jiangxin369 commented on code in PR #250:
URL: https://github.com/apache/flink-ml/pull/250#discussion_r1292699290
##########
flink-ml-iteration/flink-ml-iteration-common/src/main/java/org/apache/flink/iteration/datacache/nonkeyed/ListStateWithCache.java:
##########
@@ -62,14 +63,32 @@ public class ListStateWithCache<T> implements ListState<T> {
/** The data cache writer for the received records. */
private final DataCacheWriter<T> dataCacheWriter;
- @SuppressWarnings("unchecked")
public ListStateWithCache(
TypeSerializer<T> serializer,
StreamTask<?, ?> containingTask,
StreamingRuntimeContext runtimeContext,
StateInitializationContext stateInitializationContext,
OperatorID operatorID)
throws IOException {
+ this(
+ serializer,
+ containingTask,
+ runtimeContext,
+ stateInitializationContext,
+ operatorID,
+ 0.);
Review Comment:
With this constructor, the current algorithms using `ListStateWithCache`
like `Kmeans` would not use managed memory at all, which causes performance
regression.
Would it be better if we divide the memory evenly if multiple
`ListStateWithCache` is declared by default?
--
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]
