RFR: 8242459: ForceNUMA and only one available NUMA node hits a guarantee

[Kim Barrett]

> On Apr 27, 2020, at 3:42 AM, stefan.johansson at oracle.com wrote:
> 
> Hi Kim,
> 
> On 2020-04-27 05:21, Kim Barrett wrote:
>> Please review this small change to avoid failing assertions when
>> -XX:+ForceNUMA is used to override the decision to disable UseNUMA
>> when there is only one NUMA node being used.  There were a couple of
>> places that assumed that if UseNUMA is true then there must be more
>> than one NUMA node, but that isn't so.
>> CR:
>> https://bugs.openjdk.java.net/browse/JDK-8242459
>> Webrev:
>> https://cr.openjdk.java.net/~kbarrett/8242459/open.00/
> Looks good, might be interesting to run a few tier with +ForceNUMA, to make sure we don't hit anything else.

Thanks.  I’m running tier1 with +ForceNUMA right now.

> 
> Thanks for fixing,
> Stefan
> 
>> Testing:
>> mach5 tier1
>> Locally (linux-x64) manually ran the provided test case, verifying
>> failure before changes and working after.