Request for review: 7196080: assert(max_heap >= InitialHeapSize) in arguments.cpp

Tao Mao tao.mao at oracle.com
Mon Mar 18 11:45:45 PDT 2013


I checked that the logic flow you showed is correct. A new webrev is 
updated.
http://cr.openjdk.java.net/~tamao/7196080/webrev.02/

Thanks.
Tao

On 3/18/13 5:49 AM, Thomas Schatzl wrote:
> Hi all,
>
> On Fri, 2013-03-15 at 09:37 -0700, Jon Masamitsu wrote:
>> Looks fine.
>>
>> Jon
>>
>> On 3/14/2013 5:41 PM, Tao Mao wrote:
>>> The comment
>>>
>>> " 1172 // MaxHeapSize is aligned down in collectorPolicy"
>>>
>>> mislead here so I deleted it (because: in fact, in
>>> collectorPolicy.cpp, MaxHeapSize is aligned up). But the assertions
>>> still have some value.
>>>
>    both asserts will fail when using a small heap size, e.g. 1-3M.
>
> This is because default OldSize (and NewSize) are>=4(>=1M)
> respectively.
> (>= due to the use of ScaleForWordSize() on 64 bit)
>
> Additionally, in TwoGenerationPolicy::initialize_flags(),
> OldSize/NewSize and MaxHeapSize are intentionally synched, i.e.
> MaxHeapSize = OldSize + NewSize so that these asserts hold.
>
> My suggestion is to just remove them as these asserts seem to contradict
> TwoGenerationPolicy::initialize_flags().
> Atm these asserts cause differences in what heap sizes the VM accepts
> when running with or without debug info when using cms.
>
> Sorry for being a bit late.
>
> Thomas
>
>


More information about the hotspot-gc-dev mailing list