Request for reviews (XS): 6868486: timouts and outOfMemory in regression tests

Vladimir Kozlov Vladimir.Kozlov at Sun.COM
Tue Aug 4 16:34:20 PDT 2009

% time bin/java  -server -Xcomp -XX:-PrintVMOptions -XX:CompileThreshold=100  Test6559877
Checking all 1-byte combinations...
Checking all 2-byte combinations...
856.0u 3.0s 14:37 97% 0+0k 0+0io 0pf+0w

It does class verification for bytecodes combinations (256x256 matrix)
and it requires a new instance for each iteration. I don't see how we can
improve it.

I already using 100 as iteration count and CompileThreshold in 6826736 test.

Note, it is 296 MHz US-II box. On new machines it takes about 1 min.
Other tests take less then 1 min on new machines.


Tom Rodriguez wrote:
> 10 - 15 minute runtimes are pretty long for tests that we're running as 
> part of nightlies.  Is it possible to make the tests run in less time?  
> Ideally regression tests shouldn't take more than a minute or two.
> tom
> On Aug 4, 2009, at 4:03 PM, Vladimir Kozlov wrote:
>> Fixed 6868486: timouts and outOfMemory in regression tests
>> Problem:
>> Nightly testing uses some very old machines which are slow,
>> have one cpu(core) and <= 1Gb memory.
>> Such machines are not server class for VM so it use
>> Serial GC which also slowdown execution time.
>> Solution:
>> Increase timeout for all bug's tests.
>> Increase heap size for 6851282 test.
>> Reviewed by:
>> Fix verified (y/n): y, bug's tests
>> Other testing:

More information about the hotspot-compiler-dev mailing list