RFR: 8245002: Windows GDI functions don't support NUMA interleaving

[stefan.johansson at oracle.com]

Hi Stefan,

On 2020-05-19 13:53, Stefan Karlsson wrote:
> Updated webrevs:
>   https://cr.openjdk.java.net/~stefank/8245002/webrev.02.workaround.kims-review
>   https://cr.openjdk.java.net/~stefank/8245002/webrev.all.kims-review
> 
Looks good,
StefanJ
> StefanK
> 
> On 2020-05-18 10:17, Stefan Karlsson wrote:
>> On 2020-05-15 20:39, Kim Barrett wrote:
>>>> On May 15, 2020, at 1:46 PM, Stefan Karlsson 
>>>> <stefan.karlsson at oracle.com> wrote:
>>>>
>>>> Hi all,
>>>>
>>>> Please review this patch to turn off -XX:+UseNUMAInterleaving as 
>>>> long as Windows GDI functions don't work with memory that crosses 
>>>> memory reservation boundaries.
>>>>
>>>> https://cr.openjdk.java.net/~stefank/8245002/webrev.00.cleanup
>>>> https://cr.openjdk.java.net/~stefank/8245002/webrev.01.workaround
>>>> https://cr.openjdk.java.net/~stefank/8245002/webrev.all
>>>>
>>>> https://bugs.openjdk.java.net/browse/JDK-8245002
>>> ------------------------------------------------------------------------------ 
>>>
>>> src/hotspot/os/windows/os_windows.cpp
>>> 2753   void* verify_mem = ::malloc(4 * 1024);
>>>
>>> There's no corresponding free.
>>
>> Fixed.
>>
>>>
>>> ------------------------------------------------------------------------------ 
>>>
>>> src/hotspot/os/windows/os_windows.cpp
>>> 2776     assert(false, "Reservation failed");
>>> 2791     assert(false, "Failed to commit memory");
>>> 2802     assert(false, "Failed to commit memory");
>>>
>>> These look like leftover debugging code?
>> Not really. I left them in place so that it we would fail early in our 
>> testing, if we couldn't even create a few pages. However, I realize 
>> that this might be more problematic when we start to use this function 
>> for large pages, so I'll remove them.
>>
>> Thanks,
>> StefanK
>>
>>>
>>> ------------------------------------------------------------------------------ 
>>>
>>>
>>
>