JDK 14 RFR of JDK-8199424: consider removing ObjectInputStream and ObjectOutputStream native methods

Claes Redestad claes.redestad at oracle.com
Fri Sep 20 04:32:38 UTC 2019


On 2019-09-20 00:34, Joe Darcy wrote:
> I have not attempted to benchmark any speed differences with and without 
> the cleanup, but would welcome such an experiment being done.

I ran a quick, naive test by modifying an existing serialization
micro[1], and results are in the noise on default/tiered and C1
(-XX:TieredStopAtLevel=1) while there seems to be a 5-10% regression
running interpreter only (not verified to be statistically
significant), which is somewhat expected but inconsequential.

So no reason to worry AFAICT - ship it!

/Claes

[1]
diff -r 3386b9a8ef4d 
test/micro/org/openjdk/bench/java/io/SerializationWriteReplace.java
--- 
a/test/micro/org/openjdk/bench/java/io/SerializationWriteReplace.java 
Thu Sep 19 23:38:50 2019 +0200
+++ 
b/test/micro/org/openjdk/bench/java/io/SerializationWriteReplace.java 
Fri Sep 20 06:35:34 2019 +0200
@@ -83,8 +83,8 @@

      public static class Class3 extends Base {
          private static final long serialVersionUID = 4L;
-        private String tuto = "tuto";
-        private byte b = (byte) 0xff;
+        private float[] floats = new float[] { 1.5f, -1.43f, 
131313131313.0f };
+        private double[] doubles = new double[] { 1.5, -1.43, 
131313131313.0 };
      }

  }


More information about the core-libs-dev mailing list