RFR(s): 8171449" [aarch64] store_klass needs to use store release

Andrew Haley aph at redhat.com
Thu Dec 29 09:40:21 UTC 2016


On 28/12/16 22:59, Kim Barrett wrote:
> (3) There is a FIXME comment in the aarch64 store_klass definition
> suggesting a release_store might be needed.  That comment appears to
> be incorrect.

That comment is there because I was reading the x86 code (on which the
AArch64 code is based) and I saw this comment in tlab_refill:

  movptr(t1, ExternalAddress((address)Universe::intArrayKlassObj_addr()));
  // store klass last.  concurrent gcs assumes klass length is valid if
  // klass field is not null.
  store_klass(top, t1);

This is when recycling the free space at the top of an old tlab.  I
think that this comment is wrong.

Andrew.



More information about the hotspot-runtime-dev mailing list