RFR(XS): JDK-8199780: SetMemory0 and CopyMemory0 in unsafe.cpp need to resolve their operands (original) (raw)

Erik Österlund erik.osterlund at oracle.com
Tue Mar 20 11:00:11 UTC 2018

Hi Roman,

On 2018-03-20 11:36, Roman Kennke wrote:

Same reason as splitting resolve -> resolveforread/resolveforwrite in other routines: being able to distinguish read and write access. Also, I'd rather be careful to put this stuff in central places that might over-cover it.

It sounds like the motivation for this in my opinion more fragile call site chasing code is optimization. What is the performance difference? Has this showed up in any profiles? Whenever robustness is traded for performance, it would be great to have some understanding about how much performance was lost.

I've missed UnsafeCopySwapMemory0, good find (has this been added recently?). I'll add Access calls there, and meditate a little bit how to put this into a more central place to avoid having to fix this for every code change in unsafe.cpp ;-)

No, this has been around for at least 2 years. In this file, address resolution is consistently done with index_oop_from_field_offset_long, so I think if you want to play it safe (and I know I would), then I would put the Access<>::resolve in there, and leave the rest of the call sites the way they are.

Thanks, /Erik

Thanks, Roman

Is there a good reason why the Access<>::resolve is not performed inside of indexoopfromfieldoffsetlong instead of its callsites. For example, it looks like barriers are missing in UnsafeCopySwapMemory0, that you would get for free by putting the resolve barrier in the API used in this file for resolving addresses.

Thanks, /Erik On 2018-03-19 15:44, Roman Kennke wrote: SetMemory0 and CopyMemory0 in unsafe.cpp read and write from/to objects, and thus need to resolve their operands via Access::resolve() before accessing them.

Bug: https://bugs.openjdk.java.net/browse/JDK-8199780 Webrev: http://cr.openjdk.java.net/~rkennke/JDK-8199780/webrev.00/ I'll say again that I'd prefer resolveforread() and resolveforwrite(), but for now the strong resolve() will suffice. ;-) Can I please get a review? Roman

More information about the hotspot-dev mailing list