Problem Description:
----------------------
David experimented with some unpublished flags he knew about
(-XX:+UseSpinning -XX:PreBlockSpin=100 -XX:+UseTLAB) to force the VM to spin instead of block. These flags
allowed the VM to reach the same level of performance we saw with jrockit.
>
> Here is a summary of the two best runs we've got:
>
> JVM app.idle_time resp. times NO/Mfg TOPS IR #VMs run id.
> -----------------------------------------------------------------------------------------------
> jrockit81_141_03 8% 2.0/2.5 335.90 196 2 115
> j2sdk1.4.1_02+ ... 1% 1.1/4.25 348.11 202 1 148
----------------------
David experimented with some unpublished flags he knew about
(-XX:+UseSpinning -XX:PreBlockSpin=100 -XX:+UseTLAB) to force the VM to spin instead of block. These flags
allowed the VM to reach the same level of performance we saw with jrockit.
>
> Here is a summary of the two best runs we've got:
>
> JVM app.idle_time resp. times NO/Mfg TOPS IR #VMs run id.
> -----------------------------------------------------------------------------------------------
> jrockit81_141_03 8% 2.0/2.5 335.90 196 2 115
> j2sdk1.4.1_02+ ... 1% 1.1/4.25 348.11 202 1 148
- duplicates
-
JDK-4718952 Linux: synchronization primitive improvements
-
- Closed
-