RFR 8136924 Vectorized support for array equals/compare/mismatch using Unsafe

[Paul Sandoz]

> On 25 Nov 2015, at 18:29, Peter Levart <peter.levart at gmail.com> wrote:
> 
> 
> 
> On 11/25/2015 12:46 PM, Paul Sandoz wrote:
>> We are in Unsafe territory :-)
>> 
>> 
>> Since the Java mismatch method and the intrinsic only make temporary reads, IIUC such updates by GC to the array can only occur at a safe points, and i don’t think int loops produce a safe point check within the hot loop and nor does the intrinsic, is this really an issue?
>> 
> 
> IIUC, in interpreter, safepoint can happen between any two bytecodes:
> 
> "Safepoints are initiated using a cooperative, polling-based mechanism. In simple terms, every so often a thread asks “should I block for a safepoint?”. Asking this question efficiently is not so simple. One place where the question is often asked is during a thread state transition. Not all state transitions do this, for example a thread leaving the VM to go to native code, but many do. The other places where a thread asks are in compiled code when returning from a method or at certain stages during loop iteration. Threads executing interpreted code don't usually ask the question, instead when the safepoint is requested the interpreter switches to a different dispatch table that includes the code to ask the question; when the safepoint is over, the dispatch table is switched back again. Once a safepoint has been requested, the VMThread must wait until all threads are known to be in a safepoint-safe state before proceeding to execute the VM operation. During a safepoint the Threads_lock is used to block any threads that were running, with the VMThread finally releasing the Threads_lock after the VM operation has been performed."
> 

Ah, i see, if in the interpreter the dispatch table is switched between getting av and bv then that would be a problem:

  long av = U.getLongUnaligned(a, aOffset + bi);
  long bv = U.getLongUnaligned(b, bOffset + bi);

I will document that this method should not be used with reference component types :-) (i anyway neglected to mention about alignment in those cases).

Thanks,
Paul.

> 
> You could treat a "false positive" mismatch like you do with NaN floats and doubles (fall-back to normal code), but what about "false negatives"?
> 
> Suppose that you take a snapshot of address of object A from the 1st array, followed by safepoint which moves object A to somewhere else and moves object B to the same address A has occupied before (compacting collector). After safepoint returns, you take the snapshot of address of object B from the corresponding slot of the 2nd array. Now you compare the addresses and they are the same, but they  represent(ed) pointers to distinct objects.
> 
> Regards, Peter
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 841 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20151125/491ec55b/signature.asc>