iverase commented on a change in pull request #569: LUCENE-8687: Optimise radix
partitioning for points on heap
URL: https://github.com/apache/lucene-solr/pull/569#discussion_r255045631
##########
File path:
lucene/core/src/java/org/apache/lucene/util/bkd/BKDRadixSelector.java
##########
@@ -294,18 +335,37 @@ protected int byteAt(int i, int k) {
}
}.select(from, to, partitionPoint);
- for (int i = from; i < to; i++) {
- points.getPackedValueSlice(i, bytesRef1);
- int docID = points.docIDs[i];
- if (i < partitionPoint) {
- left.append(bytesRef1, docID);
- } else {
- right.append(bytesRef1, docID);
- }
- }
byte[] partition = new byte[bytesPerDim];
points.getPackedValueSlice(partitionPoint, bytesRef1);
System.arraycopy(bytesRef1.bytes, bytesRef1.offset + dim * bytesPerDim,
partition, 0, bytesPerDim);
return partition;
}
+
+ PointWriter getPointWriter(long count, String desc) throws IOException {
+ if (count <= maxPointsSortInHeap / 2) {
Review comment:
We divided by 2 because as we recurse, we have two `heapPointWriter` as
maximum, so they should not hold more than half of the `maxPointsSortInHeap`.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
