Hi Volker, On 2020-04-15 19:48, Volker Simonis wrote:
While doing this change, I've also realized that all the streams in java.util.zip (i.e. DeflaterInputStream, GZIPInputStream, GZIPOutputStream, InflaterInputStream, DeflaterOutputStream) use an internal byte buffer of 512 bytes by default. Looking at the benchmark attached to JDK-8242864, I think that increasing this default to a bigger size (e.g. 4096 bytes) will considerably speed up (up to 50%) read and write operations on these streams when they are created with the default buffer size. I think the value "512" is a legacy of old times when memory was more precious:) so I've opened "JDK-8242864: Increase the default, internal buffer size of the Streams in java.util.zip" to track that as as separate issue. Do you think it makes sense to increase that default value?
Seems reasonable. 8192 seems to be the buffer size we've been converging on elsewhere (see InputStream, BufferedInputStream, Files, ..). I also found an instance of 8096, which is either a typo or a clever optimization to keep the array + array object header fit snugly within 8Kb. You chose. :-)
Thank you and best regards, Volker PS: do you think it makes sense to contribute the benchmark attached to JDK-8242864 to the code-tools/mh-jdk-microbenchmarks [1] project? [1]http://openjdk.java.net/projects/code-tools/jmh-jdk-microbenchmarks/
I'd definitely welcome the micro as part of the patch under test/micro/org/openjdk/bench/java/util/zip - additionally contributing to jmh-jdk-microbenchmarks could enable you to test the micro on 8 or 11. /Claes
