MarcusSorealheis commented on code in PR #12254:
URL: https://github.com/apache/lucene/pull/12254#discussion_r1182114316
##
lucene/core/src/java/org/apache/lucene/util/hnsw/ConcurrentOnHeapHnswGraph.java:
##
@@ -0,0 +1,280 @@
+/*
+ * Licensed to the Apache Software Foundation (A
MarcusSorealheis commented on code in PR #12254:
URL: https://github.com/apache/lucene/pull/12254#discussion_r1182114316
##
lucene/core/src/java/org/apache/lucene/util/hnsw/ConcurrentOnHeapHnswGraph.java:
##
@@ -0,0 +1,280 @@
+/*
+ * Licensed to the Apache Software Foundation (A
MarcusSorealheis commented on code in PR #12254:
URL: https://github.com/apache/lucene/pull/12254#discussion_r1182114316
##
lucene/core/src/java/org/apache/lucene/util/hnsw/ConcurrentOnHeapHnswGraph.java:
##
@@ -0,0 +1,280 @@
+/*
+ * Licensed to the Apache Software Foundation (A
jbellis commented on PR #12254:
URL: https://github.com/apache/lucene/pull/12254#issuecomment-1530167402
thanks again.
and you're right, the ram usage needed a rewrite, it is passing now with the
seed you found.
--
This is an automated message from the Apache Git Service.
To respon
tang-hi commented on PR #12254:
URL: https://github.com/apache/lucene/pull/12254#issuecomment-1530005788
> Thanks! For me it is working:
>
> ```
> $ ./gradlew -p lucene/core test --tests "*Hnsw*"
> ...
> :lucene:core:test (SUCCESS): 75 test(s)
> ```
>
> Note that add
jbellis commented on PR #12254:
URL: https://github.com/apache/lucene/pull/12254#issuecomment-1529956854
Thanks! For me it is working:
```
$ ./gradlew -p lucene/core test --tests "*Hnsw*"
...
:lucene:core:test (SUCCESS): 75 test(s)
```
Note that adding these args do
tang-hi commented on PR #12254:
URL: https://github.com/apache/lucene/pull/12254#issuecomment-1529914311
> I've made a couple optimizations and merged to main.
>
> With the optimizations it's still about 38% slower than the non-concurrent
code (including the HashMap optimization for t
jbellis commented on PR #12254:
URL: https://github.com/apache/lucene/pull/12254#issuecomment-1529909165
Replaced parallel stream with ExecutorService. Only ThreadPoolExecutor is
supported, since it knows how many thread it has available. After a quick look
that is the only type I see Luc
jbellis commented on PR #12254:
URL: https://github.com/apache/lucene/pull/12254#issuecomment-1529796265
I've made a couple optimizations and merged to main.
With the optimizations it's still about 38% slower than the non-concurrent
code (including the HashMap optimization for the non
