RFR (S): 8049355: compiler/rtm/locking/TestRTMLockingThreshold test may fail if transaction was aborted by interrupt
filipp.zhinkin at oracle.com
Tue Dec 30 18:15:45 UTC 2014
Thank you for review, Vladimir.
----- Original Message -----
From: vladimir.kozlov at oracle.com
To: hotspot-compiler-dev at openjdk.java.net
Sent: Tuesday, December 30, 2014 8:33:57 PM GMT +03:00 Iraq
Subject: Re:  RFR (S): 8049355: compiler/rtm/locking/TestRTMLockingThreshold test may fail if transaction was aborted by interrupt
On 12/30/14 8:52 AM, Filipp Zhinkin wrote:
> Hi all,
> please take a look at a fix for 8049355.
> A few RTM tests verify that method was recompiled if
> abort ratio was pretty low. These tests also verify that
> before it happened, certain amount of lock attempts was done.
> In order to guarantee low abort ratio these tests force abort
> at every Nth iteration (every even iter, actually)
> or only once at all.
> However, abort may happened on iteration when we're not trying
> to force abort (due to CPU interrupt, VMM call, etc.) and amount
> of lock attempts may differ from expected value.
> To avoid such issue tests were updated to not force abort first
> L / 2 iterations and during last L / 2 iterations force it always
> (L == RTMLockingThreshold value).
> Bug id: https://bugs.openjdk.java.net/browse/JDK-8049355
> Webrev: http://cr.openjdk.java.net/~fzhinkin/8049355/webrev.00/
> Testing: manual & automated on host w/ HSW CPU, JPRT
More information about the hotspot-compiler-dev