[aarch64-port-dev ] AARCH64 stage: next steps

[Lindenmaier, Goetz]

Hi,

This is quite badly documented flag, I wouldn't want to add new
dependencies to that:

  /* UseMembar is theoretically a temp flag used for memory barrier         \
   * removal testing.  It was supposed to be removed before FCS but has     \
   * been re-added (see 6401008) */                                         

At that location, ppc needs a membar always.  If we set UseMemBar to true,
serialization page is not used any more, right?  That's not what this local barrier
is about, it's needed even with serialization pages.

Best regards,
  Goetz




-----Original Message-----
From: hotspot-runtime-dev [mailto:hotspot-runtime-dev-bounces at openjdk.java.net] On Behalf Of Dean Long
Sent: Mittwoch, 21. Januar 2015 21:12
To: Andrew Haley; Vladimir Kozlov; aarch64-port-dev at openjdk.java.net; hotspot-runtime-dev at openjdk.java.net
Subject: Re: [aarch64-port-dev ] AARCH64 stage: next steps

On 1/21/2015 7:26 AM, Andrew Haley wrote:
> On 01/21/2015 03:05 PM, Andrew Haley wrote:
>> On 01/20/2015 11:34 PM, Vladimir Kozlov wrote:
>>> I pushed 8068053 and 8068054 changes today.
>>> Are these the last changes?
>> Yes.
> Oops, I forgot one:
>
> https://bugs.openjdk.java.net/browse/JDK-8069593
>
> webrev at
>
> http://cr.openjdk.java.net/~aph/aarch64-8069593-2/
>
> Andrew.
>

Doesn't the OrderAccess::fence() take care of this when UseMembar is 
true?  If so, then
perhaps the

#if defined(PPC64) || defined (AARCH64)

build-time check should be replaced with a runtime check ofUseMembar.

dl