assignee: Ivan Fedotov
> Priority: Trivial
> Fix For: 2.5
>
>
> When creating ConcurrentHashMap:
> {code:java|title=BinaryUtils.java}
> new ConcurrentHashMap<>(U.capacity((map).size()));{code}
> [1], [2] `U.capacity` returns capacity that is sufficient to keep the map
> from being resized. In ConcurrentHashMap this is unnecessary because
> recalculation already performs in the constructor.
> Similar problem in GridConcurrentHashSet because inside it implements
> ConcurrentHashMap:
> {code:java|title=DataStreamerImpl.java}
> keys = new GridConcurrentHashSet<>(entries.size(),
> U.capacity(entries.size()), 1);{code}
> [3],[4] result of `U.capacity` is passed as `loadFactor` value. When
> loadFactor == U.capacity, initial size of table is 1.
> This leads to performance penalty due to rehashing of internal map.
>
> [1[https://github.com/1vanan/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/binary/BinaryUtils.java#L670]
> [2][https://github.com/1vanan/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/binary/BinaryUtils.java#L688|https://github.com/1vanan/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/binary/BinaryUtils.java#L670]
> [3][https://github.com/apache/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/processors/datastreamer/DataStreamerImpl.java#L633]
> [4][https://github.com/apache/ignite/blob/master/modules/core/src/main/java/org/apache/ignite/internal/processors/datastreamer/DataStreamerImpl.java#L574]
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
