RFR(XS) : 8038240 : new WB API to get nmethod

Igor Ignatyev igor.ignatyev at oracle.com
Tue Mar 25 22:03:56 UTC 2014


Thanks for review, I support you suggestion, but as you said we are 
consistent w/ what we have, also we need this in 8u20, so we don't have 
enough time.
I will file an RFE to change WB API and add version to the NMethod in 
accordance w/ your advice.

Do you agree w/ such approach? If yes, can I consider you as a Reviewer?


On 03/26/2014 01:01 AM, John Rose wrote:
> The boolean parameter is not expressive enough.  It will have to change as soon as we have more than two versions of a method.
> (In fact, we already do:  There is potentially a different OSR method for each entry point BCI.  The formulation you have returns just one of them.)
> I suggest you change "boolean isOsr" in the WB API to "long version" or even "Object version".  The version enumeration can be evolved over time.  Values 0 and 1 can be "normal at current compilation" and "first available OSR".
> Also, put the version number into the NMethod structure, so it can be requeried.  For an OSR method, the entry BCI is important to know.
> ...But I see that there is already a use of booleans like this in the WB API.  So I guess you should consistent with what you have.
> I filed a bug to track this issue of method versioning; there is technical debt here:
>    https://bugs.openjdk.java.net/browse/JDK-8038356
> — John

More information about the hotspot-dev mailing list