RFR (XXS): 8200365: TestOptionsWithRanges.java of '-XX:TLABWasteTargetPercent=100' fails intermittently

Hohensee, Paul hohensee at amazon.com
Fri Aug 10 16:24:02 UTC 2018


+1.

Paul

On 8/9/18, 9:59 AM, "hotspot-gc-dev on behalf of sangheon.kim at oracle.com" <hotspot-gc-dev-bounces at openjdk.java.net on behalf of sangheon.kim at oracle.com> wrote:

    Hi Thomas,
    
    On 8/9/18 3:34 AM, Thomas Schatzl wrote:
    > Hi,
    >
    >    can I have reviews for this tiny fix that fixes startup issues with
    > serial, parallel and CMS collectors?
    >
    > The reason is that high values of TLABWasteTargetPercent (>=26) cause
    > TLABs getting to large during startup with default ergonomics, causing
    > a GC when a second Java thread wants to allocate its TLAB.
    > However at that time we do not support GCs, and so the VM forcibly
    > fails.
    >
    > The fix is to increase the number of refills per thread for the initial
    > TLAB size calculation only.
    >
    > CR:
    > https://bugs.openjdk.java.net/browse/JDK-8200365
    > Webrev:
    > http://cr.openjdk.java.net/~tschatzl/8200365/webrev/
    > Testing:
    > Running failing test with serial/parallel/cms 400 times each without
    > failure.
    Looks good.
    
    Thanks,
    Sangheon
    
    
    >
    > Thanks,
    >    Thomas
    
    



More information about the hotspot-gc-dev mailing list