Accidentally I have stumbled into a change in Java 15 that I was not aware of. Suppose I have a very simple question: what is the size of an array of 3 integers? For this, I use JOL. The code is fairly trivial:
import org.openjdk.jol.info.ClassLayout;
import org.openjdk.jol.vm.VM;
public class Array {
public static void main(String [] args){
int [] array = new int[3];
System.out.println(ClassLayout.parseInstance(array).toPrintable());
}
}
I run this with Java 13:
java -Djdk.attach.allowAttachSelf -Djol.tryWithSudo=true -cp jol-cli.jar Array.java
I get the output:
[I object internals:
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 01 00 00 00 (00000001 00000000 00000000 00000000) (1)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 18 0e 07 00 (00011000 00001110 00000111 00000000) (462360)
12 4 (object header) 03 00 00 00 (00000011 00000000 00000000 00000000) (3)
16 12 int [I.<elements> N/A
28 4 (loss due to the next object alignment)
Instance size: 32 bytes
Space losses: 0 bytes internal + 4 bytes external = 4 bytes total
This is pretty much obvious:
12 bytes --> Object headers
4 bytes --> size of array
12 bytes --> elements of array themselves
4 bytes --> padding to align by 8 bytes
----
32 bytes total
Running this example with Java 15 yields the same output, same 32 bytes
. Expected...
For the second part, I want to disable a JVM optimization: -XX:-UseCompressedOops
. I run this with Java 13:
java -Djdk.attach.allowAttachSelf -Djol.tryWithSudo=true -cp jol-cli.jar -XX:-UseCompressedOops Array.java
[I object internals:
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 11 00 00 00 (00010001 00000000 00000000 00000000) (17)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 40 0c f0 33 (01000000 00001100 11110000 00110011) (871369792)
12 4 (object header) 02 00 00 00 (00000010 00000000 00000000 00000000) (2)
16 4 (object header) 03 00 00 00 (00000011 00000000 00000000 00000000) (3)
20 4 (alignment/padding gap)
24 12 int [I.<elements> N/A
36 4 (loss due to the next object alignment)
Instance size: 40 bytes
Space losses: 4 bytes internal + 4 bytes external = 8 bytes total
Well, sort of expected too:
16 bytes --> object headers (I did -XX:-UseCompressedOops after all)
4 bytes --> array size
4 bytes --> alignment for array headers (AFAIK this is only done for arrays)
12 bytes --> array elements themselves
4 bytes --> 4 bytes padding
----
40 bytes total
Now let's run the same example with Java 15:
[I object internals:
OFFSET SIZE TYPE DESCRIPTION VALUE
0 4 (object header) 01 00 00 00 (00000001 00000000 00000000 00000000) (1)
4 4 (object header) 00 00 00 00 (00000000 00000000 00000000 00000000) (0)
8 4 (object header) 0e 09 00 00 (00001110 00001001 00000000 00000000) (2318)
12 4 (object header) 03 00 00 00 (00000011 00000000 00000000 00000000) (3)
16 12 int [I.<elements> N/A
28 4 (loss due to the next object alignment)
Instance size: 32 bytes
Why is this 32 bytes
now? How come not 40
, just like with Java 13?
In both Java 13 and Java 15, both of these options are on by default:
When
-XX:-UseCompressedOops
is disabled, it means thatUseCompressedClassPointers
is disabled also. That is why whenUseCompressedOops
is turned off, the header size goes up by4 bytes
, becauseUseCompressedOops
turns offUseCompressedClassPointers
. At least this is how it is in Java 13:Things have changed in Java 15:
So disabling
UseCompressedOops
does not mean thatUseCompressedClassPointers
is disabled also, so it stays at4 bytes
.Though, I answered this myself, it would be nice if someone finds the relevant bug/change for this? I have not been successful in that, so far.