Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/21593#discussion_r196637144
--- Diff:
common/network-common/src/main/java/org/apache/spark/network/protocol/MessageWithHeader.java
---
@@ -137,30 +137,15 @@ protected void deallocate() {
}
private int copyByteBuf(ByteBuf buf, WritableByteChannel target) throws
IOException {
- ByteBuffer buffer = buf.nioBuffer();
- int written = (buffer.remaining() <= NIO_BUFFER_LIMIT) ?
- target.write(buffer) : writeNioBuffer(target, buffer);
+ // SPARK-24578: cap the sub-region's size of returned nio buffer to
improve the performance
+ // for the case that the passed-in buffer has too many components.
+ int length = Math.min(buf.readableBytes(), NIO_BUFFER_LIMIT);
+ ByteBuffer buffer = buf.nioBuffer(buf.readerIndex(), length);
--- End diff --
I think you can go one step further here, and call `buf.nioBuffers(int,
int)` (plural)
https://github.com/netty/netty/blob/4.1/buffer/src/main/java/io/netty/buffer/ByteBuf.java#L2355
that will avoid the copying required to create the merged buffer (though
its a bit complicated as you have to check for incomplete writes from any
single `target.write()` call).
Also OK to leave this for now as this is a pretty important fix.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
