RFR: 8209136: [JVMCI] Make sure volatile fields are read as volatile during constant reflection.

Vladimir Kozlov vladimir.kozlov at oracle.com
Fri Aug 17 20:34:53 UTC 2018


Looks good to me.

Thanks,
Vladimir

On 8/17/18 6:39 AM, Gilles Duboscq wrote:
> Please review this change to the way volatile fields are read in the `ConstantReflectionProvider`.
> 
> This should not affect normal java programs (there are no `volatile final` fields), it allows users of the JVMCI API to handle things like volatile stable fields:
> for example the compiler might want to constant-fold such field loads in program order and get volatile load ordering for the reads during constant folding.
> 
> webrev: http://cr.openjdk.java.net/~gdub/webrev-8209136
> issue:  https://bugs.openjdk.java.net/browse/JDK-8209136
> 
> Mach5 Testing: hs-tier1,hs-tier2,hs-tier3-graal
> 
>   Gilles
> 


More information about the hotspot-compiler-dev mailing list