RFR(S): 8050802: Update jprt runthese test suite to jck-8
david.holmes at oracle.com
Wed Jul 16 02:06:19 UTC 2014
On 16/07/2014 11:40 AM, Mikael Vidstedt wrote:
> From my empirical data the only test I've seen this "problem" with is
> the linux-i586-fastdebug-Xcomp; remember that there's a cost/overhead
> for setting up the individual tests too so splitting up the other Xcomp
> tests may actually make the job times longer.
> That said, if you feel that it's important for symmetry I can certainly
> do it.
Current split seems fine to me. Someone could always do additional
measurements with the other configurations to see if the split is
worthwhile - but as you say only fastdebug Xcomp is really a problem.
> On 2014-07-15 17:20, Vladimir Kozlov wrote:
>> I think you should split Xcomp on all platforms (not only for
>> linux.i586) where it runs.
>> On 7/15/14 3:59 PM, Mikael Vidstedt wrote:
>>> Please review the below change which switches the 'runThese' test suite
>>> over the new, jck-8 based 'runThese8' tests suite. The change also
>>> splits up the long running fastdebug-Xcomp test into two separate tests
>>> (fastdebug-Xcomp_lang and fastdebug-Xcomp_vm) to take advantage of the
>>> parallelism in jprt to reduce the job times further.
>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8050802
>>> Webrev (hs-rt/ (top) repo):
>>> Webrev (hs-rt/hotspot repo):
More information about the hotspot-dev