RFR: 8150183: gc/g1/plab/TestPLABResize.java - previous PLAB size should be less than current

Dmitry Fazunenko dmitry.fazunenko at oracle.com
Wed Feb 24 13:08:26 UTC 2016

Hi Misha,

The fix looks good, one minor suggestion (no extra webrev is required):


  113     /**
  114 * Returns Map of log entries which contains Long as key and Map as 
115 * Map<ReportType, Map<String,Long>>. ReportType is a key, value is 
Map of String as key and Long as value.
  116      *
  117 * @return Returns Map of log entries.
  118      */
119 public Map<Long, Map<ReportType, Map<String, Long>>> getEntries() {


* Returns the GC log entries for Survivor and Old stats. * The entries 
are represented as a map of gcID to the StatMap.
* @return The log entries for the Survivor and Old stats.
       */public Map<Long, Map<ReportType, Map<String, Long>>> getEntries()


On 24.02.2016 15:21, Michail Chernov wrote:
> Hello,
> Please review this fix for gc/g1/plab/TestPLABResize.java test.
> Webrev: http://cr.openjdk.java.net/~mchernov/8150183/webrev.00/
> Bug: https://bugs.openjdk.java.net/browse/JDK-8150183
> This test tries to check resizing PLAB. PLAB desired size calculation 
> does not depend on number of GC thread, AppPLABResize.java was fixed.
> Also checking of PLAB sizes was made less strict because we cannot 
> expect that PLAB will change exactly as amount of promoted bytes. Now 
> test checks only start and end values of test iteration.
> This was checked on a lot of VM options combination used in nightly.
> PS: Sorry for duplicate of message.
> Thanks
> Michail

More information about the hotspot-gc-dev mailing list