RFR 8204128: NMT might report incorrect numbers for Compiler area
Zhengyu Gu
zgu at redhat.com
Fri Nov 15 18:37:26 UTC 2019
Thanks, Thomas
On 11/15/19 12:59 PM, Thomas Stüfe wrote:
> Hi Zhengyu,
>
> wouldn't ssize_t not be a better choice?
You are right!
Updated: http://cr.openjdk.java.net/~zgu/JDK-8204128/webrev.01/index.html
Reran the tests.
-Zhengyu
>
> Other than that, looks good.
>
> ..Thomas
>
> On Fri, Nov 15, 2019 at 4:08 PM Zhengyu Gu <zgu at redhat.com
> <mailto:zgu at redhat.com>> wrote:
>
> I could not reproduce the problem stated in CR.
>
> The theory is that, when releasing a 2GB+ arena,
> Arena::set_size_in_bytes() passes a negative long integer to NMT, when
> it goes through long -> int -> long conversion, at the end, it
> becomes a
> positive number.
>
> This problem is illustrated in new test. Without the fix, after
> releasing a 2GB+ arena, NMT shows the arena size doubled, instead of
> going down.
>
> I am not completely sure this fixes the problem reported, but it is
> worth to cleanup inconsistent types in NMT API.
>
>
> Bug: http://cr.openjdk.java.net/~zgu/JDK-8204128/webrev.00/index.html
> Webrev: http://cr.openjdk.java.net/~zgu/JDK-8204128/webrev.00/index.html
>
> Test:
> hotspot_nmt + new test (fastdebug and release)
> on Linux x86_64
>
> Thanks,
>
> -Zhengyu
>
More information about the hotspot-runtime-dev
mailing list