Github user jlhitt commented on a diff in the pull request:
https://github.com/apache/spark/pull/14762#discussion_r78985465
--- Diff:
common/unsafe/src/main/java/org/apache/spark/unsafe/array/ByteArrayMethods.java
---
@@ -47,13 +47,20 @@ public static int roundNumberOfBytesToNearestWord(int
numBytes) {
public static boolean arrayEquals(
Object leftBase, long leftOffset, Object rightBase, long
rightOffset, final long length) {
int i = 0;
- while (i <= length - 8) {
- if (Platform.getLong(leftBase, leftOffset + i) !=
- Platform.getLong(rightBase, rightOffset + i)) {
- return false;
- }
- i += 8;
- }
+
+ // This attempts to speed up the memcmp type of operation, but there
is no way
+ // to guarantee that the offsets will be on a word boundary in order
to use
+ // Platform.getLong
--- End diff --
It should happen 1/8 of the time for left, and 1/8 of the time for right,
and those times have to match up at the same time, so 1/64, and I don't believe
it actually happens that often. We can leave the current behavior for
non-SPARC if you want.
Just FYI though, even though it "works" on non-SPARC, it is still slower
than aligned accesses as manipulations are going on under the hood to make it
"work".
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
