RFR(M) 7199175: JSR 292: C1 needs patching when invokedynamic/invokehandle call site is not linked

John Rose john.r.rose at oracle.com
Fri May 31 09:56:49 PDT 2013

On May 31, 2013, at 11:20 AM, Vladimir Kozlov <vladimir.kozlov at oracle.com> wrote:

> So you are saying that you are fixing performance issue only. Okay, may make it diagnostic then to be able use it with product VM?

The main reason for it is as Roland said, to help diagnose failures.  For that a develop switch is sufficient.  There are a number of develop flags which serve this purpose, such as UseCHA (to pick at random).

It could also be used (maybe) to evaluate performance gains, but that's a more limited use.  For that it would have eto be diagnostic (as you say).

Either would be fine with me.

In principle, it would be reasonable to age out some develop flags like that.  To pick on UseCHA again, that technique is well established, and insensitive to platform variations, so we'll probably never need to test with it flipped the other way.

— John
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20130531/c3cf410e/attachment.html 

More information about the hotspot-compiler-dev mailing list